Hildebrand CA

..
-'
..-'
~---------------...-'
AVN
e L lUS
F
APPLIC
SEl J : DI ~.
10
r
s
F. B. HILDEBRAND
Comprehensive ,
integrate d, a ~ d
clear presen ~at i on
of meth o ds and
theory in a
variety o f areas
of applicable ./..-'
mathematics..-' ..-'
..-'
..-'
..-'
Advanced
Calculus
for
Applications
SECOND EDITION
Francis B. Hildebrand
Prolessor o/ Mathematics
lvlassachusetts Instirule 01 Technology
PRENTICE-HALL, INe.
Englew oo d Cliffs, New Jersey
Library oI Congress Cataloging in Publication Data
HILDEBRA,ND, FRANCIS BEGN .~UD.
Advanced calculus for applications .
Published in 1948 under ti,le: Advanced calculus for
engineers .
Bibliography: p.
Ineludes indexo
1. Calculus. 1. Ti,le.
QA303.H55 1976 515 75-J447J
ISBN 0-13-011189-9
© 1976 by Prentiee-Hall, Ine.

Englewood Cliffs, New Jersey
AlI rights reserved. No part of this book

may be reprodueed in any form or by any means
without pennission in writing from the publisher.
109876
Printed· in the Uoited States of Amer:ea
PRENTICE-HALL lNTERNATIONAL, IN e. , Londan

PRENTTCE-HALL OF AUSTRALIA, PTY. LTD., Sydney
PRENTICE-HALL OF CANADA, LTD., Taranta
PRENTlCE-HALL Of rNDlA PRIVATE LtMITED, New Del/¡¡
PRENTICE-HALL OF JAPAN, INe., Tokyo
PRENTICE-HALL Of SOUTHEAST ASIA (PIT.) L TD . , Singapore
Contents
Preface xi
1
Ordinary Differential Equations 1
1.1 Introduction 1
1.2 Linear Dependence 3
1.3 Complete Solutions of Linear Equations 4
1.4 The Linear Differential Equation of First Order 6
1.5 Linear Differential Equations with Constant Coefficients 8
1.6 The Equidimensional Linear Differential Equation 12
1.7 Properties of Linear Operators 15
1.8 Simultaneous Linear Differential Equations 18
1.9 Particular Solutions by Variation of Parameters 24
1.10 Reduction of Order 28
1.11 Determination of Constants 30
1.12 SpeciaJ Solvable Types of Nonlinear Equations 31
2
Tbe Laplace Transform 53
2.1 An introductory Example 53

2.2 Definition and Existence of Laplace Transforms 55
v
vi Conten!s
2.3 Properties of Laplace Transforms 58

2.4 The Inverse Transform 62
2.5 The Convolution 63
2.6 Singularity Functions 65
2.7 Use of Table 01' Traosforms 67
2.8 Applications to Linear Differential Equations with
Constant Coefficients 72
2.9 The Gamma Function 76
3
Numerical Methods for Solving Ordinary Differential Equations 93
3.1 Introduction 93
3.2 Use of Taylor Series 94
3.3 The Adams Method 96
3.4 The Modified Adams Method 100
3.5 The Runge-Kutta Method 102
3.6 Picard's Method 105
3.7 Extrapolation with Differences 107
4
Series Solutions of Differential Equations: Special Functions 118
4.1 Properties of Power Series 118

4.2 Illustrative Examples 122
4.3 Singular Points of Linear Second-Order Differential
Equations 126
4.4 The Method of Frobenius 128
4.5 Treatment of Exceptional Cases 134
4.6 Example of an Exceptional Case 136
4.7 A Particular Class of Equations 138
4.8 Bessel Functions 141
4.9 Properties of Bessel Functions 147
4.10 Differential Equations Satisfied by Bessel Functions 151
4.11 Ber and Bei Functions 153
4.12 Legendre Functions 155
4.13 The Hypergeometric Function 162
4.14 Series Solutions Valid for Large Values of x 163
lts Contents vii
5
Boundary-Value Problems and Characteristic-Function Representations 186
5.1 Introduction 186

5.2 The Rotating String 188
5.3 The Rotating Shaft 192
5.4 Buckling of Long Columns Under Axial Loads 195
5.5 The Method of Stodola and Vianello 197
5.6 Orthogonality of Characteristic Functions 203
5.7 Expansion of Arbitrary Functions in Series of Orthogonal
Functions 207
5.8 Boundary-Value Problems Involving Nonhomogeneous
Differential Equations 211
5.9 Convergence of the Method of Stodola and Vianello 212
5.10 Fourier Sine Series and Cosine Series 214
5.11 Complete Fourier Series 219
5.12 Term-by-Term Differentiation of Fourier Series 223
5.13 Fourier-Bessel Series 226
5.14 Legendre Series 230
5.15 The Fourier Integral 234
6
Vector Analysis 269
6.1 Elementary Properties of Vectors 269

6.2 The Scalar Product of Two Vectors 271
6.3 The Vector Product of Two Vectors 273
6.4 Multiple Products 275
6.5 Differentiation of Vectors 277
6.6 Geometry of a Space Curve 278
6.7 The Gradient Vector 281
6.8 The Vector Operator V 283
6.9 Differentiation Formulas 284
6.10 Line In tegrals 287
6.11 The Potential Function 291
6.12 Surface Integrals 294
6.13 Interpretation of Divergence. The Divergence Theorem 297
6.14 Green's Theorem 301
6.15 Interpretation of Cur!. Laplace's Equation 302
6.16 Stokes's Theorem 303
6.17 Orthogonal Curvilinear Coordina tes 306
viii Coutents
6 .18 Special Coordinate Systems 311

6 .19 Application to Two-Dimensional Incompressible Fluid Flow 313
6.20 Compressible Ideal Fluid Flow 316
7
Topies in Higher-Dimensional Calculus 342
7 .1 Partial Differentiation. Chain Rules 342
7.2 Implicit Functions. Jacobian Determinants 347
7.3 Functional Dependence 350
7.4 Jacobians and Curvilinear Coordinates. Change of Variables
in Integrals 352
7.5 Taylor Series 354
7 .6 Maxima and Minima 356
7.7 Constraints and Lagrange Multipliers 357
7.8 Ca1culus of Variations 360
7.9 Differentiation of Integrals Involving a Parameter 364
7 . 10 Newton's Iterative Method 367
8
Partial Differential Equations 384
8.1 Definitions and Examples 384
8.2 The Quasi -Linear Equation of First Order 387
8.3 Special Devices. Initial Conditions 392
8.4 Linear and Quasi-Linear Equations of Second Order 396
8.5 Special Linear Equations of Second Order, with Constant
Coefficients 397
8.6 Other Linear Equations 400
8.7 Characteristics of Linear First-Order Equations 403
8.8 Characteristics of Linear Second-Order Equations 408
8 .9 Singular Curves on Integral Surfaces 414
8.10 Remarks on Linear Second-Order lnitial-Value Problems 417
8 .11 The Characteristics of a Particular Quasi-Linear Problem 417
9
Solutions of Partial Differential Equations of Mathematical Pbysics 439
9 .1 Introduction 439
9.2 Heat Flow 441
9.3 Steady-State Temperature Distribution in a Rectangular Plate 443
9.4 Steady-State Temperature Distribution in a Circular Annulus 446
Cootents ix
9.5 Poisson's Integral 450

9.6 Axisymmetrical Temperature Distribution in a Solid Sphere 451
9.7 Temperature Distribution in a Rectangular Parallelepiped 453
9.8 Ideal Fluid Flow about a Sphere 456
9.9 The Wave Equation. Vibration of a Circular Membrane 459
9. JO The Heat-Flow Equation. Heat Flow in a Rod 461
9.11 Duhamel's Superposition 1 ntegral 463
9.12 Traveling Waves 467
9.13 The Pulsating Cylinder 470
9.14 Examples of the Use of Fourier Integrals 473
9.15 Laplace Transform Methods 477
9.16 Application of the Laplace Transform to the Telegraph
Equations for a Long Line 480
9.17 Nonhomogeneous Conditions. The Method of Variation of
Parameters 484
9.18 Forrnulation of Problems 490
9.19 Supersonic Flow of Ideal Compressible Fluid Past an Obstacle 495
10
Functions oC a Complex Variable 539
10.1 Introduction. The Complex Variable 539

10.2 Elementary Functions o[ a Complex Variable 541
10.3 Otber Elementary Functions 544
10.4 Analytic Functions o[ a Complex Variable 550
10.5 Line lntegrals of Complex Functions 554
10.6 Cauchy's Integral Formula 560
10.7 Taylor Series 561
10.8 Laurent Series 563
10.9 Singularities of Analytic Functions 567
10.10 Singularities at Infinity 575
10.11 Significance of Singularities 578
10.12 Residues 580
10.13 Evaluatioo of Real Definite Integrals 583
10.14 Theorems on Limiting Contours 589
10.15 Indented Cootours 592
10.16 Iotegrals Involving Branch Points 595
11
Applications of Analytic Function Theory 622
11.1 Introduction 622

11.2 Inversion of Laplace Transforms 622
x Contents
11.3 Inversion of Laplace Transforms wilh Branch Points . The Loop

Integral 625
11.4 Conformal Mapping 628
1 1.5 Applications to Two-Dimensional Fluid Flow 632
11.6 Basic Flows 634
11.7 Other Applications of Conformar Mapping 638
11.8 The Schwarz-Christoffel Transformation 64/
j 1.9 Green's Functions and the Dirichlet Problem 652
lUO The Use of Conformal Mapping 658
II.! j Other Two-Dimensional Green·s Functions 661
Answers to Problems 703
Index 721
Preface
The purpose of thi s text is to present an integrated treatment of a number of

those topics in mathematics which can be made to depend only upon a sound
course in elementary calculus, and which are of common importance in many
fields of application.
An attempt is made to deal with the various topics in such a way that a
student who may not proceed into the more profound areas of mathematics
still may obtain an intelligent working knowledge of a substantial number of
useful mathematical methods, together with an appropriate awareness of the
foundations , interrelations , and limitations of these methods . At the same time,
it is hoped that a student who is to progress, say, into a rigorous course in mathe-
matical analysis will be provided, in addition, with increased incentive and
motivation . For both of these purposes , the phrase «It can be shown" is used
occas ionally, not only to exhibit a generalization of an established conclusion
or a useful related fact, but also to introduce a needed basic result for which a
rigorou s demonstration would require what is believed to be an inappropriately
excessive amount of detailed analysis or of prerequis ite preparation.
Thi s revision incorporates a large number of relatively minor changes for
the purpose of increased clarity or preci sion or to supply a previously omitted
proof, a substantial amount of added textual material (particularly in the later
chapters) , and about 250 additional problem s .
The first four chapters are concerned chiefiy with ordinary differential
equation s, including analytical , operational, and numerical methods of solution,
and with special functions generated as solution s of such equations. In par-
ticular, the material of the first chapter can be considered as either a systematic
review or an ini tial introduction to the elementary concepts and techniques, as-
xi
xii Prerace
sociated with linear equations and with special solvable types ofnonlinear equa-
tions, which are needed in subsequent chapters. The fifth chapter deals with
boundary-value problems governed by ordinary differential equations, with the
associated characteristic functions, and with series and integral representations
of arbitrary functions in terms of these functions.
Chapter 6 develops the useful ideas and tools of vector analysis; Chapter 7
provides brief introductions to sorne special topics in higher-dimensional ca\culus
which are rather frequently needed in applications. The treatment here occa-
sionally consists essentially of indicating the plausibility and practical signifi-
canee of a result and stating conditions under which its validity is rigorously
established in listed references.
In Chapter 8, certain basic concepts associated with the simpler types of
partial differential equations are introduced, after which, in Chapter 9, full use
is made of most of the tools developed in earlier chapters for the purpose of
formulating and solving a variety of typical problems governed by the partial
differential equations of mathematical physics. A new section deals with the ap-
plication of the so-called method of variation of parameters to such problems.
Chapter \O treats the basic topics in the theory of analytic functions of a
complex variable, inc\uding contour integration and residue ca\culus. Although
certain developments in preceding chapters could be made more elegant and
more complete if they were made to depend upon this treatment, introduced at
an earlier stage, it is felt tha t, in sorne cases, the knowledge based on a brief
initial study of analytic functions may not be sufficiently firm to support sign.ifi-
cantly dependent treatments of the other topics, but that such knowledge then
may better serve to c\arify the other topics when subsequently provided. How-
ever, since most of the treatments of Chapter 10, as well as most of those of
Chapters 6 and 7, are independent of the content of preceding chapters, material
from these chapters can indeed be introduced at an earlier stage in a given course,
at the discretion of the instructor. It has been considered reasonable to assume
knowledge of certain elementary properties of complex numbers in the earlier
chapters, even though the solution of the equation x' + I = O then may occa-
sion a personal review on the part of the reader.
A new Chapter 11 considers sorne applications of analytic function theory
to other fields, inc\uding the derivation of methods for the inversion of Laplace
transforms (an expansion of material previously presented in annotated prob-
lems), an indication of the properties and uses of conformal mapping (formerIy
inc\uded in Chapter 10), and a new brief treatment of Green's functions as
related to partial differential equations.
Extensive sets of problems are inc\uded at the end of each chapter, grouped
in correspondence with the respective sections with which they are associated.
In addition to more-or-less routine exercises, there are numerous annotated
problenis which are intended to guide the reader in developing results or tech-
niques which extend or complement treatments in the text, or in dealing with a
particularly challenging application . Such problems may serve as focal points
Preface xiii
for extended diseussions or for the introduetion of additional (or aJternative)

material into a ehapter, permitting the text to serve somewhat more flexibly in
eourses of varied types. New problems of this sort now permit the eonsideration
of topies sueh as one-dimensional Green's funetions and applieations of elliptie
integrals, Fourier transforms, and assoeiated Legendre funetions. Answers to
all problems are either ineorporated into the statement of the problem or listed
at the end of the book.
The author is partieularly indebted to Professor E. Reissner for valuable
eollaboration in the preliminary stages of the preparation of the original edition
and for many ideas whieh eontributed to whatever useful novelty some of the
treatments may possess, and to Professor G. B. Thomas for additional adviee
and help, as well as to a rather long list of other eolleagues and students who
have offered eritieisms and suggestions leading to many of the modifieations in-
eorporated into this revision.
F. B. HILDEBRAND
Advanced
Calculus
for
Applications
1
Ordinary Differential Equations
1.1. Introduction. A differential equation is an equation relating two or

more variables in terms of derivatives or differentials. Thus the simplest differ-
ential equation is of the form
dy
dx = h(x). (1)
where h(x) is a given function of the independent variable x. The solution is

obtained immediately by integration, in the form
y = f h(x) dx + e, (2)
where e is an arbitrary constant. Whether or not it happens that the integral

can be expressed in terms of named or tabulated functions is incidental, in the
sense that we accept as a solution of a differential equation any functional rela-
tion, not involving derivatives or integrals of unknown functions, the satisfaction
of which implies the satisfaction of the differential equation. Similarly, in an
equation of the form
F(x)G(y) dx + f(x)g(y) dy = O, (3)
we may separate the variables and obtain a solution by integration in the form
I F(x) dx '+'
f(x) '
I g(y) dy
G(y)
= e
,
(4)
if suitable account is taken of situations in which a divisor may vanish.

Usuaily we desire to obtain the most general solution of the differential
equation; that is, we require a// functional relations which imply the equation.
1
2 Ord inary Differential Equations
In the general case it may be difficult to determine when all such relations have
indeed been obtained. Fortunately, however, this difficulty does not exist in
the case of so-called linear differential equations, which are of most frequent
occurrence in applications and which are to be of principal interest in what
follows.
A differential equation of the form
d"y dn-'y dy
ao(x) dxn + a,(x) dxn , + ... + an_,(x) dx + a"(x)y =f(x) (5)
is said to be a linear differential equation of order n. The distinguishing charac-

teristic of such an equation is the absence of products or nonlinear functions
of the dependent variable (unknown function) y and its derivatives, the highest
derivative present being of order n. The coefficients ao(x), ... , an(x) may be
arbitrarily specified functions of the independent variable x.
For a linear equation of the first order,
ao(x):%: + a,(x)y =f(x),
it is shown in Section 1.4 that if both sides of the equation are multiplied by a
certain determinable function of x (an "integrating factor"), the equation
always can be put in an equivalent form
d
d)P(x)y] = F(x),
where p(x) and F(x) are simply expressible in terms of a o , a" and f, and hence
then can be solved directly by integration .
Although no such simple general method exists for solving linear equations
of higher order, there are two types of such equations which are of particular
importance in applications and which can be completely solved by direct
methods. These two cases are considered in Sections 1.5 and 1.6. In addition,
this chapter presents certain techniques that are available for treatment of
more general linear equations.
Many of the basically useful properties of linear differential equations do
not hold for nonlinear equations, such as
dY=X+y2,
dx
d 2y
dX2 + .
Sill
_
y - O, d2 + x....1:
---.l'
dX2
(ddx )2 + y = e.
A few special types of solvable nonlinear equations are dealt with briefly ID
Section 1.12.
The equations to be considered in this chapter are known as ordinary
differential equations, as distinguished fram partial differential equations,
which involve partial derivatives with respect to two or more independent
variables. Equations of the latter type are treated in subsequent chapters.
Before proceeding to the study of linear ordinary differential equations, we
next briefly introduce the notion of linear dependen ce, which is basic in this
work.
1.2. Linear Dependence 3
1.2. Linear Dependence. By a linear combination of n functions u, (x),

u 2 (x), . .. , un(x) is meant an expression of the form
n
e,u,(x) + e,u,(x) + ... -j cnu.(x) = 1: ekuk(x), (6)
k=1
where the c's are constants . When at least one c is not zero, the linear com-
bination is termed nontrivial. Tbe functions u" l/2' . . • , Un are then said to be
linearly independent over a given interval (saya -< x -< b) if over that interval
no one of the functions can be expressed as a linear combination of the others ,
or, equivalently, if no nontriviallinear combination of the functions is identically
zero over the interval considered. Otherwise, the functions are said to be
linearly dependent over that interval.
As an example, the functions cos 2x, cos 2 x, and 1 are linearly dependent
over any interval because of the identity
cos 2x - 2 cos' x + l _ O.
It follows from the definition that two functions are linearly dependent
over an interval if and only if one function is a constant multiple of the other
over that interval. The necessity of the specification of the interval in the
general case is illustrated by a consideration of the two functions x and I x l.
In the interval x > ° there follows x - Ixl = 0, whereas in the interval x <
we have x + Ixl - O. Thus the two functions are linearly dependent over any
°
interval not including the point x = O; but they are linearly independent over
any interval including x = O, since no single linear combinatíon of the two
functíons ís identically zero over such an interval.
Although in practice the linear dependence or independence of a set of
functions generally can be established by inspection, the following result is of
sorne importance in theoretical discussions. We assume that each of a set of n
functions l/" ti" . . . , Un possesses n finite derívatives at all points of an interval
l. Then, if a set of constants exists such that
e,u, + e,u + ...

2 +cnun=O
for all values of x in I, these same constants also satisfy the identities
Cl
du,
dx + e, du,
dx + ... + C n ddUxn = 0,
d'u
e, dX,1 + e, d'u
dX" + ... + Cn
d'u
dx,n = 0,
Thus the n constants must satisfy n homogeneous linear equations. However,

such a set of equations can possess nontrivial solutions only if its coefficient
determinant vanishes. Thus it follows that if the functions u" 11 " .•• , Un are
4 Ordioary Dift'ereotial Equalions
Iinearly dependent aYer an interval J, then the determinant
W(u¡, u 2 , ••• , un) = (7)

........................
dn-¡u¡ d n - ¡ U2
dxn ¡ dxn ,
vanishes identica/ly ayer J. This determinant appears frequently in theoretical
work and is caJled the Wranskian (or Wronsklan determinant) of the functions .
Thus we see that if the Wronskian af u" Uz , . . . , Un is not id?ntically zero over J,
then the fimctians are Iinearly independent a yer J.
To illustrate, since the value of the determinant
,
1 X x- xJ x·
O 1! 2x 3X2 nx,.-t
O O 2! 6x n(n - l)x n- z
W(J, x, x ' , ... , xn) =
O O O 3! n(n - IXn - 2)x n - J
O O O O n!
is merely the product of the nonvanishing constants appearing in the principal
diagonal and hence cannot vanish, it follows that the functions appearing in
the first row are linearly independent (over any interval).
Unfortunately, the converse of the preceding theorem is nat true since, in
unusual cases, the Wronsklan of a set of Iinearly independent functions also may
vanisb. Tbat is, the vanishing of the Wronskian is necessary but not sufficient
for linear dependence of a set of functions. (For an example establishing the
insufficiency, see Problem 5.)
1.3. Complete Solutions oC Linear EqnatioDS. The most general linear differ-
ential equation of the nth order can be written in the form
dny d·-' y dy_
- x,.
d
+ a¡(x) d x" -, + ... + an-¡(x)d-
x
+ a'(x)y - h(x). (8)
Here it is assumed that both sides of the equation have been divided by the
coefficient of the highest derivative. We will speak of this forro as the standard
farm of the equation. This equation is frequently written in the abbreviated form
Ly = h(x), (9)
where L here represents the linear differential aperator
dn d n- 1 d
L = d-
xl'J
+ a,(x)d~l
x"
+ .. . + an-¡(x)d-
x
+ a.(x). (10)
The problem of solving Equation (8) consists of determining the most

1.3. Complete Solulions of Linear Equations s
general express ion for y which, if substituted into tbe left-hand si de of (8), or if
operated on by (lO), gives the prescribed right-hand side h(x). When a relation-
ship of the form y = u(x) satisfies Equation (8), it is conventional to say that
eitber the relation y = u(x) or the function u(x) is a solulion of that equation . t
Ir all the coefficients a,(x) • ... , a.(x) were zero, the solution of Equation
(8) would be accomplished directly by n successive integrations, each integra-
tion introducing ao independent constant of integration. Thus it might be
expected that the general solution of (8) also would contain n independent
arbitrary constants. As a matter of fact, it is known that in any interval 1 in
which the coefficients are continuous, there exists a continuous Solulion to Equation
(8) involving exactly n independent arbitrary constants; furthermore, there are no
solutions of Equation (8) valid in 1 which cannot be obtained by specializing the
constants in any sueh solution.
1t should be noticed that tbis is a property peculiar to linear differential
equations. To illustrate, the nonlinear differential equation
2
dY )
( -dx - 2 -dy
dx
+ 4y = 4x - 1 (11 )
IS of first order. A solution containing one arbitrary constant IS of the form

y = x - (x - e)2, ( 12)
as cao be verified by direct substitution. However, this is not tbe most general
solution, since the function y = x also satisfies the differential equation but
canoot be obtained by specializing the arbitrary constant in the solution given.
The additional solution y = x is called a singular solut¡on. Such solutions can
occur only in tbe solution of nonlinear differential equations.
We consider /irsl the result of replacing the function h(x) by zero in Equa-
tion (8). The resulting differential equation, Ly = 0 , is said to be homogeneous,
since each term in tbe equation then involves tbe first power of y or of one of its
derivatives. In this case, from the linearity of the equation, it is e.a sily seen that
any linear combination of individual solutions is also a solution. Thus, if n
linearly independent solutions u,(x) , u,(x) , ... , l/,(x) of tbe associated homo-
geneous equation
LYH = ° ( 13)
are known, the general solution of Equation (13) is of the form
YH(X) = c,u,(x) + c, u,(x) + .. . + e,u.(x) = kL:-• 1 ekuk(x), (14)
where the c's are the n required arbitrary constants. That is, all solutions of the
homogeneou s equation associated with (8) are obtained by suitably specializing
tbe constants in Equation (14).
tWhereas a relationship of ¡he implicit form rp(x , y) = O also would be acceptable as a solu-
lion, there is no need for this generalization when the equation is linear.
6 Ordinary Differential Equations
In this connection, it should be explained that we refer to a function as a solu-

tion of a differential equation in a given interval J if and only if that function
satisfies the differential equation at all poinls of J. Thus, in the case of the homo-
geneous equation
d 2y
----,
d x- = °
we say that the general solution is of the form y = el + C2X. It may be argued
that, since the function y = 1xl is a linear function over any interval not includ-
ing the point x = 0, its second derivative is zero and hence it is a "solution" which
cannot be obtained by specializing CI and e2' However, it is clear that this func-
tion is not differentiable at x = O. Consequently, since the left-hand side of the
equation does not exist at x = 0, the equation is not satisfied at this point, and
y = Ixl cannot be said to be a solutioo over any interval including x = O. Over
any interval not including x = 0, the function y = 1 x I may be replaced by either
+x or -x and hence is obtained from the general solution by setting el =
and either C2 = l or C2 = -l.
°
Now suppose that one particular solution of Equation (8), say y = Yp(x),
can be obtained by inspection or otherwise, so that
LyP = h(x). (15)
Then the complete solution of Equation (8) is of the form
since trus expression contains n independent arbitrary constants and satisfies

the differential equation
Ly = L(Y1t + yp) = Ly" + LY.F = h(x). (17)
Thus it is seen that the process of solving an ordinary linear differential
equation can be divided conveniently into two parts. First, n linearly indepen-
dent solutions of the associated homogeneous equation may be obtained; then,
if any one particular solution of the complete equation is found, the complete
solution is given by Equation (16).
It is frequently convenient to say, "YH(X) is a homogeneous solution of
Ly = h," in abbreviation of the statement, " y = YB(X) is a solution of the as-
sociated homogeoeous equation Ly = O." The term "complementary solution"
also is used.
It will be sbown in Section 1.9 that, if the general bomogeneous solution of
an nth-order linear equation is known, a particular solution can always be
obtained by n integrations. In Sections 1.5 and 1.6 we consider important
special cases in which the homogeneous solution is readily obtained.
1.4. The Linear Differential Equation oC First Order. Tbe linear equation of
first order is readily sol ved in general terms, without determining separately
homogeneous and particular solutions. For this purpose, we attempt to deter-
1.4. The Linear Differential Equalion of First Drder 7
mine an intf'grating factor p(x) such that the standard form
~~ + a ¡(x)y = h( x) ( (8)
is equivalent to the equation

d
-(Py) = ph. (19)
dx
Since Equation (19) can be written in the form
dy ..L
dx'
(1P dx
dp )
y
_
- h(x),
it follows that Equations (18) and (19) are equivalent if p satisfies the equation
1 dp
-d- = a¡(x),
p x
and hence an integrating factor is
(20)
The solution of Equation (19) is obtained by integration:
py = Jph dx + e,
so that the general solution of (18) is of the form
y = -1
p
f phdx +-,
ep (21)
where p is the integrating factor defined by Equation (20), and e is an arbitrary

constant.
In particular, that solution for which y = Yo when x = X o is expressible in
the form
y(x) = IX p(x)
Xo
p(¿;) h(1;,) d¿; + Y p(X o),
o p(x)
(21 ' )
where the dummy variable io the iotegraod is denoted by ¿; in order to distio-

gui sh it from the free variable x, which is to be held fixed in the integration.
Example. To solve the differential equation
x ~~ + (l - x)y = xe>,
we· first rewrite lhe equation in the standard form,
An inlegraling factor is then

8 Ordinary Di/l'erential Equations
no co nstant being added in the integration, since only a particular integratiog factor is
needed . The solutioo is then given by Equation (21),
y =eX
-
x
f eX
xdx + C-,
x
x
•
eX
or y =2
ex + Cx'
It may be noticed that the general homogeneous solutioo of Equation (18)
is YH = Cp-I, whereas a particular solution is yp = p-I J ph dx.
1.5. Linear Differential Equations with Constant Coefficients. The simplest

and perhaps the most important differential equa tion o f higher order in practice
is the linear equation
dny d n- ' · dy
Ly = -dxn + al - -n y
dx - ' + . .. + ano, dx + a.y = h(x), (22)
in which the coefficients a k are constants.

We first attempt to determine n linearly independent solutions of the cor-
responding homogeneous equation. The appearance of the equation suggests
homogeneous solutions of the form e' x, where r is a constant, since aH derivatives
of e'X are constant multiples of the function itself,
d m rx m rx
dxme = r e .
We then have
Le" = (r n + a,r n- I
+ ... + a. _ ,r + a.)e'x. (23)
This result shows that en is a solution of the homogeneous equation associated
with Equation (22) if r is ooe of the n roots r" r" . .. , r. of the characteristic
equation
r' + a,r·- I + . . . + a. _ ,r + a. = O. (24)
It should be noticed that trus equatio n is obtaioed from the associated homo-
geoeous differential equatioo by formalty replacing dkyjdxk by r k , with the coo-
vention that dOy jdx° =y . If the n roots of Equation (24) are distinct, exactly n
independeot solutions e' '', . . . , e' oX of the homogeneous equation are so
obtained and the general horoogeneous solution is
(25)
However, if one or more of the roots is repeated, les s than n independeot

solutions are o btained in this way. To find the missing solutions we may
proceed as follows. Suppose that r = r, is a double root of Equatio n (24).
Then Equation (23) is of the form
Le" = (r - r,)'(r - r,) ... (r - r.)e' x,
and it follows that oot only the right-hand merober itself but also its (partial)
1.5. Linear Differential Equations with Constant Coefficients 9
derivative with respect to r must vanish when r = r,. The same then must be
true for the left-hand member, and hence we conclude that in this case we ha ve
both
and
when we make use of the fact that

k
~ (ake'X) _ a (ae'X)
ar axk - axk Tr .
Thus the part of the homogeneous solution corresponding to a double root r,
can be written in the form
c1e r1x + c?x~rlx = er1x(c t + c 2x).
Bya simple extension of this argument, it can be shown that the part of the
homogeneous solution corresponding to an m-fold root r, is of the form
e"x(c, + c 2x + c,x 2 + ... + cmx m-').
Hence, to each of the n roots of Equation (24), repeated roots being counted
separately, there is a corresponding known homogeneous solution, and the
general homogeneous solution is determined as a linear combination of these
n independent solutions.
Example 1. For the equation

d 3y_dy=O
dx 3 dx '
the characteristic equation is r 3 - r = r(r + I )(r - 1) = O, from which there follows
r = O, ± 1. The general solution is then
Example 2. For the differential equation

•
d 3y d 2y
dx3 - 5 dX2 + 8 dy
dx - 4y = O,
the characteristic equation is (r - 1)(r - 2)2 = O, from which there follows r = 1,2,2.
The general solution is then
If Equation (24) bas imaginary roots and if the coefficients of Equation (24)
•
are real, the roots must occur in conjugate pairs. Thus, if r, = a + ib is one root,
a second root must be r 2 = a - ib. Tbe part of the solution corresponding to
tbese two roots can be written in the form
In order that trus expression be real, the constants A and B must be imaginary.
10 Ordioary Djlferential Equations
By making use of Eu{er'sformu{a,t

e i9 = cos B + ¡sin B, (26 )
\Ve find that the solution becomes
eaX[A (e os bx + i sin bx) + B (cos bx - i sin bx)]
and hence can be written in the more convenient form
eaX(e, cos bx + e 2 sin bx),
where e, and e 2 are new arbitrary constants replacing A + B and ¡(A - B),
respectively. Accordingly, since A = ±Ce, - ieJ and B = ±(e, + ¡e 2 ), real
values of e, and el correspond to values of A and B which are conjugate com-
plexo Similarly, if a ± ib are m-fold roots, the corresponding 2m terms in the
homogeneous solution can be written in the real form
ea'<[(e, + e 2 x + ... + emx m-') cos bx
+ (c m +, +e m+2 x + ... + e 2m xm-l) sin bx).
Example 3. The equation

d 2y
+ 2ddy
dx 2 -x + 5y = O
has the characteristic equatian r + 2r + 5 = O. fram which r
2 = -1 ± 2i; hence
y = e-x(c, cas 2x + e2 sin 2x).
•
has the characteristic equation (r 2 + 1)2 = O, from wruch r = ±i, ±i; hence
y = (e, + e2x) COS x + (e] + e.x) sin x.
•
General methods for obtaining a particular solution of the complete non-
homogeneous Equation (22) are given in Sections 1.7 and 1.9. A shorter method
which can be applied in many practical cases is that of undetermined eoefficients.
This method may be used when the right-hand side of Equation (22) involves
only terms of the form x m, where m is an integer, terms of the form sin qx, cos qx,
and e Px , and/ or products of two or more such functions. The reason for the
suceess of the method is the faet that each of these funetions, or any product of
a finite number of these funetions, has only a finite number of linearly indepen-
dent derivatives.
If we define the fami{y of a funetion f(x) as the set of linearly independent
functions of whieh the funetion f(x) and its derivatives with respeet to x are
tFamiliarity with this important relatian, and with tbe elementary algebra ofcomplex numbers,
is assumed. Sueh topies are reviewed in the preliminary sections of Cbapter 10.
1.5. Linear Differential Equations with Constant CoefficieRls 11
lioear combioations, the following families may be listed:
Term Family
xm xm,xm-J , xm-2, ... ,xz.x, 1
sio qx sin qx, cos qx
cos qx sio qx, cos qx
e PX ep x
The family of a function consisting of a product of n terms of this type is readily

seeo to consist of al! possible products of n factors, io which one factor in each
product is taken from the family of each factor io the parent function . Thus it
may be verified that the famiJy of X2 sin 3x is composed of two-factor products
of terms in the families (x" x, I} and (sin 3x, cos 3x}, one term from each famiJ.y
appearing in each product:
(X2 sin 3x, x sin 3x, sin 3x, x' cos 3x, x cos 3x, cos 3x}.
The method of undetermined coefficients now may be outlined as follows .
It is assumed that the general homogeneous solution of the differential equation
already has been obtained, and that any cosh or sinh functions occurring in it,
or in the right-hand member h(x), are replaced by equivalent linear combina-
tions of exponential functions.
(1) Construct the family of each term (or product) of which h(x) is a linear
combination.
(2) If any family has a member which is a homogeneous s olution of the
differential equation, replace lhal family by a oew family in which each mem ber
of the original family is multiplied by x, or by the lowest integral power of x
for which no member of the new family is a homogeneous solution. Only
members of the offending family are so modified. It should also be noticed,
for example, that the presence of eX or sin x in the homogeneous solution does
not require modification of a family containing the product eX sin x unless that
product itself is also a homogeneous solution.
(3) Assume as a particular solution a linear combination of aH members
of the resultant families, with undetermined literal coefficients of combination,
and determine these coefficients by requiring that the differential equation be
identically satisfied by this assumed solution.
It will be found in all cases that the number of coefficients to be determined

wiU equal the number of linearly independent functions whose coefficients
must be matched, and that the resultant equations always have a solution.
The detailed proof of this general sta tement is rather lengthy and is omitted .
However, the relevant analysis of one typical case is presented , for the purpose
of illustration, at the end of Section 1.7.
It should be emphasized that ¡his procedure does not general/y apply unless
¡he differential equa/ion has constant coefficients and has a right-hand m ember
possessing a finite family.
Example 5. Consider the differential equation

d'y _ dy
dx' dx
= 2x + 1 - 4cosx + 2e'< .
The general homogeneous solution is

YH = C, + c,e x + cJe- x.
The [amUies of the terms x, 1, cos x, and eX on the right-hand side of the equation are,
respectively,
{x, I j, ( 1 J, {cos x, sin x j , {e-<J.
The second [amily is contained in the fust, and is discarded. Since the lirs.t family has
the representative 1 in tbe homogeneous solution, it is replaced by the family {x,, x j.
Similarly, the last family is replaced by (xeJ. A particular solution is then assumed in
the form
yp = AX2 + B x + Ccos x + Dsin x + Exe X.
Wheo y is replaced by yp, the differential equation beco mes
-2Ax - B - 2D cos x + 2C sio x + 2Ee'< = 2x + 1 - 4 cos x + 2e X
and, when the coefficieots of x, 1, cos x, sin x, and e X are equated, there follows
A = -1, B = -1, D= 2, C=o, E=L
A particular solution thus is
yp = - x2 - X + 2 sin x + xe x ,
and the general solution is
y = c , + C2e '< + c,e- x - X2 - X
1.6. The Equidimensional Linear Differential Equation. An equation of the

+ 2 sin x + xe.
•
forro
_ n dny n- l dn-ly dy_
Ly - x dxn + b,x dx n- ' + .. . + bn_lx dx + bny -f(x), (27)
where the b's are constants, has the property that each term on the left is
unchanged when x is replaced by cx, where c is a nonzero constant. Thus the
physical dimension of x is irrelevant in each term on the left and, if the b's are
dimensionless, each term on the left has the dimensions of y. For this reason
we shall refer to this equation as the equidímensional linear equatíon. The
equation is also variously called "Euler's equation," "Cauchy's equation,"
and the "homogeneous linear equation," although each of these terros also
has other connotations.
One method of solving this equation consists of introducing a new In-
dependent variable z by the substitution
z = log x. (28)
1.6. The Equidimensional Linear Differential Equation 13
There then follows

d dz d
- = - - = -Z - ,
1 d
dx dx dz e d::
and hence
Thus, in particular, we obtain
xdy--- dy,
" dx - dz
2
2 d y d 2y dy d (d
X dx2 = dz 2 - dz = dz dz
J dJy dJy d 2y dy
X -dJ=-dJ - 3 -2+ 2 -
x z dz dz
= !!...
dz dz
(!!... - 1) (!!...dz - 2)y,
and, in general, it is found that
xm ~:~ = ;z(;z - l)(;z - 2) ... (~ - m + l)Y. (29)
The transformed equation thus becomes linear with constant coefficients, and
y then can be determined in terms of z by the methods of Section 1.5 if the new
right-hand member is zero or if it has a finite family (with respect to z-differen-
tiation). The final result is obtained by replacing z by log x.
Example 1. To solve the differential equation

d 2y dy ,
X2 dx 2 - 2x dx + 2y = x- + 2,
we make use of Equalions (28) and (29) 10 oblain the Iransformed equation
d 2y dy
-dz 2 - 3 -dz'..L "y
~ -- e 2z + 2.
The solution is found, by the methods of Section 1.5, to be

y = e,e z + e2e2z + ze 2z + 1,
or, in lerms of Ihe original variable x,
y = e,x + e2x2 + X2 log x + 1.
•
lf the right-hand member is zero, a more convenient alternative procedure
consists of directly assuming a homogeneous solution of the form
YH=" x',
) corresponding to the assumption YH = e" III the transformed equation. By

1-'1 Ordinary Differential Equations
making use of the relationship

dmx'
xm -xm
d = r(r - 1) ... (r - m + l)x',
there follows, with the notation of Equation (27),

Lx' = ([r(r - 1) ... (r - n + 1)]
+b¡[r(r-I)···(r-n+2)] - ... + bn_¡r + bn}x'.
Hence x' is a homogeneous solution if r satisfies the characteristic equation
[r(r - 1) ... (r - n + 1)] + b¡[r(r - 1) ... (r - + 2)]
n
+ ... + b ,r + b n- n = O. (30)
This eq uation can be obtained from the left-hand si de of Eq uation (27) by
formally replacing xm(dmy(dx m) by the m-factor product
r(r - 1) ... (r - m + 1).
Let the n roots of Equation (30) be denoted by r" r 2 , • • . , r n. If these roots are
distinct, the general homogeneous solution is of the form
(31 )
In analogy with the results of Section 1.5, we find that the second homo-
geneous solution corresponding to a double root r¡ IS
[ .E....
ar
(x')]
'-'1
= x" log x
and th", part of the homogeneous solution corresponding to an m-fold root r¡ is

x"[c, + C2 log X + C3 (log x)2 + ... + Cm (log x)m-¡J.
Further, to a conjugate pair of imaginary roots r = a ± ib there corresponds
the solution
x"[c¡ cos (b log x) + C 2 sin (b log x)].
The extension to the case of repeated imaginary roots is obvious.
Except in those cases in which the right-hand member is a linear combina-
tion of powers of x (and in certain other cases of little practical interest), par-
ticular solutions of nonhomogeneous equations of type (27) usually cannot
be obtained by the method of undetermined coefficients. However, it is readily
shown by using the substitution (28) that a particular solution corresponding
to a right-hand member of the form x' is given by yp = Ax" where A is a
constant to be determined by substitution, unless x' is a homogeneous solution.
If x' is a homogeneous solution, the trial particular solution should be of the
form yp = Ax'(Iog xY, where k is the smallest positive integer for which this
expression is not a homogeneous solution. In other cases, particular solutions
can be obtained by the methods of Sections 1.7 and 1.9.
1.7. Properties of l.inear Operators 15
X O d 'y _ ~ dy ' 2 ' + 2

dX2 ~.x dx"" Y = x- I
of Example 1. Ihe c haracteristic equation (30) beco mes

r(r - 1) - 2r +2= r' - 3r +2 = O,
from which r = 1, 2. The homogeneo us solution is thus
YH = C1X + C2X2.
Since x' is a homogeneouS SOlution, we assume a solution y p = Ax' log x correspond-

ing to the right-ha~d term x' and a solution yp = B corresponding to the constant
term, and hence wnte
yp = A x' log x + B.
Substitution into the given differential equation gives A = B = 1, and the complete
solution is
y = c,x + c,x' + x' log x + 1.
•
1.7. Properties oC Linear Operators. We now consider more cr itically
certain properties of linear differential operators o f the general form
d" d"-' d
L = a o( x ) dx" + a,(x)
dx· -'
+ . .. + a.- ,(x)-d
x
+ a,.(x). (32)
An expression of ttus Sort has no intrinsic meaning by itself, but when it is

foUowed by a function u(x), the result Lu is defined to be a new function of x
defined by the relations hi p
Lu == d"
( a o -d
Xli
+ d"-'
a 1 dx" I
+ ... + G __ ,
..
d
dx +)
a" u
-
d"u
a o dx" + a,
d"-'u
+ ... + du
a ,,_, -d
+ a u
" .
dx" -' x
We speak of Lu a s the result of operating on u by the operator L.
If L, and L, are two linear operators, we say that L \ and Lz are eqtlal when
L,u = L , u for every functi o n u for which the operations are defined. In addltlOn,
we write L,L\u to IDdlcate the operation L , (L\u), that is, {he res ult of operatmg
on L,u by L " and slmllar ly for three or more successive operations. The abbre-
viatioos L ' u ==LLu, L3 U "= LLLu, and so 00, are frequeotly u sed . In particular,
if the operator d/ dx IS written as D ,
d (33)
D=- ,
dx
there follows
D' -
-
- -d (d) --
dx dx dx "
d2
D3 _ .!:!..
2
(d ) = dd '3 ' ... ,
- dx dx' x
16 Ordinary Dilfereotial Equations
and in general Dm = dm /dxm. Thus Equation (32) can be written in the equiva-
lent form
L = ao(x)D· + a, (x) D·-l + ... + a._1(x)D + a.(x)
(34)
The operations LzL1u and L1Lzu should be carefully distinguished from

each other, since the two operations are not, in general, equivalent. To iIIus-
trate, let Ll = d/dx and Lz = x(d/dx). Then
2
d u
d (du)
LzL1u=x- -
dx dx
= x--,
dX2
2
d (dU) d u du
and L1L1u = dx x dx = X dX2 + dx '
If the order in which the operators Ll and Lz are applied is immaterial, that
is, if L1Lzu = LzL1u, the two operators are said to be commutative. SimilarIy,
we say that a set of operators is commutative if each pair of operators in the
set is commutative.
It is clear that any two operators of the fonu Dm and D· are commutative;
so also are two operators of the form amDm and a.D', where a m and a. are
constant. From this fact it follows easily that the set of linear operators with
constant coefficients is commutative.
The commutativity of two linear equidimensional operators, for which
ak(x) = bkxk, is seen to depend upon the commutativity of any two operators
of the form Ll = bmxm(dm/dxm) and L 2 = b.x'(d·/ dx·). But with the substitution
(28), Equation (29) shows that Ll and L z become linear operators with constant
coefficients, and hence are commutative. Thus it follows that the set of equi-
dimensional linear operators is commutative.
It may be remarked, however, that commutativity is the exception rather
than the rule. Thus, for example, the aboye ilIustration shows that linear
operators with constant coefficients and equidimensional linear operators are
not in general commutative with each other.
The distributive property of linear operators,
(C1Ll + C L2 + ... + cnL.)u =
Z c1L1u + c 2L zu + ... + c.Lnu, (35)
as well as the distributive property of linear operations,
L(C1U 1 + CzUz + ... + cnu n) = c1Lu 1 + c 2Lu 2 + ., . + c.Lu., (36)
of wruch use has already been made, is easily established.
In many cases it is possible to factor a linear operator into the product of n
linear factors. If the factors are commutative, the result of factoring is unique,
the order in which the factors are written being arbitrary. Otherwise, the com-
ponent factors will differ in form according to the position that they occupy in
the product.
1.7. Properties oC Linear Operators 17
To illustrate, the operator DO - 3D + 2 can be factored uniquely in the

forms (D - 2)(D - 1) = (D - I)(D - 2). However, the operator XD2 + D
factors in two ways, into the products (D)(xD) and (xD + I)(D). By this state-
ment we mean, of course, that
(DXxD)u = (xD + IXD)u = (xD' + D)u
for any twice-differentiable u.
Use is frequently made of the factoring process in solving linear differential
equations. Thus, in the case of the homogeneous linear equation with constant
coefficients, the operator
L = Dn + a, Dn -, + .. . + a n_, D + an
can be factored uniquely into the linear factors
L = (D - r,XD - r z ) ... (D - r n),
where r" rZ, ... ,r n are roots of the formal equation L = O. Thus the equation
LYH = O becomes
(D - r,XD - r 2) ... (D - rn)YH = O,
where each operator operates on the express ion to its right. Hence the com-
plete expression will be zero if the result of the first operation is zero. Since any
one of the n operators can be written immediately before y, it follows that a
solution of any one of the n equations
(k = 1, 2, ... , n)
is a solution of LYH = O. But these equations are equivalent to
(k = 1,2, ... , n)
and are readily solved to give the solutions

(k = 1,2, ... , n).
By superimposing these solutions, the general homogeneous solution is ob-
tained in the case where the roots are distinct, in accordance with the results of
Section 1.5. The part of the solution corresponding to m-fold roots can be
obtained as the solution of (D - r,)mYH = O.
Analogous procedures can be applied in other cases. In particular, the
general solution of any linear differential equation with constant coefficients
and arbitrary rigbt-hand side can be obtained by a method illustrated by tbe
following example.
Example. To solve the differential equation

d 2y dy
dX2 - 3 dx + 2y = h(x),
we write the equation in the operational form

(D - 2)(D - l)y = h(x).
With the definition

y, = (D - I)y,
the differential equation becomes
(D - 2)y, = h(x) or
This linear equation is of first order and is solved, by using the results of Section 1.4,
in the form
y, = e 2x f e- 2x h(x) dx + e,e 2x .
Next, replacing y, by (D - I )y, we obtain a second first-order equation,
dy
dx _ 'v = e'·' f e-:xh(x) dx + e , el'..
with solution
Similarly, it can be shown that the linear equidimensional operator of

Equation (27) can be factored into the commutative factors
(xD - r¡XxD - ro) ... (xD - rn)'
where rJ> r 2 , • • • , r n are the roots of the characteristic equation (30). In this
connection it should be noticed that the operators
are no! equivalent.

The notion of operators ¡S useful in establishing (he general validity of the
method of unde/ermined eoeffieiel1/s described in Section 1.5. To illustrate the
argument, we here consider an equation of the form Ly = a cos qx, where Lis a
linear differential operator with constant coefficients. Since the operator D2 + q2
annihilates the right-hand member, it follows that all solutions of the given
equation are included in the general solution of the equation L*y = O, where
L* = (D2 + q2)L.
However, if D2 + q2 is not a factor of L, then the general solution of this equation
15
y = YH + A cos qx + B sin qx,
where y = YH is the general solution of Ly = O. If DO + q2 is an unrepeated
factor of L, it is a double factor of L* Since cos qx and sin qx are already present
in YH, A cos qx + B sin qx then must be replaced by Ax cos qx + Bx sin qx, and
so forth, in accordance with the rules set down in Section 1.5.
1.8. SimuItaneous Linear Differential Equations. Frequently, two or more

unknown functions are re1ated to a single independent variable by an eq ual
Dumber of linear differential equations. Thus, in the case of two unknown
1.8. 5imultaneous Linear Differential Equations 19
functioos X and y and the indepeodent variable 1, we may ha ve a pair of

simultaoeous equations of the form
L¡x + LoY = h¡(l), (37a)
(37b)
where the L ' s are linear differential operators in t. The unknown fuoction s x
and y are to be determined as functions of l.
When the operators iovolved are commutative, aU unknown fuoctions
except ooe can be eliminated successively from the given set of equations to
give a new set of hnear differential equations, each involving only one unknown
fuoctioo. We iUustrate this procedure in the case of Equations (37a, b).
If Equatio n (37a) is operated 00 by L. and (37b) by -L z , and if the resul-
tant equations are added, there follows
(L.L J - L , L,)x + (L.L , - L 2L.)y = L.h ¡ - L 2h 2 ,
or, if L , and L. are commutative,
(L.L, - L 2 L , )x = L.h, - L 2 h 2 · (38a)
Similarly, to eliminate x we operate on Equation (37a) by -L, and on (37b)
by L, . If L, and L, are also cornmutative, we then obtain, by addition,
(L,L. - L,L,)y = L,h, - L ,h, .
Finally, if L" L. and L" L, are commutative, the operators of x and y in the
last equations are identical, and we have
(L.L, - L,L,)y = L,h , - L,h,. (38b)
Equations (38a) and (38b) can be written formally ID the determinan tal
form
~x
h, L,
= 1 (39a)
h 2 L4 1 '
~y =1
L, h,
(39b)
L, h, 1'
where ~ is the operator

L,L, L'I
L4 '
(40)
if it is understood that in each term of the expansion of the right-hand sides of

Equations (39a, b) the operator is to be written befo re the function operated
upon. The formal analogy with Cramer's rule for solving linear equations by
determinants should be noticed.
Since the same operator affects x aod y in Equations (39a, b), it is seen that
the homogeneous solutions of these equation s are linear combinatioos of the
same functions, the number n of indepeodent constants in each linear combioa-
tioo being equal to the degree of the operator~. Thus the solutions of Equations
(39a, b) will contain 2n independent constants.
At this stage of the solution a certain amount of care must be taken. It is

clear that, sine e Equations (37a, b) imply (39a, b), al! solutions of the original
simultaneous equations are co ntained in the solutions of the final equations.
However, since differentiation generaUy is involved in obtaining (39) from (37),
the converse is not generally true, in the sense that the solutions of (39a, b)
may satisfy (37a, b) only if certain relationships exist among the 2n constants.
These relationships may be determined by substituting the solutions of (39)
into (37a, b) and requiring that the resultant equations be identities. However,
if the coefficients are constants, and if in one of Equations (37a, b) the two
operators involved have no common factors, the relationships are completely
determined by s ubstitution into thal. single equation (see Problem 31). If X p
and yp satisfy Equations (37), only the added Xu and YH need be so checked.
An alternative procedure consists of solvi ng only one of Equations (39a, b) for
one unknown function and of then su bstituting this re sult into whichever of
(37a, b) is more convenient fo r the subsequent determination of the second
unknown function. The expressions so obtained are then introduced into the
remaining one of Equations (37a, b) to determine possible restrictions on the
arbitrary constants.
The extension of this procedure to cases in which more than two unknown
functions are present leads to results again completely analogous to the state-
ment of Cramer's rule. Thus, assuming that aH operators involved are commu-
tative with each other, the solutions of the equations
L,x + L2y + L,z = h,(t),
L 4x + L,y + L.z = h 2 (t),
L7x + LBy + L.z = hJ(t)
are also solutions of tbree linear differential equations each involving only one
dependent variable, one of which can be written formaHy as
L, h, L,
~Y = L4 h2 L. ,
L, h, L.
L, L2 L,
where ~= L4 L, L. ,
L, L. L.
if in each term of the expansion of the first determinant the function is written
after the operators . If the operator 8 is of order n, the solutions of the three
equations so obtained involve 3n arbitrary constants, and possible restriction
on these constants must be obtained by substitution into the original equations.
This procedure may be quite laborious if several unknown functions are present.
A useful check is provided by the known fact that, if all operators involved
are commutative, the total number of independent constants present in the solu-
1.8. Simultaneous Linear DitIerential Equations 21
lion O/ a ser o/ linear differenlial equalions is equal fa the arder o/Ihe operalor ~.
This order canoot exceed the sum of the orders of the several equations and in
special cases may be less than this number.
A method of solving such sets of equations with reduced labor io certain
problems, in cases when the operators have constant coefficients, is presented
in Chapter 2.
To illustrate the preceding method, we solve the equations
d 2x
dt' - x - 2y = 1, (4Ia)
d 2y
dt ' - 2y - 3x = l . (41 b)
In operational form, these equations beco me

(D2 - I)x - 2y = 1,
-3x + (D2 - 2)y = 1.

Equation (40) gives
D'- -2
~ =1 -3 D2 - 2
1 = D4 - 3D2 - 4,
and Equations (39a, b) then become

2
(D4 - 3D2 - 4)x = 1t D2- 21 = (D2 - 2)1 + 2 = 2 - 2t (42a)
and
D2
(D 4 - 3D' - 4)y = 1
-3
~ 1 = (D' - 1)1 + 3t = 31 - 1. (42b)
We notice that the characteristic equation for both x and y is obtained by

formally replacing D by r in the express ion for ~ = O,
1'4 - 3r 2 - 4 = 0,
from which r = ±2, ± i. Hence we obtain
x H = c ,e 2' + C2e-2, + CJ cos t + C4 sin 1,
YH = d,e' t + d e - 2t + d cos t
2 J + d 4 sin l.
Parlicular solutions of Equatioos (42a, b) are read ily found by inspection
o r by the method of undetermined coefficients,
yP = -l- - tf .
To determine the relationships which must exist among the c's and d's, we
may first verify that X p and yp satisfy Equations (41a, b) as well as (42a, b) .
Hence we then introduce the expressions for X H and J'H ioto the left-hand sides
of Equations (41a, b) and require the results to be identically zero. Using first
Equation (41a), we find the conditions

dI = tel)
The same conditions are obtained by using Equation (4Ib) (see Problem 31).
Thus four of the eight constants are truly arbitrary. Retaining the four c's,
we write
YH = tecle"~ + c 2 e-") - (c) cos 1 c. sin 1),

and the final solutions are
x = c l e2' + c 2 e-o, + c) cos 1+ c. sin 1+ ±I - ±,
y = tecle2' + c,e- 2 ') - (c) cos 1+ C4 sin 1) -11 + t.
JI may be se en Ihal Equalion (42a) could be oblained direclly, in this case,
by solving (41 a) for y and subsliluling Ihe resull inlo (41 b). In more complicated
cases the present procedure is generally preferable.
If the expression for X H were introduced into the left-hand side of Equation
(41a) and the right-hand si de were replaced by zero, an expression for YH would
be found directly, in this case, in terms of the constants of X H . Substitution
of these results jnto the left-hand side of Equation (4Ib) would then show that
no further restrictions on the constants were necessary. If Equation (41 b)
were used to determine YH in terms of X H , the two new constants introduced
would be determined in terms of the c's by substitution into (41a).
The solutions of Equations (4Ja, b) can also be obtained by a slightly
different but equivalent method which is of sorne practical interest. Since the
coefficients in both linear equations are constants, jt can be assumed initially
that homogeneous solutions exist of the form
By introducing these assumptions into Equations (4Ia, b) and replacing the

right-hand sides by zeros, we obtain the conditions
(d - l)c k - 2dk = O,
-3c k + (rr - 2) d k = O.
In order that nontrivial solutions of these equations exist, it is necessary that
the determinant of the coefficients of C k and d k vanish, giving the characteristic
equation obtained previously. If r k satisfies this equation, the coefficient d k can
be expressed in terms of C k by either of the two equations. Thus, using the first
equatian, we have
dk =
rr -2 1
ck •
For the roots r = ±2 there follows d k = 1Ck' whereas for r = ±i there follows
d k = -c k • These results are easily shown to lead, by superposition, to the
previously obtained homogeneous solutions.
8
}.. S ímultuneous Línear Differential Equations 23
In order to obtain particular solutions directly from Equations (41a, b),

the method of undetermined coefficients can be applied if al! lerms on the
right-band sides 01' the equatioos are taken into account in constructing the
families . Thus, from Equation (41a) we have the family (1, l} and from (4Ib)
the family {l}, wbich is contained in the former family. Since there is 00 repre-
seotative in either bomogeoeous solution, we assume particular solutions of
the form
x p = Al + B, yp = CI + D.
Substitution inlo Equations (4Ia, b) gives

-(A + 2C)1 - (B + 2D) = 1,
-(3A + 2C)1 - (3B + 2D) = l.
lo order that these be ideotities, we must have
-A - 2C = 1, B + 2D = O, 3A + 2C = O, -'>,8 - 2D = 1,
from which there follows
A -l
- "2:' B = --1.-, C= -i, D =·h
in accordance with the previously obtained results.
lo order to iIlustrate a special situation, we coosider also tbe set
dx
2- -
dI
3x +y = 4e'
'
(43)
dy
x + 2 dI - 3y = O.
If Equations (39) and (40) are applied to these equations, there follows
2
d 2x _ 3 dx + 2x = -e',
dl dI
(44)
2
d y _ 3 dy + 2y = -e',
dl 2 dI
from which one obtains
X H = e,e" + e e', 2
(45)
YH = d,e" + d2e'
and
Xp = te l ,
t (46)
yp = te' .
But here it happens that x p and yp do nol satisfy the original Equatíons (43).
Thus it is necessary to substitute the sums X H + X p and YH + yp into Equations
s (43) for the purpose of obtaining conditions 00 the constants io Equations (45).
This process gives , fioal!y,
d2 = 2 +e 2, dI = -el' (47)
24 Ordinary Differentlal Equations
1.9. Particular Solutions bv Variation oC Parameters. We next derive a

method for determining the co~plete solution of any linear differential equation
for which the general homogeneous solution is known.
Suppose that the general homogeneous solution of the equation
clny dn-1y dy
Ly = -dxn + a (x) - -
1 dxn- 1
...!....
'
•• , + an_l(x) d-
X
+ an(x)y = h(x) (48)
has been o btained in the form

n
Yif = I:
k=1
CkUk(X), (49)
where the u's are n linearly independent homogeneous solutions and the e's
are n arbitrary constants or "parameters." We wiIl find that a particular so lution
of the complete equation can be obtained by replacing the constant parameters
e k in the soluti on of the associated homogeneous equation by certainfunetions
of x. Thus we assume that
n
yP = I:
k=l
Ck(X)Uk(x) (50)
is a solution of Equation (48) and attempt to choose the n functions C k suitably.

Since we ha ve n functions to determine, and since the requirement that Equa-
tion (50) satisfy Equation (48) represents only one condition, we ha ve n - 1
additional conditions at our disposaL
Differentiating Equation (50) and usi ng primes to denote differentiation
with respect to x , we obtain
n
CkU~ + kL- I C~ Uk ·
In order to simplify this expression, we require as our first condition that the
second summation vanish,
(5Ia)
There then foIlows
and
As the second condition, we require again tha t the sum of the terms involving
derivatives of the C's vanish,
t
k- I
C~u~ - O. (51b)
Proceeding in this way through the (n - I)th derivative, we have as our (n - l)th
condition the requirement
t
k- I
C~u1n-ZI = O (SIc)
1.9. Particular Solutions by Varialion of Parameters 25
and the (11 - l)th derivative is

d" - Iyp
dx" I
The expression for the nth derivative is then
d :
dx
fly
= i: CkUk")
k~1
i:
k- I
C~u1n-lJ.
By introducing the expressions for yp and its derivatives into the left-hand
side of Eguation (48). we find that the final condition, Ihat Eguation (50)
satisfy (48), becornes
" n n
+ an_l(x) L:
k=l
CkU~ + an(x) L: k=(
CkU k + L:
k~l
C~u1n;-IJ = h(x).
Cornbining the first surnrnations, we obtain

n [dnu dn-IU k
2:
k- t
C, d :
X'
+ al(x) dx:n 1
J
+ ...
+ a n - ¡(x) ~; + aix)Uk] + t C~u1n-IJ = h(x).
' -1
Now, since each function u, satisfies Eguation (48) with h(x) replaced by
zero, and since each bracket in the first surnrnation is precisely the result of
replacing y in the left-hand side of Eguation (48) by a function L/k' the first
surnmation vanishes identically, and the final condition becornes rnerely
i: C~ui"-I J =
k - I
h(x). (51d)
In surnrnary, the n conditions irnposed on the n unknown functions can be

written in the expanded forrn
C~ (X)uI(X) + C~ (X)U2(X)+ + C~(x)un(x) = O,
C~ (x)1I\(x) + C~(x)u~(x) +
C~(x)u~(x) = O, +
. .. ... .... .. .... .. ...... ... ........ .. . ... ... ... ... .. .. (52)
+
C; (x)u\n-2>(x) C~(x)ur-2 ) (x) + + C~ (x)u~n-2J(x) = O,
C~ (x)u\n- 1 J(x) + C~ (x)uin- 1 ,(x) + + C(x)u~n-I)(x) = h(x).
If this set of eguations is solved for C~, C~, . . . , C~ by Cramer's rule, the
common-denominator determinant is seen to be the Wronskian of ti 1> U 2 ' . . . ,
un·t
If the solutions C; , C~, ... , C~ are integrated and the results are introduced
into Eguation (50), the result is a particular solution of Eguation (48) for any
tIt is known that, if the coefficients 01,02, . . . , On in the standard form (48) are continuous in
an intervall, Ihe indicated derivatives ,,1:)
exist in 1, and furthermore the Wronskian of Ihe
linearly independent fuoclions cannol vanish in l. Henco a unique solution of (52) then exists.
26 Ordinary Oilferential Equations
choice of the n constants of integration. If the constants are lef¡ arbitrary, this
procedure yields the complete solution of Equation (48).
lt is important to notice that Equation (48) was written in "standard form."
If the coefficient of ti"y/ dx" in Equation (48) were Go(x), the last equatioD of (52)
would be modified by replacing h(x) by h(x)/ ao(x).
In particular, for a second-order linear equation of the form
d 2y
dx
2 + a ,(x)_ d-
dy...L
x
,
_
a,(x)y - h(x), (53)
there follows
y = e,(x)u,(x) + e 2(X)U 2(x), (54a)
where
h(x)u,(x)
W[u ,(x), u,(x)]
and, similarly,
e~ = .,..,..,-,-h.>..:(x7-'):';":U-'-(>..:.x'-",)~
'
W [u, (x), u,(x)]
Thus we can write
+c
e, = -
J h(x)u,(x)
W[u ,(x), l/,(X)]
dx
"
e , -- J h(x)u,(x) d
W[u,(x), u 2(x)] X
+ c,
(54b)
and the introduction of these results into Equation (54a) gives the required
solution.
If h(x) is not given expLicitly, but tbe general solution is required for an
arbilrary h(x), we may combine the result of this substitution into a more
compact form if, before substituting Equations (54b) into (54a), we denote tbe
dummy variable of integratioD by a new symbol ~, to distinguish it from the
current variable x which now will appear in the integrand (as well as in the upper
limit of integration). Substitution of Equations (54b) into (54a) then leads to
tbe result
= JX h(~)[u,(~)ulx) - u,(~)u,(x)1 dJ: + c U (x) +c u (x) (55)

Y W[u,(<;), u,{.;)] ,>, , - , , - ,
where a convenient lower limit is assumed in the integral. If h(x) is given expli-
citly, the direct evaluation of Equations (54b) and subsequent substitution into
(54a) is usually more convenient tban the use of Equation (55).
Example l. For ¡he differential equation

d 2y
dx' +y = h(x),
s 1.9. Particular Solutions by Variation oC Parameters 27
s tWO linearly independen [ homogeneous solutions are ti I = cos X, U2 = sin x. The

Wronskian is
" W(cos x, sin x) =
cos x sin x
. = sin 2 x + cos 2 :c = !. I
) -sin x cos x I
Thus use of Equations (54a,b) gives the solution
y = -cos x[f h(x) sin xdx + cl ] + sin x[f h(x) cos xdx + Cl]'
)
This form is usually most convenient for actual evaluation of the solution when h(x)
is given. The form (55), which is useful in more general considerations, here becornes
)
y = r h(e;)[cos e; sin x - sin e; cos xl de; + c, cos x + c, sin x
or y = fX h(e;) sin (x - e;)de; + C,COSX + C2SinX.

•
Example 2. For the differential equation
d'y d'y dy
dx' - 3 dx2 + 2 dx = h(x),
we may take u, = 1, u, = e, ti, = e 2 , . Equations (52) then become
C~ + C~e:c. + C~e2x = O,
C~ ex + 2C~e2 .~ = 0,
C~ eX + 4q e 2 , = h(x).
For tbe determinant of this system we find W(I, eX, e 'x) = 2e". Solving the three
simultaneous equations, we obtain
d e; = !h(x), c~ = -e-'h(x), e; = irZ,h(x).
The solution of the differential equation is then
n
'e y = J[± f h(x)dx + c,] + e[ - f e-'h(x)dx + C2] + e 2X [± f e- 2x h(x)dx + Cl]
te or, equivalently,
le
:r
o
•
It will be shown in Section 1.10 that the Wronskian of two homogeneous
solutions of Equation (53) is of the form
i)
(56)
l- where A is a specific constant depending only on the choice of tbe arbitrary
o multiplicative constants in volved in tbe homogeneolls solutions u, and u 2 •
(See also Problem 8.) It follows (see Problem 45) that W(u" L/2) can be deter-
mined if only tbe values of u I and U 2 and tbeir first derivatives are known at a
single point. This fact is useful in determining the integrand in Equation (SS)
if, for example, tbe solutions tI,(x) and u 2 (x) are expressed in terms of power
series (see Chapter 4).
28 Ordinary Dilferential Equations
It can be shownt that, more gene rally, the Wronskian of n homogeneous

solutions of Equation (48) is also given by the right-hand member of Equation
(56). The statement of this fact is known as Abe!' s formula. From the properties
of the exponential function, it follows that, if a ¡(x) is continuous in an interval
1, the Wronskian cannot vanish in 1 unless it vanishes identically.
1.10. Reduction of Order. One of the important properties of linear differ-

ential equations is the fact that, if one homogeneous solution of an equation of
order n is known, a new linear differential equation of order n - 1, determining
the remainder of the solution, can be obtained. This procedure is in a sense
analogous to the reduction of the degree of an algebraic equation when one
solution is known.
Suppose that one homogeneous solution u ¡(x) is known. We next write
y = v(x)u¡(x) and attempt to determine the function v(x). Substituting vu¡ for
y in the left-hand side of the differential equation, we obtain a new linear
differential equation of order n to determine v. But since y = cu¡(x) is a homo-
geneous solution of the original equation, v = e must be a homogeneous
solution of the new equation. Hence the new equation must lack the term of
zero order in v; that is, the coefficient of v must be zero. Thus the new equation
is of order n - 1 in the variable dv/dx.
We apply this procedure to the solution of the general second-order linear
equation
2
d y dy ( ) (
dX2 + a¡(x) dx + a2 x Y = h x), (57)
assuming that one homogeneous solution u ¡ (x) is known. Writing

y = v(x)u¡(x), (58)
and introducing Equation (58) into (57), there follows
v"u¡ + 2v'u'¡ + a¡v'u¡ + v(u'; + a¡u'¡ + a 2u¡) = h.
But since ti, is a homogeneous solution of Equation (57), the expression in
parentheses vanishes and the differential equation determining v becomes
or (v ')' + (2 -u', + a¡ ), _
U 1
V h'
--
U¡
(59)
This equation is of first order in v', with an integrating factor given by the
results of Section 1.4 in the form
210&ul+Sa¡dx _ pu2
e - l'
where (60)
tBy differentiating tbe Wronsk.ian determinant, one obtains the reJation dW¡dx = ~a ¡ W.
1.10. Reduction oC Order 29
Hence there follows
v ' = _12
PU,
f phu , dx + pU,
-.S,.
An integration then gives
f phu, dx dx + e,
V =
f PU,
2
and the introduction of Equation (61) into (58) yields the general solution
f PUf
dx
+e 2 (6J)
f phu, dx dx + e,u,(x) f d x
Y = u,(x)
f PU,
1
Thus, if U 1 is one homogeneous solution, the most generallinearly indepen-

PUl
2 + C 2U I (X). (62)
dent solution of the associated homogeneous equation is of the form
U2 = Au,(x) f dX2
PUl
+ Bul(x), (63)
where A and B are arbitrary constants with A * 0, and a particular solution

of the complete equation is
f phu dx
f
I
y = u,(x) 2 dx. (64)
PUl
We remark that no constants of integration need be added in either of the
integrations in Equation (64), since the additional terms thereby introduced
can be absorbed into the homogeneous solution. However, if arbitrary con-
stants are introduced in each of (he integrations, Equation (64) represents the
complete solution of Equation (57).
In view of Equation (63), the Wronskian of any two homogeneous solutions
of (57) is given by
U, AUI f dx
pur
...l... BU I
f
I
W(U" u,) =
-A +
PU,
A U¡' -2 I
PUl
dx , B U '1
Expansion of this determinant gives Abel's result, in this case,
TU(U u) - ~ - Ae-Ja, (X) d X (65)

" ' " 2 - p( x) - .
Example. One homogeneous solurion of (he equarion

2
X2 d y
dX2
+ (x _ ])(x dy _ y) = x2e- ·'
dx
is seen by inspection to be y = x. To fínd the complete solution, we fírst write the equa-
lion in ¡he standard form of (57),
~:.~ -L (1 _ ;) ~~ '- 1;-; x y = e,

and find
a,(x) = I -X-'
With ", = x, a second linearly independent solution is obtained from Equation (63),
taking A = 1 and B = O for convenience, in the form
f
e-X
U2 = x xdx.
The indefinite integral which appears here cannot be evaluated in terms of elementary
functions. However, the function
Ei( -x) = IX
=
e-~ dI;, =
;;,
J -<
-~
e~ dI;,
;;,
(x> O)
is a tabulated function, known as the "exponential-integral function." [See Reference

8 of Chapter 4, for example, where similar integrals involving (sin xlIx and (cos xl/x,
and known as "sine-integral" and "cosine-integral" functions, are also defined and
tabulated.] Thus the second independent homogeneous solution can be taken to be
Uz = xEi( -x),
and the general homogeneous solution is of the form
YH = x[e, + c2Ei( -x)].
A particular solution is given by Equation (64),
' f
e-X e-Xe X dx
J'
YP = x
J x dx = x e-" dx = -xe- X ,
if constants of integration are omitted. Thus the complete solution is

y = x[e, + c2Ei( -x) - e-X]
1.11. Determination oC Constants. The n arbitrary constants present in

(x> O).
•
the general solution of a linear differential equation of order n are to be deter-
mined by n suitably prescribed supplementary conditions.
Frequently, these conditions consist of the requirement that the function
and its first n - l derivatives take on prescribed values at a given point x = a,
y(a) = Yo, y'(a) = y~, ... , y"-lJ(a) = y¡,,-lJ. (66)
When such conditions are prescribed, the problem is known as an initial-value
problem. In this case it is true that if the point x = a is inc/uded in an interval
where the eoefficients a, (x), ... , a,(x) and the right-hand side h(x) of the differ-
ential equation, in standard form, are eontinuous, there exists a unique solution
satisfving Equations (66). Here, if the complete solution is written in the form
,
Y I: ekuk(x) + yP(x), (67)
k'='l
1.12. Special Solvable Types of Nonlinear Equations 31
the conditions of Equations (66) require that the constants Ck satisfy the n
equations
i; CkU~m)(a) = y&m¡ - y~mJ(a) (m = 0,1,2, ... , n - 1). (68)

k=l
We notice that the determinant of the coefficients of the constants C k is the

vaIue of the Wronskian of the linearly independent functions U k at the point
x = a, which (see footnote on page 25) cannot vanish under the specified
conditions. Thus a unique solution is assured.
Sets of conditions other than those of Equations (66) may, however, be
prescribed. For example, values of the function andjor certain of the derivatives
may be prescribed at two distinct points x = a and x = b, the probJem then
being known as a boundary-value problem. In such cases a solution may or may
not exist, and may or may not be unique.
Example. The general solution of the differential equation

d 2y
dX2 +y = O
is y = el COS x + e2 sin x.
The initial-value problem with conditions y(O) = Yo,
y'(O) = y~ has the unique solution Y = Yo cos x + y~ sin x. The condilions y(O) = 1,
y(n/2) = 1 imply the unique solution Y = cos x + sin x, while lhe only solution satisfy-
ing the conditions y(O) = y(n/2) = O is the trivial solution Y = O. However, the condi-
tions y(O) = yen) = O are both satisfied if we take C I = O, and hence in this case there
exist an infinite number of solutions of the form Y = A sin x, where A is arbitrary .
•
1.12. Special Solvable Types oC Nonlinear Equations. Although there exist
no techniques of general applicability for the purpose of obtaining solutions
of nonlinear differentiaI equations in c10sed form, there are severaI special
types of equations for which such solutions can be obtained, a few of which
are treated very briefly in this section. Here, instead of seeking y as a function
of x, we may be led to determine x as a function of y or to accept as a solution
a functionaI relationship involving the two variables in a less simple way.
(1) Separable Equations. Separable first-order equations have already been

mentioned in Section 1.1.
Example l. The equation (1 + X2) dy + (I + y2) dx = O IS separable in the form

dx dy
1 + X2 t- 1 + y' = O,
and integration gives the solution
tan-! x + lan-! y = e
or x + y = dI - xy) (e' = tan e).

•

dy ) , +
( -dx - 4y 4 = O
yields Iwo separable equations when solved algebraically for dy!dx,

dy ~ 1-2
dx - vIY - 1
,
from which +-qy- =? dx

- "¡y-l - ,
provided Ihat Ihe division by ..¡ y - J is legitima te, and hence there follows ± .- / y 1
= x - e or
y = J + (x - c)2.
Since the relation y - 1 = O has been excluded io the derivation of this solution, the
possibility that y = 1 ma,y"also be a solution must be explored separately. Direct sub-
stitution into the original differential equation shows that y = I is indeed a solution.
Furthermore. it cannot be obtained by specializing the conslant e in the relatioo y =
1 + (x - C)2. Here the complete solution consists of the ¡atter one-parameter solution
together with the "singular solution," y = 1. lt cao be verified. in this case, that all the
curves which represent particular solutions, in correspondence with particular choices
of the constant e, are tangent to the straight line representing the singular solution. so
that this line is the envelope of those curves. •
(2) Exact First-order Equations. A first-order equation, written in the form

P(x, y) dx + Q(x, y) dy = 0, (69)
where P and Q are assumed to have continuous first partial derivatives, is said
to be exact when P and Q satisfy the condition
ap aQ (70)
ay = ax'
In this case, and in this case only, there exists a function u(x, y) such that
du = Pdx + Qdy. (71)
Thus Equation (69) then IS identical with the equation du = 0, whose general
solution clearly is
u(x, y) = e, (72)
where e is an arbitrary constant.
Since Equation (71) implies
au
ax = P, ~; = Q, (73)
the necessity of Equation (70) follows from the fact that
a (aU)
ax ay = aya (aU)
ax
when the indicated derivatives are continuous. In order to obtain a function u
satisfying the two relations of Equations (73), we may, for example, start with
1.12. Specíal Solvable Types oC Nonlinear Equations 33
the first relation, integrating with respect to x with y held constant (si nce aL/fax
was formed in this way) to obtaint
u(x, y) = r.'l:"
P(x, y) dx + f(y)· (74)
Here f(y) is the added "constant of integration," to be determined by the

second relation of Equations (73), which gives
and hence
I ~Px
xo y
dx -1- I'(y) = Q
, = Q-
f(y) JX aapy dx . (75)
xo
That the right-hand member is indeed only a function of y, so that f(y) can be
determined (with an irrelevant arbitrary additive constant) by direct integra-
tion, follows from the fact that its partial derivative with respect to x is zero,
since
a
dX -
(o _IX x,
ap dx) = aQ _ ap =
ay ax ay
o
,
when Equ a tioll (70) is satisfied, so that the sufficiency of that condition is also
established.

dy I + y2 + 3x2y
dx = I - 2xy - x 3
can be written in [he form
(3x 2y + y2 + I)dx + (x 3 + 2xy - I)dy = O,
and the condition (70) of exactness is sa ti sfied . From the relation
au
ax = 3x2y + y2 + 1
there follows " = x 3y + xy2 + X + f(y). The relation

au
-ay = x' + 2xy - 1
then gives x 3 + 2xy + I'(y) = x 3 + 2xy - I or I'(y) = -1, from which f(y) = -y,
apart from an irrelevant arbitrary additive constan!. Hence the required solution 11 = e
is
tThe special notation

x 3 Y + xy2 + X - Y = c.
•
I x
x,
p(x,y)ax
is sometimes used to indica te that "partial integration" with respect lo x is lO be effected, with
y held constant in the process.
(3) Firsl-order Equations of Homogeneous Type. Afunction f(x, y) is said to

be homogeneolls of degree n ir there exists a constant n such that, for every num-
ber A, it is true that
f(Ax, AY) = Anf(x,y).
Thus, for exarnple, the functions X2 + xy and tan-ley/x) are both homo-
geneous, the first of degree two and the second of degree zero, whereas X2 + Y
is not homogeneous. The first-order differential equation
P(x, y) dx + Q(x, y) dy = ° (76)
rnay be said to be of homogeneous type if P and Q are both homogeneous of
degree n, for sorne constant n.t
Such an equation becornes separable upon the change of variables
Y = vx, dy = v dx +x dv. (77)
For, since P(x, vx) = xnP(l, v) and Q(x, vx) = xnQ(l, v), this substitution re-
duces Equation (76) to the form
xnP(l, v) dx + xnQ(l, v)(v dx + x dv) = °
or [P(l, v) + vQ(l, v)] dx + xQ(I, v) dv = 0, (78)
which is indeed separable.
The substitution
x = uy, dx = u dy + Y du
also is appropriate and may or may not lead to a more convenient forrn after a
separation of variables.

(3y2 - X2) dx = 2xy dy
is of homogeneous type (with n = 2) and, with y = vx, it becomes
(v 2 - 1) dx - 2xv dv = O
2v dv dx
(x =F O, v 2 =F 1)
v2 - 1 x
and henee lag [v 2 - 1[ = log [xl + log [e[ = log [ex[
or v2 - 1 = ex
or
The temporarily excluded relations, x = O and y = ±x, both are seen to satisfy the
given equation and hence in fact are solutions, but they need not be appended to the
solution obtained sinee they are included in it when e = ca and e = O. •
tSuch an equation is usually called a homogeneous equation. However, we will use this term
onlyas il is defined in Seetioo 1.8 and generalized io Seetiao 5.1.
LI2. Special Solvable Types of Nonlinear Equations 35
(4) lvlisce/laneous First-order Equations. Most of the other known techniques

for solving special first-order equations in c\osed form consist of reductions to
forros which are linear, separable, exact, or of homogeneous type. Sorne typical
examples follow.
Example 5. Bernoulli's equation. The equation
~~ + P(x)y = Q(x)yn (n 07= 1)

y-n ~~ + p(x)y-n+ 1 = Q(x),
which clearly becornes linear under the substitution v = y-n+'.

•
dy ax + by + c
dx = Ax + By +e
caD be reduced to an equation of homogeneous type by writing x = t + h and y =
s + k aDd determining suitable values ofthe constants h and k, provided that aB 07= bAo
In that exceptionaI case, the substitution s = ax + by + e renders the equation
separable. •
Example 7. Although the equation

y2 dx = x(x dy - y dx)
can be treated as an equation of homogeneous type or as a Bernoulli equation in y
(Example 5), the combinatioD x dy - y dx tends to suggest division by y2 or by x 2, for
the purpose of forming -d(xfy) or d(yfx). Here a division by y2, accompanied by a
division by x, is clearly indicated, leading to the form
dx + y dx - x dy = O
X y2
which is exact, since eacb term is ao exact differential, and tbere follows
log Ixl + -xy = c

x
or y = e-:----T:7T::-;
log Ix I
In trus case, multiplication of the equation by the integrating factor x- I y- 2 makes the
equation exact. In less contri ved situations, the discovery of such a factor may be much
less straightforward. (See Problem 63.) •
Example 8. The nonlinear EquatíoD (1)),
(ir - 2 i + 4y = 4x - 1,
yields the two first-degree equations
dY =l±2VX-Y.
dx
Th e prominence of the express ion x - y may suggest the substitution
x - y = ti,
1 dy _ du
- dx - dx'
which lead s to the separable equations
du
dx
± 2Vu.
If it is noticed that the process of separation necessitates the special consideration of
the relation u = O, the one-parameter solution (12) and the singular solution y = x
then are easily obtained. Equally fortuitous substitutions may suggest themselyes in
otber cases. •
(5) S econd-order Equations Lacking One Variable . The general equation of

second order is of the form
F(x, y, ~~, ~:,) = 0, (79)
containing dZy/dxz explicitly. Any such equation can be written equi yalently a s a
pair o f simultaneou s firs t-order equations, in yariou s ways . In particular, we
may write
dy = P (80)
dx
and dZy
dX2 =
¡~~
dp dy _ dp
(81 )
dy dx - P dy
in accordance with which Equation (79) can be replaced by either the set
dp) -_ 0,
F ( x,y,p, dx
(82)
or the set
F(X' y, p, p 'Z) = 0,
(83)
dy _p
dx - .
In the general case, both equations of either set invo lve the three variables
x, y, and p, and hence no one of the equation s can be sol ved independently of
the associated equation. Howeyer, if y is not explicitly invo lved in F , then the
s J.J2. Special Solvable Types of Nonlinear Equanons 37
first equation of (82) involves only x and p. If it can be solved, to pro vide a
relation between x and p, and if the result can be used to eliminate p from the
seco nd relation in Equation (82), then y can be determined in terms of x by
integration of the resulting equation.
On the other hand, if x is not explicitly involved in F, then the first equation
of (83) involves only y and p. If it can be solved, to pro vide a relation between
y and p, and if the result can be used to eliminate p from the second relation in
Equation (83), then a separable first-order equation in y and x results.
In either case, it may be more feasible to obtain x and y in terms of p, with
p retained as a para meter, than to eliminate p by using the solution of the first
,f
equation of the pair.
:t'
n
I 3
dOy = (d y )
dx2 X dx
f
lacks the variabley. With dy/dx = p and d 2 y/dx2 = dp/dx, there follows
) dp = Xp3
dx
which separa tes to give
a
+1
e P=.vc f -x o·
dy = +
) Hence
.verdx- X2
,
Y = .
-+sln-
1 X
-+c"
el -
or x = C'l sin (y - C2),
)
where C'l = ±CI'
•
2 y )2
y. dx
d y -_
2
(ddx
) lacks the variable x. With dy/dx = P and d 2 y/dx2 = p(dp/dy), there follows
p(Y~~ -p) = O
and hence
dp
y dy = P or P = O.
)
The first alternative gives
s which ineludes the second alternative p = O as a special case, and henee the general
f solution is
•
Example 1 l. The equation

d 2y
dX2 - y = O
lacks the variable x. Whereas it can be sol ved by the present method, it is linear in y
and is much more easily sol ved by the methods of Section 1.5. •

d y )2 d2y = 1
( dx + (dy )2
dX2 dx
lacks both x and y . If the absence of y is exploited, there follows
p2 ~~ = 1 + p2
and hence
x =p - tan-Ip + CI.
This relation can be used to eliminate x (rather than p) from the relation
dx
dy = p dx = P dp dp,
to give
dy = (
p 1 - 1 +1)
p2 dp = 1
p3
+ p' dp.
Hence
y = tP' - t log (p' + 1) + C,.
Here the solution provides a parametric representation of x and y. The fact that the
parameter p happens to be identifiable with dy/dx may afford a subsequent added con-
venience. •
REFERENCES
1. AGNEW, R. P ., Differential Equations, 2nd ed., McGraw-Hill Book Company, Inc.,

New York, 1960.
2. BlRKHOff, G., and G.-c. ROTA, Ordinary Differential Equations, 2nd ed., Ginn and
Company, Boston, 1969.
3. CODDINGTON, E. A., and N. LEVINSON, Theory of Ordinary Differential Equations,
McGraw-Hill Book Company, Inc., New York, 1955.
4 . DANIEL, J. W ., and R. E. MOORE, Computation and Theory in Ordinary Differential
Equations, W . H . Freeman and Company, Publishers, San Francisco, 1970.
5. INcE, ·E. L., Ordinary Differential Equations, 4th ed., Dover Publications, Inc., New
York, 1953.
6. PlAGGlO, H. T. H ., Differential Equations, Bell Mathematical Series, Open Court
Publishing Company, La Salle, Ill., 1952.
Problems 39
PROBLEMS
Section 1.1
). Differentiate eaeh of the following relations with respeet to x and, by using ¡he result
to eliminate the eonstant e, obtain a first-order differential equation of whieh the given
relation defines a one-parameter solution:
(a) y = ce"', (b) y = ce"" + 2x,
1
(e)Y=e_x' (d)y=e'"',
(e) X2 + y2 = e 2, (f) y = ex + e 2x2.
2. By differentiating eaeh of the following relations twiee, and eliminating the con·
stants e¡ and e2 from the ¡hree resultant equations, obtain a seeond-order differential
equation of whieh the given relation defines a two-parameter solution:
(a) y = e¡e' +
e2e-", (b) y = e"(e¡ C2X), +
(e) y = e¡ eos x + e2 sin x, (d) y = e¡e"",
C¡ + x
(e) y = el + x' (f) (x - C¡)2 + (y - C2)2 = 1.
3. Obtain the general solution of each of the following differential equations:

(a) dd
y
x
= 2xy, (b) ydy + vi - y 2 dx = O,
(e) (1 - y2)dx + (1 - x2)dy = O, (d) ..jI - x'ydy = ..jI - y 2 xdx.
Section 1.2
4. ]f II¡(X) and l/2(X) are lioearly independent, prove that A ¡II¡(X) + A 2112(X) and
B¡II¡(x) + B 2112(X) are also linearly independent if A¡B 2 -
A 2 B¡ O. *
5. By considering the functions II¡ = x 3 and 112 = x 2 1xl over an interval including the
origin, show that identieal vanishing of the Wronskian of l/I and l/2 over an interval
does no! imply linear dependence of II¡ and 11 2 over that interval. (Notiee that l/2 =
x 3 sgn x and l/'2 = 3X2 sgn x for aH x, where sgn x = + 1 whert'x > O, sgn x = -1 when
x < O, and sgn O = O.)
6. (a) Prove that e''', e"', and e"" are linearly independent over any interval if r l , r 2 ,
and r 3 are distinct constants.
(b) Determine whether log x, lag X2, and log x 3 are linearly independent when
x> O.
7. Determine whether each of the following sets of functions is linearly dependent over
the speeified interval:
(a) 1 + 2x, 2 + 3x, 3 + 4x (- < x < ;;o),
(b)O,I,x (O<x<I),
(e) l,x,lxl (-1 <x< 1),
(x - 1) 2 I
(d) x, x _ 2 'x _ 2 (2 < x < co).
40 Ordinary Differenlial Equalions
8. Suppose that l/, and l/o both salisfy the linear ditferential equarion
( pY'r + qy = O,
where p and q are functions of the independent var iable x. Show that there follow s
u, (PU'2)' - "2 ( pU'¡ r= O,
and that this equation is equi valent to the equarion (pW)' = O, where W is the
Wronskian of u, and U2' Hence deduce that W = A /p, where A is a constant.
Section 1.3
9. If y = u¡(x) and y = "2(X) satisfy the bomogeneous linear equarion

dny dn-¡y
- + a,(x) -
dx"
- + ... + a "(x)y
dx"-¡
= O,
prove rhat y = c¡u¡(x) + C2UZ(X) is also a solurion, for any constant values of C¡ and
c,.
10. Verify rhar 11) = 1 and U2 = log x each sarisfy rhe non linear equation
y" + y'2 = O,
but thar y = C¡lI¡ + C2UZ is not a solution unless eirher C2 = O or C2 = 1.
11 . If yj." and yj.2' are two linearly independenr particular solurions of the nonhomo-
geneous linear equarion
d 2y , d y (
dx " ,. Q¡(x) dx + a2(x)y = h x),
show thar rhe funcrion u, = yj.1> - y '/ ' satisfies the associared homogeneous equation
(in which h is replaced by O).
12. Verify rhar y = log x reduces the nonlinear expression y" + y " lO zero, and Ihar
y = x reduces ir ro uniry . To whar does the slIm y = log x x reduce it? +
SectioD 1.4
13. Solve rhe following differenrial equations:
(a) x 1x - ky = X2 (k constant), (b) ~~ - y tan x = x,
(e) ~~ + y tan x = sec x, (d) Z+ (y - 2 sin x) cos x = O,
(e) (X2 - 1) ~~ + 2y = (x + 1)', (f) X2 log x dy + (xy - 1) dx = O,
(g) 2xy ~~ - y' = x', (h) ydx = (x + y2)dy.
14. Obrain the solurion of the problem
~~ - ky = h(x),
Problems 41
in the form
when k is a constan!.
15. Show lhat the substitution u = y '-n reduces the nonlinear equation
CZ + a,Cx)y = f(x)yn (n #- 1)
to a linear equation, CThis equation is often called Bernou/lj's equatiol1,)
Section 1.5
16. 1f two roots of the characteristic equation (24) are r = ±a, show that the cor-
responding part of the homogeneous solution can be written in the forro
y = e, eosh ax + e2 sinh ax,
where sinh ax and cosh ax are the hyperbolic fllnetions defined by the equations
. eQX _ e- ax eax + e- ax
smh ax = 2 ' eosh ax = 2 '
17. If two roots of Equation (24) are r = a ± b, show that the eorresponding part of
the homogeneous solution can be written in the forro
y = eax(c, eosh bx + e2 sinb bx),
18. Solve the following differential equations:

d 2y dy d 3y d 2y dy
Ca) dX2 - dx - 2y = 0, (b) dx 3 - dx 2 - dx + y = 0,
d 2y
Cc) d----;
dy
x- - 2 d-x + 2y = 0,
,)d 4y
Cd dx' -
4d3Y+7d2y
dx 3 dX2 -
6dY+2
dx y =
°
,
.t
d 3y
Ce) dx 3 - y = 0,
d 2y
(f) dX2 - 2iy = °
(;2 = -1).
19. The following differential equations arise in dealing with the problems noted. Find
the general solution of eaeh equation, assuming that k is a nonzero eonstant:
d 4y
Ca) -dx 4 - k 4 Y = ° Cvibration of a bearo,)
4
Cb) d y4
dx
+ 4k 4 y = ° (beam on an elastic foundation),
4
Ce) d y4 _ 2k 2 d y
dx
2
dxl.
+ k 4y = ° Cbending of an elastie plate).
20. Use the method of undetermined eoeffieients to find the complete solution of eaeh
of the following differential equations:
d 2y d 2y ,
Ca) dx~ + k2y = sin X (k 2 #- 0, 1), (b) dx 2 + y = Sill X,
d 2y , d 2y
Ce) dxl. -y = SIO X, (d) dX2 - y = eX,
d 2y 2
Ce) dx" - y = xe x , (f) d y _ 2 dy
dX2 dx
+ 2y = sin x,
2
(g) d y _ 2 dy
dX2 dx
_L
I
?y
-
= e-' sin x
,
(i) ~:{ - ~~ - 2y = 3e-' + 10 sin x - 4.<.
Section 1.6
21. Solve the following differential equations:
d'
dx
d
(a) X2 --.2'2 + x J - k2y =
dx ' (b) x
2 d 2y °
d----';
x -
dy
x -d
x + 2y = 0,
d y
2
(e) X2_2 - 2y =
dx '
° , d 2y
(d) x- d x 2 -
dy
x d-
x + y = 0,
d y
(e) X 3 dx 3
3
+ 2
2
2 d y
X dx 2 -
dy -'- -
x dx ' y -
°
,
,d 2y
(f) x- d X 2 + dy
-' - n(n
2x u..'C +
l)y = 0,
2
d y dy ,d 2y dy
(g) X2 -
dX2
+ X- -
dx
y = X2
' (h) x- d x 2 + x -d
x - Y = x,
2
( -1) d y -
x 2 dx' 4 x dy
dx +6 y = 6x + 12 .
22. Solve the differential equation

d 2y
(a + X)2 dX2 - 2y = h(x)
in eaeh of the following Cases:
(a) h(x) = 0, (b) h(x) = 3(a + X)2 + 1, (e) h(x) = X2.
Section 1.7
23. Show, by direet expansion, that
(x 2D')(xD) = (XD)(X2D2)
but that
24. Use the method of the text Example to obtain the general solution of the equation
(D - rj)(D - r,)y = h(x) (r, "'" r2),
where r I and r2 are eonstants, in the form
y = e'" f e-'''[f e-"'h(x) dx ]e'" dx + c,e"x + C2 e"x.
25. (a) If the notation

1
y(x) = D _ rh(x)
is used to indieate that y(x) satisfies the equation (D - r)y = h, where D = djdx and
r is a eonstant, show that
where an arbitrary additive eonstant of integration is implied in the integral.

(b) Verify that, with the notation of part (a), the expression
Y = (D - 1 = D -1 r [D1 ]
r z )(D - r] )h(x) z - r¡ h(x)
Problems 43
satisfies the equation (D - r¡)(D - r,)y = h. Hence obtain the general solution of
that equation in the form
y = e"" f e("-"¡X [f e-"xh(x) dxJ dx,
where an arbitrary additive constant is implied in each integration. (Compare Problem
24.)
26. Yerify that, with the notation of Problem 25(a), the express ion
y = [ rl 1 ( 1
- rz D - r¡
_ 1
D - rz}
'IJ h(x)
= 1 [ I
r¡ - rz D - r¡
h(x) - 1
D - r2
h(x) J
satisfies the equation (D - r¡)(D - r,)y = h where r¡ * r2' Hence obtain the general
solution of that equation in the form
y = r¡ -
1 r [e""
2
J e-''"h(x) dx - e"X J e-"xh(x) dX]
when r¡ * rz, where an arbitrary additive constant is implied in each integration.

27. Yerify that the result of Problem 26 can be written in the form
when r¡ * r2'
d'y
dx' - y = e-'
n by use of each of the two formulas derived in Problems 25 and 26.

29. By proceeding as in Problems 25 and 26, obtain the general solution of the equation
(xD - r¡)(xD - rz)y = f(x) (rl * r2)
in each of the following forms:
(a) y = x" J [Jx',-" x-"f(x) ~ ] ~ + e¡x" + e2 x ",
Id
(b) y =
Y¡ -
1 Jr2
[X" x-"f(x) dx - x"
x
J x-"f(x) dX]
x
+ e¡x" + e2x",

d 2y
x' dX2 - 2y = x'
by use of each of the two formulas derived in Problems 29(a, b).

,
Section 1.8
31. Suppose that the eoeffieients in Equations (37a,b) are eonstants. With the symbols
R¡ = L¡x + L,y - h¡ and Rz = L,x +
L 4 y - h z , Equations (37a,b) beco me
R¡ = 0, (a)
R 2 = 0, (b)
and Equations (38a,b) or, equivalently, (39a,b) beeome
L4R¡ - L2Rz = 0, (e)
L¡R z - L3R¡ = O. (d)
Show that if solutions of (e,d) also satisfy (b), so that R, = 0, there follows also
L4R¡ = 0, LJR¡ = O. Henee, notieing that these two equations can have a nontrivial
eommon solution (R ¡ *
O) only if L. and LJ have a eorumon factor, deduce that il all
Ihe coefficienls in Equalions (37a, b) are conslanl and if, in eilher 01 Equalions (37a, b),
Ihe Iwo operalors invo/ved have no laclors in common, Ihe reslriclions on Ihe conslanlS
appearing in Ihe Solulions 01 Equalions (39a, b) are comp/ele/y delermined by SUbSlilulion
inlO Ihal equalion. Olherwise, Ihe so/ulions 01 Equalions (39a, b) musl be checked by
subslilulion inlo bOlh 01 Equations (37a, b).
32. Ulustrate the results of Problem 31 in the case of the simultaneous equations
d'y dx
dI" + dI = 0,
dy dx
dI + dI + x = O.
33. Find the solution of eaeh of the following sets of equations:
(a)
¡ 2 dx
dI'
dy
= 3x _ y
2 dI = 3y - x,
dI" dy
(b)
¡dX _ dy + x = 2e'
dI
d'x
¡dZX
dI
dy
2
1 d y
'
_
dt2 - dt + 3x - y - 4e .
,
di" - '2 dl + k x = 0,
2
2 dl'J. - dI - 4x = 2t,
(e) (d) '
dx dy _ d 2y d 2x
¡ 2 dI - 4 dI - 3y - 0, dl l - dl l + 2_
k y-O,
(e) ¡ ~;~ +
d 2x
di 1 +
x
3x
+
+
2y = 7e Z'
2y =
-
ge 2'
1
+ 1,
(f)
¡~;~ + ~~ + ~~ + 3x + y
d2x
dI 1 +
dy
dI + x + y = 0,
= 0,
dx
di = 2x,
dy
(g) di = 3x - 2y,
dz
dI = 2y + 3z.
34. Find the solution of eaeh of the following sets of equations:
(a)
¡ 21
2/ ;
~~ =
d
3x - y,
= 3y - x, (b)
I
~
~~ =
I di = 3x -
I
dz
di =
2x,
2y +
2y,
3z.
'5 probl e01S 45
Section 1.9
Is 35. Sol ve by ¡he method of variation of parameters:
(a) 'Z + a, ( x)y = h(x), (b) dx{
dO
+y = eOI x,
d 2y
(e) t!2
dx z
+y = sec x, (d) dX2 +y = log x,
2
2
y _ 2 dy
(e) dd X2 + 2y = eX tan x ( f) " d" -
X y - 4 xdy
- + 6y = " .
x-slnx
dx ' dX2 dx '
;)
1) (g) X2 ~;, - 2x ~~ + 2y = x log x,
·0
36. lf y(x) satisfies Equation (55), show that
11
1/ dy =
dx
J"< h( ';)[u,(';)u~(x) - u (';)u',(x)1 d./! +
W[u,(';) , U2(';)]
2 '() + u '()
e¡u, x e2 x .
<,. 2
),
IS
'n
37. Obtain th'e general sol u tion of the equation
2
'y d y = h(x)
dx 2
r
in the form
y = (x - c;)h(';) d'; + e,x + e2'
38. The differential equation

2
d y
dX2
+ dx
dy _ xy -
-
O
possesses solutions u ¡ (x) and u 2 (x) whieh can be represented by senes, valid near
x = O, the leading terms of whieh are as follows:
u¡(x) = 1 + ..Ix' + "', U2(X) = X - ±X2 + "'.
Use Abel's formula 10 show that W(u" U2) = e- x, and hence deduce that the general
solution of the equation
d 2y dy
- + -dx
dx 2
- xy = h(x)
is of the form
SectioD 1.10
1 39. Verify that y = e X satisfies the homogeneous equation associated with
(x - l)y " - xy' + y = 1, and obtain the general solution.
40. Verify that y = tan x satisfies the equation y " cos 2 x = 2y, and obtain the general
solution.
41. One homogeneous so lution of the equation
(1 - X2) ~;{ - 2x ~~ + 2y = 6(1 - x 2

)
IS y = x, Find the complete solution,

42. One solution of Legendre's equation ,

d' d
(1 - X2) d;: - 2x d~ + n(n + l)y = 0,
is called PnC'C). Show that a second solution is of the form

dx
P n( x)
J ~(l'----x-2"')"[pnnT(x"')"]2
43. One solution of Bessel's equation,

d 2y dy
X2 dx" + x dx + (x' - p2)y = 0,
is called Jp(x). Show that a second solution is of the form
J.(x) Jx[J~(X)F'
44. Given that the function Jo(x) in Problem 43 has a Taylor series expansion with
leading terms
x4
X2
Jo(x) = 1 - -
4 64
+ - - ...
which converges for a11 values of x, show that any independent solution of the equation
d 2y dy..L _
x d x ,+ d x ,xy - O
has an expansion of the form

X'
y = Jo(x) [ A(log x + "4 +
where A and B are constants and A 7= O, and hence becomes logarithmically infinite
as x - O.
45. Use Abel's formula (65) to show that if u, and u, are solutions of
d'y
dx ' + a,(x) dy
dx + a2(x)y = O
in an interval 1, and if we write
then the Wronskian of 1/, and U2 is given by the expression

p ea)
W[u,(x), "z(x») = p(x)'
Iu,(a)
u',(a)
u,(a)
u~(a)'
I
where x = a is a point in the intervall.
46. Removal Ol firsl-derivalive lerm. In the differential equation
d 2y dy
dx" + a, (x) dx + a,(x)y = h(x),
make the substitulÍon y = u(x)v (x) and determine v so that the coefficient of duJdx in
the resultant equation vanishes. Thus show that, if p = eSa. dx, the substitution
y = u/,.,,!p reduces the differential equation to the fo rm
d 2u 1 (
dx 2 -"4 ay + 2 da,
dx
)
- 4az u = .¡p- h(x) .
Problems 47
47. Use the result of Problem 46 to show that Bessel's equation (Problem 43) takes
the form
~ d::'u
x· dx'
I (
X2 - p' + 41 ) u = O
with Ihe substitution

y=~.
u
-/x
Section 1.11
48. A fundamental set of solutions of a homogeneous linear differential equation
Ly = O, of order n, relative to a point x = a, may be defined as a set of n solutions
vo(x), ,,¡(x), ... , "n-I (x) such that
voCal = 1, e~(a) = v~(a) = = v6n - 1 '(a) = O;
vI(a) = O, v'l(a) = 1, v';(a) = = ,,\n-I)(a) = O;
Vf7-1 ( a) = v'n- 1('a) = ... = "(n-2'(a)

n- 1 = O !
(a) Deduce that the solution of the initial-value problem, in which Ly = O and
the values y(a) = Yo, y'(a) = y~, ... , yln-j)(a) = y6n- l ) are prescribed, is then given by
n- 1
y(x) = b Ybk)Vk(X),
k=-O
(b) Show that the Wronskian of a fundamental set of solutions, relative to x = a,

is unily at x = a.
49. Show that the equation
2d'y dy _ _
x dx 2 +x dx y - O
does not possess a fundamental set of solutions, relative to x = O. Then determine a
fundamental set relative to x = a, where a "'" O.
50. Show that the functions cosh x and sinh x constitute a fundamental set of solutions
relative to x = O for the equation
d 2y
dX2 - y = o.
(See Problem 16.)
51. Obtain a fundamental set of solutions for the equation
d 3y
dx 3
+ dy
dx
_ O
- ,
i relative to x = O.
52. Obtain a fundamental set of solutions for the equation
d 4y
dx' - y = O,
relative to x = O.
53. Let vo(x), ... ,"n-' (x) comprise a fundamental set of solutions of an nth-order
homogeoeous linear differential equation Ly = O, relative to x = a, (Problem 48) and
let w(x) be a particular solution of the nonhomogeneous equation Ly = h for which
w(a) = w'(a) = ... = w'n-' '(a) = O.
(a) Show that the solution of the equation

Ly = h
for whieh
y(a) = Yo, y '( a) = y~, ... ,y' n-ll(a) = Yb"- ' )
is
n- ,
y(x) = w(x) -i- ~ Ybk )~·k(X).
k::::Q
(b) When n = 2, verify that the funetion
w(x) = IX h(¿;)(U,(¿;)U2(X) - U2(¿;)U.c¿;)] dé,

Q W[u,(é,) , ui¿;)]
has the required properties if u,(x) and U2 (X) are any two linearly independent solu-
tions of Ly = 0, provided that the integral exists.
54. (a) Determine those values of the eonstant k for which the differential equation
y" °
+ k 2 y = possesses a nontrivial solution which vanishes when x = and when °
x = a, where a is a given positive constan!.
(b) Determine those values of k for which the differential equation y" + k2y = 1
possesses a solution which vanishes when x = and when x = a . °
55. Determine the solution of the problem
d 2y
dx" +y = h(x), y(O) = 0, y(i) = 0,
when sin I *0, by the following steps:

(a) Use the result of text Example 1 to obtain
y(x) = fax h(é, ) sin (x - é,) dé, + e2 sin x,
where e2 is such that

° = S: h(é, ) sin (l - é, ) dé, + e2 sin l.
(b) Show that , if sin 1* 0, there follows
y(x) = Si~ I [f h(¿;) sin (x - é,) sin I dé, - J: h(é,) sin (l - é,) sin x dé,].
(e) By writing f ~ = f~ + f~ in the second integral, deduce that the solution can
be expressed in the forra
y(x) = S: G(x, é, )h(é,) dé"
where
rn ,,in (x - J) ( é, < x),
(x < é,),
when sin 1* O. [The function G(x , é,), or a constant multiple of it, is known as the
Green's fune/ion for the expression Ly = y" + y and the conditions y(O) = y(l) = O.]
Problems 49
56. If sin i = O, show that the problem

d 2y
- x-
d' +y = h(x), y(O) = O, y(i) = O
has nO solution unless h(x) satisfies the condition
f~ h(x) sin (i - x) dx = O,
in which case there are infiniteiy many soiutions, each of the form
y(x) = fox h(e;) sin (x - e;) de; + e sin x,

where e is an arbitrary constant. [See Problem 55(a).]
57. The Green's function. Let L be the operator such that
u·
d 2y dy
.
Ly = ao(x) d ¿
x
+ al(x) -d
x
+ a2(x)y
en and assume that the solution of the problem
Ly = h(x), y(a) = O, y(b) = O
¡
exists and can be expressed in the form
y(x) = f G(x, e;)h(e;) de;,
in terms of the Green's function

G(x, e;) = {cx(e;)u(x) (x <: e;),
/l(e;)v(x) (x :> e;).
Determine G by the fOllowing steps:
(a) Show that the assurned solution is of the form
and deduce that

y(x) = v(x) r PU;)h«(,) de: + u(x) r CX(e:)h(e;) de:
dy =
dx
lb
a
aG(x,
ax
e;) h(e;) de;
if and only if the condition
cx(x)u(x) = p(x)v(x) (A)
"
an is satisfied . [Recall that (d/dx) e.
rp(e;) de; = rp(x) if Xo = constant.]
r
(b) Show that, when (A) is satisfied,
~;, = a2~SX; e;) h(e;) de; + [/l(x)v '(x) - CX(x)u'(x)]h(x),
and hence that
Ly = f [LG(x, O]h(e;) de; + Go(x)[p(x)v'(x) - cx(x)u'(x)]h(x),
so that Ly = h(x) if
Lu(x) = O, L v(x) = O (B)
che
O.] and Go(x)[p(x)v'(x) - CX(x)u'(x)] = 1. (C)
50 Ordinary Oifferential EquatioDS
Ce) Show tha! (A) implies

"(x) L/(x)_
a(x) = f3(x) = c(x),
so that
(x -< ~),
with c determined by (C) in the form
(d) Frorn the preeeding results, deduce that l/ex) and ·v(x) must be solutions of the
equation Ly = 0, sueh that l/Ca) = °
and v(b) = 0, and that G then is defined by the
formula of pan (e), provided that both aD and the Wronskian of 1I and v do not vanish
in (a, b), and henee that l/ and vare not eonstant rnultiples of eaeh other. Show also
that if Gis eonsidered as a funetion of x, with ~ fixed, then Gis continl/ol/s when x = ~
and aGlax has ajl/mp of llaD(~) as x inereases through';.
(e) Use Abel's formula 10 show that c(,;) is a constan! if aj(x) = a~(x), so that
G(x,,;) then is symmetric in x and ,;. [Notiee that a, must be replaeed by a,laD in
(56).]
(f) Verify that the formula of part (e) eorreetly yields the result of Problem 55.
58. Generalize the results of Problern 57 as folJows: .
(a) Show that if the conditions y(a) =
homogeneous eonditions, of the forrn
and y(b) = ° °are replaeed by more general
k I y(a) + k,y'(a) = 0, k,y(b) + k.y'(b) = 0,
then l/ex) must satisfy the eondition imposed at x = a and v(x) that imposed at x = b.
(Again ¡he Wronskian of l/ and v must not vanish.)
(b) Show that, if the nonhomogeneous eonditions
k,y(a) + k,y'(a) =g" k 3 y(b) + k.y' (b) = g,
are preseribed instead , then the problem solution (if it exists) ean be written in the form
y(x) = w(x) + s: G(x, ,;)h(,;) d,;
where G is the Green's function of pan (a), corresponding to the " hornogenized" end
conditions, and where w(x) is the solution 01' the equation Ly = which satisfies the °
prescribed end conditions.
[See also Problem 80 of Chapter JI, for a different (but equivalent) approach to
the Green 's funetion .]
59. Use the results of Problems 57 and 58 to derive the solution exhibited for eaeh of
the following problems:
dZy
(a) dx Z = h(x), y(O) = A, y(l) = B;
y(x) = A (1 - ~) + B ~ + I G(x, ,;)h(,;) de;,

r
problems 51
-1).< (x -< ~),

where
(x:> ~).
(b) ~;{ = h(x),
where
d 2y y(-co) = 0, y(co) = O;
(e) dX2 - y = h(x),
d2 1 dy
y(x) = ± r- e-Ix-(Ih(~) d~.
1
(d) { - - -d = h(x), y(O) = O, y(l) = O;
ctx x x
)
y(x) = ti G(x, ~)h(~) d~,
(1 - ~2)X2
2~ (x -< ~),
j
t -
1 where G(x, ~) = _ W -2 X2)
(x:> ~).
Section 1.12
.\
60. Solve the following:
(a) (2xy J + y) dx + (3x2y2 + X - 2y) dy = 0,
(b) (3x2y - y3) dx - (3xy2 - x 3) dy = O,
(e) eX(y dx + dy) + eY(dx + x dy) = O.
(a) dy = x - Y,
dx x + y
(b) (X2 + 2xy) dy = y2 dx,
11 (e) x d
dxy
= y - V X2 + y2,
/-:-c'>""""""""
(d) x dy - y dx = x tan (~) dx.
d (a) x ~~ + y = y log (xy),
e (b) x dy + (y - X2) dx = 0,
(e) ydx + (x - y2)dy = 0,
o (d) (x + y - 3) dy = (x - y + 1) dx,
(e) x:Z +y + x2y2 = 0,
(f) (2xy2 - x) dx + (X2y + y) dy = O.
63. Integratillg faetors. Le! q be an integrating factor for the differential equation
Mdx + Ndy = O,
so that qM dx + qN dy is an exact differential.

S2 Ordinary Oifferential Equations
(a) Show that q musl be sueh Ihal
Naq _ Maq = (aM _ aN)q.

ax ay ay ax
(b) Jf
(aÑ[
ay
_ aN)/'
ax
N= ~
is independent of y, show that an integrating factor is
q(x) = eJ~dx.
(e) If
aM _ aN) 1M
( ay
= IJI
ax I
is independent of x, show Ihat an integrating factor is
q(y) = e-J~d y .
64. Use the results of Problem 63 to solve the following differential equations:
(a) (2x 2 - 4y 3) d.'C + 3xy2 dy = 0,
(b) mxm-¡ydx + [(n -p)y" -pxm]dy = O.
d2 + (ddxy )2_- 0,
y
(a) dX2
2y
d + 2x (ddxy )2 _- 0,
(b) dx '
(e) d y (dy ) = 0,
2
y dx2 + dx
2
d2y _ [ 1 + (dcJxy )2J3/2'
(d) dX2 -
2
(f) [6(d )2 + lJ d y = l.
d 2y y
(e) dX2 +y = 0, dx dX2
66. Show that the substilution y = lI'IQu , where u' = duldx, reduces the nonlinear
first-order equation
~~ + P(x)y + Q(X)y2 = R(x)
to the linear second-order equation
du -RQu = °
(p _ QQ' ) dx
2
d u +
dX2 '
and also that the substitution y = y¡ + l /v, where y¡( x ) 15 any known solution of
the given equatiQn, leads to the linear first-order equation
dv
dx - (P + 2Qy¡ )v = Q.
(The nonlinear forro is known as Riccati's equation.)

67. Use the procedures suggested in Problem 66 to obtain the general solution of the
equation
dy
X2
dx
+ xy + x2yl = 1
in the form xy = (x' - k)/( x 2 + k ), where k is an arbitrary constan!. (Take y ¡ = l /x.)
2
The Laplace Transforlll
2.1. An Introductory Example. If a function J(t) is multiplied by e - U

and the result is integrated with respect to t from t = O to t = CXJ, a new func-
tion of the new variable s is obtained when the integral exists.t This function
(when it exists) is called the Laplace transJorm of J(t). Before studying the
properties of such transforms, we illustrate one of their most useful applications
.r by considering a simple problem.
Suppose that we require tbe solution of the differential equation
dy _ y = ea, (1)
dt
for positive values of t, wh..icb satisfies tbe ioitial condition
y(O) = -1. (2)
In place of determining the general solutíon of Equatíon (1) by tbe methods of
Chapter 1, and then deterrnining the arbítrary constant by satisfying Equation
(2), we proceed as follows.
We first take tbe Laplace transform of both sides of Equation (1), by mul-
tiplying botb sides of the equatíon by e-H and integrating the results wíth
respect to t from zero to ínfinity, to obtain the equation
r~ e - " dy di _ r~ e-"y di = r~ e -He a , di. (3)

Jo dt Jo Jo
tThe variable s is coosidered to be a real variable io this chapter. However, all results which
hold wheo s is real aod s > a, for sorne real value of a, also hold wheo s is complex aod (real
part of s) > a.
53
54 The Laplace Transform
It ís assumed, of course, that the separate integrals exist for sorne range of
values of s.
The integral on the right is readily evaluated,
Jo
r= e-"e"' dt = _ e- es -aH
S - a
1=
o s- a
(4)
the integral existing when s > a. The first integral on the left ll1 Eq ua tion (3)
can be integrated formally by parts to give
j.= e-"..1:'ddt dt
o
= e-S'y(t) 1=
o
+s 1-o
e-"y dt
- y(O) + s r e-"y dt
I + s fo= e-Sly dt, (5)
assuming that e-"y(t) approaches zero, for sufficiently large values of s, as

t - , oo. Thus the transforrn of dy/dt is expressed in terms of the prescribed
initial value of y and in terms of the transform of y itself.
If the results of Equations (4) and (5) are introduced into (3), there then
follows
(s - 1)
o
=
1
e-Sly dt =
I
s - a
- 1
or r= e-Sly dt
Jo
= a
(s -
+
IXs -
1- s
a)
. (6)
The original problem is now apparently reduced to the problern ofdetermining

a function y(t) whose Laplace transform is given by the right-hand side of
Equation (6). To determine such a function, we first expand this expression by
the rnethod of partial fractions, to obtain the equivalent forrn
1 =
o
e y dt --
-SI 1
a-ls-a
I -
Reference to Equation (4) then indicates that, since I/(s - a) is the transform
a
a-Is-l
1 . (6a)
of ea'. the first term of (6a) is the transform of ea'/(a - 1) and the second terrn
the transform of -ae'/(a - 1). Thus (6a) will be satisfied if we write
y = 1 (e"' - ae'). (7)

a - 1
when a =1= l. A corresponding expression in the case a = 1 can be obtained
by taking the ¡imit as a - , 1, in the forrn y = (t - I)e'.
With this express ion for y, the validity of the transition from Equations (1)
and (2) to (6a) is readily established when s > a. Still, it is by no rneans obvious
that (7) is the only solution of (6a). That is, while (1) is known to have a unique
solution satisfying (2), it is conceivable that (6a) could have several solutions,
only one of which would then also satisfy (1) and (2).
2.2. Definition and Existence of Laplace Transforms 55
However, direct substitution shows that Equation (7) actually does repre-
sent the solution of (1) and (2). Further, it can be shown also that (7) is the
onJy continuous solution of (6a).
Although this procedure has the advantage that the panicular solution
required is obtained directly, without first obtaining the general solution, it is
clearly desirable to simplify the procedure by eliminating the necessity of
carrying out certain general integrations in each case, and to determine in
what cases such a procedure is valido
In the remainder of this chapter, certain properties of Laplace transforms
are investigated and relevant formulas are tabulated in such a way that the
solution of initial-value problems involving linear differential equations with
constant coefficients, or sets of simultaneous equations of this type, can be
conveniently obtained. Thus, for example, use of the tabulated formulas will
permit immediate transition from Equations (1) and (2) to (6), and from (6a)
to the solution (7). Use of the methods to be given will, in general, introduce
a considerable saving in labor over the alternative procedures of Chapter 1.
Laplace transforms are also useful, for example, in connection with the
solution of certain problems governed by partial differential equations (and
integral equations), as will be shown in later chapters. Certain of the properties
developed in this chapter are of principal use in these later applications.
2.2. Definition and Existence of Laplace Transforms. The Laplace trans-

form of a function f(t), defined for positive values of t, is frequently indicated
by the notation .,c[f(t)} and is defined, as a function of the variable s, by the
integralt
(8)
over that range of values of s for which the integral exists. The notation j(s),
or merely j, is often used in place of .,cU(t)}.
The integral (8) may fail to define a function of s, in particular, beca use of
infinite discontinuities in f(t) for certain positive values of t or because of
failure of f(t) to behave in a sufficiently regular way near t = O or for large
values of t. However, the presence of a finite number of finite discontinuities or
"jumps" will not, in itself, affect the existence of the integral.
A function f(t) is said to be piecewise continuous in a finite range if it is
possible to divide that range into a finite number of intervals such that f(t) is
continuous inside each interval and approaches finite values as either end of
any interval is approached froro the interior. Such functions may thus have
finite jumps at points inside the range considered. At such a point, say t = t o ,
different limits are approached by f(t) as t approaches t o from the right (that is,
tSome authors replace definilion (8) by the definition

ZU(t)} = s fo~ e-" f(l) dI.
56 Tbe Laplace Transform
from larger values of t) and from the left (from smaller values). These two limits
are called right-hand and /efl-hand Limits, respectively, and when necessary are
conveniently indicated by the respective notations
lim f(l) = f(t o+)
r-ro+
and
lim f(t) = f(t 0 - ).
(--ro -
In il.lustration, if .fCt) is defined to be unity when O < t < 1 and zero elsewhere,
then .f(t) is piecewise continuous over any range. There follows also, for example,
.f(l +) = O and .f(l-) = 1. If .f(t) is l/.vt when O < t < 1 and zero elsewhere,
then .f(/) is piecewise continuous in any range not inc1uding / = O as an interior
or end poin t.
In the developments of this chapter, we consider only funclions which are at
leasl piecewise continuous in every posilive range no! inc/uding zero as an end
point. Then, if we write (8) as the sum of three integrals,
oC{f(t)} = r e-"f(t) dI = (( + r+ S;}-"f(t) dI, (8a)
the second integral on the right exists for all positive finite values of tI and T.
If, in addition, f(t) approaches a finite limit as t ~ 0+ or if 1 f(t) 1 - , 00 as
t ~ 0+ in such a way that for sorne nurnber n, n < 1, the product tnf(t) is
bounded near t = 0, then the first integral of Equation (8a) exists.
Finally, a sufficient additional condition to guarantee the existence of the
third integral of (8a), at least for sufficiently large values of s, is the require-
ment that f(1) belong to tbe rather extensive class of "functions of exponential
order." A function f(t) is said to be of exponentia/ arder ir, for sorne nUID ber
so, the product e-'" 1 f(t) 1 is bounded for large values of t, say for t > T. lf the
bound is denoted by M, then there follows, when t > T,
e-"'lf(t)1 < M or 1 f(t) 1 < Me"'. (9)
Thus, though f(t) may become infinitely large as t ~ 00, we see that 1 f(t) 1
must not "grow" more rapidly than a multiple of sorne exponential function
of t. We say that f(t) is "of the order" of e'" and frequently write f(t) = O(e"').
In particular, if lim,_~ e-'" 1 f(t) 1 exists (and is finite) for sorne So > O, then for
sufficiently large values of t the product must be bounded, and hence f(t) is of
the order of e"'. The limit may, of course, be zero, in which case we a1so write
f(t) = o(e"').
We note that any bounded function is of exponential order with So = O. Other
examples are ea, (with So = a), ea, sin bt (with So = a), and 1" (with So any posi-
tive number no matter how small). The function e" is not of exponential order,
since e-sor e(~ = e rLsQ { is unbounded as t ~ 00 for al! values of So-
If f(t) is piecewise continuous and of exponential order, then its integral
f~ f(u) du is continuous and is o/so of exponential order. Although it cannot
2.2. Definition and Existence of Laplace Transforms 57
be said in general that derivatives of functions of exponential order have the

same property, this is true in most practical cases.
As an example of a reasonably simple exceptional case, we notice that
though sin (e") is bounded, and hence of exponential order, its derivative
2te" cos (e") is not of exponential order.
In case f(t) is of exponential order, and hence satisfies Equation (9) when
t :> T, tben there follows
I e-Uf(t)1 <e-U·ivJe'" = Me-U-,o)'.

Hence, since we have assumed f(t) to be piecewise continuous and since
f; e-('->O)< dt exists if s :> so, it follows that the third integral in (8a) then also
exists when s:> so·
Thus, in summary, ¡he Laplace transform of f(t) exists, when s is sufficiently
large, if f(t) satisfies the following conditions:
(1) f(t) is continuolls or piecewise continuous in every finite interval

ti -< t-< T, where ti:> o.
(2) t" I f(t) I is bounded near t = O for sorne number n, where n < l.
(3) e-'" I f(t) I is boundedfor large values of t, for some number so.
Although the transform may also exist in other cases, these conditions are
sufficiently weak to include most functions occurring in practice.
For reference purposes, it is noted that whenever any integral of the form
exists for s = so, it exists also for all s such that s >- so. AIso, it is then tme that
( 10)
when c >- so, that
.E...
ds
S= e-uf(O dt = S= (.E...e-")f(t)
o ods
dt (11)
when s :::: so, and that
( 12)
when So :s: IX -< fJ < oo. The remarkable fact that a1l these operations can be
effected under the integral sign, for any convergent Laplace transform, is of
frequent usefulness.
The direct ca1culation of Laplace transforms may be illustrated by the
following simple cases:
(s > O). (13a)

_ e-(,-a" 1=
(s> a). ( 13b)
s - a o s - a
.c(sin al} = l~ e-u sin al dI =

s~
,
e-"
+a 2
( .
S SID al a cos at) [
a ( 13c)
(s> O).
2.3. Properties of Laplace Transforms. Among the most useful properties

of Laplace transforms are the following:
.c(af(t)+ bg(l)} = ajes) + bg(s). ( 14)
.c{dnf(t)} = snj(s) _ [sn-If(O+) + sn-2 df(0+)

~n ~
n_Jd2f(0+)+
+ S d 1-' ( 15)
.c{I f(u) dU} = + j(s) ( 16)
.c{e"'f(l)} = jes - a). ( 17)
If f(t) = {Og(t - t < a} . h a >- O,

WJt then j(s) = e-a< g(s). (18)
a) 1 >- a -
( 19)
oC{J; f(l - u)g(u) du} =j(s)g(s). (20)
In these equations a and b are constants, and n is a positive integer.

In al! cases except Equation (15), we suppose thezt the functions f(t) and g(t)
satisfy the conditions of page 57. In the case of(l5), more slringent restrictions
are imposed. These conditions are stated in connection with the proof to be
given in this section (see page 59).
Equation (/4) expresses the linear property of Laplace transform s. Its proof
follows directly from the definition, in view of the corresponding linear prop-
erties of the integral.
Example 1. B y usin g Equations (1 4) and (13b), we obtain

1 -a, l =
n( . h
"" SID al
1_
-
n (_1 a' _ _
""1 2 e 2 e J
1
2(s _ a)
_ 1 _ a
2(s + a) - S2 _ a 2
.
•
Equation (15) s ta tes one of the most important properties of Laplace trans-
forms. It expre sses the transform of any derivative of a function in terms of
the transform of the function itself and in terms of the values of the lower-
2.3. Properties of Laplace Transforms 59
order derivatives of the function at t = °

(or, more precisely, the values ap-
°
proached by the se derivatives as I - , from positive values). We consider first
the case when 11 = 1, for which . from the definition,
.r. {df(I)}
dI
= J'=e-U df(t)
o dI
dI.
An integration by parts gives
J-o
e- U df(l) dI
dI
= e-uf(t) 1=
o
+s J-
o
e- Uf(l) dI
if f(l) is continuous and df(l) fdl is piecewise continuous in every interval (0, T).t
But since f(t) is of exponential order, the integrated part vanishes as 1 - 00
(for s > so), and there follows
.r.{d~~I)} = sj(s) - f(O+). (ISa)
Similarly, in the case 11 = 2, integration by parts gives
= e -" df(t) 1=
dI o
+s J-
o
r " df(l) dI
dI
= e-" df(t)
dI o
1- + s.r.{df(t)}
dI
,
if dffdl is continuous and d 2ffdl 2 piecewise continuous. If dffdl is also of expo-

nentiaI order, the integrated part again vanishes as 1 - 00 and, making use of
Equation (ISa), there then follows
.r. {d 2f(t)}
2
= S2 j(s) _ sf(O+) _ df(O+) . (l5b)
dl dI
Equations (ISa) and (lSb) are special cases of (15). The general proof of (15)
follows by induction from the general result
.r. { d"f(I)}
dI"
-_ s.r. {d"- If(I)} -
dl"-I
dn-'f(O+) ,
dI" I
ir d"-lf(l) fdl"-1 is continuous and d"f(l) fdl" is piecewise continuous, and ir
f(I), df(l) fdl, .. . , d"f(t) fdl" are al! of exponential order. This result is obtained,
as in the special cases n = 1 and 2, by an integration by parts.
We thus obtain the following result:
Equalion (15) is va/id if f(l) and ils first n - l derivalives are cominuotls
over every interva/ (O, T), if d"f(t) fdl" is (at least) piecewise conlintlotls over every
inlerva/ (0, T), and if f(t) and ils firsl n derivalives are of exponen/ia/ arder.
tUnless these conditions are satisfied , tbe formula for integration by parts may not be valid.
Example 2. If we take f(1) = sin al, then, rram (13c),

- a
fl.s) = s' + a2·
Equation (ISa) then gives
J3[ a cas at) = s ,~
s"' a
2 - sin O
or, using (14),
J3[casat) =
s-, ~ a-,. •
In particular, for the c1ass of functions considered, we see that if a funetion
and itsfirst n - 1 derivatives vanish at t = °
(or as t - , 0+), the transform of
its nth derivative is obtained by multiplying the transform of the funetion by s".
Applications of this fact are c10seJy related to the use of the operational nota-
tion D" f(t) to represent d"f(t)jdt".
Equation (16) is established by similar methods. Again making use of
integration by parts, and recalling that
d
dt Jor' f(u) du = f(l),
we obtain
l -
~ -f(5),
S
the integrated part vanishing at the upper limit (for sufficiently large values
of s), since f(t), and hence J~ f(u) du, is of exponential order. Thus, in general,
if a funetion is integrated over (O, t), the transform of the integral is obtained by
dividing the transform of the funetion by s.
If the lower Jimit differs from zero, the formula
J3U: f(u) dU} = + +s:

les) - f(u) du (21)
is easily established.
Equatiol1s (17) and (J 8) express the so-ealJed translation properties of the
Laplaee transformo The proof of the forroer property folJows immediately
from the definition, sinee the transform of ea' f(t) is given by
s: e-"[e a1(t)] dt = s: e-C,-a''/(t) dt,
and the last expression differs from les) only in that s is replaeed by s - a_
2.3. Properties of Laplace Transforms 6]
Thus, if a function is multiplied by e"', ¡he transforrn of lhe result is obtained by

replacing s by s - a in the lransform of the original function. It is seen that
if ¡(s ) is plotted as a funetlOn of s, the representation of the transform of
e"'j(t) is thus obtained by shirting or "translating" the transform of f(t)
through a units in the positive direetion of s.
Example 3. Ifwe take f (l) = sin bl, Equation (l3e)givesj(s) = bl(s! + b 2 ), and(l7)
then gives
"c [ea, sin bt} = ( ~" b 2
s-a",. •
Suppose now that a funetion is defined to be gel) for 1 :> O and to be zero
for negative values of t. Its transform may be denoted by g(s). If the given
funetion is translated through a units in the positive direetion of 1, and so
beeomes g(t - a) when 1 :> a and zero otherwise, Equation (18) states that the
transforrn of the translated funclion is obtained by rnulliplying the lransforrn
of the originalfunction by e-a,. To establish this property, we notiee that, sinee
the translated funetion vanishes when O -<: 1 < a, its transform ís defined by
the integral
r e'''g(l - a) di.
If t is replaeed by t + a, and the lower limit of tbe integral is ehanged aeeord-

ingly, this integral beeomes
L~ e-,"+al gel) dI = e-a, L~ e ' ''g(l) dI = e-a'g(s),
in aeeordanee with Equation (18).
Example 4. lf f(t) = sin (1 - lo) when I >- lo and f(t) = O when t < t o, there follows
= sin l, and, from (l3e), g(s) = (S2 + 1)-1 Thus Equation (18) gives
g(t)
e-Sl~
"c[f(l)} = S2 + l'
Conversely, if this relationship is known, a rever sed argument serves to determine f (t) .
To establisb the property of Equation (19), we merely differentiate both

•
sides of the equation
n times with respeet to s. Differentiation under the integral sign is va lid for al!
vaJues of s for whieb the transform exists, as was stated at the end of Seetíon 2.2.
Examp/e 5. To find the transforrn of t", EquatioD (19) states that we merely differen-
tiate the transform of unily n times with respeet to s and multiply the resuIt by (-1)".
We thus obtain
oC(t"] d"(l)
= (-1)"- - = n!
-,
ds n s sn+ I
where n is a positive integer or zero, with the usual convention that O! = 1. •
2.4. The Inverse Transform. In applications of Laplace transforms we

frequently encounter the inverse problem of determining a function which has
a given transformo The notation oC-'[F(s)} is conventionally used for the inverse
Laplace transform of F(s); that is, if F(s) = oC[f(t)}, then we write also
f(t) = oC-'[F(s)}.
The notation f(t) ~ F(s) is also frequently useful.
To determine the inverse transform of a given function F(s) it 1S thus
necessary to determine a function f(t) which satisfies the equation
r e-"f(t) dt = F(s).
Since the unknown function f(t) appears under an integral sign, an equation
of this type is called an integral equation.
In more advanced works it is pro ved that, if this equation has a solution,
then that solution is unique. Thus, if one function having a given transform is
kl1own, it is the only possible one. This result is known as Lerch's theorem.
More precisely, Lerch's theorem states that two functions having the same
transform cannot differ throughout any interval of positive length. Thus, for
example, Equation (13a) shows that the continuous solution of
r= e-"f(t) dt
Jo
= _1
s
is f(t) = 1; that is, oC - l[S- l] = 1. However, it is clear that if we take f(t) to be,
say, zero at t = 1 and unity elsewhere, or otherwise redefine the function f(t)
at a finite number of points, the value of the integral is not changed. Hence the
new function is also a solution. Such artificialities are, however, gene rally of no
significance in applications.
Although the direct determination of inverse transforms involves methods
outside the scope of this chapter,t extensive tables of corresponding functions
and transforms are available in the literature, and their use (in conjunction
with the use of the properties listed in Section 2.3) is sufficient for many pur-
poses. A sbort table of this sort is presented on pages 67 and 68.
It should be pointed out that not all functions of s are transforms, but that
the class of such functions is greatly restricted by requirements of continuity
and satisfactory behavior as s ~ oo. A useful result in this connection is the
following: lf f(t) is piecewise continuous in every jinite intervalO <: t <: T and
is of exponential order, then j(s) ~ O as s --+ 00 ; furthermore sj(s) is bounded
tSee, however, Sections 11.2 and 11.3.

2.5. The Convolution 63
as s - CQ . The proof follows from lhe fact that in such cases

1/(0 1< Me"" and le-"/(OI < Me-" -",I ,
for sorne fixed co nstants So and Al. Hence we have
M
s - So
The theorem stated then fo ll ows from the fact that M !(s - so) approaches zero
and [sj(s - so)]M is bounded as s ~ ca . Thus, such functions as 1, s/ (s + 1),
I /~s, and sin s cannot be transforms of functions satisfying the co ndition s
stated .
It should also be noted that if fU) is continuous ond df(t)/ dt is piecewise
continuous in every finit e intervo! O <: t <: T, and if f(t) and df(t) / dt are 01
exponen/ia! order, Ihen
lim des) = f(O+). (22)
This result follows from the fact that in this case the preceding theorem states
that the left-hand side of Eguation (ISa) vanishes as s - ca . Tt is useful in
those cases when only the initio! ,'o!ue of f(l) is reguired and the /rans(orm off
is known.
2.5. The CODYolution. lt frequently happens that , a1though a given function

F(s) is n o t the transform of a known functi o n, it can be expressed as the product
of two function s, each of which is the transform of a known function. Thus, it
may be possible to write
F(s) = j(s)g(s),
where j(s) and g(s) are known to be the transforms of the functions f(t) and
g(I), respectively. We suppose that these functions satisfy the conditions of
page 57. In this case, Equotion (20) states that the product j(s)g(s) is the trans-
form of the function defined by the integral f~ f(t - u)g(u) duo This integral is
called the convolution of f and g, and may be denoted by the abbreviation
I*g. It is indicated by the symmetry inj and g thatfand g can be interchanged
in the con volution, tha t is, f "g = g* f. Before ou tlin ing the proof of Eq uation
(20), we illustrate its application.
Example. To determine .)3-'(a /s(s - a)] we may refer to Equations (13a) and (13b)
and write the given function of s in the form
a 1 -
S s - a = !(s)g(s),
wheref(t) = a an d g(t) = ea,. Equation (20) states that lhe product is the transform of
the function
II the functionsfacd g are interchanged, there follows alternati vely
as before. The same result is obtained without making use of the convolution, in this
case, by using (16) acd (l3b), or by expanding the product by the method of partial
fraction s in the form
1
s - a s
and using (13a) and (l3b).
•
Equation (20) can be obtained formally as follows. From the definition,
tbe right-hand side of (20) can be written in the form
j(s)g(s) = [ [ e-,uj(v) dVJ[[ e-'"g(u) duJ
= fo~ t~ e-,Cu+ulj(v)g(u) dv du
= r g(u)[[ e-,(u+ulj(v) dvJ du
if different "dummy variables" of integration (v and u) are used in defining the

two transforms . If, in the inner integral of the last form, we replace v by a new
variable t with the substitution
v = t - u, dv = dt,
there follows
and hence
j(s)g(s) = r [f e-uj(t - u)g(u) dI] du o
Interchanging the order of integration in tbe double integral and changing tbe
limits as indicated in Figure 2.1, we then obtain formaJly
j(s)g(s) = r [S~ e - uj(t - u)g(u) du] dt
= fo=e-u[S/(t - u)g(u) dU] dt
= J2{S/(t - u)g(u)du}
n 2.6. Singularity Functions 65
io accordaoce with Equatioo (20). The interchange of order of iotegration can

be shown to be legitimate, by using appropriate limiting processes, when f
and g satisfy the assumed conditions.
u f(t)
u=t
IS
11
1
lo
• Figure 2.1
t
Figure 2.2
t
1,
2.6. Singularity Functions. Consider the function f(t) which has the yalue
lito when ° < t < t o and is zero elsewhere (Figure 2.2). We then haye
[f(i) dt = ff(i) dt = [;
tbat is, the area under tbe graph representing f(t) IS unity. The transform of
tbis function is found to be
e oC{f(t)} = _1
to
r"
.1 o
e-U dr = 1
v
Now as fa _ . O, the magnitude of f(r) oyer (O, to) ¡ncreases without limit, while,
at the same time, the length of the interyal (O, t o) shrinks toward zero in such a
way tbat the integral f¿ f(t) dt retains the yalue of unity.
In the limit we haye the occurrence of a "function" which is injinite at the
point t = ° and zero elsewhere, but which has the property that its integral
across t = ° is unity. Making use of L'Hospital's rule, we find
. 1 - e-sto . se-st~
ltm
10 ..... 0 Slo
= ltm - -
10 ..... 0 s
= l.
Tbat is, the transform of f(t) approaches unity as t o - , O.

If t represents time and f(t) force, then f¿ f(t) dt represents the impulse of
the force f(t) acting oYer the time interyal (O, to). Hence, as t o ~ 0, we may
speak loosely of a resulting "unit impulse" at t = 0, due to "an infinite force
acting oyer a zero time interyal." 1 n yiew of this interpretation, tbe limiting
form of f(t) is frequently called the unit impulse function. If we denote it by
o(t), we are led to write
oC(o(t)} = 1. (23)
gU) Tbe "function" b(t) is abo often called

the Dirac della funclion. If, for example, I
1
represented dislance along, say, the centerline
tg of a beam and f(¡) represented the intensity
of a distributed load, then b(l) could be con-
t sidered as the formal representation of a con-
1 cenlrated unit load applied at the point t = 0,
--;,
'o and analogous interpretations in other fields
are frequently useful.t
In a similar way, if g(t) has the value
Figure 2.3 -1/t6 when °
< t < to, the value + 1/15 when
lo < t < 2to, and the value zero eJsewhere
(Figure 2.3), there follows
r (t - to)g(t) dt = f'o (1 - to)g(l) dt = l.
That is, the moment of the area under the graphical representation of g(t),
about the point lo, is unity, whereas tbe (signed) area is zero. The transform
of g(t) is
Repeated use of L'Hospital's rule shows that, as t o ~ 0, oC[g(t)} ~ -s. The

limiting form of g(t), as t o ~ 0, is frequently referred to as the doub/et (or
dipo/e) function, because of certain interpretations relating to fluid f10w and
electric field theory. Denoting the negative of this limiting form by b'(t), we write
oC[b'(t)} = s. (24)
Witb the interpretation of t as distance and g(t) as load intensily, b'(t)
could be considered as the formal representation of a concentrated negative unil
moment applied at the point I = O.
Such functions, often called singularity funclions, are dealt witb rigorously
in the branch of mathematics known as the theory of distributions, and are of
frequent use in physical applications. Although they do not conform to the
restrictions of page 57, and although, in fact, they are not truefunctions, never-
theless ifformaluse of them leads to a result which is capable of physical interpre-
tálion, Ihen in practical cases the result may be accepted as correcto
tIn sorne situations it is necessary to replace the parent functionf(t) represented in Figure 2.2
by one which is conlinuolls (and perhaps also has a certain number of continuous derivatives)
for O <: t <: lo, befare effecting the relevant limit operation as lo ~ O. Clearly, this can be
done in many equivalent ways.
2.7. Use of Table of Transforms 67
If the singularity occurs at ( = 1, rather than at ( = O, we denote the coc-

responding functions by 0(1 - t ,) and 0 ' (1 - t,) and obtain (by limiting pco-
cess es analogous to those given aboye) the formal results
J?(o(t - t,)} = e - "', (25a)
(25b)
lt rnay be noticed that these results are also obtained by formally applying the
translation property (18) to Equations (23) and (24).
2.7. Use of Table of Transforms. A brief table of corresponding functions

and transforms is presented in this section , to facilitate the deterrnination of
botb direct and inverse transforms. The first ten pairs represent general rela-
tionsbips proved in Section 2.3. In pairs (T3, 4, 5) the conditions O] page 59 are
assumed, whereas in the remaining pairs the less restrictive conditions O] page 57
are implied.
Table 1. Laplace Transforms
Transform Function
TI j(s) = J:! [f(t)} = J; e-U /(t) dt Jet)

T2 ajes) + bg(s) a/(t) + bg(t)
sj(s) -/(0) d/(t)
T3 di
s2j(S) _ s/(O) _ df(Ol d 2 /(t)
T4 dt dt"
e _ • dk-I/(O) d'/(t)
T5 soles) - L: S,-k dt k- ' --¡¡¡;;-
r k- I ti tiales ti t i mes
~~
:i
e T6 ;,j(s) J~ ... j";ftr)dt .. - dI
( _1l,dnf(s)
-) T7 ds'
t'/(t)
T8 f(s - a) ea'/(t)
t) ra, fes) (a > O) f(t - a) (t > a)
T9 {O (t < a)
it
TlO j(s)g(s) J~/(t - u)g(lI)du = f~/(u)g(t - u)du
ly
Jf Tl1 1
I
s
le 1
r- T12
s+a
'e- Tl3 I _1_(e-.' _ ra,)
(s + a)(s + b) a - b
T14 s _I_(be- .' - ae-a,)
(s + ajes + b) b-a
:.2 a
Tl5 s" -r a l sin al
:s)
be T16 s cos at
s¿. + a 1.
Table 1. (Conlinlled)
Traosform Function
T17
a sinh al
TI8 cosh al
S2 - a2
2as t sin al
TI9
(S2 + 0 2)2
$2 _ a2
T20 t cos al
(S2 + a2)2
T21 2a 3 sin al - al cos al
(S2 + a2J2
2as si oh al
T22 (S2 - 0 2)2
1
T23
$2 + a2
1 cosh al
(S2 _ 02)2
20 3 cosh sinh
T24 al al - al
(S2 - a 2 )2
a e-bt sin al
T25
(s + bJ2 + a 2
T26 s+b e-be cos al

(s + b)2 + 0 2
T27 40 3 sin al cosh al - cos al sinh al
S4 + 4a4
2a:?s sin al sioh
T28 al
s' + 40'
20 3 sinh al - sin al
T29
S4 - a4
T30 2a 2 s cosh al - cos al
S4 - a4
T31 I 6(1 )
T32 6(1 - 1,)
T33 s 6'(1)
T34 Ó' (I - 1,)
I (n> O)
T35 sn (n - I)!
n! In (n > -1)
T35a
In-le-al
T36
(s + a)n (n - I)!
(n > O)
T37 S (n - J) - al ln -2r.' (n > 1)

(s + a)n (n - I )!
a 2n -¡
T38 (s2 + a2)n 2n len _ I)![.J~ (al)n-I / 2Jn_ I/ 2(al)] (n > O)
a 2 ,.-2s
T39
(S2 + a "2)n 2 n 1(:1_ l)![A(at)n-3 / 2Jn-3 / 2(ao] (n > !)
T40 (S2
a 2 ,.-1
a2)n 2n 1(; _ 1)![.J ~ (a l )n-I / 2In_1 / 2(al)] (n > O)
a 2 ,.-2 s
T41 (S2 _ a 2)n 2 n 1(:1_ 1 )![.J~(at)n- 3 /2 In- 3/2(ar)] (n > !)
68
2.7. Use of Table of Transforms 69
pairs (TI 1-30) either have been established in exarnples or can be easily
obtained frorn established results by using certain of the general properties .
Only frequently occurring basic forrns are listed; other forrns are readily deduced
from those gíven. Paírs (T3l-34) involve the singularity functions of the preced-
ing section.
Pair (T35) was derived in Example 5 ofSection 2.3 when n is a positive inte-
ger and (T36, 37) follow , in this case, by virtue of (T8) and (T3), respectively.
The formulas are valid also (under the given restrictions on n) if n is nol an
integer, if (n - 1)! is interpreted in a way to be defined in Section 2.9.
Pairs (T38-41) are included principally for future reference. In these pairs
J m and 1m are certain functions known as Besse/ funclions of order m. These
functions are to be treated in Chapter 4. If, in these pairs, n is zero or a positive
integer, then the order of the function involved is half an odd integer. In such
cases the functions in brackets, in the.right-hand colurnn, can be expressed in
terms of products of polynomials and either circular or hyperbolic functions.
These expressions are given for m = -1-,1-, ... , ! in Table 2, page 70 (where x
is written for al). The expressions for m = 4, 1f, and so on, can be obtained in
terms of these expressions by use of the recurrence formulas listed at the foot of
the table.
Although the transforms of the simpler functions of frequent occurrence
in practice can be obtained directly from the table, or by direct integration,
the determination of inverse Iransforms may frequently involve a certain amount
of manipulation. In this connection, it should be observed that if n is a posilive
inleger, al! funclions of I appearing in Ihese lab/es (except the singularity func-
tions), as wel! as al! Iheir derivalives, are continuous everywhere and are of expo-
nenlialorder. Hence it follows that ifn is aposilive inleger, all proper/ies (TI-lO)
can be applied lO all succeeding pairs in Ihe lab/e, excepl for (T3l-34).
Pair (T3) is particularly useful in the determination of inverse transforms,
when feO) = O. Reference to Equation (22) shows that this is so if sj(s) tends
to zero as s - oo. Hence it follows that iflim, _~ sj(s) = O, Ih en
,e-'(sj(s)} = di. (26)

di
Erample 1. To determine ,e -, [S2 / (S2 + 4)2), we fírst obtain from (T19) the result
n_,{
"" (S2 +S 4)2
} _
-"4/
1 .
Sin
2
/.
Hence, using Equation (26),

n_,{(s2+4)2 (1"4 /sm. 2)/ --"4I . 2 + leos 2/.
""
S2 } _ d
-dI
/ SIn /
•
When a given function F(s) , whose inverse transform is required, is the
ratio of two polynomials in s, the method of partial fractions can be used to
express F(s) as the sum of a number of terms whose inverse transforms can be
Table 2. Bessel Functions oC Order HalC an Odd lnteger
(cos x)lx (cosh x)lx

smx sinh x
sm x - x cos x x cosh x - sinh x
(3 - x2) sin x - 3xcos x (3 + X2) sinh x - 3x cosh x
(1.5 - 6X2) sin x - (l5 - x2)X cos x (15 + X2)X cosh x - (15 + 6X2) sinh x
(l05 - 45x2 + x 4 ) sin x - (l05 - IOx 2)x cos X (l05 + 45x2 + x 4 ) sinh x - (105 + IOx 2)x cosh x
Recurrel/ce Formulas
2.7. Use of Table oC Transforms 71
determined fram [he tableo In such cases, if the inverse transform does not
involve the singularity functions of Section 2.6, the degree of the denominator
must be greater than that of the numerator. In particular, if
F(s) = N(s),
D(s)
where D(s) is a polynomial of degree n wi[h n distinct real zeros s = a ¡, a 2 , . . . ,
an , and N(s) is a polynomial of degree n - 1 or less, there follows (see Problem
17)
N(s) N(a¡) 1 ,N(a n ) 1
D(s) - D'(a¡) s - a¡ ., D'(a n ) s - a n
=
m~
i:
¡
~(am)
D (a m )
1
s - am
and hence, from (Tl2),
(27)
If certain of the zeros of D(s) are repeated or complex, recourse may be had to.
conventionaJ methods of expansion in partial fractions.
Example 2. To determine .,e-¡[(S2 + 1)/(S3 + 3s 2 + 2s)}, we write

N(s) = S2 + 1, D(s) = S3 + 3s z + 2s = s(s + 1)(s + 2).
With a¡ = O, a2 = -1, a3 -2, there follows
N(a¡) = 1, D'(a¡) = 2,
N(a2) = 2, D'(a2) = -1,
N(a3) = 5, D'(a3) = 2,
and Equation (27) gives

n _
"--
¡{
s3 +
sZ
3s2
+ I
+ 2s) -
1. __
1
2
_ 2 -,
e
+..s.- - 2'
2 e .
•
Example 3. To determine .,e-1[1/[(S + 1)(S2 + 1)]}, we first assume an expansion of
the form
1 A Bs+C
(s + 1)(s2 + 1) = s + J + S2 + 1.
After clearing fractions, we require the equation to be an identity and obtain A - -B
= C = ±. Hence
1 1[ 1 1 s ]
(s + 1)(s2 + 1) = '2 s + J + S2 + 1 - S2 + J
and the use of (TI2, 15, 16) gives the inverse transform -±(e-' + sin t - cos t). •
The usefulness of (T25, 26) in determining inverse transforms should not be

overlooked.
Example 4. To determine .,e-'(s/(s' -+- 4s T 5)], we first write

s s (s + 2) - 2
S2 -7" 4s + 5 = (s + 2)2 + 1 = (s + 2)2·' 1
Pairs (T25, 26) then give the required inverse transform
e- "(cos / - 2 sin /).
2.8. Applications to Linear Differential Equations with Constant Coefficients.

•
It follows from the property (T5) and its specíal cases (T3, 4) that any ordinary
ünear differential equation with constant coefficients, with prescribed initia/
conditions at t = O, can be transformed immediately to a linear algebraic equa-
tion determining the transform of the required solution, provided that the right-
hand member of tbe equation has a transform o The solution then is to be
obtained as the inverse of the transform so determined.
If the right-hand member is of -exponential order, the same will be true of
the solution and of those derivatives whose transforms are involved, and all
tbe relations of Table 1 are appropriate.
We take as a simple example the case
of forced vibration of a mass m attached
to a spring with spring constant k . (That
is, the force exerted on the free end of the
spring is assumed to be proportional to its
{(t) displacement x from the position of equi-
~x _ _ librium, the constant of proportionality
being k.) If the applied force is f(c) and if
no damping is present (Figure 2.4), the
Figure 2.4
differential equation of motion is
d 2x
m d?/-
+ kx = j(/). (28)
Furthermore, if the mass is assumed to be at rest at equilibrium when t = O,

the initial conditions
x(O) = dx(O) = O (29)
dt
must be satisfied. With the transforms of X(I) and f(t) denoted by x(s) and
j(s), respectively, the transform of Equation (28) becomes merely
ms2x + kx = j.
Thus, if \Ve write
2
OJo
k,
=- (30)
m
the transform of the required solution is
_ 1 j
x = -m- -S'2~+';----úJ?~ (31)
2.8. Applications lo Linear Differential Equations with Constan! Coefficients 73
Since 1/(5' + W5) is the transform of (sin wot)/w o, this product can be considered
as the product of the transforms of (sin wot)/mw o and J(t), and hence use of the
convolution (TIO) gives the solution
x = -I-
mw o o
J'
J(u) sin wo(t - u) du, (32)
in terros of an arbitrary force function J(t).

However, in place of specializing this general form, it is often more con-
venient in specific cases to derive the required solution directly from Equation
(31). We consider several cases of interest.
(1) Suppose that an instantaneous impulse of magnitude [ is applied just
after the time t = O. Then J(t) = [. b(t) and ](5) = [.} = l. Hence we have
x- = -[ 1 -l
-. (t O
> ·). (33)
. m 52 + W5 ' x = SIn
mw o
wot
Thus the motion in this case is a sinusoidal vibration of amplitude l/mw o and
angular frequency W o , following the application of the impulse; W o is known
as the natural freqllency of the system. It should be noticed that here the initial
condition dx(O)/dt = O apparently is not fulfilled. However, the velocity cannot
vanish when the impulse is applied, since the momentum mv = 1 must be
imparted by the impulse, in accordance with Newton's laws of motion. Here
we may suppose that x and v = dx/dt are zero throughout an infinitesimal
interval following the time t = O, and that on the subsequent application of
the impulse the velocity abruptly takes on the value l/m and a sinusoidal
motion ensues. Interpretations of this general nature are frequently necessary
in dealing with the idealized "singularity functions."
(2) If a sinusoidal force J(t) = A sin wt is applied, there follows
- =Aw
x - l ,
m (52 + W5XS 2 + w 2 )
or, after expanding in terms of partial fractions,
_ Aw
x = m(w2 _ w~)
(1+ w~ -
S2 52
1)
+w 2 .
Hence x = Aw (sin wot _ sin wt)

m(w 2 - w~) Wo w
or x = A
(2 2) (
W '
SIn wot - W ')
o SID wt . (34)
mw o w - W o
Thus, if w * w o , the motion is compounded of two modes of vibration, one

(the natural mode) at the natural frequency w o , and the other (theforced mode)
at the frequency of the imposed force. In case the system is excited at its natural
frequency (w = w o), the motion can be determined by considering the limiting
form of Equation (34) as w ~ w o , or, more easily, by noticing that in this case
- Aw o }
x = -m- ('-S02'--+--'-::--W'--~'-;):?2
74 The Laplace Transforrn
Hence we obtain, from (T21),
x = 2 A o (sin wot - wot cos wot). (35)

mújo
Thus the last term of Equation (35) shows that, when the exciting frequency
equals the natural frequency, the amplitude of the oscillations increases in-
definitely with time. This is the case of resonance.
Similarly, if f(t) = A cos wot, there follows
A .
x=2 tSlOwot.
mw o
(3) If a constant force f(t) = A is applied when t > O, there follows
-
x
1
= -A -.,.--",...-,---",
m S(S2 + W5)
and hence, from (T 11, 16),

= m!5 ( +- S2 : - W5) ,
A (36)
x = --2 (1 - cos wot).
mw o
Thus, in this case, the mass oscillates with its natural frequency between the
points x = O and x = 2A/mW5 = 2A/k, when damping is absent.
(4) If constant force is applied only over the intervalO < 1 < lo, and no
force acts when 1 > lo, there follows j = (A/s)( 1 - e-"'), and hence
-
x =
A
m
1- +e-'"W5)
S(S2 =
A
m
[1+ w~) -
S(S2 S(S2
e-'"
+
]
W5) .
The inverse transform of the 11rst term is given by Equation (36), and, in view
of (T9), the inverse of the second term is zero when 1 < lo and is obtained by
replacing 1 by 1 - lo in (36) when 1> lo' Hence we have,. when O < 1 < lo,
A (37a)
x = --2(1 - cos Wol);
mw o
and, when 1 > lo,
A
x = --2 [(1 - cos wot) - [1 - cos wo(t - lo)]}
mw o
A
= T[cOS wo(t - lo) - cos Wol]
A
= 2k (Sin -}Wol o) sin W o (t - -}lo)' (37b)
Thus, while the force acts (O < 1 < lo), the mass oscillates at its natural fre-
quency, with amplitude A/k, about the point x = A/k; however, after the
force is removed (t > lo), the mass oscillates about the point of equilibrium
(x = O), at the same frequency, but with an amplitude (2A/k) sin 1molo. If
lo = 271/w o = T, where T is the period of the natural mode of vibration, then
2.8. APplications lo Linear Differenlial Equations wifh Constanf Coefficienls 7S
x - O when l >- lo, SO that the mass returns to its equilibrium position as the
force is removed, and then remains at that position.
It is seen that in the preceding example, and in similar cases, the use of
tables permits the determination of the transform of the solution by purely
algebraical methods. This is true, however, only in cases when the coefficients
of the linear differential equation are constants, and the usefulness of the
present methods is mainly restricted, in such applications, to such cases. (See,
however, Problems 46 to 49.)
The use of Laplace transforms is particularly advantageous in the solution
of initial-value problems associated with sets of simultaneous linear equations .
We illustrate the procedure by considering an example.
We require the solutioo of the simultaneous equations
dx ,
dl .- Y = e,
(38)
dy
di
+ x = sin l,
which satisfies the conditioos

x(O)~ 1, y(O) = O. (39)
The transforms of (38) satisfying (39) are
sx- - y- I + 1,
=
s- I
from which we obtain, algebraically,

s s 1
x = (;-s----;-j7.X;...s',-+"---'1'""') + s2 + j + (s 2 + I)2 '
1 1 s
y=
(s - j Xs 2 + 1) S2 + 1+ (S2 + 1)2·
If the first terms on the right-haod sides of these equations are expanded ID
partial fractions, there follows
x
- =""21°[ s -l+i+s
1+ + j S2 S2 1+ (S2
2
+ 1)2 ,
J
(40)
- I [ l I S 2s ]
y = 2 - s - j - s' + 1+ S2 + 1+ (S2 + 1)2 '
and refereoce to Table 1 gives the required solution,
x = -Fe' 2 sin + cos I - I COS 1),
I
(41)
y = -te -e' - sin I + cos I + sio 1).
To illustrate the existence of exceptiooal cases which may arise in connec-
tion with simultaneous differeotial equations, we next attempt to find a solution
76 The Laplace Transfonn
of the equations
dx
di
+y = o,
(42)
d'x dv
di' + di +y = e',
satisfying the conditions
x(O) = 1, x'(O) = O, y(O) = O. (43)
The tran sformed equations are
sx+.9=I,
S2 x l~ (s + 1).9 = s + I l'
s-
from which there follow
- 2 1 _ I
x=-- Y = ---':"""'" (44)
s s- s - l
The inverse transforms of these expressions are then
x = 2 - et , y = e'. (45)
However, these solutions do not satisfy the last two 01" the prescribed initial
conditions (43). It is readily shown, by methods of Chapter 1, that the most
general solution of (42) is of the form
x -- , e- e', y = e'
where e is an arbitrary constant. Hence only the initial value of x is arbitrary,
and the problem as stated does not possess a solution.
This example shows that , in the case of simultaneous equations, although
the method of Laplace transforms will yield the correct solution if it exists, it
may al so supply an erroneous solution (which fails to satisfy certain prescribed
initia! conditions) if no true solution exists. Thus, in doubtful cases, the satis-
faction of initial conditions should be checked.
2.9. The Gamma FUQctioQ. In calculating the transform of 1", where

n> -1 but n is not necessarily an integer, we encounter a function, known as
the Gamma fun ction, which also occurs frequently in many other applications.
In this section we investigate certain properties of this function.
If, in the integral defining the transform of t",
(n>-l),
we introduce a new variable of integration by setting st = x, there followst
.c(t"} = -S"T
11 J- o
e-X x" dx (n> -1). (46)
tThe restrietion n > -\ is neeessary to ensure the eonvergence of the integral. If n -< - 1,
the [unetion t n does not have a Laplace transform, as here defined.
9 The Gamma Function
2 .. 77
The integral appearing in Equation (46) depends only upon n. Although it

cannot be expressed in terms of elementary functions of n, the same integral
with n (inconveniently) replaced by n - I is a tabulated function which occurs
frequently in practice and is known as the Gamma fimction of n, written r(n):
r(n) = r e-'x"-l dx (n > O). (47)
With this notation, Equation (46) can be written in the form
\ oC(t"} = r(n +
sll+ I
1) (n > -1). (48)
By repeated use of integration by parts (or by comparison with the result of

Example 5, Section 2.3), it follows that
r(n + 1) = n! (49)
if n is a positive integer, and that
r(l) = O! = 1. (50)
Thus it is seen that, if n > -1, r(n + 1) is a continuous function of n which
takes on the value n! when n is a positive integer or zero. Por this reason, the
Gamma function is often referred to as the generalized factorial function.
When n is not necessarily integral, an integration by parts leads to the result
r(n + 1) = fo= e-Xx" dx = -x"e- X

1: + n fo= e-xx"-l dx
= n fo- e-xx"-l dx (n> O),

r(n + 1) = nr(n) (n > O). (51)
This relation displays the characteristic property of the Gamma function.
Inductive reasoning then leads to the formula
r(n + N) = (n "' N - I)(n + N - 2)·· ·(n + I)nr(n) (n > O), (52)
where N is any positive integer. AIso, if n is replaced by n - 1, Equation (51)
can be written in the alternative form
r(n - 1) = r(n) (n > 1). (53)
n-I
Four-place values of r(x) over the interval 1 <: x <: 2 are listed in Table 3.
With such atable, Equation (52) can be used to evaluate the Gamma function
for arguments greater than 2, since any such argument differs from a value in
the tabulated range by sorne positive integer N. If we write x = n + N and
replace n by x o, where 1 <: X o <: 2, Equation (52) becomes
(x > 2, I <: X o <: 2). (54)
AIso, Equation (53) serves to determine values of the Gamma function for
arguments between zero and unity.
I
Table 3. Values of r(x) = (x - 1) !
x .00 .0 I .02 .03 .04 .05 .06 .07 .08 .09
1.0 J .0000 .9943 .9888 .9835 .9784 .9735 .9687 .9642 .9597 .9555
.1 .9514 .9474 .9436 .9399 .9364 .9330 .9298 .9267 .9237 .9209
.2 .9182 .9156 .9131 .9108 .9085 .9064 .9044 .9025 .9007 .8990
.3 .8975 .8960 .8946 .8934 .8922 .8912 .8902 .8893 .8885 .8879
.4 .8873 .8868 .8864 .8860 .8858 .8857 .8856 .8856 .8857 .8859
.5 .8862 .8866 .8870 .8876 .8882 .8889 .8896 .8905 .8914 .8924
.6 .8935 .8947 .8959 .8972 .8986 .9001 .9017 .9033 .9050 .9068
.7 .9086 .9106 .9126 .9147 .9168 .9191 .9214 .9238 .9262 .9288
.8 .9314 .9341 .9368 .9397 .9426 .9456 .9487 .9518 .9551 .9584
.9 .9618 .9652 .9688 .9724 .9761 .9799 .9837 .9877 .9917 .9958
For negative values of n, the function ren) is not defined by Equa.tion (47),
since the integral does not exist. However, it is conventional to extend the
definition in such cases by requiring that the recurren ce formula (51) hold al so
for negative values of n.t Since, for any negative value of n which is not an inte-
ger, there exists a positive integer N such that n + N is in the tabulated range
(1 < n + N < 2), we may then replace n + N by X o and n by x in (52), where
1 < X o < 2, to obtain
rexo) = (xo - 1)(x o - 2)·· ·(x + l)xT(x)
or, solving for r(x),
rex) = nxo) (x < 1, 1 < Xo < 2). (55)
X(X + l)(x ...,... 2) .. '(X o - 2)(x o - 1)
Equations (54) and (55) thus serve to determine values of the Gamma
function for real arguments outside the tabulated range. It should be noticed,
however, that since the denominator of Equation (55) vanishes when x is zero
or a negative integer, the Gamma function is not defined for these values, and
becomes infinite as these values are approached (see Figure 2.5).
lt will be convenient in later work to use the notation n! even in cases when n
is not a positive integer or zero, with the convention that in such cases n! is defined
by r(n + 1).
The value of r(~) is of particular interest. A well-known but quite indirect
method of determining this value is now presented. From the definition (47),
we have
n·n = r e-'z-I/2 dz.
With the change in variables z = X2, this integral becomes
n!) = 2 r e-" dx. (56)
tA difrerent method of definilioo yields r(n) for a1l complex values of 11 except for zero aod
negative integers, al which poiots r(n) does not remain finite.
•
2.9. The Gamma Function 79
T(n)
-4 -31, -2/ -1/ 1 2 3 4 n

/ ,
, ,
, I -1
),
I I
.e I -2
o ,I
, -3
~e
I
,
, -4
,
Figure 2.5
5) If the right-hand side of Equation (56) is multiplied by itself, and if the variable
of integration is replaced by y in one factor, there follows
la
j,
'0
[r(±))2 = 4(f e- x
' dx )(f e- Y
' dY) = 4 rr e-(x'+Y') dx dy.
This double integral represents the volume under the surface z = e-(x'~Y') in
Id
the first quadrant (x ::> 0, y ::> O). Changing to polar coordinates, we obtain
Jor J'o.- e-"rdrd8

n 2
[r(1)F = 4 /
= 4 ~ e-;"1»0= n.
Thus, finally,
ct
rm = ,,/n. (57)
1),
From Equation (56) we also ha ve the useful result
r e- x ' dx = ±,,/i. (58)
We include without proof the formula
6) r(x)r(l - , x) (59) =. n
sm nx
which can be shown to be valid for nonintegral values of x (see Problem 119 of
nd
Chapter 10) and which is of frequent use in applications of the Gamma function.
I
80 The Laplace Transfol'll!
It can be shown that the Gamma function is defined alternatively by the

limit
r(x) = lim n!
+ n) n
X
n~= x(x + 1)·· ·(x .

Hence there follows also
. n !n x
~¡~~ r(x + n,- 1) = 1. (60)
Thus in limiting operations involving Gamma functions, when n ~ 00, the

expression r(x + +
n 1) can be replaced by its approximation n !n x . We
indicate this fact by the notation
(n + x)! = r(x +n+ 1) ~ n!n' (n ~ 00), (61)
where the symbol ~ is to be interpreted as indicating that (61) iroplies (60).
A further limit of importance is of the form
liro r(n + 1) 1 (62)

,,_<><> -J2n n,,+(1/2)e 11
which we can write symbolically

n! = r(n + 1) ~ ,.J2nn(nle)n (n ~ 00). (63)
If n is a positive integer, trus approximation is known as the Stirling formula
for the factorial.
Proof of these relations is beyond the scope of this chapter.
REFERENCES
1. CAMPBELL, G. A., and R. M. FOSTER, Fourier lntegrals for Practical Applications,

Van Nostrand Reinhold Company, New York, 1947.t
2. CARSLAW, H. S., and J. C. JAEGER, Operational lvlethods in Applied Mathematics,
2nd ed., Oxford University Press, Inc., New York, 1948.
3. CHURCillLL, R. V., Operational Mathematics, 2nd ed., McGraw-Hill Book Company,
Inc., New York, 1958.
4. ERDÉLYI, A., ed., Tables of Integral Transforms, 2 vols., McGraw-Hill Book Com-
pany, Inc., New York, 1954.
5. WIDDER, D. V., The Laplace Transform, Princeton University Press, Princeton,
N.J., 1941.
tThis reference contains 763 pairs, most of which can be interpreted as functions and ¡heir
Laplace transforms. For this purpose, ifthe variableg is replaced by 1, then the column headed
"Coefficient G(g)" lists the function/(/) in the present notation and the column headed "Coef-
ficient F(f)" lists the corresponding Lap1ace ¡ransform J:.[f(t)) ~ fes).
lfm Probl ems 81
the
PROBLEMS
Sectjon 2.1
1. (a) Obtain the solution of the equation
dv
(60) dt - ay = ea<
the for whieh y(O) = Yo, by the method of Seetion 2.1. Assume that a =1= a.
We (b) Verify the solution so obtained.
2. (a) Obtain the solution of the equation
(61) dy
dt - ay=ea<
(60).
for whieh y(O) = Yo, by eonsidering the limit of the solution of Problem 1 as a --+ a.
(b) Verify the solution so obtained.
(62)
Seclion 2.2
3. Find Ihe Laplace Iransform of eaeh of Ihe following funelions, by direel inlegralion:
(63) (a) e'" eos kt, (b) tne-a< (n a posilive inleger),
rormula
( ) {Sin t (O < t < 71:), {O (O < t < a),
e O (t > 71:), (d) 1 (a < t < b),
O (t > b).
Seclion 2.3
4. Find Ihe Laplace Iransform of eaeh of Ihe following funetions:
(a) t" (b) t 2 e- J " (e) eos at sinh at,
(d) te' sin 2t, (e) t 2 sin at, (f) ea, eoshbt.
,Ucatíons, 5. Find Ihe Laplaee Iransform of eaeh of the following funetions:
rhematíeJ,
(a) d;¡;~t), (b) te'f(t),
N N
(e) ~ ant n, (d) ~ aneosnt.
n=Q IJ=O
~ompanj,
6. The Laguerre po/ynomia/ of degree n is defined by Ihe equalion

.ook Corno dn
Lit) = e' dtn(tne-').
Prineeton Prove thal
J?[Ln(r)] = ns! (s ~ Ir
~s and th' 7
IUmn heade . Pro ve Ihal, if j(s) = J?[f(t)], and if a> O, Ihen also
,aded "Co< (a) J?[f(at)] = !j(;), (b) J?-I[j(as)] = ! f( ~).
82 The Laplnce TransfOrJ
8. Prove that
I( dId)n /(l)!"I =
oC 'l 1 (-
d( d)n-,
I)n ds s ds
-
[sf(sll
when 11 is a positive integer. (Use induction.)

9. Let fU) = F(I) when O < I < a, and let /(1) be periodic, of period a, so tha
f(1 + a) =f(l). By writing
.B{f(t)} = J: e-"/(l) dI + r a
e-"/(t) dI + r: e-Hf(1) dI + ...
and transforming each integral in such a way that in each case the range of integration
is (O, a), show that
.B[f(1)] = J: e-" F(l) di [1 + e-a, + e- 2a, + .. 'l,

and hence, when S > O,
.B[f(l)} = J e-"
a
O
1 -
F(l) dI
e as
•
10. Apply the result of ProbJem 9 to the "square-wave function" for which F(l) = 1
when O < 1 < a/2 and F(t) = -1 when a/2 < 1 < a. Show that the transform of this
function is
(1 - e-a",)' 1 1 - e-a,. ' 1 as
s(l - e a,) - S 1 + e a,/2 -- -S tanh -4 .
""--,..,.--=---::-:+ - -
11. Show that, if f(l) is the "square-wave function" of Problem 10, then f~ f(u) du is
a "triangular-wave function." Ske~ch it and give the expression for its transformo
12. If /(1) is the "staircase function" such that f(l) = b when O < x < a, f(l) = 2b
when a < x < 2a, and so forth, show that
b
.BU(t)J = s(1 _ e a,)
13. (a) Show thal, if ¡(s) = .B [J(I)J, and if an interchange of order ofintegrations is
valid, then
r ¡(v) dv = .BV;I)}.
[The result is valid, for s sufficiently large, whenever f(l)/I has a transform.]
(b) Use this result to deduce the transforms
sin 1)
.B { -l-f = cot-Is, n(1 - 1 e-'}
"'"'( =
(s
Iog - +
s -1) .
r r
14. (a) By formally setting s = O in the result of Problem l3(a), obtain the formula
¡(s) ds = f;t) dI.
(The result is valid when lhe inlegral on Ihe righl exisls.)

Problems 83
(b) Use lhe resull of pan (a) to oblain lhe e valualions

~ ro< - e - h '
fo
' -o sin / d/ = 3:... , =----------:---"---- d t = lo g -
b
(a, b > O).
/ 1. t a
•J o
(e) Show that, if[(I) = e' , [hen neilher side of the relation of par! (a) exist s, where-
as iCIU) = e' sin t, then the left-hand member exis[s bu[ the right-hand rnember does
no L
15. By applying [he proper!y of Equalion (16) to [he result of Problern 13, obtain the
+r
formula
oC{j~ I~u) dU } = j(v) dv.
Also, assuming tha[ [he res ult of Pro b lem 14 also applies, obtain [he formula
{f,
~ I (ui
oC -11- dl/
}
= sI~'-
l/(,') d ,·.
[In eaeh case , the gi ven formula is valid , for s suffieiently large, whenever the left-hand
member exists. The integral í ~' I(t )/t dt need not exist.)
16. Assuming the resul[s of Problern 15 , show that
oClCi(t)} = J...
s
IOg .¡ .l
5 - +
and
where
oClEi( - f)] = + log s ! l'
S 1'()
l = l' o
- -lidl/,
sin
11
'( t) = -
CI f", -eos-Udu II '
Ei(-/ ) = -
J
'~
,
e-U
udll.
Section 2.4
17. Let N(s )/ D(s) denote the ratio of two polynomials, with no common factors, such
that the degree n of D(s) is greater than that of N(s), and suppose that D(s) has n dis-
tinct real zeros s = a" a2 , "" a •. Show that the coefficients in the partial-Iraction
expansion
N(s ) A, + A 2 + ... + A • = L• A m
D (s ) = s a, s - a2 S - Gil m -= I S am
are determined by the equations
. N(s) N(a m )
Am = !~~ (s - a m) D(s) = D ' (a m)
Henee show that in this case
18. If oC[I(/)J = j(s) and if

1(t )= A o +A,f+A 2 f'+ .. . + Anln + ...
(io sorne interval about t = O) and
j(s) = Bn
S
+ B, + B +
S2 S3
2
(for sufficiently large values of s), show that
A = Bno
n n!
19. By using the ratio test, show that if limn~~ I Bn+ 1/ Bn I = So the second series in
Problem 18 converges when s> Soo Deduce that in this case limn~~IAn+l/Anl =
O, so that the first series in Problem 18 then convergesfor all values of t.
20. Use the results of Problems 18 and 19 to find the inverse transform of each of the
following fuoctioos as a power series in t:
1 1 l/s
(a) sio s' (b) ";S2 + 1 -";1 + (l/s)2°
21. (a) Show that if

d
ds F(s) = Z(g(t)}
theo
Z-I[F(s)} = _g(t).
t
[Use Equation (19) with n = 1.]
(b) Use the result of part (a) to deduce that
sin t
Z-I(cot-1S} = - - ,
t
[Compare Problem 13(b).]
22. Use the result of Problem 21(a) to deduce the following formulas:
(a) Z-I{logS -
s - a
b}
= ea, - eh',
t
(b) Z-I{tanh-l~J = sinhat.
aJ t
23. Verify Equation (22) in each of the following cases:

(a) f(t) = cos at, (b) f(t) = sinh at,
. s . 2s + 1
(e) fes) = S2 _ a2' (d) fes) = S2 + 2s + 2'
24. By starting with Equation (l5a) and eonsidering the limiting form as s ----> O, obtaio
the relation
lim sj(s) = feO)
.s""" o
+ lim f= e-"r(t) dt
s-o o
aod, by formally taking the limit on the right under the integral sign, obtain the result
lim sj(s) = feoo)
,~o
where f( 00) = lim,_= f(t). [Compare Equation (22). The result is valid when the
integral f~ r(t) dt exists. In particular, lim,_.=f(t) must exist.]
ansforrn ProbJems 85
25. (a) Show that the result of Problem 24 is nol valid, io particular, in the cases for
wbich¡(S) is given by
1 1
s - l' S2 + l' S(S2 - 3s + 2)
(b) Show that the result of Problem 24 is valid, in particular, ID the cases for
wbich¡(s) is given by
eries ID -, 1 1
,/Anl = s s + l' ses' + 3s + 2)
[The result of Problem 24 occasiooally is stated without qualifyiog restrictioos.
h of the The aboye examples show that sorne care should be taken in applying il. In practical
situatioos, it can be used if the existence of lim,_~ f(t) is assured, from physical
considerations or otherwise.]
. Section 2.5
26. Determine the eonvolutioo of eaeh of the following pairs of functions:
(a) 1, sio at, (b) t, ea"
(e) ea" e ',
b (d) sin at, sin bt.
27. Verify Equation (20) in each of the cases considered in Problem 26.
28. If ¡(s) = J:,[f(t)}, express the inverse transforID of each of the fOllowing fuoctions
as an io tegral :
(a) ¡(s) , ¡(s)
s+a (b)s'+a 2 '
¡(s) (d) ¡(s)
(e) (s + a)2' (s + a)(s + b)
29. Suppose that y(t) satisfies the integral equation
y(t) = F(t) + f: G(t - u)y(u) du,
where Fet) and G(t) are known functions, with Laplace transforms F(s) aod G(s),
respectively.
(a) Show tbat then
_ F(s) - (;(s) F()
y(s) = 1 _ G(s) = F(s) + 1 _ G(s) s.
l,obtain
(b) Deduce that the solutioo of the integral equation is
y(t) = F(t) + f: H(t - u)F(u) du,

he result
where H(t) is the function whose traosform is given by
H(s) = G(s) .
vhen the 1 - G(s)
(e) Illustrate this result in the special case when G(t) = ro'.
86 The Laplace Trans(onn
30. (a) Show that r" jes) is the transform of the function
FU) = (n _1 I)! Jor' (1 - u)"- 1 f(u) du,
when n is a positive integer.

(b) Deduce that I
n times n times I
l' l' l'
..-'''-.. .---'-..
oo o fu) dI . o o dI = (1 - u)"-' f(u) du o

o
(n - 1 1)!
o o o ...
Section 2.6
31. (a) Show that
lb ó(u - I¡)j(a)du = j(t,) if a < 1, < b,
if the integral is defined as the limit (as E ----> O) of the integral in which ó{u - t,) is
replaeed by a funetion equal to 1/ (2E) when 1, - E < u < 1, + E and equal to zero
elsewhereo
(b) In a similar way, obtain the relation
J:ó '(U-t,)j(u)du=-f'(I,) if a</, <b.
(e) Show thal the eonvolulion of ó(t - 1,) and j(/) is given by
«1
J"Ó(U-I,)jCt-U)du=IO
o
(
\j 1 - tI
) <tI))'
1> 1 , ,
when 1 and t, are positiveo
32. (a) If the Heaviside ¡mit s tep junetion H(I) is defined sueh that HCI) = 1 when
1 > O and H(t) = O when 1 < O, show that
1 e-stl
,c(H(I)} = 'S' ,c(H(t - tI)} = -s-
when 1, - O.
(b) Noticing that ,c(Ó' (I)} = s ,c(Ó(/)} = S2 ,c[H(I)}, indieate a sense in whieh, eor-
respondingly, we may be led to think of ó '(I) as a formal derivative of ó(t), and of
ó(t) as a formal derivative of H(t).
Section 2.7
33. Find the inverse Laplaee transform of eaeh of the following funetions:
Ca) S2 _ js + 2' (b) S2 _ ~s + 5' (e) :. : \,
(d) ses 2s + I ( ) 1 e-'

+ 1)(s + 2)' e S2(S2 + 1)' (f) s + 1.
34. Find the inverse Laplaee transform of eaeh of the following funetions:
S2 S2 3s + 1
Ca) (S2 + a2)2' (b) (S2 _ a2)2' (e) (s + l)(s + 2)(s + 3)'
Problems 87
(d) s' 1 a J'

(e)
1 -
s
e-.r
,
(f)
1
s~2-;--a::C2")(7:s~2-+-'----bC'2;-;)
7:
(
35. Find (he inverse Laplaee transform of eaeh of the following funetions:
s ._- 1 S2 e-'O
(a) S2 + 2s + 2' (b) S4 + 4a4 (e) S2 + l'
s' I I
(d) (s + a)J (e) (s' + 1)" (f) (S2 _ I)J
36. Determine (he inverse transform of eaeh of (he following funetions by making
appropriate use of the expansion
I
I-a I + a + a' + (lal < 1)
and o[ the property (18):

b 1 1 -e - as .' 2
(a) s(1 e 0')' (b) s I +
e 0'/ "
W w( I + e-"'-'w)
(e) (1 _ e .,/w)(s' + W')' ( d) ( 1 _ e U'w)(s' + ( 2 )
37. If F(s) = g(s)/(l - e-O' ), where g(l) = O when I >- a, show (ha( .,c-l(F(s)} is a
periodic lunelion, of period a, whieh agrees with g(l) when O < I < a.
38. Show (ha! (he (ransform s listed in par(s (b), (e), and (d) o[ Problem 36 can be
written in the form required by Problem 37, by taking g(s) = (1 - e-o,/' )'/s in part ( b)
and g(s) = (1 + e-.,·w)/(s2 + w') in parts (e) and (d), with a = 2n/w in part (e) and
a = n/w in part (d). Also verify the as sertion of Problem 37 in those cases.
39. Show that (he inverse of the transforrn
1 e- Ps/8 + e- 3Ps, 8
F(s) = 1 s +e p",
is of period P and sketch i ts graph.
Section 2.8
40. Solve the following problerns by the use of Laplaee transforms:
dy
(a) dI + ky = O, y(O) = l.
dy
(b) dI + ky = 1, y(0) = O.
dy ~
(e) dI + ky = u (1 - 1), y(0) = 1.
(d) 'J: + ky = 1(1), y(O) = Yo'

41. Solve the following problerns by the use of Laplaee transforms:
(a)d2Y + 2dY+2y=0 y(O) = 1, dy(O)= _ 1.
di' dI' dr
(b ' d2y , 2 dy
) dr 2 -r dI
+.,-y -_
-,
2
y
(O) O
= ,
dy(O)
dI
= I
.
d 2Y dy s: dy(O)
(e) dI' + 2 dI + 2y = v(1 - 1), y(O) = 1, ----¡¡¡- = -1.
88 The Laplace Transfor rn
dy(O) ,
(d) ~;; + 2 ~: + 2y = f(l), y(O) = Yo, (j( = Yo·
42. Solve the following problems by the use of Laplace transforms:

2
d4Y+4 =0
(a) dl 4
y(0) = dy(O) = d y(0) = O d'y(O) = I
y, dI dl 2 'dI 3 .
( b) d4
dl4
y + 4y= 4
,
y(O) = dy(O)
dI
= d'y(O)
dl 2
= d' y(O) = O
dl 3 .
43. Use Laplace transforms to solve the problem

d 2y
dl 2 +Y= 0(1 - a), y(O) = O, y(b) = O,
where O < a < b, by first writing dy(O)jdl = e, and finally determining e so that the
condition y(b) = O is satisfied by the inverse transformo
44. In Figure 2.6 a mass m is eonneeted to an elastic spring with spring constant k,
and to a dashpot which resists motion of the mass with a force numerically equal to e
times the velocity of motion. The applied external force is indicated by f(l) and the
displacement of the mass from equilibrium position by X.
(a) Show that the differential equation of motion can be put into the form
2
d x
dl 2
+ 2(X dx
dI
+ ((Xl + {J')x = _1 f(l)
m
with the abbreviations
k e
(X=-,
m -- (O'
o, 2m
(b) Assuming that the mass starts from
rest at its equilibrium position and is acted
on by a uniform force fa, find the resulting
motion. Consider separately the cases when fJ
is real and positive, {J = O, and {J = iy, where
[(1) y is real. Diseuss the three cases and sketch
typical curves representing the displaeement
e as· a funetion of time.
(e) Assuming that the mass starts from
Figure 2.6 rest at the position x = a aod that 00 exter-
nal force aets, investigate the resulting motion
as in part (b).
45. (a) Use tbe method of Laplaee transforrns to obtain the solution of the sirnul-
taneous equations
dx dy
dI + dI + x = -e-',
dx: dy + ix +
+ 2 dI 2 = O
dI y,
whieh satisfies the initial eonditions

x(O) = -1, y(O) = +1.
(b) Solve the same problem by one of the methods of Chapter 1.
problem s 89
46. Establish the relation

. dOy) dk "
oC { I k dl"J = (-I)k dSk[S"y - S"-ly(O)- ... _ y'n-I'(O)].
(Notice that by use of this relation a linear differential equation in y with polynomial
coefficients can be transformed into a new linear differential equation in the transform
y, which may in certain cases be more tractable than the original equation. In particular,
if the coefficients are linear functions of 1, the transformed equation is of firsl order.)
47. Use the formula obtained in Problem 46 to show that, if y(l) satisfies the differential
equation
d2v dy
(al + b) dl í + (el + d) de + (el + f)y = h(I),
then its Laplace transform y(s) satisfies the equation
(as 2 + es + e) ls - [bs
d" 2 + (d - 2a)s + (f - e)]y
= [(a - d) - bs]y(O) - by'(O) - hes).
48. Use the result of Problem 47 to transform the equation
Id2Y+2dY_C02IY=0
dl 2 de
to the equation
dy = y(O)
ds co 2 - S2
and hence deduce that

Y(I) = y(O) sinh COI .
COI
[See Problem 22(b). Notice that y'(O) cannot be assigned here and, indeed, that from the
differential equation there follows y'(O) = O if Y and y" are finile al t = O. Any solu-
tion other than the one obtained is not finite at I = O, and does not possess a Laplace
transform.]
49. Use the result of Problem 47 to transform the equation
d 2y dy..L _
I de 2 + dt ' Iy - O
to the equation
(S2 + I)
d"
ls + sy = O,
regardless of prescribed initial conditions. Hence deduce that
"(s) = y(O)
y ";s2+1
[see Equation (22)] and, by expanding y(s) in powers of l/s [Problem 20(b)], obtain the
solution
(t 2/2)2 2
Y(I) = y(O) [ I - (I !)2 + (1(2!)2
/2)4
- ....
]
[The series in brackets represents the Bessel function Jo(t). Notice (T38). The comments
on finiteness following Problem 48 also apply here.]
90 The Laplace Transforrn
Section 2.9
50. Use Table 3 (when necessary) to evaluate the following to two decimal plaees:
(a) f: e- x ,,/;;: dx. (b) r(2.7). (e) r( ~ 1.3), !
(d) (1.6)!, (e)
51. By making the indicated substitutions, transform the integral
r(n) =
(~1.3').
So e-'I"-I dI
(f) nD.
I
lo the fOllowing equivalent [orms:
(a) r(n) = l' (IOg "~ r-' dx
(b) re,,) = 21~ e-"r '"-1 dr (/ = rO).
52. (a) Obtain from StirJing's formula (63) the relation

loglo n! ~ (n + 1) loglo n ~ 0.4342945n + 0.39909 (n ~ col
[or use in numerical computation.
(b) Show that use of this formula gives the approximations
lO' "'= 3.599 )< 10 6 , 100!=9.325 X 10 '57 .
(The true values, lO four figures, are 3.629 x 10 6 and 9.333 :< 10 157 . It can be shown,
more generalJy, Ihat Stirling's [ormula [or 11! is aeeurale lO wilhin 1 percenl when
11 > 10 and wilhin 0.1 percent when 11 > 100.)
53. Use Ihe Slirling formula to show thal

a"n' ~ .../2nn
I (ae)"
-;:¡ (n ~ col,
when a is any constant, and deduce that

.
Ilm
n_'Y>
,,= O.
a"
n.
54. Investigate the following limits:

1
(a) ¡irn n. ,
"_.~ (IX + l)(IX + 2)(IX -1- 3) ... (IX + n)
" (2n) I
(b) 1
"~~ 2 2 "n!(n ~ j)!'
55. (a) Starting wirh the relation
r(x + n + 1) = (x + I)(x + 2) '"' (x + N),

r(x + 1)
for an)! positive inleger n, obtain the result
d
dx lag rcx + n + 1) = r'(x + n + 1)
r(x + n + 1)
= r'(x + 1) + _1_ + 1 + ,
, 1
_ _ _ o
rex + 1) 1 + x 2 + x n+x
Problems 91
( b) If we write 'P (z) = p ez + I)/ r (z + 1) , and set x = O in the result of part

(a), show that there follows
'P(n)
.
= 'P(O) +I+ ~
2
+ _13 + .. . + Ln
[The funetion 'P(z) is often ealJed the Digamma funclion . It take s o n the vaJue
'P (O) = -y, where y = 0.5772157 ... is a number known as El/ter's eons lt: nl .]
56. The Belafunclion. The Beta funetion of p and q is defined by tbe integra l
B(p,q) = Sal IP-l(1 - I)q-I di (p , q > O).
By writing I = sin' e, obtain the equivalent form
B(p, q) = 2 J: 11
sin 2p - I e COS 2 q- 1 e de (p, q > O).
57. Make use of the results of Problems 51(b) and 56 to verify the following develop-
ment:
Jn/1
Jo e-"r2p+1q-1 dr
~
B(p,q)r(p + q) = 4 o sin 1p - 1 ecos 1q - 1 ede
(p, q > O).

Henee sh o w that
B (p, q) = r(p)r(q) (p, q > O).
r(p q) +
58. By w nttng I = x/ex + a) in the definition of the Beta funetion , and using the
res ult of Problem 57, obtain the result
r~ x .- I dx = a -q B ( p q ) = a- q ;;r,.-'-(p~
) r:"-(ó.-'q",) (p, q, a> O).
Jo (x + a )p+ q , r(p q) +
59. U se the result s of preeeding problems to verify the following evaluations :
(a )
'JoIV i _ 1.-
1
I dI =
( T1) = v n- r (r(p
B p,
)
p + .i) . (p > O),
( b) J " ' 2 sin' e de = J".'2 cos' e de = vi r ( n t 1) (n > -1 ),

o o r (ni 2)
2
(e)
I dI
= _1 JI $( 1/ . ) -1 ds
=v r (! )
_ 1i ---,,....,...:.:.:...::...,....,..
Jo.v 1 - l' n o V1- $ n r(! + ;)
(n > O),
()J
d
o
~ x ' dx
( 1+ )2=
x
r (l -
r
e) ( 1+ e)=.
nc
sin ll C
(- 1 < c < 1),
(e) J~ d x 1 J ~ s(1 /· )- I ds n/n (n > 1),

o 1 + x' = n o 1 + $ = sin (n / n )
.. 1
" _ n /2
( f)
Jo tan e de - cos ( /m / 2) (O < n < 1).
92 The Laplace Transfonn
60. Verify the relationship
B(p,p) = 2]'ll
o
[t(l _ t)]p-, dt =,.,¡1i r(p) ,
2 2p - ' r(p + ·D
by making use of the substitution t =!(1 - cos (J).
61. By comparing the result of Problem 60 wi!h Ihe expression for B ( p , p) which fol-
lows from !he resull of Problem 57, deduce the duplication formula for the Gamma
function:
Also show tha! this result can be wrilten in !he form

2 2n
(2n)! = :;:7iin!(n - !)!.
62. (a) By writing x = l/Jt in the definition
B(m, n) = Sa' x m -' (1 - x)n-' dx,

show that
B(m, n)t m+ n-' = S; u m- '(t - U)"-1 duo
(b) By noticing that the right-hand member is the convolution of t m - 1 and t n -"
deduce that
and so obtain a simple derivation of the result of Problem 57.

3
NUDlerical Methods for Solving
Ordinary Differential Equations
3.1. Introduction. In this chapter there are presented certain rnethods of

numerically calculating particular solutions of ordinary differential equations
which cannot be readily solved analytically. The rnethods given are, in general,
step-by-step methods and are described initially for first-order equations; how-
ever, the extension of these rnethods to the solution of higher-order equations
is also indicated.
Before considering these procedures, we outline a graphical rnethod of
solving first-order differential equations which is of sorne practical interest.
If such a differential equation is solved for the derivative of tbe unknown
function, the result consists of one or more relations of the forrn
dy
dx = F(x, y), (1)
where it may be assurned that the function F(x, y) is a single-valued function

of x and y. Such an equation states that at any point (x, y) for which F(x, y) is
defined, the slope of any integral curve passing through that point is given by
F(x, y). If we plot the farnily of so-called isoc/ine curves, defined by the equation
F(x, y) = e (2)
for a series of values of the constant e, it then follows that al! integral curves
of (1) intersect a particular curve of the family (2) with the sarne slope angle rp,
where tan rp is given by the value of e specifying the isoc\ine. Thus, if on each
isoc\ine a series of short parallel segments having the required slope is drawn,
an infinite nurnber of integral curves can be sketched by starting in each case
93
94 Numerical Methods for Solving Ordinary Differential Equations
y at a given point on one isocline and sketch-

iog a curve passing through that point with
C=l ~ 1
-;;
I
8
I
10
the indicated slope and crossing the suc-
cessive isoclines with the slopes associated
with them. This method can always be
used to determine graphicalIy the particular
solution of (J) which passes through a
prescribed point (x o, Yo), when the function
F(x, y ) is single-valued and continuous. The
procedure is illustrated in Figure 3.1.
Figure 3.1 3.2. Use of Taylor Series. Suppose that

the solution of (1) which passes through
the poiDt (x o, Yo) is required. Knowiog the value of y at x = x o, we atternpt,
by a step-by-step method, to calculate successively approximate values of y at
the points
XI = X o + h, X2 = Xo + 2h, ... , Xk = Xo + kh,
where h is a suitably chosen spacing along the X axis. For this purpose we now
suppose that the value of y has been determined at x = X k , and denote this
value by Yk; that is, we writeYk = y(x k ). We then make useofthe Taylor series
representation in the form
y(x + h) = y(x) + y'(x) ~! + y"(x) ~~ +

ancl, setting X = Xk' we obtain the formula
,h + Yk"h2! +
2
Yk+1 = Yk + Yk"T! (3)
for calculating the value of y k+ l ' Primes are used to denote differentiation with
respect to x. Since y' is given by (J) in terrns of x and y, the coefficients io (3)
can be determined from (1) by successive differentiation. Thus we obtain
y' = F(x, y) y~ = F(Xk' Yk),
" _ JF + JF dy " _ JF(Xk' Yk) + aF(Xk, Yk ) ,
y - Jx ay dx' Yk - ax ay Yk,
and so forth, for the higher derivatives. Since Yo is given , Equation (3) can be
used to determine first y 1> with k = O, then Y2 with k = 1, and so on, the
number of terrns being retained in (3) at each step depending upon the spacing
and upon the accuracy desired. It is evident that a single series may, if preferred,
be used fo! several calculations in sorne cases, by assigning successively increased
values to h.
Example 1 . To illustrate this procedure, we consider the solution of the differentiaI

equation
3.2. Use of Taylor Series 95
with the initial condition that y = 2 at x = O. We choose a n interva l h = 0.1, and

hence calculate successively the approximate values al' y at x = 0.1,0.2,0.3, and so on .
By successive differentiation we obtain
y' = y - x, y" = y ' - 1, y 'f ' =y",
Hence, at x = O, we have Yo = 2 and

y~ = 2, y~' 1, y~' I, . .. ,
and, with k = O, Equation (3) gives
h' h3
y¡ = Yo + 2h + 2 +"6-'
~ 2 + 0.2000 + 0.0050 + 0.0002 + ... = 2.2052.
Next, at x = 0.1, we have y, ::::: 2.2052 and
y'¡ ::;:: 2.1052, y',' ~ 1.1052, y'¡" ::::: 1. 1052, ... ,
and, with k = 1, Equation (3) gives
Y2 ~ y¡ + 2.1052h + + 0.1842h' + .. .
0.5526h 2
::::: 2.2052 + 0.2105 + 0.0055 + 0.0002 + ... = 2.4214.
The procedure may be repeated as often as is required. In this example the exact
solution is readily found to be
y = e''C + x + 1,
and the aboye results are found to be accurate to the four decimal places retained .
Thís procedure is readily generalized to the solution of initial-value prob-

•
leros involving differential equations of higher order, as may be seen from
Example 2, below. For a seco nd-order equation ir is found ro be necessary to
calculate y~+ ¡ as well as Yk+¡ before proceeding ro the calculation of Yk+2' For
this purpose we may differentiate y(x + h) with respect to x to obtain
y'(x + h) = y'(x) + y"(x) ~! + y"'(x) ;~ +

Setting x = x" . there then follows
,
y. + ¡=Y.
I
h" jhj + h ,,, h2! + 2
... (4)
This res ult can also be obtained by differentiating Eguaríon (3) wirh respecL Lo
11, as may be seen directly from the fact thar the deri va tive s of y(x h) with +
respect to x and h are ídentical.
Example 2. Consider the nonlinear differential equation
dd2~ _ dd
y
+ xy2 = O
X x
with the initial conditions that y = 1 and y' = - 1 when x = O. We calcula te the
96 Numerical l\<Jethods for SoIYing Ordinary DifferentiaJ Equations
successive derivatives
y'" - yl/ - 2xyy' - y2,
yi' = y'" _ 2xy" - 2xyy" - 4yy',
Hence, at x = O, we have Yo = 1, y~ = -1, and
Yo' = y~ = -1, y~' = Ya' - Y5 = -2,
Then, with k = O, Equation (3) gives
h2 h'
y, = 1 - h - 2 - 3"
and, taking h = 0 .1.
y, ::::::: 1 - 0.1 - 0.005 - 0.0003 + ... = 0.8947.
Now, in order to calcula te Y2 it will be necessary next to calculate y';, y';', and so on.
However, the calculation of these values involves knowledge of the value y', in addition
to the value of y" which is now known. The value of y', can be calculated by using the
series
h'
y', = -1 - h - h 2 + 3" +
which is obtained by differentiating the series defining y, = y(x o + h) with respect
to h. Hence we obtain
y', ::::::: -1 - O. I - 0.01 + 0.0003 + ... = -1.1097.
The vaJues of y';, y';', and so on, can now be calculated from the forms given, and the
calculation of Y2 and y~ proceeds in the same way. •
A further generalizatíon to the solution of two simultaneous equations of

the form
dx
dI = F(x, y, t
) and dy
dI = O(x, y, I ) ,
with prescríbed initial values of x and y, is readily devised.
3.3. Tbe Adams Metbod. Suppose again that the solution of the problcm
dy (S)
-d = F(x, y),
x
has been determined up to the point x k = X o kh. If now we assume that +
over the interval (x k , X k + h) the derivative dy/dx changes so slowly that it can
be approximated by its value y~ at the point x k , then over that interval the
approximate increase in y is given by hy~, and we obtain
Yk+' ~ Yk + hy~ - h + hFk , (6)
where Fk is written for F(xk> Yk)' This formula obviously would give exact
results if y were a linear function of x over the interval considered.
A more nearly exact formula is obtained, in general, if we assume that the
deriva ti ve dy/dx is nearly linear over the interval (x k - h, X k + h), and hence
that the graph of y can be approximated by a parabola over this interval. We
3.3. The Adams Method 97
lhus assume the approximation

F(x, y) ~ a . b(x ~ x k)
and determine lhe constants a and b in such a way that the approximation
takes on the calculated values of F(x, y) at the points X k ~ h and x k . In this way
we obtain
Hence, integrating both sides of the differenlial equation (5) over lhe interval
(x k , x k + h), we find
Yk+1 ~ Yk ~ S:"'"[Fk + (Fk ~ Fk_J

X
h XkJdx
~hFk+ ~(Fk~Fk_l)
or Yk+1 ~ .h + hFk + ~ (Fk ~ Fk - I )· (7)
The last term of this expression is seen to be a correction to the expression

given by (6) and may be appreciable if the derivative dy/dx varies appreciably
over the interval considered. It should be noticed that the first ordinate calcula-
ble by (7) is Y2, when k = 1, and that in this case the values of YI and FI are
needed in addition to the known initial value Fa. The value of YI must be
determined by another method, for example, by the use of Taylor series; the
value of FI is then given as F(x 1> y 1)' With the notation
I'1 Fk = Fk+ 1 ~ F k,
Equation (7) can be written in the form
Yk+1 ~ Yk +
h(Fk + ±I'1Fk _ I )· (8)
Still more accurate formulas can be obtained, in general, if the derivative
of the unknown function y is approximated by a polynomial of higher degree
n, taking on calculated values at n + l consecutive points. By an extension
of the preceding method (see Section 3.7), a formula is obtained when F(x, y)
is approximated by a polynomial of fourth degree, in the form
Yk+1 ~ Yk + h(Fk + ±I'1Fk _ 1 + n1'1 2Fk_2 + t1'1 3F k _ 3 + HfiI'1 4 F k _ 4 ) (9)
with the notations
dFr = Fr+l - Fr,
1'1 2F, = I'1F,+ 1 ~ I'1F"
(10)
1'13F, = 1'12F'+1 ~ 1'12F"
1'14F, = 1'13F'+1 - 1'13F,.
These notalions define the first, second, third, and fourth forward differences
of the calculated values of F, which, as will be seen, are readily evaluated if the
calculations are suitably tabulated.
98 Numerical Methods ror Solving Ordinary Differential Equations
Formula (9) , w hich invo!ves fourth difference s, wou!d g ive exact res ult s if,
over the interval (x. - 4h , X k -+- h), the unknown fun c tion y were a polynomial
offillh degree in x. Fo r mula (6) is obtained from (9) by neg!ec tin g al! differeneces ,
and (7) o r (8) is obtained by retaining on!y nrst difference s. Formulas of inter-
mediate accuracy can be obtained from (9) by retaining only terms through the
second or third differences.
It is seen that , if, for example, it is decided that second differences are to
be retained, the calcu!ation of Yk+ 1 makes u se of the values of F k , F k _ " and
F k -" and hence of the vaJues y" Yk-" and Yk-¿' Thu s the fírst ordinate calcula-
ble in this case would be y" Since only Yo is presc ribed at the start, the va lue s
of Y, and y, mu st n.rst be determined by another method, s uc h as that of Taylor
se rie s, or the Runge-Kutta method of the followin g sec tion when initi a ! series
developments are not feasib!e. Similar1y, if third or fourth differences are
retained, three or four additiona! ordinates, re sp ec tively , must be fírst ca lculated
by another method before the present procedure c a n be appl ied.
Example 1 (contillued). We apply this method to ¡he continuation of the solution

of the problem considered in Example I (Section 3.2), retaining second differences. The
work may be tabulated as follows:
x y F=y- x áF
o 2 2 0.1052 0.0110
0.1 2.2052 2.1052 0.1162 0.0122
0.2 2.4214 2.2214 0.1284
0 .3 2 .6498 2.3 4 98
0.4 2.8917
Before the Adams [ormula is applied to calculate YJ, the two initial ordinates y I and
Y2 are required in additioo to the prescribed ordinate Yo = 2. These ordinates are
taken [rom the results of Example 1 and are entered into the second cotumn o[ the
tabte as shown, The corresponding values of F are then entered in the third columo.
Each entry in the [ourth column is obtained by subtracting the corresponding entry
in the F column [rom the succeeding entry in that column. AIgebraic signs musl, o/
course, be relained. The last coJurnn contains similar differences between successive
entries in the t1F column. In thi s way the entries aboye the division line in each column
of the table are obtained. Now , to calcula te the ordinate YJ, we make use of Formula
(9), neglecting the Ja st two lenns, and notice tha¡ the quantities needed are exactly
those which appear irnmediately aboye the di vision Jine in each coturnn, ¡hat is, the
last numbers entered in Ihe coJumns at this stage of the complltation. We IhllS oblain
y,~ 2.4214 + 0.1[2.2214 + 1CO.1162) + lzCO.OllO)) = 2.6498.
Jt 15 seeo Ihat the differences needed al a given step of the calculalioo recede along
3.3. The Adams Method 99
success ive columns in the tableo The value 01' Y 3 is now entered in the second columIl ,
the corres pond ing value of F is calculated, and additional differences are determined.
The nex t o rdinale is then calculated as before,
Y. - 2.6498 + 0 .1[2.3498 + 1(0.1284) + 1SZ(0 .0122)] = 2.8917,
and the process is continued in the same wa y. A rough check on the accuracy obtained
at each s tep can be obtai ned by estimating (he contribution of the neglected third dif-
ference. Thus retention of third differences in the calculation of Y . would , in this case,
increase the value obtained by ~(O.l)(O.OO 12) = 0.00005, and hence the fourth places
in [he res ults are in doubt. The correct value is e O. 3 + 1.3 = 2.8918. •
To apply the Adam s method to a second-order equation of the form

2
d y _ ( ,dY ) (11 )
dX2 - F x, y, dx '
we may first introduce tbe Dotation P = dy jdx aDd hence replace Equation (11)
by the simultaneou s equations
dy
dx =p , (12a)
dp ( 12b)
dx = F( x , y , p).
The initíal conditions at x = xo ,

y = Y o, (13a)
dy = p =Po, (l3b)
dx
are to be satisfied . Then , applying (9) separately to (l2a) and ( 12b), w e obtain
the two formulas
Yk+l:::: Y k + h(Pk + ±.ó.P._1 + ... ), (I4a)
Pk+1 - Pk + h(Pk + -1.ó.Fk - 1 + ... ), (I4b)
and proceed step by step as before.
Exanrple 2 (continued). We appJy this method to the continuation of the solution of

the problem considered in ExampJe 2 (Section 3.2), retaining (for simplicity) onJy firs/
differences. The work may be t a bulated as foIlows:
x y P tlp F = p - xy 2 tlF
O 1 -1 -0.110 -1 -0.190
0.1 0.895 -1.110 -1.190
0 .2 0.779 -1.23 8
The values of Y, and P" taken from ExampJe 2, as well as the prescribed values of
Yo and Po, are entered firs t. Next the corresponding values of F and the differences
6.Po and 6.Fo are calculated and entered. Equations (14a, b) ¡hen give
Y2 - 0.895 + o. I( -1.110 - 0.055) = 0.779,
P2:::::: -1.110 + 0.1(-1.190 - 0.095) = -1.238.
At this stage a second difference 6. 2 po = -0.018 can be calculated. Since its contri bu-
tion to the calculation of Y3 would be -0.00075, it may be presumed that ¡he result for
yz is also in doubt by abou! one unit in the third decimal place. •
The generalization of these methods to the solution of more general initial-

value problems involving two simultaneous differential equations offers no
difficulty.
3.4. The Modified Adams Melhod. A useful modification of the Adams

method in the case of a first-order equation consists of using an appropriate
truncation of the Adams formula
Yk+1 - Yk + h(Fk + ±6.Fk _ 1 + n6. z F k _ + i6.
2
J
Fk _ J + Hb6. 4 F_ 4 ) (15)
only as a "predictor," to provide a first approximation to the value of y k+ "
and of using a corresponding truncation of the formula
Yk+1 - Yk + h(Fk + 1 - ±Ó.Fk - fr6. 2Fk_1 - i46. J F k _ Z - lrrÓ.
7l 4
Fk _ J) (16)
as a "corrector." The derivation of the latter formula is indicated in Section 3.7.
The approximation yielded by (16) usually is better than that afforded by
(15), when both are terminated with differences of like order, since the coef-
ficients of the higher differences are smaller in (16). However, this advantage
is partly offset by the fact that, since the right-hand member of (16) involves
F k + 1 - F(Xk+I,Yk+I)' this equation expresses Yk+1 (approximately) in terms of
itself. Thus (16) generally cannot be solved analytically for Yk+1 except in
special cases, such as those in which F is a linear function of y.
Fortunately, resort gene rally can be had to a simple iterative procedure
for the solution of (16) when the spacing h is sufficiently smal!. For this purpose,
(15) is first used to determine a "predicted" value of y k+" in correspondence
with which the approximate value of Fk+ I is calculated, together with the ap-
proximate new differences Ó.Fk = Fk+ I - F k , 6. 2 F k _ " and so on. Then (16) is
used to determine a "corrected" value of y k+ ,. If this value differs significantly
from the predicted value, the entries Fk+ " Ó.Fk , and so on, then may be cor-
respondingly corrected and (16) used again, to provide a "recorrected" value
of y k+ l. Generally, the need for this recorrection is avoided either by relaining
a sufficiently large number of differences or by taking the spacing h lo be
sufficiently small.t
tWhen h is suffieiently small to avoid lhe need for large eorreetion, it frequently happens, in
faet, that the result of thefirst correetion alfords a better approximation to the true value than
that whieh would be alforded by subsequent reeorrections.
3.4. The Modífied Adams Method 101
Example 1 (continued). In the preceding tabulation of Example 1, the entry 2.8917

here would be interpreted as the predicted value of Y •. When corresponding approxima-
tions to F., !:!.F" and !:!. 2 F 2 are calculated, the tabulation appears as follows:
x Y F I'1F 1'12F
O 2 2 0.1052 0.0110
0.1 2.2052 2.1052 0.1162 0.0122
0.2 2.4214 2.2214 0.1284 0.0135
0.3 2.6498 2.3498 0.1419
0.4 2.8917 2.4917
From (16), truncated al so to second differences, a corrected value is obtained,

Y.~ 2.6498 + 0.1[2.4917 - -1(0.1419) - i-z(0.0135)] = 2.8918,
in correspondence with which the tabulated values of F., !:!.F3 , and !:!. 2 F 2 each are cor-
rected by one unit in the last place retained. Clearly, no additional corrections will
result from a repetition of the process, so tha! an advance to x = 0.5 appears to be
in order. •
When both formulas are truncated with a difference of order r (r = 2 in

Example 1), the error in the finally corrected va1ue can be estimated by the
formula
p e
E ~Yk+' -Yk~' (17)
k+' = 2 + 4r '
where y~+, is the "predicted" value O btained from (15) and Yl+ 1 is the "corrected"
value obtained from (16). (See Reference 3.) This estimate applies only to the
error introduced, by using (16) to approximate the differential equation, in
progressing from X k to X k + 1 • lt ignores the accumulated effects of errors of the
same type introduced at earlier stages, as well as the effects of "roundoff" errors;
also its validity essentially presumes that the spacing h is sufficiently small to
obviate the need for "recorrection."
Thus, for example, when only second differences are retained, the error in
the corrected value can be estimated as (yL 1 - Yl+,)1 10, provided that this
estimate is small, that the same is true of the corresponding error estimates
at al! preceding stages, that recorrection was not needed, and that roundoff
errors have been controlled by the retention of an appropriate number of sig-
nificant figures in all calculations.
lt happens that the rate of convergence of the iterative process, in the
determination of Yk+ 1 from (16), depends upon the magnitude of the quantity
_ phaF(x k, Yk) (18)
Pk - ay'
I where P IS the algebraic sum of the numerical coefficients of the differences
1
I,
102 Numerical Methods ror Solving Ordinary Dilferential Equatiol
retained In (16). (See Reference 3.) Thus, when second differeoces are retainec
P= 1 - -1- - n n· =
Uoless I Pk 1« 1, the process either wil! not converge or will converge so slowl:
¡hat many recorrections will be needed, in general. In the case of the Example
the "convergen ce factor" P3 is seen to be
P3 = nh 0.04. =
Before starting a step-by-step calculation based on (16), it is desirable to
verify that the spacing h is such that the magnitude of the initial convergence
factor Po is small relative to unity.
The application of the modified Adams method to the second-order equa-
tion (11), or to a bigher-order equation, is perfectly straightforward. In the
case of Equation (11), y" = F(x, y, y'), the approximate "convergence factor"
is found to be
(19)
3.5. The Runge-Kutta Method. The method associated with the names of
Runge and Kutta is a step-by-step process in which an approximation to y k+ I
is obtained from y k in such a way that the power series expansion of the approxi-
mation would coincide, up to terms of a certain order h N in the spacing h =
x k + I - x k , with the actual Taylor series development of y(x k + h) in powers
of h. However, no prelimioary dilferentiation is needed, and the method also
has the advantage that no ioitial values are needed beyond the prescribed values.
Such a method is particularIy useful if certain coefficients in the differentia!
equation are empirica! functions for which analytical expressions are not known,
and hence for which initial series developrnents are not feasible. In place of using
values of N derivatives of y at one point, it uses values only ofthefirst derivative
at N suitably chosen points.
The derivatíon of the basic formulas may be iUustrated by considering the
special case when second-order accuracy in h is required. Again starting with
dy
dx = F(x, y) (20)
and the prescribed iDitial condition y = Yo when x = x o, the Taylor series

expansion of Yk+1 = y(x k + h) up to second powers of h is obtained, by intro-
ducing the two succeeding relations into (3), in the form
Yk+1 = Yk
...L
I
F h
k
(aFk
ax
...L
I
aFkF ) h
ay k 2
2
+ ... (21)
with the notations
We assume an approximation of the form

h+1 ~ J'k + AlhFk + A2 hF(x k + J.i1h, Yk + J.i2 hFk) (22)
il-
r 3.5. The Runge-Kutta Method ]03
I
1
and attempt to determine the constants A" A2' J1.p and J1., in such a way that
t the expansion of the right-hand side of (22) in powers of h agrees with the
! expression glven by (21) thraugh ferms aI secand arder in h. From the Taylor
I series expanslon
+ + + Ka¡~'(, + ... ,
I,
•
¡(x -i- H, Y K) = ¡(x, y) Hale;' y)
x
for small values of H and K, where the following terms are of second order
y
y)
•
in H and K, we find
F(x k + J1. , h, Yk + J1.2 hFk)
= F(x k , Yk) + J1.,haFC~k'x Yk) + J1.2haF(~k'y Yk) Fk +
= Fk + h (J1.' a:: + J1.2 a~k Fk) + ... ,

where the omitted terms are of at least second order in h. Hence (22) becomes
Yk+l = Yk + (A, + A2)Fk h + A 2 (J1.,a:: + J1.2a~kFk)h2 + .... (23)
The terms retained in Equations (21) and (23) are brought into agreement
if we take, in particu lar,
A, = A2 = ·h J1., = J1.2 = 1.
Thus Equation (22) becomes, in this case,
Yk+, ~ Yk + ~ [F(Xk' Yk) + F(x. + h, Yk + hF.)]

or, writing
K, = hF(x k , Yk)' K 2 = hF(Xk + h, Yk + K,), (24)
this result can be put in the form
Yk+' ~ Yk + ±(Kl + K 2). (25)
It should be noticed thar Equation (23) can be brought into agreement
with (21) in infinitely rnany ways, by taking
1
J1.l = J1.2 = 2(1 - A,)
where A, *
l but is otherwise arbitrary. Thus infinite1y many other forros of
(22) could be obtained, in addition to the rather syrnmetrical form chosen here.
By methods analogous to that just given, similar formulas giving higher-
order accuracy in h may be obtained. We give without derivation two such
procedures:
Third-order accuracy:
Yk+' ~ Yk + t(a, + 4a 2 + a,), (26)
a, = hF(x., Yk)'
a 2 = hF(x. + .}h, Yk + ta,), (27)
a, = hF(x. + h, Yk + 2a z - al)'
104 Numerical Methods ror Solving Ordinary Dilferential Equations
Fourrh-order acclIracy:
Yk+' - Yk + '¡'(b, + 2b, + 2b, + b.), (28)
b, = hF(x k , Yk)'
b, = hF(x k + th, y. + tb,),
(29)
b, = hF(Xk + t h , Yk + tb 2)'
b. = hF(x k -+- h, Yk + b,) .
The close relationship between Equations (26) and (27) and [he formula of
"Simpson's rule" (see Problem 27) may be noticed.
Let the error associated with using a procedure of Nth-order accuracy k
times with spacing h be expressed in the form Kkh N +'; that is, suppose that
the correct result is given by adding Kkh N +' to the calculated result. Then it
can be shown that in ordinary cases the quantity K is not strongly dependent
on k, h, and N. Thus, if the spacing were doubled, the error associated with
the corresponding new calculation would be approximately
Since the result of subtracting the ordinate determined by the second method
from that determined by the first would then be (2 N - 1) Kkh N +', it follows
that this difference is approximately (2 N - 1) times the error in the first (more
nearly exact) calculated value. In this way we are led to the following error
estimate:
lf a procedure of Nth-order accuracy gives an ordinate y") wilh spacing h
and an ordinale y'21 wilh spacing 2h, the error in y'!1 is given approximately by
(y'I) - y(21)/(2 N - 1).
Thus the difference between the two results is divided by 3 if (24) and (25)
are used, by 7 if (26) and (27) are used, and by 15 if (28) and (29) are used, to
obtain an error estimate.
Example l. We apply (26) and (27) to integrate the equation considered in Example 1
(Section 3.2). The work may be arranged as follows:
x O 0.1 0.2
y 2 2.20517 2.42139
a, 0.2 0.21052
x +!h 0 .05 0.15
Y + ta, 2.\ 2.31043
a2 0.205 0.2\604
x+h 0.\ 0.2
JI + 2a 1 - al 2.21 2.42673
a, 0.2\\ 0.22267
Ha, + 4a2 + a,) 0.205\7 0.21622
3.6. Picard's ~..Iethod 105
Jf ¡he spacing is doubled, the resull y, ~ 2.42133 is obtained directly (with h = 0.2).
Application of the error estimate gives the approximate error (0.000006)/7 = 0.00001,
which indica tes that the first value calculated for Y2 may be one unit too small In the
last place retained. The exact value is Y2 = 2.42140 to five decimal places. •
To integrate a second-order differential equation of the form

d'y ( dy ) (30)
dX2 = F x, y, dx
with the initial conditions y = Yo and y ' = y~ when x = x o, with third-order

accl/racy in the spacing, we first calculate successively the values
k, = hy~,
k, = h(y~ + ~k',),
(31)
k~ = hF(x. + th, Y. + ~k¡, y~ + ~k ',),
k, = h(y~ + 2k', - k ',),
k~ = hF(x. -;- h, Y. -"- 2k, - k" y~ + 2k', - k',).
The values Yk+' and y~+, are then given by
Yk+, - Y. + !;(k, + 4k, + k,) (32)
and y~+, - y~ + i(k', + 4k~ + k',). (33)
A corresponding formula giviogfourth-order aCCl/racy can be written down by
analogy from (28) and (29) .
Example 2. Applying this method to calcula te y, in the case of the problem of Exam-
pie 2 (Section 3.2), we obtain successively
k, - -O.I , k,=-0.105, k,=-0.11190,
k', - -0. 1, k', = -0.1 0951, k', = -0.11982.
There then follows
y, :::::: J - i(O.1 + 0.42 + O.J 1 J90) = 0.89468,
y ', :::::: -1 - -1;(0 .1 + 0.43804 + 0.11982) = - 1.1096.
•
3.6. Picard's Metbod. In contrast with the step-by-step methods so far
considered, in which successive ordinates are calculated point by point , the
method of Picard is an itera ti ve method that gives successive functions which,
in favorable cases, tend as a whole (at Jeast over a certain interval) toward the
exact solution. Although the method is of limited practical usefulness, it illus-
trates a type of procedure which is of frequent use in other applications, and IS
in itself of theoretical importance.
Considering first an initial-value problem of first order,
dy
dx = F ( x, y), (34)
whcre y = Yo when x = xü, we formally integrate both sides of Equation (34)

over the interval (x o, x) to obtain an equivalent relation
y = Yo + IXx, F(x, y) dx. (35)
Now, to start the procedure, we take as an initial approximation to the function

y (to be determined) a suitablefimcliol1 of x, say y'('(x). If the general nature of
the required solution of (34) is known, this initial function may be chosen on
this basis. It is preferable that it satisfy the initial condition, although this is not
necessary. In the absence of further information, the initial approximation
function y' ('(x) may be taken as the constant yo. With this assumed approxima-
tion for y, as a function' of x, tbe function F(x, y' (') becomes a known function
of x, and a second approximation to the function y, say y'O'(x), is given by
y'2'(X) = Yo + IX F[x, y'('(x)] dx. (36)'

x.
A third approxirnation is obtained by replacing y by y(2'(X) in F(x, y) and using

Equation (35) to giye
y'])(x) = Yo + IXX.
F[x, y'"(x)] dx. (37)
In this way successiye approximations are obtained asfunctions ofx, accord-

ing to the formula
y'n+ ()(x) = Yo + IX F[x, y'n,(x)] dx. (38)
x.
If F(x, y) is sufficiently regular near the point (x o, Yo), the successive approxi-
mations y' J), y'2', ... , y'n; will tend toward a !irniting function y(x) oyer sorne
interva! in x about x = x o, and that function will satisfy the differentia! equation
(34) as well as the prescri bed initia! condition.
Example. Applying this method to lhe solution of Ihe equation of Example 1 (Sec-
lion 3.2), we wrile
y'n+ J) = 2 + fax (y'n) - x) dx.
Taking y' (' = 2, we oblain successively
y(2) = 2 + IX (2 - x) dx = 2 + 2x _ ~2,
y'" = 2 + r
•o
x
(2 + x _ ~2) dx = 2 + 2x + ~2 _ ~] ,
y(4J = r
2 + Jo
X
(
2 + x +
X2
2 -
x 3)
ti dx = 2 + 2x
X2
+ "2 x3
+ ti
and so forlh. In this case il is readily shown by induction Ihal
(n+ I I _ X2 x] I xn Xn+ 1
y - 2 + 2x + ?i + 3! T .. . - ¡- ti! - (n + 1)!'
3.7. Extrapolation with Differences 107
and as n beco mes large we are led lo a consideration of the infinite series
y = 1+ x + (1 + í"! + ~~ + ... + ~~ + ... ),

which may be recognjzed as the expansion of the known exact solution
y = 1 +x+ eX,
the expansion converging for al! values of x.
In practice it is usualIy no! feasible to obtain an explicit expression for the

•
!lth approximation and to proceed to the exact solution by a limiting process,
as in the example given. Furthermore, the successive approximations need not
be polynomials, as in this case. Still, if a particular approximation is used to
calcula te y, the accuracy at a given point can be estimated roughly by consider-
ing the deviation between this approximation and the preceding approximatioD
in the neighborhood of the point. UnfortuDately, it is true that in many cases
the integrals defining successive approximations cannot be evaluated analytical-
Iy, but in turn must be approximated by numerical methods.
Similar procedures can be devised for dealing with equations of higher
order, the convergence of the successive approximatioDs to the true solution
depending upon the regularity of the coefficients.
3.7. Extrapolation with DifIerences. For purposes of simplicity, the details

of the derivation of Equations (9) and (16), which specify the Adams method
and its modification for the numerical integration of differential equations,
were omitted in Sections 3.3 and 3.4. In this section we indicate this derivation
and also point out the usefulness in other connections of certain intermediate
and related results.
We again denote the value of a function I(x) at one of n + 1 equally
spaced points,
X k - 2 = X k - 2h, ... , (39)
by the abbreviation 1, = I(x,). In Section 3.3 the lorward differences were
defined by the equations
(40)
and so fonh. In sorne developments it is more convenient to use the so-called
backward difference notation, according to which we write instead
(41)
and so forth. These two notations clearJy are related by the equations
... , (42)
Thus, for example, we have
/3 - 12 = 1'1/2 = "113' 13 - 2/2 + 1, = 1'1 2/, = "1 2/3'
In particular, Equations (9) and (16) take the forms

Yk+1 ~ Yk -'- h(Fk - fV'F k + -(L:V'Fk + iVJFk + ttil-V"Fk ) (43)
ánd
Yk+1 ~ Yk -, h(Fk + 1 --}VFk +1 - nV2Fk+1
- .,}-¡VJFk+1 - ".YhV 4Fk+ 1)' (44)
with the notation (41) of backward differences, the subscripts on the right in
these equations no longer varying from difference to difference.
In order to determine a polynomial approximation of degree n to a function
f(x), having the property that agreement is exact at the n + l points defined
in (39), it is convenient for present purposes to write the approximation in the
form
f(x) ~ao + al(x - x k) + a 2 (x - xk)(x - Xk _ l )
+ aJ(x - xk)(x - xk_¡)(x - x k- Z) +
+ a.(x - Xk)(X - X k _ l ) ... (x - X k - n + I ). (45)
In this form the coefficients are readily determined by forming successive
backward differences at x = X k , as follows. First, setting x = X k , and replacing
the approximation by an equality, we have
a o = fk' (46a)
Next, if we calculate the difference Vf(x) = f(x) - f(x - h) and use (39),
there follows
Vf(x) ~ ha l + 2ha z(x - x k) + 3ha J(x - xk)(x - Xk - I) + .. .
+ nha.(x - xk)(x - X k _ 1) ... (x - x k - n + Z ).
Thus, setting x = X k and requiring equality, we obtain
l
al = ¡¡Vfk' (46b)
After calculating the second difference,

VZf(x) ~ 2·1h 2 a z -+- 3·2h 2 a J(x - x k ) +
+ n(n - l)h 2 a.(x - xk)(x - Xk_ 1) ... (x - X k - n + 3 ),
we determine a 2 ,
_ l n2,
a z - 2!h 2 v Jk'
From the way in which these results were obtained, it can be seen (induc-
(46c)
¡
tively) that the general result will be of the form
a'=r/h,V'fk (k=O,I,2, ... ,n). (47)
The resultant approximation formula (45) can be put into a convenient1y

compact form if we write
x = x k + sh, s-- x - h x k , (48)
3.7. E"trapo lation with Differences 109
so that S is distance measured from the point X k , in units of the spacing h. With
this notation, tbe introduction of (47) into (45) then gives
f( Xk -r'sh)::::::/,
- k
-L
'
sV'
Jk
+ ses 2!+ I)vu
Jk
+ ses -1- I)(s -1- 2)...,,,
3! V Jk
ifuse is made of (39). Tbis result is known as Newton's (or Gregory's) backward
difference formula for polynornial approximation.
Thus, if the values of a smooth function f(x) are known at n + 1 equally
spaced points, we may suppose that the function is approximated by the nth-
degree polynomial which agrees with f(x) at these points and , accordingly,
use (49) to determine approxirnate values of f(x) at additional nearby points .
Ibis formula is particular1y usefuI for extrapolation (prediction) beyond the
point x k , at the end of a range of tabulation, although it can also be used for
interpolation, with negative values of s.
To iUustrate the use of this formula in extrapolation, we consider the
function f(x) = Jx. Assuming known three- place values for x = 2, 3, 4, 5,
and 6, we form the following difference table:
x f Vf V2f V'f
I
I 2.0 1.414
0.318
I 3.0 1.732 -0.050
I 4.0 2.000
0.268
0.236
-0.032
0.018
0.009
5.0 2.236 -0.023
0 .213
6.0 2.449
To extrapolaté for ,./6.2, we set s = 0.2, Slllce here h = 1, and retain third
differences. Equation (49) then gives
f(6.2) - 2.449 + (0 .2)(0.213) + (0.2~ 1.2\ -0.023) + (0.2)( 162)(2.2) (0.009)

= 2.449 + 0.0426 - 0.0028 + 0.0008
= 2.490.
Similarly, to interpolate for -J5.8 we take s = -0.2 and, again retaining third
differences, obtain the value 2.408 . Both the values so obtained are correct to
the sarne number of places as the given data . In general, however, extrapolation
is less dependable than interpolation .
110 Numerical Methods for Solving Ordinary Differential Equations ,•f
Equation (49) is also useful for integrating a tabular function over an
interval near the end point X k . Thus we have, for example,
J~xk+h f(x) dx = h
f' f(x. -:- sh) ds
."( " o
... ). (50)
When f(x) is a polynomial, the sum on the right terminates and yields an
exaet result. More generally, when f(x) is nol a polynomial, the result of retain-
ing differences through the nth is identical with the result of replaeing f(x) by
the polynomial of degree n whieh agrees withfat the n + I points x k ' x k - l ' . . . ,
X k - n and integrating that polynomial over the interval (x k , X k + h).
In illustration, if we use (50) to approximate f~ .-J:X dx, with rhird differ-

enees, we obtain
'7
J 6 .-J:X dx ~ 1[2.449 + 1(0.213) + -tz( - 0.023) + i(0.009)] = 2.549.
This is in agreement with the true three-place value.

SimilarIy, from the relation
X' fO_
f x,-h f(x) dx = h I f(x k sh) ds,
we obtain the formula
r:_hf(X) dx ~ hUk - 1Vfk - TI'[V2fk - rtVJfk - ·l/uV4fk + ... ). (51)

I Equation (50), with f replaeed by F, leads to the approximate relation (43)
l or, equivalently, (9) when dy = F dx, and henee
X'+ 1
Yk+I-Yk=
f.<, Fdx .
Further, Equation (51) leads similarIy to (44) or (16) when k is replaeed by

k + l.
Finally, we may use (49) to obtain an approximate derivative formula.
Sineer(x) = h- 1 (dfjds), we obtain the relation
f'(x k + sh) ~ h
(Vfk + 2s t 2
1 V2fk + 3s + s + 2 VJfk + ... ). í (52)
Thus, to approximate (d .-JxjdX)x_6 by using the tabulated data, we take s = O,

and henee, again using tbird differences, obtain
1'(6) ~+[0.213 + 1( -0.023) + j-(0.009)]
= 0.213 - 0.0115 + 0.003 = 0.2045.
The true result is 0.204, to three places. Altbough satisfactory results were
obtained here, it should be noted that, in general, approximate differentiation
•
3.7. Extrapolation wirh Differences 111
is inaccurate. lls accuracy may be very seriously impaired by small inaccuracies

in the tabulated data. Still, there are occasions when there is no alternative
but to use an approximate method of this sort.
The formulas and methods of this section illustrate the treatment of tabulat-
ed data near the end of the range over which a tabulation exists. These data
are usually the most troublesome to deal with. This sec tion may be considered
al so as illustrating the more general use of finite differences in approximate
analysis.
Analogous formulas for interpolation and other operations near interior
tabular points may be found in texts treating numerical methods. (See also
Problems 23 to 26.) In such formulas the so-called central difference notation is
particularly useful. In this notation one writes
j, - j, - . = oj,_ ' .'0 ' oj, • •¡ 2 - Oj, - 1..2 = 02j" ... , (53)
so that the subscript of a difference is the mean of the subscripts of its parents.
Thus we have the general notational correspondence
= Oj'."2 = Vj, •• ,
jd. - j , = !!..f,
j'.2 - 2j, •. + j, = !!..2j, = 02j,+. = V2j'+2'
and so forth. The "forward" difference !!..j is sometimes caIled a "descending"
difference, whereas the "backward" difference Vj sometimes called an "ascend-
ing" difference. Although this practice leads to a certain amount of confusion,
its motivation may be realized by an inspection of the following table of differ-
ence notations.
)
jo
)
!!.. 'jo = O'j3/2 = V'j,
y
!!..j2 = oj'l2 = Vj,
j,
!!"/3 = 0/'/2 = V/4
It is apparent that the differences !!..j" !!..2j" !!.. 3j" and so fortb , "descend"
.) to the rigbt, whereas the differences Vj" V2j" V3j" and so forth, "ascend" to
the rigb t. The central differences 02j" 04j" and so forth, remain in tbe same
" horizontal line as the entry j" whereas the differences Oj,. ' / 2' 02j" 03j,+ 1/ 2' 04j"
and so forth, form a " forward zigzag" set of differences in increasing order,
remaining as c10se as possible to a horizontalline.
More detailed treatments ofthe errors associated with formulas for approx-
imate interpolation, numerical integration and differentiation, and the numerical
e solution of differential equations may be found in many sources, inc1uding the
n reference s listed at the end of trus chapter.
REFERENCES
1. COLLATZ, L., The Numerical Treatment 01 Differential Equations, 3rd ed., Springer_
Verlag New York, Ine., New York, 1960.
2. HENRICI, P., Discrete Variable Methods in Ordinary Differential Equations, John
Wiley & Sons, Ine., New York, 1962.
3. HILDEBRAND, F. B., Introduction to Numerical Analysis, 2nd ed., MeGraw-HiII Book
Company, Ine., New York, 1974.
4. MILNE, W. E., Numerical Solution 01 Differential Equations, 2nd ed., Peter Smith
Publisher, Gloucester, Mass., 1970.
5. STEFFENSEN, J. F., Interpolation, 2nd ed., Chelsea Publishing Company, Ine., New
York,1950.
PROBLEMS
Section 3.1
1. Use the method of isoclines to sketch the integral curves of the equation
dy
--y=x
dx
in the first quadrant.
2. Proeeed as in Problem 1 with the equation
~ _ y2 = X2.
3. Use the method of isocIines to obtain a sketch, in the first quad.rant, of the integral
curve of the equation
dy _ x2y = x
dx
whieh passes through the point (O, 1).
Section 3.2
4. Use the method of Taylor series to determine to four places the values of the solu-
tion of the problem
dy _ x2y = x y(0) = 1
dx '
at the poiots x = 0.1,0.2, and 0.3. (The true values at x = 0.1, 0.2, 0.3, 0.4, and 0.5
round to 1.00533, 1.02270, 1.05428, 1.10260, and 1.17072.)
5. Use the method of Taylor series to determine to four plaees the values of the sol u-
tion of the problem
dy
dx
+ xy2 = O
'
y(O) = 1
problems 113
at the points x = 0.1, 0.2, and 0.3, (The true values at x = 0.1, 0.2, 0.3, 0.4, and 0.5
ro und to 0.99502, 0.98039, 0.95694, 0.92593, and 0.88889.)
6. Use the method of Taylor series to determine to four places the values of the sol u-
tion of the problem
d 2y dy
dX2 - X2 dx - 2xy = 1, y(O) = 1, y'(O) = O
at the points x = 0.1, 0.2, and 0.3. (The true values are the same as those in Problem
4.)
Section 3.3
7. Assuming the values of y in Problem 4 for x = O, 0.1, 0.2, and 0.3, use the Adams
method to calcula te the values for x = 0.4 and 0.5, using third differenees.
8. Assuming the values of y in Problem 5 for x = O, 0.1, 0.2, and 0.3, use the Adams
method to calcula te the values for x = 0.4 and 0.5, using third differenees.
9. Assuming the results of Problem 6 for x = O, 0.1, 0.2, and 0.3, use the Adams
method to ealculate the values of y for x = 0.4 and 0.5, using third differenees.
Section 3.4
10. Use the modified Adams method, retaining only seeond differenees, to effeet the
determinations of (a) Problem 7, (b) Problem 8, and (e) Problem 9. At eaeh stage, use
Equation (17) to estimate the error introduced.
11. If no differences beyond the second are retained, show that the formulas (15) and
(16), of the modified Adams method, can be written in the explicit forms
~ h
Yk+1 = Yk + 12(23Fk - 16Fk _ 1 + 5Fk _2)
h
and Yk+1 :::::: Yk + l 2(5Fk+ 1 + 8Fk - F k _ I ),
respeetively. (These formulas are easily used, since differeneing is not involved. How-
ever, the use of differences is usually preferable in questionable cases, sinee then gross
errors may be indicated by irregularities in the eolumns of differenees, and also warn-
ings may be served by trends observable in these eolumns.)
1-
12. Making use of the faet tbat, for small values of the spaeing h, there follows
!J.'f= V'f= J'f= h,d'f,

dx'
show tbat the errors introdueed into the formulas of Problem 11 by negleeting the
.5 first omitted differenee are respectively approximated by
8
4
3h (d 4y )
dx 4 k'
_ h24 (ddx
4 4y
4
)
k
.
}-
[It is known that the true truneation error in eaeh of these cases is given by the result
of replacing (d 4 yfdx 4 h by the value of d 4 yfdx 4 at sorne point between Xk_2 and Xk+l
in tbe expression so obtained.]
114 Numerical lVIethods for Solving Ordinary Differential Equatio ns
Section 3.5
13. Use the Runge-Kutta method, with third-order aeeuraey, to determine the ap_
proximate values of y at x = 0.1 and 0.2 if y satisfies the eonditions of (a) Problem 4 ,
(b) Problem 5, and (e) Problem 6.
14. (a-e) Reealculate the value of y(O.2) determined in the eorresponding part of
Problem 13 by taking only one step, with spaeing 2h = 0.2, and estimate the error in
the two-step ealculation by use of the proeedure suggested in the text.
15. A funetion y(x) satisfies the equation
d 2y
dX2 + yrp(x) = O
and the initial eonditions y(O) = 1 and y'(O) = O. The following approximate values of
the funetion rp(x) are known:
x o 0.05 0.10 0.15 0.20
~(x) 1.000 1.032 1.115 1.249 1.434
Use the Runge-Kutta method, with third-order aeeuraey, to determine approximate

values of y at x = 0.1 and 0.2.
16. Reealculate y(0.2) in Problem 15 by taking one step, with spaeing 2h, and estimate
the error in the two-step ea1culation.
Section 3.6
17. Apply Pieard's method to the solution of the problem

dy _ x2y = x y(O) = 1,
dx '
taking y(l)(x) = 1 and making two sueeessive substitutions, and compare the ap-
proximations with the series expansion,
y = 1 + ·iX2 + .jX3 + r\,x 5 + T1rX6 +
of the exaet solution.
18. Apply Pieard's method to the solution of the problem
y(O) = 1,
taking y' II (x) = 1 and making two successive substitutions, and compare tbe ap-
proximations witb the series expansion,
y = 1 + x + ix 2 + 4X3 + Hx + 4
of the exaet solution.

Probl ems 115
SectiOD 3.7
19. Suppose Ihat the following rounded values of a eertain funetion J(x) are known:
x 1.0 2.0 3.0 4.0 5.0
J(x) 1.2840 1.3499 1.4191 1.4918 1.5683
By making use of formulas deríved in Seetion 3.7, obtain approximate values of the
following quantities as aeeurately as possible:
(a) J( 4.8), (b) J(5.2), (e) 1'(5.0), (d)' J, J(x) dx.

'6
20. Derive from Equation (49) ¡he formula

(3 +2)
J.,
."+'h [ .
J(x) dx ~ th Jk + ~ VJk + t 12 t V'Jk +
.
21. Use the result of Problem 20 te obtain, from the data of Problem 19, approxímate
vaJues of the following quantities:
(a) f" J(x)dx,

5,0
(b) f'o ',8
J(x) dx.
22. Establish the following notational relations:

(a) 11J, = OJ,+ 1 / 2,
(e) 11 V11J, o'J,. = 1/2.
23. Determine the eoeffieients speeifying the polynomial approximation of degree

n = 2k of the form
J(x) ~ao + al(x - xo) + a2(x - xo)(x - XI)
+ a,(x - xoXx - x,Xx - x_,)
+ a.(x - xo)(x - x¡)(x - x_,)(x - X2) +
+ anCx - xoXx - x,)(x - X_I) . .. (x - Xk)(X - X_k).
so that the two members are equal at the n + 1 equally spaeed points
X-k = Xo - kh, ... , . .. , Xk = Xo + kh.
by ealculating the differenee 11J(x) = J(x + h) - J(x), then the differenees V[11J(x)],
11[V11J(x)l. and so on, and equating the two members at X = Xo after eaeh sueh dif-
fereneing. Thus show that ao = Jo, that
11J(x)~ h[al + 2a,(x - xo) + 3a,(x - xo)(x - X_I)
+ 4a.(x - xo)(x - x_¡)(x - Xl) + .. ·l.
and henee that al = 11Jo/h = OJ, / ,/h, and so forth. Make use of the results of Problem
lló Numerical Methods for Solving Ordinary Differential EquatiollS
22 to express the result in the form

2
- {'
I (x)=.to+ (x-xoTT7l+
) b/' !2 (X-Xo(X-x'2!h2
)' ) 15 /0
J
, (X-Xo )( x - x , Xx
T - x_, )b3!h
/I/2
J
b 4J¡
+ (x - xo)(x - x,)(x - x _ ,)(x - x2)4!h~ +
24. By writing x = Xo + sh, express the result of Problem 23 in the form

1) 5(5 2 - 1)
I(xo + sh) - lo + sb/' !2 + 5(5 2!- 15 2/ 0 + . 3! 15 3/ 1/ 2
+ 5(5
2
- !~(s - 2) 15 4/0 + '"
(This is the interpolation formula of Gauss. The central differences involved remain as
near as possible to the horizontalline through the tabular value/o , comprising a "for-
ward zigzag. ")
25. By integrating the result of Problem 24 over appropriate intervals, obtain the
integral formulas
X O+ h
(a) Jx,- h I(x) dx ~ 2h(lo + 1/52/0 - rhb4/0 + ... ),
2"
Jx, - 2h I(x) dx -
.%0+
(b) 4h(lo + 1152/0 + M'/o + ... ),

26. Make appropriate use of the Gauss interpolation formula of Problem 24 to obtaio,
from the data g:¡ven 10 Problem 19, approximate values of each of the following
quantities:
(a) 1(2.8), (b) 1(3.2), (e) 1'(3.0),
(d) 1'(2.8), (e) s: I(x) dx, (f) f I(x) dx.
27. By retaining only second differenees in the result of Problem 25(a), deduce the
formula
x,+h h
I
x,-h I(x) dx - 3(1-, + 4/0 + 1,)·
[This is the celebrated formula of Simpson's rule. Notiee that, sinee the eoefficient of
the third differenee in the more general formula is zero, the formula is exact when
I(x) is a cubic polynomial.)
28. Show (as in Problem 12) that the error associated with Simpsoó's rule is ap-
proximately
h' (d 4/ )
-90 dx 4 x,'
[The Irue error is known to be expressible in this form as well, but with d 4 y/dx4
evaluated instead at sorne unknown point between Xo - h and Xo + h.]
29. Show that, if Simpson's rule, of Problem 27, is applied suecessively over !he
adjaeent double intervals (a, a + 2h), (a + 2h, a + 4h), ... , (b - 2h, b) in the ap-
lí - -
Pro b1ems 117
proximate evaluation of the integral f:/(x) dx, there follows

h
f
b
a I(x) dx ~ "TUo + 4/, + 21z + 4/3 + ... + 21n-2 + 4/.-, + In),
where h = (b - a)/n and n is an even integer. [This formula, often known as the
parabolic rule, is probably the most widely used formula for numerical integration.
11 is exacl when I(x) is a polynomial of degree not greater than three. As is suggested
by n/2 applications of the result of Problem 28, for a given interval (a, b) the error in
the approximation is nearly proportional to h 4 and henee to l/n 4 • From this faet it
follows that, if two ealculations are made, one (In) with n subdivisions and one (IZn)
with 2n subdivisions, the error in I Zn may be estimated by (I2. - I n)/15.]
30. Evaluate the following integrals approximately by use of the parabolie rule (prob-
lem 29) first with n = 2 and then with n = 4, estimate the error in 14 in eaeh case, and
compare the results with the given rounded true values:
(a) f 1 +dxX2 = 0.7854, (b) f si: x dx = 1.8518,

nn
(e)
I
o ~1 - 1 sin 2 x dx = 1.3506.
4
Series Solutions of Differential
Equations: Special Functions
4.1. Properties of Power Series. A large class of ordinary differential

equations possesses solutions expressible, over a certain interval, in terms of
power series and related series. Before investigating methods of obtaining such
solutions, we review without proof certain useful properties of power series.
An express ion of the form
~
Ao + A¡(x - xo) + ... + A.(x - x o)· + - L.:o A.(x

n=
- x o)· (1)
is called a power series and is defined as the limit
N
lim L.:
N-o<> n=O
A.(x - xo)·
for those values of x for which the limit exists. For such values of x the series
is said to converge. In this chapter we suppose that the variable x and the coef·
ficients are real; complex power series are dealt with in Chapter 10.
To determine for what values of x the series (1) converges, we rnay make
use of the ratio test , which states that, if the absolute value of the ratio of the
(n + I)th term to the nth term in any infinite series approaches a limit p as
n - . 00, then the series converges when p < 1 and diverges when p > 1. The
test fails if p = 1. A more delicate test states that, if the absolute value of the
same ratio is bounded by sorne number a as 11 - 00, then the series converges
when a < 1. In the case of the power series (1) we obtain
p = ~i_,::: I A"t I1 x - x o1 = L 1x - Xo 1,
where L = lim I A.+ ¡ 1, (2)

11 _0<> An
118
4.1. Propertíes of Power Series 119
if the last limit exists. In this case it follows that (1) converges when
Ix - xol < yI
and dil'erges when
1
Ix-xol>y'
Thus, when L exists and is finite, an interval of convergence
(xo - l, :c o+ l)
is determined symmetricalJy about the point x o, such that inside the interval
the series converges and outside the interval it diverges. The distance R = I/L is
frequently calJed the radills of con vergen ce.
The behavior of the series at the end points of the interval is not determined
by the ratio test. Useful tests for investigating convergence of the two series
of constants corresponding to the end points x = X o + R are:
(1) If, at an end point, the successive terms of the series alternate in sign
for sufficiently large values of n, the series converges if after a certain stage the
.1 successive terms always decrease in magnitude and if the nth term approaches
f zero. and the series diverges if the nth term does not tend to zero.
h (2) If, at an end point, the successive terms of the series are of constant
.. sign, and if the ratio of the (n + l)th term to the 11th term can be written in
the form
) 1 _ ~ + en,
n n2
where k is independent of n and en is bounded as n ~ =, then the series con-
verges if k > 1 and diverges if k -< 1. It should be noticed that this test is
applicable even in the case when k =,1, so long as an expression of the indicated
s form can be obtained. We shaU refer to this test as Raabe's test.t
Example. In illustration, we consider the series

e = (1·2·3··· n)(x - a)n
e n~l (ct + I)(ct + 2)(ct + 3) ... (ct + n)'
.s where ct is not a negative integer. In this case we obtain
e
e L=~i~lct: ~ ~ 11 = 1.
's Hence the interval of convergence is given by Ix - al < 1 or
a-l<x<a+1.
tThis very useful test is also assoeiated wi!h the name of Gauss. The !erm en /n 2 ean in fae! be
replaced by en/n1+p, where it is required only tha! p > O and en be bounded, and still more
delicate modifieations exist. (See, for example, Referenee 10.)
120 Series Solutions oC Differential Equations: Special Functions
When x = a - 1, the signs of successive terms alternate when n > - a. Apart from
algebraic sign, the nth tenn is then
(a + I)(a +
n!
2) . ..
,,'na
(a + 11) = na + J)
+ n + 1)'
Reference to Equation (61) of Chapter 2 shows that this ratio is approximated by
r(a + I)n-- when n is large, and hence approaches zero as n increases only if a > O.
Thus the series converges at x = a - l if a > O, and diverges at x = a - l otherwise.
When x = a + 1, the terms are ofconstant sign when n > -a. The ratio ofconsecu-
tive terms is then
n + 1 a a(a + 1)
n + a + l = 1 - n + n(n + a + 1)
Hence, by Raabe's test with en = faCa + 1)nl/(n + a + 1), the senes converges at
x = a + 1 if a > l and diverges at x = a + 1 otherwise .
lt may be noticed tbal if L is zero the interval of convergence ineludes aU

•
values of x. However, if L is infinite, the series converges only at the point
x = X o' Whether or not the limil L exists, it is known tbat always, for tbe
power series (1), either the series converges only when x = x o, or the series
converges everywhere, or there exists a positive number R such tbat the series
converges when Ix - Xo I < R and diverges wben Ix - X o I > R.
It may happen tbat the series (1) contains only terms for which the sub-
script n is an integral multiple of an integer N> 1, and hence is of tbe form
~
Ao + AN(x - x o)H + A 2 ,v (x - X O)2N + ... = :E Ak~X

k- O
- xoyN, (3)
or is a product of a power of x - X o and such a series. Examples are afforded by

the series
Xl x4 _ k X 2k
+ + .. . - (Ixl <
OC>
cosx = 1 - 2! 4! {;0(-1) (2k)! 00 )
and
(Ixl < l),

for whicn N = 2 and 4, respectively. In such cases, tbe ratio An+l/An is undefined
for infinitely many values of n. Tbe Iimiting absolute value of the ratio of suc-
cessi ve terms is
where LN=limIA,k+l 1N I, (4)

k- _ AkN
if tbat limil exists, and tbe series converges when L N Ix - X o IN < 1, or

1 (5)
Ix - xol < VL N '
.1.1. Properties oC Power Series 121
'os
'm A particularly useful property of power series is the fact that convergent
power series can be treated, for many purposes, in the same way as polynomials.
Ioside its ioterval of coovergence, a power series represents a continuous
function of x with cootinuous derivatives of all orders. Inside this interval,
a power series can be integrated or differentiated term by term, as in the case
by
of a polynomial, and the resultant series will converge, in the sarne interval, to
O.
5e.
the integral or derivative of the functioo represented by the original series.
:u- Further, two power series in x - X o can be multiplied together" term by term
and the resultant series will converge to the product of the functions repre-
seoted by the original series, inside the cornrnon intervalo/ conl'ergence. A
similar statement applies to division of one series by aoother, provided that the
at denominator is not zero at Xo . Here the resultant series will converge to the ratio
• in sorne subinterval of the common interval of convergence .

By the statement on term-by-term multiplication of power series, we mean (tak-
all
~nl
ing Xo = ° for simplicity) that, if
~
he f(x) = L: a,x' ([xl < R,)

(= o
les ~
les and g(x) = L: bjx j ([x 1 < R 2 ),

1- 0
where R, and R 2 are positive, then
.b-
·m f(x)g(x) = (i: a,xi)(f b1x
1=0 J- O
l) = i:
ko:O
c.x',
at least for Ixl < min (R" R 2 ), where

c. = aob. + a,b._, + ... + a.b o
=
• a,b._ I .
L: (6)
i= o
With respect to division of power series, we mean that under the same assump-
tions, together with the requirement that b o 01= 0, we have
~
L: a,x i ~
(x = ':0
f)
= L: d.x',
g(x) L: bjXI .~o
1=0
in sorne interval 1x 1 < R, where the d's are to be determined by the relations
ed
lC- (i = 0, 1, 2, ... ). (7)
Since these conditions take the form

ao = boda,
a, = b,do + bod¡,
a2 = b 2d o + b ¡di + b od 2,
it follows that, with b o 01= 0, the first equation yields do, the secood yields di'
and so forth.
122 Series Solutions of Differential Equ3tions: Special Functions
Now suppose that the series 2::;;0 An(x - x o)" converges in a nonzero
interval about x = X o and hence represents a fUllction, say f(x), in that interval,
~
fex) = 2:: A,.(x

n= o
- xo)n. (8)
Then, differentiating both sides of Equation (8) k times and setting x = X o !n

the result, we obtain
f'k'(X O) = k! Ak (k = 0, 1, 2, ... ),
and hence (8) beco mes
~ f 'n'(x)
f(x) = I:
n""O n.
, o (x - xo)". (9)
This is tbe so-called Tay!or series expansion of fex) near x = xo.t It is clear
that not all functions possess such expansions, since, in particular, in order
that (9) be defíned, al! derivalíl'es of f(x) must exis t at x = X o . A function
which possesses such an expansion is said to be regular at x = X o. The aboye
derivation shows that, if a function is regular at x = x o, it has only one ex pan-
sion in powers of x - X o and that expansion is given by (9).
If f(x) and all its derivatives are continuous in an interval including x = x o,
then f(x) can be expressed as afinite Taylor series plus a remainder, in the forrn
f(x) = 2::o f'n'(x

N-I
n= n.
( )
o (x - x o)" + R¡v(x) . (10)
Here R N , the remainder after N terms, is given by

_ fU"(O N
Rr/..x) - N! (x - x o) , (11)
where e; is sorne point in the interval (x o , x). To show that the expansion (9)
is valid, so thar f(x) is regular at x o , we must show that RN(x) - + as N - + CX)
for values of x in an interval including x = xo. A test which is much more
°
easily applied and which is sufficient in the case of most functions occurring
in practice co n s ists of determining whether the formal series in Equation (9)
converges in an interval abour x = x o .
It is apparent, in particular, that any poly nomial in x is regular for al! x.
Further, any raliona! funelion (ralio of polynomials) IS regular for all values
of x which are not zeros of the denominator.
4.2. IlIustrative Examples. To illustrate the use of power series in obtaining

solutions of differential equations, we fírst consider tbe solution of three specific
linear equations of second order.
(1) To solve the differential equation
d 2y (12)
Ly - dx2 - y = 0,
tWhen xo = o, ¡he expansion is oflen called a Maclaurin series.

4.2. IIIustrative Examples 123
we assume a solution in the form

y = Ao + A,x + Azx z + AJx J + A.x· + A,x' + ...
and assume that the series converges in an interval including x = O. Differen-
tiating twice term by term, we then obtain
d 2y
d = 2A,
-x- ? + 6AJx + l2A.x2 + 20A,x' +
With the assumed form for y there follows
Ly == (2A 2 - Aa) + (6A) - A,)x + (12A " - A 2 )X 2
+ (20A s - A»)x' + ... = O.
In order that this equation be valid over an interval, it is necessary tbat the
coefficients of all powers of x vanish independently, giving the equations
2A,=A o, 6A J =A" l2A.=A" 20A,=A J, ... ,
froro which tbere follows
A. = -nAz = -r'¡¡-Aa ,
As = .Jo-A) = Tto- A "
The solution then becomes
y = Ao(l + ix' + rtx' + ... ) + A,(x + kxJ + Thx' + ... ).
It is seen that the coefficients A o and A, are undetermined, and hence arbitrary,
but that succeeding coefficients are determined in terms of tbem. The general
solution is thus of the form
y = Aou,(x) + A,u,(x),
where u,(x) and u 2 (x) are two linearIy independent solutioos, expressed as
power series, of which the first three terms ha ve been obtained. Tbe terros
found may be recognized as the first terros of the series representing tbe known
solutions cosb x and sinh x, respectively.
A more compact and convenient procedure uses the summation notation
in place of writing out a certain number of terms of tbe series. Thus we write
tbe assumed solution in tbe forro
and obtain, by differentiation,

d'y ~
:Ek(k - I)A,x' - z .
dX2 ,-o
Tbere tben foJJows
~ ~
Ly == k;O
:E k(k - l)A.x'-z - :E A,x'
k- O
= O.
lo order to collect the coefficieots of like powers of x, we next cbange the indices
124 Series Solutions of Differential Equntions: Specia.1 Functions
of summation in such a way that the exponents of x in the two summations are
equa!. For this purpose we may, for exarnple, replace k by k - 2 in the second
summation, so that it becomes
and hence
~
Ly _ I:
k=O
k(k - I)A.x'- 2 - I:
k~ 2
A'_2X'-2 = O.
Since the first two terms (k = 0, 1) of the first surnmation are zero, we may
replace the lower ¡imit by k = 2 and then combine the summations to obtain
~
Ly _ I: [k(k - l)A. - A k _ 2 ]X·- 2 = O.

k =2
Equating to zero the coefficients of all powers of x in volved In this sum, we

obtain the condition
k(k - ¡)A. = A' - 2 (k = 2,3, ... ).
This condition is known as the recurrenc e formu/a for A •. It expresses each coef-
ficient A. for which k >- 2 as a multiple of the second preceding coefficient
A k - 2 , and reduces to the previously determined conditions when k = 2,3,4,
and 5.
(2) As a second example we consider the equation
Ly = X2 ddX2y + (X2 + x) dydx -

2
y = O. (13)
Assurning a solution of the form
we obtain
and hence
Ly _
~
I: k(k - I)A.x' + I:
~
kA.x'+1 + I:
- kA.x' - I: Ak X'.
k- O k- O k- O k=O
The first , third, and fourth surnrnations may be combined to give

-
I:
k~ O
[k(k - 1) + k - IJAkx' =
k=O
~
I: (k 2 - l)A.x',
and hence there follows

~
Ly == I: (k 2 -
k=O
l)A. x' + kI:
- O
kA.x'+ l.
In order to combine these surns, we replace k by k - 1 in the second, to obtain
Ly
-
= kI:- O (k 2 -
l)A.x' + k=1
I: ~
(k - l)A._lx'.
4.2. IHustrative Examples 125
Since the ranges of summation differ, the term corresponding to k = must

be extracted from the first sum, after which the remainder of the first sum can
°
be combined with the second. In this way we find
Ly = -A o + -
I; [(k' - l)A k
k=l
+ (k - I)Ak_¡)X k •
In order that Ly may vanish identically, the constant term, as well as the coeffi-
cients of the successive powers of x, must vanish independently, giving the
condition
Aa = °
and the recurrence formula
(k - l)[(k + l)A k + A k -¡) = ° (k = 1, 2, 3, ... ).
The recurrence formula is identically satisfied when k = J. When k >- 2, it
becomes
I Ak _ 1
k+l
(k = 2, 3, 4, ... ).
I
t
Hence we obtain
f,
A2 = -~" A,= A2
"4 =
A,
3·4'
--= -
A,
5
A,
3·4·5
,
Thus in this case Ao = 0, A, is arbitrary, and all succeeding coefficients are

determined in terms of A,. t The solution becomes
1)
If this solution is put in ¡he form

_ 2A, (X 2 x' x· x'
y - x2! - TI + 4! - TI +
the series in parentheses in the final form is recognized as the expansion of e-x,
and, writing 2A, = c, the solution obtained may be put in the closed form
y = e
e-X - 1 +x.
x
In this case only one so/ution was obtained. This fact indica tes that any
linearly independent so/ution cannot be expanded in power series near x = O;
that is, it is not regular at x = O. A/though a second solution could be obtained
by the method of Section 1.10, an alternative procedure given in a foJlowing
section is somewhat more easily applied .
tClearly, the templation lo "cancel Ihe common factor" k - 1 before setting k = 1 in the
recurrence formula must be resisled.
126 Series SolutioDS of Differential Equations: Specíal Functíons
(3) As a final example we consider the equation

3 d'y
_
Ly = x -x-
d' +y = 0, ( 14)
Again assuming a so lution of the form
we obtain
= 2: k(k -
o, ~
Ly
k=O
l)Akx k +' + 2:
k- O
AkXk
Jf k is replaced by k - 1 in the first sum, there follows
- Ao + L: [A k + (k
k=l
- IXk - 2)A k _ .J Xk.
The condition Ly = ° then requires that

Ao = °
and that the succeeding coefficieots satisfy the recurrence formula
Ak = -(k - lXk - 2)A k _, (k = 1,2, . .. ).
For k = 1 and k = 2, the recurreoce formula gives A, = A, = O; and sioce,
from this point, the remaining conditions express each A as a multiple of the
precediog one, it follows tbat all the A's must be zero. Hence tbe only solution
obtained is the trivial one y = O. It thus follows that the equatíon possesses no
nOnlrivial solutions which are regular at x = O.
Next we proceed to a c1assification of types of linear differeotial equatioos
of second order, and to a study of the basic differences among the three pro blems
so far considered.
4.3. Singular Points oC Linear Second-Order Differential Equations. If a

homogeneous second-order linear ditferential equatioo is written in the standard
form
d 'y dy
-d
x-
' + a,(x)-d
x
+ a,(x)y = 0, (15)
the behavior of solutioos of tbe equation near a point x = X o is fOllOd to

depeod upon the behavior of the coefficients a,(x) and a,(x) near x = x o' The
point x = X o is said to be ao ordinary poinl of the differential equation if both
a,(x) and a,(x) are regular at x = x o, that is, if a, and a , can be expanded in
power series io an interval including x = x o. Otherwise, the poiot x = Xo is
said to be a singular point of the differential equation. In such a case, if tbe
products (x - x o)a,(x) and (x - x o)2a , (x) are both regular at x = x o , the
~.3. Singular Poinls of Linear Second-Order Differenlial Equ31ions 127
point X = X o is said to be a regular singular poinl ; otherwise, the point is

called an irregular singular poinl.
In illustration, we notice that for the fír st differential equation of the
preceding section, a I C-~) = °
and ao(x) = - l. Thu s aIl points are ordinary
points. In the second example the coefficients a I (x) = 1 + x- I and a 2 (x) =
_x- 2 cannot be expanded io powers of x but can be expaoded in power series
oear any other point x = x o. Thus the point x = °
is the only singular poin!.
Since the products xa l (x) = x + 1 and x a,(x) = -1 are regular at x = 0, it
2
follows that the point is a regular singular point. In the third case the point
x = O is readily seen to be an irregular singular poin!. Similarly, for the equation
d 2y dy
x 3(1 - X)2 dx' - 2X2(1 - x) dx + 3y = 0,
it can be verified that x = °

is an irregular singular point and x = 1 is a regular
singular poio!. AII other points are ordinary points.
When the coefficients a I (x) and a,(x) in the standard form (15) are ral¡os
01 polynomials, singular poiots can occur ooly wben a denominator is zero,
so that, unless the numerator also vanishes tbere, the correspoodiog coefficient
is oot finite. Most of tbe equatioos considered in this text will be of trus type.
However, a singular point also may occur when al (x) or a 2 (x) beco mes
iofioite in sorne otber way, or even io the absence of sucb behavior. For exarnple,
if a,(x) = (x - 1)5/3, it is seen tbat a'¡'(x) becomes infinite as x - > 1, so that
a ,(x) cannot be expanded in a series of powers of x - 1. Since the function
(x - l)a ,(x) = (x - 1)'· 3 also cannot be so expanded, it follows tbat the
differential equation (15) has, in ract, an irregular singular point at x = 1 when
al(x) = (x _ 1)5-' 3.
It w ill be shown that, if x = X o is ao ordinory point of (15), then the equation
possesses two linearly independent solutions which are regular at x = x o, and
heoce are both expressible in the forrn ¿:k'- o A.(x - xoY. If x = X o is a regular
.' singular point of (15), it wiIl be showo that the equation does not necessarily
possess any nootrivial solution which is regular near x = x o, but that at least
one solution exists of the forrn
~
y = (x - xo)' ¿:
k=O
A.(x - XO)k,
where s is a determinable number which may be real or imagioary. Such a

solution is regular at x = X o only if s is zero or a positive integer. If x = X o 1s
ao irregular singular point of (15), the problem is more involved; a nontrivial
solu ti on of tbis type may or may not exist.t
In illustration, the equation
x,d'y + (1 + 2x)dy = O
dx' dx
nt is known, however, that in this ease the differeotial equation io faet eannot have />vo in-
dependent solutions of this type.
has an irregular singular point at x = O. The general solution is

Y = e 1 e l:x ....L
I
e 2,.
thus the solution y = constant is the only solution regular at x = O. The equa-
tion
x, d y
dX2
2
+2 X 3 dy
dx
+y = O
also has an irregular singular point at x = O and has the general solution
Y = C I SID -
. 1 + C2 cos - l .
x x
It foIlows froro the nature of these functions that this equation has neither a
nontrivial regular solution at x = O nor a solution expressible in the more
general form
00
y = x' L; Akx k .
k=O
4.4. The Method of Frobenius. In this section we restrict attention to

solutions valid in the neighborhood of the point x = O. Solutions valid near
a more general point x = X o may be obtained in an analogous way, although
for this purpose it is frequently more convenient first to replace x - X o by a
new variable t and then to determine solutions of the transformed differential
equation near the point t = O.
In place of reducing a second-order linear equation to the standard form
(15), it is frequently more convenient to use a forrn which is, to sorne extent,
cleared of fractions, particularIy if a I (x) or a 2 (x) is the ratio of two polynornials.
For this reason, to investigate the nature of solutions valid near x = O, we sup-
pose that the equation has been put in the forrn
d 2y 1 1 ) _ O
Ly
_
= R(X)d 2
x
+ -P(x)d-
x
dy
x
+ xzQ(x y - , (16)
where R(x) does not vanish in sorne interval including x = O. We also suppose
that P(x), Q(x), and R(x) are regular at x = O. Then, with the notation of (15),
the products
_ P(x)
xa¡ ( x ) - R(x) and x2a 2 (x) = Q(x)
R(x)
are regular at x = O, and this point is either an ordinary point or, at worst, a
regular singular point of the differential equation.
It is convenient to suppose also that the original equation has been divided
through by a suitable constant so that R(O) = l. Then we may write
p(x) = Po + Plx + P 2X 2 + "',
Q(x) = Qo + Qlx + Q2X2 + "', (17)
R(x) = 1 +R¡x+R2x2+ " ' ,
the series converging in sorne interval including x = o.
4.4. The Method of Frobenius 129
We attempt to find nontrivial solutions which are in the form of a power

series in x multiplied by a power of x,
~
y = x' 2: AkXk
k=Q
= Aox' + A,x'+' + A,X'+ 2 + .. . ) (l8)
where s is to be determined. The number Ao is now, by assumption, the co-
efficient of thefirst term in the series, and hence must not vanish. Substitution
into the left-hand member of (16) then givest
Ly = (1 -i- R¡x + ... ) x [s(s - I)A ox ' -2
+ R,x'
+ (s + l)sA¡x' -¡ + (s + 2Xs + I)A 2x ' + ... J
+ (Po + P,x + P X + ... ) X [SA X'-2 + (s + I)A,x'-'
2 2 O
+ (s + 2)A x' + ... J 2
+ (Qo + Q,x + Q,X2 -1- ... ) X [.1o x '-Z + A,x'-' + Azx' + ... J,
or, after multiplying term by term and collecting the coefficients of successive
powers of x,
Ly =" [s(s - 1) + Pos + + ([(s + I)s + Po(s + 1) + QoJA,
QolAox'-'
+ [R,s(s - 1) + P,S + Q,lAo)x'-¡
+ ([(s + 2Xs + 1) + Po(s + 2) + QolA2
+ [R,(s + l)s + P,(s + 1) + Q,lA,
+ [R,s(s - 1) + P 2s + Q2lAoJx' + (19)
In order to abbreviate this relation, we next define the functions
f(s) = s(s - 1) + Pos + Qo = S2 + (Po - I)s + Qo (20)
and gn(s) = Rn(s - nXs - 11 - J) + Pn(s - 11) + Qn
= R.{s - n)Z + (Pn - RnXs - n) + Qn· (21)
With this notation, Equation (19) then becomes
Ly =. f(s)Aox'-Z + + g,(s + I)Ao]x'-'
[f(s + I)A,
+ [f(s + 2)A + g,(s + 2)A, + g2(S + 2)A o]x' +
2
+ [f(s + k)A k + gn(s + k)Ak_nJX'+k-Z + t, (22)
In order that (16) be satisfied in an interval including x = 0, this expression

must vanish identically, in the sense that the coefficients of all powers of x in
(22) must vanish independently. The vanishing of the coefficient of the lowest
power X'-2 gives the requirement
f(s) = ° or s, + (Po - l)s + Qo = O. (23)
This equation determines two values of s (which may however be equal) and
is called the indicial equa/ion. The two values of s, which specify the exponents
tA more compact development is obtained by using summation notation. (See Problem 10.)
130 Series Solutions of Differential Equations: Special Functions
of the leading terms of possible series solutions of ¡he form ( 18), are called the
exponenls of Ihe differenlial equalion at x = o.
For each s uch vaLue of s, the vanishing of the coefficient of x' - ' in (22) gives
tbe requirement
fes + I)A, = -g,(s + I)A o
and berrce determines A, in terms of Ao if fes + 1) o. Next, the vani shing *
of the coefficient of x' in (22) determines A, in terms of A, and A o,
fes + 2)A, = -g ,(s + 2).4, - g,(s + 2)Ao,
and hence in terms of A o, if fes + 2) o. *
In general, the vanishing of the coefficient of X ,+ k-' in Equation (22) gives
the recurrence formula
k
fes + k)A k = - L: gis + k)A k _n (k :> J), (24)
n- '
whicb determines each Ak in terms of the preceding A's, and hence in terms of
Ao, if for each positive integer k the quantity fes + k) is not zero.
Thus, if two distinct values of s are delermioed by (23), and if for each such
value of s the quantity fes + k) is never zero for any positive integer k, the
coefficients of two series of the form (18) are determined and these series are
solutions of (\ 6) in their interval of convergence. In such cases, if the solution
obtained with s = s, is denoted by A ou, (x) and that with s = s, by A ou 2(x),
then the general solution can be expressed in the form y = c,u,(x) + c,u,(x).
We next investigate the exceplional cases.
Let the roots of Equation (23) be s = s, and s = S2' where
SI =
1 -
2
Po !-. -}-v"(I - Po)2 - 4Q o,
(25)
- Po - 1
52 =
2
2,J( I - P O)2 - 4Q o·
The first exceptional case is then the case when the exponents s, and S2 are equal,
(L - Po)' - 4Qo = o. (26)
In this case only one solution of the form (18) can be obtained.
Now suppose that the two exponents are distinct. The second exceptiooal
case may then arise if fes I + k) or f(s2 + k) vanishes for a positive integral
value of k, say k = K , so that (24) cannot be solved for the coefficienl Ax. Witb
the notation of (25), we have
fes) = (s - s,)(s - S2) '
and hence
fe s + k) = (s +k - s,)(s +k - s,),
fes, + k) = k[k + (s, - S2)]' f(s2 + k) = k[k - (S, - S2)]' (27)
-'.4. The Method of Frobenius 131
lf s, i$ imaginary and the coefficients P k , Qk, and Rk are all real, then S2 is the
conjugate complex number. Hence in this case SI - S2 is imaginary and the
expressions in (27) cannot vanish for any real values of k except k = O. Next
suppose that SI and S2 are real and distinct and that s, > s,. Then, since
SI - s, > 0, it follows that fes, + k) cannot vanish for k >- 1 and that
¡(S2 + k) can vaoish only when k = s, - S2' Since k may take on only posi-
tive integral values, lhis condition is possib/e on/y if s I - S2 is a posili ve inleger.
If s , = S2' then fes, + k) = k', and hence fes, + k) cannot vanish when k >- l.
We thus see that if lhe lwo exponenlS s, and S2 do nol differ by zero or a
positive inleger, two distincl so/ulions of lype (18) are obtained. if the exponenlS
are equa/, one s uch so/ution is oblained, whereas if the exponenls differ by a
positive inleger, a so/ulion of lype (18) corresponding lo lhe /arger exponenl is
oblained.
lt is known (see, for exam ple, Reference 6 of Chapter 1) that lheinterva/
of convergence of each series so oblained is al /easl lhe /argesl inlerva/, cen lered
al x = 0, inside which lhe expansions of xa,(x) and x 2 a'(x) in powers of x bOlh
converge, with the natural understanding that the point x = O itself must be
excluded when the exponent (s, or S2) is negative or has a negative real part.
When x is complex, each infmi te series converges to a solution in a circle in ¡he
complex plane, with center at the origin and radius at least ¡he distance to the
nearest singularity of a,(x) or a2(x), and with ¡he center deleted when necessary.
The solution theo is said to have a pole at the origin wheo the associated exponent
is a negative integer, and a branch poinl at the origin when the exponent is non-
integral, as well as in the exceptional cases (Section 4.5) when the function log x
is involved (see Chapter 10).
If s, - S2 = K, where K is a positive integer, then when k = K the recur-
rence formula (24) becomes
K
(s - s,)(s - S2 + K)A K - - L:
n= I
gn(s + K)A K - n· (28)
Thus, as we have seeo, the left member vanishes when s = s" and the equation
canoot be satisfied by aoy value of A K uoless it happens that the right member
is also zero,t in which case the coefficient A K is uodetermined, aod hence arbi-
trary. If this conditioo exists, a solution of t yp e (18) is then obtained, correspond-
ing lo the sma/ler exponenl S2, which cootaios lwo arbilrary conSlanls A a and Ax,
and heoce is the comp/ele so/uNon. Thus we cooelude that, if the exponents
differ by a positive ioteger, either no so/ulion of type (18) is obtained for the
smaller exponent or lwo independenl so/ulions are obtained.
In the latter case the two solutions so obtained must then inelude the
solution corresponding to the larger exponent as the coefficieot of A K • It is
important to notice that this is the s ituation, for example, when x = O is an
tHere it is to be understood that the recurrence fo rmula has been used 10 express A" A2, ... ,
A K - , as multiples of Aa, so Ihat the right-hand member of (28) is expressed as Aa multiplied
by a specific funclion of s.
132 Series Solutions oC Differential EqualioDs: Special Functions
ordinary p oint . For in lhis case one has Po = Qo = Q, = O, and hence s, = 1,

S2 = O, and K = s, - S2 = 1. Thus Equation (28) here beco mes
ses + I)A, = -g,(s + I)A o
-(R,s2 + (P, - R,)slA o
-s[R,s + (P, - R , )lA o ,
and when s = S2 = O the recurrence formula is identically satisfied , leaving A,
as well as Ao arbitrary. Thus, when x = O is an ordinary point, two Iinearly
independent solutions which are regular at X = O are obtained.
The preceding detailed derivation was intended for the purpose of in-
vestigating the existen ce of series solutions of the assumed type. Although the
formulas obtained caD be used directly for the determinatioD of the coefficients,
once the functions fes), g,(s), g2(S), .. . are identified, it is usually preferable to
obtain the indicial equation and the recurrence formulas in actual practice by
direct substitution of the assumed series into the differential equatioD, written
in any convenient formo
To iilustrate the application of the preceding treatment, we consider again
the second example, Equation (I3), of Section 4.2. We thus seek solutions of
the equation
d2 d
Ly
dX2
==
X2 --..X + (x + x) ~ - y = O
2
dx
of the form
~ ~
y = x'
k =
I:o AkXk = I:o Ak Xk +,.
k-
By direct substitution in the differential equation, there follows
Ly = (S2 - l)A ox'

~
+ I: [[(s +
k=1
k)2 - 11Ak + (s + k - l)Ak_,}X k +,.
Hence the indicial equation is

S2 - 1= O
and the recurren ce formula is
[(s + k)2 - + (s + k - l)A
11Ak k _ , = O
or (s + k - I)[(s + k + I)A + A -¡) = O k k (k >- 1).
The exponents s ¡ and S2 are + I and -1, respectively.
Since they differ by an
integer, a solution of the required type is assured oruy when s has the larger
value + l.
With s = + 1, the recurrence formula becomes
k[(k + 2)A k + Ak - ,1 = O,
or, since k * O,
(k >- 1).
4.4. The Method of Frobenius 133
Thus one has

A - Ao A - _ Ao ... ,
1 - 3.4' J - 3.4.5'
and hence the solution corresponding to s = I is
y = X(A o - A ox
3
+ Ao x' _. An xJ
3·4 J·4·)
+ ... )
... )
_ 2A e-X -
- o
1 + x
'
x
in accordance with the result obtained previously.
With s = -1, the recurrence formula becomes
(k - 2)(kAk + A k _,) = O (k :> 1).
It is important to notice that the factor k - 2 cannot be canceled except on
the understanding that k "*
2. That is, when k = 2, the correct form of the
recurrence formula is O = O. If k = 1, there follows
A, = -A o .
If k = 2, the recurrence formula is identically satisfied, so that A 2 is arbitrary.
If k :> 3, the recurrence formula can be written in the form
A - _A k _, (k:> 3).
k- -k-
If we take A 2 = O, then there follows AJ = A. = ... = O, and the solution

corresponding to S = -1 becomes simply
l-x
y = x-'(A o - Aox) = Ao .
x
The general solution of the given equation is then a linear combillation of
the two solutions so obtained, and hence can be taken conveniently in the form
y = c,x(l - 2-
3
+ 3·4
x' - ... ) +c 2 x-'(1 - x)
= 2c,
e-X - 1 + x + C2-----
1- x
X X
or, alternatively,
e- X I - x
y = C,-
X
+ C2
X
'
where C, = 2c" and C , = c, - 2c, . The only solution regular at x = O is the

one formerly obtained,
e-X - 1+ x
y=c .
x
Jt can be verified that if Al is left arbitrary, the solution corresponding to

tbe exponent -1 is the sum of Aa times the two-term solution obtained and A,
times the infinite series solution corresponding to the exponent +].
For the purpose of computational efficiency, when SI and S2 differ by a
posi/ive in/eger K, it usually is desirable to explore first the situation correspond-
ing to the smaller exponent, since then if a nontrivial solution is obtained with
S = S2 it will be permissible to leave both A K and Aa arbitrary in that solution,
and so to obtain the general solution without needing to proceed also to a
determination with s = SI' A frequently committed blunder, which vitiates tbis
possibility, consists of dividing both members of the recurrence formula by
k - K and then setting k = K in the result (and hence dividing by zero).
4.5. Treatment oC Exceptional Cases.t We consider first the case of equal

exponen/s, and attempt to determine a second solution which is independent
of the one obtained by the method of Frobenius. Although this result could be
accomplished by the methods of Section 1.10, the method to be given is usually
more easily appJied.
In place of first introducing the value of the repeated exponent SI into the
recurrence formula (24) and then determining Al' A 2' • . . , A k , ... directly in
terms of Aa, we suppose that the coefficients A 1> Al' ... , A k , . . • first are
expressed, by use of the recurrence formula, in terms of Aa and s. We indicate
tbis fact by writing Al = AI(s), A 2 = A 2 (s), ... , where, in fact, each Ak(s) will
be of the form Aack(s), with ck(s) a specific function of s. With these values of the
A 's, as functions of s, a function y depending upon S as well as x is determined
and is denoted here by y(x, s);
=
y(x, s) = x" 2: Ak(s)X k . (29)
k ~ a
Reference to Equation (22) shows that satisfaction of the recurrence formula,

for k ::> 1, brings about vanishing of all terms in (22) except the first, and so
tbere follows
Ly(x, s) = A a f(s)x'-2, (30)
or, since in the case of a repeated exponent SI we have fes) = (s - SI)',
Ly(x, s) = Aa(s - SI)'X,-2. (31 )

The fact that the right-hand mem ber of (31) vanishes when s = SI is in accor-
dance with the known fact that (29) beco!J1es a solution of (16), say y¡(x), wben
s = SI; that is,
Yl(X) =y(x, SI)'
However, since s = SI IS a repeated zero of the right-hand member of (31), it
tSections 4.5 and 4.6, together with the derivation of the series for Ya(x) in Section 4.8, can
be omitted without JogicaI difficulty. However, in this event a consideration of the last para-
graph of Section 4.5 (and/or the working of Problems J6 and J7) is suggested.
4.5. Trealmenl oC Exceptional Cases 135
follows also that the result of differentiating either member of (31) with respect
to s (holding x constant),
a Ly(x,
as s) = A o[2(s - s,) + (s - S,)2 log X]X'-2,
is zero when s = SI ' But, since the operator alas and the linear operator L are
commutative, there then follows also
[ ~Ly(X,
as
s)] S-S I
= L[ay(X,
as
s)] S~.I'J = O. (32)
Hence a second solution of (16), when SI = S" is of the form
h(X) = [ay(x, S)] (33)

as S =S I
The second exceptional case is that in which the exponents differ by a

positive integer K,
SI -s,=K > 1,
but where the recurrence formula is not identically satisfied when k = K and
S = s" that is, when the right member of (28) does not vanish when S = S2'
In such a case, Equation (28) can be satisfied only if Ao = Al = ... = A K _ I = 0,
and hence Equation (16) does not possess a solution of type (18) beginning
with a termo of the form Aox".
In this case we suppose again that the recurrence formula is satisfied when
k > l for all values of s, so that with each Ak expressed as the product of Ao
aod a certain fuoction of s, we again define a fuoction y(x, s) of form (29). lo
this case, however, it is clear from (28) and from the nature of the recurrence
formula (24) tbat the expressions for tbe coefficieots AK(s), Ax+l(s), ... oow all
will ha ve a factor s - S2 in a denominator, and hence will not approach finite
limits as s - > 52' Ifwe consider the product (s - S2)y(X, s), we see that as S - > S2
terms witb coefficients Ak for which k < K will vanish and the remaioing terms
will approach finite limits, thus giving rise to an infioite series of powe'rs of x
starting with a term involving X,,+K = x". Thus the limiting series must be
proportional to tbe series for which s = SI ' In this case, however, since again
satisfaction of the recurrence formula for k > 1 causes al.l terms of (22) except
tbe first to vanish, we have
L (s - S2)Y(X, s)} = Ao(s - S,)2(S - SI)X'-2. (34)
But since the ríght member bas a double zero S = S2' tbe partÍal derivative of
either member must vanish as S ~ ..1'2' and, by an argument similar to that
leading to (32), we conclude that
L{ts[(S - S2)y(X, s)]} , _" = °

so that the function
(35)
is a solution of (16), in addition to the solution y I (x) = [y(x, s)], ~ " correspond_
ing to the larger exponent SI'
From Equations (29) and (33) it follows that when S2 = SI the second
solution )' 2 is expressible as
(36)
and the coefficient of log x is seen to be YI(x). Further, when s, and SI differ
by a positive integer but there is no Frobenius series solution with S = s"
Equations (29) and (35) show that the missing solution Y2 is expressible as
Yz(x) = [1=0 (s - s1)Ak(s)xk+'I~" log x
+ 1;0 {fs [(s - s2)A k(s)] L" Xk+". (37)
The coefficient of log x is lim,_" [(s - S2)Y(X, s)], which has been seen to be
proportiona/ to y Jx).
Hence it follows that, in all cases when the differentia/ equation, having
x = O as a regular singular point with exponents SJ and s" possesses on/y one
solution
~
YI(X) = L:
. ~ o
AkXk+" = Aoul(x)
of the form (18), any independent solution is of the form

~
Yz(x) = Cul(x) log x + L:k- O

Bkx k +" , (38)
where e is a constant. Thus, in place of using the reslllts of Equations (33) or

(35) in such cases, a second so/ution may be obtained by direct/y assuming a
solution of this last form and by determining the necessary relationships between
the coefficients Bk and an arbitrarily chosen constant c"* o.
4.6. Example of an Exceptional Case. To illustrate the procedure's devel-

oped in the preceding section, we consider the equation
_ d 2y
Ly = x d 2 - Y = O. (39)
x
With the notation
~
y(x, s) = L: A.(s)xk+',
k - O
(40)
there follows
~
Ly(x, s) = ses - l)A ox<-l + L:

k=1
[(k + s)(k + s - l)A k - Ak_t]xk+,-I.
Hence we obtain the recurrence formula

(k + sXk + s - I )A k = A k_ I (k >- 1) (41)
4.6. Example oC an Exceplional Case 137
and tbe two i ndices

s,=I, S2=0. (42)
Since the indices differ by unity, a solution of the form (40) is assured only for
s = 1.
From 41 there follows
A - Ao A - Ao ... ,
'-(s+l)s' 2 - (s + 2)(s + 1)2s'
and in general, by ind uctive reasoning,
Ak(s) = Ao (k >- 2). (43)

(s + k)[(s + k - 1) ... (s + 1)]2S
The solution for which s = S¡ = l then becomes
Xk+ 1
= Aou,(x).
00
yJx) = Ao k~O (k + l)! k! (44)
However, since Ak(s) - , 00 when s ~ O, for all k >- 1, there IS indeed no

solulion of Iype (40) for which s = O.
In order lO obtain a second solution, we may refer to Equation (37). The
coefficient of log x is seen to be
(45)
when account is taken of the vanishing of (sAo),~o' The coefficient d[sAk(s)]/ds,

involved in the second series of Equation (37), is conveniently evaluated by
logarithmic differentiation. For this purpose, we first deduce from (43) the
relation
k-'
10g [sAk(s)] = log Ao - log (s + k) - 2 :L; log (s + m),
m=1
then differentiate the equal members and resolve the result in the form
1 +2I;' 1
s + k m~' s + m A (k >- 2).
(s + k)[s + k - 1) ... (s + 1))2 o
Thus there follows
rp(k) + rp(k - 1) A (k >- 2), (46)
k!(k - 1)! o
with the abbreviation

k 1 1 1 +_1
:L;-=I+-+-+ (k = 1,2, ... ),
rp(k) = m- ' m 2 3 k (47)
{
O (k = O).
lt is found by direct calculation that Equation (46) aIso hoJds for k = 1, whereas
the right-hand member must be replaced by Aa when k = O. Hence (37) gives
- ~ rp(k) + rp(k - 1)
Xk]
l
Yo(x) = Aa u,(x) log X + 1 -"J;I k! (k _ 1) ! (48)
or, in expanded form,

h(x) = Ao[(x + tx2 + ¡~x' + ... ) log x + (1 - x - "ix 2 - n,x' - ... )].
The alternative procedure described at the end of the preceding section
consists of substituting the relation of Equation (38) in the form
(49)
directly into (39), to obtain the condition
c[ (x ~;2' - u,) 10g x + 2 ~; - ~ U 1]

=
+ kI:- O [(k + l)kB k+, - Bk]X k = O. (50)
Since u, satisfies Equation (39), the coefficient of log x in (50) vanishes, and the
introduction of (45) reduces (50) to the form
to [(k + l) kB k+1 - Bk]X

k
+ C .to 2
[(k !)2 - (k + \)! k !Jxk = O.
The requirement that the coefficient of the general power Xk vanish beco mes
, _ 2k + 1
(k , l)kB k +¡ - Bk - - (k + 1)! k ! C (k >- O).
By setting k successively equal to O, 1, 2, ... , we obtain

Bo = C, B2 = tB¡ - -¡c, B, = ¡1z-B¡ - ...Jr;C,
Here both B¡ and C are arbitrary and alI the other coefficients are expressible
in terms of these two constants, yielding the solution (49) in the form
Y2(X) = C[(x + tx 2 + tix' + ... ) log x + (1 - -¡X2 - ...Jr;x' + ... )]
+ B¡(x + tx2 + tix' + ... ). (51)
The coefficient of B, is seen to be u¡(x). Thus, if C and B¡ are left as arbitrary
constants, this expression for y,(x) in fact represents the general solution of(39).
The particular express ion for h obtained in (48), by the first method, is obtained
I
from (51) by choosing C = -B , = Aa.
4.7. A Particular Class oC Equations. Many important second-order

equations, of frequent occurrence in practice, can be obtained by specializing
the constants in the equation
2 y
(1 + RMXM) dd {
x
+ _1 (Po
x
+ PMX M) dd
x
+ -;(Qo
x
+ QMXM)y = O, (52)
4.7. A Particular Class of Equations
139
where "1 is a positive integer. (In the case when NI = O, the equation is equi-
dimensional.) Here the introduclion of the assumption
y(x) = L: Akx k +' (53)

k=Q
leads lo the condition
where
fes) = S2 + (Po - I)s + Qo
and g(s) gM(S) = RM(s - M)2 + (P M - RM)(s - M) + QM'
Thus, for each exponent satisfying the indicial equation fes) = O, the recUf-
rence formula is
fes + k)A k = O (k = 1,2, ... ,M-l),
fes + k)A k = -g(s + k)A k - M (k >- M).
The first M - l conditions are satisfied by taking
Al = A2 = ... = AM - 1 = O,
after which the recurrence formula for k >- M shows that all coefficients Ak
for which k is not an integral rnultiple of NI can be taken to be zero.
Accordingly, it is convenient to write
~
y(x) = L: Bkx Mk +'

k=O
(54)
when seeking a solution of a special case of Equa/ion (52). Here k bas been
replaced by Mk in (53) and Bk has been written for AMk'
Further, it is seen tbat here an exceptional case can occur only when Ihe
exponents SI and S2 are equal or when SI - S2 = KM, I1'here K is a positive
integer.t In such a situation, when only one solution of type (54) is obtained,
a second solution can be found, as usual, by use of Equations (33) or (35).
Since the expansion of (1 + RMxM)-1 converges when
(55)
the solutions obtained for (52) also wiII converge in that interval. In particular,
if RM = O, the series wiII converge for aH finite values of x (the value x = O
itself being excepted, as usual, when the real part of S is negative).
Among the many important specializations of Equation (52), we note
Bessel's equation,
(56)
tThe present procedure c1early bypasses the possibility of determining a Iwo-parameter solu-
tion with s = S2 when s 1 - S2 is a positive integer other than a multiple of M, but it provides
one solution for each exponent in such special cases.
for which M = 2; Legendre's equation,

d 2y dy
(1 -
2
x )d 2
x
- 2x -
dx
+ p(p + I)y = 0, (57)
for which M = 2; and Gauss's equation,

d 2y dy
x( 1 - x) dX2 + [y - (a + f3 + I)x] dx - txf3y = 0, (58)
for which M = l. The solutions of these equations, in the neighborhood of

x = 0, are studied in the following sections .
Other notable special cases may be listed as follows.
(1) The eq uation

d 2y dy_
X dX2 + (e " x) dx -ay - 0, (59)
for which M = 1, is satisfied by the confiuent hypergeometric function of K u m-

mer, y = M(a , e, x) (see Problem 8). If e = 1 and a = -n, where n is a positive
integer or zero, one solution is the nth Laguerre polynomial, y = L.(x). If
e = m + l and a = m - n, where m and n are integral, one solution is the
associated Laguerre polynomial,
dm
y = L';:(x) = dxm L.(x),
if m -<: n.
(2) The equation
2
d~ d
- 2x ~ + 2ny = 0, (60)
dX2 dx
for which M = 2, is satisfied by the nth Hermite polynomial, y = H.(x), when

n is a positive integer or zero .
(3) The equation
d2 d
(1 - X2) ~ -
dX2
X ~
dx
+ n2y = O, (61)
for which M = 2, is satisfied by the nth Chebyshev polynomial, y = T.(x), when

(4) The equation
d 2y dy
x(1 - x) dx 2 + [a - (1 + b)x] dx + n(b + n)y = O, (62)
for which M = 1, is satisfied by the nth Jacobi polynomial, y = J.(a, b, x), when
The functions mentioned are useful in many applications. lt is a curious

fact that they all satisfy equations which are special cases of (52),
4.8. Bessel Functions 141
4.8. Bessel Functiolls. Solutions of the dilferential eguation
X 2 d2
2
d y + X dy
- + ( x2 - p2)y =O, (63)
X dX
or, equivalently,
X!(X~~) + (x 2- p2)y = O, (63a)
are knoWD as Besselfunctions of order p. These functions are of freguent use in

the solution of many types of potential problems involving circular cylindrical
boundaries, as well as in other applications, in such fields as elasticity, fluid
flow, electrical field theory, and aerodynamic flutter analysis. We suppose that
the constant p is real. Since only the quantity pl appears in Equation (63), we
may also con s ider p to be nonnegative without loss of generality.
Since Equation (63) is of type (52), with Jv[ = 2, we may seek a solution of
the form (54),
~
y(x) = I: B k x 2k +'. (64)

k- O
Substitution into (63) yields the indices

SI =p, S2 =-p (65)
and the recurrence formula
(s + 2k + p)(s + 2k - p)B k = -B k - 1 (k >- 1). (66)
The exceptional cases may arise only if s. - S2 = 2p is zero or an integral
roultiple of M = 2, tbat is, if pis zero or a pos itive integer. In such cases we can
be certain only of one solution of type (64) .
In correspondence with the exponent SI, = p, repeated use of Equation (66)
gives
Bk(p) = (-l)k (2 + 2p)(4 + 2~) ... (2k + 2p) 2 'k! f

=(-I)k(1+P)(2+~) ... (k+P)2f.k! (k>-l), (67)
so that a series of type (64) is determined, for s = p, in the form

~ (_1)kx2k+p ]
P
y.(x) = Bo [ x + k~. (1 + p)(2 + p) . . . (k + p)2 2k k! '
or, with use made of Equation (52) of Chapter 2,
~ (-I)k x 2k +P
y .(x) = Bor(l + p) k~ 2 2k k! r(k + p + 1)'
This result is put into a more compact form if we use the abbreviatioD
r(k + p + 1) = (k + p)! and write
(68)
142 Series Solutions of Differential Equations: Special FunctiOIlS
Tbe series multiplying 2'p'B o in Equation (68) is known as the Bessel

funclion of /he firs/ kind, of order p, and is denoted by Jp(x),
_ - (-I)"(x/2)2k+,
Jp(x) - k~O k! (k +
p)! . (69)
In particular, when p = O and 1, we obtain the series in the forms
(70)
X7
2 7 3! 4! + .... (71)
With s = S2 - - p, Equation (66) yields the result of replacing p by - P In

(67),
Bk(-p) = (-I)k (1 _ p)(2 _ ~) (k :> 1). (72)
Thus, if p is a positive integer, al! coefficients Bk for which k :> P beco me

infinite, and no Frobenius solution is obtained, in such a case, corresponding
to the exponent s = -p. However, if p is not zero or a positive integer, a
second solution is obtained by replacing p by -p in the first solution, and hence
may be taken in the form
~ (-l)k(x/2)2k-,
J_i x) = I:
k-O
k' (k _
•
), .
P . (73)
Thus, if p is no/ zero or a posi/ive in/eger, /he comple/e solu/ion of Besse/'s equa-
/ion (63) is a linear combina/ion of /he solu/ions (69) and (73), of the form
y = CJp(x) + C 2 J_ P (x). (74)
If p = O, the two solutions are identical. Moreover, if pis a positive integer,
the second solution J _p(x) is not independent of Jp(x). This statement is a
coosequeoce of the fact that if p is a positive integer n, the factor J /(k - n)! in
(73) is zero when k < n, and hence (73) is then equivalent to
_ ~ ( - l )k(x/ 2)2k-·
J_.(x) -~. k! (k - n)!
or, with the index k replaced by k + n,
- (_l)k+n(x/2)2k+.
J_n(x) = ~
k-O
k'. (k + ), .
n .
Hence, if n is an in/eger, we obtain
J_.(x) = (-I)'J.(x). (75)
lt should be noticed that, although the higher coefficients io the series yz(x)
woulct beco me infinite asp ~ n (n = J, 2, 3, ... ) if the coefficient Bo were held
fixed, we have obtaioed (73) by setting Bo(-p)! = 1 or Bo = l/r(1 -p) in
that series and, since this quantity tends to zero as p - n, we have obtained a
solution J _ ,(x) in which the coefficients which previously beca me infinite as
p .-+ 17 now approach finite limits, and the remaining coefficients tend to zero.
4.8. Bessel Functioos
143
To fiod a second solutioD complementiog JAx) iD the exceptional cases,

recourse may be had to the methods of Sect iOD 4.5. We illustrate the procedure
in the case of equal exponents, p = O. In this case we obtaio a fllnction y(x, s)
by determining Bk(s) from (66) and introducing the result into (64). The
required second solution is then
y,(x) = [ay(x,
as
s)] ,
"0
where here
"
y(x. s) = I:
k=O
B k (s)X 2k +' ,
s~ that
(76)
The recurrence formula (66) first gives
B
Bk(s) = (-1)' [(s + 2)(s + 4) o... (s + 2k)]2
To calculate B~(s) it is again convenient first to take the logaritbm of the two
sides of the equation ,
log [( -IY BB~S)] = -2 log [(s + 2)(s + 4) ... (s + 2k))

k
= -2 I:
m _ 1
log (s + 2m).
Differentiation with respect to s then gives

B~(s) = -2
Bk(s) m~ l S
± + 2m
1
aod hence
B~(O) _ -2
Bk(O) -
_1_
m_ 12m
± = k
I:
m=1
l
m
Thus. ir we agaio introduce the abbreviation
rp(k) = t
m~ l
_1 = 1
m
+ _1
2
+ . .. +_1
k
(k >- 1), (77a)
with rp(O) = O. (77b)
we obtain
B~(O) = - rp(k)Bk(O) = -rp(k)[(-IY[2.4.6 ~~. (2k))2]
( - I)krp(k)
= - 2 Zk (k !)' B o ,
aod (76) then gives the required second solution in the case p = O in the form
h(x) = B.[Jo(x) log x + ti (-lrlrp(k)«(k~~:kJ

The coefficient of Bo is thus a second solution of Bessel's equation (63)

wheo p = O,
d 2y dy _
x d x 2 + -d x + xy - O.
It was taken as the standard form of the second solution by Neumann, and is
us ually denoted by y rOl(x). Thus any linear combinatio n of J O(x ) and y rOl(x) is
al so a solution . The standard fo rm chosen by Weber is defined in term s of Jo
and y rOl by the equation
2
Yo(.'C) = ¡r[yro)(x ) + (y - log 2)J o(x )], (78)
where y is Eu/er 's conSlGnt , defined by the relation

y = lim [tp(k) - log k]
++
k- -
= ~i~ (1 + + + . .. + l- logk) = 0.5772157.... (79)
We thus obtain a second solution in the form
Yo(x) = n2 [( log""2X+ )Y J o(x) + f;o

- (-I)k+ 'tp(k) (x/2)2k]
(k !)2
= ; [(IOg ~ + y) J o(x ) + {~: - 2.~'!)2(1 + +)

+ 26(;!)2 ( 1 + ~ + +) - ... }1 (80)
The function Yo(x) is known a s Weber' s Besse/ fimcrion of the second kind, 01
order zero. In German texts it is frequently denoted by No(x). The complete
solution thus can be written in the form
(81)
Weber's definition of the function of the second kind [Equation (78)] is more
convenient than that of Neumann becau se of the fact that the behavior of the
fUDction Y O(x ) so defined,lor /arge va/ues 01 x, is more nearly comparable with
the behavior of J o(x) [see Equatíon (88)].
A similar but more involved calculatíon leads to expressions for Weber's
form of the Bessel fun ction of lhe second kind, of order n,
Yn(x) = ; [ (log ~ + y) Jn( x ) - ~ %0 (n - k - i~! (x / 2)2k - n
I - (x / 2)2k+n ]
+ ""2 .&0(- I) k+ '
[cp(k) + tp(k + n)] k! (n + k) ! (82)
when n is a posilive integ er. Thus, in particular,
Y,(x) = ; [(IOg ~ + y)J¡(x) - ! - ~ + {I + (i + -D}2~;!

_{(I+±)+(I+± + 1)} x' + ... ] . (83)
2 2'2 ! 3 !
4.8. Bessel Functions 145
Ir follows that if p = n, where n is zero or a positive integer, the general solution

of (63) can be taken in the form
y = C,J"(x) + C 2 Y"(x). (84)
If pis not zero or a positive integer, the function Y/x) is defined by the equation
Y (x) = (cospn)Jp(x) - J_/x). (85)
p SIn pn
This definition can be shown to be consistent with Equation (82) as p - n, and

it defines Yp(x) as a linear combination of Jp(x) and J_p(x) otherwise. It should
be emphasized, however, that the second solution Yp(x) is not needed unless
p= n.
The general solution of Bessel's equation is frequently abbreviated by use
of the notation
(86)
with the convention that (86) stands for (74) unless pis zero or a positive integer,
and for (84) in these cases.
The transformation y = u(x)/,Jx transforms (63) to the equation
2
d u +
dX2
p2 - (1 _U=
X2
*) O (87)
(see Problem 47 of Chapter 1). For large values of x the term (p2 - *)/x 2 is
negligible in comparison with unity. Thus it may be expected that for large
va)ues of x the behavior of solutions of (87) will be similar to that of correspond-
ing solutions of the equation
d 2v
dX2 + V = O.
Since such soIutions can be written in the form v = A cos (x - rp), where A and
'P are constants, we are led to the possibility that for large values of x any solu-
tion of (63) behaves like tbe function
v(x) A
r = reos (x - 'P),
,..,¡X ,..,¡X
for properly chosen values of A and 'P. A rather involved analysis shows that
for the function Jp(x) one has
Al=J; and 'Pl=(2p+l)~,

whereas for Yp(x) there follows
A2 = J; and
n
'P2=""2+'Pj'
Thus we may write
Jp(x)
Y/x)
~
~
J n: cos (x -
/ 2 sin (x -
IX p)
IX p)
¡ (x -, C<J), (88)
'V 7r.X
146 Series SolutioDS of Differential Equations: Special Functions
where (X,p = (2p + 1) ~ . (89)
T he nota ti on of (88) denotes ¡hat t be ra tio of t wo expressions connected by the

sym bol ~ approaches unity as x - oo . We say that Jp(x) b eh aves ClSymplolically
like -/2/ 1tx cos (x - (X, p).
Jt follows from (88) that the complex function Jp(x) + i Yp(x) ha s the
asymp lOtic behavior
(x - 00 ), (90)
whereas the conjugate complex func t ion has the behavio r
J p(x) - iYp(x) - .J 2 e-i(x-,.)

1t X
(x - 00 ). (9 1)
These complex func tio n s are known as Bessel funclion s of Ih e Ihird kind, or,
m o re generaJly, as the Hankelfun ctions of lh e first and second kinds , respectively,
and the abbre viation s
H ~1J(x) = Jp(x) + iYp(x),
(92)
H ~2 ) (X) = J p(x) - iYp(x)
are conventional. The Hankel fun c tions are particularly u seful in studying cer-
tain type s of wa ve propagation (see Section 9.13) .
The differential equation
(93)
wruch differs from Besse I's equation (6 3) only in the s ign of X2 in the coefficie nt
o f y, is tra n sformed by the substitution ix = I to the equation
2
t2 d y
dt'
+ t dydi + ( 12 _ p2)y = O,
"'
which is in the form of Bessel's equation (63). Hence, the general solution is of
the form y = Z.(t), or, in term s of the original variable x, the general solution
of (9 3) is
y = Ziix). (94)
That is, if p-is not zero o r a positive integer, the general soluti o n is o f the forro
y = c .J p(ix) + c, J_.cix ),
whereas o therwise it may be ta ken in the form
y = c.Jn(ix) + c 2 Y.(i x).
From Equation (69) we have
. _ ~ ( - I )ki 2k + p (x / 2)2k+ P _ 'p ~ (x / 2)2k+' .
J p(IX) - {;o k! (k + p)! - 1 {;o k ! (k + p)!
4.9. Properlies of Bessel Functions 147
In place of usiog this function as a fundamental solution of (93), it is preferable

to use the function I .(x) = ¡-'Jp(ix),
_ - (xí 2)2k+P
l /x) - k~O k ! (k +
p) ! ' (95)
since tbis function is real for real vaJues of p. This fuoction is known as tbe
modified Bessel function of the fi rst kind, of order p. The term s in the series
representing [p(x) differ from those in the series for Jp(x) only 'i n that the terms
are all posi tive in the I p series, whereas they alternate in sign in the J . series.
Tbus, if p is not zero or a positive integer, the general solution o f (93) can be
take n in the real form
(96)
As a seco nd real fundamental solutioo of (93), in the case wheo p = n,
wbere n is zero or a positive intege r, it is rather conventional to define the fuoc-
tioo K.(x) by the equatiooT
(97)
leacliog to the general so lution

y = Z.(ix) = cJ.(x) + c 2 K.(x), (98)
wbeo n is zero or a positive ioteger. Tbe fuoction K.(x) is koowo as tbe modified
Bessel fimction of the second kind, of order n.
If p is not zero or a positive ioteger, tbe functioo K.(x) is defioed by tbe
equation
K (x) = !!... I .ex? - leC x ), (99)
p 2 SID p1l
wbich is coosisteot witb (97) wbeo p - n.

For large values of x the m odified fuoctions bave tbe asymptotic bebavior
e-X (x - 00). (100)

K.(x)~ ,J ;x
It is important to ootice tbat the rigbt members of (lOO) are iodepeodeot of p.
4.9. Properties oC Bessel FunCtiODS. It is readily verified direc tly tbat aU

power series involved in the defioitions of all Bessel fuocti o o s cooverge for aU
finite values of x. However, io consequence of the fact tbat tbese series are io
tSome references omit the Caetor " in (97) [and (99)]. in order to equalize the two nurnerical
Caetors in (lOO).
148 Series Solütions of Differential Equations: Special FunctioDS
maoy cases multiplied by a negative power of x or by a logarithmic term, it is

found that on{y the functions Jp(x) and ¡p(x) are finile al x = O (when p :> O).
For small values of x, retention of the leading terms in the respective series
leads to the approximations
(p =F n), (01)
2
(p =F O), Yo(x) ~ - log x, (102)
re
l
Ip(x) ~ '5I'I xP , (p =F n), (103)
- p.
Kp(x) ~ 2 P - 1 (p - I)! [P (p =F O), Ko(x) ~ -Iog x, (104)
agaio with the usual implicatioo that the ratio of two quantities connected by
the symbol ~ approaches unity as x tends to the relevant limit (he re zero).
For large values of x (x ~ 00), we recapitulate the results listed in the
preceding section:
J p(x) ~ / 2 cos
~ rex
(x _ ~4 _ pn),
2
y (x)
p
~ jv 2 sin (x _
rex
~
4
_ pn)
2
, (105)
(106)
( 107)
The following derivative formulas are of frequent use:
~[P
x yp ( ax )] -_ { axPYp_¡(ax) (y = J, Y, 1, H'I', HIZ'),
(l08)
dx -axPy P_ Jax) (y = K);
~[ _P ( )] _ {-ax-pyp+l(ax) (y = J, Y, K, H' 1', H'2'),

( 109)
d x Yo ax -
x . . ax-pYp+I(ax) (y = l).
These formulas are established for J p and Y p by considering their series defini-
tions, and for the remaining functions by considering their definitions in terms
of J p and Yp- Thus, to prove (108) for J p , we note that from the definition (69)
we have
C-~ (_1)ka2k+Px2k+2P-1
= ,{;O 2 2k + p -¡k! (k +p - I)!

_ p.~ (-I)k(ax/2)2k+ P -I
-ax t;o
k!(k+p-I)!
= axPJ p_ Jax).
4.9. Properties of Bessel Functions 149
From (l08) there follows
IXy p- 1(IXX) ~ l!... y p(IXX) (y = J, y, 1, H' 1', H(2J),

d) =
_Yp(IXX x
(J 10)
dx
1 ~ IXy p_ 1 (IXX) ~ -
Py p(IXX)
x
(y = K),
(y = J, Y, K, H ( 1 " H (2!),
( 111)
(y = 1).
By addition and subtraetion of (110) and (111) we also obtain the relations
2dd Y.(IXx)
X
= IX[YP_I(IXX) ~ Yp+,(ax)] (y = J , y, H'I ' , Hl2J), (112)
~ 2p
Y p+ 1 (IXX ) ~ - y p(IXX) ~ y p_ /IXX) (y
. = J Y H ' " , H(ZI) ,
) , ( 113)
IXX
2;Xlp(IXX) = IX[lp_I(IXX) + Ip.;.,(IXX)] , (112a)
21xKp(IXX) = ~ IX(Kp _ ,(IXX) + Kp.;.,(IXX)], (l12b)
(113a)
(l13b)
Relations (108-112) are u seful in evaluating eertain integrals and deriva-

tives involving Bessel funetions. In particular, setting p = O in (109), we obtain
the relations
d ( ) {-IXY 1 (IX x) (y = J, Y, K, H'IJ, H'2J),
-Yo IXX = (114)
dx exy, (IX x) (y = 1).
The relations (113) are useful in expressing Bessel funetions in terms of eor-
responding functions of lower order.
The Besselfunctions of order p, where pis half an odd integer, can be expressed
in closedform, in terms of elementary functions. To establish this faet, we eon-
sider first the case p = 1- and denote the general Sollltion of (63) by y = Z 1/2(X).
I If we write
--! xZ I/'(X),
1, u(x) =
I Equation (87) shows that the fllnetion u(x) satisfies the equation
d 2u
-+u = O
dx'
150 Series Solutions of Differentilll Equations: Specilll Functions
and hence is of the form 11 = A cos X + B sin x. Takin g

1/ = ,../ XJ ,:2(X),
we have also, from ( 108),
du -
- X = .../x J _ 1. 2(X) .
d
Hence, using (101), we find that
, 2 1.2 ..J2 /2
1/(0) = O and !I (O) = (-t) ! = r(t) = V n'
Thus we must have A = O, B = .../2/71., and there follows
/2 .
!I = V n SIn x,
or, finally,
J, 2(X) = / 2 sin x, (115)

V 71.x
in accordance with (88), which here becomes an equality. AIso, since 1' /2(x) =
i -1.·2J' /2(ix) and L 1.' 2(X) = i 1/2J_ I .'2(ix), we have
1"2(X) 2 sinh x, L 2(X) = 2 cosh x. (116)

= /
V 71.x ", V/ 71.x
From (113), witb ex. = 1 and p = n - ±, we obtain the recurrence formulas
J n+ 1¡ 2(X) = 2n x- 1
J n_ 1.'2(X) - J n- 3.'2(X) ( 117)
2n - 1
and I n+ ¡¡2(X) = - x I n- 1/2(X) + l n_ 3.'2(X), (l18)
which permit the determination of J n + ' /2(X) al)d In+ ' / 2(X) for all integral values
of n, in terms of the functions in (115) and (116). Certain of those functions are
listed in TabJe 2 of Chapter 2 (page 70).
As is indicated by the asymptotic approximations (l05), the functions Jp(x)
and Yp(x) are oscillatory in nature, the amplitude of oscillation about a zero
value tending to decrease with .../2/71.x and the distance between successive zeros
of the function decreasing toward 71. .t It can be shown that the zeros of J/x)
separate the zeros of Jp+,(x); that is, between any two consecutive zeros of
J p < 1 (x) there is one and only one zero of J/x) . (See Problem 36.) The sa me
applies to the zeros of the functions of the second kind. The functions Ip and
Kp are not oscillatory. It is found that the former function essentially increases
exponentially with x, whereas the second essentially decreases exponentially.
Sketches of tbese functions are presented in Figure 4. J.
tA brief table of values of zeros ofcertain Bessel functions is presented iD SectioD 5.13.
4.10. Differential EquatioDs Satisiied b y Bessel Functions lSI
~w ~W
1.0 LO
10 x x
-0.5 -0.5
(a) (b)
10 (x) a nd lOOKo(x)
100
lOOK o
50
5 x
(e)
Figure 4.1
4.10. Differential Equations Satisfied by Bessel Functions. The solution of

t be djfferential equation
2
X' d y + X dY + (X' _ p ') Y = O (119)
dX 2 dX
can be written in tbe form
y = ZP(X). (120)
lf we make tbe substitutions
X= I(x) (121)
aod notice that then
) d 1 d
f dX= f'(x) dx ' (122)
Equation (119) becomes
j
s p{J..,~[/l
1 dx , dx
~(L)J}
g
+ 1[_1~(L)J
f' dx g
+ (/2 - p2) Lg=
'
O
or 1 ~[fx
dx
(f )] + ~dx (L)g + f'1 (/2 _
1
p 2)L
g
= O. (123)
152 Series Solutions of Dilferential Equntions: Special Functions
Reference to (120) and (121) shows that (123) is the differential equation satisfied
by
y = g(x)Z.[f(x)]. (124)
In particular, if we set
g(x) = xAe- S ,', f(x) = Cx", (125)
Equation (123) can be reduced to the form
x, d y
dx'
2
+ x[(1 - 2A) + 2rBx'] ddxy
+ [A' - p'S2 ..,- S2C2X2S - rB(2A - r)x r + r2B2x2'Jy = O.
This equation is somewhat simplified if we write
1- 2A = a, rB = b,
fram which there follows
A =
I-a
2 '
b
B=-, C = ",-/d 1
p = _~A2 - C. ( 126)
r s ' S
With this notation, the differential equation takes the form

d 2y dy
X2 dX2 + x(a + 2bx') dx + [c + dx 2s - b(l - a - r)x r + b 2x2r]y = O,
(127)
and the solution of (127), obtained by introducing (125) and (126) into (124),
is of the form
y =
I -a
x- bx' (.Jd )
2- e --;:- Z p -s- x' , (128)
where p -
_ s1 J (I -2 a)2 _ c. (129)
If ~ djs is real, Z p is to be interpreted by (74) or (84), whereas if ~ djs is

imaginary, Zp stands for (96) or (98), the choice in either case depending upon
whether p is not or is zero or a positive integer, respectively.t
Thus, if it is possible to identify a particular differential equation with
(127), by suitably choosing the constants in (127), the solution is given im-
mediately by (128) in terms of Bessel functions of order p, where p is given by
(129). We list here certain useful special forms with corresponding solutions
readily obtainable in this way.
x2y" +
xy' +
(k 2 x 2 - p2)y = O, y = Zp(kx) (k O). (130a) *
(k * O). (130b)
tBecause of the ambiguity of the signs of Ihe radicals in (128) and (129), both the order p and
the coefficient of X S can be replaced by their negatives, and hence can be taken to be nOflnegative
when they are real, in (128) and in ils specializations (l30a-e).
4.11. Ber and Bei Functions 153
.:c'y" + :cy' ~ ([3' ~ a:c")y = 0, (as * O). (130c)
y k:c'"y = 0, y =
.-
---/ X Z_,_
(?
-.~
/7(; m- ' )
2x ' (mk * ~2k). (l30d)
m+2 m I
!!... (xn d ) +
y
kxmy = 0,
I -n
Y = x"'T'" Z p
(17(;
_v-- x' )
dx dx s
where p=
l~n
2s (ks * O). ( 130e)
The restricted cases in (130a-e) are all readily solvable as equidimensional

equations (see Section 1.6).
4.11. Ber and Bei Functions. In certain problems it is convenient to obtain

a desired solution as the real or imaginary part of a complex function, in terms
of which a simplified formulation of the problem is possible. In such cases
differential equations with complex coefficients may occur. In particular,
equations are rather frequently obtained which are reducible to the form
X2 dd2.~ +x dc/y ~ (p2 + iX2)y = 0, (131)

x x
where pis a real constant. Since (131) is eq uivalent to (130b) if k 2 = i or k = i 1.2,
the solution of (131) can be expressed in the form

y = Zp(i3 'x). ( 132)
The solutions of (131) which are finite at x =
3 2
are thus multiples of the °
function J p (i / X). This function is a complex function whose real and imaginary
parts are written ber pX and beipx, respectively,
(133)
The polar notation
(134)
is also used.
We consider, in illustration, the case p = O. In this case the zero subscripts
are conventionally omitted in the notation for the real and lmagmary parts,
J o(i3/2 X ) = ber x + i bei x. (135)
From the definition (69) we obtain, with p = 0,
'3 2 ) ~ ~ (~I)mi3m(x/2)2m
J o( 1 X ~ L.. (')2 '
m""O m.
independently of the choice of the two possible interpretations of i 3 2. We now

separate this series into two parts, in the first of which In takes on even values
and in the second of which only odd values of In occur. lf we re place m by 2k
in the first series and m by 2k + 1 in the second, we obtain

.3.2 _ ~ i6k(X/2)4k ~ ¡6 k +3(X/2)4k+2
1 0(1 x) - ¡';O [(2k) !F - ¡';O [(2k + 1)!)2 .
Noticing that
i 6k = (_1)3k = (-ly,
we obtain finally
(X/2)4k ~
J oCi
3 2
/ y) =
-
¡';O (-1)
k
[(2k)!J2 T
I
1
.
bo (-1) k [(2k
(xj2)4k+
+ 1) !]"
2
(136)
A comparison of (135) and (136) gives the results

- k (X/2)4k
ber x = ¡';o (-1) [(2k) !JO
_ x4 I x8 (137)
- 1 - 224 2 T 224 26 2 8 2
and bei x = ko- (-IY[(2k

(X/2)4k+2
+ 1) !j2
X2 x6 x lO
22 - 224 2 6 2 + 224 26 2 8 2 10 2
(138)
Analogous series expansions defining the functions berpx and beipx for general
values of pare obtained by similar methods.
Similar functions of the seeond kind, which are not finite at x = 0, are
defined by the relation
ker p x + i kei p x = i- P KP(i l / 2 X). ( 139)
The general solution of (131) then can be written in the form
y = (e, ber p x + e 2 ker p x) + i(e, bei p x + e 2 kei p x). (140)
To i!lustrate the occurrence of Bessel functions of order zero in practice,
we consider particular solutions of the partial differential equatiolls
(141)
and (142)
It is found that in many physical problems in various fields, a physical property

U depending on a single distance variable x measured from a reference axis of
symmetry, and on a time variable t, must satisfy one or the other of these
equations, the quantities ;t and f.l involving known physical constants indepen-
dent of time and position. It is frequently important to determine solutions of
the form
¡(x) sin wt + g(x) cos wt,
where w is a constant. However, instead of proceeding directly to such a deter-
4.U. Legendre FUDctions 155
mination, it is more convenieot to consider the required solution as Ihe real 01'
imaginary parl of a complex soll/Iion of the form
u= y(x)e''''', (143)
where y(x) is a complex function of the form y(x) = F(x) + iG(x) .
Substitution of (143) ioto (141) and subsequent cancellation of the factor
e''''' show that the function y(x) must satisfy the differential equation
x + -d
2
d y2
dy +..1.w 2 xy -_
x d
x . ° (144)
If ..1. > 0, the general solution of (144) is given by (l30a) in the form
y = Zo(,./Iwx), whereas if ..1. < 0, the general soJution is of the form
y = ZoCi"/ -..1.wx). In particular, if physical considerations require the solution
to be finite when x = 0, the solutions must be of the form CJo(,./Iwx) wheo
..1. > O and Cl o (,./ - ..1.wx) when íl < 0 , where in either case C is a constant
(which may be complex) .
In a similar way, substitution of (143) into (142) gives the equation
d 2y dy.
X dX2 + dx - If.lwxy -
_
° (145)
to be satisfied by the complex amplitude function y(x). If f.l > O, comparison

of (145) and (131) shows that the most general solution of (145) which is finite
when x = °
is of the form
y = CJO(i 3 , 2,./ f.lw x) = C(ber,./f.lw x +i bei ,./ ¡.¡,W x). (146)
If f.l < O, the corresponding solution is of the form
y = CJo(iJ/2,./_¡.¡,WX)
or, equivalently,
(147)
as may be seen by taking the complex conjugate of the equal members of (145).
Useful tables of Bessel functioos are included in Tables of Functions,
complied by Jahnke, Emde, and L6sch (Reference 8), the ootation Np(x)
being used io place of Yp(x). In addition, the real and imaginary parts of the
functions J OCil/2 X), H'o11(il/2X), and i 1/ 2 J 1 Cil/2 X ), il/ 2H\11(i 1l2 x) are tabulated
thereio. The first two functions are independent solutions of the equatioo
X2 d y
2
dx 2
+ x dy
dx
_ (p2 _ iX2)y = ° (148)
when p = 0, and the last two fuoctions are solutions when p = 1. Supplemeo-
tary material, includiog tables of zeros of Jp(x), is also included. Brief tables of
the functions appeariog in (140) are included in Dwight's Tables (Reference 3).
4.12. Legendre Functions. Solutions of the differential equation
(1 -
d2
X2) ~ -
dX2
dy
2x-
dx
+ p(p + l)y = ° (149)
156 Series Solutions of Differential Equ!ltions: Special Functions
or, eq ui valen tly,

:X[ (l-- X2) ~:~J + p(p + I)y = O ( 149a)
are known as Legendre fllnctions of order p, where p is assumed he re to be real

and nonnegative.t They are of particular use in the solution of potential prob-
lems involving spherícal boundaries, when rotational symmetry is present.
We may notice that x = O is an ordil1ory point for (149), so that one could
assume y = ¿; Akx k , with the knowledge that both Ao and A, will be arbitrary.
However, since (149) also is of type (52), wíth M = 2, it is somewhat more
convenient to use the method of Section 4.7. Thus the introduction of
=
y(x) = ¿;
k,.O
B k x 2k +' (150)
into Equatíon (149) readily yields the exponents

S2 = O ( 151)
(as was anticipated) and the recurrence formula
(s + 2k)(s + 2k - I)B k = -(p - s - 2k + 2)(p +s+ 2k - I)B k _ 1• (152)
With s = O, this formula yields the result
B (O) - (-IY[P(P - 2)(p - 4)··· (p - 2k + 2)]
k - [1.3.5 ... (2k - 1)]
x [(p + 1)(p + 3)(p + 5) ... (p + 2k - 1)]B
[2.4.6 ... (2k)] o,
or
B k (O) -- ( - 1Y B o [ ( - 2) . . . ( - ") k I 7)]

(2k)! P P P - --r-
X [(p + 1)(p + 3) ... (p + 2k - 1)]. (153)

Similarly, when s = 1, Equation (152) gives
( -I)kB
Bk(l) = (2k + 1)![(P - l)(p - 3) ... (p - 2k + 1)]
x [(p + 2)(p + 4) ... (p + 2k)]. (154)
The solutions corresponding to the exponents s = O and s = 1 then are of
the respective forms
Bo + ¿;
k=l
B k (0)X 2k , Box + ¿; k=l
B.(I)x 2k +'.
tSince p(p + J) is unchanged when pis replaced by -(p + 1), the solutions for p ~ -Po are
(he same as those for p = Po - J.
4.12. Legendre Functions 157
The coefficients of Ba in these expressions are here denoted by up(x) and vp(x),
respectiveJy, so that we write
Up(x)
= J - p(p + l)x2 + p(p - 2)(p -i-- I)(p + 3)x4
2! . 4! "
-
p(p - 2)(p - 4)(p +
6!
I)(p + 3)(p + 5)
x
6
-,-
;
(155)
and
(p - l)(p - 3)(p
_,
+ 2)(p + 4) X ,
).
-
(p - I)(p - 3)(p - 5)(p
7!
+ 2)(p -;- 4)(p + 6)
x
7
,
•
... (156)
It may be seen that, if pis an el'en posilive inleger n (or zero) the- series (155)
terminates with the term involving x n , and hence is a polynomial of degree n.
Similarly, if p is an odd positive inleger n, the series (156) terminates with the
term involving xn. Otherwise, the expressions are infinite series. The results of
Section 4.4 show that the series converge when - I < x < 1; they diverge
otherwise (unless they terminate), as can be verified directly.
Thus the general solution of Equation (149) could be expressed in the form
y = c,up(x) +c 2 v p(x)
when -1 < x < l. However, a different terminology IS conventional, for
reasons which are now to be explained.
We consider first the cases when p = n, where n is a posltlve integer or
zero. These are the cases commonly arising in practice. When p = n, one of
the solutions (155) or (156) is a polynomial of degree n, whereas the other is
an infinite series. That multiple of the polynomial of degree n which has the
value unity when x = I is called the nth Legendre polynomial and is denoted
by Pn(x).
Thus we have, when n is even,
P( ) = un(x) (1 57a)
nx un(l)
and, when n is odd,
P ( ) = vn(x). (157b)
n X vil)
The first six Legendre polynomials are readily found to be

Pa(X) = 1, P, (x) = x, P 2 (x) = t(3x 2 - 1),
P3(X) = ±(5x 3 - 3x), P 4 (x) = S·(35x 4 - 30X2 + 3), (158)
P,(x) = i(63x' - 70x 3 -1- 15x).
For later reference, it is noted that the functions u.ex), with neven, and
v.ex), with n odd, can be shown to have the following values at x = 1 :
uo(l) = 1, u (1) = (- 1)n2 2·4·6· .. n (n = 2, 4, 6, ... ),

n 1.3.5 ... (n-l)
( 159a)
v¡(I) = 1, v.(l) = (_1)'n-¡)/2 2·4·6· ._. (n - 1) (n = 3,5,7, ... ).

1·3·) , ··n
(l59b)
When p is an even integer n, the solution vn(x) is in the form of an infinite
series, whereas if p is an odd integer n, the solution u.(x) is an infinite series.
Suitable multiples of these solutions are called Legendre functions of the second
kind and are denoted by On(x). It is conventional to take the multiplicative
e
factors as (-l)nun(l) and _l)nvnO), respectively, leading to the definition
Q.(x) = {-Vn(I)Un(X) (n odd), (160)
u.(l)vn(x) (n even),
where the constants UnO), neven, and vn(l), n odd, are defined by (l59a, b).
However, since the series appearing in (160) converge only when Ixl < 1, the
functions Qn(x) are defined by (160) only inside this interval.
Thus, when p is an integer n, a certain multiple of thar solurion (155) or
(156) which is a polynomial is written as Pn(x), and a certain multiple of the
other (infinite series) solution is written as Q.(x), so that the general solution
of (149) in this case is written in the forrn
y = c¡P.(x) +c 2 Qn(x). (161)
It can be shown (see Problem 64) that Pn(x) and Qn(x) both satisfy the
recurrence formllla
(162)
Equation (162) permits the determination of expressions for Legendre func-
tions in terms of corresponding functions of lower degree.
We next express Qo(x) and Ql(X) in cJosed form, by using the methods of
Section 1.10. Reference to Problem 42 of Chapter 1 shows that the functions
Q.(x) are expressible in the form
Q.(x) = AnPn(x) J (1 _ X~)[Pn(x)J2 + BnPn(x),

where An and Bn are suitably chosen constants. In particular, since Po(x) = 1,
there follows
or, iflxl < 1,

Ao I + x
Q o(x) = T10g 1 _ x + Bo·
4.12. Legendre Functions 159
From (160) we oblain Qo(O) = lI o(l )'u o(O) = O and Q'o(O) = lI o(l)v'o (O) = 1.
Hence there follows A o = 1, Bo = O, and we have the result
_ o()
O X = 21 1og 11 +
_ xx = t an h - ' x. ( 163)
Similarly, when PI = 1, since P,(x) = x there foIlows
Q,(x) . A,x f x2(ld~ X2) + B,x

or, if Ixl < 1,
Q,(x) = A¡ (~ log : ~~ - 1) + B¡x.

From (160) we have
Q¡(O) = -v¡{l)u¡(O) = -1 and Q',(O) = -v,(I)u',(O) = O.
Hence we must take A ¡ = l and B, = O, and so obtain
x
Q¡(x) = 210g l1 +
_ x
x - 1 = xQo(x ) - 1. (164)
The series expansions of Qo(x) and Q¡(x), in powers of x, are readily shown to
be in agreement with the series indicated in (160).
Use of the recurrence formula (162) now permits the determination of
Qn(x) for any positive integral value of n. In this way one obtains, in particular,
the expressions
3
Q 2(X) = P 2(x)Qo(x) - 2x,
5
QJ(x) = PJ(x)Q o(x) - 2X2 + ]2 '
(165)
35 55
Q.(x) = p.(x)Qo(x) - 8 x
J
+ 24x,
63
Q,(x) = P,(x)Qo(x) - 8 x' + 849 X2 -
8
13·
For values of x such that Ixl > 1, the integral J dxj(l - X2) takes the form
2I 1og xx + I
_ 1+ e = coth-¡ x + C.
Thus, if Ixl > 1, the function
Qo(x) = 2l 10g xx _+ 11 = coth-¡ x (166)
is a solution of (149) which complements the polynomial solution Po(x) = 1.

Corresponding solutions Qn(x) for integral values of n, in the range I xl> 1, are
obtained by using the notation of(166) [in place of(163)] in Equations (164) and
( (65).
If p is not an integer, a certain combination of the series (155) and (156)
can be determined so as to remain finite and take on the value unity at x = l
(see Problem 65). This funcrlOn is called P p(x), and a second independent com-
binarion is denoted by Qp(x), so that the general solurion of (149) in the general
case is wrirten in the form
(167)
The function Pp(x) so defined, however, will not also remain finite at the
point x = -1 unless p is integral, and the function Qp(x) cannot be finite at
x = l. Thus the only Legendre fimctions which are finite at both x = 1 and
x = -1 are the Legendre polynomials Pp(x), for which p is integral. (This
fact will be of importance in Section 5.14.)
Rodrigues' formula, which expresses Pn(x) in the alternative form
n 2
P (x) = _1_ d (x - I)n, (168)
n 2nn! dxn
is particularly useful in dealing with certain integrals involving Legendre poly-
nomials. Proof that (168) is indeed consistent with (157) for all positive integral
values of n is omitted here (see Problem 60), but it is readily verified that (168)
reduces, in the special cases of (158), to the forms given.
From (168) it can be deduced that all the zeros of P'(x) are real and un-
repeated, and lie in the interval -1 < x < 1 (see Problem 61). Another important
fact (Problem 62) is that, in the interval -1 <: x <: 1, the magnitude of each
Legendre polynomial is maximum at the end points, so that
IPnCx) I <: 1 when Ixl <: 1,
Outside the interval (-1, 1), each polynomial Pn(x) increases or decreases
steadily, without maxima or minima or turning points (see Figure 4.2).
Figure 4.2
4,12, Legendre Functions 161
If Ixl < 1, the substitution x = cos rp transforms Legendre's equation from

the form (149a) into the form
- .- 1 , rp -d y )
-d ( SIn + n(n ,,- l)y = O ( 169)
SIn rp drp drp
or, equi va le ntly,

d'y dv
drp' + d~ cot rp + n(n + l)y = O, (l69a)
when p = n, and hen ce (169) ha s the generaL solution

y = c , Pn(cosrp) + c,Q n(cosrp) , (170)
Equations of sllch a form frequently arise when spherical coordinates are
introduced in the solution of a pote ntial problem with rotational sy mmetry, We
note that Qn (cos rp) is not finite when cos rp = ± 1, that is, when rp = kn ,
whereas P n (cos rp) is merely a polynomial of degree n in cos rp, In particular,
we have the expressions
Po(cos rp) . ~ 1, P,(cos rp) = cos rp,
P,(CO S rp) = 1(3 cos' rp - 1) = teJ cos 2rp + 1), (171)
Pl(COS rp) = t(S cos' rp - 3 cos rp) = t(5 cos 3rp + 3 cos rp),
When Ixl < 1, the functions
(172)
are called the associaled Legendre fl/nclions of degree n and order m, of the firsl
and second kinds, respectively, They can be shown to satisfy the differential
equation
2
2 d'y dy [ m
(1 - x ) dx2 - 2x dx + n(n + 1) - I _ x' y -
] _
O, ( 173)
which differs from (149) only in the presence of the term involving m, and
properly reduces to (149) when m = O, When Ixl > 1, the definitions (172) are
modified by replacing 1 - x' by X2 - 1.
For any nonnegative integral vall/e of ni, lh e eq ualion (17 J) possesses a non-
trivial solutiofl which isfinire ar bor/¡ x = 1 and x = -1 only when n is an ¡nteger,
and thal sollltion is a mU/liple of P:;'(x ), wilh n :> m.
The substitution x = cos rp tran sform s (173) into the equation
d 2y dy
drp' + drp cot rp + [n(n + 1) - m 2 ese' rpJy = O, (174)
which thus is satisfied by P:;'(cos rp) and Q;:O(cos rp). These fun ctio ns are often
useful in the solution of potential problems involving spherical boundaries, in
the absence of rotational syrnmetry.
162 Series Solutions of Differential EQuations: Special Functions
4.13. The Hypergeometric Function. Solutions of the differential equation

d 2y dy
x(1 - x) dX 2 + [y - (a + f3 ..L I)x] dx - af3y = O (175)
are generally called hypergeomelric functions , since their series representalions

are, in a sense, generalizations of tbe elementary geometric series. Since (175)
is of type (52), with IV! = 1, the series (54) reduces here lO the usual form
(176)
The exponents are found to be

s = O, I - y, (177)
so that only one solution of the assumed form can be expected when y is integral.
Tbe recurrence formula is obtained in the form
(s + k)(s + k + Y- I)B k = (s +k+ a - I)(s +k+ 13 - I)B k _¡ (k >- 1),
(178)
B (s) - [(s + + a + 1) .. . (s + a + k - 1)]
a)(s
k - + I)(s + 2) ... (s + k)]
[(s
[(s + f3)(s + 13 + 1)·· · (s + 13 + k - l)]B (k >- 1). (179)
x [(s + y)(s + y + 1) ... (s + y + k - 1)] a
Corresponding to the exponent s = 0, we thus obtain the solutíon
y = Ba {I + f; laCa + + k - 1)] [13(13 + 1) ... (13 + k -

l) ... (a I)]xk}.
k- l [1·2···kJ[y(y+I)···(y+k-I)]
(180)
Tbe coefficient of Ba in (180) is written as F(a, f3; y; x), and the series is known
as the hypergeometric series or funclion,
F(a,f3;y;x) = 1 +~x + a(a + 1)13(13+ l)x 2 + .... (181)
I·y 1·2·y(y + J)
It is found tbat tbe series (181) converges in the interval Ixl < 1 and also tbat,
wben x = + 1, tbe series converges on1y if Y - a - 13 > O and, when x = -1,
the series converges on1y if Y - a - 13 +1> O.
It may be noticed that if a = 1 and 13 = y, the series becomes the elemen-
tary geometric series
F(l, 13 ; 13; x) = 1 +x+ X2 + ... + x· + ... = 7'1_1---

-x
(Ixl < 1). (182)
lt is seen also tbat, because of the symmetry in a and 13, these parameters are
in terchangea ble,
F(a, 13 ; y; x) = F(f3, a; y ; x) . (183)
The so/ulion (181) does not exist (in general) when y is zero al' a negative integer.
4.14. Series Solutions Valid for Large Values of x 163
Corresponding to the exponent S = 1 - y. we obtain the solution

+ f; [(a - y + 1)(a - y + 2) ... (a - y + k)]
k=' [1·2···k]
x [(p - y + I)(P - y + 2) ... (P - y + k)] xk} . (184)
[(2 - y)(3 - y) ... (k + 1 - y)]
The series in braces in (184) is seen to differ from that in (180) only in that a, p,
and y in (180) are replaced by (a - y + 1), (P - y + 1), and (2 - y), respec-
tively, in (184). Hence (184) can be written in the form
y = Box'-'F(a - y + 1, P- y + 1; 2 - y; x). (185)
The so/utÍon (185) does not exist (in general) when y is a positive integer greater
than IInity. When y = J, the solution (185) becomes identical with (181).
Thus, if y is not zero or an integer, the general solution of (175) can be
expressed in the form
y = c,F(a, p;y;x) + c 2 x'-'F(a - Y + I,p - y + 1;2 - y;x), (186)
when Ixl < l. The exceptional cases can be treated by the methods of Section
4.5.
Many elementary functions are expressible in terms of the hypergeometric
fUllction (181), including the following exaroples:
(1 - x)-a = F(a, p; p; x),
(l + X)-k + (1 - X)-k = 2F(;, k ~ 1 ;f;x 2

),
(1 +,y"l - X)-k = 2- kF(; ,k ~ l;k + 1 ;x),
log (1 - x) = xF(I, 1 ; 2; x),
Iog 11 + x _?
_ x - -x
F(2'1 l', 23 '. x 2) .
4.14. Series SolutioDs Valid for Large Values of x. In the preceding
sections we have considered series solutions valid in an interval centered at
the point x = O, and have noticed that if solutions va lid near a point x = X o
were desired, such solutions could be conveniently obtained by first replacing
x - X o by a new independent variable t and then seeking series solutions of
the forro
I; Ant n+, = I; An(x - xo)n+.
froro the new equation. In such cases, the point x = X o naturally should be not
worse than a regular singular point.
Thus, if series solutions of Bessel's equation of order zero in powers of x - 1
were required, we could set 1 = X - 1 and thus transforro that equation to the
form
d 2y dy
(1 + 1) dl 2 + dI + (1 + I)Y = O.
164 Series Solutions of Differential Equations ; Special Functions
r
Since the point I = O is an ordinary point, two solut io ns oC [he form 2: Antn can
be obtained and rewri tten finally in the desired form 2: A n(x - l)n.
In order to investigate the behavior of sollltions for large values of x, we
are Ied to the possibility of replaeing l/x by a new independent variable 1, and
then of studying the behavior of solutions of the new equation for smal! values
of 1, sinee 1 - 0 as Ixl-' oo. If the new equation has the point 1=0 as an
O/'dinar)" point, we obtain two sollltions of the form I: Ap = I: Anx- n, whereas
for a regular singular point at least one solution I: A"l n+ s = I: Anx- n-, is
obtained .
With the substitution x = I / t, the equation
d'y dy
dx
.2 + a,(x) -d
x
+ a,(x)y = O (187)
2
beeomes --
d y + -I
dl 2
t 2
[ 21 - a ( - I )] --
'1
dy
dt
+ I4
-a,
1
( - I ) y = O.
- t
(188)
If the point 1=0 i s an ordinary point (or a singular point) of (188), it is

eonventional to say that " the point x = co is an ordinary point (or a singular
point) of (187)." The u se of sue h a phrase is motivated by the faet that, if t = O
is an ordinary point (or regular si ngular point) of (188), then (187) possesses
solutions of the form I: A"x- n (o r I: A"x-"-').
Equation (188) shows th at in order that x = co be an ordinary poinl of
(187) the funetions
_1 a 2
[4
(-I)[
(189)
mu st be regular at t = O, whereas In order that x = 0:0 be a regular singular

point of (187) the fllnetions
_1 a
t '
(_1)
[
and
12
I a ( I)
2 t
( 190)
mu st be regular at I = O.
To illustrate, we notiee that, if a, and a 2 are constanls in (1 $7), the point
x = 00 is an irregular singular point unle ss a, = a 2 = O, si nce the functions
aJ / I a nd a 2 / [' are not regular at I = O.
Bessel's equalion (63) also has un irregular s ingular point at x = 00 , Slnee
the function
/, a,(+) = 1; - p2
is not regular at t = O. Thus Be ssel' s equation has x =~ O as a regul a r singular
point and x = CQ as an irregular singula r point. AIl other points are orcJinary.
Legendre's equalion (149), however, has a regular singular point at x = 00,
slnee
_1 a (_1 ) = p(p
[2 2 I 12 _
+ I1)
4.1-1. Series Solutions Valid for Large Yalues of x 165
are regular at 1 = O, whereas the function s (189) a re nOl regul ar at 1 = O. Thu s

it is seen that Legendre 's equation has regular singu lar point s a t .\' = 1 and
at x = co and ordinary paint, el,ewhere.
The expansion 01' º"(x) in inverse powers of x when Ix l > 1 can be exp ressed in
terms of a hypergeometric series in ¡he form
Q(x)= n!,Jn F ( ~ ~.1I + 3.J...).
11 (n+-t>!(2xyn+1 2' 2' 2 'x~
For the hypergeomelric equaliol1 (175) we find that x = 15 an ordinary

poinl only if either a = O, j3 = I or a = 1, j3 = O, and is a regular s ing ular
poinl otherwise. Thus, except for the cases noted, the hypergeometri c ee¡ uation
has regular singular poi,lts at :c = O, x = 1, and x = 00 and ordinary point s
e[sewhere. The general solution when Ix. > 1 is of the form
y = e 1 x -, F (a, a - Y '1 1; a --' j3 + I ; ,~ )
+ c,x- P F (j3. j3 - y + 1 ; j3 - a + 1; ,~) ,
provided that j3 - a is nonintegral. (See Problem 76.)

If, in the hypergeometric eqLlation ([ 75), we replace:c by a new independent
variable xl j3, we obtain the equation
x (1 -
X)d2V
7f d,~! +
[ Y - x - (1 a)~Jdy - ay = O, (19 1)
ji d:c
which h as regular sing ular points at x = O, x = j3, and x = 00 , and whose
general so [ution , when 1:(1 < 1, is given by ([86) with x rep[aced by x/j3.1f now
we let j3 ~ 00 , Eguatio n ([ 9 1) formally beco mes
dOy , ,_ dy _ . =
X dx! -r ( y x) d.L ay O. (192)
In the transition from ([91) to (192), we have moved the singular point at
x = j3 i n (191) into coincidence o r "confluence" with [he second singular point
a t ,'1: = oo. For thi s reason, (192) is known as the confluem hypergeol11elric
eqlloliol1 . Thi s equa[ion is a s pecial case of (52), as was pointed out at the end
of Section 4.7 , a nd is of so me importance in applications. It is of intcrest to
notice that x = O~ is an irregular s ingular point of ([92), formed by the conflu-
ence of two originally distinct regular s ing ular points . (See al so Problem 14.)
In cases where x = ce is an irregular s ingular point , it is frequentl y poss ible
to obtain series of the type
. A
.
v( x ) - e'X L: _ k ,
k = O X k +s
( 193)
s uch that/orl11al s ubs tituti o n of the series iM o th e diffe re ntial equation reduce s
the equation to an identity. Ho wever, the se rie s so obta ined generally do nol
cO/1l'crge lo/' al/y jinile I'alues o/ x Still, they ge nerally do have the property
th'at if a finite num bcr, N , of term s is ret a ined, the s um , SAx), of these terms
approximates a solution y(x) in such a way that not only the difference
y(x) - SN(X) but also the product x,vlv(x) - S,v(x)] approaches zero as 1x 1- ' oo.
Such series, called asymplolic expansions of a solution y(x), are of use not ol1ly
in studying the na/l/re of solutions for large values of 1 x 1 but also in actually
calculnling values of such solutions to within a predictable accuracy. The reason
is that in such cases the error associated with calculating y(x) by using N terms is
of the order of magnitude of the next following (neg1ected) term, prol'ided that
Ihot lerm is numerically smaller than (he lasl lerm retained. [This, of course, is
not a general property of all divergent (or convergent) series.] Although this
error eventually increases without limit as N increases, che first few successive
term s frequently decrease rapidly in magnitude when x is large, so that it may
be possible to stop with a term preceding a term of the order of magnitude of the
tolera ble cerror. t
The asymptotic approximations given in Equations (lO5-107) for the
Besse l functions represent in each case the leading term of asymptotic expan-
sions of this sort , which may be 1isted as follows:
Jk<) ~ Jn 2
x [U/x) cos (x - ~ - p;) - Vp(x) sin (x - ~ - p;)}
(194)
/2
Y/x)~-Vnx . ( x - n - pn)
[ U/X)SlD + v p(x) cos (x - ~ - P;) }
4 2
(195)
(196)
H p(2 )(x) ~
V/ 7tX
2 [U
p
(x) _ iV (x)]e-ilx- ' •• , -
P
( p.: 2>1, ( 197)
(198)
( 199)
where U p , V" and W p denote the respective asymptotic series

(4p2 - J2)(4p 2 - 32 )
U/x) = 1 - 2 ! (8x)'
(4p2 12)(4p2 _ 3 2 )(4p2 _ 5 2 )(4p2 _ 7 2 ) _
+ 4! (8X)4
(200)
_ 4p2 - ]2 (4p2 - !2)(4p2 - 3 2)(4p2 _ 52)

V/x) - I! 8x - 3 ! (8x)3 + (201)
tFor further information on asymptotic expansions, see Refe re nces 4 and 9.

References 167
Wix) = Uiix) - iVp(ix)

_ 4p2 - 12 I (4p2 - J2)(4p2 - 32 )
1 J ! 8x'- 2! (8X)2 (202)
For bOlh U)x) and Vix) it is known (Reference 17) that the error committed
when the expansion is terminated with the kth term is not greater in magnitude
than the absolute value of the (k + I)th term, provided that k > (2p - 1)/4.
The same statement applies to ~Vp(x) if k > (2p - 1)/2. These expansions are
frequently useful in the numerical evaluation of problem solutions involving
Bessel functions of large argument, when the accuracy afforded by the Jeading
term is insufficient or su bject to question.
Asymptotic series, as defined aboye, are sometimes also called semiconvergent
series.
REFERENCES
1. References at end of Chapter l.

2. DAVJS, H. T., Tables 01 Higher Mathematical FlInctions, Principia Press, Blooming-
ton, Ind., 1960.
3. DWIGHT, H. B., Tables ollntegrals and Other Mathematicul Data, 3rd ed., Dover
Publications, Inc., New York, 1958.
4. ERDELYI, A., Asymptotic Expansions, Dover Publications, Inc., New York, 1956.
5. FLETCHER, A., J. C. P. MILLER, L. ROSENHEAD, and L. J. COMRrE, An lndex 01
Mathematical Tables l. 2nd ed., Addison-Wesley Publishing Company, Inc., Read-
ing, Mass., 1962.
6. GRAY, A., G. R. MATHEWS, and T. M. MACRoBERT, A Treatise on Bessel FlInctions,
2nd ed., SI. Martin's Press, Inc., New York, 1952.
7. HOBSON, E. W., Theory 01 Spherical and Ellipsoidal Harmonics, Chelsea Publishing
Company, Inc., New York, 1955.
8. JAHNKE, E., F. EMDE, and F. LOSCH, Tables 01 Higher Functions, McGraw-Hill
Book Company, Inc., New York, 1960.
9. JEFFREYS, B., and B. S. JEFFREYS, Methods 01 Mathematical Physics, Cambridge
University Press, New York, 1956.
10. KNoPp, K., lnfinite Seqllences and Series, Dover Publications, Inc., New York,
1956.
11. MACRoBERT, T. M., Spherical Harmonics, 2nd rev. ed., Dover Publications, Inc.,
New York, 1948.
12. MAGNUS, W., and F. OBERHETTINGER, Special FlInctions 01 N[athematical Physics,
Chelsea Publishing Company, New York, 1949.
13. McLACHLAN, N. W., Bessel Functions lor Engineers, 2nd ed., Oxford University
Press, Inc., New York, 1955.
14. RAINVILLE, E. D., Special FUllctions, Macmillan Publishing Co., Inc., New York,
1960.
15 . SZEG O, O., Orthogonal Polyncmials, 3rd ed., Ameriean Mathematieal Society,

Providenee, R . I., 1967.
16. WATSON, G. N ., A Trea/ise on/he Theory 01 Bessel Func/ions, 2nd ed., Macmillan
Publishing Co ., Inc., New York, 1945.
17 . WHITTAKER , E . T., and G. N. WATSON, /vfodern Analysis, 4th ed., Cambridge
Univer sity Pre ss, New York, 1958.
PROBLEMS
Section 4.1
1. Determine the interval of eonvergence for each of the following series, ineluding
consideration of the behavior of the series at eaeh end poinl:
(a) ~ 2 x ,
11=0
n n
n!
(b) i: (_J)nn(x;:- J ) n,
n= I -
(e) ~ ( _I )n(x -, 1) 2n , (d) i: k(k + 1) ... ~k + n - 1) xn,

"=1 3n n-l n.
(e) ( f) ~ n'x',
n "= 1
~ (n - c.:) ! ( _ )n - (n !)2 n
(g) ~ "x
,, "" I n . IX .
(7, (h) ,~o (2n) !x ,
(i) -
~-,
,, -= I
1
n .."'("
(j) i:
n =O
(~)n.
X + 2
2. Slarting with the e xpansions
_.
¡ .(x ) - _ x' , x'
smx-x-TI'5! g(x) = eos x = J -
X2
2! +
x4
4' -
and assurning Ihat

xJ 2x'
tan x = x + 3' + 'T5 +
verify through three nonvani s hing lerm s tha t term-by-term operations on the series
properly yield the following results:
(a) f'(x) = g(x), ( b) fax g(u) elu = f(x),
(e) f(x)g(x) = ±f(2x) , (d) :~~~; = tan x (Ixl < ~).

3. Show that
o·,
+ TI + 2! + .. . + "n! + (n+I)!X n+ ,
X x2 eS' X
(a) e' = J I
(b) eos x
X2 x4
= J - 2! + 4 I - + (-I)m x2m + (_ r)m+ I eos «()2 X ) X2m+2
(2m)! (2m + 2) l '
where O < (), 2 < 1.

Problems 169
4. Prove that the remainder term in eaeh of the represe ntations in Problem 3 tends to
zero as the number of terms increases, for all values of x. (See Problem 53 of Chapter
2.)
Seclion 4.2
5. Obtain the general solution of eaeh of the following differential equations in terms
of MacJaurin series:
dZy d 2y dy
(a) dX2 = xy, ( b) d x 2 +x d-x - y = O,
( e) xd2y _ dy _ 4 x 3 y = o.
dxZ dx
6. For ea eh of the following differential equations , obtain tbe most general solution
whieh is representable by a MacJaurin se ries :
(a) ~;: + y = 0, (b~ ~;~ - (x - 3)y = 0,
I
(e) I 1 -
1 ) d 2y
-2 x 2 dX
- X
dy
+ ax - Y = °' d 2y
(d) X 2 -dX2 -
dy
- +y = °
\ 2 dx '
d 2y dy
(e) (X2 + x) dX2 - (x 2 - 2) dx - (x + 2)y = 0,
(f) X dX2
2
2d y _ dy _
dx -
° .
Obtain three nonvanishing terms in each infinite series involved.
Section 4.3
7. (a-f) Loeate and elassify the singular points of the differential equations of Prob-
lem 6.
8. Loeate and elassify the singular points of the following differential equations:
d 2y -
(a) (x- j)dx 2 +'¡xy =O (x > 0),
d 2y
(b) d x 2
dy
+ d-
x log x + xy = ° (x > O),
d 2y .
(e) X dx 2 + Y S10 X = 0,
d 2
(d ) ---.X.
dX 2 -11 - x 2 1Y = °'
(e) dd2~
x
+ yeos'¡x = ° (x > O).
9. Show th at the change of variables x = / 2, with / = "/x > 0 , transforms the

equation
d2y -
dx 2 + ·" / XY =0 (x > O),
with an irregular singular point at x = 0, into the equation

,d2y_dy , 414y=0 (t > O)
dl 2 dI""
with a regular si ngular point at I = O.
170 Series Solutions of Oilferential Equations: Special Functions
Section 4.4
10. Deri ve Equation (22) by usi ng summation notation, as follows:

(a) Introduce lhe expansions
' .0
P(x) = ~ Pnx n , Q( x ) - ~ Qnx n, R(x)

n~ o n= Q
"
a nd y(x ) = 2: Amx m+ "
m - O
with Ro = 1, into the expression for Ly and obtain

=
Ly = ~
n ~ o
~
m= o
[Rn(s + m ) (s + m - 1) + Pn(s + m) + QnlAmxn+ m+' -2.
(b) Write n + m k, using n and k a s new s ummation indices and noticing that
then n may vary from
s ien for Ly into
°
=
through k as k varies from Oto = . Thus transform the expres-
where g n(s ) is defined by (21) when n = 1,2, ... and go(s ) = fes), and verify that this
result is equivalent to (22).
11. Use the method of Frobenills to obtain the general solution of each of the follow-
ing differential equations, valid near x = O:
d 2y dy
(a) 2x dX2 + (1 - 2x) dx - y = 0,
dy
1- (b) X1d2y1
dx
+ x dx + (X2 _l...)y
4
= 0,
,
I d 2y
l .,.. (e) x - d
X·
o + 2 dy
x +
-d xy = 0,
d1y dy
(d) x(1 - x) dX2 - 2 dx + 2y = O.
12. Use the method of Frobenius to obtain the general solution of eaeh ofthe following
differential equations, val id near x = O:
d 2y
(a) X2 -
dX2
- 2x -
dy
dx
(2 - X2)y =
'
+ °
2
d y dy
(b) (x - 1) dX2 - x dx + y = 0,
d 2y dy 3_
(e) X dx 2 - dx + 4x y - O,
d 2y dy
(d) (1 - cos x) dX2 - dx sin x + y = O.
Obtain three nonvanishiog terms in eaeh infinite senes of parts (a-e). 10 part (d),
obtaio two such terms io each series.
13. (a-d) For eaeh of the equations in Problem 12, gíve the largest interval inside
which eonvergeoee of the Frobeoius solutioos is guaranteed (exeept possibly at x = O)
by the theorem stated 00 p il ge 131.
Problerns 171
14. Find the general solution of the differer.tial equation

2
x d y ..L (e _ x) dy - ay = O
dx2. I dx '
valid near x = O, assuming that e is nonintegral. The solution which is regular at

x = O and which is unity at that point is callee! the eon/fuen/ hypergeomelrie ¡une/ion
and is usually denoted by M(a, e; x). Show then that, if e is nonintegral , the general
solution is of rhe form
y = clll/f(a, c;x) + czxl-'M(l + a - e,2 - c;x).
15. Determine the two values of the constant IX for which it is true that al! solutions
of the equation
d 2y dy
x -
dX2 + (x - 1) -
dx - lXy = O
are regular at x = 0, and obtain the general so lution in each of these cases.
16. (a) Show that the equation
dZ d
x~+~-y= O
dxz dx
possesses equal exponents SI = S2 = O at x = O.
(b) Obtain the regular so lution, and denote by u, (x) the res ult ofsetting the lead-
ing eoeffieient Ao equal to unity.
(e) Assume a seeond sol ution o f the form
yz(x) = ClI 1(x) log x + v(x),
where C =F 0, and show that v(x) must satisfy the equation
x d2V
dxz
+ dv _ v _ _ ?C du ,.
dx - - dx
(d) Obtain one solution of this equation in the form

= ~
v(x) = ~ Bkx"''' = ~ Bk x k ,
k=O k ~ O
showing that C and B o are arbitrary, but taking C = 1 and Bo = for eOllvenience.
Hence obtain the general solution of the original equation in the form
°
y = clu¡(x) + cz[u,(x) log x + v(x) ].
xdx'
d'y
- - - xdx
- -y
dy
= °
possesses exponents s, = 1 and s, = O at x = O.
(b) Obtain the regular solution, corresponding to SI - 1, a nd denote by lI,(X)
the result of setting the leading coefficient equal to unity.
(e) Assurne a second solution in the form
y,(x) = Cu,(x) log x + v(x),
172 Series Soiutions 01' Differemial Equarjons: Special Functions
where e* 0, a nd show that ¡'(x) mus! satisfy th e equati on

d',· d,) e dll,~
x dx' - x dx - l' ~ X (1 -'- X)lI, - :Zx d.\" .:
(el) Obtain one solution of this equation in the form
showing that e and 8, are arbitrary, but taking e = 1 and E, =

Hence ob tai n rhe general s olution of rhe original equation in rhe form
for co n ve nience. °
y = "lI,(X) + ,,[lI,(X) log x + l'ix) j.
18. Nonho m offeneolls linear equolion5.

(a) Show that a particular so lurion of rhe nonhomogeneous linear difrerential
e qua t ion
d'y dy
x'( l + R, x ... . . ' ) -d '
x
+ x( P o + P,x + .. ')-d
x
+ (Qo + Q, x + .. . )y
= x'(do + d,x + ... ),
where r is any cons tao t, and w here a l! se ries converge in so rne interval about x =
o r terminate . can be obtained in the form
°
y = xr(Ao + Alx + ... ) = ~ A kx k + r ,
k=O
if neither of the exponent s 5, a nd s" for which

s' + (Po - 1)5 + Qo = 0,
equal s r or exceeds r by a positive intege r .
(b) For an equation of the fo rm
l· (1 +o,x + "')dx'
dZy
+ (b o + b , x+ " ' )dx
dy
+ (ca c,x + ... )y
= do + d,x + " ',
w hic h hen ce possesses an ordinary point at x = and a regular right-hand member,
deduce the existence of a particular so lution of the form
°
y = 1{X) = x'(A o + A,x ... ) = ~
"
k = O
A k x k + 2•
[If the equation is w ritt en in the form y " + o¡y' + U2.l' = x'-'d, the series solu-
tion s wi ll converge at leas t imide the larges r interval in which the senes representing
the result a nt function s xO " X202, an d d would 0 11 converge. In the exeeptional cases
noted in part (a), Joga rithmic terms may be in vo lved .]
19. Determine a particular so luti on of eac.h of (he fo llowin g equations, in (he forro
of a series valid near x = O, by lhe method of Problem 18. In each case, obtain four
nonvanishing terms .
d'y d 2y
(a) dx' + y = e" ( b) dx' + xy = 1,
d'y d2 eX
(e) x dx' .- y - .\", (d) x'--...X+y= ~ ·
dx' "';x
Problems 173
Sections 4.5 and 4.6
20. Obtain ¡he general soluti on of the equati on

xd'y'+dY_y=O
dx' dx
by use of the formula (36).
21. Obtain the general solution of the equation
d 2y dy
xd--,-x--y=O
x- dx
by use of the formula (37).
22. Obtain the general solution of the equation of Problem 14 when a = c = 1 by use
of either of the two methods of Seetion 4.5.
23. Obtain the general solution of Problem 14 w hen a = 1 and c = O by use of either
of the methods of Seetion 4. 5.
Section 4.7
d 2y dy
x -xd2 + (l - x) -d
x + ny = O
is known as Laguerre's equation.

(a) Verify !ha! !his equation is a speeial case of Equati on (52), with M = 1, and
show that the exponents a! x = O are both zero.
(b) Obtain the regular solu!ion in the form
_ B [1
Y 1( X ) - o
+ ~ (_l)k n (11 -
~
1)(11 - 2)··· (n -
(k .')'
k + 1) k].
x
k=t
(e) Show tha! this solution is a polynomial of degree 11 when n is a nonnegative

integer, and verify tha! the choice Bo = 1 leads to the Laguerre polynomial of degree 11,
with the definition
Lb) = 1 - (nt, + (~)~~ - ... + (~?,

where (Z) represe nts the binomial coefficient n! ¡ [en - k)! k !J.
d 2y dy
-xd2 - 2x d-
x + 211y = O
is known as Hermite's equatiol1 . Verify tha! this equation is a special case of Equation
(52), with M = 2, and ob!ain the general so)ution in the form
y(x) = c ¡u"(x) + C2Vn(X),
where
() 1
- .!!...x 2 + n(n - 2)x' _ /l(n - 2)(11 - 4) x· +
Un X - 1 1! 1 ·3 2! I ·3·5 3!
174 Series Solutions of Differential Equations: Special Funclioos
and
_ _ 71 - 1 x 3 , (n - J)(n - 3) x' _ (n - l)(n - 3)(n - 5) x' '.
v" ( x ) - x 3 1!" 3.5 2! 3. 5.7 3 ! "l
[Henee verify that the solution unIx) is a polynomial of degree n when n is a positive
even integer or zero, whereas vn(x) is a polynornial of degree n when n is a positive
odd integer. That multiple of the nth-degree polynomial for whieh the eoeffieient of
x n is 2 n is ealled the nth Hermile polynomial and is often denoted by Hn (x). ]
, d 2y dy
(1 - x-) dx2 - ex dx + n(n + e - l)y = O
is a speeialization of Jacobi's equation (see Problem 68), referred to the interval ( -1,1 ).
(a) Show that if n is zero or a positive integer it possesses a polynomial solution,
of the form
""(X) = 1 _ n(n +2~ - l )x2 + [n(n - 2)][(n + ~! 1)(n + e + 1)]x' _
when n is even , and of the form

(n - l)(n + 2) 3 , [en - l)(n - 3)][(n + eXn + e + 2)] 5
V n( X ) = X - 3! x "T 5~ x
when n is odd.
(b) When e = l, that multiple of the nth-degree pol ynomial for which the coef-
ficient of x n is 2"- I (n > 1) is called the nth Chebyshev polynomial, and is often denoted
by Tn( x). lt is conventional to take To(x) = l. Obtain the results
T o(x) = 1, TI (x) = x, T 2 (x) = 2x 2 - 1, T 3(x) = 4x 3 - 3x
and verify in these cases that
Tn(x) = cos ( n cos-, x).
(e) When e = 3, that multip1e of the nth-degree polynomial for which the eoef-
fieient of x" is 2" is caBed the nth Chebyshev polynomial 01 ¡he second kind, and is often
denoted by Sn(x), Obtain the res ults
So(x) = 1, S, (x) = 2x, S2(X) = 4X2 - 1, S 3(X) = 8x 3 - 4x
and verify in these cases that
S( )=_I_T' ()=sin[(n+ I)eos-Ix],
"x n+l n+'X ../1-x2
given that T.(x) = 8x' - 8X2 + 1.
(When e = 2, the L egendre polynomials are obtained. They are eonsidered in sorne
detail in Seetion 4.12.)
Section 4.8
27. Evaluate the following quantities, from the series definitions, to three-plaee ae-
euraey:
(a) J,(ü.3), (b) Y o(0.2), (e) JQ.7 ,(0.2),
(d) 1 2 (1), (e) H~')(ü.2), (f) J'I (0.5).
Problems 175
28. Find rhe general solution of [he simultaneous equations

dx = O.
y -r ldi
dy
dI -/x=O.
29. By m a king an appropriate change of variables, obtain the general solution of the
differential equation
2
(Ax + B) d y,
dx"
+ A dy
dx
+ A'(Ax + E)}" = O.
Section 4.9
30. Use E q uations (10 1-1 07) to evaluate the following lim its:
(a ~ ' Iim xl. JJ _ , ' J(x), ( b) lim x Y, ( x),
, -o .x . · 0
(e) lim x 2K, ( x ), (d) lim Jn(x) ,

x -o )c .o xn
(e) lim x[ [J p(x)J2
)e _ O'"
+ [Yp(x)J2 J.
31. Show that the definitions (85) and (99) y ield the special relations
Y, 2( X) = - J_ ' . 2( X) = - / 2 eos x
-V 7rx
and
Kl.2 (.<) = ~ [I -1. 2(X ) - ' l.2(x)] = J":xe-x
and verify that these result s are cons istenl with (88) and (100) .
32. Prove from the series definition that
33. Use Equations (112) and (113) to prove that, if e p(x) = [J p(x)j2, then
de/x) = ; [e p_, (x) - e p+ 1(x)].

X _p
34. Use Equations (108) and (109) to deduce (by induction) the formulas
(-+ ::X)"' [xPJp(rxx)] = rxmxp-mJp_m(rxx)
and
(~ :xr[X-PJp(rxX)] = (-IX)mX-p - noJp+m(IXX).
35. Use Equations (112) and (113) to show that if J.cm = O and J~(p) = O, so that
p is a zero of J p(x) of multiplicity a t least two, then J p +, ( p) = J p - l (p) and hence,
from (110) and (111), J ~+, (P) = J~_, (p) . Then deduce that a lso J~(Pl = O, so that
the multiplieity of P is at least three. Noticing that this argumeot can be repeated
indefinitely , prove tha! ¡he zeros o/ J p(x) are all simple; tha! is, if JiP) = O, theo
J~(p) *
O
36. Use Eguatioo (111) and lhe result of Problem 35 to show that J p (x) has opposite +,
signs al IWO eonseeutive zeros of Jp(x), so that J p+, (x) has at least one zero between
suecessive zeros o[ J.(x). Then use (110) with p replaeed by p + 1, similarly, to show
that Jp(x) has at least one zero between sueeessive zeros of J p+, (x). Henee prove that
Ihere is exaclly one zero 01 J.(x) belween successive zeros 01 J p +, (x).
37. Pro ve that 1.(0) = O when p > O, but that loex) has no real zeros and l.(x) has
no real zeros other than x = O when p > O. [Inspeet the power-series definition of
x- p 1 p (x).]
38. By using Eguations (l08) and (109), together with integratioo by parts, deduce
the following reduetion formulas:
(a) f xmJn(x)dx = xmJn+,(x) - (m - n - 1) f xm-'Jn+,(x)dx,
(b) f xmJn(x)dx = -xmJn_,(x) + (m + n - 1) f xm-'Jn_¡(x)dx.
[Notiee that the first reduetion eventually yields a elosed form, by jteration, when
m - n is an odd positive ioteger, whereas the seeood reduetion does so wheo m + n
is an odd positive integer. When m and n are both even or both odd integers, appro-
priate use of one or both of the reduetions wiU yield the sum of a elosed forrn and a
multiple of the ter m J 1,,(x) dx. Whereas this integral eannot be further simplified,
the funetion f~ 11)(t) dI is a labula!ed funetion.]
39. Use the results of Problem 38 to deduce the following formulas:
(a) I x p+' Jp(ax) dx = ~ x p+ 'Jp+, (ax) + e,

(b) I x,-pJp(ax)dx = - ~ x'-PJp_'(O:x) + e,
(e) f x 3 J o(x) dx = x 3 J, (x) - 2x2J2(X) + e,
(d) f x 6J,(x) dx = X6J 2(X) - 4x'J)(x) + 8X 4J 4(X) + e,
(e) I J)(x) dx = -J 2 (x) - ; J,(x) + e,
(f) f xJ,(x) dx = + Jo(x) dx + e,
-xJo(x) f
(g) f x-¡J¡(x) dx = -J,(x) + f
Jo(x) dx + e,
(h) f J (x) dx = -2J,(x) +

2 f
Jo(x) dx + C.
1'/2
2 o c a s (x sio
Jo(x) = n e) de
by verifying that the right-hand member satisfies Bessel's eguation of order zero and
investigating its value when x = O. (The other Bessel funetions have sirrtilar integral
represeotations. See Problem 41, below, aod Problem 65 of Chapter 5.)
Probl ems 177
41. (a ) Deduce fro m Pro blem 40 that
J ,(x) = n2 f.n:'
o sin (x sin O) sin O dO.
( b) By integra ting the r ight-hand member of this relation by part s, s how that
1
X-J, (x) = n2 f.on12eos ( x sin O) eos 2 O dO.
(e) Deduce fro m Problems 40 and 41 (b) that

n
J, ( x) = n2 f.o ;2cos (x sin O) cos 20 dO.
42. E s tablish the relation
n=->o<>
by firs t eons idering the expansion
eÍ'e -;' = [i:

}~o
(xr/,2)1J[~
J. k- o
(_l)k(X/2~)kJ _ ~ ~ ( _ I)k (X/,2)/; \ j - k,
k. j ~ O k~O J .k .
replacing j by n + k, where n is a new index of summation, and then idemifying the
coeffieient of r n in the resultant series. [The funetion e 1C' - f) is kno wn a s the g enerafing
funetion for Jn(x).]
43. Let U(x, r) = eÍC'-f)

(a) Show that
au
ax = ~(r
2
- ~)u
r
and, by using the result of Problem 42, deduce the relation

d 1
dxJn(x) = z[J.-,(x) - J n+, (x )].
(b) Show that
au
ar
= ~(I
2
+ _1
r
)U
2
and, again using Problem 42, deduce that

x
(n + I ) Jn +,(x) = z [Jn(x) + J n + 2 (x) ] .
[Note that these results are specializations o f (112) and (l 13) .]

44. (a) U se Abel's formula, Equat ion (65) o f Chapte r 1, to show that the Wrons kian
of any two solution s of Bessel' s equati o n (63) or o f the m o dified equa ti o n (93) is always
of the form A /x , where A is an appro p riate constan!.
( b) Fro m thi s result deduce that
Jp(x) Y~(x) - Yp(x) J 'p(x) = ;-

178 Series SolutioDS of Differential EquatioDs: Special Functions
and, by eonsidering rhe limiring form as x-O, show rhat e = 2/n and henee
Jp(X) Y~(x) - Yp(x)J'.(x) = ;".
(e) In a similar way, show that
Jp(x)J'-i x ) - J_p(x)J~(x) = - ;x sinpn,
Ip(x)K~(x) - Kp(x)I~(x) = - ; ,
Ip(x)I'-ix) - Lp(x)I~(x) = - nx
2. sinpn:.
[Use Equation (59) of Chapter 2.]
45. Use the resuJt of Problem 44(b) , and the methods of Seetion 1.9, to obtain the
general solurion of the equation
X2 ----f
d
dx
2
+ d
x -.-2'
dx
+ (X2 - p2)y = f(x)
in the form
[Notiee that here h(x) = f(x)/x 2, with the notation of Seetion 1.9.]
Section 4.10
46. Obtain the general soluti on of eaeh of [he following equarions in terms of Bessel
functions or , if possible , in terms of elementary fune[ions:
d 2y dy
(a) x - ' - 3 d- + xy = 0,
d x- x
d2y dy 3_
(b ) X dx2 - dx + 4x y - O.
2
(e) X 2 ddX2y + x dy
dx
_ ( 2 ,
x .. 4 Y - ,
_1) - °
( d) x ~;{ + (2x + 1) (~; + y) = 0,
d 2y dy
(e) X dX2 - dx - xy = 0,
d 2y
(f) x' - - + a 2y =
dX2 ° ,
d 2y
(g) dx' - x2y = 0,
(h) xdd2~
x + (1 + 2x)ddY
X
+ y = 0,
(i) x d;i
d'
+ (1 + 4x d;
d
+ x(5 + 4x
2
)
2
)y = O.
47. The two fOllo'Ning equations each have arisen in several physical investigations.
Express the ge neral solution of each equation in terms of Bessel funetions and also
J'roblems 179
snoW that it can be expressed in terms of ek,mentary fun c tions when m is an integer :
d'y _ a,l _ m(m + 1)
(a) dX2 y - X2 y,
dly _ 2mdy _ a,ly
(b) cJxl
= O
X dx .
48. Show that for the differential equation
d'y, dy _
X dX2 .., 3 dx + 4xy - O
tbe condition y(O) = 1 determines a unique solution, and hence that y ' (O) cannot also
be prescribed. Determine tbis solution.
49. Find the most general solution of the equation
dly y
X 2 dX2 + X dcJx +(?
x- - 1)y = O
for whicb
lim 2nxy(x) = P,
x-o
where P is a given constan!.
50. The differential equation for small deflections of a rotating string is of the form
fx(Ti) + pW 2 y = O.
Obtain the general solution of this equation under the foJlowing assumptions:
( a) T = Toxn, p = poxn; T o = ¡lpOW 2 •
(b) T = To x n, P = Po, fl;>'= 2; T o = flpow ' .
(e) T = T ox2, P = po; T o = 4Pow 2 •
Section 4.11
51. From the series definitions (137) and (138), and the definition (134), calcula te the
value of each of the following quantities to four significant figures:
(a) ber 1, (b) bei 1, (e) Mo(l), (d) 8 0 (1).
52. EstabJish the following relations, assuming tbat ¡3 /l = e," l/ 4:
(a) ber' x + í bei' x = _¡ 3IlJ¡(í3 Il X) = M,(x)e iI8 ,(x l- n / 41,
(b) ber¡ x = "i2 (ber' x - bei' x),
(e) bei, x = "i2 ( ber' x + bei ' x).
53. (a ) Use Equations (J 33) and (107) to obtain the res ults
as x -jo 00 , where
180 Series SoJutions of DifferentiaJ Equations: Special Functions
(b) Use Equations (139) and (107) to obtain the res ults
e- ·'t .' ,, ' 1"
ker p x ~ I 2 cos I¡J p ,
I-
'Y n'x
as x ---+ co, where
x n pn
I¡J p = -/2 + 8 + "2 .
54. (a) Show that Equation (131) can be put into the form
d
dx(xy ') =
(P2)
x + ix y .
( b) Notieing that this equatíon is satisfied by the complex function
y = berpx + ¡ beipx,
deduce the res ults
d
-d (x bel ~
. x) = x ber x ~i x
+ p2 _ _P_.
p
X X
(e) Show simiJarly that the funetions ker p x and kei p x satisfy eompletely analog-
ous equations .
Section 4.12
55. Determine particular solutions of the partia] differentiaJ equation

I a ( . a
sinrparp smrp arp
U) + al'a ( ,2
a U)
al' = o
of the form U(r, rp) = rny(rp), where n is a constant and y is independent of r.
56. Establish the followíng properties of the Legendre poJynomials:
(a) P.e-x ) = (_l)npnC-"), P~(-x) = (_l)n+lp~(x),
O ( n even),
(b) P~ (O) = . ( 1.3.5 ... n)2 (n odd),
{ (_1)(n-I)/ 2-,-,--=-=-,_-<-
n!
(e) P~ ( l) = (_l)n+lp~(-l) = ±n(n + 1).
[Use Equation (149) in part (e).]
57. By díreet integration when n = O, and otherwise by integrating the equal members
of Equatíon (189) and using Problem 56(b) when necessary, establish the following
result s:
2 (n = O),
J
I
(a) -1 Pn(x) dx = {O (n *- O),
I (n = O),
(b) f p .(x) dx = O (n = 2,4,6, ... ),

o
J(_ 1)(.-1)/ 2n(n I+ 1) (1·3·5···
n'
n)2 (n = 1, 3,5, .. . ).
probl ems 181
58. Show tha t the following differential equations have the indieated general solutions:
(a)
d'
d~{ + lsd eoths - n(n + I)y = 0, y = c'?n(eoshs) + c2Qn(eoshs);
d' d
(b) (1 - X2) dO { - 4x dY + (n + 2)(n - 1)y = 0, y = c,P~(x) + C2Q~(X);
x x
2
d y dy
(e) x(1 - x ) - + (l - 2x)d- + n(n + l)y = 0,
dx ' X
y = c,Pn(l - 2x) + C2Qn(1 - 2x).
59. (a) Use Equations (157) and (159) to obtain an expression for Pn(x) in deseending
powers of x in the forrn
_ (2n - 1)(2" - 3)··· 3.1[ n _ n(n -1) n-2
Pn(x) - n! x 2(1)(2n - 1 r"
n(n - 1)(n - 2)(n - 3) x n- 4 - •.• ].
+ 22(1. 2)(2n - 1)(2n - 3)
(b) Yerify that the eoeffieient of X n- 2k in the expression for Pn(x) is
_ k (2n - 2k)!
Cnk - (-1) 2nk! (n _ k)! (n - 2k)!
where k may vary from ° to n/2 or (n - 1)/2, aeeording as n Is even or odd.

60. (a) Use the binomial expansion to obtain the result
n ni
(X2 - l)n = L
k~O
(-l)k I
k.(n-k).
. , x 2n - 2k •
(b) Henee show that, with the notation of Problem 59(b),
wbere N = n/2 or (n - 1)/2, aeeording as n is even or odd. [This estabJishes Rodrigues'

formula (168).]
61. Use Rodrigues' formula (168) to prove that Pn(x) has n distinet zeros inside the
interval (-1,1). [Notiee that (X2 - l)n is a polynomial of degree 2n with zeros of
multiplieity n at the end points, so that d(X2 - l)n/dx has zeros of multiplieity n - 1
at the end points and one interior zero, and so forth.]
62. Prove that 1 Pn(x) 1 < 1 when 1 x 1 < 1 by the following steps:
(a) With the definition
M(x) = [Pn(x)]2 + n~n-+x;)[P~(x)F,

when n = 1,2, ... , show that
max [Pn(x)]2 < max lv/ex).
Ix[~l [xl;:;;¡
[Notiee that M(x) = [Pn(x)]2 when Pn(x) takes on a relative maximum or minimum
and when x = ":"1.]
(b) Show that
M'(x) =
n(n + 1)[P~(x)J2.
2x
(e) Deduce that in the interval -1 <: x <: 1 the function M(x) takes on its maxi-
mum value at the end points, and henee complete the required proof.
(Ixl <: l,lrl < 1)
by first obtaining the binomial expansion

r 1· 3
[1 - r(2x - r)]-1/2 = 1 + 2(2x - r) + 2 22!r 2(2x - r)2 +
.l 1·3 ... (2n - 3)rn-I(2x _ r)n-l
, 2 n 'en - 1)! .
1·3 ... (?n - 1)
+ 2 nn-! rn(2x - r)" + ....
when Ir(2x - r)1 < 1, and then picking out the coefficient of r n in thisexpression and
using the result of Problem 59(a). [Show that the coefficient of r k in (2x - r)m is
(_ 1)' m(m - 1) ..~ !(m - k + 1)(2x)m-k
when k > 1 and is (2.x)m when k = O, and notice that no terms beyond those written
can involve xn. Since I r(2x - r) I < 1 when both O <: x <: 1 and O <: r < 1, and also
when both -1 <: x <: O and -1 < r <: O, in particular, the desired result is thus
established in those cases. But since the series itself is a power series in r, the region
inside which it converges must be symmetric in r, and hence must admit Ir I < 1 when
Ixl <: 1. The function (1 - 2rx + r ')- 1/2 is called the generaling funclion for Pn(x).]
64. Let V(x, r) = (1 - 2rx +r 2 )-1 '2.
(a) Show that

aV
(1 - 2rx + r2)-¡¡¡: ~ (x - r)V
and, by using the result of Problem 63, deduce the relation

(n + l)Pn+¡(x) - 2nxPn(x) +(n - l)Pn-¡(x) = xPix) - Pn-l(x)
when n > 1, and hence establish the recurrence formula (162) for Pn(x).
(b) Show tha t
(1 - 2rx + av
r 2) ax = r V
and, again using Problem 63, deduce the formula

d d d
dxPn(x) - 2x dxPn-l(x) + dxPn-2(x) = Pn-l(x)
when n > 2.
[By combining the results of parts (a) and (b) with the differential equation satisfied
by P.(x), a variety of other differential recurrence formulas can be derived.]
65. (a) Show that the substitution t = 1 - x transforms Legendre's equation to the
form
d2 d
1(2 - 1) d/. + 2(1 - I)d~ +p(p + I)y = O,
problems 183
that this equation has one solution regular at t = O, with exponent zero, and that all
other solutions beeome 10garithmieally infinite at I = O.
(b) Determine the regular solution, and henee show that the solution P p(x) of
Legendre's equation whieh is finite at x = 1 and whieh is uoity at that point is given by
P (x) = I
l'
+ ~ [(p
k- t
+ IXp + 2) ... (p + k)]
x [(-p)(l - p) ... (k - I _p)](l - x)k

2k(k !)2
near x = l.
(e) Verify that this result is in aeeordanee with Equations (158) when p = O, 1,
and 2 by writing out the terms io the series.
Section 4.13
66. Show that
:XF(rx" p; y; x) = rx,! F(rx, + 1, P+ 1; Y + 1; x) .

•
61. (a) Show that the ehange io variables x = 1 - I traosforms the hypergeometric
equation (175) to the form
1(1 - /) ~;:i + [(rx, + p - y + 1) - (rx, + p + 1)/] '!!; - rx,py = O.
(b) By eomparing the eoeffieients in trus equation with those in Equation (175),
deduee that the general solution of (175) eao be writteo in the form
y = c,F(rx" p; rx, + p - y + 1;1 - x)
+ c2x,-~-PF(y - p, y - rx,; 1 - rx, - p + y; 1 - x)
near the second singular point at x = 1, jf rx, + p - y is noniotegral.
68. (a) Show that Jacobj's equalion,
d 2y dy
x(1 - x) dX2 + [a .,- (1 + b)x] dx + n(b + n)y = 0,
is a hypergeometric equatioo.
(b) Deduce that the general solution, near x = 0, ean be expressed in the form
y = c,F(n + b , -n; a; x) + c2x'-aF(n + b - a + 1,1 - n-a; 2 - a; x),
if a is nooiotegral.
69. Show that a solution of Jacobi 's equation (Problem 68) which is regular at x =
ean be written in the explicit forro
°
J( b )=I_(n+b)nx+(n+b Xn+b+l)n(n-l)x 2 +
na, ,x al! a(a + 1) 2!
if a is not zero or a negative integer, where
Jn(a, b, x) = F(n + b, -n ; a; x).
(Notiee that this solutioo is a polynomial of degree n when n is a positive integer or
zero.)
184 Series Solutions oC Dilferential Equations: Special Functioos
70. (a) Show that the substitution x = 1 - 21 transforms the equation
(1 - x 2
)'!.:;- - e x : + n(n + e - l)y = O
(see Problem 26) to the speeial Jaeobi equation
t(1 - {
t)~; + ~ (1 - 2t) CZ + n(n + e - I)y = O.
(b) Deduce that, when n is a nonnegative integer, the polynomial solution f/J.(x)
of the original equation ean be expressed in the fonn
f/J.(x) = f/Jn(l)F ( n + e - 1 ; -n; 2e ; 1 -2 x) .

(e) Use this result and the results of Problems 26 and 69 to obtain the fOllowing
relations:
Tn(x) = (
F n, -n; l1-X)
2; - 2 - = (1
J n 2' O, - 2 - , l-X)
Pn(x)=Fn+
(
1 -
1,-n;1;-2- x) =J (1,1'-2-'
1 - X)
n
Sn(x) = (n + l)F ( n + 31-x)

2, -n; 2; 2 = (n + (3 l-X)
1)Jn 2,2, - 2- .
Section 4.14
71. Show thal the following differential equations eaeh have an ordinary point at
x = 00 and in each case obtain two independent solutions expressed in the form
y = ~ Akx- k :
d 2y dy 2 3dy
(a) x2 dx,' + (1 + 2x)dx = O, (b) x .d
dxY+2
2 X dx, + Y = O.
7'1.. (a) Show that the differential equation
d 2y
x' dx,2 +Y = O
has a regular singular point at x = 00 and obtain two independent solutions directiy
in the fonn y = ~ Akx- k -,.
(b) Obtain (he same solutions by fust making the ehange in variables x = l/t.
73. Verify that the s ubstitution x = 1/1 transfonns the hypergeometric equation (175)
to the form
d 2y dy
1(1 - t) dl 2 + [(1 - IX -
fJ
) - (2 - y)t] di + IXfJ
-t y = O,
and that this equation possesses a regular singular point at t = O unless IXfJ = O and
IX+ fJ = 1, in whieh ease t = O is an ordinary point.
74. Show that the exponents of the differential equation obtained in Problem 73 are
IX and fJ at t = °
and deduee thal the transformed equation possesses solutions of the
fonns t~u(t) and tPv(t), where u(t) and v(t) are regular at t = 0, unless IX and fJ are
equal or differ by an inleger.
Problems 185
75. Verify that ¡he substitution y = I'u(l) transforms ¡he equarion of Problem 73 lo
the form
dOu
1(1 - 1) dl 2 + [(1 + a -
f3.
) - (la - y + 2)1] dI -
du
a(1 + a - y)u = O.
76. Show that Ihe equation obtained in Problem 75 is of the form of Equation (175) if
pis replaced by 1 a - y and y is replaced by 1 + a - f3 in (175), and hence deduce
that ¡he functions
I ;C"F(a, 1 + a - y; 1 + a - fl; ;), :c P F(fl, I + fl - y; 1 - a + fl ; ; )

are independent solu¡ions of the hypergeometric equation (175) for large values of x
unJess either a and fl differ by an integer, in which case one of these expressions is
undefined, or a = fl, in which case the two expressions are identical.
77. Use Equations (194) and (195) to verify ¡he three-place values
(a) J o(10) = -0.246, (b) YoCIO) = 0.056.
78. Show that the relations (194-199) are equali¡ies for p = i.
79. Assume as a formal solution of ¡he modified Bessel equation of order zero ,
2
d y
dX2
+ l.. dy
X dx
_
y -
- O
,
an expression of the form
~
y "" e TX
L Akx- 1k+,rl
k=O
and obtain the formal requiremeot

(,2 _ I)Aox-' + [(,2 - I)A, +,(l - 2s)A o]x-'-'
-
+ k~= O [(" - I)Ak+2 - ,(J + 2k + 2s)A k +, + (k + s)'A k]x-,-k-2 = O.
Hence deduce ¡hat, if Ao ;: O, there must folJow either
,= 1,
1 (2k + 1)2 > O)
s=2' Ak+1 = 8(k + 1) Ak (k
I (2k + 1)2
or r= -1, s=2' Ak+l - - 8(k + 1) Ak (k > O).
Thus obtain the formal solutions
y,(x) ~ e, 1
~;[l + /(~X) + 2]1~~:;2 - ... ]1 (x ~ col.
e-xl-
Y2(X) ~ e 2"¡ x I I ! (8x) + 2! (8X)2
j2 ]2.3 2
...
]
[Witb e, = 1/"¡2 n , e, = "¡n/2, these results are respectively the asymptotic (but
divergenl) expansions of ¡ o(x) and K o(x), in agreement with (198), (199), and (202).
In accordance with ¡he s¡atement at the end of Section 4.14, retention of only the fust
term would introduce an error of less ¡han approximately 1 percent for values of x
larger than about 12.]
5
Boundary-Value Problellls and
Characteristic-Function
Representations
5.1. Introduction. In many problems the solution of an ordinary differential

equation must satisfy certain conditions which are specified for two or more
values of the independent variable. Such problems are called boundary-vallle
problems, as distinct from initial-value problems, wherein all conditions are
specified at one point.
A linear condition or equation is said to be homogeneolls if, when it is
satisfied by a particular function F, it is also satisfied by eF, where e is an arbi-
trary conSlant. For example, the requirement that a function or one of its
derivatives (or sorne linear combination of the function and/or certain of its
derivatives) vanish at a point is a homogeneous condition. In the present chapter
we are mainly concerned with homogeneous linear ordinary differential equa-
tions and associated homogeneous boundary conditions.
In illustration, we may require a solution of a homogeneous linear equation
of second order, of the form
d 2 y2 + dy
al(x)-d + _ O
a 2 ( x)y- (1 )
dx x
which vanishes at the two points x = a and x = b,
y(a) = O, y(b) = O. (2)
Since lhe general solution of the differential equation is of the form
y = elul(x) + e 2 u 2 (x), (3)
where lit and U2 are linearly independent solutions and el and e 2 are constants,
186
5. t. lntroduction
187
the boundary conditions constitute the requirements

clul(a) + c 2 u,(a) = 0,
clul(b) -i- C2 u 2 (b) = O. (4)
One solution of these equations is e, = e 2 = 0 , leading to the trivial Soft/lion

y = O. If the determinant of the coefficients of el and c 2 does not vanish, tben
(by Cramer's rule) this is the only solution. Hence, in order that nontrivial solu-
tions exist, it is necessary that the determinant of coefficients vanish,
u 2 (a)
u 2 (b) -
1_ O. (5)
If this condition exists, the two equations in (4) are in general equivalent and
one constant can be expressed as a multiple of the second by use of either
equation, the second constant then being arbitrary. Thus, ¡f (5) is satisjied, the
second equation of (4) may be discarded and the first equation then gives
e 2 u 2 (a) = -elu,(a). If we write el = Cu 2(a), there follows e 2 - -Cu,(a), and
the solution (3) becomes
(6)
with C arbitrary.
al
Jt should be noticed that (6) is a non trivial solution onJy i.f li I (a) and u2(a) are
le not both zero. If u¡(a) = u2(a) = 0, the fust equatioo of(4) is the trivial identity,
re and the seeond equatioo must be used to relate el and e2' This process leads to a
oontrivial solution of the form
lS y = C[u2(b)u¡(x) - U, ( b)U2(X»),
'1- assumiog that u,(b) and I/2(b) are not also both zero. If u¡(x) and I/2(X) should
ts both vanish at x = a and at x = b, theo (3) would satisfy (2) for arbitrary values
ts of both e, and e2'
er In many cases one or both of the coefficients a,(x) and a 2 (x) in (1), and
a- hence the solutions u,(x) and lI 2(X) , depend upon a constant parameter A wruch
may take on various constant values in a particular discussion. In such cases
the determinant (5) may vanish for certain specific values of A, say A = A"
A2 , • • • • For each such value of A a solution of type (6), involving an arbitrary
multiplicative factor, is tben obtained. Problems of this sort are known as
1) eharaeteristie-value problems; the values of A for which nontrivial solutions
exist are called the eharaeteristie values of A, and the corresponding solutions
(with convenient choices of the arbitrary multiplicative constants) are called the
2) eharaeteristiefunetions ofthe problem. Tbe terms "eigenvalues" and "eigenfunc-
tions" are also frequently used.
In the remainder of this chapter we first consider certain examples of sucb
problems, and then establish and investigate certain rather general properties of
3)
characteristic functions which are extremely useful in a wide dass of related
~ s, problems.
188 Boundary-VaJue Problems and Characteristic-Function Representations
5_2_ The Rotating String. We consider the problem ol' determining the form
assumed by a tightly stretched flexible string of length 1 and linear density p,
rotating with uniform angular velocity úJ about ¡ts equilibrium position along
the x axis. It is assumed that tbe initial tension in tbe string is so large that
additional nonuniform stress introduced by tbe curvature of tbe string is rela-
tively negligible. Denoting tbe displacement from the axis of rotation by y(x)
and tbe uniform tensile force by T, and considering only small displacements
and slopes, the condition of force equilibrium is of tbe form
d 2y
T dX2 + púJ2y = O. (7)
If we consider a small element o[ the de[ormed string, projecting into the interval
(x, x + Óx), the y component of the tensile force is given by - T dyjds at the
end (x, y) and by (T dyjds)., + ó.x at the end (x + Ó.x, y + Óy), where s is are length
measured along the string. The differential resultant force on the element is
thus given by (djdx)(T dyjds) &. If the distributed external force, per unit dis-
tance along the string, in the y direction is denoted by Y, then the requirement of
differential force equilibrium is
:!x(T~:)& + YÓs = O.
By dividing by Óx, letting Óx _ O, and replacing Y in the present case by the
inertia force púJ2y, we thus obtain the equation
:!s(T'Z) + púJ'y = o.
For small slopes we have
de- .. )
d(· .. ) dx d(· .. )
ds = JI + C7xr= dx
With this approximation and the assumption of nearly uniform tensile force, we
obtain Equation (7).
Restricting attention to tbe case of a string of uniform density, we define
the constant parameter
(8)
and write (7) in the form

(9)
If one end of tbe string is attacbed to tbe axis of rotation at tbe point x = O,
tbe end condition
y(O) = O (10)
must be satisfied. If tbe otber end is also attacbed to the axis, the second end
condition is
y(l) = o. (l1a)
5.2. The Rotating String
189
If, however, the second end is attached lO a yielding support equivalent to an

elastic spring which exerts a restoring force proportional to its stretch, toward
the axis of rotation, the second end condition is of the form T dy/dx = -ky,
where k is the "spring constant." With the abbreviation
T
a=-,
kl
where a is a dimension/ess constant inversely proportional to the elastic modulus
of the spring support, this end condition can be written in the more convenient
form
a/y'(l) = - y(l). (11 b)
In the limiting case where a = O, Equation (llb) reduces to the condition of
fixity (1 la); in the limiting case a = 00, the end at x = / is not restrained from
motion normal to the axis of rotation, and we are led to the "free-end" condition
y'(l) = o. (11 e)
In order that the hypothesis of nearly uniform tension be realized, it mus! be
supposed that the "free end" is, however, restricted from appreciable movement
in the direction of the axis of rotation.
The most general solution of Equation (9) satisfying (10) is of the form
y = C sin ~Ix. (12)
If the end x = / isfixed, the condition of (1 la) becomes
Csin ~I/ = O. (13)
Hence nontrivial solutions of this problem exist only if A has a value such that
(n = 1,2, ... ). (14)
If these values of A are ordered as Al> Az , ... , An , ••• , with respect to their
magnitude, we can then write
(15)
where n is an integer. For each such value of A, the solution (12) can be written
in the form
y = C'Pn(x) ,
where . nnx (16)

'Pn (x) = SIn - / _ .
It is clear that only positive integral values of n need be considered.

Thus the boundary-value problem .consisting of the differential equation (9)
and the homogeneous boundary conditions (10) and (1 la) has no so/ution other
than the trivial solution y _ O, unless A has one of the characteristic va/ues
given by (15). Corresponding to each characteristic value of A, there exists a
characteristic function 'Pn(x). given by (16), such that any constant multiple of
this function is a solution of the problem.
190 Boundary-Value Problems and Characteristic-Function Representations
Each characteristic value ).n corresponds to a specific value of the angular

velocity w (for given T and p), according to the notation of (8),
( n=I,2, . . . ). (17)
These velocities are known as the critica/ speeds of the rotating string.
The preseot analysis indicates that for speeds s maller than w"
O < w < .!E.. / T,

I V p
the only stable positioo of the string is its undeformed position along the axis
of rotation. However, if the speed is continuously increased until the first critical
speed is attained (w = w,) , a possible oew equilibrium form or deflection mode
y =
· n-/-
C SIO x
may exist. Corresponding to the second critical speed (w = W2 = 2w,) another

mode
y = e sin 2n2.
/
may exist, and so on.
It may be seen that only the shapes of the string modes are determined, the
amplitudes being apparently arbitrary. This iodeterminacy is a result of the
approximations made io deriviog the lioearized formulation of the problem aod
can be removed by a more nearly exact analysis of the true nonlinear problem.
Thus, if use is made of the additional relation T dx/ds = T o = constant, which
follows from the condition of force equilibrium in the x direction (in the absence
of external force components in that direction), the tension T can be eliminated
fram the relation
.!L(T dy ) + pW 2 y= O
ds ds
to yield the nonlinear equation
To~;' +PW 2y Jl
When p is constant, the appropriate solution of tbis equation can be obtained
by use of so-called e//iptic illtegrals,t after wrucb T is given by the relation
tThe integrals
F (k , rp) = f: vI dO . 2 O and
- k 2 sin
E(k, rp) = f.o vi k 2 sin 2 8 dO,
where O < k < 1, are known as elliptic i1ltegrals of rhe firsr and second kind, respectively, and
are tabulated. For reductions expressing related integrals in terms of ¡hese integrals, see
Reference 4 of Chapter 7. t
5.2. The Rotatiog String 191
If the end X = 1 is attached to the axis of rotation by an elastic spring, the

requirement that (12) satisfy (i lb) becomes
-IX..JII cos .JII = sin .JII.
With the introduction of a dimensionless parameter Ji, of the form
Ji = ../XI, (18)
this condition can be written
tan Ji = - IX.Ji. (19)
That there are an infinite number of values of Ji which satisfy (19), for a given
value of IX., is readily seen if the two curves y = tan Ji and y = -IX.Ji are plotted
together and the desired roots are recognized as the values of Ji corresponding
to the intersections.t If we order the positive roots with respect to magnitude
as Ji" Jiz, ... , Ji., ... , the corresponding characteristic vall/es of A are given by
z
An = Ji.
IZ (20)
and the corresponding deftection rnodes are of the form
rp.(x) = sin Ji. ~ . (21 )
Finally, the corresponding critical speeds are given by
(J). = ~• .¡ ~ . (22)
In the limiting case IX. = 00, when the end x = 1 is not restrained from
moving normal to the axis of rotation, Equation (19) gives
2n - I
(23)
cos Ji = O, Jin= 2 1C.
The characteristic values of A are thus

Z
= (2n - l)Z1C (n = 1, 2, ... ), (24)
An 2 12
with corresponding characteristic modes
. 2n - I x (25)
'P.(x) = Sin 2 1CT'
and the critical speeds are given by

(J) = 2n - 1 1C ff . (26)
• 2 /V/i
tOnce approximale values of rools of lranscendental equations such as (19) have been ob-
tained graphically, improved values oflen can be obtained by a "feedback" method, in which
an approximation is inserted in one s ide of the equalily and the resultant equation is then
solved for the next approximation. NeWlon's melhod (see Section 7.10) is also useful. '.
5.3. The Rotating Shaft. We next consider the determination of the possible
deflection modes of an originally straight shaft of length 1, rotating with uniform
angular velocity w about its equilibrium position along the x axis. According
to the elementary theory of bending of beams, the deflection y from the straight
form is determined (approximately) by the differential equation
2 2
d
dX2
(
El d Y)
dX2 -
, = O,
pcu-y (27)
where E is Young's modulus, 1 is the moment of inertia of a cross section about

an axis perpendicular to the xy plane, and p is the linear density of the shaft
material.
If we consider the portion of the shaft to the left of an arbitrary cross sec-
tion (x positive to the right, y positive upward), the influence of the right-hand
portion of the shaft on the portion considered can be resolved into a vertical
shearing force S and a bending moment M. We choose the convention that S is
then positive downward and M is positive
when counterc!ockwise (se e Figure 5.1).
Thus, if the portion considered is to be
1 ~s
~ ¡7/I77// / Y/i , /T7/1 , I / i ,
M
in equilibrium, S must equal the alge-
braic sum of the actual upward forces
/ / f / / f / J / ' L / / / / J ' / / ;' " , ' L
..... (áistributed aná concentrated) acting to
L x
the left of the section considereá, and
¡\.f must equal the resultant c!ockwise
moment, about the section eh osen, of
Figure 5.1
" al! [orces acting on the portion of the
shaft to the left of the section. With these conventions, there fol!ows also
_. "'" d2
M = El y (28)
dX2
2
and s=-
d ( EdI -
dx dX2
Y) . (29)
If we consider a small element of the shaft, in the interval (x, x + Ax), and
assume small deflections and slopes, the differential unbalanced vertical force
acting on the element is given by -(dSjdx) Ax and the differential unbalanced
counterclockwise moment is (dMjdx) fu - S Ax. With the distributed external
load intensity (positive in the positive y direction) denoted by Y(x), the condi-
tions of differential force and momen! equilibrium become
dS di'VI
- dxti>: + Yfu = O, -tiT
dx
A .
- Sax = O
'
and there follows
dM -s dS = y
dx - , dx .
The elementary theory of bending leads to the approximate relationship
El
R=M,
5.3. The Rotating Shaft 193
where 1/ R is the curva ture oC ¡he shaCt and is approximated by d 2 y/dx2 iC the
sJope oC the shaCt is small. We Ihus o btain the relations ( 28) and (29), in addi-
tion to the equalion
2 1y
d ( El ddX' ) =
dX' Y(x),
which reduces 10 (27) iC y is replaced by the inertia loading pcv 2 y.

In add ition to the differential equation (27), four boundary conditions must
be prescribed, two conditions being imposed at each end of the shaft. Types of
homogeneous end conditions of particular interest are the following:
(1) Hinged End. In this case the lateral displacement y and the bending
moment IV[ must vanish, at the end considered, giving the two conditions
d 2y _
y = O, El dx 2 - O. (30a)
(2) Fixed End. Here the displacement y and the slope dy/dx must vanish,
y = O, dy = O (30b)
dx .
(3) Free End. If no moment or shearing force is applied at an end, there

follows
d ( 2y
dx El ddX2 ) _
- O. (30c)
(4) Sliding Clamped End. If one end of the shaft is constrained to retain
zero slope but is completely free to move in the y direction, the slope dy/dx and
the shear S must vanish at that end,
dy = O ~(Eld2y) = O. (30d)
dx ' dx dX2
(5) Elastically Supported End. If the motion of one end in the y direction is
partially restrained as by an elastic spring, with modulus k" the magnitude of
the shearing force mustbe k, times the displacement,
2y
d ( El ddX2 ) = ±k ,y.
dx (30e)
lf the change in slope of the end is also partially restrained as by a spring sys-
tem such that the restraining moment is proportional to the slope, the second
condition at that end is of the form
d 2y dy (30f)
EI - = ±k2 dx
dX 2
- ·
By taking the four possib1e combinations of the limiting form s of these con-
ditions when k I and k 2 are zero or infinitely large, the four preceding cases are
attained. Specifically, k, = 00, k 2 = O gives (30a); k, = k 2 = 00 gives (30b);
k, = k, = O gives (3Oc); and k, = O, k 2 = 00 gives (30d).
194 Boundary-Value Problems and Characteristic-FunctioD Representations
We restrict attention to the case of a uniform shaft, for whicb El and pare
constant , and introduce the constant parameter
2
pw- .
A.= (31)
El
The differential equation (27) then becomes
d'y
dx' - A.y = 0, (32)
with general solution of the form

y = e, sinh"yIx + e 2 cosh --YXx + e 3 sin "yIx + e. cos --YXx. (33)
In the special case of a rotating shaft hinged al bolh ends, we take the origin
at one end of the shaft and impose the boundary conditions
y(O) = y " (O) = 0 , y(l) = y"(1) = O. (34)
From the conditions at x = °we obtain
e2 + e. = 0, e2 - e. = 0,
e 2 = e. = O. (35)
,. To simplify the notation in the remaining work, it is convenient to introduce
the dimensionless parameter Il, where
4/ -
Il = v' A. 1, 11 2 = w[2 /...P.... . (36)
'V El
Making use of (35) and (36), we then write (33) in the fOrIn
. h I lx-
Y = e, SIO + e, SID
. IlT'
X (37)
l
Tbe last two condition s of (34) then beco me
e, sinh Il + e, sin Il = 0, (38a)
e J sinh Il - e 3 sin Il = O. (38b)
These equations are compatible only if the determinant of coefficients vanishes,
sinb Il sin
.
I
Il = - 2 smb Il . .
Sill Il = O.
I sinb Il - 510 Il
Since the value Il = °leads to a trivial solution y = O, the only permissible

values of Il wbich are of interest are the positive roots of the equation sin Il = O,
Iln = n7t (n = 1, 2, ... ). (39)
For such values of Il, Equations (38a, b) require that
e, = O. (40)
5.4. Buckling of Long Columns under Axial Loads 195
Hence the characteristic functions corresponding to permissible values Jln are of

the form y = c 3 IPn(x), where
. nnx
x = sm
IPn () -,-o (41)
These deflection modes are of the same form as the modes of a rotating
flexible string with fixed ends. The characteristic values of A. in the present case
are gi ven by (36),
n4 n 4
A. n = --¡.--' (42)
and the critical speeds are given by

_ n 2 n 2 IEl
COn - --¡z -Y p (n = 1, 2, ... ). (43)
The smallest critical speed corresponds to n = 1. If we denote thisfundamental

critica' speed by n,
n= n2 I EI ,
1 V P 2
(44)
then the nth critical speed is given by

con = n 2 Q. (45)
Thus, whereas Equation (17) shows that for a rotating string with fixed ends
the successive critical speeds increase in proportion to the integers, Equation
(45) shows that for a sbaft hinged at both ends the speeds increase in proportion
to the squares of the integers.
5.4. Bllckling oC Long Columns under Axial Loads. As a final example, we

next consider a long shaft or column subjected to an axial compressive force P
applied at an end and investiga te possible modes of lateral deflection from tbe
initial equilibrium position along tbe x axis. Tbe differential equation determin-
ing tbe deflection y, according to the elementary theory, is then of the form
2 2
d ( d 2y ) d y _
dX2 El dX2 + p dX2 - o. (46)
The bending moment }vlp at an arbitrary section x, due to a compressive force P

applied in the negative x direction at the end x = L, is given by P(y(L) - y(x)].
Hence (see Section 5.3) the effect of the axial load can be taken into account if a
fictitious distributed lateralloading of intensity
d 2M p d 2y
Yp(x) = dX2 = -P dX2
is introduced. If the actual distributed lateral load intensity is denoted by Y(x),

the differential equation governing lateral deflection of the column is then
2 2
d ( d 2y ) d y
dX2 El dX2 = Y(x) - P dx2·
This equation reduces to (46) in the absence of distributed lateral loading.

196 Boundary-Value Problems and Characteristic-Functioll Representatiolls
The boundary conditions to be imposed depend upon the nature of the end
restraints, as was outlined in Section 5.3.
Restricting attention to a uniform column, \Ve write
P,
A=- (47)
El
and Equation (46) then becomes
d'y d 2y _
dx' + A dX2 - O, (48)
with general solution of the form

y = Cl sin -JIx + C2 cos ",/Ix + c,x + C,. (49)
In the special case when the two ends x = O and x = / are both hinged, the
conditions
y(O) = y"(O) = O
at the end x = O give c, + C4 = O and (.'2 = O, from which there follows
c, = c, = O, (50)
and Equation (49) takes the form
y = Cl sin -JIx + c,x. (51)
The conditions
y(l) = y"(I) = O
at the end x = / then give
Cl sin -JI! + c,/ = O, Cl sin -JI/ = O (52)
so that non trivial solutions of the problem exist only ir sin -JI/ = O and c, = o.
Thus the characteristic values of A are of the form
(n = 1,2, ... ), (53)
and the corresponding characteristic functions are the deflection modes

. nnx
'Pn ()
x = sm -/-. (54)
The values of the axial load P corresponding to these modes are known as the
critica/ buckling /oads, aod are given by (47) and (53) io the form
(n = 1, 2, ... ). (55)
For axial loads smaller in magnitude than the smaIlest critical value
_ 2 El (56)
Pr - 7r
f2'
the column is stable only in its unbent positioo. That is, if the column is artifi-
cialIy imparted a small amount of bending, it will tend to returo to its initial
5.5. The lVIethod of Stodola and Vianello 197
straight formo When P = PI' however, this theory prediets that under a small
disturbanee it wilJ assume a form
y =
. nx ,
e sm (57)
T
known as the fundamental buckling mode. Although the linear theory does not
prediet tbe amplitude e of the sinusoidal deformation, a more nearly exaet
analysis sbows that tbe amplitude will inerease with a small inerease in the axial
load P aboye the eritieal value PI' until bending failure of the eolumn oeeurs.
The eritieal value (56) is known as the Euler load of a eolurnn binged at botb
ends.
In practice, failure is not caused entirely by bending except for very long columns,
and the structure may fail before the Euler load is attained. However, the Euler
load does give an upper ¡imil of stability in the undeformed position.
The higher modes cannot be attained in practice unless additional con-
straints are introdueed.
If Ihe end x = O is hinged and the end x = l is fixed, we find, by imposing
the eonditions
y(l) = y'(l) = O
on Eguation (51) and introdueing the dimensionless parameter
(58)
that tbe eritical loads are then of the form
Pn = Ji?;~[' (59)
where Jin is a positive solution of the transcendental eguation

tan Ji = Ji. (60)
The eorresponding deflection modes are of tbe form
rp (x) = sin ~Jinxll) _~. (61)
n sm Jin l
Since the smallest positive root of (60) is approximately
Jil = 1.43n,
the Euler load in this case is approximately
PI = 2.05n 2 ~: . (62)
s.s.The Method of Stodola and Vianello. An iterative procedure, which is

sometímes known as the method of Stodola and Viane/lo, often is useful for the
approximale determinatíon of the ebaracteristic numbers and functions of a
boundary-value problem. Before outlining the application of tbis procedure to
problems of a rather general type, we nrst illustrate its use by considering a

prevjously sol ved problem .
In rhe case of a rotating string fixed at its ends (Secrion 5.2), the boundary-
vaJue problem consisting of the differential equation
d'
-1'. + Ay = O (63)
dX2
and the end conditions

y(O) = O, y(l) = O (64)
was found to have the characteristic values
n 2 n;2
An =12 (n = 1, 2, ... ), (65)
with corresponding characteristic functions
rpn = Sin _,_o

. nnx
(66)
The method to be presented is a method of successive approximations leading to

approximate expressions for the sma/lest characteristic value and the correspond-
ing characteristic function, although the method can be modified in such a way
that the higher modes are obtained.t
We first transpose the term involving A to the right in Equation (63), to
obtain
d2
---.l' = -Ay. (67)
dX2
Next, we replace the unknown function y on the right-hand side of (67) by a

convcniently chosen first approximation y, (x). It is preferable that this function
satisfy the prescribed end conditions, although this is not necessary. The result-
ing equation,
d2
---.l' = - Ay l
(x) (68)
dX2 '
then can be sol ved directly and the two constants of integration can be deter-
mined so that the solution satisfies the prescribed end conditions. It is clear
that the solution so obtained will contain the parameter A as a factor, and hence
y = Af,(x). (69)
Now, if the originaUy assumed function y, (x) were actually a characteristic
function of the problem, the function y(x) given by (69) would necessarily
become identical with the assumed function y,(x) if A were assigned the corre-
sponding characteristic value. That is, the corresponding characteristic value of
A would be given by the constant ratio y,(x)/f,(x).
tSee ProbJem 47 and Rererence l.

5.5. The Method of Stodola and Vianello 199
Thus, if we take
. 7tx
y¡ (x) = sm T'
the so lutioo of (68) for wruch y(O) = y(l) = O is readily found to be
, / 2 . nx
yx
() ="7t ,smT'
In order tbat y¡(x) = y(x), we must then have A = 7t 2 /f2, in accordance with
(65) and (66).
However, since y ¡ (x) is generally not a cbaracteristic function, tbe functions
y.(x) and¡;(x) will in general not be in a constant ratio. Sti11, as wi11 be sbown
in Section 5.9, if a convenient multiple of ¡; (x) is taken as a new approximation
y/x) and the process is repeated indefinitely, the ratio y.(x)lfix) tends toward
a constant value over the interval (O, 1) as tbe number n of cycles is increased,
and this constant vaJue is exactly the sma/lest characteristic value Al' . Also, the
sequence of successive functions y.cx), Y2(X), ... ,y.(x), ... converges to the
corresponding characteristic function.
Successive estimates of the characteristic value A¡ may be obtained after
each cycle by requiring that the functions y.(x) and y(x) = A!.(X) agree as well
as possible (in sorne sense) over the interval (O, 1). In particular, a simple pro-
cedure consists of determining A so that the integral, over (O, 1), of the difference
between the functions be zero. Thus the nth approximation to tbe sma11est
perm iss ible value of A is given by
A(.) -
r o
y.(x) dx
(70)
¡ - f~ !.(x) dx '
To illustrate this procedure in the present problem, we choose as our first

approximation the function
y¡(x) = x(l - x),
which satisfies tbe end conditions (64). If y is replaced by y¡(x) on the right-
hand side of (67), there results
d'y
- = A(X 2 - Ix)
dX2 '
and the solution of this equation satisfying (64) is found by direct integration
in the form
y = ¡A (x 4
2
- 21x' + I'x) = A!¡(X).
If the integrals of y¡(x) and y(x) = A!¡(X) over (0,1) are equated, there follows
l' l'
6 = AÓO'
from w hich the first approximation to lbe sma11est characteristic value of A is
A\¡) = ~~.
Comparison wirh the exact value 70 2 // 2 = 9.870// 2 shows thar the error is about
1.3 percent.
If now we take
y,(x) = x, - 2/x' + /'x
as the initial approximation of the next cycle, we obtain
A .
y = Af2(X) = - 3U(x 6 - 3/x' + S/'x' - 3/'x),
and Equation (70) gives the second approximation to A¡,
A12 ) = 9},82,
which differs from the exact value by abou r 0.12 percent.

The procedure as illustrated in this problem applies without essential modi-
fication to somewhat more involved boundary-value problems. Thus, for exam-
pie, we may have differential equation s of the form
1xCp (x) ~~] + Ár(x)y = O (7 la)
or d2[d
dX
2 2y
p(x) d ] + Ar(x)y -_
X2 O, (7lb)
with appropriate homogeneous end conditions. If the ter m involving A is trans-

posed to the right and if the factor y in its coefficient is then replaced by a
suitable approximation y ¡ (x), the resultant differential equation can be solved
by direct integration for a new function of the form y = Af¡ (x) and the remain-
ing procedure outlined aboye is directly applicable.
In place of using Equation (70) to determine successive approximations to
Al> one may make use of either of the following formulas (see Section 5.9):
s: r(x)[Yn(x)J2 dx
A1¡") -
r r(x)fnCx)Yn(x) dx
(72a)
AIn) -
r r(x)f.(x)y.(x) dx
s:
or a (72b)
r(x)[fn(x»)2 dx
Either of these formulas will in general give a better approximation to the tme
value of A¡, the last form being in general the most accurate.t In illustration,
tIt can be shown that . in the genera l case of (7 \ a) or (71 b), the result of using (72a) in the
11th cycle is co mparable with the result of u s ing (70) in the (2n) th cycle, while the result oE
using (72b) in the 11th cycle is compara ble with that oE using (70) in the (2n + t )th cyc\e.
In the case o[ Ola), if y, is suc h that (py;)' = const ant, these pairs of approximations are
respect ively equaL (See Problem 43.) The 'ame is true of (7 1b ) if (py;')" = cOnstant.
5.5. Tbe Method oC Stodola and Vianello 201
we list the results of using the three formulas to estimate 12 A, in tbe preceding
example.
n (70) (na) (nb)
1 10.00000 9.88235 9.87096

2 9.8 8235 9.86975
00 9.86960
The tabulated values of 12 A\") show that in this case only one approximation,
using (72b), is needed to obtain an approximate value of A, correct to within 2
parts in 10,000.
As a further application we investigate the Euler load of a colurnn of length
1, so constructed that its bending stiffness factor El is of the form
EI(x) = C ~ , (73)
",here C is the maximum stiffness factor (at x = 1). We suppose that both ends
of tbe column are hinged. Then the bending moment M(x) at a section x is due
entirely to the axial load P and is given by M = -Py(x). Hence, from Equation
(28), the relevant differential equation is of the form
x d 2y _
C
T dX2 + Py - O. (74)
(See also Problem 15.) If we introduce tbe dimensionless parameter

PI2
J1.2 e'
= (75)
Equation (74) beco mes
d y
X dX2
2 J1.2_
+ -I- y - O. (76)
The end conditions are of the form

y(O) = O, y(/) = O. (77)
According to Equation (l30d) of Chapter 4, the general solution of (76) is
of the form
y JxZ, (2J1.j
= n
or y = c,JxJ,(2J1. j;) + c2--1 xy ,(2J1.j;) .
Since, for smaIl values of x, we have
__ --1',
nJ1.
the c o ndition y(O) = O requires that
and tbe rema ining condition y (l) = O requires that ¡J be a solutio n o f the equa-
tion
J,(2¡J) = O. (78)
A rouDded value of the s malIest positive root of this equation (se e Table 1 of
Section 5.13) is
¡J, = 1.916. (79)
Correspondingly, the Euler load is given by
_ z e _ 3 670 e
P , -¡J,--¡z (80)
-. 12'
and the fundamental buckling mode is given by
q¡(x) = ",/ xJ, (3.832J ~ ). (81)
To find aD approximale value of ¡J, we take as our first approximation the

function
y,(x) = xCI - x). (82)
From (76) we then have
d 2y ¡J2
- , = -1 (x - 1),
dx
,
y = ~¡<X 3 - 31x' + 2J2x) - ¡J'f,(x). (83)
The condition
s: y ,(x) dx = ¡J' J~ f,(x) dx (84)
then gives, as a first approximation,

,, ( 11 -
rl -
2, (85)
leading to an approximation
PI' ) = 4~ (86)
l'
for the Euler load. A second approximation gives
(87)
If the condition
J~ r(x)[Yn(x)p dx = ¡J' J~ r(x)fn(x)y.(x) dx
with r(x)=~,
x
5.6. Orthogonality of Characteristic Functions 203
corresponding to (na), is used in place of (84) to determine approximatioDs to

¡.t" the improved values ¡.t\' ) = 1.936 and ¡.t\' ) = 1.917 are obtained. Use of
(nb) leads to the successive approximatioos 1.922 and 1.916.
An investigation of the convergence of the iterative procedure to rhe lowest
characteristic mode, as well as ao indicatioo of modifications leading to the
determination of higher modes, is preseoted io Section 5.9 .
5.6. Orthogonality of Characteristic Functions. Two functioos 'Pm(x) aod

rpn(x) are said to be orthogonal on an interval (a, b) if the integral of the product
rpm'P. over that interval vanishes:
r rpm(x)rpn(x)dx = o. (88)
More generally, the fuocrions rpm(x) and rp.(x) are said to be orthogonal with
respect to a weighting function r(x), on an interval (a, b), if
r r(x)rpm(x)rp.(x) dx = o. (89)
Finally, a set of fuoctioos is said to be orthogooal 00 (a, b) if al! pairs of distinct

functioos in the set are orthogonal 00 (a, b) .
An extremely useful property of bouodary-value problems of a rather gen-
eral type consists of the fact that the sets of characteristic functions correspond-
ing to such problems are orthogooal with respect to a weighting function. In
order to establish this fact, we consider first the boundary-value problem con-
sistiog of the linear homogeneous second-order differeotial equation
~ [p(x) ~~J + [q(x) + Ar(x)]y = O (90)
and suitably prescribed homogeoeous bouodary cooditions, to be specified

preseotly, at tbe eods of ao interval (a, b). The functioos p, q, and r are assumed
lO be real. lo terms of the operator
L = fx (p fx) + q p :;2 + ~~ fx + q,
= (91)
Equatioo (90) can be written in the abbreviated forro

Ly + Ar(x)y = o. (92)
We notice that any equalion of Ihe form
ao(x)dd2~
x
+ a,(x)ddYx + [a2(X)
·
+ Aa,(x)]y = O (93)
can b e wrilten in Ihe form of (90) by setting
p = ef a .tlJo d ;c , q = a?p, r = a,p. (94)

ao ao
Suppose now that A, and A2 are any two different characteristic values of
the problem considered and that the correspondiog characteristic functioos are
204 Boundary- Value Problems and Characteristic-Function Representations
y = tpl(X) and y = tpz(x), respectively. There then follows
d ( P dtp¡)
dx dx ,.I ( q ..L
I
')
Alr tpl -- O,
(95)
ix (p":fx2) + (q + A2 r )tp2 = O.
If the first of these equations is multiplied by tp2(x)and the second by tp¡(x),

and the resultant equations are subtracted from each other, there follows
and hence
(A 2 - A¡) f rtpltp2 dx = f [ tp2ix(p":fx') - tp¡:x(p1:)]dx. (96)
Integrating the right member by parts, we obtain
(A2 - A¡) f rtpjtp2 dx = [tp2(p":fxl) - tpl(p1:)I
_lb a
[dtp2(p dtp
dx dx
1) _ dtp
dx
1(p dtp2)]
dx
dx.
Since the last integrand vanishes identically, there follows finally
(Az - A¡) f rtp¡tp2 dx = [p(x){tp2(x)dtpd~x) - tpJx)dtpd~x)}l (97)
In view of the fact that both tp ¡ (x) and tp2 (x) satisfy the conditions prescribed
in connection with (90) at the points x = a and x = b, the right-hand member
of (97) clearly vanishes if at each end point a prescribed condition is of one of
the forms
y=O (98a)
or dy = O (98b)
dx
or y + ydy = O (98c)
dx
when x = a or x = b. Here the vanishing of the right-hand member of (97) at
an end point for which tptCx) and tp2(X) satisfy (98c), for any value of y, follows
from the identity
tp2tp', - tp¡tp~ - (tp2 + ytp~)tp'¡ - (tpl + ytp'¡)tp'2'
Further, if it happens that
p(x) = O when x = a or x = b, (99)
then vanishing of the right-hand member of (97) at x = a or x = bis assured if
only y is required to be finite at that point and if either dy/dx is finite or p dy/dx
tends to zero at that point.
5.6. Orthogooality of Characteristic Functions 20S
Finally, if
p(b) = pea), (100)
the right-hand member 01' (97) vanishes ir the conditions
y(b) = y(a), y'(b) = y'(a) (101)
are satisfied.t It should be noticed that (I01) will be satisfied, in particular, ir
the solutions /P, (x) and /P2(X) are required to be periodic, 01' period b - a.
In the cases listed, there follows
r r(x)q¡,(x)/p,(x) dx = O (102)
that is, if the characteristic functions correspond to different characteristic num-

bers, then they are orthogonal with respect to the function r(x).
A boundary-value problem consisting of a differential equatio[).. of type (90),
together with two homogeneous conditions of the types noted, is called a Sturm-
LiouvilIe problem. In all the cases considered here, except only when the period-
icity conditions (lOO) and (101) are involved, it can be shown that to each
characteristic number An there corresponds only one characteristic function /Pn(x).
(See Problem 27.) From the homogeneity of the problem it is clear, however,
that each such function involves an arbitrary multiplicative factor.
Example l. We have seen that the problem

d'y
dx
2 + Ay = 0, y(O) = y(l) = °
has the characteristic numbers An = n 2 n 2 /f2, with corresponding characteristic func-
lions proportional 10 the functions /P. = sin (nnx/l). Since in this case r(x) = 1, there
follows from (102)
I /Pm(x)/Pn(x) dx = I sin ~x sin n~x dx = O (m"" n),
when m and n are positive integers. This faet is readily verified independently by direct
integration. •
In most appLications, the functionsp(x) and r(x) in Equation (90) arepositive

throughout the interval (a, b), including the end points, while the function q(x)
is nonpositive in that interval. Furthermore, when one or both of tbe end con-
ditions is of type (98c), with y "" O, normally y is negative when the condition
is imposed at the lower limit x = a but is positive when the condition is imposed
at tbe upper l.imit x = b. We will speak of a Sturm-Liouville problem satisfying
these restrictions as a proper problem.
tBoundary conditions such as those of (101), which each involve data at /wo boundaries, are
sometimes called mixed bOtmdary condi¡ion::. [The same phrase is also used to describe a condi-
tion such as (98c), which relates ¡he fimc¡ion and i¡s deriva/ive at one boundary.) More general
mixed conditions of two-point type, which also lead to orthogonality when (lOO) holds, are
considered in Problem 26.
206 Boundary~Value Problems and Characteristic-Function Representations
For a proper Sturm-Liouville problem, it is found (see Pro blems 28 and 29)
that a ll ch aracteristic numbers are real and nonnegati ve, and that th e correspond-
ing charac teris tic functions are real (or can be made real by rejecting a possible
comple x constant multiplicative factor) . In addition, it can be s hown in thi s case
that there are infinitely many characteristic numbers, that they are discret ely
d is tributed (and hence do not fill out any interval), and that their array is
unbounded .
Furthermore, the integral of the weighted square of a characteristic function
qJn(x),
( 103)
then has a pOSltIve numerical value, known as the norm (or sometimes as the
square of the norm) of qJn' with respect to the weighting function r. If the arbi-
trary multiplicative factor involved in the definitio n of. qJn(x) is so chosen that
this integral has the value unity, the function qJn(x) is said to be norma/ized with
res pect to r(x). A set of normalized orthogonal functions is said to be or/ho-
normal.
Example 2. In the aboye Example 1, we find by direct integration that
Hence, in order to normalize the functions si n Cn7T.xll) over (0, 1), we would divide them
by the common normalizing factor ,.,¡ 1/ 2. The set of functi o ns
'Pn*C)
x = JT
2 .sm -,-
n7T.x (n = ),2, . . . )
is thus an orthonormal set on (O, 1) .
Orthogonality properties of sets of characteristic function s generated by

•
certain boundary-value problems of higher order are readily established. Thus,
in the case of a problem invol ving the four/h- order differential equation
2 2
d ] + dx
d [ s(x) d!z
dX2 d [ p(x) dd~ ] [q(x) +
Ar(x)]y = O + (104)
and appropriate hom oge neou s boundar y conditions, the methods used abo ye
lead to the equation
(..1. 2 - A,) frqJ,qJ2dx = [(qJ2(SqJ ';)' - qJ,(sqJ ~n

- s(qJ~qJ': - qJ ',qJ~) + P(qJ2qJ', - qJ,qJ'2)]~' (lOS)
where qJ ¡(x) and !P 2(X) are characteri s tic functions corresponding to distinct
characteristic numbers A, and ,1.2' and primes indicate differentiation with re-
spect to x. It follows, in particular, that the functions qJ, and qJ2 are orthogonal
5.7. Expansion of Arbitrary Functions in Series of Orthogonal Functions 207
with respect to r(x) on (a, b) if at each end of the interval there IS prescribed
one of the following pairs of homogeneous conditions:
y = O, dy = O (l06a)
dx
O, d2y
or y = s(x)- = O (106b)
dX2
dy = O 2y
or -d [ s(x)-
d ] = O (l06c)
dx ' dx dX2
when x = a or x = b.
5.7. Expansion of Arbitrary Functions in Series of Orthogonal Functions.

Suppose that we have a set of functions [IPn(x»), orthogonal on a given interval
(a, b) with respect to a certain known weighting function r(x), and desire to
expand a given function f(x) in terms of a series of these functions, of the form
f(x) = AolPo(x) + A,IP,(x) + A2IP2(X) +
(107)
If we assume that such an expansion exists, and multiply both sides of (107)
by r(x)IPk(X), where IPk(X) is the kth function in the set, we have
=
r(x)f(x)IPk(X) = L:o Anr(x)IPn(X)IPk(X).
n~
Next, if we integrate both sides of this last equation over the interval (a, b) and
assume that the integral of the infinite sum is equivalent to the sum of the
integrals,t there follows formally
lbr(x)f(X)IPk(X) dx
a =
=
n~ An
lb r(x)IPn(X)IPk(X) dx.
a (108)
But by virtue of the orthogonality of the set (IPn(x)}, al! lerms in the sum on the
right are zero excepl that one for which n = k, and hence (108) reduces to the
equation
An r r(x)[IPn(x»)2 dx = r r(x)f(x)IPn(x) dx. (109)
Thus Equation (109) determines each constant involved in the required series
(107) as the ratio of integrals of known functions.
With these values of the constants, a formal series L: AnIPn(x) is determined.
It should be emphasized, however, that we have not established the fact that
this series actually does represenl the function f(x) in the interval (a, b). In fact,
we have not even shown that the series converges in (a, b) and hence represents
any function in that interval. We may, however, speak of the series so obtained
as the formal representation of the function f(x).
tThis assLlmption is justified, in particular, ir the series (107) is uniformly cOllvergenl in (a, b).
208 Boundary-Yalue Problems and Characteristic-Functioll Representations
Examp/e l. In illustration, we have seen that the set of fune[ions [sin (nnx//)} is
orthogonal on (O, 1), with the weighting funetion r(x) = 1. In Ihe formal representalion
of the funelion ¡(x) = x in a series of these funetions, the eonstants An are given, in
aecordanee with (109), by the equation
l • ? nnx
JI . n7tx
An
r s1o- -/-dx
~ o
=
o
x sm -/-dx
2/ 2/ (- 1)n+ ,
or A
n
= --cosnn
nn =-
n n .
Henee we obtain the formal expansion

_ 2/ ~ (- I)n+ 1 . nnx
x - - ¿",
n n=' n/n
2/ (1 . nx
l /
l . 2nx
s10 - - = - - s10 - - - sm - -
2 /
+
•
The general problem of determining whether the formal expansion obtained
aetually converges lo the given function in the interval involved is a relatively
difficult one and is beyond the scope of this work. In connection with this
problem, it may be noticed that if the function f(x) to be expanded were not
identically zero in (a, b), but were orthogonal to al! the functions 'Pn of the set,
with respect to the weighting function r(x), all coefficients in the formal expan-
sion would vanish and no expansion would be obtained. However, it can be
shown that, except for artificial funetions of no physical interest, there are no
nontrivial functions which are orthogonal to al! members of any set generated
as characteristic functions of a proper Sturm-Liouville problem. In this sense
we say that the sets so generated are complete.
If the Sturm-Liouville problem is proper, and if also the relevant functions
p(x), q(x), and r(x) are regular in the interval (a, b), then it is known that the
formal representation of a pieeewise differen/iablet fune/ion f(x) in a series of the
generated eharaeteristie funetions converges to f(x) inside (a, b) at al! points
where f(x) is eontinllous, and converges to the mean value ·Hf(x+) + f(x-)]
at points where finitejumps oeeur. In the sequel, to simplify the notation, we wiIl
tacitly assume that, when a function f(x) has a finile jump at an interior point
x = e in (a, b), the mean of the right- and left-hand limits at that point is
definedto be f(e), so that the series also converges to f(x) at such points.
Certain cases in which one or both of the functions p and r vanish at an
end point of the interval are also of importance, as has been noted; although
they require individual treatment, it has been shown that in the special cases to
be treated here the convergence of the formal representation is again described
by the aboye statement.
When the conditions imposed on the generating Sturm-Liouville problem
require that the characteristic functions vanish at one or both end points of the
t A funclion f(x) is said to be piecewise differen/iable in (a, b) if that interval can be divided
into a finile number of subintervals in such a way [hat, in each subinterval, f(x) has a deriva-
tive at each interior point, a right-hand derivative at the initia! point, and a left-halld derivative
at the terminal poin!. Such a function thus may have a finite number offinite "jumps" in (a, b).
5.7. Expansion of Arbitrary Functions in Series of Orthogonal Functions 209
interval (a, b), or relate their values at the two ends [as in ( 101)], convergence to
f(x) at ¡he end point or points in question wiU ensue only if f(x) also satisfies
the respon sible conditions.
Examp/e 2. In the case of Example 1, the lheorem stated asserts thal ¡he expansion
obtained converges to the function [(x) = x at all points inside the interval (0, 1). Since
all terms in the series vanish at bOlh end points x = °
and x = l. whereas f(x) vanishes
only at x = 0, it is see n that the series also represe nts [(x) = x at x = O, but does not
represent it at the second end point x = l. From the convergence at the point x = 1/2
we obtain, in particular,
and hence establish the useful result
•
It is important to notice the very great generality of such expansions. In
the case of the power series expansions of Taylor and MacIaurin, a function
cannot be representable over an interval unless thatfunction and all its derivatives
are continuous throughout that interval, and even these stringent conditions
are not sufficient to ensure representation. However, in the present case of
expansions in series of characteristic function s, a represe ntable function may
itself possess a finite number of fínite discontinuities, and may even be defined
by different analytical expressions over different parts of the interval of repre-
sentation.
On the other hand, whereas it is true that a power series representation of
a function f(x) can be differenliated term by terro with the result assuredly
converging to f'(x) at all poiots inside the interval of convergence of the parent
series, the same statement does not apply to the present expansions. If the
generating Sturm-Liouvi lle problem is proper, it is known, however, that if f(x)
is continuous and has a derivative f'(x) which is piecewise dijferentiable in (a, b),
and if f(x) s atisfies the boundary conditions relevant to the generating character-
istic-value problem, then th e res ult of dijferentiating the series (107) term by term
will converge to f'(x) inside (a, b) at all points where f'(x) is continuous. When
the conditions specified in this statement are n ot all satisfied, term-by- term
differeotiatioo ma y or may not be legitimate. (See also Section 5.12.)
Examp/e 3. If the series representat io n

21 ~ ( - 1)n+ 1 • n1CX
x = - ¿;
1!n at n
SI n - -
I
( O <: x < 1),
obtained in Example J, is formally differentiated term by term, the queried resu lt

? nnx
1 ~ 2 ¿;
QQ
( -I)n+ 1 cos--
n-l 1
210 Boundary-Value Problems and Characterislic-Function Representations
is 1701 valid onywhere, since [he nth term of the series does nOI tend to zero as n ~ CQ
and hence the series diverges everywhere. Here Ihe function f(x) = x violales only
the requirement that i t vanish at x = l. •
An operation which is rather frequently needed (as in the two following

sections), and which involves [IVO differentiations, is the term-by-term app1ica_
tion of the operator L defined by (91) to the associated series (107) . In order to
investigate its justification, we notice that the result of formally evaluatiog Lf,
by aUowing L to operate on the series (107) term by term, is the expression
Lf(x) ~
-
I:o AnLrpn(x) = -r(x)
~
I:o AnAnrpn(x), (110)
II=" /1 -
since Lrpn(x) + A"r(x)rpn(x) = O. On the other hand, if [Lf(x)J /r(x) is piecewise

differentiable in (a, b), then its true representation as a series of the rp's is of ¡he
form
Lf(x) =
( ) = I: B"rp"(x), ( 111)
r x 11=0
where
Bn f brrp;; dx = fb r -rpn
a
Lf
r
dx
a
= fb rpnLf dx
a
or Bn r rrp; dx = r rpn(Pf')' dx + r rpnqf dx. (112)
If the first term on the right is integrated twice by parts, the result can be written
in the form
B" r rrp;; dx = [p(rpJ' -frp~)J~ + f."fLrp" dx
or, equivalently,
B" l
a
b
rrp',; dx = [p(rpJ' -frp~)J~ - A" 1 rfrp"dx.
-b
(113)
Hence it follows that B" = -A"A", where A" is defined by (109), and accord-
ingly that (I 10) properly agrees with (111), if and only if f(x) is such that
(p(rpJ' -frp~)J~ = o. (114)
Thus we conclude that if [Lf(x)J/ r(x ) is piecewise differenliable in (a, b), and if
f(x) satisfies the end conditions relevanl lO the generating proper characteristic-
value problem, then the operalor L can be applied term by term to the series (107):
L
- A"rp.(x) = I:- AnLrp"(x) =
I: - r(x) I:
~
A"A"rpn(x). ( 115)
11 = 0 11 = 0 11 = 0
The series (llS) theo represents Lf(x) in the usual sense inside the interval
(a, b) and may or may not converge to Lf(x) at the end points.
Expansions of the type considered here will be of particular usefulness in
the solution of boundary-value problems involving partial differential equalions,
as wilI be seen in Chapter 9 . We next point out certain other useful applications
of the theory.
5.8. Boundary-Value Problems Involving Nonhomogeneous Differential Equations 211
5.8. Boundary-Value Problems Iuvolving Nouhomogeueous Differential Equa-

tions. Suppose that we require the solution of the nonhomogeneolls differential
equation
[fx (p ~~) + qy] + Ary = h(x) ( 116)
which satisfies prescribed homogeneous boundary conditions of the types listed

in Section 5.7, with A a prescribed constant. Using the operational notation of
(91), we may write (116) in the form
Ly + Ary = h(x). (117)
The most important special case is that in which A = O, so that we require the
solution of the equation Ly = h(x), in which case the function r(x) can be
chosen at our convenience unless other considerations dictate the choice.
With the given problem we may associate the boundary-value problem
consisting of the homogeneolls equation
Ly + Ary = O ( 118)
and the prescribed boundary conditions. This problem determines a set of
orthogonal characteristic functions {<Pn(x)}, such that the nth function <Pn(x)
corresponds to a characteristic number An ,
L<p.(x) + Anr(x)<p.(x) = o. ( 119)
Assuming that the required solution of Equation (117) exists, we express it
as a series of the form
(120)
and attempt to determine the coefficients an0 For this purpose we introduce
(120) into the differential equation (117). Then, since (119) glves
L<pn = - Anr<pn'
Equation (117) becomes formally
r(x) L: (A - An)an<Pn(x) = h(x). (l21)
Hence, if we use (109) to calculate the coefficients An in the expansion
f (x) -- r(x)
h(x) -- '"
....... An<Pn ( x ) , (122)
assuming that h(x)/r(x) is piecewise differentiable, a comparison of Equations

(121) and (122) determines the required coefficients a n as solutions of the equa-
tions
(A - An)a n = A n, (123)
so that, provided that A is not one of the characteristic numbers, the solution
(120) of the nonhomogeneous problem becomes simply
y -_ '"
....... A -An A <Pn()
x _- A Ao
_ A <Po ( x ) ,T A Al
_ A <P 1( X ) + (124)
n o 1
The term-by-term differentiation of (120) which leads to (121) is justified ,

when the Sturm-Liouville problem is proper, by reference to the result obtained
at the end of the preceding section, since the function y in (120) satisfies the
relevant boundary conditions as well as the equation
Ly = _Ay + .!!...,
r r
where h/r has been assumed to be piecewise differentiable. The particular excep-
tional cases which wi\l be encountered in later sections require special treatment,
but justification again is possible. In fact, because of the growth of the factor
A - A. o in the denominator of (124), as n ~ 00, the solution series (124) often
will converge even when the series (122) does not.
It is immediately verified that the aboye derivation, beginning with Equation
(117), is unchanged if the operator L in (117) is interpreted as the fourth-order
opera/or
L =
2
dX2 d J+
2
d [ s(x) dX2 dx dJ +
d [ p(x) dx q(x) (125)
in Equation (104), and if boundary conditions of the type listed III (106) are
then assumed.
From these results one may draw certain important conclusions. We have
seen that if h(x) is identically zero, the problem consisting of (116) and the
prescribed homogeneous boundary conditions has non trivial solutions only if
A has a characteristic value A = A.k' Equation (124) shows that if h(x) is not
identically zero, the corresponding problem has a solution in general only if A
does not take on one of the characteristic values of the homogeneous problem.
More specifically, the solution (124) becomes nonexistent as A - A.k unless it
happens that Ak = 0, that ¡s, unless
r r(x)J(x)!Pk(X) dx = r h(X)!Pk(X) dx = O,
so that h(x) is simply orthogonal to !Pk(X). In that special case, Equation (123)
shows that the coefficient of 'Pk(X) in (120) is arbitrary.
In connection with the physical problems of rotating strings and shafts (Sections
5.2 and 5.3), the presence of a prescribed function of the axial distance x on the
right-hand side of the relevant differential equation would correspond to the
presence of a distribution of transverse load. The aboye conclusions indicate
that, for noncritical rotation speeds, definite deflection shapes are determined,
but that (according to the linearized theory) as a critical speed is approached, the
amplitude of the defle~tion will in general increase without limito
5.9_ Convergen ce of tbe Metbod of Stodola and Vianello. We may also make
use of these developments in investigating the convergence of the iterative
methods of Stodola and Yianello (Section 5.5), as applied to the Sturm-Liouville
problem (with q = O) consisting of equation (71a),
fx[p(x) ~~J + A.r(x)y = O, (126)

s 5.9. Convergence oC Ihe MelhOO oC 51000la and Vianello 213
1, and appropriate homogeneous boundary conditions. We suppose that the prob-

d lem is a proper one. In addition, we assume that 1 = O is not a characteristic
e number. The first cycle in the procedure consists of assuming an initial approxi-
mation y,(x) and determining the solution of the differential equation
dx d y ] = -lr(x)y,(x),
d [ p(x) dx (J27)
)-
which satisfies the bouodary cooditions. The initial approximation may be
t, iroagined as expressed in terros of a series of the exact characteristic fuoctions
lr of the problern, say
n ~
y,(x) = I:
TI= 1
A.rp.(x), (128)
o
?r where ~[p(x) drp.]
dx dx
= -1 •
r(x)rp (x)
•
(n = 1, 2, 3, ... ) (129)
and where O < 1, < 1 2 < ....

i) The solutioo of (127) can aIso be written in the form
~
re y(x) = I: anrp.(x). (130)
"=1
le To determine the coefficients in (130), we introduce (130) into (127) and obtaint
le
if
~ d [ <!.P..,,] _
n~ a ndx p(x) dx -
_ 1r(x).~
~
A.rp.(x),
?t
or, using (129),
i\. ~ ~
n. - I:
n~ I
a.1.r(x)rpn(x) = -lr(x) I:
n== I
A.rp.(x).
it
Hence the coefficients in (130) are determined in the form
1
a. = yAn (131)
•
3) aod (130) becomes
~ A
y(x) = 1 .~ l:rp.(x)
or y(x) = l[trp,(x) + trp2(X) + 1:rp,(x) + ... J

= 1 [ A,rp,(x)
I;" + 11~ A 2rp2(X) + 11: A 3rp3(X) + ....
] ( 132)
If the last expression in brackets is taken as the initial approximation Y2(X)

in the second cycle, the resuIting solulÍon is found, by a repetition of the aboye
,e process, in the form
ve
le y(x) = ; , [ A,rp,(x) + (tr A 2 rp2(X) + (~y A,rp3(x) + ... }
6) tTerm-by-term application of (he operator

end of Section 5.7.
:x (p :f.x) is justified by the result oblained at Ihe
21'¡ Boundary-Value Problems and Characteristic-Function Representations
and, similarly , In the Nth cycJe we have the initial approximation

A),vo, +
(AA: )NO' +
Y v(x ) = A ,'P,(x) ( A~ A,'Pl(X) AJ'PJ(x) (133)
and the corresponding solution
y(x) = ).fN(X) = : . [ A,<p,(x) + (tr A!Ip,(x) + (l:)"A J 'P,(x) .. .. -l (134)
Since A, is the smal/esl characteristic number, the ratios A, /Az, A, /A" ... ,
are smaller than unity and increasing powers of these ratios tend to zero as
N ~ =, showing that successive approximations always tend to a multiple of
the characlerislic funclion corresponding 10 Ihe smallesl characlerislic number A,
unless A, = O, that is, unless Ihe inilial approximalion y,(x) is orthogonal lo
'P ,(x) wilh respecI lO r(x). 1t is also cJear that the ratio YN(x)/fN(X), where
'!'v(x) = y(x)/A, then tends to the limit A" as was stated in Section 5.5.
I! is seen further tha t, if the condi tion
r r(x)Ip,(x)y(x) dx = r r(x)'P,(x)y,v(x) dx (135)
is imposed in the Nth cycJe, there follows from the orthogonality of the system
TA A , lb r(x)[<p,(X)P dx = A, J·-b r(X)[Ip,(x)J2 dx

, a a
or A = A,.
Hence, if the characteristic function Ip ,(x) were known exactly in any cycJe of
the calculation, Equation (135) would identify A with the exact value of A,.
Although 'P,(x) is not known exactly, an approximate determination of A, may
be accomplished by replacing Ip ,(x) by its approximation YN(X) or fN(X) in (135).
This consideration is the motivation of Equations (72a, b).
Modifications of the iterative procedure which permit numerical determi-
nation of the higher characteristic modes depend essentially on approximate
methods of "subtracting off" the terms involving the lower modes in successive
cycJes.
5.10. Fourier Sine Series aod Cosioe Series. Sínce the boundary-value prob-
lem
y(O) = O, y(l~ O (136)
has the characteristic functions

. nnx (137)
Ipn ()
X = sm -1- (n= 1,2,3, ... )
corresponding to the characteristic numbers An = n2n' /I', the results of the

5.10. Fourier Sine Series and Cosine Series 215
preceding section 5.7 state that, if a function f(x) can be represented by a series
of the form
x = A I Sln
f() . -nx + A 2 Sln
. -
2nx...l...
- I .••
/ /
~ A . nnx
-.:.... n Sln - / - (O < x < /), (138)
n= I
then the coefficients An are given, in accordance with (109) with r(x) = 1, by
An L sin 2 n7 dx = Lf(x) sin n~x dx.

x
(139)
This formula is a consequence of the orthogonality of the characteristic func-

tions 'Pn = sin (nnx//) with respect to the weighting function r(x) = 1,
Jor
l
. mnx . nnx
Sln ' - / - Sln - / - dx = ° (m *- n). (140)
Since, by direct integration,

I
Jo
. 2 nnx
sm -/-dx = ¿
/
(n=1,2,3, ... ), (141)
there follows
An = - /
2 r sm nnx
l
J/(x) -/-dx.
.
(142)
It may be noticed that the coefficient An in (142) is given by twice the average
va/ue of the product f(x) sin (nnx//) in the interva/ (O, /).
The average value (with respect to x) of a function F(x) in (a, b) is defined by
the integral
F = b _
1 a Jb F(x) dx.
a
I.f f(x) is piecewise differentiab/e in the interva/ (O, /), the series (138) con-
verges to f(x) at points of continuity, and to the mean of the two va/ues approached
from the right and left at afinite discontinuity.
The series (138) is known as the Fourier sine series representation of f(x)
in the interval (0,/). It is seen that aH terms or "harmonics" in (138) are periodic
and have the common period 2/; that is, the common period is twice the length
of the interval of representation (O, 1). AIso, if x is replaced by -x, the algebraic
sign of each harmonic is merely reversed. Functions having this 1ast property
are known as odd functions of x.
In general, a function F(x) is said to be an odd function if F( -x) = -F(x) and
an even function if F( -x) = F(x). In a graphical representation, y = F(x), an
even function is symmetrical with respect to the y axis, whereas an odd function
is antisymmetrical; that is, the graph of an odd function is symmetrical with
respect to the origin of the coordinate system. Thus the functions sin mx and xk,
216 Boundary-Yalue Problems aud Characterislic-Fuuclion Represeulalions
where k is an odd integer, are odd functions, and the functions cos mx and Xk ,
where k is an even integer, are even functions. Such functions as eX and log x are
neilher evell llar odd. 1t is eJear that the product of two even functions or of two
odd functions is an even function, whereas the product of an even function and
an odd function is an odd function. Further, the derivative of an even function
is an odd function and converse ly; the integral of an odd function is an even
function, whereas the integral of an even function is the sum of a constant and
an odd function.
Jt follow s tbat in tbe interval (-1, O) tbe series (138) represent s the function
- fe-x). Since all terms have the common period 21, tbe bebavior of the series
in (-1,1) is repeated periodicaUy for aH values of x.
Jf f(x) is an odd funetion of x, tbe series (138) accordingly represents f(x)
not only in the interval (O, 1) but also in the larger interval (-1,1) . Jf, in addition,
f(x) is periodie, of period 21, the series represents f(x) everywhere (witb the
usual convention regarding points of discontinuity) .
Thus the sine series represe ntation of tbe odd function f(x) = x, given as
an illustration in Section 5.9, represents x in the larger interval (-1,1) except
at the end points x = ±l, and repeats this behavior periodieally for aH values
of x (see Figure 5.2).
/
./ '"
'" '" '" '" '"
- 3/
"'"
/ ' - 2/
'" '"
/'2/ 31 x
" " "
'" " '"
" " ""
Figure 5.2
Example 1. To obtain the Fourier sine series representing ¡(x) = eX in the interval
(0,71:), we set 1 = 71: in Equatiens (138) and (142) and obtain
An = -
71:
21" o
e' sin IIX dx = - 2 2:
71:11
1 (1 - en ces 117t).
Thus the desired expansion is of the fonn

2 [en + 1 . 2(e n - 1) . 3(e n + 1) . 3
e' = n 2 Sln x - 5 Sln 2x + 10 Sin x + ... ] (O <x < 71:).
(143)
The series (143) represents e·' when °< x < 71:, and vanishes at the end points x = O
5.10. Fourier Sine Series and Cosine Series 217
and X = 71:. In the interval -71: < x < O the series represents the function -e-'" and,
in general, [he series represents an odd periodic function which coincides with eX when
O < x < 71:, with -e- when -71: < x < O, which is zero when x = -n, O, n , and
X
which is of period 2n. •
Similar series developments involving cosine terms, rather than sine terms,
may be obtained by considering the boundary-value problem
d 2y
dX2 + Ay = 0, y'(O) = 0, y'(l) = 0, (144)
which has the characteristic functions
( ) = nnx (n = 0, 1, 2, ... ) (145)

!po:c cos -,-
corresponding to the characteristic numbers Ao = n 2 n 2 /F. It is important to

notice that If'o(x) = 1 is a mem ber of the set (145), corresponding to Ao = O.
The functions (145) possess the orthogonality property
Jor cos mnx

l
cos nnx dx =
/
° (m =7'= n) , (146)
as can be independently verified by direct integration. The constants in a series

represe ntation of the form
xn 2nx
f(x) = Ao +
Al cos-/- A 2 cos-/-+ + ",
= Ao + n~- Aocos-/-
nnx (O <: x <: /) (147)
are given by (109), with r(x) = 1, in the forms
Ao L dx = Lf(X) d:c, An L cos 2 n~x dx = Lf(X) cos n~x dx.

{1/2 (n=l,2,3, ... ),
Since
there follows
J I
0 / /
nnx
cos 2 --dx =
(n = O),
(148)
Ao =
1
T rl
J/(x) dx, A. =
2
T
r' nnx
J/(x) cos -/-d:c (n = 1,2,3, ... ). (149)
Hence the coefficient A o in (147) is the average va/L/e oll(x) in (O, /), whereas the
coefficient An, when n =7'= O, is twice the average va/ue 01 f(x) cos (nnx//) in
(0, /).
With these values of the coefficients, the series (147) is known as the Fourier
cosine series representation of f(x) in (O, /) . 1t is seen that a11 harmonics in (147)
are even functions of x and are again periodic, of period 2/. Thus, if f(x) is
piecewise differentiable in (O, /), the series (147) will c onverge to a function f*(x)
which is periodic, with period P = 2/, and which is such that f*(x) = f(x)
218 Boundary~ Value Problems and Characteristic-Function Representations
for °-<: X -<: ( and f*(x) = f( -x) for -1 -<: X -<: 0, again with the adopted
convention regarding discontinuities. Since f " (x) = f(x) when X = and when
x = 1, tbe series will in fact converge to f(x) at those end points [as is asserted
°
in (147)]. In addition, if f(x) is an even f/./n ction, that is, if f( -x) = f(x), (he
series will converge LO f(x) in the double ¡nterval -1 -<: x -<: l. Finally, if f(x)
also is periodic, with period 2/, the series will converge to f(x) everywhere.
Thus the cosine series representation of the odd function f(x) = X in (O, 1)
°
would represent x only for -<: x -<: I and would represent f( -x) = -x for
°
-1 -<: x -<: (see Figure 5.3).
'" "
"
A
" "", "" " ,

A
/
"
./
" ""
""
/
" "" , " , "", / "
- 3i
"- 2/"" -i
"" "2/ 3/ x
Figure 5.3
Example 2. To obtain the Fourier cosine series representing [(x) = eX in the interval
(O, n), we set 1 = n in (147) and (149) and obtain
Ao = -1
7t
1"
o
eX dx = en - 1 ,
7t
An = n2 Jor" eX cos nx dx = n2 n2 +1 1 (en cos n7t - J) (n,,: O).
Thus ¡he desired expansion is of the form

2
e x = 7t (en 2- ¡_en +2 1 ces x
"--"¡-":' + en 5- 1 cos 2x - en 10
+ J ces 3x + .. .)
(O -<: x -<: n). (ISO)
The series (150) represents an even function of period 27t which is ideotified witb eX
when O -<: x -<: 7t and with e-X wheo -7t -<: x -<: O. •
We notice tha! the two expansions (143) and (150) both represent e" in tbe
interval (0, n) but represent entirely distinct functions outside this interval. It
may be remarked tbat half the sum of the two series will represent a. periodic
function which is zero when -n < x < and which coiocides with eX when °
° < x < 7t. At x = O this composite series, consisting of both sine and cosine
5.11. Complete FOUJ"ier Series 219
temu, converges to the value 1, whereas at the points x = ±:n: the series con-
verges to the value :l-e n •
5.11. Complete Fourier Series. It has now been se en that any reasonably
well-behaved function f(x) can be represented in the interval (0,1) by a series
consisting either of sines or of cosines with common period P = 2/, and that
for an odd function f(x) the sine series representation is valid in (-1,/), whereas
for an even function f(x) the cosine series representation is valid in (-1,/). We
next show that, if both sines and cosines of common period 21 are used, a repre-
sentation valid in the interval (-1,1) can be obtained for any function f(x)
which is piecewise differentiable in (-1, 1).
It is clear that any function f(x) can be expressed as the sum of an even
functionfe(x) and an odd functionJ:,(x),
f(x) = fe(x) + fo(x), (151)
since if we write
f(x) = ·Hf(x) + fe-x)] + ±lf(x) - fe-x)], (152)
the first bracket is an even function of x, whereas the second bracket is an odd
function. The preceding theory states that the expansions
nnx
fe(x) = Ao + 00
~ An cos -1- ,
(153)
00 nnx •
fo(x) = ~ Bn srn -1-
are valid in the interval ( -1, 1), if the coefficients are caIculated by the formulas
Ao =
r'
T1 J/e(X) dx,
I
2 n:n:x
An = T
J /e(X)cos-l-dx,
Bn =
2 r'
T J/o(x) srn-I-dx.
. n:n:x
Since aIl integrands are even functions of x, we clearIy may replace f~ by

± f'-I in each case. FinalIy, if we use (151) to replacele (x) by f(x) - lo (x) in the
first two integral s and replace f.(x) by f(x) - h(X) in the third integral, and
notice that the integral of an odd function over a symmetrical interval (-1, l)
vanishes, there folIows
Ao = 21
1JI_/(x) dx - 211 JI_/o(x) dx 211 JI_/(x) dx, =
I I n:n:x 1 JI n:n:x
An = T
J f(x) cos -I- dx - T _/o(x) cos -I- dx
._1
I
1 n:n:x
= T
J f(x) cos -I- dx ,
_1
220 Boundary-Value Proolems and CharaClerislic-FunClioD Representations
Bn = -l
1
J'
_lf(X)
I
. nnx
Sin-l-dx - T
I JI_/,(x) SIn
. nnx
- l- dx
I . nnx
= T
J-1
f(x) SlD-l-dx.
Hence, by introducíng (153) into (151), there follows

~
f(x) = Aa + n-=-t ( An cos -l-
nnx + Bn Sin
. -l-
nnx) (-l < X < l), (154)
where Aa =
I
2l JI f(x) dx,
_1 An = 1I J'
_1
nnx dx,
f(x) cos -l-
Bn = +JI -1
f(x) sin n7 dx.
x
(155)
The series (154) is the complete Fourier series representatíon of f(x) in the
interval (-1, 1). It is seen that again Aa is ¡he average value of f(x), the average
now being calculated over ¡he inrerval of representation (-1,1), whereas An and
Bn are twice the average values of f(x) cos (nnx/ l) and f(x) sin (nnx/l), respec-
tively, over (he same interval. For an even function there follows Bn = O and the
cosine series previously obtained results; for an odd function Aa = An = O
and the previously obtained sine series results. If f(x) is neither even nor
odd, the series (154) will contain both sines and cosines of period 21 and will
represent f(x) , in the sense described, in the interval (-l, l) of length 2l. Because
of the period.icity of all terms in the series, trus representation is repeated period-
ically for aH values of x.
Example 1. To obtain a Fourier series of period 2n, representing f(x) = e'< In tbe
interval (-n, n), we set ¡ = n in (\ 54) and (155) and obtain
1
Aa = 2n J"-. 1 (e' - e-')
eX dx = 2n = I sinh n,
Ji"
1
An=1i J" 2 cosnn
_.ez cosnxdX=Ji"n .
2+ 1smhn (n *" O),
B" = -nlJ" ex . d = - -n2ncosnn'
Sin nx
n + x 2 1 SID h n.
-"
The desired expansion is thus of the forro
eX = sin: n (1 - cos :, + 5"2 cos 2x 1 cos 3 x
- 5" + ...
+ . X -
SlIl 4Sin
"5
3 ' 3 + .. .
' 2X+ "5 Sill X ) (-n < x<n)
or
sin b n [ ~ (_I)n+ I ]
eX = n 1 - 2 '~I n 2 + 1 (cos nx - n sin nx) (-n < x < n). (156)
This series converges to eX at all points inside the interval (-n, n). Since
eX = cosh x + sinh x,
5.11. Complete Fourier Series 221
it follows that the even terms represent the function cosh x in (-n, n), whereas
the odd terms represent sinh x:
sin h n [ ~ (- 1)n. 1
¡ + 1 CQS nx
]
cosh x = 1 - 2 2: (-n < x <n),
n ~- l n
. sinh n ~ (_l)n. 1n .
smh x = 2 - - 2:
n
¡ . 1 Sin nx
,, =1 n + (-n < x < n).
•
If we consider tbe periodic extension of f(x), of period 2/, tbat is, the func-
tionf*(x) which is of period 21 and wbicb agrees witb f(x) wben -1 < x < 1,
we see that at botb the points x = ±I the rigbt- and left-hand limits of f"(x)
are f( -/) and f(l), respectively. Since tbe. series (I54) converges everywhere to
f*(x) when f(x) is piecewise differentiable in (-1,/), with the usual convention
regarding points of discontinuity, we deduce that the series (154) converges at
both end points x = ±I to the value Hf(1) + f( -1)), that is, to the ·mean of the
two vall/es of f(x) at the end points. We recall, in summary, tbat also the sine
series (138) vanishes identically at the end points of the interval (O, 1), and hence
converges to f(x) at those points only if f(x) vanishes there, and that the cosine
series (147) converges to f(x) at both end points of the interval (O, 1).
We next show that the constants in (I54) can be determined in sucb a way
that the given function f(x) is represented in any specified interval of lengtb
2/, that is, of lengtb equal to tbe common period of tbe trigonometric functions
involved. Suppose that a function f(x) is specified in an interval (a, a 21) +
and tbat it is required tbat tbe coefficients in (154) be determined so that (154)
represents f(x) in tbe usual way in that interval. We next again consider the
periodic extension f*(x) defined for all values of x in such a way tbat f*(x)
coincides with f(x) in the given interval, and sucb that f*(x) is a periodic func-
tion, of period 2/. Then if f(x) is replaced by f*(x) in tbe definitions of tbe
constants (155), tbe corresponding series (I54) c1early will represent the periodic
function f*(x) for all values of x (in the usual sense), and hence, in particular,
will represent f(x) in tbe interval (a, a· + 21). But since the integrands in (155)
are tben periodic functions, of period 2/, tbe range of integration may equally
well be taken as any interval of lengtb 2/. In particular, we may cboose as that
range the interval over wbicb representation is desired, where f"(x) = f(x).
It follows that if a representation of the form
~ ( 2nnx . 2nnx)
f( x) = Ao + n7! An cos ---¡¡- + Bn S l O - p (157)
is required over the interval (a, a + P), the coefficients are to be determined by
the equations
fO+P
Jor + n;x
o p
1 2 2
Ao = P o f(x) dx, An = p f(x) cos dx,
(158)
Bn = p2JHP
o f(x)
. 2nnx
SlO--¡¡-dx.
In particular, representation in the interval (O, P) is obtained by setting a = O.

These general results are also obtainable directly by considering the bound-
ary-value problem
2
d y'
- +
dX2 1v
~
= O, y(a) = y(a + P), y'(a) = y'(a -,- P) (159)
for which the characteristic values 1n = 4n 2 n 2 jP2 correspond to the two-para-

meter characteristic functions
( 160)
The orthogonality of these functions for arbitrarily assigned values of both en

and d n , on the interval (a, a + P) , corresponds to the relations
"'a+P
. 2mnx . 2nnx d - O
P sm ----¡¡- x-:-
J° sm (m =;t= n),
o+ P ?_mnx ')
_nnx O
J o
P
cos
P
cos-- d x
P
= (m =;t= n), (161)
O+ . 2mnx 2nnx d - O
J• Sin P COS ----¡¡- x - ,
where the validity of the third relation when rn = n is not guaranteed by the
results of Section 5.6 (since the two functions involved then correspond to the
same characteristic number), but is directly verified.
Example 2. We suppose that it is required to find a Fourier series of period 6 which

in the interval (1 , 7) represents a function fCx) taking on the constant value + 1 when
j < x < 4 and the constan! value -1 when 4 < x < 7. With a = 1, P = 6, the series
(157) becomes
~ ( Ancos-
rmx . rmx)
f(x) = Aa + n~'
3- + B n sm- 3- ,
where, in accordance with (J 58),
Aa= ~ff(X)dx= ~[f'dX+ fC-l)dX]=O,

7 []4 17 rmx ]
An = 3" ] f(x) cos
1 rmx rmx 1
1 dx = 3" cos
-3- dx - • cos--dx I -3- 3
=
1
rm -
[? . SID 34rm - . rm
Sin 3 - . 7 rm]
SIn 3 Cn =;t= O),
Bn=~ ff(x)sin~xdx = j [fsinn~xdX- fSin~Xdx]

1 [
= - rm 4nn
2 cos -3- - cos 3
nn - cos 7nn]
3 .
5.12. Term-by-Term Differentiation of Fourier Series 223
By virtue of the relations

. 4n71 . n71 . 7n71 . n71
-3- = cos n71 S in 3' sln-- 3'
3 =
Sin Sin
4n71
cos -3- = cos n71 cos 3n71 ' cos -3-
7n71
= cos)'
n71
these results take the form

,1-cosn71. n71
Ao = 0, A ¡t = -- nlt sin),
B = 2 I - cos n71 cos n71 .

" n71 3
Since ¡he factor 1 - cos n71 is zero when Il is even and 2 when n is odd, it follows
Iba! ¡he required series cOOlains only harmonics corresponding to the odd integers.
The series can be written in the formt
J x = -4
r () ""
..... - 1 ( cos nn. nnx
- Sin - - - . -n71
SIO cos -.n71X)
- .
71 " odd n 3 3 3 3
We notice that this series is equivalent 10 ¡he form
r ( X ) -_ - 4
J
""
.....
1 '. nn(x . . . -
-Sln
1)
1t n odd n -'
which could have been obtained more directly, in this case, by first ¡ransforming the
variable by moving the origin to the point x = 1. •
Any function f(x) which is of period P and which also satisfies the relation
f (x + n = -f(x) (162)
is called an odd-harmonic fun ction, since it is readily shown that the Fourier
series representation of suc h a function pos ses ses only harmonics corresponding
to odd integers. The function in the preceding example is of this type.
5.12. Term-by-Term Differentiation oC Fourier Series. Because of the fact

that the formal (term-by-term) derivative of a Fourier series is another Fourier
series, and also that the formal derivative of a sine series is a cosine series, and
conversely, it is possible to obtain conditions permitting term-by-term differen-
tiation of those series by use of a relatively simple argument. Since the need for
such operations arises fairly often, we now deal briefly with their justification.
Suppose that f (x) possesses a Fourier series of period P which converges
to f(x) inside the interval (a, a + P). Then that series is specified by Equation
(157), where the coefficients are determined by (158). The result of formally
differentiating that series term by term is ¡hen
~ (2n71 B cos-- -A . I2n7tx)
.-1
L... - -
p.
2n71x - -
2n71
P P"
S O--'
P
(163)
tThe ph.rase " n odd" is used in summations lo indicate that n is to ¡ake on ¡he values n = 1,
3~ S, . ...
224 Boundary-Value Problems and Characteristic-Function Representatiol1s
On the other hand, if f'ex) exists and is piecewise differentiable, its Fourier
senes of period P in (a, a + P) is
f '( x ) = ao + n~¡
~ (a n CQS ~
2mrx +b n
. ----¡¡-
SIn
2nnx)' , (164)
where
J.r + f J.r
a P a
1 , 2 +P , 2nnx
ao = y (x) dx, an = y f (x) cos ~dx,
(165)
b. =
21
Y
Q
a
+
P
f'(x) sin 2n;x dx.

If f(x) is continuous and f'(x) is (at least) piecewise continuous, we may
transform the formulas (165) to the forms
1
ao = y[f(a + P) - fea)],
Qp
a. = y
2 {[ 2nnxr+p
f(x) cos ----¡>
2nn r
+ -p JQ + f ( x) sm
• 2nnx }
- p dx
2 2nna 2nn
= p[f(a + P) - fea)] cos ~ + -pBn,
b - 2 {[fe ) . 2nnx]Q+P
•- p
2nn + f()
x sm ~ Q - -p J.
2nnx d }
x cos ~ x
ra P
= ;, [fea + P) - fea)] sin 2n;a _ 2,; An,
by use of integration by parts. Consequently, the formal senes (163) then is

identified with (164) if and only if the additional condition
fea + P) = fea) (166)
is satisfied.
Furthermore, if f'(x) is piecewise differentiable in (a, a + P), its expansion
(164) will in fact converge to f'(x) at each interior point of that interval at which
f'(x) is continuous. When these facts are summarized, we are led to conclude
that if fex) is continuous and f'(x) is piecewise differentiable in (a, a + P), and
if fea + P) = fea), then the result of term-by-term differentiation of the Fourier
series of period P representing f(x) inside (a, a + P) converges to f'(x) at each
interior point at which f'(x) is continuous.
This result is in accordance with the more general result quoted in Section
5.7, except that here f(x) need satisfy only the first of the two conditions
involved in the generating Sturm-Liouville problem (159). It may be seen that,
if the condition f'(a + P) = f'ea) is also satisfied, then the differentiated series
also will converge to f'ex) at the two end points x = a and x = a + P.
Analogous considerations of the sine series (138) and the cosine series (147)
lead easily to the conclusion that, if f(x) is continuous and f'(x) ís piecewise
5.12. Term-by-Term Dift'erentiation of Fourier Series 225
differentiable in (O, (), then the result of differentiating ¡he sine series (138) term
by term converges to f'(x) at each interior point of(O, 1) at which f'(x) is continu-
ous, as well as at the end points x = O and x = 1, provided that f(x) also satisfies
the end conditions feO) = O and f(l) = O. For the cosine series (147), the same
statement applies, with two exceptions: convergence of the differentiated series
to f'(x) at the end points generally does not follow and also no supplementary end
conditions need be satisfied by f(x). In the case of the cosine series, it is seen
that each term of the differentiated series vanishes at both end points and,
accordingly, that satisfaction of the end conditions 1'(0) = O and/or 1'(1) = O
would in fact bring about convergence of the differentiated series to f'(x) at
one or both of the end points as well.
Example 1_ From the readily established representation

l·· 41 1 mr.x
x = -2 - -2 ~ -2 cos -1- (O < x < 1), (167)
11: noddn
we deduce the representation

1 = -±-
n 11
~
odd
1..- sin mr.x
n 1
(O < x < 1), (168)
by differentiation. The validity of this result is easily confirmed independently by ob-

taining (168) directly, by use of (138). •
Example 2. The function

f(x) = {X1- x (O < x <1/2),
(1/2 < x <1)
has the sine-series expansion
f(x) = 4/ ~ sin (n7t/2) sin n7tx
n 2 11 =z1 n2 1
I
(O < x <1)
I [see Problem 49(b)]. Since f(x) satisfies the stated conditions [in spite of the discon-
'1 tinuity in f'(x)], we may differentiate term by term, when O < x < 1/2 and when
1/2 < x < 1, to produce the verifiable representation
4 ~ sin (mr. /2) mr.x
g ()
x = - .:." cos - -
11: n~ I n 1
= ~ ( + cos ~x _ -} cos 3~x + ... ),

(O < x < 1/2),
where g(x) = f'(x) = { 1
-1 (1/2 < x <1),
and g(l/2) = O.
Since, for Fourier series, the associated operator L is such that Ly = y",
•
the result obtained at the end of Section 5.7 applies to this operation, with
respect to the interior of the relevant interval. (See also Problems 67 and 68.)
226 Boundary·Value Problems and Characteristic·Function Representations
5.13. F ourier-Bessel Series. Expansions in terms of Bessel funetions arise

most frequently in eonneetion with a boundary-value problem involving the
X2 dJ
d' dy
+ x dx + (¡t2X2 - p')y = O (169)
or, equivalently,
.!!..(x
ax
dy
dx
) + (_P'x + ¡t2X)Y = O. (170)
Referenee to Equation (130a), Chapter 4, shows that the general solution of

this equation is of the form y = Z /¡tx) or
y = {C¡J/¡tX) + c,J_p(¡tx) (p not an integer),
(171 )
c¡J/¡tx) + c,Y/¡tx) (p an integer).
Equation (170) is identified with (90) if we write
2
p(x) = x, q(x) = _l!..., r(x) = x, (172)
x
Restrieting attention to the interval (O, 1), we see that sinee p(O) = O, the results
of Seetion 5.6 show that the eharaeteristie funetions of the problem are orthogo-
nal on (0,/), with respeet to the weighting funetion r(x) = x, if at x = O it is
required merely that y be finite and x dy/dx be zero and if at x = 1 a homo-
geneous eondition is preseribed.
The eonditions at x = O require that we take
c2 = O ( 173)
in (171). If at x = 1 the eondition
y(l) = O (1 74a)
is preseribed, the eharaeteristie values 01' ¡t are deterrnined as the roots of the
equation Jp(¡t) = O and henee, if we write ¡t) = a., so that a n is dimensionless,
there follows
(1 74b)
so that a. is the nth zero of Jp(x). Similarly, the end eondition

y'(l) = O (175a)
requires that the eharaeteristie numbers ¡tn satisfy
(175b)
whereas the more general eondition

ky(l) + Iy'(l) = O (k > O), ( 176a)
5.13. Fourier-Bessel Series 227
where k is a prescribed dimensionless constant, determines fin in the form
11 n
__ = IX1n (176b)
In all these cases the characteristic functions are of the form
(177)
where fin and IX n are defined by (1 74b), (17 Sb), or (l76b), and the results of
Section 5.6 show that these functions are orthogonal on (O, 1) with the weighting
function r(x) = x,
(m *' n). (178)
Although Equation (170) is one of the exceptional cases described in Section

5.7 in that p(x) and r(x) vanish at the end point x = O, and also q(x) is not
regular at .'C = O unless the order p of the Bessel functions is zero, stilJ it can be
shown that the formal series (107), with coefficients determined by (109), will
converge to f(x) inside the interval (0,1) under the same conditions as those
stated with reference to the Fourier series.
We must, of course, be careful to inelude all the characteristic functions of
the problem involved. From the definition of the Bessel functions Jp(x), it is
elear that the relation
Jp(-X) = (-I)PJp(x)
is satisfied. Thus solutions of (174b), (17 Sb), and (1 76b) exist in pairs, symmetri-
cally located with respect to the point x = O. However, since replacing fin by
-!1,. in (177) either does not change rpn(x) or rnerely rnultiplies it by a numerical
factor, the negative values of fin need not be considered. If fio = O were a char-
acteristic number, the corresponding function 'Po (x), as given by (177), would
be identically zero (and hence not a characteristic function) except in the case
when the order p of the relevant Bessel functions is zero. When p = O, only
(l7Sb) permits the solution fio = O. Thus it is necessary to consider only the set
of functions ['Pn(x)} corresponding to positive characteristic numbers fin (n = 1,
2, 3, ... ) in all cases except in the case of (l7Sb) when p = O, in which case the
function 'Po = 1 corresponding to fio = O must be added to the set.
Temporarily exeluding this exceptional case, the coefficients in the senes
(O < x < 1), (179)
where the numbers fin and IX n are the positive quantities defined by one of Equa-
tions (174b), (17Sb), or (176b), are given, by virtue of(109), by the equation
An S: x(J.(finx )]2 dx S: xf(x)Jp(fin

= X) dx.
228 Boundary-Value Problems "nd Characterístic-Function Representations
Tbe coefficient of A. is independent of f(x) and can be determined once and foc
aH. Denoting it by C., we then have
A. = C1
•
l'
o
xf(x)Jh.t"x) dx, ( 180)
where en = {x[JhlnX)j2 dx. (I81)
To determine the factors Cn> we proceed by an indirect methoo as follows .

With the notation of (177) we have
d (drp.)
dx x dx + (2
f.1.. x - p2) rp. -- O,
X (182)
since the function rp.(x) satisfies (170) when f.1. = f.1.., If both sides of (182) are
multiplied by 2x drp./dx, the resulting equation can be written in the form
(f.1. 2X2 - p2)_(rp2)

d = d
-_ ( X d rp. )2 .
• dx' dx dx
When both sides of this equation are integrated over (O, 1), and the resulting
left-hand side is simplified by an integration by parts, there follows
[f.1.;x
2
- p2)rp;1~ - 2f.1.; f>rp; dx = -[x2(~~'rl
If we notice that rpn(O) = J p(O) = O when p > O, that Jo(O) = 1, and that
xJ~(x) = O at x = O, it is readily seen that the integrated terms vanish at the
lower limit. Thus we obtain the important relation
Cn = S: x[J/f.1.n x )j2 dx
= 2~;{(f.1.~¡Z - p2)[rpn(l)j2 + ,{d~;I)T} ( 183)
where rp.(x) = J /f.1.nx). ( 184)

We also have
rp.(/) = J if.1..1)
and, making use of Equation (111), Section 4 .9,
~_d ( )X_ _ f.1. . J p+' (

f.1.n)
X + -pxJ if.1.. x ) ,
d x - d-
x J p f.1.. -
so that there foJlows
dp;;1) = - f.1.Jp+,(f.1.. / ) + frp·(I)·
Thus, if f.1.. satísfies (I74b), there follows rp.(I) = O and

¡z ¡z
C n = 2[Jp+ ,(f.1..!)]2 = 2[Jp+,(1X.)]2. ( 185a)
5.13. Fouricr-Bcsscl Series 229
Ir p. satisfies (175b), rhere Jol/ows drp.(I) / dx = o and

p2/2 _ p' 12 et2 _ p2
Cn = n7 2 [Ji,u.!W -""7 n , [Jp (Ct n»)2. (l8Sb)
-J.1.n - an
Finally, if ,un sarisfies (176b), rhere Jol/ows Idrpn(l)/dx = -krpn(l) and
C = ,u~/2 - p2
n 2 J..l;
+ 2
k [J (,u 1))2
P ti
= f22 Ct; - p2
IX;
+ k'[J (et))2
JI ti •
(18Se)
In the exceptional case noted aboye, when an expansion in terms of the

functions Jo(P.x) is required , where ,un = et n/ ' is a solution of the equation
(186)
the presence of the additional characteristic number ,uo = O introduces a con-
stant term Ao, leading to the series
J(x) = Ao + n~~ AJo(,un x ) = Ao

~
+ "~, AJo
(X)
n-
Ct ' (187)
I
where Ao = I~ f xJ(x) dx ( 188)
and the rematntng coefficients are glven by Equations (180) and (18Sb) with
p = O.
Although the explicit evaluation of the integral appearing in (180) often is
not feasible, even when J(x) is of a simple form, in practice the integral can be
evaluated approximately for as many values of n as are required, by numerical
or graphical methods.
Expansions of the types considered are of frequent use in connection with
the solution of certain boundary- va lue problems involving partial difrerential
equations, the particular type of expansion to be used being determined by the
nature of the problem.
Example. As an application, we determine the expansion of the function J(x) = 1

in the interval (O, 1), in a series of the form
1 = A,Jo(,u,x) + A 2 J O(,u,x) + ...
~
= ~ AnJo(,un x ) (O < x < 1), ( 189)

n= I
where J o(,unl) = O, so that

Ct n
,un = T with JO(et n ) = O. (190)
In this case Equations (180) and (1 85a) give
(191)
where ( 192)
230 Boundary-Yalue Problems and Characteristic-Function Representations
AIso, making use of Equation (l08), Section 4.9, we have
and hence
(193)
With the results of (192) and (193), Equation (191) becomes
...4 =
2 , (194)
n (jinl).!,(jin l )
and the desired series is of the form
1 = -.2.. .!o(ji,x)..L 2 .!o(jiox) + (195)
jiII.!, (ji,/) , ji2 IJ I(ji ol)
where .! o (ji.!) = 0, or, equivalently,
(196)
where IX. is the nth zero of .!o(x).

Since the Bessel functions involved are even functions of x, the series must repre-
sent 1 not only in (0,1) but also in the syrnmetrical interval (-1,1). At the end points
x = ±I the series clearly does not converge to the given function, since all terms in the
series vanish. •
Three-place values of the first five zeros of sorne Bessel functions are
inc!uded in Table 1. Much more elaborate tabulations appear in the literature.
Table 1. Zeros or Bessel Functions: J p(!!..) = O
n p=O p=1 p=2 p=3 p=4 p=5

1 2.405 3.832 5.135 6.379 7.586 8.780
2 5.520 7.016 8.417 9.760 11.064 12.339
3 8.654 10.173 11.620 13.017 14.373 15.700
4 11.792 13.323 14.796 16.224 17.616 18.982
5 14.931 16.470 17.960 19.410 20.827 22.220
5.14. Legendre Series. Legendre's differential equation,
d Y
X"
2
( 1 - x 2) d-----,- 2 x dy
-
dX
+ pp
( + 1) y- - , ° (197)
or, equivalently,
d:[(l - X2)~~J + p(p + 1)y = 0, (198)
can be identified with (90) if we write

p(x) = 1 - x 2, q(x) = 0, r(x) = 1, A = p(p + 1). (199)
5.14. Legendre Series 231
Since the function p(x) vanishes at the points X = ± 1, the results of Section 5.6
state that any two distinct solutions of (197) which are finile and have a finite
derivative at x = ± 1 are orthogonal, with respect to the weighting function
r(x) = 1, on the interval (-], 1). Since (197) possesses solutions wbich are
finite at x = ± 1 only if p is a positive integer or zero (see Section 4.12), this
condition of finiteness determines permissible values of p in the form
p=n (n = O, 1, 2, 3, ... ). (200)
Tbe corresponding solutions which are finite when x = ± 1 are proportional
to the Legendre polynomials
'P. = p.(x), (201)
and the aboye reasoning leads to the conclusion
(m =1= n). (202)
Just as in the preceding cases, a function f(x) which is piecewise differen-

tiable in the in terval (-1, 1) can be represented by a series of the functions 'P.,
=
= 2:o A.P.(x)
n=
(-1 < x < 1). (203)
The coefficients A. are given by the equation
A. [ [p.(x)P dx = [ f(x)P.(x) dx. (204)
In order to evaluate the integrals appearing in (204), it is frequently useful

to express P .(x) by Rodrigues' formula [Equation (168), Section 4.12],
(205)
Hence we may write
J'-1
f(x)P.(x) dx = 1I
-2'
n.
J' -1
f(x)dd'.(x 2
x
- 1)' dx. (206)
If the right-hand member of (206) is integrated by parts N times, where N <: n,

assuming thal the first N derivatives of f(x) are continuous in (-1, I),t and if it
is noticed that the first n - 1 derivatives of (X2 - 1)' vanish when x = ± 1,
J' (-l)NJ' [d -I[d.-

N N
2 (N <: n).
_, f(x)P'(x) dx = 2'n! -1 dxNf(x) _ dx' N(X - 1)' ] dx
(207)
tOtberwise, integration by parts may not be valido

In particular, when N = n there follows
fl f(x)P'(x) dx = (;~r fl (x
2
-
1)" d" f(x) dx
dx"
(208)
ir f(x) and its jirst n deriva ti ves are continuolls in ¡he interval (-1, 1).
Replacing f(x) by P,,(x) in (208) and noticing tbat, from (205),
2
d" P,,(x) = _1_ d " (X2 _ 1)"
dx" 2"n! dx 2 "
_ 1 d 2 " ( 2" _ 2"-2 + ... ) = (2n)!, (209)
- 2"n! dx 2" x nx 2"n!
we obtain, from (208),
JI -1
[F,,(x)J2 dx = )2:(
--
(?)'
n.
;)2 JI
-1
(l - x 2 )" dx. (210)
The integral on the right can be evaluated by successive reductions (involving

integration by parts) in the form
= 2 2"+I(n!)2.
JI
_1 (
1 _ X2 "dx = 2 (2n)(2n - 2) .. ·4-2
) (2n + 1)(2n - 1)· . ·5- 3
Thus Equation (210) takes the simple form
(2n + 1)!
(211)
OI
J_1 [F,,(x)J2 dx = 2n
2
+ 1
(212)
and Equations (204) and (208) give for the coefficients in the expansion
f(x) = AoFo(x) + AIFI(x) + ...
~
= I:
n=O
A"F,,(x) (-I<x<1) (213a)
the alternative forms
A .. 2n+lJI
2 _1 f(x)F,,(x) dx,
" j + JI
2n
2 n+ 1 n.,
1
-1
(1 _
X
2)"d"f(x)d
d"
x x,
(2l3b)
where, again, the second form in (213b) is to be used only if f(x) and its first
n derivatives are continuous in (-1, 1).
Since F,,(x) is an evell function of x when n is even, and an odd function
when n is odd, it foHows that if f(x) is an even function of x, the coefficients A"
will vanish when n is odd; whereas if f(x) is an oddfimction ofx, the coefficients
A" will vanish wben n is even.
Thus, for an even function f(x), there follows
O (n odd),
A
"
=
{ (2n + 1) f f(x)P,,(x) dx (n even),
(214)
5.14. Legendre Series 233
whereas for an odd function f(x) we have
A = f(2n + 1) fa' f(x)Pn(x) dx (n odd),

(215)
n lo (n even).
If we are interested in representing a given function f(x) merely over the
interval (O, 1), it is then clear that a series containing only even polynomials can
be obtained by using (214), or, if preferred, a series containing only odd poly-
nomials can be obtained by using (215). Either of these series will then represent
f(x) inside the interval (O, 1), and the first series will represent f( -x) and the
second -fe-x) in the interval (-1, O).
Finally, we remark that expansions valid in the more general interval (-1, l)
are readily obtained by replacing x by xiI in the preceding developments. This
procedure leads to the expansion
f(x) = AoPo(~) + AIPI(n + ...

=nt.,AnPn(~) (-I<x</), (216)
where
A = j2n 2~ 1L f(x)Pn ( ~ ) dx,
(217)
n 2n
2n+1n!/n+(
+ 1 JI (12 _
-1 X
2)n dnf(x) d
dxn x,
which generalizes (213).

It is clear that if f(x) is a polynomial of degree k, al! derivatives of f(x) of
order n vanish identically when n > k. Hence it follows from (213) or (217)
that any polynomial of degree k can be expressed as a linear combination of the
jirst k + 1 Legendre polynomials.
Example 1. As an application, we determine Ao, Al, and A 2 so that

X2 = AoPo(x) + AIPI(x) + A 2P 2(X),
the higher coefficients vanishing by virtue of the aboye remark. Since X2 is an even
function of x, the remaining odd coefficient Al must also vanish:
Al =0.
To ca1culate Ao and A 2, we use the second form of(21~b) and obtain
Ao = T1 JI
- 1
l·x 2 dx =
1
3' 5
A 2 =S.2 J'
-1
(l-x 2)2·2dx =3'
2
and hence
X2 = HPo(x) + 2P 2(x)]. (21S)
The truth of this identity is verified by reference to the expressions for Po (x) and P 2(x)
in Section 4.12 [Equation (158)]. •
234 Boundar)'-Value Problems and Characteristic-Function Representations
Example 2_ We determine the Legendre expansion in (-1, 1) of the unit step function
f(x) which vanishes when - 1 < x < O and which takes on the value unity when
O < x < 1,
(-1 < x < O),
f(x)
.
= {O1 (O < x < 1).
(2 19)
Since f( x) is discon linuou s al x = O, Ihe second express ion in (213b) cannol be used .
The basic relation becomes, in this case,
A" =
2n +
2
1
Jor' p"(x) dx.
Hence, referring to Equation (J 58), Section 4.12, \Ve find that
Aa = 1 So' 1 dx = 1,
A, =!So'xdx=,t,
A2 = ~'! So' (3x 2 - I)dx = O,
A, = ! . ! So' (5x 3 - 3x) dx = --?o,

and so forth. Since f(x) can be considered as the sum of the constant 1; and an odd
,. function of x, it is seen that Aa is the only nonvanishing even coefficient. The first terms
of the desired series are of the form
'. f(x) = 1;P o(x) + iP,(x) - T7oP,(x) + -HP,(x) + ... (-1 < x < 1). (220)
The determination of the general coefficient requires knowledge of the formula
i (n = O),
So' p.(x) dx =
jO
(
(n =
_1)(.-') / 2
n!
2 , 4, 6, ... ),
n(n
1
+ 1)
(1-3-5·· ·n) 2
(n = 1,3,5, ... )
(see Problem 57 of Chapter 4). By use of this result, the expaosion (220) can be writteo
(221 )
in the form
f(x) = 1. + ~ (_ 1)(.-1)/2 2n + 1(1·3·5 · · · n)2p.(x) (-1 < x < 1). (222)
2 .odd 2n + 2 n· n!
5.15. Tbe Fourier Integral. We have seen in Section 5.10 that the functions
•
CPn(X) = sin n~ x (n = 1,2,3, ... )
form an orthogonal set in the intervalO < x < 1 and that a piecewise differen-
tiable function f(x) can be represented as a series of these functions in that
interval. The question arises as to the possibility of a similar representation in
the semi-infinite intervalO < x < oo . However, since the functions q>.ex) vanish
when 1 ~ 00, it is clear that such a representation cannot be obtained by merely
replacing 1// by zero in the series (138).
5.15. The Fourier Integral 235
If we consider the relation

Sin (U2 u,)1 _ sin (u 2 + u,)1
J' . '
-
?
::..
I D
Sin u,x Sin UzX dx =
l
I
(u z -
-
u,)1
si n? 2u,l
_ti!
I (
(u z + u,)1
_ u, ) ,
Uz -
when u, and U z are posltlve, we notice that the right-hand member vanishes if
(223)
ti, and tlz are distinct integral multiples ofn/I and is equal to tlnity if u, and U z
both take on the same integral multiple of n/I. This is, of course, in accordance
with the results of Section 5.10. In addition, however, we notice that as 1 - 00
the right-hand member approaches zero for any positive values of u, and u 2 ,
so long as U z 7'= u" but that if l/z = u" the right-hand member approaches unity.
That is, if we write
ifJ.(x) = sin ux, (224)
we obtain
. 2 J' ifJu,(x)ifJu,(x) dx = {O1

ltm -1
1-- D
(u z 7'= u,),
(u z = u,),
(225)
when u 1 and l/2 are positive. Thus two functions of form (224), corresponding
to any two positive values of ti, have a sort of orthogonality property in the
semi-infinite interval (O, <Xl), and we are led to suspect that a representation of
a function f(x) in that interval generally must involve all possible functions of
type (224), where ti is any positive number and is not restricted, as in the finite
case, to a set of discrete values.
In the finite case f(x) is expressed as a linear combination of functions of
the latter type, that is, as an infinite series,
~ . nnx
f(x) = n~ An SlD -1- (O < x < 1),
where the constants An specify the contributions of the successive harmonics

in the series. In the limiting case 1 - 00, if we write the contribution of sin l/X
to the representation of f(x) in the form A(u) sin l/X , we may expect that the
complete representation of f(x) will be obtained by superimposing these con-
tributions by means of an infinite integral,
f(x) = fD~ A(u) sin ux du (O < X < 00). (226)
Assuming that such a representation exists, for a sufficiently well-behaved fune-

lion f(x), it remains to determine A(u) for ail positive values of u, that is, as
a function of u. In the foIlowing developments we indicate a formal method of
determining A(ll), and afterward state eonditions under which the result obtained
is correet.
In determining the funetion A(u), we are guided by the analogous procedure
of Section 5. 10. Thus we first multiply botb sides of (226) by sin uox, where UD
236 Boundary-VaJue Problems and Characteristic-Function Representations
is any positive value of ti,
J(x) sin uox = sin u ox L~ A(u) sin ux duo
We next integrate both sides of the result with respect to x over the interval
(0,/), where I is destined to beco me infinite, and so obtain
f: J(x) sin uox dx = f~ sin uox [f A(u) sin ux du Jdx

or, assuming that the order of integration may be interchanged,
f: J(x) sin uox dx = fa~ A(u) (f~ sin ux sin UaX dx) duo (227)
We denote the right-hand member of (227) by R,; then, making use of (223),
we obtain
R, = 21 r~ ~u) 1
sin I(u - u o) du - -2 r~ A1.u ) sin I(u + u o) duo (228)
Jo u Uo Jo u U o
With the substitution I(u - u a ) = t in the first integral of (228) and the
substitution I(u + u a ) = t in the second integral, this expression takes the form
R, = J.. f~ A(U a + _t ) sin t dt - J.. f~ A(-u o + _t_) sin t dt. (229)

2 -Iu, 1 t 2 ,,,, 1 t
Now, if tl o > O, we see that as 1 - + 00 the second integral Jormally approaches
zero,t whereas the first integral takes the value
J..
2
f~ _~
A(u o) sin t dt = A(u o) f~ sin t dt = E...A(uo).t
t 2 __ t 2
Tbus, as 1-+ 00, Equation (227) gives formally
Jor~ J(x) sin UaX dx = lim

1_00
R, = 7t
2
A(ua) (u a > O). (230)
Replacing U o by ti in (230), we bave determined tbe function A(u) required In

(226) in tbe form
A(u) = - 2 i~ J(x) sin ux dx. (231)

7t o
To distinguisb tbe dummy variable x in (231) from the current variable x
in (226), we re place x by t in (231) before introducing (231) into (226), and so
obtain
J(x) =.2-
7t
r- sin ux [J(t)
Jo o
sin ut dt du (O < x < 00). (232)
tSince A(II) is not yet koowo, the vanishing of trus term as 1-+ 00 cannot be guaranteed at this
stage. However, the plausibility of this occurrenee follows from the faet that (sin 1) / 1 exeeutes
damped oseiUatioos over che cange of integracion.
tSee Problem 14(b) of Chapter 2.
5.15. Tbe Fourier Integral 237
This expres sion is known as the Fourier sine integral representatíon of f(x) and
can be shown , by more rigorous analysis, to be valíd when x > O if f(x) is piece-
wise differentiable in every finite positive interval and if the integral f ~ I f(x) I dx
exists. As in the Fourier series tbeory, the integral represents f(x) at points of
continuity and, in order that the same be true at a point x = e at which f(x) has
a finite jump, we again as sume the definition f(e) - -Hf(e+) + f(e- )].
In a similar way, the Fourier eosine integral representation
f(x) = -21~ cos uxl~ f(t) cos ut dt du (O < x < 00 ) (233)

1t o o
can be obtained , subject to the same restrictions.
It is clear that (232) represents - f( -x) wben x < O, whereas (233) repre-
sents f( -x) wheo x < O. Thus, if f(x) is an odd function, the representation
(232) ioeludes aH values of x, whereas if f(x) is an even function , tbe same is
true of (233). As iD the finite case, we can make use of tbese fact s to obtain a
representation valid for aU values of x by using both sine and cosine harmonics .
Thus, if we write f(x) = ¡;(x) + fo(x), where ¡; is even and fo is odd, we verify
that
-1
7r
l~ cos ux I~
o __ 00
f(t) cos ut dI du = - 1
1t
l~ cos ux
o
[_00
le(t) cos ut dt du
= -21~ cos ux l~ I,(t) cos ut dt du

1t o o
=fe(x) (- 00 < x < 00), (234a)
and, similarly,
- 1 l~ sin ux J~ f(t) sin ut dt du = f.(x) (- 00 < x < 00), (234b)

7t o _ 00
under the assumption that both f ~ If.(x) I dx and f ~ lfo(x) I dx exist or, equiva-
lently, that f==I f(x) I dx exists. Hence, adding (234a) and (234b), we obtain
the representation
f(x) = r [A(u) cos ux + B(u) sin u x ] du (- 00 < x < 00), (235)
where A(u) and B(u) are defined by the expressions
A(u) = n1 J~_ ~ f(t) cos ut dt, B(u) = n1 f~_~ f(t) sin ut dt. (236)
By introducing (236) ioto (235), we obta in also
f(x) = ! l~ [L~ f(t)COS U(t - x) dtJdU (- 00 < x < 00), (237a)
or, equivalently,
1 f~
f(x) = 21t _ ~ f~
_ ~ f(t) cos u(t - x ) dt du (- 00 < x < 00). (237b)
238 Boundary-Value Problems aod Cbaracteristic-Fuoclion Representations
Tbis express ion ís known as the complete Fourier integral representation of f(x)
and represents f(x) for al! values of x in the usual sense if f(x) is piecewise
dífferentiable in every finite interval, and if ¡he integral f:::~ I f (x) I dx exists .
By writing
we can show that (237a) also can be written in tbe alternative comp/ex form
(-00 < x < 00), (238a)
or, equivalently,
(-00 < x < 00) . (238b)
FinaUy, if we write u = 2n:w, Equation (238b) becomes
(-00 < x < 00). (239)
Equations (235), (237), (238), and (239) are al! equivalent forms of tbe Fourier
integral.
If we denote the inner integral in (238b) by j(u) = ~{f(x)}, there follows
(-00 < u < 00), (240a)
f(x) = -1
2n: _~
-
f~ eiuxf(u) du (-00 < x < 00). (240b)
The functioD j(tI) defined by (240a) is called the Fourier transform of f. If the
Fourier transform of f is known, Equation (240b) permits the determination
of the function [in terms of an integral involving the tranSform j.
For a function f(x) which vanishes when x < O, tbe Fourier transform
beco mes formally identical with the Laplace transform (Chapter 2) if we replace
¡u by s and x by t.
In a similar way, we obtain from (232) and (233) the inversion formulas
fs(u) - S{f(x)} = r f(x) sin ux dx (O < u < 00), (241a)
f(x) = -2 f fs(u) sin u x du (O < x <oo ) (241 b)

n: o
and
fc(u) - e(f(x)} = r f(x) cos ux dx (0 < ti < 00), (242a)
f(x) = - 2
n:
1-
o
f c(u) cos ux du (O < x < 00). (242b)
5.15. The Fourier Integral 239
The func¡ions defined by (24Ia) and (242a) are called the Fourier sine and cosine
transforms of J, respectively.t
Sorne of the useful properties of these tran sforms are exbibited in Problems
89-92.
Example. To illustrate the definitions, we consider the function

< O),
G
(x
f(x) = (O < x < a), (243)
(x> a).
From (232) we obtain the Fourier sine integral representation,
x
f() =-
2 l~ 1 - cos au SlDUX
. du
(O < x < 00); (244a)
n o ti.
Equation (233) gives the Fourier cosine integral· representation ,
f(x) = 2 r~ sin au cos ux du (O < x < 00); (244b)

n Jo u
and Equations (235), (238b), and (239) give (hree equivalent forms of the complete
Fourier integral representation,
f()
x = n1 [(Sin
o -u-
au COS XII + 1-cosau.
u
)
SIn xu du (-co < x < co), (245a)
1 J~ 1 - e- Iou
f(x) = 2ni _ ~ "---=II"--e'u x du ( - co < x < co), (245b)
and
1 f~ 1 - e- 2 • lo",
f(x) = 2ni _ ~ ':"""--'w;';-'--e hiX ", dw (-co < x < co). (245c)
Also, from (240a), (241 a), and (242a) we obtain the transforms of f(x),
- 1 - e-jau
f(u) = . , (246a)
tU
- COS au
f s(u) - (u> O), (246b)
u
JCU
r ( ) _
---
sin au (u> O). (246c)
U
When X < 0, the integrals appearing in (244a) and (244b) represent - f( -x)
•
and f( -x), respectively. Thus, for example, taking ioto account the discon-
tAn important property of the transforms/s(u) and/cCu) is that both lend lO zero as ti --+ 00 if
/(x) is such thatthe integral S~ I/(x) I dx ex.ists. In addition, j(u) lends lo zero as u --+ ± 00 if
S':_I/(x) I dx exists. Clearly, corresponding statements apply to intcgrals of the form
S;;, F(u) cos L/X cW, SO' F(u) sin ux du, and S':- F(u) e iux du as X --+ 00 or x --+ ±oo.
240 Boundary-Value Problems and Characteristic-Function Representalions
tinuities in f(x), we have from (244b) the useful result

n
2 (Ixl < a),
r ~ sin au cos ux du =
n (247)
Jo u 4
(Ixl=a),
O (Ixl> a).
It should be noted that minor va riation s of the definition s used here also
appear in the literature . Specifically, e-Iu, and e'U' sometimes are interchanged
in (240a , b) and the fact or 1/(2 n) in (240b) may appear instead in (240a), or the
factor ]/--/2n may appear in both definit io ns , for the purpose of increased sym-
metry. Similarly, in (24Ia, b) and (242a, b), the factor 2/n may be tran sfe rred
t o the definition o f the transform or the factor --/2/ n may appear in both the
transform and its inverse, to bring about complete symmetry.
ID this text we denote by j both the Laplace transform (Chapter 2) and the
Fourier transform , continuing to use s to denote the transform va riable in the
former function and using a different letter (usually u, as here) for that purpose
in the latter one. The remaining notational inconsistency is not trouble so me
since the t wo transforms are no t used together.
Except when f(x) van ishes outside a finite interval, as in (243), the actual
evaluation of an integral specifying a transform (or an inverse transform) usuall y
must be effected by methods of so-called residue calculus (see Sections 10.13-
10.16), or by reference to table s.
,l' Analogo u s integral representations involving Bessel functions can be ob-
tained. In particular, it can be shownt th a t the equation
(O < x < 00) (248)
is valid when p > -1, with the usual convention regarding discontinuities, if
f(x) is piecewise differentiable in every finite positive interval and if the integral
f; xl f(x) I dx exists . This expression ts known as the Fourier-Bessef integral
representation of I(x), of order p.
The Hankel (or Fourier-Bessel) transform, of order p, of a function f(x)
may be defined by the relation
fH,(u) = <BAf(x)} = r xf(x) J .(ux) dx (O < u < 00 ), (249a)
after which the function f(x) is expressi ble in terms of its transform by use of
the equation
f(x) = r ufH,(u)J/ux) du (O < x < 00), (249b)

by virtue of (248) .
tSee Reference 5 of Chapter 4.

Problems 241
REFERENCES
1. BESKIN, L., and R . M. ROSENIlERG, "Higher Modes of Vibrarion by a Method of

Sweeping," J. Aero. Sci., 13: 597-604, 1946.
2. CARSLAW, H. S., In/roduclion lO Ihe Theory 01 Fourier's Series and Inlegrals, 3rd
ed., Dover Publications, Tnc., New York, 1952.
3. CHURCHILL, R. V., Fouriá Series and Boundary Value Problems, McGraw-Hill
Book Company, lnc. , New York, J941.
4. JACKSON, D. , Fourier Series and Orlhogonal PoIYllomia(s, Sixth Carus Mathe-
matical Monograph, Open Court Publishing Company, La Salle, 111., J941.
5. SNEDDON, 1. N., FO/l/"ier Translorms , McGraw-Hill Book Company, Tnc ., New
York, 1951.
6. TlTCHMARSH, E. C., Eigenlunc lion Expansions Associaled wilh Second-Order Dlf-
leren/ia( Equa/ions, Oxford University Press, lne ., New York, 1946.
7. WHlTTAKER, E. T., and G. N. WATSON, Modern Ana(ysis, 4th ed., Cambridge Uni-
versity Press, New York, J958.
PROBLEMS
Section 5.1
1. Show that the boundary-value prob1em
d 2y
- 2 - k2y
d.x = O1
y(O) = y(f) = O
canno! have a nontrivial solution for real values of k.

2. Determine those values of k for which the partial differential equation
a2 T a2 T
ax2 + ay' = o
possesses nootrivial so lutions of the form T (x, y) = I(x) sinh hy whieh vanis h when
x = o and when x = l .
d 2y
dX2 + Ax 2y = o
can possess a non trivial solution which vanishes at x = O and at x 1 only if A is
sueh that
JI .• CPI 2) = O
and that, eorresponding to sueh a charaeteristic number Ab any multiple of the fune-
tion
I/h(X) = Xl 2J, '(1 A1 'X2)
is a solution with the required properties.
242 Boundary-Value Problems and Characlerislic-Funclion Represenlalions

dOY..Ldy-,-, -O
x d x-' ' dx AXy- I
can possess a nontrívial solution which is finite at x = O and which vanishes when
x = a only if A is such that JoCA' °a) = O, and obtain the form of the solution in this
case.
Section 5.2
5. Suppose that the uniform string considered in Section 5.2 is unrestrained from
transverse motion at the end x = O, and is attached at the end x = / to a yielding
support of modulus k, the ends of the string being constrained against appreciable
movement parallel to the axis of rotation. Show that the nth critical speed W n is given by
where cot ¡.Ln = ct.¡.Ln and ct. = Tlkl.
6. Suppose that a uniform rotating string has both ends attached to yielding supports,
the modulus at x = O being k, and that at x = I being k" so that the end conditions
ct.,ly'(O) = y(O), ct. 2 Iy'(l) = - y(l)
are to be satisfied, where ct., = TI(k,/) and ct. 2 = TI(k 2 /). Show that the nth deflection
mode is a multiple of the function
where ¡.Ln is the 11th positive solution of the equation

(ct.,ct. 2 ¡.L;; - l}tan ¡.Ln = (ct. j + ct.,)¡.Ln
and that the nth corresponding critical speed is
wn = 11:; J~ .
(Notice that combinations of "fixed" or "free" ends correspond to combinations of
ct., and/or ct. 2 zero or infinite.)
7. The linear density of a flexible string of length I varies according to the law
p = Po(l + xl/)2, where po is a constant and x is measured from one end. The ends
of the string are attached to an axis rotating with angular velocity W.
(a) Show that the governing differential equation can be written in the form
2
ddr2
y + 4 fA'
11 'r2y = O,
where r = 1 + xII and ¡.L' = (Pow 2 / 2 )/(4T), and that the end conditions then require
that y vanish when r = 1 and when r = 2.
(b) Show that the 11th critical speed w n is given by
w n 2#o~n
=
problems 243
where J1 n is che nth solution of the equacion

JI .(J1) YI " .(4J1) - JI ·. (4J1) YI '.(J1) = o.
8. (a) lf a flexibl e s tring is caused lo vibrate transversely in aplane wich respect to a
st raight form along tbe x axis, the displacement IV of a point is a function of x and of
time l. Noticing that the inertia force per unit length is then given by -p(a 2IVla,2), and
denoting the actual transverse external force by f(x, r), show tbat IV satisfies the partial
%x (T~:) - p ~:: + f = O
when small slopes and displacements are considered.

(b) ]n the case of free periodic vibrations of circular frequency w, with f = O,
we must have
w(x , 1) = y(x) sin (WI + rp),
where rp is constant and the amplitude y depends only on x. Show that y(x) then
satisfies the same ordinary differential equation as the displacement of a s tring rotating
about the x axis with angular velocity W (Section 5.2) , and hence that the results of
Section 5.2 can be interpreted in cenns of natural vibration modes and natural fre-
quencies in the analysis of period ic vibrations.
9. Deal with the problem
T o ~~ + pW2yJ 1 + (~~r = O,
y(O) = O, y(/) = O
(see page 190) as follows:
(a) _By writing dyl dx = p and d 2yldx 2 = p dpl dy (se e Section 1.12), separating,
and integrating the resultant equal members, deduce the relation
-VI +p< = 1 + ex(l - ;D,

where Ym is the value of y when p = O and
= pw 2y;'.
ex 21'0
(b) By solving for p and defining a parameter k such tbat
k 2 = ex 2k 2
2 + ex' oc = 1 _ k<'
obtain the equation
~~ = ± 1~\2J(1 - ;D(l - k2;~).

(c) Restricting attention to tbe case when y has only one extreme value, and
noticing from the symmetry that then y = Ym when x = 112, show that
'J dy
y2)( y2) = ±
2k ( 1)
'2 '
J(
Y. 1 - -
y;'
1 - k 2 - ,
y;;,
1 - k' x -
244 Boundary-Value Problems and Characteristic-Function Representalioll.s
(d) Make Ihe substitution

y = Ym sin rp, rp = sin-t L
Ym
and obtain Ihe relation
drp _
- k 2 sin 2 rp-
± 2k
J-k 2
(x _~).
2
(e) By expressing Ym in terrns of k and using the notation
F (k, rp) = s: ...;--r1~---';'~~~:-s"in-"'2

=rp ,
rewrile Ihis resull in the form
where
rp = sin- tZ ,
Ym
(f) Show Ihat the end condition y(Q) = Q requires that k satisfy the equation
"';1 _ k K(k) 2 = .jpC:;:2,

where
K(k) = F(k, 1l )
2
= JO':2~
v
/ 1 ~dkrp""2"""'-""?-
- smrp
JI
[Thus the required solution is defined by the equation obtained in part (e), with the
constant k determined by the equation in part(f). The parameter k is called the modullls
of the elliptic integral F(k, cp); K(k) is called the complete elliptic integral of the tirst
kind.]
(g) Verify that, if we assume that Ym /I « 1, it follows that k « 1, and also verify
that,if k is replaced by zero in the relations obtained in parts (e) and (f), the resultant
approxima te . relations are
x = ; ± .f!j;(~ - sin- t :J,

where / pC0 21 2 = !!...
...¡ 4To 2
or, equivalently,
. 1lX
y = Ym SIn -1-
and CO = ~ J;,
in accordance with (16) and (17) when n = l.
Secllon 5.3
10. Determine expressions for the critical speeds and deflection modes of a rotating
uniform shaft of length I fixed at both ends, taking the origin at one end of the shaf!.
Problems 245
11. Sol ve Problem 10, taking the origin al Ihe cenler of the shaft. (Express the general
solution in terms of circular and hyperbolic functions and, by suitably combining the
four equations deterrnining the constants of integration, obtain an equivalent set
consisting of two pairs of equations each involving coefficients of either only even or
only odd functions. Notice that symmetrical and antisymmetrical modes are thus
readily distinguished.)
12. Proceeding as in Problem 8, show that the displacement w(x, 1) of a shaft or beam
executing smal! transverse vibrations in aplane satisfies the equation
a (a
2 2
a w .
2W)
ax2 El a X2 + p al 2 - j = O.
Show also that in the case of free periodic vibration the amplitude function satisfies
Equation (27).
13. A uniform beam is hinged at the end x = 0, whereas the end x = 1 is free. Show
that the natural frequencies of free transverse vibrations of the beam are given by
0). = J1.?;1-2,,¡' EI/p, where J1.. is a root of the equation tanh J1. = tan J1., and the deflec-
tion in a natural mode is given by
W n( X,
1) -
-
C [Sinh (J1.n x / l )
n cosh J.ln
+ sincos
(J1.n x / I )] .
J.ln sin ron!·
(See Problem 12.) Notice that the problem also admits the solution w = Cx when
O) = 0, corresponding to a rigid-body rotation about the end x = O.
14. Suppose that a mass kpl, equal to k times the mass of the entire beam, is attached
to the free end of the beam of Problern J 3.
(a) Verify that the condition of vanishing shear at x = 1 then is replaced by the
requirement 13 y "'(l) + kJ1.4y(l) = 0, where J1.4 = p0)214/EI, and wherey(x) is the
amplitude function.
(b) Show that J1. then must satisfy the equation
tanh J1. = tan J1. + 2k J1. tan J1. tanh J1.
or cot J1. = coth J1. + 2kJ1..
(Notice that this condition reduces to that obtained in Problem 13 when k = O.)
Section 5.4
15. By integrating Equation (46) twice and determining the two constants of integra-
tion, show that the differential equation for buckling of a long column hinged al bolh
ends can be taken in the form
d 2y
El dx2 + Py = 0,
with the two end conditions y(O) = y(l) = 0, if no transverse loads are acting.
16. An axial load P is applied to a column of circular cross section with linear taper,
so that I(x) = lo(x/b)., where x is measured from the point at which the column would
taper to a point if it were extended, and lo is the value of I at the end x = b.
(a) If the column is hinged at the ends x = a and x = b, show that the governing
differential equation can be put in the form
x' d y2
dx
2
+ "'" H2y = ° ,
where J1.2 = Pb 4 /El o, and where the conditions y(a) = y(b) = ° are to be satisfied.
246 Boundary-Value problems and Characleristic-Function Represenlations
(b) Show that the general solution of this equation can be expressed first in terms
of Bessel functions and finally in the elementary form
Y=x(clcos ~ +cosin ~)-

(e) Show that the critical loads are of the form
_ 2 2(
P n-n 1T. b
a) o Elo
/2'
where l = b - a is the length of the column, and that the buckling modes are given by
q!n(x) = x sin [nn ~ (1 - ~)}

17. An originally vertical colurnn of length lis assumed to bend under its own weight.
If its mass per unit length is denoted by p(x), where x is measured from the top of the
column downward, show that the shearing force S at distance x from the top is given
by the component of the load w(x) = g f~ p dx in the direction normal to the deftected
axis of the column, and hence is of the form S = - w tan where e, e
is the angle
of deflection from the vertical. Noticing that the bending moment is given by
M = El dejdx, and that S = dMjdx, obtain the governing differential equation in
the forro
d ( El dx
dx de) + w(x)e = o,
when small deflections and slopes are assumed, where, if the end x = O is unrestrained
whereas the end x = lis built in, the end conditions Ele'(O) = O and e(i) = O are to
be satisfied.
18. In the case of a uniform column, for which El = constant and w' = gOA = con-
stant, where O is the volume mass density and A the cross-sectional area, show that
the differential equation of Problem 17 takes the form
d2
dX2
e + gOA
El xe = O,
and that the solution of th.is equation for which e'(O) = Ois of the forro
e(x) = I 2
eX / J_ I !3( ~ Jgl: x3/2)-
where e is an arbitrary constan!. By imposing the condition e(l) = O, and making use
of the fact that the smallest zero of J- 1I3 (z) is z ~ 1.866, deduce that the smallest value
of l for which buckling may occur (the "critical length" of the column) is given by
. ( El ) 1!3
1" = 1.986 gOA .
Section 5.5
19. Determine an approximation to the Euler load PI of a column of length 1, hinged

at both ends, with bending stiffness
l
E/(x) = el + x'
Problems 247
where e is a eons tant, by the melhod of Stodola and Vianello. Take Equation (82)
as a first approxímation to the defleetion curve and compare the results of usíng Equa-
tions (70) and (72a) or (72b) in ealculating A\" = f2P ',Il/ C.
20_ Find approximately the smallest eharaeteristie value 01' A for the problem
dZy
dxZ + Ay = O, y'(O) = y(1) = O
by using the method of Vianello and Stodola as follows:

(a) Determine a Brst approximation, by first replaeing the term Ay by a eon-
veoient eonstant in the given differential equation, in the form y, = 1 - X'.
(b) Show that there follows f, = TS:(5 - 6x 2 + x·).
(e) Show that the use of Equation (70) gives A, =
2.5, whereas the use of (72a)
gives A, = H"':" 2.471.
(d) Show that the true value of A, is 7t 2 /4"':" 2.467.
21- Find approximately the smallest value of J,l for whieh the problem
d 2y dy
x -- - -
dX2 dx
+ J,l 2xy = O
'
y(O) = y(1) = O
possesses a non trivial solution, by using the method of Stodola and Vianello as follows:
(a) Show that the differential equation can be written in the forro
.!!...
dx xdx
(1.. dY ) + J,l'x Y = O.
(b) Determine the tirst approximation y,(x) in sueh a way that
.!!..(_l
dx x ,be
dY J) = eonstaot
'
and also y,(O) = y,(l) = O, and show that there follows y, = xl(l - x), with a eon-
venient choice of the constan!.
(e) Show that then f,(x) = rh-(7x ' - 15x· + 8x').
(d) Verify that the use of Equation (70), with A = J,l2, gives
J,l, = ../I5 ...:.. 3.873,
whereas the use of (72a), with p = l/x, gives
J,l, =,.¡w ..:. 3.839.
(e) Show that the true value of J,l, is the smallest zero of J, (x) and henee (see
Table 1 of Seetion 5.13) is given by J,l, ..:. 3.832.
SeC(jOD 5.6
22. Reduce eaeh of the fOllowing dilferential equations to the standard form
:)p ':tJ + (q + Ar)y = O:
d2y dy
(a) x dX2 + 2 dx + (x + A)Y = O,
d y 2 dy
(b) dx 2 + dx eot x + Ay = O,
2
(e) d y
dX2
+ a dx dy + (b + A)y = O (a, b eonstants),
y
. ~ X)ddX - ay + Ay = O
d2y
(d) -x dX2 .".' (e (a, e constants).
23. By eonsidering the charaeteristie funetions of the problem

d2y ,
dX2 + Ji.- y = O, y(O) = O, aly'(l) + y(l) = O,
with a :> O, and using the results of Seetion 5.6, show that
S: sin Ji.,x sin Ji.2X dx = O
when Ji.l and Ji.2 are distinet positive solutions of the equation tan Ji.{ + aJi.1 = O.
24. By eonsidering the eharaeteristie funetions of the problem
d 2Y. dy
X dx 2 dx + 2_
+
Ji. xy - O, y(O) finite, y(l) = O,
and using the results of Seetion 5.6, show that
S: XJ O(Ji.1 X )JO(Ji.2 X ) dx = O
when Ji.l and Ji., are distinet positive roots of the equation Jo(Ji.I) = O.
25. By eonsidering the eharaeteristie funetions of the problem
d 2y dy
(1 - X2) - - 2x- + Ay = O y(± 1) finite
dX2 dx '
and using the results of Seetion 5.6, show that
f, P,(x)P,(x)dx = O
when r and s are distinet nonnegative integers. [Notiee that the equation is Legendre's
equation, and that A must be of the form n(n + 1), where n is an integer, in order that
solutions finite at both x = 1 and x = -1 exist, aeeording to Seetion 4.12.]
26. If p(b) = pea), and if the eonditions
y(b) = a"y(a) + a ,2 y'(a),
y'(b) = a 21 y(a) + a 22 y'(a)
are assoeiated with the differential equation (90), show that the orthogonality property
(102) holds for two eharaeteristic funetions, eorresponding to distinet charaeteristic
numbers, provided that
a" aI21=1.
l C(21 C(22
27. Pro ve that if an end eondition imposed in a proper Sturm-Liouville problem req/lires
that
at an end poinf,
when rp 1 and rp2 are characteristic functions, then no characteristie number can correspond
to two linearly independent characferistic functions. [Suggestion: Assume, on the con-
trary, that rp 1 and rp, are linearly independent and that both eorrespond to the same
Problems 249
characteristic number (and hence both satisfy ¡he same differential equation). Show
that then p(x)[rp,(X)rp'2(X) - rp2(X)rp',(x)] is a cons tant (Abel's theorem), so that its
vanishing at a point implies its identical vanishing in (a, b). Then prove that rp2(X) is a
constant multiple of rp ,(x) in (a, b) and note the resultant contradiction.]
28. Prove, by the following steps, that the characteristic nllmbers of a real Stllrm-
Liouville problem with r(x) > °
are all real:
(a) Show that if ..1., is asstlmed to be nomeal, with an associated characteristic
function rp, (x), then X, also is a characteris tic number with an associated characteristic
function rp ,(x), which is the complex conjugate of rp ,(x).
(b) Deduce from (97) that if rp, (x) = ti, (x) + iv, (x), where 11 , and v, are real,
¡here follows
(A, - X,) s: rqJ,tP¡ dx = (..1., - X,) S: r(u; + vDdx = 0,
and hence, since r (x) > 0, conclude that ..1., is in fact real. [Thus rp, (x) also can be
taken to be real , by suppressing a possible imaginary constan t multiplicative factor.]
29. Prove, by the following steps, that the characteristic numbers of a proper Sturm-
LiOllville problem are real and nonnegative:
(a) Let A, and rp, (x) be corresponding real characteristic quantities (see Problem
28) and show that
A, f brrp¡ dx =
a -
[drp
prp, d.../ a
J
Jb + bP (dJx' )2dx
a - fb qrp'¡ dx.
a
(b) Show that the first term on the right vanishes unless a condition of type (98c)
is imposed at x = a and/or x = b, that a condition y + y,y' =
introduce the term -y¡p(a)[rp',(a)p, and that a condition y + Y2Y' =
at x = a would
at x = b
° °
would introduce the term +Y2P(b)[rp'(b)]2. Thus, by imposiog the requirements that
¡he problem be proper, deduce that A, :> O. Show also ¡ha¡..1., = O is a characteristic
number of a proper problem if and only if q(x) = O and the botlndary conditions are
satisfied by a constant.
30. Derive Equation (105) from (104).
Section 5.7
31. Determine the coefficients in the representation
f(x) = ¿;,
n~
A" sin nx (O < x < n)
in the following cases:
(a) f(x) = 1, (b) f(x) = x, (c) f(x) = E (x < 1</2),

(x = 1</2),
(x> 1</2).
32. (a-c) Show that the series obtained in Problem 31 cannot be differentiated term
by termo (Investigate the convergence of the result in each case.)
33. (a) If the expans ion f(.~) = ¿;;;'_, An sin nx is valid in (0,1<), s how formally that
~ i: A;.
I n
O
[f(x»)2 dx =
n-]
250 Boundary-Value Problems and Characterisric-Function Representations
(b) From this result, and from the results of Problems 31(a, b), deduce the fol-
lowing relations:
J j , l _ n 2
)2+3 2 1 52 ' ... ---g-'
34. Expand the funetion f(x) = 1 in a series of the eharaeteristie funetions of tbe
boundary-value problem
d 2y
d,+ Ay
- x- = 0,
y(O) = 0, ly '(I) + ky(l) = ° (k >- O),

over the interval (0, / ).
35. Consider the Sturm-Liouville problem
, d 2y dy
x- d J\:. 2 + X d-
X
+ 1y = 0, y (l ) = y(b) = 0,
where b > J.
(a) Show that the eharaeteristie numbers and funetions are
l. = (1~:b;2' rp.(x) = sin(n7l:~: ~) (n = 1,2, ... ).
(b) Ir f(x) is pieeewise differentiable in the interval (l, b), and if
f(x) = -
~ A. SIn
. ( nn
log -b X) (1 <x<b),
n=1 I og
show that
bf(x)sln. (n n X) -dx
A
f
. f,
,
log -
log b X
= "-'-"'b;---:----:-:--:---:--
si n 2 (nn _to_g_x) dx
log b x
and, by introdueing an appropriate ehange of variables, reduce this result to the form
A. = 2 So' f(b') sin nnt dI.
(e) Evaluate A. when f(x) = 1 and when f(x) = x.
Section 5.8
36. Obtain the solution of the problem
d 2y
dX2 + Ay = h(x) , y(O) = y(l) = °
in the form
..~ A" . n1tx
y(x) -
n_ I
~ A - n 21l 2/'2 s,n -,-
(O <: X <: ')
when A * (nn/')2 (n = 1,2, ... ), where A. is the nth eoeffieient in the expansion
~ A . nnx
h( x ) = n~1 n sIn - , - (O < x < 1),
assuming that h(x) is pieeewise differentiable in (0, 1).

Problems 251
37. Obtain the eondition whieh must be satisfied by a pieeewise differentiable fune-
tion h(x) in order that the problem
d 2y 2
n'
dX2 + p----¡-z- y = h(x). y(O) = y(l) = O
possesses a solution, when p is a positive integer, and express the eorresponding most
general solutian as a series 2: On sin (nnx/I).
38. Let h(x) sin (pnx/I) in Problem 36, where p is a positive integer.
=
(a) Show that the solution of that problem then is
sin (pnx/ 1)
y = A - (p2n'/12)
if A"'" (nn/l)2 (n = 1,2, ... ).
(b) If A = (rn/lp, where r is a positive integer but r "'" p, show that the solution
of part (a) beeomes
1 2 sin (pnx/I)
Y=1t2 r2_p2
and also show that to this solution ean be added any eonstant multiple of sin (rnx/I).
(e) Aeeount for the nonuniqueness of the solution in par! (b).
39. By introdueing the value of An (as an integral) inlO the result of Problem 36 and
interehanging the summation and integration (this ean be shown to be legitimate),
show that the solution of the problem
d 2y
- d2
x
+ Ay = h(x), y!O) = y(l) = O
ean be expressed in the form

y(x) = J: G(x, ';)h(';) d';
where G(x, .;) possesses the expansion

G(x ~) = 2 ~ sin (nnx//) sin (nn';fI) (O <: x <: 1, O <: ~ <: 1),
, 1 n~I A - (n 2n 2/12)
provided that A "'" n 2n 2/12 (n = 1,2, ... ). [He re G(x, ~) is the Green's function for the
expression d 2 y/dx 2 + Ay and the eonditions y(O) = y(l) = O. lt is given in elosed
form when A = l.and when A = O in Problems 55 and 59 of Chapter l.]
E.. (x ddxy ) -'--" ,Ax y --

dx h( x,
) y(O) finite, y(l) = O
111 the form
~ A
y(x) = 2: A n JO(J1.n x ) (O <: x <: 1),
n=1 - /J.n
when A "'" J1.r.

J1.L ... , where J1.n is the nth root of the equation J o(J1.l) = O, and
where An is the nth eoeffieient in the expansion
(O < x < 1),

assuming that h(x)/x is piecewise differentiable in (O, [). Show also that the coefficient
A. is 10 be delermined by the equation
' lo-,
An
f
o x[J O(.u nx)]2 dx = h(x)Jo(f./.nx ) d:c.
Section 5.9
41. With the abbreviation
Yk = f rrp¡ dx,
show formally that, with the notation of Sections 5.7 and 5.9,
f· rfJ dx = 1
12
Al k= 1
L~
•
and hence deduce that, if we write
f./.., = i; (~J)" YkAl;,

k == I Ak
then, in the Nth cycle, the approximation 10 A, yielded by EquatioD (72a) is

(f./.2N-2 / f./.2N-¡)A" whereas that yielded by (72b) is (f./.1N-, / f./.2N)A¡. (These results are
val id when the generating Sturm-Liouville is proper, as is assumed in SectioD 5.9.)
IX k = f:!Pk dx
and the notation ofProblem 41 and Sections 5.7 and 5.9, show formally that, in thenth
cycle,
f· 1 ~
fndx=,L
Al k-l
•
and hence, with the additional abbreviation
deduce that the approximation lO A, yielded by (70) in the nth cycle is (vn_tlvn)A,.
43. Wilh Ihe nolalion of Problems 41 and 42, suppose that the initial approximation
y,(x) is such that
~(p d Y ,) = 1.
dx dx
Problems 253
(a) Show that Equations (128) and (129) then imply also that
~ y l
dx ( P ddx ) = L;1
- n~ AnAnr(x)¡Pn(x),
and hence also
=
r(x) L; AnAnqJn(x) = -1,
n=l
when a < x < b. Thus deduce that in this case there follows
(J,k = -AkYkAk,
and hence also
(b) Deduce that, with the assumed determination of y,(x), the use of Equation
(na) gives the same approximation to Al in the Nth cycle as does the use of (70) in the
(2N)th cycle, and also the use of (nb) in the Nth cycle gives the same approximation
as does the use of (70) in the (2N + 1)th cycle.
44. Use results of Problems 41 and 42 to show that, when A, O and A 2 O in a * *
Stodola-Yianello sequence, the error in the approximation afforded to a multiple of
qJ,(x) by a multiple of JN(X) in the Nth cycle is small of order (A,/A2)N, whereas the
error in the approximation afforded to A, by use of either Equation (na) or (72b) in
that cycle is small of order (AI/A2)2N. (lf A 2 = O but A, *
O, the ratio A,/A 2 is to be
replaced by A,/A" and so forth.)
45. Jf the initial approximation y,(x) is orthogonal to the first characteristic function
qJ 1 (x) with respect to r(x), show that the method of Stodola and Yianello leads to the
second characteristic quantities A2 and qJ2(X) unless y, (x) is also orthogonal to qJ2(X).
46. Show that zero is a characteristic number of the problem

d 2y
dX2 + Ay = O, y'(O) = y'(l) = O,
corresponding to the characteristic function qJ(x) = 1, in addition to the characteristic

numbers An = n 2:n: 2 (n = 1, 2, ... ), corresponding to the respective characteristic func-
tions qJnCx) = cos n:n:x. Hence deduce that the method of Stodola and Yianello would
lead to A = O and y(x) = 1 in this case unless the initial approximation y,(x) were
orthogonal to the function y = 1 over (O, 1), that is, unless
So' YI(x)dx = O.
47. Apply the method of Stodola and Yianello to the approximate determination of
the smallest nonzero characteristic number of the problem considered in Problem 46,
assuming y, (x) in the form
YI(x) = Co + c,x + C2X2 + c,x J ,
and determining the constants such that
y',(O) = O, y'I(l) = O,
and I y,(x) dx = O.
Show that the resultant initial approximation y, = 1 - 6X2 + 4x', corresponding to

254 Boundary-Value Problems and Cbaracteristic-Function Representations
a eonvenient choice of lhe arbilrary multiplieative factor, leads to the result

J, = ~1l( I - 5X2 + 5x' - 2x')
if the same three eondilions are imposed on J,. Notice thal Equation (70) cannot be
applied here, bur thar the simple proeess of equaring y, (x) ro )'J, (x) at either x = O
or x = 1 gives).l "'" 10, whereas the use of (72a) gives)., "'" V,6 ~ 9.871. (The exact
result is )., = n 2 ~ 9.870.)
Section 5.10
48. Establish, by direer integration, rhe truth of (a) Equation (140), (b) Equation
(146), and (e) Equations (161).
49. Expand eaeh of the following funetions in a Fourier sine series of period 2/, over
the interval (O, 1), and in eaeh case sketch the funerion represented by the series in the
interval ( - 3/, 3/):
<
(a) J(x) = eU- x)
(x <
(x> O),
O),
(b) J(x) = {~
1- x
(x
(O
(x> 1/2),
O),
< x < 1/2),
< 1/2),
{~
(x
(e) J(x) = (d) J(x) = sin ~-~,
(x > 1/2),
(e) f(x) = '1

Isin nx (O < x < 1/2),
0
1
( otherwise),
(f) f(x) = j i (O < x <
(olherwise).
E < 1),
50. (a-f) Expand eaeh of the funetions listed in Problem 49 in a Fourier eosine series
of period 2/, over the interval (0,/), and in eaeh case sketch the funetion represented
by the series in the interval (-3/,3/).
51. (a) Obtain the expansion
sin na ~ 2a sin na
eosax =
na + ~ (-l)n
n= I n (a-' - n ")eosnx
(-n -< x -< n),
when a is nonintegral.
(b) Deduce [rom thi s result that
eot na = -1r.1(1
-a - ~ 2
~
,,""In-(J.
2a)
2'
when a is nonintegral.
52. Plot the fírst three partial sums Sn(x), for whieh
S,(.':") = sin x, 5 2(X) = sin x - i sin 2x,
and s,(x) = sin x - 1: sin 2x + j sin 3x,
of the righl-hand member of the relation
x <o ( - I)k+' .
-2 = k-~, k S1l1 kx (O -< x < n),
over the interval (0, n), and compare lhese plots with a plot of the limit function.
Problems 255
53. Plot the first three distinet partial sums of the series representation
x(n - x) = ~ (sin x + 3\ sin 3x + 5\ sin 5x + ... ) (O <: x <: n),
over the interval (O, n), and compare these plots with a plot of the limit funetion.
54. Use (124) to solve the problem
d 2y .
dx2 + Ay = SIn wx, y(O) = y(l) = O
when A "'" p 2n 2 /12 and w 2

"'" p 2n 1
//2 (p = O, 1, 2, ... ).
55. Investigate the excluded cases in Problem 54, assuming ¡hat w >- O.
56. Show that, if we determine the eoeffieients in an N-term approximation of the form
~ . nnx
f (x) = "'" a
n- 1
n SIn -I- (O <: x <: 1)
in sueh a way that
JJf(X) - ntl a nsin n~xr dx = minimum
(that is, by the method of leas! squares), the eoeffieients are determined in the form
2
ak=T J' . knx
of(x)sIn-,-dx (k=1,2, ... ,N),
that is, as the first N eoeffieients in the Fourier sine series representation. (Differentiate
the quantity to be minimized with respeet to a general eoeffieient ak and equate the
result to zero.)
Section 5.11
57. (a-f) Expand eaeh of the funetioDs listed in Problem 49 in a Fourier series of
period 2/, over the interval (-1,/), and in eaeh case sketch the funetion represented by
the series in the in terval (- 3/, 3/).
58. Expand eaeh of the following funetions in a Fourier series of period equal to the
length of the indieated interval of representation:
(a) f(x) = a + bx (O < x < P),
O ( O)"
(b) f(x) = { x < (-1 <x< 1),
1 (x> O)
(e) f(x) = sin x (O <: x <: n),
(d) f(x) = x (1 < x < 2).
59. Prove that the Fourier series of period P whieh represents an odd-harmonie fune-
tion of period P [satisfying (I62)] involves only the odd harmonies.
60. Obtain lhe Fourier series of period 2n whieh represents the solution of the problem
d 2y . _
y( -n) = yen), y'( -n) = y'(n)
dx2 .., Ay - h(x),
when
< x < O),
h(x ) = E (-n
(O < x < n /2).
(n / 2 < x < n).
assuming that A * p2 (p = o, 1,2, ... ).
61. (a) If the representation
n7rx
+ ~I
00
f(x) = Ao An cos -/- (O -< x-< /)
is valid, show formally that
T2 f'
_o
[f(x)]2 dx = 2Ai) -
+ n~ A;.
(b) If the representation
.• rm.x
f(x) = ~ En sin -/- (O < x < 1)
n- I
is valid, show formalIy that
T2
-o
f' [f(x)J2 dx = n~
= B;.
(e) If the representation
f(x) = Ao + J; (An cos n~x + Bn sin n~x) (-/ < x < /)
+r
is valid, show formally that
[f(x)]2 dx = 2A~ + n~ (A; + Bn.

[It is a remarkable fact that the reslI/ts of parts (a), (b), and (e) are valid whenever
í f(x) is bounded and integrable over the relevant interval, whether or not the associated
Fourier representations referred to are valip . These results are often called the Parseva/
I eqllalities.)
62. (a) Sbow that the Fourier series expansion described by Equations (154) and (155)
f(x) =
I
2/ r' f(t) dt =
+ n~ T1 r' ( f(t) cos rm.x
-/- cos nn
-/-t + . n7tx . rm.t) dt
Sin - / - SID - /
... -1 .... -1
= 2/ r'
I _ -1 f(t) dt - J'
+ n~' TI -1 f(t) cos nn(x/ - t) dt
when - / < x < 1.

(b) By making use of Euler's formula
n1t ( X - 1) 1 inrd x -r } _i"nb:-r)
cos / = Z[e I + e I J,
show t hat tbe las t form can be t ransformed to
f(x) = n.'-+:- 2/I J' -1 f(t)e

innlx-')
1 dt
Problems 257
or. eq uivalently,
¡nlT ...
f(x) = ¿; On e-'- ( - / < x<1)
fI "'" - .,
where 0n = 2/1 J'/ .- ~ dx.

f(xle'
-/
(This is known as the comp/ex form of (he FOllrier series.)

63. Determine the eoeffieients in the eomplex form of the Fourier series of period
p = 2/ (Problem 62) representing eaeh of the foJlowing funetions in (-1,/):
(a) f(x) = 1, ( b) f(x) = eos x,
-1 (-/ < x < O),
(e) f(x) = e"X, (d) f(x) = { 1
(O < x < 1).
64. (a) Ass uming the relation
(see Problem 42 of Chapter 4), replaee r by e lu and deduee the Fourier expansion
~
e{x sin (J = ¿ J,,(x)é"fJ,
(b) By equating separately the real and imaginary parts of the two sides of this
equation, and reealling that J_n(x) = (_l)nJn (x), show that
eos (x sin e) = Jo(x) + 2e + J.(x) cos 4e + ... ]
2[J,(x) eos
and sin (x sin e) = 2[J,(x) sin e + J,(x) sin 3e + ... ].
65. (a) From the result of Problem 64(a) deduee the integral formula
Jn(x) = 2~ f~" el/x ,1" .-nUI de
= nl rn
j o eos (x sin e - ne) de,

when n is an integer.
(b) From this result or from the result of Problem 64(b), also obtain the relations
Jn(x) r"
= n1 jo cos (x sin el eos ne de
= ; In. 1
COS (X sin e) eos ne de (n even);
Jn(x) = ~ 1" sin (x sin e) sin ne de
= ;; 1" 2 sin (x sin e) sin ne de (n odd).
[Notiee that the integrands in part (b) are symmetrieal with respeet to e= Te/2, with
the stated restrietions on n .]
(e) In particular, show that
Jo(X ) = -:¡¡
2 f"
o 2 cos ex sin e) de
2 J'~ l cos xt
= -:¡¡ oV I - lid !.
Section 5.12
66. Derive the statements made in Section 5.12 relative to term-by-term differentiation
of (a) the Fourier cosine series and (b) tbe Fourier sine series.
67. If f(x) and f'(x) are continuous and f"(x) is piecewise differentiable in
(a, a + P), show that the series obtained by two term-by-term differentiations of the
complete Fourier series (157) converges to f"(x) at each interior point of (a, a + P)
at which f "(x) is continuous, provided that
fea + P) = fea), f'(a + P) = f'(a),
and also converges to f"(x) at the end points if also f"(a + P) = f"(a).
68. If f(x) and f'(x) are continuous and f"(x) is piecewise differentiable in (O, [),
derive the following statements:
(a) The series obtained by twice differentiating the Fourier sine series (138) term
by terro converges to f"(x) at each interior point of (0,1) at which f"(x) is continuous
provided that
feO) = f(l) = O,
and also converges to f"(x) at an end point if f"(x) = O at that point.
(b) The series obtained by twice differentiating the Fourier cosine series (147)
term by term converges to f"(x) at each interior point of (O, [) at which f"(x) is con-
tinuous, and at both end points, provided that
1'(0) = 1'(1) = o.
69. Show that the results of Problem 68 do not permit two term'by-term differentiations
of the expansions of text Examples 1 and 2 and verify that in fact the resultant series
would diverge at all interior points in both cases.
70. Term-by-term integration of FOllrier series. Show that the result of term-by-terro
integration of the formal series
,,.~ nnx
f(x) = Ao + n~' An cos -,-
over the interval (O, x),

x
I
00 ,
f(¡)dl = Aox + ~ - An sin fT7C,X,

o n~ 1 nn
is valid when O <: x <: I provided that the A's are calculated from (148) and that f(x)
is piecewise continuous in (O, '). [Let F(x) = f ~ f(l) di - (x/I) g f(l) dI and expand
F(x) in a Fourier sine series over (O, [), noticing that F(O) = F(I) = O and that
F(x) = f(x) - - Oi') Jlo f(l) di, where f(x) is continuous. Here the origina' cosine
Problems 259
series need not converge lo j(x). Similar conclusions follow for Ihe sine series and for
the complete series.]
Serlion 5.13
71. Expand the function j(x) = xP In a senes of Ihe characteristic functions of the
boundary-value problem
x:!x (x ~~) .L (¡1h 2 - p2)y = O, y(O) finite, y(l) = O
over the intervalO < x < 1, where p is a given nonnegative constan!. [Make use of
Equation (108), Section 4.9.]
72. Expand j(x) = x P in a series of the characteristic functions of the boundary-
value problem
dy
x..E... (x ) + (¡.t2 X ' - p2)y = O y(O) finite, y'(I) = O
dx dx '
over the intervalO < x < 1, where p is a given nonnegative constan!. (Consider the
case p = O separately.)
;,(xCZJ + Axy x, = y(O) finite, y(l) = O
in the form
y = 2: 2 o JO(¡.tn x ) (O -< x -< 1),
n~ ¡ A - ¡.t; ¡.tn LJ ¡ (¡.tnl)
where JO(¡.tnl) = O (n = 1,2, ... ), if J o("'; Al) 7"= O.
74. If A = O in Problem 73, show that the solution is of the form y = i(x 2 - /2).
Hence deduce the representation
I - (~)2 = 2: J,Jo((Xn.y)
(O -< x -< 1),
1 n~¡(Xn J¡«(Xn)
where (Xn is the nth zero of Jo(x).
Section 5.14
75. Show that the coefficients in the expansion
F(rp) = ¿; AnPn(cos rp) (O < rp < n)

n- o
are of the form
An = 2n i 1 I F(rp)Pn(cos rp) sin rp drp (n = O, 1, ... ).
e
76. Find the fust three coefficients in the expansion of the function
(-1 < x < O),
j(x) =
(O < x < 1)
in a series of Legendre polynomials over the interval (-1, 1).
77. Find the lirst three coefficients in the expansion of the function
(O < rp < n/2),
F(rp) = {~OS rp
(n/2 < rp < n)
in a series of the form
~
F(rp) = I: AnP n(cos rp) (O < rp < n).

n- a
78. Show that if ¡he coefficients Aa, A" ... , An are determined in such a way that
1= [,(f(x) - [AaPa(x) + A,P¡(x) + ... + AnPn(x)])2dx = minimum,
so ¡hat the integral of the squared error in the approximation
(-I<x<l)
is least, ¡he coefficients are obtained (by requiring that al/aA k = O for k = O, 1, ... ,
n and making use of the relevant orthogonality) in the form
(k = O, 1, ... , n).
Hence, recalling ¡hat any polynomial of degree n can be expressed as a linear com-
bina¡ion of Pa(x), P,C-"), ... , Pn(x), deduce that the polynomial 01 degree n which best
approximates a lunction over (-), 1) in the least-squares sense consists 01 the sllm olthe
terms 01 degree not greater than n in the Legendre expansion 01 that IUllction over that
in ter val.
79. Determine the first three nonvanishing terms in the Legendre expansion, over the
interval (-1,1), of the function
h(X)={~E (1 x I < E),
(E<lxl<I),
in ¡he form
h (x) = AaPa(x) + A 2 P 2(X) + A.P.(x) + ... (-I<x<l)
where
80. Use the results of Section 5.8 to obtain the solution of the problem
d~ [O - X2) ~~J + ky = h(x), y(±I) finite,
where h(x) is defined in Problem 79, in ¡he form

- A
Y = ~a k - n(~ + l)Pn(x),
if k =;t= O, 6, 20, ....
81. As E - O in Problems 79 and 80, the function h(x) tends toward the "unit impulse
func¡ion.'· lt can be shown ¡hat ¡he series in Problem 79 does not converge when
E = O. However, the series solution in Problem 80 does then converge.
Problems 261
(a) Show that if h(x) tends toward the unit impulse function Ó(x), the solution
of Problem 80 becomes
-i 5 27
Y = TPo(X) - k ~ 6 P2 (x) +k ~ 20 P .(x) + .. ..
(b) Show that the same result is obtained if the formal expansion
is dealt with as though it were truly a convergent representation.

[In the procedure of part (a), the limit is taken after the coefficients of the expan-
sion have been substituted into the general solution, whereas in that of part (b) the
limit is taken formally befare substitution. The former is mathematically sound; the
second, strictly speaking, is not. However, both give the same final result. Since in
many similar problems the second procedure may be much more simply applied than
the first, it is rather frequently used in practice and can be justified from a practical
point of view by an analysis similar to the one just considered.]
82. (a) With the notation of Section 4.12, show that the solution of Legendre's equa-
tion of order p which vanishes when x = O must be a multiple of vp(x). Making use of
the fact that vp(x) is not finite at x = 1 unless p is an odd integer, deduce that the
problem
!!.-[(1 - X2) dx
dx
d Y ] + Ay = ° '
Y(O) = O. y(1) fini te
possesses the characteristic numbers An = 2n(2n - 1) and the corresponding charac-

teristic functions !Pn(x) = P 2n -, (x), where n = 1,2, ....
(b) In a similar way, deduce that the problem
.!!...
dx
[(1 - X2) dx
y
d ] + A.y = O
'
y '(O) = 0, y(l) finite
possesses the characteristic numbers A. n = 2/1(2n + 1) and the corresponding charac-

teristic functions !Pn(x) = P 2n(x), where /1 = O, 1,2, ....
83. (a) Obtain the solution of the problem
!!.-[(I -
dx
y
X2) d ]
dx
+ Ay = h(x)
'
y(O) = O, y(1) fini te
in the forrn
y = n~' A _ 2~2n _ l)P2n -'(x), Cn = (4/1 - 1) l' h(x)P 2n _,(x) dx,
if A =7'= 2, ]2, 30, ... .

(b) If the condition al x = O in part (a) is replaced by the condition y'(O) = O,
obtain the solution in the form
y = n~o A _ 2~n2n + OPd x ), Cn = (4n + 1) l' h(x)P 2.(x) dx,
if A =7'= 0, 6, 20, ....

262 Boundary- Value Problems and Characteristic-Function Representations
Section 5.15
84. With the symbo ls

nn n
Un = -,- = n .ó.u. ó'u = T
where Un is a funetion of the integer 11, inereasing by jumps of magnitude Ó.u, show
that the eomplex form of the Fourier series representation of f(x), as obtained in
Problem 62, can be expressed in the form
L: Ó.u
f(x) = - l ' ·
2n n---
S'
-1
f(/)e-;u"U-x) dI.
[Notiee the formal similarity between this form (as I ~. =) and the Fourier integral
representat ion expressed in the form (238a).]
85. Use (244a) 10 evaluate the integral
lo
~ .!.-.---===
- eos au. d
sin ux u
u
for all real va lues of x when a > O.
86. (a) Ir a is a positive real eonstant, determine the Fourier sine and eos ine integral
representations of e-U in the forms
r«< = 1=(; a' ~ uo ) sin l/X du = 1-c; a' ~~ U 2) eos l/X du (a> 0, x> O) .
(b) Use these results to determine funetions A(u) and B(u) sueh that
a2 ~ X2 = S:~ A(u) sin ux du, 2 a

a + x
2 = f- o
B(u) cos ux dl/ (a> 0, x > O).
(e) Deduce from pre ceding results that

c>{'a2 +
X } n - xu , u
O) X2 = "Te
and
ef a '1 _ !!... - au
ta'+x 2 f-2 e ,
when a > O.
87. Show that a can be replaeed by a + ib, where a and b are real and a > 0, in Prob-
lem 86(a). In particular, show that
e
-ax( b
eos x -
.'
1 Sin
b)
X = n2 Jor~ (a2 _ u sin ux
b2 + u2) + d
2iab u
when a > °
and x > O and, by equatlng real and imaginary parts of the equal me m-
bers, deduce the sine integral representations of e- ax eos bx and e- ax sin bx when a > O.
88. (a) If a is a positive real constant, determine the comp lex form of the Fourier
integral representation of e- alx ) in the form
e-xix ) = S~ (...!.. a )e'ux du (a > O).

_= n a' + u'
Problems 263
(b) Deduee from this result the funetion C(u) sueh that
""'_ac----, = J~ C(u)e'ux du (a> O).

a" + X" -0
(e) Deduee that
g: j a 1. = ne-"iul
lIX 2 + X2J '
when a> O.
89. Assuming that r(x) and f"(x) are eontinuous (or, more generally, that integra-
tion by parts is permissible when needed), and that the relevant transforms exist, obtain
the following relations:
(a) (f[J'(x)} = iuí(u), (f[J"(x)} = - U 2í(u)
if f(±co) = r( +co) = O.
(b) S[J'(x)} = -ufc(u), S[J"(x)} = uf(O+) - ulfs(u)
if f(co) = r(co) = o.
(e) e[J'(x)} = -f(O+) + ufs(u), e[J"(x)} = -r(O+) - u 2 fcCu)
if f(co) = r(co) = O.
90. Transforms of the delta funetion.
(a) By replaeing f(x) by f(:e)/a in the text Example and eonsidering the limit of
the Fourier transform of the result as a - , O, indieate the sense in whieh it is said that
"the Fourier transform of the delta funetion .5(x) is l."
(b) Generalize the resuIt of part (a) to show that
g:¡ .5(x - a)} = e-la"
and, similarly, that
S¡O(x - a)} = sin au (a> O), e[o(x - a)} = eos au (a > O).
91. The eonvolution. Let the eonvolution of f(x) and g(x) in (-ca, col be defined as
the funetion
F(x) = [0 f(x - ,,;)g(,,;) dI;,
assuming the existenee of the integral.

(a) Assuming also that F(x), f(x), and g(x) have the Fourier transforms F(u),
í(u), and g(u), respeetively, show ¡hat
F(u) = í(u)g(u),
and henee ¡hat if í and g are the Fourier transforms off and g, then íg is the Fourier
transform of the eonvolution off and g. (Notiee that f and g ean be interehanged in the
definition of the eonvolution F.)
(b) If f(x) and g(x) both vanish when x < O, and F(x) is their eonvolution, show
that also F(x) vanishes when x < O and that
F(x) = f: f(x - ,,;)g(,,;) dé, (x> O).
(Compare the definition in Seetion 2.9.)

264 Boundary-Value Problems and Characteristic-Function Representatioos
92. Convo/u/ion proper/ies of sine and eosine /ransforms.

(a) Show that the result of Problem 91(a) can be written in the form
f
~
_~, f(x - ';)g(';) el'; =
1 f~
2n _~ e'""f(u)g(u) duo
-
(b) When f(x) and g(x) are both even funetions of x, show thatj and g are even
funetions of 11, and also that j(u) = 2fcCu) and g(u) = 2gcCu), where fe aod ge are the
eosine transforms of fand g. Thus deduce that, when x > O,
fJ(x - ';)g( -.;) d,; + J: f(x - ';)g(';) d'; + r 1('; - x)g(';) d';
=..! r~ eos uxfe(u)gcCu) du

n Jo
and henee
1
"2 Jor~ [f(1 x - .; 1) + f(x + ';)]g(';) de; = n2 Jor~ eos uxfe(u)gcCu) du,
so that, if fe and ge are the eosine transforms off and g, then fege is the eosine transform
of /he funetion
-1 r[f(1 x - .; 1) + f(x + ';)]g(';) d';.

(e) By supposing that f(x) is even aod g(x) is odd, aod proeeeding as in part (b),
show that if fe is the eosine transform 011 and gs is the sine transform of g, /hen fegs is
the sine translorm of the jime/ion
-1 f: [f(1 x - .; 1) - f(x + ,;)]g(';) d';.
[Notiee that the sine and eosine transforms require the definition of land g only
for positive argllments, and that the extended definitions for negative arguments in
parts (b) and (e) are merely for the purpose of appropriately using the result of part
(a).]
93. Sol ve the problem
y(±co) = O
by introdueing the Fourier transformo using Problem 89 to show that

_( ) _ _ h(u)
y u - u" + k"
and using Problems 88 and 91 to deduce that
when k> O.
94. Solve the problem
d'y
dX2 - k2y = h(x) • y'(O) = y~, y'(=) = O
Problems 265
by introducing the Fourier cosine transform, using Problem 89 lO show that

_ y~ + hC(l/)
Yc ( l/ ) - - l/2 + k2 '
and using Problems 86 and 92 lO deduce Ihat
when k> O.
95. Solve Ihe pro blem
d 2y
dX2 - Py = h(x) y(O) = Yo, y(co) = O
by inlroducing Ihe Fourier sine transform, using Problem 89 10 show that

"Yo - hs(u)
Ys ()
u = l/2 + k2 '
and using Problems 86 and 92 lo deduce Ihat
y(x) = yoe- kx - 2~ l~ [e-klx-~I - e-k(x+~)lh(';) d';

when k> O.
96. Solve Ihe problem
dy
.!!...(x
dx dx
) _ k 2 xy = xh(x) y(O) finile, y( co) = O
in Ihe form
()
Y x = - Jor~ u'2uhB,(u)
+
J (
k2o UX
)d
u,
where hH.(U) = L~ xh(x)Jo(ux) dx.

97. Tt is required to determine C(u) in such a way that the representalion
¡(x) = n21~o C(u)(a: sin ux + flu cos ux) du

is valid when O < x < co, where a: and fl are prescribed constants and a:fl > O.
(a) Assuming that such a representation exists, and that differentialion under the
integral sign is permissible, show that, if we write
F(x) = n21~o C(u) sin l/X du,

there follows
{1F(x) + a:F(x) = ¡(x), F(O) = O
and C(u) = L= F(x) sin ux dx.

266 Boundary-Value Problems and Characteristic-Function Re(r,sentatioDs
(b) Deduce that
F(x) = - 1 IX f(t)e- (<< /Pl(x -t ) dt

f3 o
and, by formally interehanging the order of integration in the result of introdueing
tbis expression ioto the expressioo for C(u), and ehanging the dummy variable of
integration in the result, obtain the evaluation
C( ) = I~ f( ) IX sin ux + f3u cos ux d x.

u
o
X
IX
2
+ f32 u 2
(e) Verify that the representation obtained reduces to the Fourier eosine and
sine representations when IX = O and when f3 = O, respectively. [The result can be
shown to be val id under the same restrietions on f(x) as those whieh apply in those
speeial cases when a.f3 > O. Ao additional lerm is needed when 1Xf3 < O.]
(d) Show that the representation obtained expresses f(x) as a superposition of the
eharaeteristie funetions of the problem
d 2y
dX2 + Ay = o, f3y'(0) = lXy(O), y(=) finite
when 1Xf3 > O, bul Ihat when 1Xf3 < O Ihere is a missingeharaeterisliefunelion e-I«/Plx,
corresponding to the characteristie number .A. = -1X2/ f32.
(e) Verify Ihat, with the definition
qu) = f",,(u) ,
IX + f3u
there follows
fM(U) = IX~ : /':. [ f(x)(1X sin tIX + f3u eos ux) dx
aod
f(x) = 2 r~ 1.\1(U) IX sio ux + ~(eos ux duo

7t Jo IX + u
Show also Ihat 1.\1(u) reduces to feCu) when IX = O and to fs(u) when f3 = O. [Since
fM(") eorresponds lo the "mixed" end condition f3y' (O) - lXy(O) = O, it might be called
a Fourier mixed transform .]
98. Solve the problem
~;: - k2y = h(x) , lXy(O) - f3y'(O) = y, y(=) = O,
when 1Xf3 > O, by use of Ihe Fourier "mixed transform" (Problem 97), as follows:
(a) Show tha! the mixed transform of f"(x) is
~~ ~ ~~~~[lXf(O+) - f3 /'(0+)] - u 21.,/(u).
(b) Use this result 10 show that

YM(U) = (IX + f3u)u _ hM(u) .
1X2 + f32u2 ,,2+ k2
Problems 267
(e) Deduce Ihal the solution can be written in the form
.(x) = 2y r~ l/(a sin l/X + flu cos ux) du _ 2 r~ h.,>,(u)(a sin l/X + flu cos ux)
y 7t Jo (a 2 + fl 2 u 2 )(U ' + k 2 ) 7t Jo (a + flu)(u 2 + k2) du,
where
hM(l/) = a~ : t:21~ h(x)(a sin ux + fll/ cos ux) dx.
[See footnote on page 239 with respect to the condition y( co) = O. The result of part
(e) can be reduced to the form
y(x) = y e-k.< - ..!... r~[e-klx- O. This form can be deduced directly (and more easily) from the result of
part (b), by making use of results corresponding to those of Problems 86 and 92 re!e-
vant to the mixed transform.]
99. Show formalIy that the function
T(x, y) = Sa~ e-Y"A(u) sin ux du (y> O)
satisfies the partia! differential equation

aZT aZT
axz + ay' = o
and vanishes when x = o and when y ----> co. Determine A(u) so that T(x, y) formalIy
reduces to a prescribed function f(x) when y = O and x is positive.
100. Show formally lhal the function
T(x, y) = ; [ l~~ e-Y" f(l) cos u(t - x) di du (y> O)
satisfies t he partia! differen tial equation

aZT aZT
ax Z +a y 2=0,
vanishes when y ---+ co, and reduces to f(x) when y = O.

101. Show formalIy thal the function
rp(r, z) = Sa~ Sa~ e-'"ulf(t)Jo(ut)Jo(ur) di du

satisfies the partia! differential equation
vanishes when z ---+ co, and reduces to f(r) when z = o.

268 8oundary-Value Problems and Characleristic-Function Representations
J02. (a) If ¡(x) = x P when O < x < a and ¡(x) = O when a < x < ce, show that
the H anke l tran sform of ¡(x), of order p, is
1
CBp(xpl = - a P + I J p+ Jau).
u
(b) Deduce that
! (~r (O < x < a),
2
O
1
a(; r (x
(a <x < ce).

= a),
103. Verify that the result of Problem 102(b) beco mes equivalent to Equation (247)
whenp = -1/2.
6
Vector Analysis
6.1. Elementary Properties of Vectors. A vector quantity is di s ti nguisbed

from a scalar quantity by the fact tbat a scalar quantity possesses only magni-
tilde, whereas a vector quantity possesses both magnitude and (except for the
zero vector) direction . It is conventional to represent a vector geornetrica lly as
an arrow, pointing in the direction associated witb the vector, and having a
length proportional to the associated magnitude. Thu s, in particular, the posi-
tion of a point B relative to a point A can be completely described by a vector
AB [rom A to B, in the sen se that the vector AB specifies both tbe distance a nd
direction from A to B and, in fact, indicates the displacement from A to B.
Since motion from A to a tbird point C can be accornplished along the vector
AC or, alternatively, along tbe vector AB to B
and thence along OC to C (Figure 6.1), it is natu-
ral to extend the concept of addition to vector
quantities by writing
---> ---> ~
AC= AB + BC.
We say that two vectors are equal if tbey
have the same direction and tbe same magni-
tude. In accordance with tbis definition a vector
-.
is unchanged if it is moved parallel to itself in AB
A
any way ; that is, the actualposition of tbe vector
in space can be assigned at pleasure. It should Figure 6.1
be noted, however, that in sorne applications it
may be necessary to specify the position of a vector (as, for exarnple, wben
a momenl associated with a forc e vector is to be determined).
269
270 Vector Analysis
\Vith this convention,t any two vectors a and b can be represented as

arrows so placed that the terminal point of a coincides with the initial point of
b. The sum of a and b ,
e = a --;- b,
is then defined as the vector extending from the initial point of a to the terminal
point of b. From the definition it follows readily that vector addition is com ..
mlltative,
a+b = b+a (1)
and also tha t the associative la w
(a + b) + e = a + (b + e) (2)
is satisfied. Tbe common value of the two expressions in (2) is written as the
sum a + b + e of the three vectors, and the definition is readily extended to
th~ sum of any number of vectors.
If the initial and terminal points of the arrow representing a vector are
interchanged, the resulting vector is called the negative of the given vector.
Thus, for example EA = -AB. To Sl/btract a vector b from a vector a, we
add the negative of b to a,
a - b = a + (-b). (3)
lf a vector a is multiplied by a scalar m, the result is a vector ma whose
magnitude is the arithmetic product of m and the magnitude of a and whose
direction is that of a if mis positive or that of -a if m is negative.
A vector of unit length is called a l/nit vector. lt is c1ear that any vector can
be written as the product of its length and a unit vector.
L
A vector of zero length (and arbitrary direction) is called a zero vector and is
denoted by O. The difference between two equal vectors is a zero vector.
We now consider a right-handed rectangular coordinate system, in which
the coordinates x, y, and z are measured along three mutually perpendicular
axes such that rotation of the x axis into the y axis about the z axis is accom-
plished by the right-hand rule, and define unit vectors i, j, and k having the
directions of the positive x, y, and z axes, respectively (Figure 6.2). Then it is
I readily verified that any vector v whose projectioos on these axes are v" V y '
and v" respectively, can be written as the vector sum
The oumbers v" v y , and v, are called the scalar components of v in the x, y ,
(4)
and z directions (Figure 6.3). If, when the initial point of v coincides with the
origin of the system, the angles measured to v from the positive x, y, and z axes
tIn printed work a vector quantity is oflen denoted (as he re) by a boldface lelter. Thus a repre-
senls a vector, and a represents a scalar quantity. In wriuen work, the use of an arrow (o)
or of an underline (q) is convenienl; Ihe nolalion Ú (or iI) is often used lO denote a vector of
l/ni! length.
6.2. The Scalar Producl of Two Vectors 271
Iz
I
I
I
I
I
k
--/"""""---
r--~----- / Y
j y /
, r-----;----.Y
/
/
/ / uyj
/ /
/x / x
Figure 6.2 Figure 6.3
are denoted by (J" p, and y, respectively, .. there follows

V)f = v cos cx, vy = v cos p, v, =vcosy, (5)
where v is written for the length or magnitude of the vector v. The notation 1 v 1
is also frequently used in place of v. The numbers cos (J" cos p, and cos y are
called the direclion cosines of v and are frequently denoted by 1, m, and n,
respectively. Any three numbers (A, B, C) proportional to (/, m, n) are known
as "direction ratios." From the geometrical relationship
v = ,.Jv; + v; + v; (6)
relating v to the scalar components of v, it follows that the direction cosmes
satisfy the equation
f2 + m' + n = 2
cos 2 (J, + cos' P+ cos 2 Y = j. (7)
We see also that if we write Equation (4) in the form
v = v(i cos (J, +j cos p+ k cos y), (8)
the coefficient of v is a unit vector, and hence the vector
v, = i cos (J, +j cos p+ k cos Y (9)
is a unit vector in the direction ofv. Combining Equations (5) and (6), we have
Vx cos p _ vy
cos (J, = ,.Jv; + v; + v;' ,.Jv; + v; + v; (10)
cos y = / 2 • 2
""v x + v + v;
y
6.2. The Scalar Product of Two Vectors. Two types of products of two
vectors a and b are conventionally defined. The first type of product is called
the scalar, dot , or inner proc/L/Cl and is written as a • b or (a b). This product is
defined to be a scalar equal to the product of the lengths of the two vectors
and the cosine of the angle () between the positive directions of the two vectors
272 Vector Analysis
(see Figure 6.4), and hence is glven by the

I equation
1
,
,
1
1
a •b = ab cos (J. (1 1)
~"
1,
Since b cos (J is the projection of b on a,
whereas a cos (J is the projection of a on b, it
I ' follows that a • b is numericalIy equal to the
1 '
a
"
length of b times the projection of a on b, and
also is equal to the length of a times the pro-
Figure 6.4 jection of b on a. If (J is obtuse, the projec-
tions considered are to be taken as negative.
In particular, if two vectors are perpendicular, their dot product is zero. From the
geometrical definition it is readily seen that the dot product is commutative,
a •b = b • a, ( 12)
and distributive,
a • (b + c) = a • b + a • c. ( 13)
For the unit vectors i, j, and k there folIows immediately
j . j = j . j = k · k = 1,
(14)
j • j = j • i = j • k = k •j = j • k = k • i = 0,
and hence, if
a=a) + ayj+akk, b=b)+byj+b,k, (15)
we ha ve, using Equation (13),
a· b = axb x + ayby + a,b,. (16)
The cosine of the angle B between 11 and b is given by Equations (J 1) and (J 6)
in the form
_ axb x + ayb y + a,b, , (17)
cos (J - A,/I a 2 + ay2 + a,2 A,/l b'x + b';: + b2,
x
or, if we denote the direction cosines of a and b by (1" m" n,) and (/" m 2 , n 2 ),
(18)
The abbreviation a' is sometimes used to indicate a • a. Then we also have
a'=a.a=Ja J2=a 2. (19)
The dot product is particularly useful in expressing a given vector v as a
linear combination of three mutually perpendicular unit vectors u" U 2 , and u J •
For ir we write
v = c,U, + C'U 2 + CJU J , (20)
we may successively take the dot product of u" u" and 1IJ into both sides of
(20) and so obtain
6.3. The Vector Product of Two Vector s 273
Thus Equation (20) becomes

v = (v • u,)u, + (v • u,)u, + (v • uJ)u J . (21 )
lt is seen that this representation could fail only if a component of v were
perpendicular to u" U 2 ' and U J • This condition, however, is clearly impossible
in space of three dimensions, since u" u" and U J are assumed to be mutually
perpendicular.
6.3. The Vector Product of Two Vectors. The second conventional type of
product of two vectors a and bis called the vector, cross, or outer product and
is written as a x b or [a b]. It is defined to be a vector having the properties
that (1) the length ofax b is the product of the lengths of a and b and the
nu merical value of the sine of the angle (J between the vectors and (2) the vector
a x bis perpendicular to the plane of a and b and is so oriented that a is rotated
into b about a x b by the right·hand rule, through not more than 180°.
According to the definition, there follows (see
Figure 6.5)
la x bl = ab I sin (JI. (22)
Thus the length ofax b is equal to twice the are a of
the triangle of which a and b form coterminous sides.
It is seen that the cross product is not commutative,
since, from the definition, b
b x a = -(a x b). (23)
However, the cross product is distributive, so that
a x (b + c) = a x b +a x c. (24)
This relation can be established by geometrical Figure 6.5
considerations. It is seen that the cross product of
two parallel vectors is zero.
For the unit vector S j, j, and k there follows, from the definition,
i x j = j X j = k x k = O,
(25)
ixj=-jxi=k, jxk=-kxj=i, kxi=-ixk=j.
These relations are easily remembered in terms of the cyclic arrangemeot
j k i j k
if we notice that the cross product of a vector ioto its neighbor is the following
vector when reading to the right and is the negative of the following vector
when reading to the left.
To calculate the cross product of two vectors a aod b in terms of their
components as given in (15), we make use of Equations (24) and (25) and
obtain
274 Vector AnalySis
This result can be written conveniently as the determinant

j k
a x b = Gx a y Gz (27)
bx by b,
The usefulness of the cross product may be illustrated by two physical
applications. First, we consider a point P in a rigid body rotating with angular
velocity of magnitude w about a fixed axis. Let O be a poiot on the axis of
rotation and represent the angular velocity by a vector ro of length w, extend-
ing along the axis of rotation in the sense determined by the right-hand rule
(Figure 6.6) . Then , if the position vector from Oto P is written as r , the velocity
vector associated with the point P is given by the equation
"= ro x T, (28)
slnce the magnitude of this vector is wr I sin (J 1, where r I sin (J I is the distance
from the axis of rotation to the point P, and the direction of the vector is as
indicated in the figure .
As a second application, we consider the moment vector M at a point O,
associated with a force F acting at a point P (Figure 6 .7). If the vector from O
,"
v
p
o -
P'
to P is denoted by r, the moment of F about the point O is defined as the vector

M = r x F. (29)
We see that the vector 1\1 is perpendicular to the plane of OP and the vector F
and has as its magnitude
1M I = M = Frl sin (J 1,
where () is the angle between r and F, and hence r I sin () I is the perpendicular
distance from O to the line of action of F. We may thus speak of M as the
scalar moment of the force about the axis ofM. Geometrically, Mis seen to be
twice the area of the triangle determioed by r and F. Further, the scalar moment
MOA of the force about any axis OA through O is seen to be twice the area of the
6A. l" luJliple Products 275
projection of this triangle on ¡he plane through O perpendicular to OA. But

since ¡he angle be¡ween ¡he lwo planes considered is the same as the angle
be¡ween the corresponding axes, ¡he projected area must be equal to the scalar
projec¡ion of M on OA. Thal is. lh e scalar momenl 01 F abolll any axis through
O is numerically equa/ 10 the /englh of the projeclion of M on that axis.
6.4. Multiple Produets. Three type of products involving three vectors are
of importance, namely, those of the respective forms (a. b)e, (a x b) • e,
and (a x b) x e. The first type ,
(a • b)e,
is merely the product of the scalar a • b and the vector e.
The second type,
(axb) .e,
which is called the triple sea/ar product, is seen to be the dot product of ¡he
vector a x b and the vector e, and hence is a scalar quantity. The value of this
product is .given by the produc¡ of the length ofax b and the projection of e
on a x b (Figure 6.8). But since a x b is a vector
perpendicular to ¡he plane of a and b, having a aXb
length numerically equal to the area of the paral-
lelogram of which a and b form coterminous si des, \'
\ "
and since the projection of e on a x b is the altitude \ "
of the parallelepiped with a, b, and e as coterminous \ '
\
edges, it follows ¡hat (a x b) • e is Ilumerical/y equal
to the volllme of this parallelepiped. Alternatively,
we see that (a x b) • e is numerically equal to six
times the volume of the tetrahedron determined by a
a, b, and e as edges. The sign of the product depends
upon the re la tive orientation of the three vectors, Figure 6.8
and is positive if and only if a, b, and e form a
right-handed system, in the sense that e and a x b lie on the same side of
the plane determined by a and b. From these facts it follows easily that
(a x b) • e = (b x e) • a = (e x a) • b (30)
and that the other products
(a x e) • b = (b x a) • e = (e x b) • a (31)
have the opposite algebraic sign. We see that the triple scalar product is I/ot
changed by a cyc/ic permutation of the three elements. Since also, from Equations
(30) and (12),
(a x b) • e = (b x e) • a = a • (b x e),
it follows that the dOl and cross can be interchanged in a triple scalar prOd/lCI.
For thi s reason, the notation (a b e) is frequently used to indicate the common
value of the products listed in Equation (30) .
276 Vector Analysis
ff we write
a = a) -!- ayj + a,k, b = b) + byj + b,k, e = c) + cy.i + c,k,
we find that
i j k
(a b e) = a • (b x e) = a· b x by b,
= axCbyc, - b,cy) + aib,c, - bxc,) + az(bxcy - byc x ),

or, in determinant form,
(32)
From Equation (32) we see tbat if two vectors in the product (a b e) are parallel,
rhen the product vanishes, since in this case the corresponding elements in two
rows of the determinant (32) are proportional.
The third type of triple product,
(a x b) x c,
is clearly a vector. Since it is perpendicular to a x b, which is itself perpen-
dicular to the plane of a and b, and is also perpendicular to e, it follows that
(a x b) x e is a vector which is in the plane o/ a and b and perpendicular lo c.
Thus this product must be expressible as a linear combination of a and b,
(a x b) x c = ma + nb.
In order to determine the scalars m and n, we first form the dot product of c
into the equal members of this relation and deduce that
ma • c + nb • e = O.
If we then write n = la • e, there follows also m = -lb. c, and hence we
now know that
(a x b) x e = l[(a • e)b - (b • e)a], (33)
for sorne scalar A.. Substitution of the relations a = a ,i + a j + a,k,
2 and so
forth, finally yields the determination
1 = 1,
(a x b) x c = (a • e)b - (b • e)a. (34a)
(A somewhat less tedious determination of 1 is indicated in Problem 22.) In a
similar way, the vector a x (b x e) is seen to be expressible as a linear com-
bination of b and e, and the identity
a x (b x e) = (a • e)b - (b • a)e (34b)
can be deduced from (34a).
6.5. Differentiation of Vectors
277
The two formulas (34a, b) are easily rememberect if ít is noticed that in

each expansion the middle factor on the left is multiplied by the dot product of
the other two factors , whereas in the term with the negative sign the o/her factor
in parenlheses on the left is multipliect by the dot product of the remaining
fac tors.
It should be noticed that such combinations as ab, (a x b)e, and a x b x e
a re here left undefined.
Vector products involving more than three vectors are readily evaluated in
terms of the products considered aboye. For example, the product
(a x b) • (e x d)
can be considered, say, as the triple scalar product of a, b, and (e x d) and hence,
if we write temporarily u = e x d, there follows
a x b··. u = a •b x n
= a • [b x (e x d)]
= a • [(b. d)e - (b • e)d],
and hence
(a x b) • (e x d) = (a • e)(b • d) - (a • d)(b • e). (35)
In particular, if we take
e = a, d = b,
(a x b) • (a x b) = (a • a)(b • b) - (a • b)2. (36)
This relationship is known as the iden/i/y of Lagrange. The truth of Equation
(36) fo[[ows aIso from the fael Ihal it ean be written in the form
la x bl 2 = a 2b 2 - a 2b 2 cos 2 e = (ab sin e)2,
which result is a eonsequenee of the definition ofax b.
6.5. Differentiation of Veetors. If the definitíon of a vector quantity v

involves a parameter /, the derivative ofthe vectorv(t) with respect to lis defined
as the limít
dV(I) = liro v(1 + ~I) - V(I) , (37)
dt ",,-o ~I
when that limit exists.

From this definition it foIlows that the derivative of the product of a scalar
s(t) and a vector v(t) is given by the familiar product law
d
-sv dv
= s -.L-v
ds (38)
di di ' dt .
Hence, if a vector 1S expressed in terms of its components along the fixed
coordinate axes,
v = f(t)i + g(t)j + h(l)k,
278 Vector Analysis
there follow s
dv = df i + dg. + dh k (39)
di di di] di'
since i , j, and k are constant vectors.
It follows also, from the definition, that the derivative of a product in-
volving two or more vectors is defined as in the corresponding scaJar case if
the order of the factors is retained. Thus, for example, we obtain the formulas
!{(a • b) = a • db + da • b
di di dI '
!{(a x b)
dI
= a x db
dt
+ da x b
dI '
!{(a • b x c) = da • b x e
dt dt
+ a • db x e
dt
+ a • b x dc.
dt
In the first case the order of factors in the separa te terms is irrelevant. This is,
however, not true in the second and third cases.
The derivative of a vector of constant length, but changing direction, is
perpendicular to the vector. This may be seen by noticing that if a has constant
length there follows
!{(a. a) = !{a 2 = O
dt dt
and also
d da
dt (a • a) = 2a • dt·
These results are compatible only if either da/ dt is zero or da/ dt Ís perpendicular
to 8.
6.6. Geometry of a Space Curve. Tbe equations

x = x(t), y = y(t), z = z(t) (40)
define a curve in space as the parameter t varíes over a specified range. We
denote by r the position vector from the origin O to the poínt P(x, y, z), corre-
sponding to a specífied value of t,
r = xi + yj + zk. (41)
If t is increased by !1t, and x,y, z increase accordingly, the vector from O to the
new point Q is given by
r + !1r = (x + !1x)i + (y + !1y)j + (z + !1z)k
and hence there foIlows
M = (!1x)i + (!1y)j + (!1z)k. (42)
This vector 1s c1early the vector PQ, and hence is of length equal to the chord
PQ. (See Figure 6.9.) Tf both sides of Equation (42) are divided by the íncrement
6.6. Geometry of a Space Curve 279
!:if, there follows z

!:ie _ !:ix ¡ + !:iy. + !:iz k
!:it - !:it !!.t J !:il'
If s represents arc length along the curve, we
can artificially rewrite Ihis equation in the
forro
(43) Q=p(t+At)
r+Ar
y
The vector !:ir/!:is in parenlheses is in the direc-
x
tion PQ if s increases with t, and is in the
opposite direction otherwise; it has a length Figure 6.9
equal lo the ratio of chord length to arc
length. Tbus, as !:it --+ 0, we have in the liroit
dr = (dx¡ + dy.J + dz k ) ds, (44)

dt ds ds ds dt
assurning that the derivatives exist, where the expression in parentheses is
clearlya unit vector tangent to the space curve at the point P. Since this expres-
sion is equivalent to dr/ds, we have the result that Ihe derivative of a position
vector to a space curve, with respect to arc length along Ihe curve, is a unit vector
tangent to the curve, pointing in the direction of increasing arc length. We denote
this unit tangent vector by u,
u = de _ dx¡ + dy.J + dZ k (45)
ds - ds ds ds'
Since u is a unit vector, there follows
(¿:r + (1sr + (~~r = 1.

This equation can be written in the differential forro
ds = --/dx 2 + dy2 + dz 2 = j (~~r + (1rr + (::r dt. (46)
giving the eleroent of arc length for a space curve.

If t represents time, the vector de/dt is the velocity veclor associated with a
point rnoving with speed d~/dt along the curve. Thus Equation (44) then becornes
ds
v = u - = vu (47)
dt '
where v = ds/dt.
If Equation (47) is differentiated with respect to 1, the acceleration vector a
is obtained in the forrn
dv du dv
a = -u
dt
+ v-
dt
= -u
di
+ v 2du
-'
ds
(48)
280 Vector Analysis
Since U is a unit vector, the derivative du/ds is perpendicular to the tangent vector.
Making use of Equation (45), we obtain
du _ d 2 e _ d 2 x¡ + d 2 y. +d 2
z
ds - ds 2 - ds 2 ds 2J ds 2 k '
The length of this vector is called the curvature of the curve and is a meaSUre
of the rate at which the tangent vector changes its direction with distance along
the curve. The reciprocal of this value is the radius of curvature, denoted by p.
Hence we may write
du 1
=-n. (49)
ds p
where o is a unit vector, perpendicular to the tangent vector u at P, and known
as the principal normall'ector.t There follows also
_1
P
= J(dds2X)2
2 + (d 2y
ds 2
)2
'
(d 2
Z)2.
ds 2
I (50)
By introducing Equation (49) into (48), we may express the acceleration

vector in the form
v2 dv
a = - o + -u. ( 51)
p dt
,.
Thus, if a particle moves along a space curve with speed v, its acceleration can
be resolved into a component dv/dt along the curve and a component v 2 / p along
the principal normal to the curve.
In addition to the two perpendicular vectors u and n at a point P of a curve,
a third vector, known as the binormal vector, is defined by the equation
., b = u x n, (52)
and hence is a unit vector perpendicular to both u and o. Thus, with each point
P moving along a space curve, we may associate a moving and rotating triad
of mutually orthogonal unit vectors. For a plane curve, the tangent vector u
and the principal normal vector o lie in the plane of the curve, whereas the
binormal vector is a constant unit vector perpendicular to tbat planeo
If we differentiate (52) and make use of (49), there follows
db du I do do
-=-xn,ux-=ux-·
ds ds ds ds
Thus db/ds is perpendicular to u and, since bis a unit vector, db/ds is also per-
pendicular to b. Hence db/ds
is a scalar multiple of n and can be written in the
form
db 1
ds
- --o.
r
(53)
tEquation (49) states that n aod dulds point either io the same direction or in opposite direc-
tions, depending upon the sigll of p. We here define p to be always positive, so that o and
dulds always poiot in the same direction. However, other conventions also appear in the litera-
ture. For a straight line, IIp = O and n can be taken to be any unit vector normal to U.
6.7. The Gradient Vector 281
The sea lar I/r: is called the lorsion of the curve, the ncgative sign having been
iotroduced so that the torsion is positive when the vector triad rotates in a right-
handed sense about the tangent as it progresses along the curve. The length Ir: I
is called the radius of lorsion. A curve whose torsion is not idcntically zero, and
which hence does not líe in aplane, is often called a Iwisled curve.
Equations (49) and (53) give the derivatives of u and b with respect to arc
!englh. To calcula te dn/ ds, we write o = b x u and differentiate, making use of
(49) and (53), lo obtain
~ db ~
ds = ds x u -j- b x ds = .- r1 (n x u) _1 (b x
P
)
D,
or dn = _1 b _ _1 u. (54)
ds r: p
Equations (49), (53), and (54) are known as Frenel's formulas.
If we form the dot product of b with the two sides of (54) and use (52) and
(49), we obtain
2
_1_ = b • do = u X o • dn = pZ (u x du • d u) . (55a)
r: ds ds ds ds 2
This result can be written in the determinant form
dx dy dz
ds ds ds
2 d 2y d2z ,
_1 = pZ d x2 (55b)
r: ds ds 2 ds 2
d'x d'y d 3z
ds' ds' ds'
where p is defined by Equation (50).
Formulas express ing p and ... in terms of a general parameter 1, rather than
are length s, are deduced in Problems 36 to 39.
The plane determined at a point Po on a curve by u and n is called the
osculaling plane, that deterrnined by n and b the normal plane, and that deter-
mined by u and b the reclifying plane. Hence, if ro is the position vector to Po
and r = xi + yj + zk , the equations of these planes are
b o • (r - ro) = O, DO • (r - ro) = 0,
respectively, where b o , u o , and Do are evaluated at Po'
6.7. The Gradient Vector. If rp is a scalar function of the coordinates

.\", y, and z, the values of the partial derivatives arp/ax, arp/ay, and arp/az at a
point represent the rates of change of rp with respect to distance in the x, y, and
z directions, respectively, at that point. If we consider the vector
V = i arp + j arp + k arp , (56)

ax ay az
282 Vector Analysis
we see that the scalar components of this vector in the x, y, and z directions at
any point Pare then exactly the respective rates at which rp is changing with
respect to distance in those directions at P. To determine the component of V
in any direction at P, we consider the position vector r from the origin to P and
indicate a differential displacement from P in any chosen direction by dr,
dr = i dx + j dy + k dz. (57)
The sea lar component of V in this direction is then obtained as the dot product
of V and the unit vector u = drjds, where ds = I dr 1, and hence has the value
V • u = ~ dx + arp dy + arp dz = drp . (58)
8x ds ay ds az ds ds
Thus we see that the component of V in any direction is the rate of change
of rp with respect to distance in that direction, so that, in fact, V is a vector
function which is associated with thescalar function rp in a way which is inde-
pendent of the coordinate system employed for their specification. It is called the
gradient of rp, and accordingly we may write
grad rp = i arp
ax
+ jarp
ay
+ k arp
az
(59)
'. when we employ a rectangular coordinate system. In the general case, if r

represents the position vector to a point P and if dr represents differential
displacement of length ds from P, Equation (58) provides the basic definition
," dr . grad rp = drp (60a)
., ds ds
or, in terms of differentials,
dr . grad rp = drp. (60b)
'"
Since the component of the vector grad rp in any direction is the derivative of
I rp in that direction, it follows that grad rp must point in the direction in which the
derivative of rp is numerically greatest, and must have a length numericallyequal
to that maximum derivative. In particular, grad rp must have no component in
the directions in which the derivative of rp is zero, and hence must be perpendicu-
lar to the surfaces rp(x, y, z) = constant.
Example. If the temperature at any point P(x, y, z) is given by

T = C(X2 + y2 + Z2),
the gradient vector has the form
grad T = 2C(xi + yj + zk).
This vector is clearly a radial vector pointed outward from the origin, if C > 0, and
hence is perpendicular to the spherical equithermal surfaces
X2 + y2 + Z2 = constan!.
The magnitude of grad T is given by
Igrad TI = 2Cr,
6.8. The Vector Operator V 283
where r --:::: vi X2 T y2 + 2 2, and, since T ~-= Cr 2, we have
I oorad TI = dT.
dr
This result is, of eourse, in aeeordanee with the faet that the preseribed temperature
changes most rapidly in ¡he r direction. The rate of ehange of the temperature at P
with respect to distance in a direetion specified by the direction cosines (l, m, n) is given
by
(I¡ + mj + nk) . grad T ~ 2C(lx + my + nz).
•
6.8. The Vector Operator V. lt is conventional to write
grad rp - V rp, (61)
where the symbol V, called de!, represents a vector operator which, accordingly,
is of the form
V = ¡...eL
ax
+ j...eL
ay
+ k...eL
az
(62)
in rectangular coordinates.t That is, we adopt the convention
Vrp = (¡...eL
ax
+ j...eL
ay
+ k~)
az
rp = ¡arp
ax
+ jarp
ay
+ k arp
az
, (63)
where rp is any differentiable scalar function of x, y, and z. In this notation,

Equations (60) beco me
dr drp
_·Vrp=-, (64)
dr • Vrp = drp,
ds ds
independently of the coordinate system.
If F is a vector function of x, y, and z, we may define the dot and cross
products of the operator V into F. These products are called the divergence
and cur! of F, respectively, and are of the form
V • F = ¡ • aF
ax
+j • aF
ay
+k • aF = div F,
az
(65)
V x F = i x ~~ + j x ~~ + k x ~~ = curl F. (66)
If F = F,¡ + Fyj + Fzk, we then have

V
•
F -
-
d'
LV
F _ aFx , aFy
- ax T ay
+ aF=
az (67)
j k
a a a
and V x F = curl F = (68a)
ax ay az
Fx Fy F,
tExpressions for the operator V, and for the related quantities introdueed in this section, are
obtained in terms of other eoordinate systems in Seetion 6.17.
284 Vector Analysis
or
vX F = i (aF z _ aFy ) + j (aF x _ aFz ) + k (aF y _ aFx ) • (68b)
ay az az ax ax ay
Just as the product ah has not been assigned a meaning, the combination VF
is here left undefined.
Tbe total differential of a vector function F is given in rectangular coordi-
nates by
aF aF aF
dF = ax dx + ay dy + az dz,
or, in operational form,
dF =
aa
( dx ax dy ay + + dz aza) F.
Since the operator is the dot product of tbe differential vector dr = i dx + j dy
+ k dz and the operator V, we may write tbis relation in the invariant form
dF = (dr. V)F. (69)
The derivative of F in the direction of dr is then
dF = (dr. V)F. (70)

"
ds ds
More generally, if u is any unit vector, then the derivative 01 F in the direction
01 u is given by
,"' dF (71)
.. 1
ds = (u. V)F.
Since VF is undefined, the parentheses in Equation (71) are generally

omitted and we foIlow the convention
v • VF = (v • V)F, (72)
where, if vis any vector with rectangular components Vx> v y , and v" the product
v • V is tbe scalar operator
(73)
6.9. Differentiation Formulas. The following identities are of frequent use:

V • I¡:>U = I¡:>V • u + u • VI¡:>, (74a)
V x I¡:>U = I¡:>V x u + VI¡:> x u, (74b)
V • u x v = v • V x u - u • V x v, (74c)
V x (u x v) = v • Vu - u • Vv + u(V • v) - v(V • u), (74d)
V(u • v) = u • Vv + v • Vu + u x (V x v) + v x (V x u), (74e)
V x (VI¡:» = curl grad I¡:> = O, (74f)
V • (V x u) = div curl u = O, (74g)
6.9. Differentiation Formulas 285
v X (V X u) = curl curl U = V(V • U) - V • Vu

= grad div U - V'u, (74h)
v . (Vq¡l X Vq¡z) ~ O. (74i)
In these formulas ti and vare arbitrary vectors and 91,91 l' and q¡z arbitrary
scalars for which the indicated derivatives exist. In formulas (74f) and (74g),
we assume in addítiol1 that 91 and ti have contínuous partíal derívatives of the
second order.
These identities can al! be verified by direct expansion in terms (say) of
components along i, j, and k. In sorne cases, however, the proof can be con-
siderably shortened by making use of the fact that the operator V may be sub-
stítuted for a vector ín any vector identity provided that it operates on the same
factors in al! terms.
It should be noticed that the operator V is a distributive operator, in the
sense that the equations
V • (u+ v) = (V • u) + (V • v),
V x (u + v) = (V x u) + (V x v)
are true.
Also, direct expansion shows that the identity
(u • V)q¡ = u • (Vq¡) (75)
is valido For this reason the parentheses are conventionally omitted in this case
and the product is written in the form U • Vq¡. Further, as has been noted, no
parentheses are needed in a product such as u· Vv, since Vv is undefined and
hence this notation can only mean (u • V)v.
In general, we adopt the convention that in a multiple product V operates
on alI terms to its right which are not specificalIy excluded from its inftuence by
the use of parentheses, brackets, or braces. If this is not convenient, we wilI
here indicate a term which is to be treated as a constant in the differentiation,
and hence is l10t to be operated on by V, by underlining that termo Thus we
use the notations V • (q¡~) and V X (q¡~) to indicate that u is to be treated as a
constant vector in the differentiation implied in V, so that V operates only on
91. Jt is readily verified that we then have
V • (q¡~) = u • Vq¡, V x (q¡~) = - u x Vq¡. (76)
From the nature of V we see that
V • (q¡u) = V • (~u)+ V • (q¡!l),
V x (q¡u) = V x (~u) + V x (q¡ll).
That is, the derivative of the product is the sum of the two derivatives obtained
by holding one of the factors constant and allowing the other to be operated
on by V. These considerations lead directly to formulas (74a) and (74b).
286 Vector Analysis
Formula (74c) is obtained in a similar way if we first notice that

V • u x c = -V· c x u = c • V x u
when c is a constant vector.
We see that if the product
V x (u x v)
isformally expanded by substituting V, u, and v for a, b, and c in Equation (34),
the result
(V • v)u - (u • V)v
is not necessarijy equivalent to the given product, since there V operates on
both u and v, whereas in the two terms of the formal expansion, V operates
ooly on v. However, if u is treated as a constant vector in the given product,
the expansion is valido That is, the expansion repre sents V x (!! x v). By adding
to this expansion the corresponding expression for V x (u x ~), we obtain
formula (74d).
Formula (74e) may be established by an indirect method in which we first
obtain the results
u x (V x v) = Ve!! • v) - u • Vv ,
v X (V x u) = V(u • y) - v • Vu,
and tben form tbe sum of these equations.
~ .. Equations (74f) and (74g) are easily established by direct calculation. Thus,
.. ' to establish (74f), we write
j k
, a a a _ . ( a2rp 2p 2p )
V x Vrp - aY: ay az - I ay az - aza2pay )+.J ( aza ax a
- ax az
arp arp arp 2rp
ax ay az +k (axa ay a2rp ) .
- ayax
If the order of differentiation is irnmaterial, all componeots on the right vaoish.
This situatioo exists, in particular, if the second crossed partíal derivatives of rp
are continuous. Similarly, Equation (74g) is readily shown to be valid if the
correspooding derivatives of the compooents of u are continuous.
Equation (74h) is obtained by replacing a, b, e, by V, V , u in Equation (34).
The scalar operator
,= a2 + -a2 + -a2
V- V •V =
ax2 ayZ az 2
- (77)
is of frequent occurrence, and is known as the Lap/acian operator. Equation

(74i) is established by making use of (74c) and (74f).
If r is the position vector, so that
r = xi + yj + zk
6.10. Une Integrals 287
in rectangular coordinates, we find by direct calculation that

v •r = 3, (78a)
V x r = O, (78b)
and a150 that
u • Vr = u, (79)
where u is any vector.
6.10. Line Iotegrals. Let F represent a vector function of pOSltlon in a

space of three dimensions and let e represent a space curve, each point P of
which is specified by its position vector r frorn an origin O to the point P. With
each point P of the curve we associate a differential distance vector dr. Then,
since
dr
dr = ds ds = u ds, (80)
where u is a unit tangent vector to e at P, the differential vector dr has as its

length the differential arc length ds along e at P and as its direction the direc-
tion of the curve at P. The scalar differential F • dr = (F • u) ds is then numeri-
callyequal to the product of the component of F in the direction of e at P and
the differentiallength ds. Finally, the integral
fe F • dr fe (F. u) ds,
= (81)
taken along the curve between two specified points Po and PI' is known as the
line integral of F along C.
In particular, if F representsforce acting on a partic1e, the line integral (81)
c1early represents the work done by the force in moving the partic1e along e
from Po to PI'
We assume bere, and in the sequel, that any curve e along whicb a line
integral is to be evaluated is made up of a finite number of arcs, along each of
which the tangent vector u not only exists, but also varies continuously witb s,
so that the integrand F • u is continuous when F is continuous. Sucb a curve is
said to be piecewise smooth. In particular, when u is continuous over al! of e,
we say that e is a smooth curve. For sorne purposes, we will require also that a
curve not intersect itself, that is, that no point be encountered twice as the curve
is traced out. Such a curve is called a simple curve.
If we write
F = P(x, y, z)i + Q(x, y, z)j + R(x, y, z)k,
(82)
r = xi + yj + zk,
tbe tine integral (81) takes the form
fe F • dr fe (P dx + Q dy + R dz).
= (83)
On each arc of the curve e, the variables x, y, and z, as well as the correspond-
288 Vector Analysis
ing differenrials, are to be expressed in terms of an appropriate single variable,

and the limits of the integral along that are must be determined accordingly.
It should be pointed out that the line integral fe F • dr, taken along the
portion of a curve between two points A and B, may be defined more funda-
mentally as the limit of a sum of the form
i: Fk' (~r)k
k=O
= Fa' (~r)a + F¡· (~r)¡ + ... + Fn' (~h,
where Fa, F¡, ... , Fn are the values of F at A and at n points PI' P 2 , • . . , Pn,
arbitrarily chosen along the arc of e between A and B, and where (~r)a = AP¡,
(~-;\ = pJ\, ... , C~;)n = PnB. The limit is taken as n ~ 00 in such a way
that all the chords tend to zero. This definition is analogous to the definition
of the ordinary integral. When F is continuous and e is piecewise smooth, the
integral so defined can be evaluated by the method described aboye.
Example. As an illustration, we take

F = yzi + xyj + xzk
and fírst ca1culate the line integral of F from (0, 0, O) to (1, 1, 1) along the path e con-
sisting of the curve x = y2, Z = °
in the xy plane from (0,0, O) to (1, 1, O) and the
line x = 1, Y = 1 perpendicular to the xy plane from (1, 1, O) to (1, 1, 1). In the first
,.
part of the parh, e¡, we have
x = y2, Z = O; dx = 2ydy, dz = 0,
aod henee
F . dr = yz dx + xy dy + xz dz = y3 dy.
Thus there follows
r F.
JCI dr = fo
¡ y 3 dy = :l:-
lo the second part of the path, e2, we have
x=l, y=l; dx = dy = 0,
and henee
F· dr = z dz.
Thus there follows
r F.
JC2 dr = f[ o
Z dz = ·h
and, fínally,
Se F . dr = í + 1- = ,t.
If we integrate instead along the path e' eonsisting of the straight Ene x = y = z,
directly from (0,0, O) ro (1, 1, 1), we have, on e'.
x = z, y = z; dx = dz, dy = dz
aod henee
r
Jc'
F. dr = r
Jo
¡ 3z 2 dz = 1.
•
6.10. Line IOlegrals 289
In this e'\ample the value of the line integral between the given points
depends upon the palh chosen. However, in certain cases this is not so. For if
the e.xpression P dx + Q dr - R d:: is the differential of a function rp(x, y, z),
Pdx -;- Qdy --;-- Rdz = drp. (84)
Ihen the integral (83) becomes merely
(85)
and it s value accordingly is Ihe change in rp a/ong fh e curve C. Thus the line
integral depends only on the location of the end points Po and PI if Ihefunclion
rp is sing/e-I'o/Lled, so that the value approached by rp at a point in space does not
depend upon the manner of approach.
Since the differential of rp is al so given by
drp = arp dx
ax
+ arp dy
ay
+ arp dz,
az
(86)
there follows, by comparing (84) and (86) and noticing that x, y, and z are
independent variables,
arp arp
ax = P, ay = Q. (87)
By appropriately differentiating the first two equations, we find also that

a rp
2 ap
ayax=ay'
and hence. if the derivatives involved are continuous, we cOnclude that P and
Q mus[ satisfy [he condition
(88a)
In a similar way we find that the two conditions
(88b)
(88c)
must also be satisfied.

Hence, if a function rp exists so that Equation (84) is true , that is, if F • dr =
Pe/x + Q dy + R d:: is an e,\act differential, the functions P, Q, and R must
satisfy (88a, b, c). Converse /y, when P, Q, and R and their first partial deriva-
tives are continuous in a region <R, it can be shown (see Problem 63) that, if
these conditions are satisfied, a function rp exists for which (84) is true in that
reglOn .
290 Vector Analysis
W'e notice next that since
(aR
ay _ aaz ) + j (ap ax + k (aax _ ap),
az _ aR)
v x F = i Q Q
ay (89)
the satisfaction of (88a, b, c) is equivalent to the vanishing of V x F. Thus we

conclude that, if F is continuously differentiable in a region CR and if V x F = O
everywhere in CR, then a scalar function rp exists such that drp = F • dr. Further,
if e is any curve Iying in CR and joining the points Po and PI' then
fe F • dr fe= drp = f).crp,
where f).crp is the change in rp corresponding to a transition along e from Po to

PI' However, if rp is not single-valued in CR, this change still may depend upon C.
To exclude the possibility that rp be multiple-valued, we henceforth require
that the region CR ha ve the property that an arbitrary closed curve Iying in CR
can be shrunk continuously to a point in CR without passing outside of CR. Such a
regio n is called a simply connec/ed region or, more briefly, a simple region.
Thus the plane annular region included between two concentric circles is not
a simple region, since a curve which surrounds the inner circle and lies in the
annulus cannot be shrunk to a point without passing outside the annulus, and
the same conclusion can be drawn for the region between two coaxial cy/inders
in space. The region between two concen/ric spheres is simple; the interior of a
/orus is no!.
To show rigorously the relationship between simplicity of CR and single-
valuedness of rp in CR is beyond the scope of this work. (See also Section 6.16.)
It may be remarked, however, that the proof depends upon the fact that by
restricting CR to be a simple region we ensure that any closed curve e in CR is
the complete boundary of sorne open surface S in CR, to which S/okes's th?or2m
(Section 6.16) can be applied.
For any curve e joining the points Po and PI and Iying in a simple region CR
throughout which F is continuously differentiable and also V x F = 0, we
may write
f F • dr
e
=
SP' F
Pu
• dr = SP' drp
Po
= rp(P I ) - rp(P o), (90)
where rp is single-valued in CR. That is, in such a case /he fine integrallrom Po
to PI is independen/ 01 /he pa/h, so long as that path remains in CR. In particular,
for a closed path Iying in CR the initial and terminal points coincide and the line
integral vanishes. We may denote the line integral once around a closed curve
by the symbol § and write
(91 )
in this situation.
If the conditions stated are not satisfied, the integral around e may or may
not vanish. The value of that integral is called the circula/ion of F around C.
6.11. Thc Potential Function 291
When a closed cur ve e lies o n a s urface S, one side of which is regarded a s

its posi tive side, it is conventional to define the positive diree tion on e ( that is ,
the direction of u = dr /ds) as the direction along which an observer, travelin g on
the positive side of S , would mo ve while keeping the area enclosed by e to hi s
lef!. In this connection, for a coordinate plane or surface (on which a coordinate
is co nstant) it is to be understood in what follows that the positi ve side is the one
from which the relevant coordinate increases , unle ss the contrary is stated.
6.11. The Potential Funetion. Since the existence of a single- valued func -
ti OD rp such tha t
drp = F • dr (92)
at all points of a simple regio n is guaranteed by the condition V x F = O, and
since Equation (64) gives also
drp = Vrp • dr,
there follows by su btraction
(F - Vrp) • dr = O.
Hence the vector F - Vrp must be perpendicular to dr; but since the direc tion
of dr is arbitrary, we conclude that F - Vrp must vanish,
F = Vrp. (93)
Thus, if V x F = O in a simple region, th en F is the gradient of a single-valued
sealar funetion rp in that region. Conversely, Equation (74f) states that if F = Vrp
and if rp has continuous second partial derivatives, then V x F = O.
The function rp defined by Equation (93) is known as the potential of F .
If F represents force, the negative of rp is called the potential energy associated
with F. When such a function rp exists, and is single-valued in a region, the
force F is said to be eonservative, since in this case the total work done in moving
a particle around a closed contour in that region is zero .
To illustrate tbe determination of the potential, we consider the force field
F = y2i + 2(x y + z)j + 2y k . (94)
lt is readily verified that V x F = O everywhere. Hence the force is eonservative
and the work done by F in moving a particle between two points is independent
of the path. Further, a sealar potential funetion rp exists su eh that
F • dr = drp (95)
and also
F = Vrp . (96)
To determine rp , we write Equation (95 ) in the form
drp = y2 dx + 2(xy + z) dy + 2y d z.
Comparison of thi s equation with (86) s how s that rp must satisfy tbe three
292 Vector Analysis
conditions
arp = y2 (97a)
ax '
~~ = 2(xy + z), (97b)
~~ = 2y. (97c)
Tf we integrate (97a), holding y and z constant, we obtain

rp = xyZ + I(y, z), (98)
where I(y, z) is an arbitrary function of y and z. In order that (98) satisfy (97b),
we must then have
2xy + al(y, z) = 2xy + 2z or al(y, z) _ 2
ay - z.
ay
Integration gives
I(y, z) = 2yz + g(z), (99)
where g(z) is an arbitrary function of z. Introducing (98) and (99) into (97c),
,. we then obtain the final condition
2y + d~~z) = 2y,
g(z) = e , (100)
where e is an arbitrary constant. Thus we have finally
rr rp = xyZ + 2yz + e, (101)
where the constant e can be chosen arbitrarily, so that the potential at a conve-
nient reference point is zero. (See also Problem 64.)
The surfaces rp = constant are called equipotential surlaees. We see that the
work done in moving a particle from a point on the surface rp = el to a point
on the surface rp = e z is merely e z - e,.
To illustrate the existence of unusual cases, we notice that if
F=
there then follows
F • dr = x dy - Y dx = d tan -11:'.. = de
x, + yZ x'
where e is angular displacement about the z axis. Thus it seems that we have
F= ve,
and hence that, from (74f), there follows
V x F = O.
6.11. The Polential Function 293
Hence it might appear that the circulation of F around any closed curve e
would be zero. However, we notice that F is discontinuous at points on the
z axis, in the sense that the magnitude of F,
1
IFI = ~. X 2 f y 2'
~ ~
becomes infinite as (x, y) - , (O, O). Thus V x F is nol defined along the z axis.
At al! other points, however, V x F is defined and is zero. Any closed curve
which does not surround the z axis can be included in a simple region where F
is continuously differentiable and V x F = O, and hence the circulation around
any such path must vanish. To verify this statemen t, we consider an arbitrary
closed path e beginning and ending at a point P. Since
t F • dr = fe de = !lee,
e
we see that if e does not enclose the z axis, the angle may alternately increase
and decrease along e until a maximum value is reached, after which eventuallye
returns to its original value at P and the total net increase in e is indeed zero.
That is, over such a closed path the circulation is zero.
However, if the path e surrounds the z axis once, and the path is described
in the positive (counterclockwise) direction, the angle e experiences a net
increase of 2:n: when the complete path is described, and the circulation is 2:n:.
Similarly, the circulation along a closed curve encircling the z axis k times in
the positive direction is 2k:n:.
It is important to notice here that e is, in fact, not a single-valued function,
since at any point P not on the z axis e has an infinite number of admissible
values, differing from each other by integral multiples of 2:n:. (On the z axis, {}
may have any value.) In a simple region <R not including any point on the z
axis (and hence also not including any closed curve which surrounds that axis),
we may define a single~valued interpretation of e such that the interpretation so
defined is continuous (and differentiable) along any smooth curve e in <R, so
that de exists everywhere on e, and so that fe de then is dependent only upon the
end points of C. If there were any closed curve in <R which surrounded the z
axis, such a definition no longer would be possible. It should be noted that the
shaded two~dimensional region indicated in Figure 6.10 is an example of a
permissible region, for present purposes.
If we dele te the regio n in the neighborhood of the z axis by cutting out an
infinitely extended right circular cylinder of arbitrarily smal! radius, with its
axis along the z axis, the remainder of space is a region <R' inside which V x F = O
at al! points. However, the regio n <R' clearly is not a simple region. Those
closed curves in <R' which can be shrunk to a point in <R' without passing outside
<R' are said to be reducible. Thus any closed curve in <R' which does not enclose
the z axis is reducible, and we see that in the present example the circulation
around any reducible closed curve in <R' is zero.
294 Vector Analysis
Figure 6.10
It can be shown, more generally, that if F is piecewise differentiable and

V x F = O in any regíon CR, whether simply coonected or not, the circulation
of F around any reducible closed curve e in that region is zero.t
6.12. Surface Integrals. A surface S in three dimensions is usually specified

by an equation of the form
g(x, y, z) = O, (102)
where restrictions may be imposed on x, y, and z. If we think of S as one of the
family of surfaces g(x, y, z) = c, as c takes on aH possible values, we see that at
any point where the vector Vg exists, it is normal to the particular surface of the
family which passes througb that point, and heoce, in particular, Vg is normal
to S at points of S. At such points Vg can then be expressed as the product of a
scalar and a l/nit vector n normal to S. We can thus write
Vg (103)
D = ±IVgl'
where the choice of the ambiguous sigo depends upon the desired orientation
of D.
For a closed surface, it is conventional to define the orientation of D at any
point as outward. In any case , we suppose that S actually possesses two sides,
tThis resul! can be established by applying Stokes's theorem (Section 6.16) to a surface S
which is Iraced out as the reducible curve e is shrunk 10 a point in <R, while remaining in <R.
6.12. Surface Inlegrals 295
which can be distinguished from each other, and that one side has been selected
as the positive side, from which n is to point.
We will suppose also ¡hat S is a so-called simple surlace, having the property
that it does not intersect itself, so that there cannot be more than one vector n
at a single point.
Further, we suppose here, and in the remainder of the text, that any surface
S over which an integration is to be effected can be subdivided, by smooth
curves, into a finite number of parts such that on each part the normal n exists
and also varies continuously with position. Such a surface is said to be piecewise
smoolh. When n is continuous over all of S, we say that S is a smooth surlace.
In the cases when S is described by an equation of form (102), the preceding
requirement is merely that ag/ax, ag/a y, and ag/az exist and be continuous
over S, or over each of a finite number of parts into which S is divided.
Thus, for the unit sphere .'(2 + y2 + Z2 = 1, we can write
g(x, y, z) = X2 + y2 + Z2 - 1,
and hence
Vg = 2xí + 2yj + 2zk.
It follows that at any point (x o , Yo' zo) on the sphere we have
n -_ ± xoí
/ +
2
yoj 2+ zok2 -- ±(Xol. + YoJ. + '
"O'
k)
-v Xo + yo + Zo
The positive sign must be chosen if n is to point outward from the surface of
the sphere.
A t a point P(x, y, z) on S we define a differential surlace area vector dcr
whose length is numerically equal to the element of surface area da associated
w.ith P, and whose direction coincides with the normal vector n at P. If the
element da is constructed by projecting a rectangular element dA = dx dy in
the xy plane vertically onto the surface S at P, it is seen that the projection of da
on the xy plane is numerically equal to the projection of n da on the vector
which is normal to the xy plane. (See Figure 6.11.) Hence we have
In. kl da = dxdy.
If the direction cosines of n are denoted by (cos ft, cos p, cos y), this equation
becomes
da = I sec y I dx dy (104)
and from Equation (103) we obtain
cos y - n • k - ± ag/az (105)

- - ~(ag/ax)2 + (ag¡ay)2 + (ag¡az)2
In particular, if the surface is given in the special form
z = I(x, y), (106)
296 Vector Analysis
y
077~71';y;
I I
I
/ ..
,J.L
dy
x
Figure 6.11
l' there follows g(x, y, z) = z - f(x, y), and Equation (lOS) becomes
1
( 107)
cos y = ± -J I + (af/ ax)' + (af/ ay)'
With the definition da = n da, ¡he integral
( 108)
carried out o ver a surface S, 1S called the surface integral of F over S.t We notice
that the element o f integration is the product of the component of F normal to
the surface at a point P and the scalar element of surface area as soc iated with P.
Example. We ea1culate the surfaee integral o f the vector funetion

F = xi + yj
over (he po rtion of the su rfaee o f the unit sphere
S: X2 + y' + z2 =
aboye the xy plane, z ~ O. The unit normal at a point of S is given by
n = xi + yj + zk .
Also, we obtain
cos y = n . k = z,
tSimilarJ y defined integrals of the more general form f f s h da (w here the sea lar funetion h
is not necessarily expressed as F . n) or of the vector fo rm f f s V da also may arise in practice.
6.13. Interpretation of Divergence. The Divergence Theorem 297
and hence Equation (104) gives

du = e/xdy
z
Thus the required surface integral is of the form
JI F . dfJ = fL (x' + yl) dXzdy .

where D denotes the projection of S on the xy plane, and hence D is the interior of the
unit circle X2 + y2 = 1 in the xy plane. In this integral, the variable z must be expressed
in terms of x and y by using the equation of S. Hence we obtain the form
JL F· dfJ = 4 f f"-'" ~1 x' (~/~ yl) dydx.

or, after changing for convenience to polar coordinates and evaluating .the result as
a repeated integral,
•
lt should be noticed that, if lines parallel to the z axis intersect the surface S
in more than one point, the surface must be considered in two or more parts,
the formulas resulting from the use of (104) being applied separately to the
individual parts. In such cases it may be more convenient to project tbe element
da onto the xz or yz plane, or to proceed by a more direct method.
6.13. Interpretation of Divergence. The Divergence Theorem. Let V represent

the velocity of flow of a fluid in three dimensions. and denote the rnass density
of the fluid at a point (x , y, z) by p(x, y, z). Then the vector Q = pV points
in the direction of flow and has a rnagnitude Q numerically equal to the rate of
flow of fluid ma ss through unít area perpendicular to the direction of flow. Tbe
differential rate of flow tbrough a directed elernent of surface area da = n da
is then given by Q • da = Q • n da, this quantity being positive if the projection
of Q on the vector n is positive. Jn particular, if da is an elernent of a dosed
surface, then Q • da is positive if the f10w is outward from the surface. We write
Q = Qxi + Qyj + Q,k. (109)
Now consider a small closed differential element of volurne cODsisting of a
rectangular parallelepiped with one vertex at P(x, y, z) and with edges dx, dy, dz
parallel to the coordinates axes. Then the left-haDd face is represented by the
differential surface vector - j dx dz and the differential rate of flow through
this face is given by
Q. (-jdxdz) = -Q,dxdz,
the negative sign indicating that, if Q, is positive, tbe flow through tbis face is
¡nto the volume elemenl. Similarly, the differential rate of flow through the
298 Vector Analysis
right-hand face is given by
QI . (j dx dz) = Qyl
y+dy y+ d y
dx dz = (Qy + aaQy
Y dY) dx dz
within infinitesimals of higher order. If the remaining four faces are treated
similarly, we find that the net differential rate of flow dF outward from the
volume element dr = dx dy dz is given by
dF = [(Qx + aa~ dx) - Qx] dy dz + [(Q, + aa~y dY) - Qy]dx dz
+ [(Qr + aa~rdz) - Qr] dXdy
- (aa~x + aa~ + aa~;) dx dy dz,
or dF = (V • Q) dr. (110)
Thus we may say that the divergence of Q at a point P represents the rate
of fluid flow, per unit volume, outward from a differential volume associated
with P, or that the divergen ce ofQ is the rate of de crease of mass per unit volume
in the neighborhood of the point. Thus, if no mass is added to or withdrawn
from the element dr, we have the relation
(111)
which can also be written in the form
\7 • (pV) + !Jj¡ = O
or, using Equation (74a),
pV • V = - (V . V P + !Jj¡) . (112)
For an incompressible fluid, p = constant and there follows

V •Q = pV • V = O. (113)
Thus the velocity of an incompressible fluid has zero dil'ergence.
It has been assumed here that no mass is introduced into the system or
taken from the system, that is, that there are no points in the element dr where
fluid is added to or withdrawn from the system. A vector V with nonzero
divergence can be considered as a velocity vector of an incompressibJe fluid in
a region only if such points are assumed to be present. We speak of points
where fluid is added to or taken from the system as sources and sinks, respec-
tively. Thus the presence of a source distribu tion is associated with a positive
divergence of V.
Consider now a c10sed region CR in three dimensions and suppose that in
each elementary volume dr of this region there are present sources through
6.13. lnlerprelalion of Divergence. The Divergence Theorem 299
which an amount dF = pV • V d1: of incompressible fluid is introduced into

the region per unit time. A sink is considered to be a negative source. Then the
net mass of fluid introduced into CR in unit time through sources and sinks is
given by the integral
(1 14)
where the integration is carried out over the volume of the region. If the total
masS in <R is conserved, this fluid clearly must escape from the regio n through
the surface S which bounds it. If we cons ider a vector surface element da = D da,
where n is the outward unit normal, the rate of mass flow outward through
the element da is given by pV • n da = pV • da. Thus the total rate of flow
outward through S is given by the surface integral of V over S,
p .§s V • da = p.§ s V • n da, (1 I 5)
where the symbol <tJi indicates that the integration is carried out over a closed
surface.
The statement that (114) and (115) must be equivalent,
SSSen V • V d1: = .§s V • o da, (116)
that is, the consequence of conservation of mass in <R, is known as the divergence
theorem (or as Gauss's theorem) . This theorem can be established without refer-
ence to the physical considerations used here, t and is true if V and its partial
derivatives are continuous in <R and on S and if S is simple and piecewise smooth.
The region <R need not be simply connected, so long as its complete boundary is
taken into account in the surface integral.
Jf we write
V = Pi + Qj + Rk, D = i cos rJ. + j cos P + k cos y,
Equation (116) takes the form
JJ1(~: + ~; + ~~) dxdydz
= .§s (Pcos rJ. + Q cos P+ R cos y) da. ( 117)
This theorem is useful in many applications. In particular, it serves the

purpose of avoiding the direct evaluation of a surface integral of a vector over
a closed surface S, by replacing that integral by a volume integral over the
interior <R, if the divergence of the vector is of a convenient form in <R. When the
original integration is over an open surface SI' one may first introduce an ad-
ditional surface S2 which, together with SI> forms a closed surface S bounding a
convenient regio n <R. The surface integral over SI then can be evaluated as a
tFor an indication of lhe analylical proof, see Problem 76.

300 Vector Analysis
yolume integral oyer CR reduced by the surface integral oyer S2' if the latter
computation is preferable to the former.
Example. In illustration, we consider the example treated in Section 6.12. There we

calculated directly ¡he surface integral
JLF. da,
where F = xi + yj
and S is the upper half of the unit sphere X2 y2 + + Z2
= 1. If we close the surface of
integration by adding the portion of the xy plane which spans the hemisphere, we
notice that the surface integral of Foyer the added surface is zero, since
F • n = F • (-k) = O
oyer this area. Thus the diyergence theorem states that we may ca\culate the required
surface integral of F by eyaluating
JJJ(J\ V· Fd,
throughout the interior of the hemisphere. Since V • F = 2, the result is merely twice
the yolume of the unit hemisphere, or 4n/3, as was obtained by direct integration .
•
The surface integral of a vector F over a closed surface S is sometimes
called the flux of F through S. Thus the divergence theorem states that the flux
of F through S is a measure of the divergence of F inside S. In particular, if F
is continuously diff'erentiable alld V • F = O in the region bounded by S, the flux
of F through S is zero,
ff s F • da = ff s F • u da = O.
Thus, remembering the result of Section 6.10, we see that, for a continu-
ously differentiable function F, the circulation f F • dr ofF around a closed curve
e is zero if e is in a simple region throughout which V x F = O, and also the flux
cf:ji F • da of F through a closed surface S is zero if V • F = O in the region bounded
by S.
We may expect that the circulation of F around e will be in sorne sense a
measure of the curl of F at points on a surface of 'lYhich e is a boundary, in
analogy with the relation (116) which connects flux through , a surface to diver-
gence inside the surface. A relationship of this sort, kno~n as Stokes's theorem, is
considered in Section 6.16.
The two-dimensional form of the divergence theorem is
JL V • V dA = fe V • uds, (118)
where D is a region in the xy plane, e is its complete boundary, and u is the unit
outward normal along e (see Problem 78). In particular, for a simply connected
6.14. Green's Theorem 301
region D \Ve may write

_ dy. dx.
O--I--J
ds ds ( 119)
along e and, wilh
v = Pi + Qj, ( 120)
(1 18) then takes the form
f L(~~ + ~;) dx dy = fe (P dy - Q dx) . (121)
6,14. Green's Theorem. If. in the divergence theorem (116), we write

V=q¡,Vq¡2'
where q¡, and q¡2 are scalar functions of position, we obtain the result
f f t V • q¡,Vq¡z dT: = ft o • q¡,Vq¡z da .

By making use of (74a), this result takes the form
ff t [q¡,V2q¡Z + (Vq¡,)· (Vq¡z)] dT: ft o • q¡,Vq¡ 2 da.

= ( 122)
This equation is known as thefirstform ofGreen's theorem. A more symmetrical

form is obtained if q¡1 and q¡z are interchanged in Equation (122) and the result-
ant equation is subtracted from (122) to give
(123)
This equation is known as the second form of Green's rheorem, or frequently

merely as Green's theorem, and is frequently useful in applications. In both
(122) and (123), the region Cll. is the region bounded by the c10sed surface S.
It is assumed that q¡ I and q¡z are twice cootinuously differentiable io eR.
Two special cases of these theorems are of particular interest. If we take
q¡1 = q¡z = q¡ in (122), there follows
1'1' t [q¡v zq¡ + (Vq¡)Z] dT: = 3t q¡o • Vq¡ da, (124)
where (Vq¡)2 = (Vq¡) • (Vq¡). We notice that the product D • Vq¡ is, according to
(64), the derivative of q¡ in the direction of o, that is, in the direction of the
outward normal to S at a point on S. Hence we may write
~~ = o • Vq¡ (125)
and speak of this quaotity as the normal derivative of q¡ at a point on S. Wilh
Ihis notation , Equation (124) becomes
f
J J<R [q¡V2q¡ + (Vq¡)Z] dT: = ft q¡~~ da, (126)
and Equations (122) and (123) can be rewritten in a similar way.
302 Vector Analysis
A second important special case of Green's theorem is obtained by taking

9, = 9 and 90 = l in (123), Since then V9 2 = O and V292 = O, there follows,
with the notation of (J 25),
¡JI
- <l1
V09 dr: = JI: ~ dlJ,
lf san
( 127)
The relations (126) and (J 27) will be useful in certain applications which fol-
low (Section 9.2). Their two-dimensional specializations are considered in
Problem 80.
6.15. Interpretation of Curio Laplace's Equation. Suppose that the motion

of a fluid is simply a rotation about a given axis fixed in space. Then we may
represent the angular velocity by a constant vector 0>, as in Section 6.3 , where
the length w is the scalar angular velocity and the direction of O> is along the
direction of the axis of rotation in accordance with the right-hand rule. If we
take the origin of a rectangular coordinate system on this axis and denote the
positio n vector to a point P(x, y, z) by r, then the results of Section 6.3 show
that the velocity of P is given by the vector
v = O> x r. (128)
.. We ha ve then, using Equation (74c),
V • V = V • O> x r = r • (V x 0» - O> • (V x r),
and hence, since O> is constant and V x r = O, from (78b), there follows
V • V = O. (129)
We also have, using Equation (74d) and again noticing that O> is constant,
V x V = V x (o> x r) = o>(V • r) - (o> • V)r,
or, by virtue of (78a) and (79),
V x V = 20>. (130)
Thus, if a fluid experiences apure rotation, the divergence 01 ils velocity vector
is zero, and the curl 01 the velocity vector is a constant vector equal to twice the
angular velocity vector.
More gene rally if V is the velocity vector of a fluid flow , and if V x V =
20>0 *' O at a point Po, then a component of V corresponds to a local tendency
of the fluid to rotate about the vector 0>0' with angular velocity \ 0>0\'
If V x V = O in a region, we say that the flow is irrotatiorml in that region.
The results of Section 6.10 state that the circulation around a closed curve in a
simple region where the f10w is inotational is zero. If the fluid is incompressible
and there is no distribution of sources or sinks in the region, we have also
V • V = O. Since the condition V x V = O implies the existence of a potential
9 such that
V = V9, (131)
6.16. Stokes 's Theorem 303
we see that if also V • V =c O there follows V • Vrp .~ V2rp = O. That is, in the
flow of an incompressible irrotational fluid without distributed sources or sinks,
the ve/ocit;: vector is the gradient of a potential rp which satisjies ¡he equation
a 2rp . a'rp azrp _
V'rp = O or ax2 T ay2 + az 2 - o. ( 132)
This equation is known as Laplace's equation.

We see, more generally, that, in any continuously differentiable vector field
F with zero divergence and curl in a simple region, the vector F is the gradient
of a solution of Laplace's equation. Solutions of this equation are called
harmonic functions, and their determination is studied in Chapter 9.
6.16. Stokes's Theorem. We next indicate that if 5 is a two-sided surface in

three dimensions having a closed curve e as its boundary, then the circulation
of a respectable vector V around e is equal to the flux of the curl of V over 5.
It is assumed throughout that 5 and e are simple and piecewise smooth.
The relation to be considered is thus of the form
f J> .(V x V) da t = V • dr, ( 133)
where ° is the unit vector normal to 5 on that side of 5 which is arbitrarily

taken as the positive side. The positive direction along e is then defined as the
direction along which an observer, traveling on the positive side of 5, would
proceed in keeping the enclosed area to his left. Equation (133) is known as
5tokes's theorem, and is true if 5 is contained in a simple region where V is
continuously differentiable. In the following considerations, however, we
assume also the existence of contínuous partíal derivatives of the second order.
Suppose that Equation (133) has been established for one surface 5, having
e as its boundary, and consider any other surface 52 having the same boundary.
If we denote the closed region included between 5, and 52 by <R, the divergence
theorem sta tes that
f L, o, • (V x V) da + f L, O2 • (V X V) da = f f f<lt V • V x Vd, = O,
(134)
if V is twice continuously differentiable in <R, according to Equation (74g).
In this equation the vectors o, and O 2 are headed outward from the enclosed
region <R. Hence if we choose a positive direction around the common boundary
e so that o, is on the positive si de of 5" and hence o, = ° on 51> it follows that
O2 is then on the negative side of 5" and hence o, = -o on 52· Thus (134)
becomes
ff S'I
o· (V x V) da = ff
S2
o· (V x V) da. (135)
Accordingly, if Equation (133) is true for one surface having e as its boundary,
304 Vector AnalYSis
it is also true for any other such surface, if both surfaces he in a simple region
where V is twice continuously differentiable.
We prove (133) in the s pecial case when
e is a plane curve by taking S as the plane
region D bounded by e in its planeo The
vector n normal to D is then a constant
vector. The proof is obtained most readily
by an indirect method in which we apply the
divergence theorem to a vector V x n, which
Do< is thus parallel to the plane of D, in a closed
three-dimensional regio n eR which is the
interior of a right cylinder of constant
Figure 6.12 height h having D as its lower base (Figure
6.12). We then have
(136)
where the surface integral is extended over the complete boundary S(R of eR,
and O(R is the outward unit normal vector on S(R. On the upper and lower faces of
S(R we have O(R = ±o, and hence the integrand in the surface integral vanishes.
On the lateral boundary of S(R we have n(R = u x n, where u is the unit tangent
vector to the curve e which bounds D, and hence here the integrand becomes
(u x o) • (V x n) = (V • u)(n • n) - (u • n)(n • V) = V • u,
by virtue of (35) and the fact that u • n = O. Also, since n is constant, we have
V • (V x n) = o • V x V, from (74c), and hence Equation (136) becomes
{' (f In n. V x V du) dh = S: (fe u

V • ds) dh.
Since this relationship is true for arbitrary values of h, we conclude that
f In n • V X V da = fe V • dr, ( 137)
as was to be shown.
To extend this result to the case of an arbitrary closed curve e in space, we
first approximate e by a space polygon en whose n sides P,P2 , P 2 P 3 , . . . , PnP,
are formed by joining n successive points p¡. P 2 , . • • , P n on e by straight
lines. A surface Sn having en as its boundary can be defined as the polyhedron
whose triangular faces are determined by the 11 sides of en and the n - 3
straight Unes joining p¡ to the points P 3 , . • • , P n - , . If we apply Equation (137)
to each of the triangular faces of Sn and add the results, we notice that the line
integrals along the tines P IP 3 , PIP., ... , P,Pn-, are taken twice in opposite
directions and hence cancel, leaving only the line integral around the polygon
en' Thus (137) is also true for Sn and its boundary en' As n becomes infinite in
such a way that all the chords PkP k + 1 approach zero, the polygon en ap-
proaches e, and Sn approaches a surface S for which the relation (133) is true.
6.16. Stokes's Theorem 305
From (135) we conclude finally that Stokes's theorem is valid for an arbitrary
piecewise smooth surface S with e as its boundary, if S is in a simple region
where V is twice continuously differentiable. If we write
V = Pi + Qj - Rk, n = i eos a + j eos jJ + k cos y,
rhis theorem takes rhe form
fe (P dx + Q dy -¡- R d7) = JJJ (~~ - ~~) cos a
-1- (ap _ aR)

az ax. cos jJ + (a Q
ax _
ay eos y] da.
ap) (138)
In addition ro serving as an important theoretical tool, Stokes's theorem can

be used to avoid the caIculation ofa surface integral by substitutinganequivalent
line integral. However, it is perhaps more frequently employed in the opposite
direction, in cases when the hne integral is not easily evaluated but the quantity
n • V X V is of relatively simple form on sorne open surface which spans C.
Example J. In order ro evaluare the integral
1 = fe [(e-x' - yz) dx + (e-Y' + xz + 2x) dy + e-·' dz]
when e is the circle

x = cose, y - sin e,
oriented in the direction of increasing e, we may note that
V x V = -xi -yj + (2 + 2z)k.
Hence, ir we take S to be the circular diskt bounded by e in the plane z = 2, so thar
n = k, there foIlows n • V X V = 60n S and accordingly
1 = 6 SSs da = 6n.
•
For aplane region D in the xy plane, rhere follows dz = 0, eos a = cos jJ = 0,
cos y = 1, da = dx dy, and hence Stokes's theorem reduces to the two-dimen-
sional form
L (Pdx + Q dy) = JI (~~ - ~~) dx dy. (139)
Equation (139) is also frequently referred ro as Green's rheorem, although we

here reserve this name for rhe theorem of Section 6.14. It may be noticed that in
faet the two-dimensional form (l39) of Stokes's theorem is equivalent to the
two-dimensional form (121) of the divergen~e Iheo/'"m, since (139) becomes (121)
with a notational change in which Pis replaced by - Q and Q by P in (139).
t A circular disk is a pla ne region bounded by a circle.

306 Vector Analysis
As a special case of ( 139) we take P = - y and Q x. The result,
±t(xdy - ydx) -~ A, ( 140)
IS useful in finding ¡he area enclosed by a simple plane curve e whose equalion
1<; given in paramelric formo
Example 2. For the ellipse (x2/a 2 ) + (y'/b 2) = 1, we can write

x = a cos 8, y=b sin8, (O < 8 <2n),
and Equation (140) gives for its area
A = ±So'" ab(cos 2 8 + sin 2 8)d8 = nabo

•
A useful interpretation of the curl vector, supplementing that suggested in
Section 6.15, can be obtained frorn Stokes's theorem (133) by taking e to be a
circ1e with center al a point P and S lo be the circular disk bounded by e, and
by then letting the radius of e tend to zero. Since at each stage of the limiting
process the left-hand member of (133) is equal to the product of the area of the
disk and the value of n • V x V at sorne interior point, we may deduce in the
limit that the component ofV x V in the dire ction of a vector n, at a point P, is
the cireulation per unit area ofV aroune! P, in a plane normal to n.
The rigorous proof of Stokes's theorem (133) in full generality involves sub-
tleties which are beyond the level of the present treatmeots. However, il is of
,. some importance here to notice that from that theorem one may deduce, in
particular, that ij'V is continllously differentiab/e in a simply connected region <R,
and ir v x V = o everywhere in <R, then
fe V • dr = O
jor any piecewise smooth closed Cllrve C in <R. This is a result made plausible in
Section 6.10 [Equalion (91)].
From that result , in lurn, one can reason in a direclion opposile lo Ihal
followed in Seclion 6.10 to deduce Ihal, lInder the conditions just stated, rhe line
integral fe V • dr between two points in <R is ¡ndependen! ojrhe path Cjoining those
poinrs, provided only that C is piecewise smooth and lies in <R, and it then jollows
a/so that V • dr = dlp, where Ip is single-valued in <R.
The specialization of these results lo ¡he case when <R is a two-dimensional
region, in the xy plane, and in which Stokes's theorem consequently reduces lo
(139), will be of particular significance in Section 10.5.
6.17. Orthogooal Curvilinear Coordina tes. Suppose ¡hat the rectangular

coordina tes x, y, z are expressed in terms of new coordinates u 1> u 2 , l/, by tbe
equations
x = X(II" u 2 , !l,), y = Y(!l¡, u 2, u,), Z = z(u" U2, u , ), (141)
and ¡hat, conversely, these relalions can be inverted to express II¡, u2, U, 1ll
6.17. Orrhogonal Curvilinear Coordina tes 307
terms of x. y,::, when x, y,:: and/or 11" u 2 ' u) are suitably restricted. Then, at
least in some region, any point wilh coordinates (x, y,::) has corresponding
coordinates lu" u" u». We assume that the correspondence is unique. If a
particle moves from a point P in such a way that u, and 113 are held constant
and only u, varies, a curve in space is generated. We speak of this curve as the
11, curve. Similarly, two other coordinate curves, the U 2 and U 3 curves, are
determined at each point (Figure 6.13). Further, if only one coordinate is held
constant, we determine successively three surfaces passing through a point of
space, these surfaces intersecting in the coordinate curves. lt is often convenient
lO choose the new coordinates in such a way that the coordinate curves are
mutually perpendicular at each point in space. Such coordinates are called
orthogonal currilinear coordinates.
z z
/ y
y
() r /
/
x x
As an example (see Figure 6.14), in circular cylindricaJ coordinates (1', e, z)

we have
x = r cos e, y = r sin e, z z,
where r::> o and o <:: e < 2¡¡;. If we take
U¡ = r, tlz = e,
then at any point P the U z curve is a circle of radius r and the tI¡ and u) curves
are straight lines.
Let r represent the position vector of a point P in space,
r = x i -j yj+zk. (142)
Then a tangent vector to the u, curve at P is given by
V _ ar _ ar ds], (143)
, - au, - as, dtl]
where s, IS arc length along the u, curve. Since ai:las¡ is a unit vector, we can
write
V, = h,u" (144)
where u, is the unit vector tangent to the ti, curve in the direction of increasing
308 Vector Analysis
arc length and h, = ds ,/ du, is the len glh of U,. Considering the olher coordinate
curves similarl y, we lhus write
V J = hJu J , ( 145)
where Uk (k = 1,2,3) is the unit vector langent to the l/k curve, and
h, = ds,
du,
= Iau,
ar 1, h, = ds z =
- duz
Iau,
ar 1, hJ = dS J =
dU J
I~
aU J l·
(l46)
These equalions can be wrillen in the differential form

dS J = h J dUJo (147)
We thus see thal h" h" hJ are of lhe nature of sca fe factors, giving the ralios of
differential dislance s to lhe differenlials of lhe coordinate parameters. It should
be noticed that the calculation of V" from Equation (143), determines both
the scale factor h, = IV,I and the unil vector u , = V,/h,.
We adopt the convention that u " U z, and u J form a right-handed sys tem in
¡he order written , in consequence of an appropriale numbering of the coordi-
nales.
If s is arc len g th a long a c urve in any direction, the vector
d r = ar du, -+- ar duz + ar dU J
ds au, ds . au z ds ¡¡u;ds
_ V du, . V duz . V dU J (148)
- 'ds --¡--Z ds --¡-- J ds
!.
has unit length . Thus, if th e coo rdinate curves are orthogonal, so that
V, • V z = V z • V J = VJ • V, = O, (149)
follo~s
'0 0
there
V, • V, du} + V 2 ' Vzdu~ + V J ' V J du~ = ds 2
or, using Equation (145),
ds 2 = h} duf + hi du~ +h5 du ~ . (150)
To find the element of volume, we notice that the vector s V, du" V 2 duz,
and V J dU J are mutually perpendicular vectors ha ving as their lengths the arc-
le ngth differentials ds" ds 2 , a nd ds J . Thus the elemenl of volume d. is given by
the volume of the rectangular p a rallelepiped determined by these vector s,
d • . = V, du, X V 2 duz' V J dU J = (u, X U2 • u J ) h,h 2 h J du, duz du J ,
and hence
(151)
The fact that V, x V 2 • V J is positive and that u, x U 2 • u J = + 1 is a consequence
of the assumed right-handedness.
On the surface u, = consta nt , the vector element of surface area dcr, is, in a
s imilar way, given by the vector product V 2 du z x V , du J , and hence
(152)
6.17. Orthogonal Curvilinear Coordinates 309
Analogous expressions are obtained on the other coordina te surfaces. In

particular, the sealar surface elements on the eoordinate surfaees are of the form
(153)
We next proceed to the determination of expressions for the gradient,
divergence, and cud, and for the Laplacian operator V2 in the present co-
ordinates. To determine the gradient of a scalar functionf, we make use of the
relation (64),
df = Vf' dr (154)
which specifies the basic geometrical property of the gradient. In terms of the
coordinates U¡_ Uz , u] we have
al ,af af
df = -a'
u]
du¡ , -a
U
duz
z
+ -a dU J UJ
and dr = V¡ du, + V dU + V dU
2 2 J j
= h,u¡ du¡ + h,u z dU + h]u] dUJo

2
If we now write
(155)
where the A's are to be determined, the substitution of these expressions into
(154) gives
~ ~~
au, du, , aU duz + ~
au] duo = h,A, du, + h,A 2 du, + ,
h]ItJdu J,
2
and hence, since the variables tl p u 2 , u] are independent, we have
(k = 1, 2, 3). (156)
Thus Equation (155) beco mes
Vf = ~ af + u, af + u J af (157)
h, au, h, au, h J aU J
and the operator V accordingly has the form
V = ~~ + U2 ~ + UJ ~ . (158)
h,au¡ h 2 aU 2 hJau J
Equations (157) and (158) can be written in the equivalent forms
_ af, af af
Vf - u¡-a
s, ' u'-a
s, + u]-a
s] '
(159)
V = u¡~
as] + u,~
-as, + u]~,
as]
(160)
wh ich display the intrinsic nature of the operator V and which, indeed, could
have been anticipated from the results of Section 6.7.
310 Vector Analysis
In particular, there follows

U,
VU¡ =71' Vu, u, (161)
, =]/' ,
for the gradients of the independent variables u" lI z , lI,.
From these special results we now obtain two further results which will be
useful in the following work. First, since V x VU k = O, from (74f), we conclude
that the expressions u,/hl> u 2 /h 2 , and u,/h, have zero curl,
V x .!!..!. = V X Uz = V X u' = o. ( 162)
h, hz h,
Second, slllce u, = U2 X u" there follows
Uh'
h2 J
= hUz x ?, = (Vuz) x (VU3)'
2 n3
and two analogous expressions are similarly obtained. Reference to Equation
(74i) then shows that (he expressions u,/(hzh,), uz/(h,h ,), and u,/(h ,h z) have zero
divergence,
(163)
It should be kept in mind that while tbe vectors u" U z, and u, are of constant
length unity, the directions of these vectors will in general change with position
in space, and hence their divergence and curl will not, in general, vanisb.
To find an expression for the divergence of a vector F in u" u 2 , u, coordi-
nates,
F = F,u, + Fzu z + F,u" (164)
we first write
V • F = V • (F,ua + V • (F 2 u 2 ) + V • (F,u,).
If we now write the first term in the form
V • (F, u,) = V • [(hzh,F,) (h~h,) ]

and notice that, according to Equation (163), tbe second factor in the brackets
has zero divergence, we obtain from (74a) the result
1
V • (F,u,) = hUh' • V(hzh,F,) = h h h d. a (h zh 3 F,).
2 J 1 2 3 U 1
Treating tbe otber terms in a similar way, we obtain finally
V • F = h h1 h [a
1
-a (hzh,F,)
2. 3 U J
+ a-(h
a
U2
3 h¡Fz) + -a
a (h,hzF,) ] .
U3
(165)
In particular, by combining Equations (158) and (165) we find the expres-

sion for the Laplacian operator,
VZ = V • V = 1 [~(h,h,~)
h,h 2 h, au, h, au,
( 166)
6.18. Special Coordina te Systems 311
In a similar way, to find an expression for the curl of F,

V x F = V x (F,u,) -;- V x (F,u 2 ) + V x (F,u,),
we write the first term in the form
VX (F,uJ = Vx [(h,F,)(~:)J
so that, according to Equation (162), the second factor in the brackets has zero
curl. Then, from (74b), we obtain the result
V x (F,u,) = - (~: x V)(h¡F,)
- hUh'
I 2
aaU2
(h,F,) + hUh'
) I
aau) (h,FJ
= h,h1 h,
2
(h,ua
zau , - h,u a) (h¡F,).
3auz
If the other terms are treated in a similar way, the result can be written in the
form of a determinant
h,u, h 2 u2 h,u,
1 a a a
V x F = h h h, l2 au , au z au, (167)
h,F, h,F2 h,F,
For the rectangular coordinates (l/I' l/2' l/,) _ (x, y, z) we ha ve h, = h, =
h, = 1, and al! the preceding results are seen to reduce to the forms originalIy
given in that case.
6.18. Special Coordinate Systems. In circlllar cylindrical coordina/es (r, e, z)
we havet
x = r cos e,
y = r sin z = z, e, (l68a)
where r :> o and o <: e< 2n. Since r, e, and z in Ihal order form a right-handed
system, we may take
(r, e, z) (u" u2, u,).
The position vector r has the form
r = ir cos e + jr sin e + kz.
Thus we have
U, = ar
ar = I. COS e..L··
,J SIn e, Ve
ar =
= a(J -ir sin (J + jr cos (J,
From Equation (146), there follows
h, = 1, he = r, (168b)
and hence also
tWe notice that the coordinate r in this system is distance fram lhe z axis. It must not be con·
fused with Ir 1, where r is the position vector and Ir I is distance from the origino
312 Vector Analysis
u, ~ i cos e + j sin e, Ua = -j sin e + j cos e, (l68c)

Eq uations (l SO), (l S 1), and (153) give
ds = .y dr'- + rO de 2 + d~2,
d. = r de dr d:: , ( l68d)
da, = r de d::, dae = dr d~, da, = r de dr.
Finally, Equations (157), (165), (166), and (167) give the results
af 1 af af
Vf = u, ar + u e¡:- ae + u, a:!'
v.F = _1 .2...(rF) + _1 aFe T aF"

r ar ' r ae az
1 a ( af ) 1 a 2f I a 2f
V2f = ¡:- ar r ar T r2 ae2 T az2' (168e)
1 a a a
v x F = ¡:- ar ae az
rFe F, F,
In spherical coordinares (r, rp, e), where r is distance from (he origin, e is the
polar angle measured from the xz plane, and rp is the "cone angIe" measured
from the z axis (Figure 6.15), there followst
x = r sin rp cos e, y = r sin rp sin e, z = reos rp, (169a)
where r ::> O, O < rp < 71:, O < e< 271:, and the following resuIts are obtained:
hrp = r, he = r sin rp, (169b)
u, = ¡sin rp cos e + j sin rp sin e + k cos rp,
u. = i cos rp cos e + j cos rp sin e - k si n rp, (l69c)
Ue = - i sin e + j cos e,
z p
Figure 6.15
tThe angles ip and (J are defined in different ways by different writers. The present convention
is consistent wirh tha! used for circular cylindrical coordinates in the plane z = O.
6.19. Applicalion to Two-Dimensional lncompressible Fluid Flow 313
ds =, ,b' z + rO drpz -+- rO sin Z rp d(J2,

d, = rO sin rp dr drp de, (I69d)
d(J, = r2 sin rp drp de, d(J. = r sin rp dr de, d(Je = r dr drp,
and also
VI = u al -'-- ~ al..L u. al,
, ar ' r arp , r sin rp ae
V • F = ~2
r aar (r2FJ - r sm
l
rp aarp (F_- sin m)
r
+ r si~ rp aaFee,
V21 = _12 ~ (r2 al)
r ar ar
..L l .E...- (sin al)
'r 2 sin rp arp rp arp
+ r 2 sinl 2
aOI,
rp ae 2
(16ge)
u, ru. r sm rp U e
1 a a a
V xF=
r' sin rp ar arp ae
F, rF. rsinrpF.
6_19_ Application lo Two-Dimensional Incompressible Fluid Flow. In certain

types of fluid flow it is possible to choose a system of orthogonal curvilinear
coordina tes in such a way that the flow depends only upon two of the coordi-
nates and is independent of the third. The simplest example is that in which flow
is parallel to aplane, say the xy plane, and is such that its properties everywhere
are then specified by its properties on the xy plane. In the more general case,
we assume that the flow is independent of U J and is then specified by its properties
on any surface for which U J is constant. We suppose also that the coordinate
system is such that the scale factors h" h" and h J are independent of U J .
If the flow is divergenceless and irrotational, that is, if there are no distribu-
tions of sources or sinks or of vortices, about which the fluid tends to rota te,
and if also the fluid is assumed to be incompressible, we have seen that the
velocity vector V is the gradient of a Cunction rp, called the ve/oei/y potentia/,
and that rp satisfies Laplace's equation.
Since in the 'present case V has no component in the U J direction, we may
write
V = V,u, + V 2 u" (170)
where V, and V 2 are the components oC V in the directions of u, and u 2 , respec-
tively. Since the velocity is to be independent of ¡¡J' the relation V = Vrp becomes,
by virtue of (158),
V _ ~ arp ..L u" arp , (171)

- h, au, ' h2 au,
and hence we ha ve
(172)
31~ Vector Analysis
A Iso, the equation V 2rp = ° becomes

( 173)
When the solution rp of the partial c!ifferentiaJ equation (173), satisfying

appropriate conditions on the boundary of the region considered, has been
obtained, the equipotentiallines in the lI, surfaces are curves given by equations
of the form
(174)
as e ¡ takes on successive constant values. The veJocity vectors, and hence the
corresponding Iines of flow or streamlines, are normal to these curves. If we
express the streamlines by equations of the form
(175)
then, since we must have
Vrp • V1p = 0, (¡ 76)
t here follows
1 arp) (1 alf/) + (1 arp) (1 a1p) 0,

( --¡¡; au¡ --¡¡; au¡ hz au z h z au z
.'
alf/ a1p
or '}U2 au¡
h z arp h¡ arp
h¡ du¡ h z au z
r
If we denote the common value of these two ratios by ¡.¡., where ¡.¡. 15 an
unknown function of U¡ and l/z, there follows
.'
1.1
( 177)
Todetermine ¡.¡., we make use of the equation

aZIf/ = az1p ,
aUZau¡ au¡aU Z
assuming appropriate continuity, and so obtain from (177)
~
au¡ (¡.¡.h
z arp) + ~ (¡.¡.h¡ arp)
h¡ au¡ au z h z au z = ° .
lf this equation is compared with (173), one obtains the result ¡.¡. = eh" where e
is an arbitrary constan!. As will be seen, it is convenient to take e = 1, so thatt
¡.¡. = h,. (178)
"iEquation (178) defines a particular solution of the equation determining J-i.. The mast general
solution is !l ~ h,f(IjI), wherefis an arbitrary function of the expression ljI being determined.
The more general soJution Jeads only to the obvious ract that any function or ¡he ljI deter-
mined by Equarion (79) will also be constant along the streamlines.
6.19. Applicalion lo Two-Dimensional Incompressible Fluid Flow 315
Thus, from Equations ( 177) an d (178) we obtain the relationship
d/{/ = _ h,h, arp du, _~ h 2hJkdu" ( 179)

h, du , h, au, -
from which /{/ can be determined by integration when rp is known. By making
use of (172), Equation (179) can al so be written in lhe form
d/{/ = hJ(-h, V, du, + h,V, duz). ( 180)
By using Equation s ( 160), (171), and (172), we obtai n the relation
V/{/ = -!!.J. ~u, +!!.J. ~U2 = h,(-V2 u, + V,u 2 ), (1 8 1)

h 2aU2 h, au,
which can also be written in the form
V/{/ = .h , u, x V. ( 182)
Hence lhere follows al so
vz/{/ = V • (h,u J x V) = V • V x (hJu J) - hJu, • V x V
= V • V x (h~ ~ ~) = V • ( Vh 1 x ~ :) = ~~~ • hJu J x V
or V 2/{/ = (V log h D • V/{/, ( 183)

when use IS made of Equation s (74c), (74b), and (162), and of the fact thal
V x V = O in con sequence of the assumed irro tational nature of the ftow .
Thus we conclude that if h, is con s tant then the function /{/ in the fto w
under consideralion also satisfies Laplace's equation, and hence may be ' con-
s idered a s the velocity potential co rrespo nding to a second f10w having the
curves rp = e, as streamlines. The two flo ws so related are said 10 be conjugate.
If h J is not constant, the fun c tion /{/ doe s not satisfy Laplace's equation,
and hence no conjugate (nondivergent) ftow exists.
In any case, the function /{/ is known as the strearn lunction of the f10w for
which rp is the velocity potential. We notice from (179) that the expression for /{/
involve s an arbitrary additive constant (as well as the assigned multiplicative.
constant c). The additive constant can be chosen so that along a particular
reference streamline we have /{/ = O.
In order to establish another useful property of the stream function /{/, we
consider a curve e in a u, surface, joining two points Po and P,. Then the rate
of mass flow, in the direction of the normal D, through a "rectangular eJement"
based o n a linear element dr of this curve, and ex tending in the U 3 direction a
distance corresponding to a unit increment in u 3 , is given by
di = pV • D h , ds = pV • (~: x u ,) h, ds = ph , u, x V • dr.
If use is made of Equation ( 182), this relation take s the form

di = P V/{/ • dr = p d/{/ .
316 Vector Analysis
Thus, since here p is constant, we obtain by integration along e from Po to P,

the result
( 184)
where ~c indicates change along e, so that the difference between the values of
ljI at two points in a 113 sU/jace is numerically equal to the rate 01 mass flow of a
fluid with unit density "across any curve in that sU/jace which joins these points."
Example. We verify that the function

rp = X2 _ y2
is a solution of Laplace's equation in the xy plane,

2 _ a2 rp a2 rp_
V rp - ax2 + ay2 - 0,
and hence is the potential function of a flow of the type considered. In this case we
have
li l = x, 112 = y, h3 - 1.
The velocity vector V = Vrp becomes
V = 2xi - 2yj,
."
and the equipotential curves in the xy plane are the hyperbolas
X2 _ y 2 = el.
The stream function ljI IS determined from

Equation (179),
y
'" = constant
dljl = 2y dx + 2x dy = 2d(xy) ,
and hence the streamlines are the hyperbolas
'" = constant
/ /
/ / ljI = 2xy = C2'
/
/
/
/
/
/
/
/
/
Since, in particular, the lines x = 0, y =
streamlines, the f10w in the first quadrant is a
are °
plane flow around a right-angled comer (Figure
6.16). Also, since ljI = 2xy also satisfies Laplace's
equation, ljI can be considered as the velocity
potential in a conjugate flow where the curves
x
X2 - y2 = c, are streamlines, and the function
Figure 6.16 rp = X2 - y2 is the stream function. •
6.20. Compressible Ideal Fluid Flow. To further illustrate the use of vector
methods, we now treat briefly the basic equations governing the flow of non-
viscous (ideal) compressible fluids.
We consider first an element of fluid mass dm in the form of a rectangular
parallelepiped with edges dx, e/y, dz parallel to the coordinate axes at the time t
and moving with velocity V at that instant. If we denote the pressure by p, the
differential force exerted by the pressure distributions on the two faces parallel
6.20. Compressible Ideal Fluid Flo,," JI7
[O ¡he y::: plane is readily found to be -[i(ap/ ax) dx] dy dz. Considering the forces
acting on the other four faces similarly, we ¡hen find that ¡he resullant differ-
ential force due to fluid pressure on ¡he element dm is given by - VP dx dy dz
or, equivalently, by -(Vp / p) dm. The same expression can be shown (by use of
the divergence theorem) to be valid for the differential force due to pressure
acting on an element of volume (dm)/ p of any shape.
As an element moves in space, its shape and vo!ume generally will change.
However, its mass dm must be conserved. For an element of constan! mas s dm
the ra¡e of change of momentum is given by (dV/dl) dm. Hence, if external forces
are omitted, Newton's second law of motion leads !O ¡he equation
dV
PIJI + Vp = O. (185)
The velocity V is a function of position x, y, z and time 1, whereas the

position coordinates x, y, z depend not only upon time 1 but also upon the
initial values X o , Yo, LO at a reference time, say 1 = O. Thus the time derivative
dV/ dl in (185) actually is not a total derivative in the strict sense, but is calculated
only for fixed initial coordinates. That is, the time differentiation follows a
given particle in its motion. For this reason this derivative is frequently termed
the "particle," "substantial," or "material" derivative of V, and is often indicated
by the special notarion DV / Dr.
For fixed initial coordinates, V depends directly on x, y, z and 1, and x, y,
and z themselves depend upon the time 1, in such a way that dx/ dl = V... and
so on. Hence we may write the substantial derivative of V in the expanded form
(see Section 7.1),
dV _ av dx + av dy -L avaz + av
dI - ax dI ay dI ' az al al
V av + V av + V av + av
"ax Yay 'az al
or dV = (V • V)V + av. (186)

dI al
In terms of differential operalOrs we may write also
d
-dt = V • V + -aat =
a
V y -C1X
_ + a
Vy - y
a
+ a
V'-az + a
-aI . (187)
Thus Equation (185) can be written in the form
p(v. VV + ~~) + Vp = O ( 188)
and is equivalent to three scalar equations, the first of which is of the form
p(v av x
ax
•y
+ V avx
y ay
+ V avy
, az
av. +
+ TI y) ap =
ax
o.
These equations of motion are known as Eu{er's equalions.
318 Vertor Analysis
[n addition to the vector equation (188), the condition of conservation of

mass must hold, so that Equation (112),
pV. V -(V. Vp -'- aalp )

=
I
= _dp,
di
(189)
must be satisfíed. This equation is known as the equalion 01 conlinuity.

Finally, lO complete the system of basic equations, \Ve may take the effect
of compressibility into account by assuming a suitable relationship between
pressure p and density p. (For an incompressible fluid, p is constant.) Assuming
such a relationship, we may then write
V = ¡ap + .ap + k ap = dp (¡a p + .ap + k ap ) = dp Vp.

p ax J ay az dp ax J ay az dp
The quantity
.j~~ = Vs (190)
is known as the local sonic velocity, and is seen to vary with p from point to
point in the fluid, and with time.
With these results Euler's equations in the vector form (188) beco me
pV.VV+paa; + VjVp=O. (191 )
Equations (189) and (19 J) are sufficient lo determine p and V if suitable bound-
ary conditions and¡or initial conditions are prescribed.
.• ~ o, ' In many problems the flow is nearly uniformo That is, the vector V and the
scalar functions p and pare near[y constant. In such cases, Equations (189)
L'I~
1.1,
and (191) can be /inearized by first writing
V = D + u, p = Po + J, (192)
I where D, Po, and V so are the constant values corresponding to the uniform
flow and the quantities u, J, and V s are considered as small deviations. With
this notation Equations (189) and (191) become
(Po I J) V • u (U + u) • V J
aJ
+ TI = O,
(Po + J)(U + u) • Vu + (Po + J) ~~ + (Vso + V S )2 VJ = O.
If products of small deviations are assumed to be relatively negligible, these

equations are reduced to the linear forms
Po V . u + (U • V + :J J = 0, (193)
( U. V + ~)u + Vjo VJ = O. (194)

at Po
For an irrotational flow (see Section 6.15), the velocity V, and hence also u,
can be expressed as the gradient of a scalar function. Hence, if we write
u = Vrp, (195)
6.20. Compressible Ideal Fluid Flow 319
Equatíon (194) becomes
V[O+ t;o(UoV+ ;J91J=O,

from which the den si ty deviation o is expressed in terms of lhe ve/ocily deviation
poten/ial91 in the form
o -- -...J!.Q..
Vio
(u o V + .i.)
al ro "1'>
(196)
excepl for an additive function of time only, which can be considered as incor-
porated in 91. It should be noticed that all the preceding relations involve only
the invariant operator V, insofar as space differentiation is concerned, and
hence may be readily expressed in terms of any convenient coodinate system.
We now s uppose that the uniform fiow is parallel to the x axis , and so write
u = Vi, (197)
in which case Equation (196) becomes
o= - 4-
V so
(V 0P. + a91).
ax al (198)
If Equations (195), (197), and (198) are introduced into (193), we obtain the
equation
V291 __1_2 (v.i.
Vs o a.'C + .i.)291
at = O, (199)
which must be satisfied by the potential function 91. For an incompressible

fluid V s = 0 0 , and the equation reduces to Laplace's equation, in accordance
with the results of Section 6.15.
In the special case of steady flow in two dimensions, where the fiow is parallel
to the xy plane and the velocity and density do not vary with time, Equation
(194) reduc·e s to the form
a~ a 91_ O, 2
x + -ay
_ 2
(l !vi) 2 2 - (200)
where Mis the so-called local Mach number,

V (201)
!vI=-V '
so
and is the ratio of the speed of the uniform fiow to the sonic velocity corre-
sponding to rhis flow. A f10w is said to be subsonic when M < 1 and supersonic
when M> l. It is important to notice that the coefficient of 2 2 in Equation a 91¡ax
(200) changes sign when the sonic velocity is passed.
In ·the special case of nons/eady one-dimensional flow, when the velocities
in the y and z directions are assumed to be negligible, Equation (199) becomes
(202)
320 Vector Analysis
lf the aboye developments are still assumed to be satisfactorily approximate

when small deviations are measured from a stafe ofrest, so that U = O, we may
replace u by V and rewrite (198) and (199) in the form
o= P ~ Po = ~ P~ aq; (203)
V so al
2 _ 1 a 2 rp
and v q; - V2
so
-a
I
2' (204)
where V = Vrp. (205)

Bere V]o is essentially an effective "mean value" of dp/dp. Thus we have also,
approximately,
p ~ Po = V]o(p ~ Po)
so that the pressure change due to ¡he disturbing flow is given approximately by
p ~ Po = ~p o~
arp. (206)
To conclude this section, we obtain an explicit exact integral of Euler's

equation, when external forces are absent, in the case of irrotational flow
(V x V = O). From the identity
V(V • V) = 2V • VV + 2V x (V x V)
[see Equation (74e)], we here obtain the result
V • VV = V(-± V2),
,-
where V = 1V 1 = -JV • V. Thus Euler's equation (188) can be written in the
form
V (~V2) + av
2 al + Vp
p
= O. (207)
If we now write
V = Vrp (208)
and define a function P by the relation
VP= ~ Vp or P= Jet;. (209)
Equation (207) takes the form
v(; V2) + %1 Vrp + VP=O
or V (; V2 + ~~ + p) = O. (210)
Bence the quantity -±V2 + (arp/at) + P must be independent of the space co-
ordinates and, accordingly, a function of time only,
; V2 +~+ P = f(I). (211)

Problems 321
This is one form of B ernou/li's equation. For s teadyflow (in wbich the f10w
IS independeot of time), thi s equation beco mes
tV2 -i- P = constan!. (212)
In particular , for an in co mpressib le fluid (p = constant) we may take P = pi p,
in accordance with (209), and hence (212) can be written in the form
ipV2 + p = constant (213)
in thi s case.
rn illustration, for the example at the end of Section 6.19 involving f10w
around a right-angled comer, Equation (213) determines the fluid pressure p in
the form
p = p, _ 2p(x.2 + )/2),
where p, is the pressure at the comer.
REFERENCES
1. BRAND, L., Vector Analysis, John Wi[ey & Sons, Inc., New York, [957 .
2. KELLOGG, O. D., Foundations of Potentia/ Theory, Dover Pub[ications, Inc., New
York, 1953.
3. PHILLlPS, H. B., Vector Ana/ysis, John Wi[ey & Sons, Inc., New York, 1933.
4. STRU1K, D. J., Lectures on C/assical Differentia/ Geometry, Addison-Wes[ey Pub-
lishing Company, Inc., Reading, Mass., 1950.
PROBLEMS
Seclion 6.1
1. Find tbe 1ength and direction cosines of the vector a [rom the poin! (l, -1, 3) to
the midpoint of the line segment from the origin to the point (6, - 6,4).
2. The vectors a and b extend from the origin O 10 the points A and B. Determine
the vector e which extends from Oto the point C which divides the [ine segment from
A to B in the ralio m : n.
3. (a) Ir O denotes the ang[e between the vectors a and b, use a theorem of e[emenlary
geometry to show that
la + bl 2 lal 2 + Ib l' + 2lallblcosO.
=
(b) Ir a = axi + aA + a,k and b = bxi + bA + b,k, use the preceding resu[¡ to
show that
- a .,bx + a,by + a,b,
cos O - lallbl .
4. Prove tha! la + bl <: lal + Ibl and la + bl :> lal - Ibl.
322 Vector Analysis
5. (a) Show that a straight line with direction angles ex, p, y can be s pecified by the
equations
x - Xo _ y - Yo z - Zo
cos ex - cos p - cos y
where (xo. Yo, zo) is a point on the lineo
(b) Show that the above equations can al so be written in the parametric form
x = xo + / cos ex, y = Yo +I cos p, z = Zo + I cos y.
where I is a parameter.
(Notice that in either set of equations cos ex, cos fJ. and cos y can be replaced by
direction ralios A , B, and C which are proportional to them.)
6. The equations of a straight line are of the form
x-l y+2 z
I - 2 = -2 '
(a) Show that the line ineludes the point (1, - 2, O) ando determine its direction
COSiDes.
(b) Determine a unit vector in the directioo of the line.
Section 6.2
7. Pro ve geometrically, from the definition of the scalar product, that the distributive
law a • (b + e) = a • b + a • e is valid.
8. The vectors a and b are defined as follows:
a = 3i - 4k , b = 2i - 2j + k.
(a) Find the scalar projection of a on b.
(b) Find the angle between the positive directions of the vectors.
9. Find the magnitude of the scalar component of the force vector
F =i+ 2j + 2k
iD the direction of the straight line with equations x = y = 2 •.
10. A plaoe is determined by a point Po(xo. Yo. zo) on it and by a vector
N = Ai + Bj + Ck
normal to il. Show that the requirement that the vector from Po to a point P(x, y, z)
10the plane be perpendicular to N determines the equation of the plane in the form
A(x - xo)+ B(y - Yo) + C(z - zo) = O.
11. (a) Show that the vector Ai + Bj + Ck is normal to the plane
Ax + By + Cz = D.
(b) Pro ve that the shortest distance from the point poexo, Yo, zo) to the plane
Ax + By + Cz = D is given by
d = l Axo + B yo + CZ o - DI.
,,¡ A 2 + B2 + C2
[Le t P ¡(x 1, Y 1, Z 1) be any poin t on the plane and determine t he projection of the vector
from Po lO P, on Ihe normal 10 the plane.]
& ,'oolems 323
J2. Find the angle between the planes 2x - y + 2= = I and x - y = 2.
Seclion 6.3
13. (a) Determine a unir vector perpendicular to the plaoe of the vectors a = i + 3j - k,
b = 2i + j + k.
(b) Find the area of the triangle of which these two vec tors forrn coterminou3
sides.
14. (a) Determine a unit vector normal to the plane determined by the points (O, O, O).
(1.1,1), and (2, 1. 3).
(b) Find the area of the triangle with vertices at the points defined in pan (a) .
15. A rigid body rotates with angular velocity úJ about the line x = y = z. Fi!ld the
s peed of a particle at the point (1, 2, 2).
16. Find the scalar moment of the force F = i - 3j + 2k, acting at,the point (1, 2, 1),
abou t the z axis.
17. Show that the shortest distance from a point P o to the line joining the points P,
and P, is given by
d = I v, X V2 I ,
I v31
~ ---> --->
where v, = PoP" V 2 = P O P 2 , and v, = P,P 2 • (Notice that Iv, x vol is twicethearea
of the triangle PoP, P 2 , and that the desired distance is the altitude of that triangle
normal to the base P, P 2 .)
18. Show that the shortest distance between the lines AB and CD is the projection of
~
AC (or of the vector joining any point of AB with any point of C D) on the vector
~ ~
AB x CD.
Section 6.4
19. From the last result of Section 6.3, deduce that, if r denotes the vector from a
point O to the point of applica,tioo of a force F, then the sea lar moment of F about
an axis OA is given by MOA = (r Fu), where u is a unit vector in the directioo of OA.
20. Find the volume of the tetrahedron with vertices at the points (O, O, O), (1, 1, 1) ,
(2,1,1), and (1, 2,1).
21. Show that the assertion
(abcXdef)= b-d b-e b-f

c-d e-e e-f
can be interpreted as a sta tement of the fact that the determinant of the product of
two 3 x 3 matrices is the product of their determinants.
22. Given that
(a X b) x e = A(la • e)b - (b - e)a],
for some scalar A lsee Equation (33)], determine A as follows: Let u be a unit vector
parallel to a, v a second unit vector perpendicular to a and such that b is in the plane
324 Vector Analysis
of u and v, and w = u X v. Substitute ¡he relations a = a,u, b = p,u + P2V, and

e = y¡u + yzv + Y3W into the given relation and show that A = 1.
23. Deduce Equation (34b) from (34a).
24. Determine a unit vector in the plane of the vectors i +j and j +k and perpen-
dicular to the vector i + j + k.
25. If u is a unit vector, prove that u x (a x u) is the vector projection of a on aplane
perpendicular to u.
26. Prove that a x (b x e) + b x (e x a) + e x (a x b) = O.
27. (a) Prove that
(a b d)e - (a b e)d,
(a x b) X (e X d) = {(a e d)b - (b e d)a.
(To obtain the first form, write temporarily u = a X b.)

(b) Show that this vector, if not a zero vector, is in the direction of the intersection
of aplane including the vectors a and b with one including e and d.
Seclion 6.5
28. Jf F is a function of 1, find the derivative of
• F . dF x d 2 F
di dl 2
29. At time 1, the vector from the origin to a moving point is

r = a cos ()JI + b sin 0]1,
where a, b, and O] are constants.
(a) Find the velocity v = dr/dl and prove that r x vis constant, so that the curve
traced out lies in aplane.
rI (b) Show that the acceleration is directed toward the origin and is proportional
to the distance from the origino
30. Let r represent the vector from a fixed origin °
to a moving particle of mas s m,
subject to a force F.
(a) If H denotes the moment of the momentum vector mv = m(dr/dl) about 0,
prove that
dH d
-dI = m-(r X v)
dI
= r x F = M
'
where M is the moment of the force F about O.
(b) If the force F always passes through the fixed point 0, show that
d
dI (r x v) = O.
and hence deduce that r X v = h, where h is a constant vector. Deduce also that the
motion is in aplane.
(c) Show that 1 r x vi is twice the rate dA/dI = ir 2 de/dI at which area A is swept
over by the vector r, and hence deduce that, when a mass at P is subject to a "central"
force, which always passes through a fixed point 0, the vector OP moves in aplane
and sweeps over equal areas in equal times.
Problems 325
31. If u is a unit vector originating at a fixed point O, and rotating about a fixed vector
ro through O, with angular velocity of constant magnitude w, show that
du
dt = O) x u.
32. Let r = xi + yj + zk represent the position vector from a fixed origin O to a

point P, and suppose that the xyz axis system is rotating about a fixed vector O) through
O, with angular velocity of constant magnitude w.
(a) By calculating dr/dt, and noticing that di/di = O) x i, and so forth, obtain
the velocity vector in the form
v = Vo + O) X r,
where the vector
Vo = ~;i + ~i + ~:k
is the velocity vector which would be obtained if the axes were fixed.
(b) Obtain the acceleration vector in the form
a = ao + 20) x Vo + O) x (O) X r),
where Vo is defined in part (a), and where
'd 2 x. d 2 y. d 2z
ao = di 2 I + di 2 J + dt 2 k.
33. lf the system of Problem 32 is rotating with angular velocity of constant magnitude
w about the z axis, the z axis being fixed, show that the equations of motion for a
point mass m are of the form
m(~;~ - 2W~ - W2 x) = Fx>
d 2y dx )
m ( dl Z + 2w di - co y
2
= F"
d 2z
m di z = F"
where Fx> F y , and F: are the components of the external force along the respective
rotating axes. [Notice that the mass hence behaves as though the axes were fixed, with
an additional force 2mw(dy/dl) +
mw 2 x acting in the positive x direction and an addi-
tional force -2mw(dx/dt) + mw y acting in the positive y direction.]
2
SectioD 6.6
34. Ca) lf e is polar angle, show that the vectors
U1 = i cos e + j sin e, = U2 -isine + icose
are perpendicular unit vectors in the radial and circumferential directions, respectively,
in the xy plane, and that
dUI dU2
de = uz, de = -UI'
(b) For points on a plane curve in polar coordinates, the position vector is of the
form r = ru l ' By differentiation with respect to time t, obtain expressions for the
326 Vector Analysis
veetors of veloeity and aeeeleration of a point moving along the curve and show that
the radial and eireumferential eomponents are of the form
L'r = f, VB = re,
-
a. = re + 2f () = r1 d .
at(rZe),
where a dot denotes time differentiation. [Notiee, for example, that ü¡ = (du¡/de)().]
35. (a) Verify that the parametrie equations
x=aeost, y = a sin t, z = et
speeify a right circular helix in spaee.
(b) Show that for this curve there follows
dt, ds = va 2 +e 2
where s is are length.

(e) Determine the unit tangent, principal normal, and binormal vectors, and
show that the radii of eurvature and torsion are given by
p= a
2 + el
, T= a
2
+e 2
.
a e
(d) Show that the oseulating, normal, and reetifying planes at the point (a, O, 2ne)
are speeified by the respective equations
az - ey = 2nea, ay = 2ne 2 , x = G.
In Prob/ems 36-39, the position vector to a curve e is assumed to be expressed in

the form r = x(t)i y(t)j + +
z(t)k, and a prime is used to denote d!fferentiation with
respeet to the parameter t, whereas s denotes are /ength a/ong C.
36. Establish the relations
r' r'
( a) s' = I r' I = V r' • r", (b) u = f?l
37. (a) Show that
1 du u' (r' • r')r" - (r' • r ")r' (r' X r") X r'
pD = ds. 7 = Ir' 14 = Ir' 14
(b) Use the identity of Lagrange [Equation (36)] to obtain the resuIt
I (r' X r ") X r' I = I r' x r" I I r' l.
(e) Deduce the following results:
1 I r' x r" I D=
(r' X r") X r'
.
p = Ir'I' ' Ir' x r"llr'l
38. Use Equation (52) and the results of Problems 36(b) and 37(e) to obtain the resuIt
r' x r"
b = Ir' X r" I
39. (a) Verify that
u X du. d u =
ds ds 2
2
(~) f X'3
(s'rn - s"r') . (r'" + ... ) _
I J -
(r' r" r"')
'6 '
S S S S
where the omitted terms in the third factor of the seeond expression involve r' and
r" linearly, and henee do not affeet the value of the produet.
Problems 327
(b) Using the relation (55a),

2U) ,
+=P'(u x du . d
ds ds'
and the results of Problems 37(c) and 39(a), obtain the result
(r ' r rUI)
N
1: I r' X r" 12
40. Determine u, n, b, p, and 1: for the right circular helix

x = a cos 1, y = a sin 1, z = el
by using preceding results, and compare with the results of Problem 35. (The formulas
obtained in ProbJems 36-39 are particuJarly useful when t cannot be simply expressed
io terms of the are Jength s.)
41. Use appropriate results of Problems 37 and 38 to show that the curve for which
r = X(I); + Y(I)j + z(l)k is a straight line if r' x r" = O and is a plane curve if
r ' X rl! . r / /! = o.
42. Determine the curvature and torsion of the twisted cubic x = 1, Y = 1', Z = 13
at the point (l, 1,1).
43. Bending 01 a rod. Suppose that a thin rod of uniform circular cross section is bent
and twisted in such a way that its axis coincides with a space curve C. Let the intensity
of the applied load per unit distance along C have components Pu, Pn, and P. along the
tangent, principal normal, and bioormal, respectively, of the deformed rod and write
p = PuU + Pnn + p.b. Similarly, write m = muu + mnn + m.b for the vector whose
components are intensities of distributed applied couples along the deformed rod. Let
the influence of that part of the rod beyond a section at distance s from one end upon
the remaining part be resolved into a force vector F = Tu + Qn + Rb and a moment
vector M = Hu + Ln + Mb, so that T is tension, Q and R shear forces, Ha twisting
moment, and L and j'vf bending moments.
(a) By considering static equilibrium of a section between s and s + ~s, and
proceeding to the limit as ~s ~ 0, show that the conditions
~; + p = O,
dM
--+uxF..Lm=O
ds '
must be satisfied.
(b) Show that these two vector equations imply the following six equations of
equilibrium :
dT
ds -
1
¡¡ Q + Pu = 0,
dQ
ds + P1 T - -:r1 R + Pn = 0, dR
ds
+ _1 Q
1: + _
Pb -
° ,
dH 1 dL I 1
-ds - - p
L + mu
=0, ds + ¡¡H - -:cM - R + mn = O,
dM 1
-d
s + -L
1:
+ Q + m. = O.
[Notice that this set comprises six equations in the eight unknown quantities T, Q, R,
H, L, M, p, and 1:. For an e/aslic rod, these equations are conventionally supplemented
by tbe two equations
H = B(1--
1:
1-),
1:0
328 Vector Analysis
where Po and 'o

are the radii of curvature and torsion of the undeformed rod, and
A and B are bending and torsional stiffnesses associated with rhe material and cross
section of the rod.]
(e) Suppose that the rod is originally straight and untwisted, so that l / po and
1/'0 = 0, and is subjected only to a transverse load distribution of intensity Pn (and to
end constraints). If, in addition, small deformations are assumed, so that 1/ p and 1h
are assumed to be negligible in the equilibrium equations for the deformed rod, show
that the equations of part (b) are satisfied by T = R = H = L = if M and Q satisfy °
the equarions
~~ + Pn
dM
= 0, ds + Q = O.
[With the supplementary equation M = El/p, where El is the bending stiffness, these
are the basic equations of the elementary theory of small deflections of laterally loaded,
originally straight rods or beams. (See Seetion 5.3, where S -Q and where s is =
replaced by its projeetion x along the undeformed axis.)]
Section 6.7
44. The temperature at any point in spaee is given by
T = xy + yz + zx.
(a) Find the direetion eosines of the direction in which the temperature ehanges
most rapidly with distance from the point (1, 1, 1), and determine the maximum rate
of change.
(b) Find the derivative of T in the direetion of the vector 3i - 4k at the point
,. (1,1,1).
45. lf r and (j are polar coordinates in the xy plane, determine grad r and grad (j.
46. Determine the direetion eosines of grad rp(x. y, z) .
47. Determine a unit vector normal 10 the surface
X' - xyz + Z3 = 1
at the point 0,1,1).
Section 6.8
48. lf V = grad rp, where rp = xyz, determine div V and curl V.
49. Show Ihal

v . (xv) = xV • v +i.v
and V x (xv) = xV x v + i x v.
50. Prove that u x V . v = u • V x v.
SI. Prove tbal (u . V)rp = u • (Vrp).
52. (a) Determine the derivative of the funetion x, + y 2 + Z2 in the direction spe-
eified by the direetion eosines 1, m, and n.
(b) Determine the derivative of the vector x 2 i -T- y2j + z 2k in the direetion spe-
eified by the direction cosines 1, m, and n .
Problems 329
53. Pro ve that V x F is nor necessarily perpendicular ro F, by gi v ing a suitable exam-

pIe .
Seclion 6.9
54. Prove t hat V • (V ({J) = (V • V)({J == V '({J.

55. (a) Show that
V(v • v) = 2v • Vv + 2v x (V x v).
and deduce that the relation
v • Vv = J:Vv l
is true when V x v = O or, more generally, when V x vis parallel to v.
(b) Verify the general result of part (a) when v = xyi + y'j.
56. Show that V 2 u = O if V x u = V({J and V • u = O, where ({J is a scalar function.
57. If v = ({J,V({J2' prove that V x v is perpendicular to v.
58. If there exists a family of surfaces ({J(x, y, z) = constant such that v(x, y, z) is
perpendicular at every point to the surface ({J = constant which passes through that
point, show that v • (V x v) = O. (Use the result of Problem 57.)
59. If u and v have zero divergence and w has zero cur!, show that
V • [(u X v) x w] = w • [(v· V)u - (u • V)vJ.
SeclioD 6.10, 6.11
60. For each of the following vector functions, determine whether the equation
V({J = F possesses a solution, and determine that solution if it exists:
(a) F = 2xyz 3 i - (X 2 Z 3 + 2y)j + 3x l yz l k,
eb) F = 2xyi + (x' + 2yz)j + (yl + 1)k.
61. For each vector function F defined in Problem 60, determine the value of the
integral Je F • dr from the origin to the point (J, 1, 1) along the curve specified by
the simultaneous equations y = x 2 , Z = x'-
62. For each vector function defined in Problem 60, determine the value of the integral
fe F • dr around the unit circle with center at the origin, in the xy plane.
63. (a) Jf P dx + Q dy + R dz = d({J in a region <R including a point (xo, Yo, zo), show
that the result of integrating d({J along straight line segments from (xo, Yo, zo) to
(x, Yo, zo) to (x, y, zo) to (x, y, z) is
({J(x, y, z) = f XP(I, Yo , zo)dt + fY

Xo Yo
Q(x, 1, zo) dI + f' ~o
R(x, y, 1) dI + constant,
if those segrnents remain in eR, where x is held constant in the second integral and both
x and y are held constant in the third one .
(b) Jf ({J is defined by the expression obtained in part (a), and if the first partial
derivati ves of P, Q, and R are continuous and such that the conditions (SSa, b, c) are
satisfied in eR, show that
a({J = P
ax '
330 Vector Analysis
so that drp = P dx -i- Q dy + R dz in eR. [Notice, for example, that then
rp.,(x, y, z) = P(x, Yo, zo) + s:, QxCx, r, zo) di + 1: Rx(x, y, 1) dI
and that Qx(x, 1, zo) = P,(x, 1, zo) and Rx(x, y, 1) = P,!x, y, 1).]
64. Show that the use of the formula obtained in Problem 63, with (xo, Yo, zo) =
(0,0, O), for the determination of rp such that drp = y2 dx + 2(xy + z) dy + 2y dz,
leads to the resul t
rp(x, y, z) = f: ° dI + J: 2xI dI + J: 2y dI +e = xy2 + 2yz + c.
(Compare the determination in the text, page 291.)

65. Evaluate the integral fe h ds, where s is arc length along e, in each of the follow-
ing cases:
(a) h = X2 + y2 + Z2; e is the helical arc x = cos (), y = sin (), z = ()/(2n)
between (O, 1, O) and (O, 1, 1).
(b) h = x + y + z; e is the circ1e X2 + y2 = 1, z = 1.
(c) h = xyz; e is the are of the twisted cubic x = 1, Y = 31 2 , Z = 61 J between
(0, 0, O) and (l, 3, 6).
Section 6.12
66. (a) If S is a portion of a surface specified by an equation of the form z = J(x, y),
r show that the unit normal vector n which points in the positive z direction from S is
given by
n =
-i(aJ/ax) - MJ;ay) + k .
-VI + (aflax)2 + (aJ/ay)2
(b) Deduce that, in this case, if F = Pi + Qj + Rk, there follows
r
where D is the projection of Son the xy plane, and where z is to be replaced by J(x, y)
I in the expressions for P, Q, and R.

67. Evaluate the surface integral of the vector F = xi + yj + zk over that portion of
the surface z = xy + 1 which covers the square <: x <: 1, <: y <: 1 in the xy° °
planeo
I
68. Evaluate the surface integral of the vector F = xi + yj + zk over that portion of
I the paraboloid z = X2 + y2 which is inside the cylinder X2 + y2 = 1.
69. Evaluate the surface integral of the vector F = xi + yj + zk over the c10sed sur-
face of the cube bounded by the planes x = ± 1, y = + 1, z = ± 1.
70. Evaluate the surface integral of F = yzi + xzj + xyk over the c10sed boundary of
the region bounded below by z = X2 + y2 and above by z = l.
71. Evaluate the integral f f s h du in each of the following cases:
(a) h = X2; S is the portion of the plane x + y + z = 1 inside the cylinder
X2 + y2 l.
°
=
(b) h = xy + -vz; S is the portio n of the cylinder z = x2/2 for which <: x <: 1
and ° <: y <: l.
ProbJems 331
Section 6.13
72. Evaluate tbe integral of Problem 69 by using the divergence theorem.
73. Determine the value of the surface integral SSs F • n da in each of the following
cases, by use of the divergence theorem:
(a) F = xi + yj + zk; S is the closed spherical surface X2 + y2 + Z2 = 1.
(b) F = xyi + xzj + (1 - z - yz)k; S is the closed surface composed of the
portion of the para bolo id z = I - X2 - y2 for which z >- O and the circular disk
XZ + y2 < 1, z = O.
(c) F = xyi + xzj + (1 - z - yz)k; S IS the portio n of the paraboloid
z = 1 - X2 - y2 for which z >- O.
(d) F = x 2 i - (1 + 2x)j + zk; S is the lateral surface of that portion of the
cylinder X2 + y2 = 1 for which O < z < 1.
74. If r is the position vector xi + yj + zk, show that
(a) ffs r· nda = 3V, (b) ffs xr· nda = 4VX,
where Vis the volume enclosed by S and x is the x coordinate of its center of gravity.
75. By using the physical argument ofSection 6.13, sbow tbat, if<R is the region inside
a closed surface S and outside a closed surface I:, then the divergence theorem takes
the form
JJJ(!\V.Vd7:=ffsv.nda+ff~v.nda,
where n points outward from <R along both S and I:, so that n points into the de1eted
region enclosed by I: in the last integral.
76. Analy/ical deriva/ion 01 the divergence /heorem. Suppose that the closed boundary
S of a region <R is cut by lines in tbe z direction in not more than two points. Denote
the upper part of S by S+ and the lower part by S-, and denote the projection of
either part on tbe xy plane by D.
(a) Show that
f f L~~ dx dy dz = f Iv Rs+ dx dy - f Iv Rs- dx dy,
where Rs+ and Rs- denote the value of R(x, y, z) on S + and on S -, respectively.
(b) Noticing that
dx dy = da cos y on S+
and dxdy = -da cos y on S-,
where cos y is the z direction cosine of the outward normal, deduce the relation
SS J(!\ ~~ dxdydz = ffs Rcos y da
in this case. [Corresponding results in the x and y directions estabJish the divergence
theorem (117) analytically when <R is a convex region. Proofs for other regions es-
sentially depend upon subdividing them into convex regions or taking limits of results
corresponding to such subdivisions. '
77. Show that the product da cos IX in the right-hand member of the divergence
theorem (117) can be replaced by ±dy dz, where the sign chosen at a point of S is to be
332 Vector Analysis
the sign of cos IX at that point, and similarly that we may write da cos = ±dx dz p
and da cos y = ±dx dy. (Notice that proper choice of each sign may diRá from one
part of S to another.)
78. (a) If V is a two-dimensional vector in tbe xy plane, show that
f In V • V dx dy = fe V • nds,
where D is a simply connected region in the xy plane with a simple closed boundary e,
and n is the unit outward normal vector along the curve C. (Apply the divergence
theorem to the vector V over a three-dimensional regioo which consists of the ioterior
of a right cylinder of unit height having the regioo D as its lower base.)
(b) If D is the region inside a simple closed curve e but outside a simple closed
interior curve r in the xy plane, show that tbere follows
JIn V • V dx dy = fe V • nds + frV • nds,

when n points outward from D aloog both e and r, so that n poiots into the deleted
region enclosed by r in the last integral.
79. Let CR be a regio n bouoded by a closed surface S.
(a) Establish the "gradient theorem,"
in rectaogular coordinates by applying the divergeoce theorem (117) to each com-

ponen!.
(b) lo a similar way, establish the "curl theorem,"
f H. V x F d. = § "dG' X F.
(c) Show that
]JI.,. (rpV • F + F· Vrp) d. = #s rpF. dG',
by applying (116) to the vector V = '7'F.
SectiOD 6.14
80. Show that Equations (123), (126), and (127) imply the relations
JL ('7'1 V2'7'2 - rp2 V2'7'1) dx dy = fe ('7'1 al¡2 - rp2 aa~¡1 ) ds,
JL [rpV2rp + (Vrp)2] dx dy = ferp~~dS,
JL V2rp dx dy = fe ~ ds,
where rp = rp (x, y) and V2 = (a 2/ax2) + (a 2/ ay2), and where D is the reglO o m the
xy plane bounded by C.
Problems 333
81. Verify the validity of the second equation of Problem 80 when rp = x and D is
the circular disk of radius a with center at the origino Use polar coordinates in the
right-hand member and notice that, on the boundary C, there follows
~~ = ~~ = cos 8.
82. Suppose that V2rp = °
everywhere in a region <R bounded by a closed surface S.
Establish the following results in that case:
(a) .#J~ da = 0, (b) .#S rp~da f H"

= (Vrp)2 d"t".
83. By writing rp, = rp and taking rp2 = V2rp in Equation (123), obtain the relation
ff fffi [rpV4rp - (V2rp)2] d"t" = .#J rpa ;:rp - (V2rp)~:J da.

84. Derive the following generalizations of Green's theorems (122) and (123) from the
divergence theorem, assuming that p is continuously differentiable:
(a) Jf t [rp,V • pVrp2 + pVrp, • Vrpz] d7: = .#sprp, a:"2 da,

(b) fft[rp,v 'pVrp2-rp Z V .pVrp.Jd7: = l
.#sp(rp,a:nz - rp2a n') da.
[Write V = prp,Vrpz (in place of rp,Vrp2) in the divergence theorem.]
Section 6.15
85. Suppose that rp satisfies Laplace's equation everywhere in a regio n <R bounded by
a closed surface S.
(a) If arp/an vanishes everywhere on S, deduce from Problem 82(b) that rp must
have a constant value in <R .
(b) If rp vanishes everywhere on S, deduce from Problem 82(b) that rp must vanish
everywhere in <R.
86. Suppose that rp satisfies the equation
4 _ a rp4 a4 rp
V rp = ax 4 + 2ax Z ay2
known as the bi-Laplacian equGlion, everywhere in a regio n <R bounded by a closed
surface S, and that both rp and arp/an vanish everywhere on S. Deduce from Problem
83 that rp then must also satisfy Laplace's equation everywhere in <R and hence, by
virtue of the result of Problem 85(b), must vanish everywhere in <R.
Section 6.16
87. If S is a closed surface in a region <R where the vector V is continuously differen-
tiable, show that
ir s n • V X V da = O.
88. Verify the truth of Stokes's theorem, as given in Equation (133), in the case when
V = yi . 2xj -4- zk, if C is the circle Xl + y2 = 1 (or x = cos 1, Y = sin r) in the xy
plane, and S is the plane area bounded by C.
334 Vector Analysis
89. U se Stokes's theorem to determine the value of the integral f f s n • V x V da over

the part of the unit sphere X2 + y' + Z2 = I above the xy plane , when V = yi.
90. Use Stokes's theorem to determine the value of the integral
t[O +y)zdx+ ( 1 +z)xdy+(1 +x)ydz ]
with eaeh of the following definitions of C:

(a) The eircle x = eos e,
y = s in e,
z = 1, oriented in th e direetion of inereas-
iog e.
( b) The triangle with vertiees at P , (l, 0 , O), P 2 (0, 1, O), and P) (O, 0 , 1), oriented
from P, to p •.
(e) A elosed curve io the plane x - 2y -+- z = 1.
91. Show t hat the formulas
fe Xdy = -feYdx=A,
and infinitely many others, can be use d in place of Equation (l40) for the purpose of
determining the plane area bounded by a simple closed curve.
92. Show that
C
a)1fe x , dy = -fexy dy = 1fe(x' dy - xy dx) = Ax,
t fe
( b) x) dy = -fex2y dy = t fe (x) dy - x2y dx) = I y,
where A is the area in the xy plane bouoded by a simple closed curve C, (x , y) is its
eenter of grav ity , and I y its moment of ioert ia about the y axis .
93. For any si mple closed eireuit C bounding an opeo surface S io a region <R, the
eleetrie intensity vector E and tbe m agnetie intensity vector H sat isfy the relations
i
Je
E. dr = -(1.~
al f f. s
H· deJ , fe H • dr = P :1 f LE. deJ,
where (1. and pare eertaio eoos taot s. Use Stokes's theorem to transform these rela-
tions to tbe form
f fJv x E + (1.aa~)· deJ = 0, fL (V x H - p~~) . deJ = O.
From the arbitrarines s of S, deduce Max well's equalions in the form
V x E = -
aH
(1. - ,
al
94. In a regioo free of electrie and magnetic eharges, it is true that V • E = and °
V . H = O. By eliminatiog E aod H s uecessively between the equations of Problem 93,
and using this faet, deduce that E and H both satisfy the veclor wave equalion
V 2V = Cl.P~:~
in sueh a regioo.
Problems 335
95. Let S be an open surface bounded by a simple closed curve C.

(a) Establish the relation
in rectangular coordina tes by applying Stokes's theorem (138) to each componen!.

(b) Show tha t
te rp,Vrpz' dr = -te rpoVrp¡ • dr,
by applying (133) to the vector V = V(rp¡rpz).

(c) Show that
Jt (Vrp x+ x
F rpV F) • dG = te rpF • dr,
by applying (133) to the vector V = rpF.
Section 6.17
96. Ellip/ical cylindrical coordina/es may be defined by the equations
x = a cosh u cos v, y = a sinh u sin v, z = z,
° °
where 11 :> and -< v < 2n.
(a) Show that this system of coordinates (11, v, z) is orthogonal [by verifying that
Equation (149) is satisfied].
(b) Show that in the xy plane a curve u = constant is an ellipse with semi-axes
a cosh 11, in the x direction, and a sinh u, in the y direction; also that a curve v = con-
stant is half 01 one branch of an hyperbola with sem1-axes a cos v and a sin v. In par-
ticular, show that the locus u = °
degenerates into the segment (-a, a) of the x axis,
whiJe the Joci v = ° and v = n are respectively the positive aod negative exteriors of
this segment; also that the loci v = n/2 and v = 3n/2 are respectively the positive
and negative portions of the y axis. Sketch and label in a single diagram the curves
u = 0, 1 and v = 0, n/4, n/2, 3n/4, 5n/4, 3n/2, and 7n/4.
97. For the coordina tes of Problem 96 derive the relations analogous to those of
Equations (l68b-e) for circular cylindrical coordinates. In particular, verify that
hu = h" = a,-/cosh 2 u - coso v,

Ul =
i sioh u cos v + j cosh u sin v
,
,-/cosh 2 U - cos 2 V
U2 = - i cosh u sio v
--~~~~~~~~~~~~
+ j sinh u cos v
,-/cosh
2 U - 2 V cos
336 Vector AnalYSis
Show also that for large vallles of u there follows

U¡ ......., ieos 1) + jsin L', U2'"'" -isinv+jcosv,
and U1 cos 'v - U 2. sin v "'-' i u L sin 'v + U2 CQS v ,...., j.
98. Parabolic cylindrical coordinares may be defined by the equations

x = :l:{u 2 _ ,.2 l, y = uv, z = z,
where - co - O.
Ca) Show that this system is orthogonal.
(b) Show that in the xy plane a curve v = eonstant is a parabola symmetrieal
about the x axis and opening to the right, while a curve u = eonstant is one half of
a similar parabola opening to the left. In particular, show that the loeus v = O is the
positive x axis, while the loeus u = O is the negative x axis, and that the positive y axis
is given by u = v and the negative y axis by u = - l: . Sketch and label in a single
diagram the curves u = O, ± J and v = O, l.
Ce) Perform the ealculations oeeessary to show that the Laplaeian is of the forrn
2 2
V2f = 1 (a f + a2f) + a f .
u + v au
2 2 2 av 2 a;¡;2
99. Parab%idal coordina/es eorrespond lO a system in which the plane configuralion
of Problem 98 is rotated about the axis of symmetry of the two seIs of parabolas. The
axis of rotation is then conventionally taken as the z axis.
Ca) In Problem 98 replace x b y z and y by"; X2 + y2, the distance from the Z axis;
then write y = x tan B, so that B is the cireumferential angle, and show that the
coordina te transforma tion beeomes
x = uv eos B, y = uv sin B, Z = :!-(u 2 - v 2 ),
where now u >- O and v :> O.
(b) Obtain the Laplacian in the fonn
V'f -
-
[.l. ~ (u al)
u' + v' u au
1
au
+ .l. ~ (val)]
v av av
+ u'v
_1 2
2
a f
aB2
100. In a rranslarion and ro/arion ofaxes, in which the origio in the new x'y'z' plane
is taken at the point (a, b, c) io the xyz plane, and in which the directions of the x',
y', and ;¡; ' axes are speeified by the direetion cosines (1" m" n,), (/" m" n,), and
(/" mJ, nJ) relative to the original axes, the traosformation of coordinates is of the
form
x = a + I,x' + I,y ' + IJz ' ,
y = b + m,x' + m,y' + mJz' ,
z = e + I1¡X ' -1- 1l2Y' + nJz' .
Show that Laplace's equation is of the form
a 2f a 2f a'f
ax" + ay' ! + az' z = O
in terms of the new variables.
101. Suppose that the coordinates u, and tlz are rela ted to x aod y by an equation of
the form
x + iy = F (u, + iuo),
Problems 337
where i2 = - 1, so that x is the real part of F (u, + iuz) and y tbe imaginary part, and
that "" uz, and z are ehosen as eurvilinear eoordinates in spaee.
(a) Show that , if F is a differentiable funetion of the argument ", + there iU2'
follows
ax + .ay F '( U j +.lU2,) ax + . ay
-a = 1'F ' ( " , + /tI
')
au, au,
- 1- =
U2 a 1-
U2
z,
where a prime denotes differentiation witb respeet to the complete argument u, + jU2,
and henee that
~
/(~)Z +
au, (aa",y )2 =
~
/(aX)2
a"z + (ay)'
auz = IF(u, + i"2 ) 1.
(b) Deduce al so that

ax + i ay = -i ax + ay ,
o,au, au, all z au,
and henee, by equating real and imaginary parts, that
ax ay ay ax
all, = aU2' au, = - a1l 2'
(e) With tbe notation of Seetion 6.17, deduce that the veetors U" U" and U) are
mutually ortbogonaI, and that
h, = h2= 1F(lI, + ;U2) 1, h) = 1.
Henee sbow that Laplaee's equation is of tbe fonn
a2! + a2! + h2 a2! = O

aUT aU! az 2 '
where h = 1 F(u, + jU2)!.
102. (a) If x and y are related to u I and u, by the equation

x + iy = ±(u, + iU2)2,
so that x = ±(UT - uD and y = U,U2, use the result ofProblem 101 to obtain Laplaee's
equation in U,U2Z eoordinates in the fonn
a2!z + -a
a'! + ('" , + uz )aaZ!' -_ O. 2
a U 1 U2
2
Z
(b) Obtain the same result by using the formulas of Seetion 6.17. (Se e al so Problem
98.)
Section 6.18
103. If u, and Us are the unit tangent veetors in the r and {J direetions, in circular
eylindrieal coordina tes, show that
i = u, ces {J - Us sin {J , j = u, sin {J + Us cos {J.
104. If r, (J, and z are circular e yli ndrical coordina tes, evaluate the following quantities:
(a) V{J, (b) Vr" , (e) V x us,
(d) V· [r"- '(u,sinn{J + uscosn{J») , (e) V2 ( r 2 cos{J),
n
(f) V'(r cos n{J).
338 Vector Analysis
105. Ir u" u.' and u. are the unit veetors tangent to the eoordinates curves in spherieal
coordina te s, show that
(u, sin rp + u. eos rp) eos B- u. sin B,
i (u, sin rp + u. eo> rp) sin B+ u. eos B,
k = u, eos rp - u. sin rp.
106. Ir r , rp , and B are spherieal eoordinates, evaluate the following quantities :
(a) V rp, (b) VB,
(e ) V • [u , eot rp - 2u.], (d) V2[(r + },)cosrp]
107. Show that the unit tangent veetors in spherieal eoordinates satisfy the following
relations :
au, au,
aB = u.
.
Sin rp,
au. _
aB - u. eos rp,
arp -
au.
aB =
.
- u, Sin rp - u. cos qJ.
108. By writing the position vector in the form r = ru" and using the results of Prob-
lem 107, obtain expressions for components of acceleration along the coordinatecurves
in s pherical coord inates a s follow s :
a, = f - rrp 2 - rO' sin 2 rp,
a. = 2frp + r;p - r0 2 sin rp cos rp,
a. = 2fO sin rp + rÓ sin rp + 2rrpO cos rp.
109. Prove that
V2[r'P" (cos rp)] = 0,
where r and rp are spherical coordinates and P, is the Legendre polynomial of order n.
[See Equation (169) of Chapter 4 .]
110. (a) Evaluate the surface integral
f.Is F· da,
where F = xi - yj + zk and where S is the lateral surface of the cylinder X2 + y2 - 1
between the planes z = °
and z = 1, using right circular cylindrical coordinates.
(b) Check the result by use of the divergence theorem.
111. (a) Evaluate the surface integral of F = xi - yj + zk over the closed surface of
the sphere X2 + y2 + Z2 = 1, using spherical coordina tes .
(b) Check the result by use of the divergence theorem.
Section 6.19
112. Suppose tha! a flow of an ideal incompressible fluid is free of distributions of
sources and sink~, and of vortices, and that it takes place parallel to the xy plane.
Problems JJ9
(a) Show that the veJoeity potentiaJ rp(x, y) and the stream funetion 'I/(x, y) are
sueh that
and that rp and '1/ satisfy Laplaee's equation.

(b) Show that
'(X'Yl e,.,)
q¡ (x , y) = j (Vx dx + Vy dy), 'I/(X, y) =
f (- v, dx + V x dy),
where the line integrals are eaeh evaluated along an arbitrary path from a fixed point
to the variable point (x, y).
113. (a) Verify that the funetion q¡(x, y) = x' - 3xy2 is a veloeity potential funetion.
(b) Determine the veloeity vector V, and its magnitude V.
(e) Determine the stream funetion 'I/(x, y), $ubjeet to the eondition '1/(0, O) = O,
and obtain the equation of the streamlines.
(d) If the uniform density of the fluid is p, determine the rate of f10w aeross an
are joining the points (1, 1) and (2,2) (that is, through a eylindrieal surfaee of unit
height having this are as i ts base).
114. (a) If elliptieal eylindrieal eoordinates (Problems 96 and 97) are used to deseribe
the flow of an ideal fluid parallel to the xy plane and independent of Z, show that the
velocity potential satisfies the equation
a2rp a q¡_ 2
-a
II + a 2 O. V
2 -
(b) Show that the corresponding stream funetion '1/ satisfies an equation of the
same form and that '1/ may be related to q¡ by the equation
dl{/ = _0P.
av du + 0P.
au dv .
115. (a) Show that
q¡ = e cosh II sin v
is a permissible potential function in Problem 114 and that Ihe corresponding stream
function is then of the form
'1/ = - e sinh u eos v,
if an arbitrary additive constant is discarded.
(b) Show that the streamline '1/ = o consists of the combination of the loci u = O
and both v = n/2 and v = 3n/2, and henee deduce from continuity considerations
that neighboring streamlines follow near the y axis toward the x axis, thence around a
perpendicular barrier (plate) of breadth 2a baek toward the y axis and onward nearly
paralJel to the y axis .
(e) Show Ihat the Ilow velocity is given by
V = a.V COS h 2 eII - eos v

2 (U¡ sinh u sin v + U2 cosh II cos v),
and that at large distances from the barrier (u ---> 00) the velocity vector tends to the
340 Vector Analysis
constant value
V ~ ea (U1 sin v + U2 cos v ) _ ea j (u ~ ca).
Deduce that the potential

rp = a V o cosh 11 sin v
corresponds to a flow, around a plate of breadth 2a, which tends to a uniform flow ,
with velocity V o at right angles to the plate, at large distaoces from il.
(d) Show that at a point (u, v, z) the flow velocity is of magnitude
V = V o / cosh 2 U - si0 2 v .
..¡ cosh 2 U - cos 2 V
In particular, verify that at points on the plate (u = O) there follows
Ixl (Ixl < a),
V = V0.y' a 2 - x 2
while for points on the x axis outside the plate (v = 0, n)

_ Ixl (Ixl > a),
V - v0.y' X 2 - a2
and for points on the y axis (v = n /2, 3n /2)
V = Vo Iy l
.y'a 2 +y2
116. Suppose that a source-free and vortex-free flow of an iocompressible fluid pos-
sesses axial syrnmetry about the z axis, so that its velocity can be expressed in the form
V = V:k + V,u"
in terms of circular cyliodrical coordina tes.
(a) With (u l , u 2 , U3) = (z, r, (J), show that ht = h2 = 1 aod h3 = r, and that
arp
V, = ar'
where rp(z, r) satisfies the equation
l.. ~ (r arp) +
2
a rp = O.
r ar ar az 2
(b) Show that the stream function r¡¡ is deterrnined by the relation
dr¡¡ = r( - V, dz + V z dr)
aod verify that the right-hand member is indeed an exaet differential.
(e) Show that the rate of flow through the surfaee obtained by rotating about the
z axis an arc e joioing two points Po and PI in aplane (J = constant is giveo by
2np Xc dr¡¡ = 2np[r¡¡(P l ) - r¡¡(P o)].
117. (a) Verify that the function rp(z, r) = 2z 2 - r 2 is a potential funetion corres-
ponding to the type of flow considered in Problem 116.
(b) Determine the velocity vector V, and its magnitude V.
Problems 341
(c) Determine the stream function r¡;, and show that the streamlines are the curves
in the planes e = constant for which r 2 z = constan!.
(d) Show that the rate of f10w through lhe surface obtained by rotating about the
z axis an arc joining Po and P, in a plane e
= constant is given by 4np(rrz, - r5zo).
Section 6.20
118. If, in the developments of Section 6.20, a body force F per unit mass is assumed
to be active, show that pF must be added to the right-hand members of Equations
(185) and (188).
119. If a fluid is viscol/s, the force on a surface element da, due to internal friction, is
usually assumed to have a component in any direction equal to ¡.t times the product
of da and the derivative, normal to da, of the velocity component in that direction,
where ¡.t is a constant known as the coefficient of viscosity.
(a) Show that the viscous force in the x direction on an element da then is of the
magnitude
¡.to· V V x da,
and that the net viscous force, in the x direction, on the c10sed boundary S of a regíon
<R is of the magnitude
Hence deduce that the viscous force on an element do is equivalent to a body force
with component ¡.t V2 V x do in the x direction, and analogous components in the y
and z directions, and hence to a body-force vector ¡.t V2V do.
(b) Deduce that effects of viscosity tben may be taken into account by adding the
term ¡.t V2V 10 the right-hand members of (185) and (188).
120. Ir a body force F per unit mass is conservative, and hence can be written in the
form
F = -VU,
where U is a potential energy function, use the result of Problem 118 to show that
Equation (211) is replaced by the equation
V2 +
..!..
2 at + P +
arp U = f(t)
'
when F is present, but viscosity effects are neglected.
121. If the Mach number M is greater than unity, show that any expression of the
form
rp(x, y) = f(x + at) + g (x - at) (a = "";M2 - 1)
satisfies Equation (200), where f and g are any twice-differentiable functions.
122. Verify Ihat the expression
rp(x, t) = F[x - (U + Vso)t] + G[x - (U - Vso)t]
satisfies Equation (202), where F and G are any twice-differentiable functions.
7
Topics in Higher-Dhnensional Calculus
7.1. Partial Differentiation. Chain Rules. In this seetion we review and dis-
euss eertain notations and relations involving partial derivatives whieh will be
needed in the seque!.
The more general case may be illustrated here by eonsidering a funetion f
of three variables x, y, and z,
f = f(x, y, z). (1)
Ir yand z are held eonstant and only x is allowed to vary, the partial derivative
with respeet to x is denoted by af/ax and is defined as the limit
af(x, y, z) = lim f(x + ~x, y, z) - f(x, y, z). (2)
ax "X~O ~x
Similarly we define the funetions af/ay and af/az. In all cases two of the three
variables explieitly appearing in the definition of f are held eonstant, and f is
differentiated with respeet to the third variable. The total differential of f is
defined by the equation
df = af dx
ax
+ af dy
ay
+ af
az
dz, (3)
whether or not x, y, and z are independent of eaeh other, provided only that
the partial derivatives involved are eontinuous. Several types of dependenee
among x, y, and z are now eonsidered. In eaeh of the formulas to be obtained,
the eontinuity of all derivatives appearing in the right-hand member is to be
assumed.
(1) If x, y, and z are aU funetions of a single variable, say l, then the depen-
denl variable f may also be eonsidered as truly a funetion of the one independent
342
7.1. Partial Differentiation. Chain Rules 343
variable 1, and . we may conveniently speak of x, y, and z as inlerm.ediale vari-

ables. Since only one independent variable is present, df/dl has a meaning and
it can be shown, by appropriate limiting processes , that
df _ af dx I af dy + af dz (4)
dI - ax dI ... ay dI az dI
This result is formally obtained by dividing the expression for df by dI. We
notice that dI/dI is the sum of three terms, each of which represents the con-
tribution of the change in I through the corresponding change in one of the
inlermediale va riables.
(2) More general/y, the intermediate variables x, y, and z may be functions
of two (or more) independent variables, say s and l . Then if we consider f as a
function of s and 1, we may investigate the parlial derivative of f with respect
to I when s is held constant. If we denote this function by the notation (dI/dI)"
then Equation (4) must be modified to read
af) _ af (aX) I af (ay ) af(aZ) (5a)

( at , - ax at , -¡- ay Oí , + az al "
The derivatives with respect to I are nOW not total but partia!. The subscript
indicates the variable held constant. With this COnvention we should perhaps
also write (af/ax)y., in place of af/ax, and so on, in (4) and (5a). However, we
wil! follow Ihe convenlion Ihal af/ax, wilhoul subscripls, indicales Ihe resull of
differenlialing f with respecl 10 Ih e explicilly app earing variable x, holding al!
olher explicilly appearing variables (here y and z) conslanl. Frequently, we also
omit the subscript s in (5a) and write merely
(Sb)
if it is clear from the context that s and I are to be associated with each other
as the independent variables, with x, y , and z as the (explicit) intermediate
variables. Alternatively, we could write F(s, 1) for the result of replacing x, y,
and z by their equivalents in f(x, y, z), so that
f[x(s, 1), y(s, 1), z(s, 1)] = F(s, 1),
in accordance with which (5a, b) could be written in the form
(5c)
without any poss ible ambiguity. Whereas this is the most elegant way of pro-
ceeding in such situations, it is often inconvenient in practice to use two different
symbols (here f and F ) to represent the same physical or geometrical quantity.
In the preceding cases the independent variable I does not appear explicitly
in J, and its changes are reflected in f only through the intermediate changes
in x, y , and z. However, it may be convenient to take an explicitly appearing
variable as an independent variable . We again distinguish two cases.
344 Topics in Higher-Dimensional Calculus
(3) If we suppose that y and z are functions of x, then I is a function of

the one independent variable x, and y and z are intermediate. Also, identifying
I with x in Equation (4), we obtain
t!i.. _
al + al dy + al dz . (6)
dx-ax aydx azdx
The term al/a x in (6) is obtained, as before, by holding the other two explicit
variables (y and z) constant, and it represents the contribution of the explicit
variation of x. The other terms add the contributions of the intermediate
variations in y and z.
(4) If we suppose that x and y are independent but that z is a function of
both x and y, then I can be considered as depending upon x and y directly and
also intermediately through z. AIso, identifying t and s with x and y in (Sa),
we obtain
(7a)
Here the notation is rather treacherous. On the left-hand side of (7a) we think
of las being actually expressed in terms of x and y, the variable z having been
,1, replaced by its equivalent in terms of these variables. Then we imagine that y
is held constant in the x differentiation. On the right-hand side I is expressed
in its original form, in terms of x, y, and z. The first term on the right is again
calculated with the other explicit variables y and z held constant, and it represents
the contribution due to the explicit variation of x. The other term adds the
contribution of the only intermediate variable z. Since z depends only on x
and y, the last subscript may be omitted without confusion. However, the sub-
script on the left is clearly essential. Thus we may write Equation (7a) in the
form
(7b)
These formulas are useful when we deal with a functionl in abstract terms.
If lis given as a specific function and the dependencies are specifically stated,
such formulas usually are not needed. In illustration we consider the function
I(x, y, z) = X2 + xz + 2y2.
If we consider x, y, and z as functions of t, and possibly other independent
variables, we merely differentiate term by term and obtain
al
at
= 2.x ax
at
+ x az
at
+ z ax
at
+ 4y ay
at
= (2x + z) ax
at
+ 4y ay
at
+ x az.
at
This result is the same as that given by Equation (Sb). If we consider x as inde-
pendent and as sume that y and z are given by other equations as functions of
x, we again differentiate ter m by term and obtain
7.1. Partíal Dilferentiatíon. Chain Rules 345
df _ 2
-
+ xd-'~
dI + I _ 4 dy
dX - X x
y-dX
=
dx
+ z -+-
4y dy + X dz ,
2x
dx
in accordance with the result of using Equation (6). If z is given by another
equation in terms of X and y, and if x and y are independent, then we obtain
directly, holding y constant,
(%\ = 2x + z + X ~~,
in accordance with the result of using (7b). The term 2x + z is equivalent to
af/ax, where y and z are held constant. The term X az/ax corrects for the fact
that he re z cannot actua//y be held constant but must vary also with x.
As a further example, suppose that we ha ve the relation
X2 + xz + 2y2 = O.
We again denote the function of x, y, and z on the left by f,
f = X2 + xz + 2y2.
We may consider the given relation as determining, say, z in terms of X and y,
·. both of which may then be taken as independent. Then, holding y constant
and differentiating with respect to x, we obtain
az
2x +z +X ax = O.
Hence we must have
2x +z
x
We may al so arrive at this result by considering the left-hand member of
the given equation as a function of the independent variables x and y and the
intermediate variable z. Then, since f constantly satisfies the equation
f(x,y, z) = O, (8)
the partial deriva ti ve of f with respect to either independent variable must
vanish. But Equation (7b) then gives
af) _ af
( ax y - ax
+ af az _ O
az ax - .
This equation states that the contributions of the explicit variation of x and the
intermediate variation of z must cancel. In order that this be so, we must then
have
az af/ax (9)
ax = - af/az'
in accordance with the result obtained above.t
tEquation (9) tends to illustrate the dangers associated with routine symbolic manipulations,
since a formal (but unjustified) inversion and cancellation in the right-hand member mighl
suggest that the prefixed sign is incorrecl.
346 Topics in Higher-Dimensional Ca\culus
Partial derivatives of higher order, of a function f(x, y, z), are calculated

by successive differentiation. Thus we write, for example,
alf a af
ay ax = ayax'
and so fortb. In tbis connection, we review the important fact that the crossed
partial derivatives are equal,
alf ay
ayax=axay' (lO)
that is, the order of differentiation is irnrnaterial, if the derivat/ves involved are
continuous. This statement is true for derivatives of any order if they are con-
tinuous, but it may not be true otherwise.
For the purpose of obtaining analytical formulas for bigher-order partial
derivatives , it is often convenient to use operational notation (see also Problems
3, 4, and 5). As an example, we suppose that f is a function of x, y, and z, with
x and y independent and z a function of x and y, so that Equation (7) applies.
Here, in order to avoid the complexities involved in unambiguously generalizing
the notation (affax), to higher-order derivatives, it is particularly desirable to
introduce the special notation
f[x, y, z(x, y)] = F(x, y), ( 11)
so that F(x, y) is the result of replacing z by its equivalent in terms of x and y
in the expression for f(x, y, z). The formulas (7a, b) then can be rewritten in the
form
aF _ af
ax - ax az ax
+
afaz = (~+ az .E....)f
ax ax az ' (12)
from which there follows, by iteration,
alF (a az a)(af az af)
ax + ax az ax --, ax az
I
axl =
a 2f az a 2f ) az (a 2f az a2f) [( a az a) aZ] af
(
- ax2 + ax ax az + ax ax az + ax di2 + (Ix + Tx Tz ax az
(13)
Here, as before, the convention is tbat, in eacb of tbe indicated partial deriva-
tives, aH variables explicitly involved in the function being differentiated are
held constant except the one with respect to wbich the differentiation is being
effected. Thus, since F = F(x, y) and z = z(x, y), tbere foHows
~: = (~) y and ~: = (~:t

On the other hand, since f = ¡(x, y, z), tbere follows
af = (af ) .
ax - ax Y.'
7.2. lmplicit Functions. Jacobian Determinants 347
7.2. Implicit Functioos. Jacobiao Determioants. An equation of the form

I(x, y, z, ... ) = O, (14)
involving any finite number of variables, where I possesses continuous first
partíal derivatives, can be considered as determining one of the variables, say z,
as a function of the reroaining variables, say
z = rp(x, y, . .. ), (15)
in sorne region about any point where Equation (14) is satisfied and where the
partial derivative of Iwith respect to that variable exists and is not zero,
~*o. (16)
In such a case we say that Equation (14) defines z as an implicit function of the
other variables, in the neighborhood of that point. If we consider ali the other
variables as independent, we can determine the partial derivative of z with
respect to any one of thero, without solving explicit/y for z, by differentiating
(14) partially with respect to that variable. Thus, to determine az¡ax, we obtain
froro (14)
al + al az _ o (17)
dx diTx-'
the denominator differing from zero by virtue of Equation (16).
If n + k variables are related by n equations, it is usually possible to con-
sider n of the variables as functions of tbe remaining k variables. However, trus
is not a/ways possible.
As an illustration, suppose that x, y, u, and vare related by two equations
of the form
I(x, y, u, v) = O,
(l8)
g(x, y, u, v) = O.
If these equations determine u and v as differentiable functions of the variables
x and y, we may differentiate the system with respect to x and y, considering
these two variables to be independent, and so obtain the four relations
al + al au + al av _ o
ax "Tu ax dv ax - ,
(19)
ag + agau +agav_ o
Tx au ax avax - ,
al
Ty
+ g¡ au + al av
au Oy av ay
_
-
o
,
(20)
ag + ag au + ag av _ o
ay Tu ay av ay - .
For brevity, we use the conventional subscript notation for partial derivatives
so that, for example, L/x is written for au¡ax. Then, if (19) is solved for au¡ax
348 Topics in Higher-Olmensional Calculus
and av/ax and (20) is solved for au/ay and av/ay, the expressions for these
partial deri vatives can be written in terms of determinants as follows:
~
fx f,
g,
u x = - gx Vx = _ gu gx ,
Ifu f, I Ifu f, I
gu g, gu g,
(21)
Ify f, I fu fy I
_ gy gu = - gu
gy
uy =
ft
J
Vy
~
gu g,
Ifu
gu g,
It must be assumed, however, that the common denominator in (21) does not
vanish, that is, that
af
av
(22)
ag
av
I Unless (22) is satisfied, the desired partial derivatives cannot exist uniquely,
so that u and v cannot be differentiable functions of x and y_ However, if (18)
and (22) are satisfied at a point, and if the first partial derivatives of f and g
are continuous at and near that point, it can be shown that Equations (18)
determine u and v as implicit functions of x and y in sorne region inc1uding that
point, with partial derivatives given by (21).
The determinant in (22) is known as the Jacobian off and g witb respecI lo
ti and v, and the notation
af af
a(f, g) _ Tu av afag afag
(23)
a(u, v) = ag ag = Tu av - l1v au
au av
is frequently used. In a similar way we write, for example,
af af af
au av aw
a(f, g, h) ag ag ag
(24)
a(u, v, w) - au av aw
ah ah ah
au JV aw
and proceed in the same way to define the Jacobian of any n functions with
respect to n variables.
In this notation, if
a(f, g) =1= O (25)
a(u, v) ,
7.2. Implicit FunCtiODS. Jacobian Determinants 349
the first equation of (21) becomes, for example,

au, g)
au a(x, v)
dx= -au, g)' (26)
a(u, v)
More generally, if n +
k variables are related by n equations of the form
I I = 0,/2 = 0 , ... ,l. = 0, where the functions Ik each have continuous first
partial derivatives, then any set of n variables may be considered as functions
of the remaining k variables, in some neighborhood of a point where the n
equations are satisfied, if the Jacobian 01 the j ' s with respect lo the n dependent
variables is not zero at that point.
Example. We consider the system

x +y + z = 0,
(27)
X2 + y2 + Z2 + 2xz - 1 = O.
To investigate whether x and y can be considered as functions of z, we denote the left-
hand members by f and g, respectively, and ca1culate the Jacobian
a(f. g) = I 1 1 I= -2(x + z - y). (28)

a(x, y) 2x + 2z 2y
Thus, except on the surface x + z - y = 0, x and y can be considered as functions of
z. That is, z can be taken as the independent variable. When y = x + z, the equations
°
become 2(x + z) = and 2(x + Z)2 = 1 and are hence incompatible. To investigate
whether x and z can be taken as the dependent variables, we calculate the Jacobian
a(f. g) I 1
1 -O I (29)
a(x. z) = 2x + 2z 2x+2z - .
Since this determinant is identically zero, we see that x and z cannot be taken as the
dependent variables. lt is readily verified directly that the system (27) cannot be solved
for x and z in terrns of y . This situation follows from the fact that both equations
involve only y and the combination x + z, and hence canno! be solved for x and z
separately. •
By direct expansion we can verify tbat if u and vare functions of r and s,

and also r and s are functions of x and y, then tbe relevant Jacobians satisfy
the equation
a(u, v) a(r, s) a(u, v)
(30)
a(r, s) a(x, y) = a(x, y)'
As a special case of this result, we find tbat if u and vare functions of x and y,
and conversely, tben
a(u, v) a(x, y ) _ 1 (31)
a(x, y) a(u, v) - .
Analogous identities bold for Jacobian s of any order. Thus Jacobians bebave
in certain ways like derivatives, as is sugges ted by the notation used .
The Jacobian notation is useful in many other applications. In tbis connec_

tion, the identity
au au au
ax ay az
av av av _ a(u, v, w)
(Vu) . (V v) x (Vw) =
ax
ay az - a(x,y, z) (32)
aw aw aw
ax ay az
may be noted. From this relation we conclude that if a(u, v, w)/a(x, y, z) = O at
a point, then the surfaces u = el' v = e 2 , w = e 3 which pass through this point
have coplanar normals.
7.3. Functional Dependence. The general solutions of certain types of partial

differential equations, to be dealt with (n Chapter 8, are of the form
z = f[u¡(x, y)] + g[u 2(x, y)], (33)
where u ¡ and u 2 are independent particular solutions and f and g are arbitrary
functions of these expressions. If U 2 were a function of u l' both terms would
then be functions merely of u 1> and (33) would not be the required general
solution. Thus in the expression
z =f(x + y) + g(x 2 + 2xy -;- y2 + 1) =f(u¡) + g(u 2)
we have U 2 = uI + 1 and hence both terms are functions of the same combina-
tion u ¡ = x + y. It is thus important in more involved cases to have a criterion
for determining whether one function u¡(x, y) is a function of a second function
u 2 (x, y). Jf such a functional relationsbip does exist, the two functions are said
to be fLlnetionally dependen!.
Suppose that such a relationsbip does exist, so that, for sorne F, not identi-
eal!y zero, it is true tha t
F(u p u 2) O, (34)
where u¡ and l/2 are functions of the independent variables x and y. Then, if
we calculate the partial derivatives of (34), we obtain
aF au¡ + aF aU 2 = O
au¡ ax aU 2 ax - ,
(35)
aFau¡..L aFau 2 =0
au ¡ ay I au, ay - .
These two linear equations in the quantities aF/aLl¡ and aF/au 2 can have non-
trivial solutions only if the deterrninant of their coefficients vanishes. But this
deterrninant is precisely the Jacobian of LI¡ and Ll 2 with respect to x and y. Hence
if LI Jx, y) and u,(x, y) are funetionally dependent in a region, their Jaeobian must
vanish identieally in that region,
a(u¡, u 2) = O (36)
a(x,y) - .
7.3. Functional Dependence 351
Conversely, it can be shown that if the partial derivatives are continuous and
if the Jacobian vanishes identically in a region, the two functions are function-
ally dependent in that region.
Completely analogous statements apply in the more general case of n func-
tions of n variables. Thus if the functions ul> u l , . . . , Un' are functions of n
variables, and if their partial derivatives are continuous, then the functions are
functionally dependent, that is, there exists a non trivial F such that
F(u¡, Uz , ... , un) O
in a region, if and only if the Jacobian of these functions with respect to the n
variables is identically zero in that region.
Example. For the t\Vo linear functions

u¡ = ax + by + e,
Uz = dx + ey + /,
the Jacobian is
a(u¡, "z) = lad ble =
a(x, y)
ae _ bd.
Thus U¡ and IIZ are functionally independent unless ae = bd. If ae = bd, the functional
relationship
ell¡ - buz = ec - bf
exists between U¡ and Uz.
•
When there are fewer functions than variables, several relations of form
(36) must hold. For example, in the case of two functions of three variables,
U ¡ (x, y, z) and uz(x, y, z), the assumption
F(u ¡, u z) - O (37)
leads to the three equations
aF au¡ + aF auz = O
au¡ ax aU 2 ax - ,
aF au¡ + aF auz = O
au¡ ay auz ay - ,
aF au z = O
aF au¡ +
au¡ az
au z az - ,
from which, by considering the equations in pairs, we may deduce that the three
conditions
a(u¡, u 2 ) = O a(u¡, uz) = O (38)
a(x,y) - , a(y, z) - ,
must be satisfied, by the same argument used in deriving (36). Conversely, if
the partial derivatives are continuous and if the three conditions of (38) are
satisfied identically in a region, then the functions u¡(x, y, z) and u 2 (x, y, z) are
functionally dependent in that region.
-~
3 =>- Topies in Higher-Dimensiooal Calculus
The ge nera liza t io n to m functions of Il variables, when m < n, is straight-

forward . When m > /1 , the m functions are a/ways functionally dependent.
7.4. Jacobians and Curvilinear Coordinates. Change oC Variables in Integrals.

Ir the equations
x = x(u 1 , u" u 3 ), y = y(u" u z , u 3), z = z(u" u 2 , u 3) (39)
are interpreted as defining curvilinear coordinates u" u 2 • and U 3 m space, and
if we write
(k = 1, 2, 3), (40)
then, as has been shown in Section 6.17, the vectors U" U z , and U 3 are vectors
tangent to the three coordinate curves at any point, with lengths given by
ds,/du" dSzldu z , and ds 3/du 3• where s" .'z, and S3 represent arc length along the
coordinate curves. Then [compare Equa.tion (151), Section 6.17] the element of
volume in the new coordinate system, whether or not the system is orthogonal,
is seen to be given by
d-r = (U, ,U z X U 3) du, duz du 3,
~ ,
k if the coordinates are so ordered that the right-hand member is positive. But
from (40) we obtain
l. ax ay az
au, au; au¡
ax ay az a(x. y. z)
U"U Z xU 3 - au q¡¡; - a(u" u z, u 3)'
z au z
ax ay az
Tu; Tu; aU 3
since the determinant is unchanged ir rows and columns are interchanged. Thus
we may write
d -r =
I
a(a(x, y, z) )
U , U , U
1 2 J
Id
U¡ d Uz d u 3· (41)
It is seen that the requirement that U, ' U z X U, be different from zero is

necessary in order that (39) be solvable for u" u z ' and u 3' In the special case of
orthogonal coordinates, U, ' U 2 X U 3 has the value h,h 2 h 3 • with the notation
of Section 6.17.
Accordingly, we have the change-of-variables formula
ff L w(x,y, z)dxdydz = f f L. W(u,. uz.u3)lat;;~'::.~,) ldU, dU z du 3,

(42)
where
W(u" u 2 , u,) = w[x(u" UZ , u 3),y(u" Uz , U3)' z(u" uz , U3)]
and where CR* is the u,U Z u 3 regio n into which (39) transforms the xyz region CR.
7.4. Jacobians aod Curvilinear Coordina tes. Change oC Variables in Integrals 353
Here it is assumed that the Jacobian a(x, y, z)/a(L1" L1o, L1,) is continuous and
nonzero in CR *.
In a similar way, the equations
x = x(L1" L1z), y = y(u" u z ) (43)
can be interpreted as defining curvilinear coordinates L1, and L1z in the xy plane.
The vectors
V ¡ -_ .Iax ay
J -a'
+. v z = l' au
ax ay
J au (44) +.
a-
u, u, z z
are then tangent to the coordinate curves, with lengths ds¡/du, and dSzlduz.
The vector element of plane area is then given by
i j k
ax ay
dA = (V, x V z ) du, duz -
au, au,
o du, duz,
ax ay
au z au z
o
and this relation gives the result
dA = IdAI = la~S~:~~)ldu, duz· (45)

Accordingly, we ha ve
fL w(x, y) dx dy = f L. W(u" u z ) I a~S~: ~~) I du, duz, (46)
where W(u" Hz) = W[X(L1" Hz), Y(H" u z )] and where (43) transforms D into D*,
if a(x, y)¡a(u" u z ) is continuous and nonzero in D*.
Example. We consider the coordinates u and rp defined by the equations

x = au COS qJ, = bu sin rp
y (u ::> O, O -< rp < 27t). (47)
The curves u = constant are the ellipses
Xz ~-1
(au)Z + (bu)Z -
with semi-axes au and bu, whereas a curve rp = constant is the portion of the straight
line y = (b/a)x tan rp in the quadrant determined by rp. The element of area in urp
coordina tes is given by (45),
dA = Ia(x, y) Idu drp
a(u,rp)
= abu du drp. (48)
Thus, for example, any integral of the form f fD w(x, y) dx dy, where the integration is
carried out over the interior of the ellipse
XZ
aZ
+ yZ _ 1
bZ -
corresponding to u = 1, can be written in the form
St w(x, y) dx dy = ab
ln
Sa So' w(au cos rp, bu sin rp) u du drp. (49)
354 Topies in Higher-Dimensional Calculus
In particular, to calculate the moment of inertia Ix of the afea about the x axis, we
write
w( x, y ) = y2 = b 2u 2 sin' q;,
and Equation (49) gives
1'" l' •
3
Ix = ab ' o o u' sin 2 q; du dq; = n~b .
Formulas permitting the determination of area and arc length on surfaces

are derived in Problems 25 and 26.
7-5_ Taylor Series. Functions of two or more variables often can be expand-
ed in power series which generalize the familiar one-dimensional expansions.
The more general situation may be illustrated here by a consideration of the
two-variable case.
For this purpose, we begin by defining a function F(t), such that
/(x + ht, Y + kt) = F(t), (50)
where x, y, h, and k are temporarily to be held fixed and, in any case, are ro be
independent of t. Then, if F(t) has a continuous Nth derivative in sorne interval
:'l . about t = O, we may write
(51)
for sorne value of, between

Chapter 4.
° and t, by virtue of Equations (10) and (11) of
Now, since
~F(t) = ha/ex + ht,y + kt) + ka/ex + ht, Y + kt)
dl ax ay
= (h:'C + k¿)/(x + ht,y + kt),

there follows also
;;,F(t) = (h;x + k;y)"/(x + ht,y + kt) (n = 0, 1, ... ).
Hence we have the results
F'n,(O) = (h ;x + k ;y)"f(X, y) (52)
and F'N'(-r) = (h ;x + k :yr f(x + ,h, y + -rk). (53)
If we introduce Equations (51), (52), and (53) into (50), and specialize the
result by taking t = 1, we thus obtain the form
/(x+h,y+k) = ¿;-,l (a
N- 1
n~on.
h - + k -a)"/(x,y) +
ax ay RN , (54)
7.5. Taylor Series 355
where R N , the "remainder after N terms," is given by
RN = N'I (a a)N
h ax + k ay f(x . ,h,y • ,k) (O < , < 1), (55)
for sorne, between O and J.

For example, when N = 3 this result becomes
f(x -c h, y + k) = + [hfx(x, y) + kj,(x, y)]
f(x, y)
+ i,Wfxx(x,y) + 2hkfxy(x,y) + k'fyy(x, y)] + R J, (56)
where
It is seen that the contents of the brackets in RJ are evaluated at"some point on
the straight line segment between the point (x, y) and the point (x h, y k). + +
More generally, Equation (54) represents an expansion of f(x h, y k) + +
in powers of h and k through the (N - l)th, with an error term, and is one
form of the two-dimensional Taylor formula. Although the form given is perhaps
tbe most compact one, a form which more c10sely resembles the most familiar
one-dimensional form can be obtained from Equation (54) by first replacíng
(x, y) by (x o, Yo) and then replacing h and k by x - X o and y - Yo, respectívely.
When N = 2, for example, there then follows
f(x, y) = f(x o, Yo) + (x - xo)fx(x o, Yo) + (y - Yo)fix o, Yo) +R 2 (58)
with
R2 = i, [(x - XO)2!.,x(¿;' 17) + 2(x - xo)(y - Yo)JXY(¿;' 17)

+ (y - YoFfyY(¿;, 17)], (59)
where the poínt (¿;,17) is somewhere on the line segment joining the points
(x o, Yo) and (x, y).
When f(x, y) is sufficiently well bebaved, the remainder RN tends to zero
for sufficiently small values of the increments, yielding a (gene rally infinite)
power series of the form
f(x, y) = f(xo, Yo) + [(x - xo)!.,(xo, Yo) + (y - Yo)f/x o, Yo)]
+ d, [(x - x o)2fxx(xo, Yo) + 2(x - xo)(y - yo)Jx/xo, Yo)

+ (y - Yo)2fyy(xo, Yo)] + "', (60)
which converges when Ix - X o I and Iy - Yo I are sufficiently smal!. Witbin its
region of convergence, the series can be differentiated or integrated term by
term and the result will converge to the derivative or integral of f inside the
same region.
It can be pro ved that tbe expaas ion (60) is unique, in the sense that if an
expansio n of the form
f(x, y) = a o + b¡(x - xo) I b,(y - Yo) + c¡(x - xo)2 + .. "
which converges to f near (x o, Yo), can be obtained by any method, it is neces-
sarily the same as that defined by (60). For the elementary functions , alternative
methods which are preferable to the use of (60) are usuaIly evident.
7.6. Maxima aud Miuima. The developments of the preceding section are
helpful in studying maxima and minima of functions of several variables. We
again restrict attention here to the two-dimensional case.
If we write
6.f(x, y) = f(x + h, Y + k) - f(x, y) (61)
for the increment of f, corresponding to the increments h and k in x and y,
respectively, we say that f has a re/alive minimum at P(x o, Yo) if 6.f(x o, Yo) >- O
for al! s ufficiently small permissible increments h and k, and thatf has a re /alive
maximum at P if instead 6.f(x o , Yo) < O for aIl such increments in x and y.
If the point P is an interior point of a region in whichf, af/ ax, and af/ ay
exist, Equation (56) shows that a necessary condilion Ihal f assume a re/alive
maximum o/' a relalive minimum al (x o, Yo) is Ihal
fx =fy = O (62)
For, when h and k are sufficiently small, the sign of 6.f(x o, Yo) will be the same
as the sign of hf.(x o, Yo) + kfy(xo,yo) when this quantity is not zero , and clearly
the sign of this quantity will change as the signs of h and/or k change unless
Equatioo (62) holds.
Suppose now that the condition of (62) is satisfied at a certain point P.
Then, from (56), there follows
sig n [6.f(x o, Yo)] = sigo [h 2f ,,(xo, Yo) + 2hkf.y(xo, Yo) + k 2fyy(x o, Yo)] (63)
when h and k are sufficiently smaIl, unless the bracketed quantity on the right
is zero. That quantity is a quadratic expression in h and k, of the form Ah' +
2Bhk + Ck 2 • When the discriminant B2 - AC is posilive, and only in that case,
there will be two distinct values of the ratio k /h for which the expression is zero,
the expression having one sign for intermediate values of k / h and the opposite
sign for aIl other values. Hence a necessary condilion Ihal f ha ve eilher a relalive
maximum or a relalive mínimum al P(x o, Yo) is Ihal
Ó =fxxf,. - fA >- O al (xo, Yo)' (64)
1f Ó < O at a point P(x o, Yo) where Equation (62) is satisfied, then 6.f is
positive for some increments in x and y and negative for others, and the point
P is said to be a sadd/e poinl or a minimax.
If Ó = O at a point P(x o, Yo) , then the bracketed espression on the right in
(63) is a perfect square, of the form (och - fik) 2, and hence either is identically
7.7. Conslrainls and Lagrange Multipliers 357
zero or is zero along a line

~ = y - Yo IX
h x - Xo 7f
passing through P. Thus in this case the sign of l:!..f(x o' Yo) may not be com-
pletely determined near (x o, Yo) by the quadratic terms when (62) is satisfied,
and the terms involving higher-order partial derivatives must be considered.
If c5 > O at (x o, Yo), then clearly fxx and fyy must be either both positive or
both negative at that point. Since l:!..f(x o, Yo) is of constant sign in either case,
when h and k are sufficiently small, it follows from (63) that the former case
corresponds to a relative minimum (l:!.. >- O) and the latter to a relative maxi-
muro (l:!.. <: O).
Thus, in summary, iffx = O and fy = O al a point P, Ihen al Ihal poinl f has
(a) a re/alive maximum if fxx < O,fufyy > f}y al P,
(b) a re/ative minimum if fx. > O,fxxf" > f}y al P,
(e) a sadd/e point if fxxfyy < f}y al P.
H/hen fx .• f" = f}y at P, further investigalion is necessary.
Cases (a), (b), and (e) are illuslrated by the funetions 1 - X2 - y2, X2 + y2,
and xy, respeclively, at (O, O). For the funetions 1 - x 2y2, x 2y2, and x'y>
Ihe poinl (O, O) is exceptionaI, sinee for eaeh not onIy l. = fy = O bul aIso
Ixx = Ixy = /,y = O al the point. However, it is obvious, by inspection, that
these funclions have a maximum, a minimum, and a saddIe point, respectiveIy,
al Ihe origin o
Very frequently, in practice, the use of tbe preceding criteria is too involved
10 be feasible, and a direet study of the behavior of l:!..f(x o, Yo) for small h and k
may be necessary. Often physical or geometrical considerations make such
investigations unnecessary.
lt should be noted that a relative maximum or minimum may also be
attained at a point where fx and/ or fy fai/ 10 exist and that, when attention is
restricted to a region <R with a finite boundary, it may happen that an extreme
value is taken on at a boundary point , at whichfx and fy may or may not exist
and may or may not differ from zero. In order to locate an abso/ute maximum
or minimum (that is, the largest or smallest value taken on) in a finite region,
it is neeessary to explore all these possibilities.
7.7. Constraints and Lagrange Multipliers. Situations also may occur in

which a function f, to be maximized or minimized, depends upon variables
which are not independent, but are interrelated by one or more constraint
conditions. The more general situation may be illustrated by the problem of
maximizing or minimizing a function f(x, y, z),
f(x, y, z) = relative max or min, (65)
subject to two constraints of the form
g(x, y, z) = O, (66a)
h(x, y, z) = O, (66b)
where g and h are not functionally dependent, so that the constraints are neitber
equivalent nor incompatible. We suppose that tbe functions f, g, and h bave
first partial derivatives everywbere in a region which ineludes the desired point.
An obvious procedure consists of using Equations (66a, b) to eliminate two
of tbe variables from f, leading to a problem of maximizing or minimizing a
function of only one variable, without constraints. However, this elimination,
by analytical methods, will be feasible only if the functions g and h are of
relatively simple formo
Alternatively, we may notice thatfcan have an extreme value at P(x o, Yo, zo)
only if the linear terms in the Taylor expansion of I1f about Pare zero . This
condition can be written in the form
fx dx + fy dy + f, dz = O (67)
Here, however, the increments dx, dy, and dz are not independent, so tbat bere
we cannol conclude that!"fy, andf, must vanisb separately at P. Furtber, since
g and h are eacb constant, the differential of eacb must be zero, ~o tbat we must
have
gx dx + gy dy + g, dz = O al (x o, Yo, zo), (68a)
hx dx + hy dy + h, dz = O al (x o, Yo, z o). (68b)
Now Equations (68a, b) are linear in dx, dy, and dz. They can be solved uniquely
for two of tbose differentials in terms of the tbird when and only when g and h
are functiooa\ly independent (see Section 7.3). Thus two of the differentials
(say dx and dy) can be eliminated from (67), leaving an equation of the form
F(x, y, z) dz = O, in whicb the one remaining differential can be arbitrarily
assigned. The condition F(x, y, z) = O, together witb the conditions (66a, b), at
(xo, Yo' zo), constitute three equations in the three unknowns x o, Yo ' and zo·
A third alternative, of frequent usefulness, is based on the observation that
one may multiply the equal members of (68a) and of (68b) by any constants A,
and A2' respectively, and add tbe results to (67) to yield the requirement
(fx + A,gx + A2hx) dx + (fy + A,gy + A2hy) dy + (f, + A¡g, + A2hJ dz = O
at (xo,Yo, zo) for any values of A, and A2 . Now it is possible to determine A, and
A2 so that tbe coefficients of two of the differentials are zero. For if this were
not so it would follow that
h,
h.<
1=0'
and beDce g and h would be functioDally dependent, so tbat the two constraints
would be eitber equivalent or inconsistent. If we imagine that A, and A2 bave
been so determined, then tbe remaining differential can be arbitrarily assigned,
so that ils coefficient a/so must vanish. Hence we obtain the three equations
fx
fy
+ A,gx + A2hx _
+ A,gy + A2hy - O
O 1 (69)
f, + A¡g, + A2 h, = O
1.1. Constraints and Lagrange Multipliers 359
which . rogether with the conditions

g = O, h = O (70)
compnse five equations in the five unknown c¡uantiries xa,Ya, Za, Al> and A2 .
The paramerers A, and Al are called Lagrange mu/lip/iers. If they were
eliminated from (69), the result would be the relation F = O obtained by
eliminating dx, dy, and dz from (67) and (68a , b). However, the new system of
nve equarions may be preferable because of the additionaJ flexibility which ir
affords . For example, it may be more convenient to solve (69) for x, y, and z
in terms of A, and A2 and ro introduce the result s into (70) for the determination
of A, and A,.
The equarions (69) are easily remembered if one notices that they are the
neces s ary condirions rp ., = O, rpy = O, rp, = O that the ';auxiliary function"
rp =f + A,g + A,h (71)
attain a relative maximum or minimum at (x a, Yo, za) when no constraínts are
imposed .
More generally, if f were to be maximized subject to n independent con-
straints g, = O, g, = O, ... , gn = O, one would form the auxiliary function
rp=f + A,g,+A l g 2 "'+Angn, (72)
where A" . . . , An are unknown constants, and write down the necessary con-
ditions for rendering rp a relative maximum or minimum with no constraints.
Example, As a very simple illustration, we seek the point on the plane

Ax + By + ez = D
which is nearest the origino Thus we are to minimize x' + y' + z', subject to the
single constraint Ax + By + ez - D = O. With
rp = (x' + y2 + Z2) + A(Ax + By + ez - D),
the conditions rp, = rp, = rp, = °at (xa, Ya, za) become
2xa + AA = 0, 2Ya + AB = 0, 2za + Ae = 0,
Xo = -FA, Yo = - ±AB ,
and substitution into the constraint condition determines A,
-FCA' + B2 + e') = D.
Thus, fínally, the coordinates of the desired point are found 10 be
AD BD eD
Xa = A' + B' + e" Yo = A' + B' + e" A2+B'+e'·
The corresponding minimum distance from the origin is
•
7.8. Calculus of Variations. An important elass of problems involves the

determination of one or morefunctions, subject to certain conditions, so as to
maximize or minimize a certain definite integral, whose integrand depends upon
the unknown function or functions and/or certain of their derivatives.
For example, to find the equation y = u(x) of the curve along which the
distance from (O, O) to (1, 1) in the xy plane is least, we would seek u(x) such
that
1= fal..jl + u' 2 dx = min
with u(O) = O, u(l) = 1.

This section presents a brief treatment of some of the simpler aspects of
such problems.
We consider first the case when we are to attempt to maximize or minimize
an integral of the form
1 = r F(x, u, u') dx, (73)
subject to the conditions
u(a) = A, u(b) = B, (74)
• where a, b, A, and B are given constants. We suppose that F has continuous
second-order derivatives with respect to its three arguments and require that
the unknown function u(x) possess two derivatives everywhere in (a, b). To fix
ideas, we suppose that 1 is to be maximized.
We thus visualize a competition, to which only functions which have two
derivatives in (a, b) and which take on the prescribed end values are admissible.
The problem is that of selecting, from all admissible competing functions, the
function (or functions) for which 1 is largest.
Under the assumption that there is indeed a function u(x) having this
property, we next consider a one-parameter family of admissible functions
which ineludes u(x), namely, the set of all functions of ·the form
u(x) + 01(X)
where 17(X) is any arbitrarily chosen twice-differentiable function which vanishes
at the end points of the interval (a, b),
17(a) = 17(b) = O, (75)
and where ~ is a parameter which is constant for any one function in the set
but which varies from one function to another. The increment ~17(X), represent-
ing the difference between the varied function and the actual solution function,
is often called a variation of u(x).
If the result of replacing u(x) by u(x) + ~17(x) in 1 is denoted by I(~),
I(~) = r F(x, u + ~17, u' + ~17') dx, (76)
it then follows that I(~) takes on its maximum value when ~ = 0, that is, when
7.8. Calculus of Varialions 361
[he variation of u is zero . Hence i[ must follow that
when E = O. (77)
The assumed continuity of the partial derivatives of F with respect to its three
arguments implies the continuity of dF/ dE, so that we may differentiate 1(E)
under the integral sign (see Section 7.9) to obtain
dl(E) = lb [aF(x. u + 01 . u' + 0'0 + aF(x, u + 01, u', + o¡') ,] d x .

dE a a(u + 0'/) r¡ a(u' + El] ) 1]
Hence, by setting E = O, we obtain an expression for the condition (77) In the

form
reo) = f [~~ r¡(x) + ~~ I]'(X)] dx = O. (78)
Here we write F F(x, u, u'), noticing that the partial derivatives aF/au and
aF/au' have been formed with x, u, and u' treated as independent variables.
The next step consists of transforming the integral of the second product
in (78) by an integration by parts, to give
b aF , [aF]b lb d (aF)
a au'" (x) dx = au,r¡(x) a - a dx au' ,,(x) dx
l
= -
la
b
in consequence of (75) . Hence Equation (78) beco mes

d aF)
dx (au' I](X) dx,
b [d
- (aF) aF] r¡(x) dx = O.
--, - -. (79)
la dx au du
It is possible to prove rigorously that, since (79) is true for any' function
,,(x) which is twice differentiable in (a, b) and zero at the ends of that interval,
consequently the coefficient of ,,(x) in the integrand must be zero e verywhere in
(a, b), so that the condition
(80)
must be satisfied. This is the so-called Euler equation associated with the prob-
lem of maximizing (or minimizing) the integral (73), subject to (74).
lf we recall [hat F, and hence also it s partial derivatives, may depend upon
x both directly and indirectly, through the intermediate variables u(x) and u'(x),
we deduce from the chain rule that
dM = aM + aM du + JM d 2u , (81)
dx dx du dx du ' dX2
where the function M 'may be identified with F or with one of its partial deriva-
tives . Thus, in particular, we can use (81) with M = aF/Ju ' to write the Euler
equation (80) in the expanded form
Fu'''' ~:~ + Fuu' ~~ + ( F,<u' - Fu) = O. (82)
AI¡hough the original form (80) usually is more convenient 10 practice, the
expanded form shows that, except in the special cases when Fu'u' =
a 2 F /au'2 is
zero, the equation is in fact a differential equation of second order in u, subject
to the two boundary conditions u(a) = A and l/(b) = B.
Since the coefficients in Equation (82) may depeod oot only on x but also
on u and du/dx, ¡he equation is not necessarily linear. However, it iovolves lhe
highest-order derivative d 2 u/dx 2 in a linear way and hence (as is customary) it
may be described a s a quasi-linear equation in tho se cases when it is not in fact
linear.
It may be noticed that from (80) there follow s
~~ = constant when F = F(x, u' ), (83a)
so that, wheo F doe s not involve u explicitly, the firs t-order equation aF/ au ' =
constant comprises a "first integral" of the Euler equation. In additioo, it is
shown in Problem 45 that (80) also implies the relation
~~ u' - F = constant wh en F = F(u, u'), (83b)
so that a fir st iotegral is also available when F does not involve x explicitly
(see also Problern 47).
It is of sorne importance to notice that we have not shown that (80) r':1s a
solution sa tisfyiog (74) or, if it has such a solution , that this solution does indeed
maxirnize or rninimize 1. We have indicated only that (80) is a neeessary condi-
tion , which mu st be sa tisfied by u if u is to qualify. The formulation of sufficient
cooditions, which ensure that a functi o n u(x) so obtained truly ma ximizes or
mioimizes 1 (or makes it a relative ma ximum o r minimum in so rne sen se), is
much more difficult.
Not infrequently, in practice, ooe can be certain in advance that an admis-
sible maxirnizing (or minimizing) function exists . lo such a case that fuoctioo
oecessarily will be obtaioed as the solution of (80) which satisfies (74), or will
be one such solution if there are seve ral. Solutioos of (80) are ofteo called
extremals of the variational problem, whether or not they satisfy (74) and maxi-
mize or minimize 1. A solution of (80) which also satisfies the prescribed end
cooditioo s (74) is said to make the integral (73) stationary, whether or not the
correspoodiog s tationary value of the integral is a maximum or minimum.
In the case of the example cited at the beginning of this section , where
F = (1 + U '2) 1/2, the fact that F depend s only upon u' shows that the relevaot
Euler equation (8 2) reduces to the form u" = O, so that (as wa s to be expected)
u must be a linear function, ti = elx + e 2 . The given end coodilioos yield
el = I and e 2 = O and hence u(x) = x.
7.8. Calculus of Variations 363
Example 1. We seek lO minimize the integral
1= r' [(~y - yZ + 21yJ dI
with y(O) = O, y(n/2) = O.

The Euler equation (80), with u and x replaced by y and 1, respectively, becomes
y dZy
-d ( 2d- ) -(-2y+21)=0
ell ell
or dI 2 +y =l,
from which there follows y = e, cos I + e2 sin I + l. The end conditions then give
e, = O, e2 = -n/ 2, and hence
n .
Y = I - -
2
510 I
'
in correspondence with which
•
Generalizations, in which more dependent and/ or independent variables are
involved or which involve other modifications, as well as formulations of suffi-
ciency conditions, may be found in the literature.
Two such generalizations, which are partieularly straightforward, may be
deseri bed he re :
(a) If (73) is replaced by the integral
1 = r F(x; u" ... , un; u'" ... , u~) dx, (84)
where values of the n independent unknown functions ti, (x), ... , un(x) are eacb
given at the end points x = a and x = b, we obtain an Euler equation similar
to (80) io correspondence with each un
~(aF)
e/x au:
_ au,
aF -_ O (r=I,2, ... ,n). (85)
Example 2. The Euler equatioos associated with the integral

Lb (u',' + u~z - 2u,uz + 2xu,) dx
are obtained by use of (85) in Ihe form
1x(2U',) - (-2uz + 2x) = O
or d; + U2 = x
(b) Suppose that we are to maximize or minimize (73),

and u/{ + ti¡ = O.
•
r F(x, u, l/') dx = max or min, (86)
where u(x) is to satisfy the prescribed eod conditions

u(a) = A, u(b) = B, (87)
as before, but Ihat also a constraint condition is imposed in the form
s: G(x, u, u') dx = K, (88)
where K is a prescribed constant. In this case, the appropriate Euler equation

IS found to be the result of replacing Fin (80) by the auxiliary function
H = F+ 1G, (89)
where 1 is an unknown constant. This constant, which is of the nature of a
Lagrange multiplier (Section 7.7), thus generally will appear in the Euler equa-
tion and in its solution, and is to be determined together with the two constants
of integration in such a way that the three conditions of (87) and (88) are
satisfied.
Example 3. To minimize the integral

fa' y'2 dx,
subject to the end conditions y(O) = O and yO) = O and al so to the constraint
fa' ydx = 1,
we write H = y'2 + 1y, in correspondence with which the EuJer equation is

2y" - 1 = O.
Hence y must be of the form y = ;\1x 2 + c,x + C2' The end conditions and the con-
straint condition yieJd c, = 6, C2 = O, and 1 = -24, and hence there follows
y = 6x(l - x).
7.9. Differentiation oC Integrals Involving a Parameter. Rather frequently it

•
is necessary to deal with a function rp(x) defined by an integral of the form
Be:.;)
rp(x) =
S A(x)
j(x, t) dt, (90)
wherejis such that the integration cannot be effected analytically. In particular,

an expression for the derivative rp'(x) often is required.
If the limits A and B are finite constants, differentiation with respect to x
under the integral sign can be justified for all x in an interval (a, b) when I and
the resultant integrand aj/ax are continuolls lor a <: x <: b and A <: t <: B.
More generally, when the limits are not constant we can think of rp as a
function of x directly and also indirectly, through the intermediate variables
A and B, and hence write rp = rp(x, A, B). Jt then follows as an application of
(6) that
drp _ arp arp dB + arp dA, + (91)
dx - ax aBdx aA dx
if the deriva ti ves on the right are continuous, where arp/ax is to be calculated by
treating A and B as constants, and hence by merely differentiating with respect
7.9 . Differenlialion oC Integrals Involving a Parameler 365
to X under the integral s ign (when f and af¡ax are continuous). To e valuate the
other partia l derivative s of rp in (91), let F(x, t) be a function such that
f( x, t) -_ aF(x,
al
t) .
There then follow s

B aF
rp(x, A, B) =
JA dI dI = F(x, B) - F(x, A)
a nd hence, when x is held constant as A and B are imagined to vary, there

follows
arp = aF(x, B) = f( B) arp _ _ aF(x, A) _ -fe A)
dA - aA - x,.
dB aB x"
By introducing these results into Equation (91), we thu s obtain the useful
formula
d
dX
fB f(x, t) dt = fB af~x,x t) dt + f(x, B) ddB -
A X
f(x, A) ddA,
x
(92)
A
whicb is valid for al! values of x in an interval (a, b) when f and af/ ax are con-
linuous for a < x ~ b and A < t < B, and also A'(x) and B'(x) are continuous
in (a, b). This formula is often known as Leibnilz's rule.
Example 1. If
y(x) = r h(t) sin (x - t) dI, (93)
then the repeated use of Equation (92) yields

y'(x) = IX h(l) cos (x - 1) di
and y"(x) = - r h(l) sin (x - 1) di + h(x). (94)
Henee it folfows that y(x) satisfies Ihe differential equalion

y " (x) + y(x) = h(x ). (95)
In fact, by setting x = a in the expressions for y and dy/cIx, we find that (93) defines
that solution of (95) whieh satisfies the initial conditions
y(a) = O,
In the case of a funetion defined by an improper integral,

y ' (a) = O.
•
rp(x) = f ~ f(x, t) dt, (96)
A. (x }
wbere tbe range of integration is infinite, the situation is somewhat more com-
plicated . Here, assuming tbat the integral (96) in faet converges, it is known
tbat the formula
~ f~ f(x , t) dt = f~ af(x, t) dI - f(x , A) dA (97)

dx A (x) A ax dx
366 Topies in Higher-Dimensional Caleulus
is validfor x in (a, b) whenf and af/ax are eon/inuousfor all/ ::>-

- A and a -<:: x <::
- b,
and A'(x) is con/inuolls in (a, b), and when also /here exisrs a fune/ion M(t),
independent of x, such /hat
af(x, t) I <: M(r)
I ax
for al! t -== A and a <: x <: b, and such rhat the integral
rA
M(t) dt
converges. The situations In which A = - ca In (96) are included as specia!

cases.
Example 2_ We consider the evaluation of the integral

rp(x) = t- e-" cos (2Ix) di. (98)
Formally, Equation (97) gives
~': = -21= te-" sin (2Ix) di, (99)
and the validity of the use of that formula is established for al! real x when we notice
that I fe-" sin (2/x) I <: /e-" for al! real x and tha! the integral S;;' /e-" d/ converges.
Further, if Equation (99) is integrated by parts, there follows
~': = [e-" sin (2tx)l~: ;;' - 2x f: e-" cos (2/x) d/
= -2x S; e-" cos (2/x) di.
Hence rp satisfies the differential equation (drpldx) + 2xrp = O, and therefore is of the
forro rp(x) = ce-x'. But when x = O Equation (98) gives
rp(O) = Jor= e-"d/ =

1
2,../7i,
by Equation (58) of Chapter 2, so that the constant c is deterroined and there fol!ows
1= e-" cos (2/x) d/ = ~ "';7i e-X'. (100)
If we introduce the change in variables / = au, where a is real and positive, and
write x = bl(2a), this result takes the useful form
(a> O). (l01)
Jt should be Doted that the stated conditions are sufficient, but by no means
•
necessary, for the applicability of (97).
7.JO. Newton's Iterative Method 367
r :;:'
Example 3. For the integral
rp(x) = di (x > O), (¡ 02)
the conditions stated in connect ion with (97) are strongly violated. Formal use of that
relalion would give
rp ' (x ) = - So- e-x'v i dt. (103)
Jf the substitution xl = u 2 is made before ¡he differentiation , there follows

? r-
rp(x) = ';xJo e-u' du = J~ '
- (l04)
and hence the correct formula is
rp ' (x)
__
1 In. (lOS)
2x V x
The same substitution in (J03) yields a result equivalent 10 (JOS) (see Problem 57) and
hence shows that differentiation under the integral sign was indeed permissible, in
spite of the unpleasant behavior of the integrando •
Example 4. For an integral of the form
rp(x) = 1=:;;' g(l) dI (x > O), (J06)
the substitution I = v 2 leads to the equivalent form
rp(x) = 2 Sa~ e-""'g(v 2 ) dv (x > O), (107)
for which differentiation under the integral sign may be more easily justified. •
Problem 56 iUustrates the existence of situations iu which the interchange

of t integration and x differentiation cannot be justified.
7.10. Newton's Iterative Metbod. The use of the procedure known as New-
ton 's method (or as the Newton-Raphson method) in obtaining successive
approximations to roots of algebraic and transcendental equatious in oue vari-
able is introduced in courses in elementary calculus. Iu this section we rederive
the basic equation and indicate the role of Jacobian determinants in tbe general-
ization to the solution of simultaneous equations in several variables.
Suppose that we ha ve obtained a first approximation to a certain root x = IX
of an equation of the form ¡(x) = O and require a more nearly accurate value.
If we denote the first approximation by x o, we then attempt to determine h so
that ¡(x o + h) = O. But for small values of h we have the Taylor series expan-
Slon
¡(x o (108)
If the initial approxirnation is sufficiently accurate and if the higher derivatives

of f are not excessively large at x o, we may neglect terms involving higher
powers of h in the equation f(x o + h) = O and hence obtain for h the approxi-
mation
( 109)
The next approximation for x is then taken as x, = X o + ho, and the process
is repeated. GeometricalIy, this procedure consists of approximating the curve
representing y = f(x) by its tangent line at x = x o , and of deterrnining the
intersection of the tangent line with the x axis (see Figure 7.1). In unfavorable
cases the process may not converge.
Figure 7.1
Ir the curve representing y = f(x) is concave away from the x axis at the
point for which x = x o, and if the curve has no turning points or infiections
in the closed interval between X o and ex (as, for example, in Figure 7.1), it is
obvious geornetricalIy (and easily proved analyticalIy) that X l wiII faII between
X o and ex, X 2 between x, and ex, and so forth, and tbat convergence of the
sequen ce of approximations x o, x 1> • • • , x., ... to ex tben is guaranteed. Tbe
specified geometrical conditions correspond to the mathematical requirements
that fexo)f"ex o) > O and that f'ex) and f"ex) each be of constant sign in the
closed interval between X o and ex, and are known as the Fourier conditions.
When they are not satisfied, convergence to ex eor to another root) may or may
not folIow.
It can be proved (see Problem 59) that, when f " ex) is continuous, the
error e. in the kth approximation x. satisfies the equation
- - f " (l k - , ) 2 (110)
e. - 2f'ex.- ,) ek_ l '
where 1.- 1 is sorne number between x k _ 1 and the true root ex-
7.10. Newton's Iterative Method 369
Example l. To illustrate Newton's method, we seek the real root of the equation
f(x) = x3 + 3x - 1 = O.
Since feO) = -1 and f( 1) = 3, and since .!'(x) > O for all x, there is one and only
one real root and it lies in (O, 1). Further, a consideration of the form of fsuggests the
starting value Xo = 1, for which it is also found that the desirable condition ff" > O is
satisfied. Since also f"(x) > O for ail x > o, the Fourier conditions are satisfied and
convergence is guaranteed. The formula
h - _f(Xk) _ _ xk + 3Xk - 1
k - .!'(xk ) - 3(x~ +
1)
then yields ho = -,Ja when Xo = 1, so that x, = 0.3333 - 0.0111 = 0.3222 when

four significant figures are retained. Next, with k = 1, there follows h, = -0.00001465.
10 four significant figures, and hence X2 = 0.32218535, to eight places.
Equation (110) here is of the form
and, since here tk~ ', certainJy is in (O, 1) when Xk-' is in that interval, it follows that
certainly lekl < d-, in the present case. Ifwe estimate e, as approximately X 2 - X,
"'= 2 X JO-s, we deduce that probably le21 < 4 X 10-'0, indicating that the eight
digits in X2 are correct and, indeed, that a ninth digit might have been proper!y retained
in its calculation . •
Now suppose that a certain so!ution of the simultaneolls equations

f(x , y ) = 0, g(x, y) = ° (1 I 1)
is required, and that a reasonably accurate initial approxirnation (x o , Yo) has
been obtained by sorne rnethod. We next attempt to determine values of h and
k such that the equati o ns
f(x o + h, Yo + k) = 0, g(x o + h, Yo + k) = °
are simultaneously satisfied. If the left-hand members are expanded in Taylor
series about the initial point and if only linear terms are retained, these equa-
tions become
fo + hf.o + kfyo = 0,
(112)
go + hg xo + kgyo = 0,
where the zero subscripts indicate that the functions involved are evaluated at
the point (x o, Yo). Thus approximate corrections ho and ko are given by the
solution of these equations, in the form
fo fyo I f .o fo I
ho - I go gyO
ko - I g. o go ( 113)
a(f, g)o a(f, g)o
J(x, y) a(x, y)
Tt is seen that the success of this method depends in part upon the magnitude
of the Jacobian determinant in the neighborhood of the desired solution.
To see the geometrical significance of this dependence, we recall that the

vectors
Vf =f) + fyj, Vg=g)+gyj
are normal to the curves representing f = O and g = O in tbe xy plane. Further,
we obtain tbe result
(Vf) x (V g) = kUxgy - fygx) = k ~~~ ;~, (114)
and it follows from the definition of the cross product (Section 6.3) that the
magnitude of the Jacobian is proportional to tbe magnitude of the sine of the
angle between the normals (and hence also between the tangents) to the curves
at points of intersection. Thus small values of the Jacobian may be expected
to correspond to cases in which the intersecting curves have nearly equal slope
at the intersection, in whicb cases slow convergence (or divergence) of the itera-
tive process is to be anticipated.
Example 2. To iIlustrate this procedure in a favorable case, we consider the simul-

taneous eq uations
f = x 3 - y2 - 3x + y + 2 = O, g = X2 + y2 - 4 = o.
By rough graphical methods it is found that the circ1e g = O intersects the cubic curve
f = O at a point with approximare coordinates (1.5, 1.5). With the choice X a = Yo = 1.5,
Equations (113) give ha = -0.080 and ka = -0.087, leading to the first improved
coordinates XI = 1.420, YI = 1.413. The exact values are x = Y = -vil = 1.414 ....
The methods of tbis section c1early can be generalized to the solution of n

•
equations in n variables.
REFERENCES
1. ApOSTOL, T. M., lVlarhemariea/ Ana/ysis, Addison-Wesley Publishing Company, Inc.,

Reading, Mass., 1957.
2. Buss, G. A., Ca/eu/us of Variarions, First Carus Mathematical Monograph, Open
Court Publishing Company, La Salle, 111., 1925.
3. BucK, R. c., Advaneed Ca/eu/us, McGraw-Hill Book Company, Inc., New York,
1956.
4. FRANKLIN, P., A Trearise on Advaneed Ca/eu/us, John Wiley & Sons, Inc., New
York, 1940.
5. HARDY, G. H., A Course of Pure Marhemaries, Cambridge University Press, New
York, 1959.
6. KAPLAN, W., Advaneed Ca/eu/us, Addison-Wesley Publishing Company, Inc., Read-
ing, Mass., 1952.
7. TAYLOR, A. E., Advaneed Ca/eu/us, Ginn and Company, Boston, 1955.
Problems 371
8. WEINSTOCK, R., Ca/Cl//us al Varialion s: Wilh App/icalions lO Physics and Engineer-

ing, MeGraw-Hill Book Company, Inc., New York, 1952 .
9. WI DD ER, D. V., Advallced Ca/cu/us, 2nd ed., Prentiee-Hall. Ine., Englewood Clifrs,
N.J., 1961.
PROBLEMS
Section 7.1
1. Ir x = reos 8 and y = r sin 8. determine expressions for eaeh of the following and
express eaeh result as a funelion of r and 8 :
(a)
(~:).. ( b) (aX)
a8 .-
(e)
(~n.. (d)
(~e).-
(e) (~~t, (fl
(~:); (g)
(~~t, (h)
(~~);
2. Ir x = reos 8 and y = r sin 8, express eaeh of the following as a funetion of r and
8:
(a)
(~~): (b)
(~;).- (e)
(~~)¡ (d) (~;).,
(e)
(~~),; (f)
(~~t, (g)
(~~): (h)
(~~);
3. If s and 1 are funelions of x and y, say s = f(x, y) and 1 = g(x, y), show that
aaF
x =
(s a
., as + l ., a) F,
al
4. With Ihe nOlalion of Problem 3, show Ihal
and henee obtain the result

a'F a 2F a2F a'F aF aF
a x ay = SxSy as2 + (sx t, + S,lx) as al + lx t , al2 + Sxy as + I xy al'
[Notiee that the variables (x, y) and (s, 1) may be interehanged throughout.]
5. (a) By identifying x and y in the result 01' Problem 4, obtain the relation
a' F a 2F
2 a 2F a2F aF aF
a x2 = Sx as' + 2sxtx aSal + t; al' + sxx as + Ixx Tr'
as well as an analogous express ion for a2 F;a y2.
372 Topies in Higher-Dimensional Calculos
(b) If Iy = Sx and Ix = -s" show !ha!

2 2 2
V 2F = a F
- ax2
+ aay2
F = (s2z--'-
' Sy
2) (a F
as2
+ aal2F).
2
6. If f(p, v, T) = O, show !ha!
and deduce tha!
7. Ir E = f(P, T) and T = g(p, v), show tba!
(a) (~~t = frg. = (~;t (~~t

(b) (~!). =fp + frgp = (~!t + (~;t(~~);
Section 7.2
8. If u and vare funetions of r and s, and also r and s are funetions of x and y, prove
that
a(u, v) aCr. s) a(u, v)
a(r, s) a(x, y) = a(x, y)'
9. The variables x and y are expressed in terms of the variables u and v by the equa-
lions
x = F(u, v), y = G(u, v).
(a) By writing
f(x, y, u, v) = x - F(u, v), g(x, y, u, v) = y - G(u, v)
and thus redueing the given equations to a speeial case of Equations (18), show tbat
the statement following Equation (22) implies that u and v may be eonsidered as fune-
tions of x and y if the Jaeobian J = a(F, G)¡a(u, v) does not vanish, and that this
eondition is equivalent to the requirement
J = a(x, y)
a(u, v)
*' O
when x and y are eonsidered as functions of u and v.
(b) Assuming tbat J *'
O, show that Equations (21) then beeome
u = y.
J'
u ==. x
----..!.
= -7'
Yu x".
= y'
."1: y J' Vx Vy
(e) Obtain the same results direetly by differentiating the original equations
partially with respeet to x and with respeet to y, in eaeh case holding the seeond
variable eonstant, .t o obtain the relations
1 = Fuux + F vx,
lI o= Fully + Fvvy,
O = Guu x + Gvv x , 1 = Guu y + G.v"
and by solving these equations for the desired quantities.
Problems 373
10. The five v¡¡.riables x, y, z, u, and vare related by three equations of the form
J(x, u, v) = O g(y, u, v) = O, hez, u, v) = O.
(a) State eonditions under whieh the first two equations determine 1I and v as
funetions of x and y and the third equation determines z as a funetion of u and v, and
henee then as a funetion of x and y.
(b) Considering z, u, and v as funetions of the independent variables x and y,
differentiate the three equations partially with respeet to x, holding y eonstant, and
henee show that Ux> Vx> and Zx satisfy the equations
J.u x + J,v x = -J"
guu x + gvv;x = O,
h,z.< + h.u x + hvv x = O.
(e) By solving these equations for Zx (by determinants), obtain the result
fx aCh, g)¡aClI, v)
z = -. .
.< h, a(f, g)fa(u, v)
(d) Show also, by syrnrnetry or otherwise, that
z = _ &. . a(h,f)fa(u, v) .
y h, a(f, g)¡a(u, v)
(e) Verify these results in the case when
2X2 - U +V = O, 2y J - U - V = O, z + u - v2 = O
by first obtaining z explicitIy as a funetion of x and y and differentiating, and also
using the resuIts of parts (e) and (d), to determine Zx and zy-
11. Show that the equations
x = F(u, v), y = G(u, v), z = H(u, v)
determine z as a funetion of x and y sueh that
a(z, y)¡aCu, v) a(z, x)¡a(u, v)
Zx = a(x, y)¡a(u, v)' Zy = - a(x, y)¡a(u, v)'
if a(x, y)¡a(u, v) 7"= O. (Write J = x - F, g = y - G, h = z - Hin Problem 10.)
12. If f(x, y, z) = O and g(x, y, z) = O, we may in general eonsider any two of the
variables as funetions of the third. Show that
dx dy dz
a(f, g)fa (y, z) - a'(-;-/,""',g--c)C7¡a'('-z,-x""') a(f, g)fa(x, y)
if no denominator vanishes.
13. Show that the tangent line, at a point (xo, Yo, zo), to the curve of interseetion of
the surfaees J(x, y, z) = O and g(x, y, z) = O is speeified by the equations
x - Xo y - Yo z - Zo
[au; g)¡a(y, z)]o ¡a(/, g)¡a(z, x)]o [a(/, g)¡a(x, y)]o
Section 7.3
14. Prove that the functions
x+y , u _ xy
u¡ (x _ y)2
x-y 2 -
are funetionally dependen!.

15. Pro ve that the functions u , = y + 2, U2 = X + 2z 2 , and u, = X - 4yz - 2y 2 are

functionally dependent.
16. Determine whether the functions
x-y
and x +z
U I -
x+z y +z
are functionally dependen!.
17. Determine whether the functions
lfl =
yz - X
,
xyz - y2z2 + X2
X xyz ,- y2z2
are functionally dependent.
Section 7.4
18. The coordinates l/, rp, and e are defined by the equations
x = au sin rp cos e, y = bu sin rp sin e, z = eu cos rp,
where u :> O, o -< rp -< n, and o -< < 2n. e
(a) Show that the surfaces U = constant are the ellipsoids
whereas the surfaces rp = constant and e

= constanl are elliptical eones and planes,
respectively.
(b) Show that the element of volume is of the form
d7: = abeu 2 sin rp du drp de.
19. Use the result of Problem 18 lo show that the z coordinate of the center of gravity
of that half of the ellipsoid
X2 y2 Z2 1
a' + b' + e' =
which lies aboye the xy plane is given by
Vi =
r
abe' Jo
I r
u' du Jo
n
/2
sin rp cos rp drp
f,n
Jo de =
n be ' ,
4"a
where
V = abe
'J o
u' du
J"!2sin rp drp f,n de = 23n abe,
o .. o
and hence has the value i = 3e/8.
20. To generalize Problem 18, let
x = f(rp, e), y = g(rp, e), z = h(rp, e)
define a simple c10sed surface S which surrounds the origin, with O -< rp -< n and
O -< 8 < 2n, and introduce the coordinates
x = uf(rp, 8), y = ug(rp, 8), z = uh(rp, e),
where O -< u -< 1, O -< rp -< n, and O -< e < 2n.
Problems 375
(a) Show tha! Ihe element of volume is

d, = I Q(rp, e) luz du drp de,
f g h
af ag ah
where Q(rp, e) = arp arp arp
af ag ah
ae ae ae
(b) Show tha I
J'J'J's w(x, y, z) dx dy dz = j'2o" ...r"o ...r'o W(u, rp, e) I Q(rp, e) lu 2du drp de,
where
W(ll, rp, e) = w[¡t{(rp, el, ug(rp, 8), uh(rp, e)],
and, in particular, that the yolume of S is
V = I
'3 f2"o f"o I Q(rp, e) I drp de.
(Notice that only in special cases are rp and e the cone angle and the polar angle
of spherical coordinates.)
21. Suppose that a simple closed curve e, surrounding the origin in the xy plane, is
specified by the parametric equations
x = f(rp), y = g(rp), (O < rp < 2n).
(a) Show that, if coordinates u and rp are defined by the equations
x = ¡t{(rp) , y = ug(rp),
Ihen the plane region D encJosed by e is specified by the ranges O < u <1 and
O -<: rp < 2n.
(b) Show that the element of area in urp coordinates is given by
dA = I Q(rp) I u du drp,
where Q(rp) is the Wronskian (se e Section 1.2) of f(rp) and g(rp),
Q(rp) = f(rp)g'(rp) - f'(rp)g(rp).
Deduce that
ft w(x, y) dx dy = Sa2n Sal w[uf(rp), ug(rp)] I Q(rp) lu du drp,
and, in particular, that the area of D is
A =
1
'2 f2"o I Q(rp) I drp.
22, (a) Verify that the results of Problem 21 specialize to those of Ihe lext Example
when f(rp) = a cos rp and g(rp) = b sin rp.
(b) If b = a in part (a), so that e is the circle X2 + y2 = a 2 and D is ils interior,
show that the result of Problem 21 (b) takes Ihe form
r r w(x, y) dx dy
..... D O
= a2 S2n ...ro1 w(au cos rp, au sin rp)u du drp
and that it reduces to a more familiar form with the substitutions e ~ rp and r = au.
(See also Problem 23.)
23. If rp can be identified with the polar angle e
~ cos-' (x!r) in Problem 21, verify
that the result of Problem 21 corresponds simply to replacing the polar coordinates
(r, e) by new coordinates (ti, rp) such that
e= rp, r = tlp(e),
with
(e) = p(e) cos e, g(e) ~ p(8) sin e, p(e) = ../[/(e)]2 + [g(e)]2,
where r ~ p(e) on C. [Show that here the Wronskian of /and gis Q(e) = [p(e)]2 and
that the integral
.o
2• fP{(J)
J o w(r cos e, r sin e)r dr de
takes the form provided by Problem 21 (b) with the change of variables e = rp, r ~ up(e).
Also specialize to the case when /(rp) ~ a cos rp and g(rp) ~ a sin rp. See also Problem
22(b).]
24. If D is the plane region bounded by the curve
x ~ a(2 cos rp - cos 2rp), y = a(2 sin rp - sin 2rp) (O <: rp < 2n:),
use the result of Problem 2J(b) to show that
f lo w(x, y) dx dy = 6a 2 Sol. So' W(u, rp)(1 - cos rp)u du drp,

where
W(lI, rp) ~ w[au(2 cos rp - cos 2rp), au(2 sin rp - sin 2rp)].
In particular, show that the enclosed area A is 6n:a 2 (Here rp clearly is not identifiable
with the polar angle e,
in contrast with the cases considered in Problem 23. The curve
is an epieycloid. traced out by a point on the circumference of a circle of radius a as
that circle rolls around a fixed circle with the same radius.)
25. (a) If a surface S is defined by the equations
x = x(u" U2), y = y(u" U2)' z ~ z(u" U2),
show that the directed element of sur/ace area on that surface is given by
da ~ U, X U 2 du, dU2'
where U! ~ ar/au, and U 2 ~ ar/au2' and where r is the position vector xi + yj + zk.
(b) Use Lagrange's identity [Equation (36) of Section 6.4] to show that the scalar
element of surface area is of the form
da = ../ EG - F2 du! dU2
where
26. Show that the element of are length ds along any curve Con the surface considered
in Problem 25 can be obtained, from the fact that
dr = ar du! + ar dU2
ds au, ds aU2 ds
problems 377
is a uni! vector tangent to the curve e, in the form

ds = V E duy + 2F du, duz + G duÍ,
where E, F, and G are defined in Problem 25.
27. Speeialize ¡he results of Problems 25 and 26 to the ease when the surface S is spe-
eified by the equations
x = U¡, z = U¡UZ,
showing that
da = V 1 + u} + u~ du, duz
and
ds = ·.,10 + uD du} + 2u,uz du, duz + (1 + uD du¡
on S. Also show that S is the hyperbolie paraboloid
z = xy,
and rederive the expression for da on S by use of Equation (04) of.Chapter 6 and that
for ds on S by starting with
ds = .v dx2 + dy' + dz 2 .
(Notiee that here x and y are not orthogonal eoordinates on S.)
28. lf S is the ellipsoid
x = asinu l COSU2, y = b sin u, sin U2, z = eCOS U¡,
where O -< u, -< 71. and O -< U2 < 271., show that
E = a 2 cos 2 lit cos 2 lIz + b 2 cos 2 U¡ sin z U'2, + e 2 sin z U}l
F = (b 2 - aZ) sin u, eos u, sin U2 eos u,-,

G = a 2 sin 2 U¡ sin 2 llZ + b 2 sin 2 U¡ cos 2 UZ,
with the notation of Problems 25 and 26, and henee that

da = sin u, .ve 2 sin' u, (a Z sin 2 U2 + b' eos' u,) + a 2 b 2 eos' u, du, dU2
and ds = VE duy + 2F du, dU2 + G dur
on S. In pariieular, when a = b show that u, and u, are orthogonal eoordinates on S,
with ti, = (J, where (J is the polar angle, and that then
da = ave 2 sin 2 ", + a 2 eos 2 u, sin u, dtl, dU2,
ds = "v(a 2 cos 2 II I + e 2 sin 2 ud dUT + a 2 sin 2 U 1 du!¿,
whereas when a = b = e the ellipsoid beeomes a sphere and the eoordinates beeome
the spherieal eoordinates u, = rp, U2 = (J, with r = a.
Section 7.5
29. Obtain the terms of degree two or less in the Taylor expansion
e X + Y = Qo + b¡x + b2 y + C 1X 2 + C2xy + C3y2 + ...
(a) by direet use of the two-dimensional Taylor formula,
(b) by first expanding e X + Y in a series of powers of x + y,
(e) by multiplying together appropriate expansions of eX and eY.
30. Obtain the terms of degree two or less in the Taylor expans ion
1
r x - y
= 00 + b,(x - 1) + b , (y - 1) -r e,(x - 1) :
+ c,(x - I )(y 1) +e J (y - 1)' + ".

(a) by direet use of the two-dimensional Taylor formula,
(b) by any shorter method.
31. For a eertain funetion f(x , y), it is known that
feo. O) = J, IJO, O) = 0 .25, /,/0, O) = 0 .50
and that 1/,., I <: 0.15,lfx,1 <: 0.05, and Ifn I S 0 .05 e verywhere along that segment of
the line y = x whieh joins (O, O) and (0.1,0.1). Show that
1.0735 <: f(O.l, O. J) <: 1.0765.
32. Obtain a three-deeimal place approximation to the value of
t' fo' e -"'" dx dy

by use of Taylor series.
Section 7.6
33. Loeate and identify a11 relative maxima and minima and saddle points (if sueh
exist) of the following funetions:
(a) X2 + 2bxy + y2, (b) X2 - xy + y2 - 2x + y,
(e) x2y2 - X2 _ y2, (d) x 3 y3 - 3x - 3y.
34. Loeate and identify all points at whieh absolute maxima or minima are assumed
by the following funetioos:
(a) x 2 ; J + y': 3, (b) (X2 + y2 - 2x + 2y + 2)1 .' 2,
(e) (y2 - X 2 )' / \ (d) (2x - 2y - x' - y2)' , 1.
35. Loeate and classify a11 maxima and minima and determine any saddle points for
the foJJowing fuoetioos:
(a) 2x' - x ' y' + 2y', (b) x 4 - 2x'y' + y4 ,
4
(e) x _ 4x2y2 +y4 , (d) x 4 + y4.
36. Find the absolute maximum and mlnlmum of the funetion X2 - y1 - 2x in the
region x ' + y' <: l. (Write x = eos y = sin on the boundary.) e, e
Section 7.7
37. Determine when, if ever, X2 + y1 takes on a minimum or a maximum value under
the eonstraint xy = J.
38. Determine when, if ever, .'(y takes on a maximum or a mioimum value under the
eonstraint X2 + y2 = 2.
39. Minimize X2 + 4y2 + 16z 2 under eaeh of the following eODstraints:
(a) xyz = 1, (b) xy = 1, (e) x = 1.
Problems 379
40. Derermine the points on the ellipse, defined by the intersection of the surfaces
x + Y = I and x' - 2y' + z' = 1, which are nearest tO and funhest from the origino
Section 7.8
41. Obtain the Euler equation associated with maximizing or minimizing S: F dx In
the following cases:
(a) F = pex)y" - qex)y2 + 2f(x)y,
(b) F = A(x)y" --'- 2B(x)yy' + C(x)y' + 2D(x)y' + 2E(x)y.
42. Obtain ¡he two Euler equations associated with maximizing or minimizing j! F dx
in the followi ng cases :
(a) F = Aex)u'¡' + 2B(x)u'¡u', + C(x)U~',
(b) F = a¡¡(x)u'¡2 ..,.. 2a¡ ,(x)u'¡u', ' a"ex)u'¡'
+ b¡¡(x)u; + 2b¡,(x)u¡1/ 2 + b,,(X) Uf.
43. Geodesics on a surfaee are the extremals of the problem of minimizing the distance,
along that surface, between two points on the surfaee . Determine the geodesics on the
right circular eylinder x' + y' = a', by writing x = a eos e
and y = a sin e
and
proeeeding by eaeh of the following methods:
(a) Assume z = u(e), so that ds = -../-0'2-.-'--'-,-u---,."2 de,
(b) Assume e = vez), so that ds = -../ a 2 v'2 + 1 dz,
(e) Assume z = f(t) and e = g(r), where I is a parameter, so that
ds = -../1'2 + alg" dI .
[Notice that in (a) we assume that z can be expressed as a single-valued differen-
tiable function of e,whereas in (b) the roles of the variables are reversed. In (e) neither
assumption is made.]
44. Determine the geodesies on the sphere x' + y' + Z2 = 0 2 , using the spherieal
coordinates defined by Equation (169) of Chapter 6 and assuming that the required
equation can be written in the form e
= u(rp). (Take eot rp as a new variable in evaluat-
ing a relevant integraL)
45. Transforma/ion of the Euler equa/ion.
(a) By multiplying the equal members of the Euler equation (80) by L/' = du/dx,
transforming the first resultant left-hand member by using the relation UV' =
( UV)' - U ' V, and referring to (81) with M = F, show that the Euler equation (80)
implies the equa/ion
!!..(a~u' _ F) + aF O.
dx au ax =
[Apart from the fact that thi s equation admits the general1y extraneous solution
u = constant, the derivation shows that it is equivalen/lO (80) .]
(b) Deduce that if F does not depend explicitly upon x, then a firSI in/egral o( /he
Euler equa/ion is
~~ u ' - F = constant,
with the understanding that the solution u = constant may not be admissible. (The
use of this result may Or may not be advantageous in a specific case, as is illustrated by
Problems 46 and 47.)
46. (a) Use the result of Problem 45 to obtain the extremals associated with the
integral
s: ,.¡ u2 + U'2 dx
In the form
u(x) = el sec (x - e2)'
(b) Obtain the Euler equation (80) when F = ,.¡ ,,2 + U'2 and compare the labor
required for its solution with that needed in part (a).
[Note that the solution u = e, discarded in part (a), in fact does not satisfy the
Euler equation unless e = O, in which case it is a specialization of the basic solution.]
47. Compare the determination of the extremals associated with the integral
s: (u 2 + U'2) dx
by the method ofProblem 45 with that based on (80). (In general, if the Euler equation
is a linear eql1ation for which the solution is known, the "short-cut"' method of Prob-
lem 45 is not to be recommended.)
48. (a) Determine the minimum and maximum values of the integral
S: yy' dx,
subject to the end conditions y(O) = 0, y(l) = l. Also account for the degenerate form
of the Euler equation in this case.
(b) Show that, of all the twice-differentiable functions such that y( -1) = -1,
y(l) = 1, there is no one for which the integral
S I
-1
x2y'2 cix
takes on either a minimum value or a maximum value.

49. Show that the Euler equation corresponding to minimizing the integral
1 =s: [p(X)y'2 - q(X)y2] dx,
subject to the end conditions y(a) = 0, y(b) = °and to the constraint

s: r(x)y2 dx = 1,
is
:X(p ~~) + (q + A.r)y = 0,
when the Lagrange multiplier is denoted by -A., [Notice that here A. is a characteristic
number associated with a Sturm-Liouville problem (see Section 5.6).]
50. Find the minimum value of ti y'2 dx, subject to the conditions
y(O) = y(1) = ° and So' y2 dx = 1.

Problems 381
51. Jf the length of a curvilinear arc joining the points (O, O) and (1, O) is prescribed as
1, and if the area bounded by this arc and the x axis in the upper half-plaoe is to be
maximized, show that the arc must be circular, under the assumption that we may
write y = ll(X) along the arc. Show also that this assumption is contradicted unless
1-< n/2, and that the additional assumption that u'(x) exist at the ends requires in fact
that 1 < n/2.
52. To avoid the restriction 1 < n/2 in Problem 51, use Equation (140) of Chapter 6
to write
~
l
A = Jo (xy' - yx') dI
where
x(O) = O, x(1) = 1, y(O) = O, y(1) - 1
and where
t' ,..¡ X'2 + y2' dI = l.
By introducing a Lagrange multiplier l, integrating the two Euler equations once, and
then eliminating l between the results, show again that the arc must be circular, but
now with no restriction on l. [Omit the determination of X(I) and Y(I).]
53. Determine y(x) and z(x) such tha!
y(O) = z(O) = O, yen) = z(n) = 2
and such that the integral
1= t" (y'2 + Z'2 + 2y'z) dx
is stationary, and aIso determine the corresponding stationary value of l.
Section 7.9
54. If
1
xa _ Xb
where a> b> -1, show that

I(a) =
J o
1
ogx
dx,
1
J
I
1'(0) = o x dx
a
= a + l'
and hence, noticing that I(b) = O, obtain the result
J'
o
x" - xb
log x dx =
Ja b II
du
+ 1 =
a
log b
+
+
1
1 (a> b> -l).
55. (a) If
1(0) = J- o
e-
ax
x
sin x dx
'
where a > O, show that I'(a) = -1/(0 2 + 1). Hence, noticing that 1(=) = O, obtain
the result
-
e-ax sin x n - tan-' a
o J
c-_=.:...c.: dx = -
x 2
(a> O).
(b) By eonsidering the limil of this resull as a ~ O, and making the justifiable
assumption that the limit can be taken under lhe integral sign, deduce also that
l
~sinxdy -~.
" -- /
o X -
56. (a) By replaeing x by ax in the result of Problem 55(b), obtain the result
l( a ) -= l~ sin ax d x = ~
1. (a> O).
o x
(b) Show that, although henee /'(a) = O when a> O, the derivative cannot be
detennined by differentiating under the integral signo
(e) Show that 1(0) = O and also that l( -a) = -I(a), so that l ea) must be given
by -n/2 when a is negative.
57. Given the evaJuation
L~~ e-x' dx = .";7i.
[se e Equation (58) of Chapter 2], deduce the formula
by replaeing x by l/X in the original integral, dividing by l/, differentiating with respeet
to l/, and finally setting u = l.
58. Obtain a differential equation, logether with appropriate initial eonditions,
satisfied by the funetion
1
y(x) = ;-¡
1..
I X(x - t)2h(t) dt.
a
Section 7.10
59. (a) If Xk is the kth approximation afforded by Newton 's method to a zero IX of
f(x), show that
(b) If f(x) has a eontinuous se eond derivative in the interval between Xk_' and
IX, show that
f(lX) - f(Xk-.) = (IX - Xk _. )/'(x._.) + (IX - X k- I )2
2 f "(1
. k-I
)
,
where 1k - I is some number between X'_1 and IX .

(e) By introdueing the result of part (b) into the res ult of part (a), and writing
ek = IX - Xk for the error in Xk, deduce Equation ( 1 JO).
60. Determine to fíve plaees the real root of the equation x 3 - 2x - 5 = O.

61. Determine to three plaees the smallest positive root of the equation
tan x = x
by Newton 's method. [Suggestions: Notiee that x = 3n/2 is a fair approximation,
Problems 383
obtained by sketchiag the Curves y = lan x and y = x on Ihe same graph and e s timat-
ing the abscissa of Ihe interseclion. Take J(x) = sin x - x cos x, rather Ihan
J(x) = tan x - x, to avoid the infinil y of the ¡alter funclion.]
62. Determine lO three places Ihe smallesl posilive root of the equation
tanh x = tan x
by Newlon's method.
63. Delermine 10 Iwo places the smallest characleristic value of ¡.¡. for Ihe problem
d 2y y'(O) = y(O), y'(I) = y(J).
dX2 - ¡.¡.2y = 0, -
64. Determine lO three places the real solution of the simultaneous equations
X2 + y2 = 1, xy = x - y
in the first quadrant.
65. Determine to three places the real solútion of the simultaneous equations
4x 3 - 27xy2 + 25 = 0, 4x2y - 3y 3 - l = O
in the fust quadrant.
66. The path of a projectile movlng In the xy plane is specified by the paramelric
equations x = J(I), y = g(I), where I is time. It is required to determine Ihe time at
which its Irajectory will intercept a curve specified by the equation rp(x, y) = O. Ir the
approximate time is lo and Ihe approximale coordinales of the interception are (xo, Yo),
show Ihat Newton's melhod yields a new estimate 1= lo + T, where
T=
rpo + (Jo - xo)rpxo + (go - yo)rpyO
J~rpxo + g~rpyO
and where the zero subscripls indicale evalualion for I = lo, X = xo, and Y = Yo .
[Linearize Ihe relations Xo + h - J(lo + r) = 0, Yo + k - g(lo + T) = 0, and
rp(xo + h, Yo + k) = O and solve for T.]
8
Partial Differential Equations
8.1. Definitions and Examples. A partial differential equation is said to be

linear if, when the equation has been rationalized and cIeared of fractions, no
powers or products of the unknown function or its partial derivatives are pre-
sent. Ir it is true only that no powers or products of the partial derivatives 01
highest order are present, the equation is said to be quasi-linear. Thus, for
example, the equation
xaz
- + yaz
-=z
ax ay
is a linear equation in z, of first order, whereas the equation
a2z + (aZ)2
zax2
_
ay - o
is a quasi-linear equation of second order.
By a solution of such an equation we mean either an explicit expression for
z, of the form z = rp(x, y), which reduces the equation to an identity, or a rela-
tion F(x, y, z) = O which determines z implicitly as such a function of x and y.
The term integral surlace is frequently used to denote either tbe geometrical
representation of a solution or the solution itself.
In this work we deal principaIly witb linear equations, although certain
procedures to be developed can also be applied to the solution of certain quasi-
linear equations.
We recaIl that in the case of an ordinary differential equation of order n,
a general solution involves n independent arbitrary constants. In particular, the
general solution of a linear ordinary differential equation expresses the unknown
variable as a linear combination of n independent functions, tbe n arbitrary
384
8.1. Definitions and Examples 385
constants appearing as the coefficients of the n functions in the linear combi-

nation. For nonlinear ordinary equations the constants in general appear in a
more complicated way in the solution, and there may exist so-called "singular
solutions" which do not involve the arbitrary constants and cannot be obtained
from the first solution by specializing these constants.
In the case of partial differential equations, general solutions are found to
involve arbitrary functions of specific functions. As a simple example, we readily
verify that the equation
(1)
is satisfied by the expression

z = f(x + 2y), (2)
no matter what functional relationship is indicated by f, so long asfis differen-
tiable. For from Equation (2) we obtain
az = df(x +
2y) a(x + 2y) = f'(x + 2 ),
ax d(x + 2y) ax y
az = df(x + 2y) a(x + 2y) = 2'f'(x + 2 ) = 2 az ,
ay d(x + 2y) ay y ax
and hence (2) implies (1).
To see that (2) is indeed the most general solution of (1), we introduce the
new variables
t = x + 2y, s = x. (3)
There then follows
az _ az!!.!.. + az as _ az + az ,
ax - at ax as ax - at as
az _ az at + az as = 2 az ,
ay - at a y a s ay at
and (1) becomes merely
(4)
when z is considered as a function of s and t. The general solution of this equa-

tion is clearly z = f(t) = f(x + 2y), where f is an arbitrary function, in
accordance with (2). We see that such expressions as z = x + 2y + 1,
z = sin (x + 2y), z = 4"jx + 2y + cos (x + 2y), ... , are aH particular solu-
tions of the partial differential equation (1).
As a further example, we determine a partial differential equation which
has the expression
z = f(2x + y) + g(x - y) - xy (5)
as its general solution. The procedure consists of attempting to obtain, by
differentiation, sufficiently many relations to permit elimination of the arbitrary
386 Partial Differential Equations
functions. With the notations

l' = dJ(2x y), + , dg(x - y)
d(2x + y) g = d(x - y) ,
the first and second derivatives of (5) are obtained in the forro
~~ = 21' + g' - y, az
ay = J' - g, - x, (6a)
a:2~y
2 2
a
ax2
z = 4r "
~
+ '",," , = 2/" - g" - 1, a z =J"
ay2 + g" . (6b)
Equations (6b) constitute three relations involving the two arbitrary functions
J" and g". If two of these relations are used to determine J" and g", and if the
results are introduced into the third relation, there foIlows
(7)
This is the diffetential equation of lowest order satisfied by (5), with J and g
arbitrary functions.
We may verify that, in fact, Equation (5) defines the masl general solution
of (7) by first noticing that, since z = -xy satisfies (7), the function
w = z + xy (8)
is to be the general solution of the equation
a 2w a 2w a'w (9)
-
ax2 - ax ay - 2ay2
-=0.
Now, if we roake the substitutions

2x +y = s, x-y=l (10)
suggested by (5), there follows
~ _ as ~ ~~ _ 2~ ~,
ax - ax as + ax al - as + al
~ _ as ~ + al ~ _ ~ _ ~,
ay - ayas ay al - as al
and (9) takes the form
(11)
from which the general solution

w = J(s) + g(l) = J(2x + y) + g(x - y) (12)
is obtained by direct integration. The introd uction of (12) into (8) then leads
indeed to (5).
In a similar way, the expression
z = J(ax + by) + g(cx + dy), (13)
8.2. The Quasi-Linear Equation of First Drder 387
where a. b, e,_ and d are constants such that ad =;Oc be, is readily shown to be
the general solution of a linear partial differential equation of tbe form
( 14)
wbere A, B, and e are constants depending upon a, b, e, and d.

In more involved partial differential equations, the arbitrary functional
relations may enter into a general solution in much more complicated ways,
and additional solutions may exist which cannot be obtained by specializing
the general solution so obtained. Because of the fact that "general" solutions
of partial differential equations involve arbitrary functions, the specialization
of such solutions to particular forms which satisfy prescribed boundary condi-
tions involves the determination of funetional relations, rather than merely the
determination of eonstants, and is usually not feasible. For this reason, we gen-
erally prefer to determine a set of particular solutions directly, and to attempt to
combine these solutions in such a way that the prescribed boundary conditions
are satisfied. The development of such procedures, in cases of linear equations,
forms the basis of much of Chapter 9. However, since certain general properties
of solutions of certain equations are most readily obtained by studying the
general solutions, we show in Sections 8.2 through 8.5 in what way such solu-
tions can be obtained in certain simple cases of frequent occurrence.
8.2. The Quasi-Linear Equation of First Order. The most general quasi-
linear partial differential equation of first order can be written in the form
P(x, y, z) ~~ + Q(x, y, z) ~; = R(x, y, z). (15)
where z is the dependent variable and x and y are independent. If (15) is to be

truly linear, tben P and Q must be independent of z and R must be a linear
function o(z, say R = R,z + R 2 , so that (15) becomes
az
P(x, y) ax + az
Q(x, y) ay = R,(x, y)z + R 2 ( X, Y ) (16)
in this special case.

Suppose that the equation
u(x, y, z) = e (17)
defines a solution (or "integral surface") of (15), in the sense tbat (17) determines
z as a function of x and y which satisfies (15). Then, by partial differentiation
we obtain the two results .
aU...Lauaz_O (18)
ax ' az ax - ,
where in the derivatives aulax, aulay, and aulaz the variables x, y, and z are con-
sidered as independenl. Thus we may write

au/ax az au;ay
- au/az' ay = - au/az' (19)
assuming, of course, that au/az =1= o. If these expressions are introduced into
(15), an equation governing the function u is obtained in the form
pau+Qau+Rau=O. (20)
ax ay az
This form has the advantage that in it the variables x, y, and z play COrn-
pletely symmetrical roles, and also we are readily led to a geometrical inter-
pretation of the equation. Equation (20) obviously can be written in terrns of a
dot product,
(Pi Qj + +
Rk) • Vu = O. (21)
Since Vu is a vector normal to the surface u = c, Equation (21) sta tes that the
vector Pi + Qj + Rk is perpendicular to the normal to that surface at any
point on the surface, and hence lies in the tangent plane. Thus we see that at
any point the vector Pi + Q j + Rk is tangent to a curve in the integral surface
u = C which passes through that point.
Thus the differential equation (15) may be considered as determining at any
point in sorne three-dimensional region a direction, specified by the vector
V = Pi + Q j + Rk. If a particle moves from a given initial point in such a way
that its direction at any point coincides with the direction of the vector V at
that point, a space curve is traced out. Such curves are called the characteristic
curves of the differential equation.
Tn a similar way, the ordinary differential equation of first order, dyjdx = f(x. y).
defines an angle rp = tan-¡ f(x, y) at any point in sorne region in the xy plane,
which the tangent to an integral curve must make with the x axis, and the general
solution of the equation is represented by ¡he set of curves having the prescribed
direction at any point in the two-dimensional region of definition.
We see next that any surface which is bui!t up frorn such characteristic
curves in space will have the property that the tangent plane at any point
contains the vector V = Pi + Qj + Rk, so that the normal vector at any point
of such a surface is perpendicular to V, as is required by (21). That is, any
smooth surface built up frorn characteristic curves is an integral surface of the
differential equation (15), provided only that its equation is not independent
of Z.
If r is the position vector to a point of a characteristic curve, and if s
represents arc length along the curve, then the unit tangent vector to the curve
at that point is given by
dr = dX i + dy. + dz k .
ds ds ds J ds
The requirement that this vector have the same direction as the vector
8.2. The Quasi-Linear Equation of First Order 389
Pi + Qj + Rk yields the conditions

dx dy dz
P = j1. ds ' Q = j1. ds ' R = j1.-' (22)
ds
where j1. may be a function of x, y, and z. These relations can be written in the
differential f orm
(23)
and hence are equivalent to two ordinary differential equations.

Let the solutions of two independent equations which imply (23) be denoted
by
u ¡(x, y, z) = el> u 2 (x, y, z) = el' (24)
where e, and e 2 are independent constants. Then these equations represent two
families of surfaces, such that a surface of one family intersects a surface of tbe
second family in a characteristic curve. If we consider the one-parameter family
of characteristic curves whicb are tbe intersections of those pairs of surfaces for
which e, and e 2 are related by an equation of tbe general form F(el> e 2 ) = O,
we see that the locus of these intersections will determine a surface which is an
integral surface, for any differentiable function Ft That is, any surface specified
by an equation of the form
F[u¡(x, y, z), u 2 (x, y, z)] = O (25a)
wil! be an integral surface of the partial differential equation (15) provided,
again, that F is differentiable and that the left-hand member of (25a) actual!y
depends upon z. In many cases [particularly, if u, does not involve z, so that
u, = e, is not an integral surface of (15)], it is more convenient to write the
solution in the alternative form
u 2 (x, y, z) = f[u,(x, y, z)]. (25b)
The procedure il!ustrated in Section 8.1 wil! confirm the fact that (25a) or
(25b) does indeed define the most general solution of (15), so that the result
can be summarized as follows:
The general solutíon of the equatíon
(26)
is of the form
F(u"u 2 )=0 or u 2 =f(u , ), (27)
where ul(x, y, z) = el and u 2 (x, y, z) = e 2 are solutions of any two independent
ordinary differential equations whieh imply the relationships
dx dy dz 28)
P=Q=R' (
tIn particular, ti I = e, itself is an integral surface of (I5) unless ti, is independent of Z, and
the same statement applies to the surface tl2 = e2.
The inlerseclion of any Iwo of Ihe surfaces u ¡ = c ¡ and l/o = c, is a characteristic

curve whose langenl al any poinl has Ihe direclion ratios (P, Q, R).
We notice that, since (28) is a cODsequence of (22), in case P, Q, or R is
identically zero we must take dx, dy, or dz, respectively, equal to zero. Thus,
the relations dx/P = dy/ Q = dz/O imply the two equations Q dx = P dy and
dz = O.
We may now establish the validity of the general result by an argument
which is independent of geometrical considerations. Let u¡(x, y, z) = Cl and
u,(x, y, z) = c, be solutions of two independent equations which imply (23).
Then since (23) implies (22), we have
0= du¡ = aU I dx
ax
+ au¡ dy
ay
+ au¡ dz
az
=(p au¡
ax
+ Q au¡
ay
+ R aU¡)dS.
az f.I.
Hence u = u ¡ and (similarly) u = Uz are solutions of (20). If follows also that,

if u¡ = c¡ and l/z = c, determine z as functions of x and y, these functions are
solutions of (15). We next re place the independent variables x, y, and z in (20)
by the new variables r = u¡(x, y, z), S = uz(x, y, z), and I = rp(x, y, z), where
rp is any function which is independent of U¡ and u, and which does nol satisfy
(20). There then follows
au = aliar + auas + aual = auau¡ + auauz + auarp,

ax ar ax as ax al ax ar ax as ax al ax
together with similar expressions for aulay and aulaz. With these substitutions
(20) takes the forrn
where u is now considered as a function of the independent variables r = U¡,

S = Uz, and I = rp. But since U¡ and Uz satisfy (20), the coefficients ofaular and
aulas vanish, and since rp does nol satisfy (20), Equalion (20) is Ihus equivalenl lo
Ihe equa/ion au/al = O. Hence the mosl general solution of (20) is u = F(r, s) =
F(u¡, u z ), where F is an arbitrary differentiable function. Thus, finally, the most
general solution of (15) is u = c or F(u¡, u,) = c, where c can be incorporated
into F, and hence can be replaced by zero, as was to be shown.
Example 1. We consider Equation (1),

2 az _ az = O (29)
ax ay ,
and deduce the solution (2) by the method derived aboye. In this case (28) becomes
dx dy dz
T = -) = O'
or dx + 2 dy = O, dz = O.
8.2. The Quasi-Linear Equation or First Qrder 391
The sol utio ns of these equatio ns a re

x + 2y = cI , z = eh
a nd hence, with f./ 1 = X + 2y, f./2 = z, Equation (27) gives the general solution of (29)
in the alter native forms
F (x + 2y, z) = O or z = f(x + 2y), (3 0 )
in accordance with (2). In this case the su rfaces f./ I = C I and f./ 2 = Cz are planes, and
the characteristic curves are the straight line s parallel to the xy plane, given by the
intersections of tbe planes x + 2y = e l with the planes z = Cz. The direct io n cosines
of these lines are proportional to the coefficients P, Q, R in (2 9), that is, to the values
(2, - 1, O). In this case the characteristic "curves" coincide with the ir tangents. Any
surface built up from straight lines of the type no ted is of the forrn (30), for som e choice
of f o r F, an d is an integra l surface of (29) unless it is a plane x + 2y = e. Jt should be
no ticed tha t any so lution Z of (29) is cons tan t a lo ng any line x + 2y = constant in
the xy plane. •
To generalize the re s ult of Example 1, w e may verify that the genera l so lu-
tion of the equation
(31)
where a and b are constants, defines all cylindrical surfaces with element s parallel
to the direction (a, b, O), and is o f the form
z = f(bx - ay) . (32)
lt should be n ot iced that thi s solu ti o n can also be expressed in o ther related
forms such as
z = F (ay - bx),
a nd so o n , if ex ceptions when a or b vao ish are taken int o account.

az az =
x-
ax +y-
ay z (33)
the associa ted equations become

dx dy dz
-=-=-,
x y z
an d solutio ns of two of them are obtained in the form
Z
Lx = el
'
-
x
=c."
-
The general soluti on of (33) then can be written in the form
Fe, ~) = o or z = xf( ~) (3 4)
392 Partial Dilferential Equations
or in several Qther equivalent fonns, such as
z = yC(;).
Since the characteristic curves are straight lines with the direction ratios (x, y, z), it
follows that any surface (34) contains the straight lines from the origin to points on
the surface, and hence is a conieal surface with vertex at the origino •
8.3. Special Devices. lnitial Conditions. Although the geometrie interpreta-

tion ofthe general solution ofan equation oftype (26) usually is readily obtained,
the determination of explicit solutions of two of the associated equations (28)
may present difficulties. In the linear case,
az az
P(x, y) ax + Q(x, y) ay = R Jx, y)z + R 2 (x, y), (35)
since P and Q are independent of z, the equation

dy Q
dx= P
involves only x and y. If it can be solved in the form ul(x, y) = el' then this
result may be used to express, say, y in terms of x and el' Then the equation
dz Rlz + R2
dx P
becomes a linear ordinary differential equation In z.t The solution of this
equation can be written in the form
U2 = ZIXI(X, el) + IX 2 (X, el) = e 2·
If C I is then replaced by its equivalent in terms of x and y, the result is of the
general form
U2 zp¡(x, y) + P2(X, y) = C 2·
The general solution of the partial differential equation then becomes U2 = f(u 1)
or
1 f[ ( )] P2(X, y)
z=PI(x,y) ulx,y -PI(x,y)'
Thus, the general solulion of a linear equation offirst order can be put in the form
z = SI (x, y)f[S2(X, y)] + s,(x, y), (36)
where SI' S2' and s) are speeifie fune/ions and f is an arbitary fune/ion,
Example l. In the case of the equation

az,
y-,. x- = z -
az 1 (37)
ax ay ,
tI! is proper to hold el constant in solving this equation, since the condition 1I1(X, y) = el is
to be combined with its solution to define a characteristic curve.
8.3. Special Devices. Initial Condilions 393
Equations (28) become

dx dy dz
y-=-X =Z=¡ '
Equality of ¡he first two members gives

y dy - x dx = 0, Y!_X2=C¡o
Equality of the first and third members then gives

dz dx z-] z-l
z-] ./x2+e,' --~~~~-- =
:c + ./Xl + e,
c! or x +y = e2'
Thus, with 11, = y2 - x2 and 1/2 = (L - I)/(x + y), the general solution of (37) be-
comes U2 = f(u,) or
z = (x + y)f(y2 - X2) + l. (38)
In other cases, if one integrable equation can be obtained by suitably

•
rearranging (28), a similar procedure can be followed. In particular, sinee the
equal ratios in (28) are also equal to the ratio
k, d:c + k 2 dy + k3 dz (39)
k,P + k 2 Q + k3R
where k" k 2 , and k 3 are entirely arbitrary, any one of the given ratios can be
replaeed by a new ratio of this formo If k" k 2 , and k 3 can be ehosen in sueh a
way that tl'te denominator of the new ratio vanishes and the numerator is an
exact differential du" then the corresponding requisite vanishing of the numera-
tor leads to the faet that one integral is ti, = e, .
Example 2. As a elassic example, we consider the equation

az az
(mz - ny) ax + (nx - Iz) ay = Iy - m x. (40)
The associated equations are

dx dy dz
mz - ny = nx - Iz = Iy - mx'
and no pairs are irnrnediately integrable. We next determine k" k 2 • and k3 so that the
ratio (39) has a zero denominator and hence also a vanishing numerator, and so write
k,(mz - ny) + k 2 (nx - Iz) + k3(ly - mx) = O.
If we require the coefficients of 1, m, and n to vanish independently, we obtain

k l = x, k 2 = y,
and if we require the eoeffieients of x, y, and z to vanish independently, we obtain

k, = 1,
394 Parlíal Dífferenlíal Equalions
Wilh these particular choices of (he k 's , we obtain the two additional equi valent ratios
x dx + y dy + z dz I dx + m dy + n dz
O O ·
Since the numerators happen to be the exact differentials d(x 2 + y2 + z2)/ 2 and
dUx my + nz), their vanishing leads immediately to the integrals
ll¡ = Xl + y2 + Z2 = el > U2 = Ix + my + nz = C2,
and hence the general solution of (40) can be written in the form
F(X2 + y2 + z2, Ix + my + nz) = O. (41)
The characteristic curves in this case are the circles in space determined as the inter-
sections of the spheres "I = el and the planes U2 = C2. •
In the case of an ordinary differential equation of first order, the arbitrary

constant can, in general, be determined so that the integral curve passes through
a specified point in the xy planeo In a similar way, the arbitrary function in the
general solution of a first-order partial differential equation can be determined,
in general, so that the integral surface inc1udes a specified curve in xyz space. If
the equation of the curve is given by the pair of equations
tpl(X, y. z) = O, tp2(X, y, z) = O, (42)
that is, as the intersection of the two surfaces tp I = O and '1' 2 = O, and if the
solutions of two of the associated equations (28) are written in the form
ul(x,y, z)=c l , u,( x,y,z)=C2> (43)
the determination of the function F in the general solution (27) is equivalent
to the determination of a functional relationship between C I and C 2 in (43) such
that (42) and (43) are compatible. Thus, if the elimination of x, y, and z from
the four equations involved in (42) and (43) leads to an equation of the form
F(c l , c,) = O, the required integral surface is given by F(ul> u,) = o.
Example 3. We require the solution of (37) which passes through the curve
z=x 2 + y+l, y=2x . (44)
Firs/ method. We eliminate x, y, and z between these equations and the equations
z-I
y2_X 2 =CI, --=c."
x+y -
The result of the elimination is
C,
and hence the desired solution is

_ x +
z - 3
Y(.../ y2 -~3
X2 + 2) 1. (45)
Second me/hod. If we take x as the independent variable along the curve, we have,
from (44), y - 2x, z = ( x + 1)2, and the introduction of these results directly into the
r
8.3. Special Devices. ¡nitial Conditions 395
general solution (38) gives

x" + 2x = 3xf(3x 2 ), f(3x') = ~ + ;.
Ir we write u = 3x 2 , this equation becomes
"';u 2
f(u) = 3"';3 + '3
aod so determines the function fin (38) in accordance with the resuJt given by (45).
Third me/hado In an equivalent procedure, which may be a bit longer, but in which the
steps are more easily interpreted geometrically, we first determine the characteristic
curve passing through [he point PoL~o, Yo, zo), in the form
z - 1 Zo - 1
= ,
x + Y Xo + Yo
assuming that Xo + Yo =1= 0, then require that Po lie on the specified curve (44), so that
Zo = x¡l + Yo + 1,
and hence
Y o = 2xo ,
Thus the charac teristic curve passing through a point Po on e is defined by the simul-
taneous equations
z _ 1 = Xo 3+ 2(x + y.)
The requ ired surface is the locus traced out by this curve as Xo varies through all real
values, and is obtained in the form (45) by eliminating Xo. •
It should be noticed tbat if a unique integral surface is to be determined,

the prescribed curve cannat be taken as a characteristic curve, since, in general,
injinítely rnany integ ral sl/rfaces include a specified characterislic curve. Further,
in general there exist exceptional curves through which no integral surface passes
(see Section 8.7).
For example, in the case of (37), if the prescribed curve e projects onto the
straigbt line y = x in the plane z = O, then along that curve (38) becomes
Z = 2xf(0) + l.
Consequently, unless al so z = A x + I alo ng e, where A is a constant, there is
na integral surface passing through that curve. If e is defined by the require-
men! that z = Ax +
l when y = x, the above condition becomes
A = 2f(0)
and is satisfied by any function f for which feO) = A / 2, so that for any such f
which is differentiable tbe surface (38) passes through the curve C. In trus last
case, we verify tha! the prescribed curve is indeed a characteristic curve (11 1 = O,
u, = A / 2).
The procedures given above can be ex tended to the solution of analogous
equations in which there are more than two independent variables. Thus, for
example, if x, y, and tare independent variables, the general solution of the
quasi-linear equation
(46)
is of the form
(47)
where u¡(x, y, z, t) = ej> u 2 (x, y, z, t) = C and ¡¡,(x, y, z, t) = c, are indepen-
dent solutions of three of the associated "ordinary equations
(48)
8.4. Linear and Quasi-Linear Equations oC Second Order. The general

quasi-linear equation of second order, involving two independent variables x
and y, is of the form
(49)
where a, b, e, and F may depend upon x, y, z, az/ax, and az/ ay. In particular,
when a, b, and c are independent of z, az/ ax, and az/ ay, while F is a linear
function of those quantities, the equatio!1 is linear in z. Such an equation thus
is of the form
(50)
where the functions a, ... , g depend only upon x and y.

It happens that the terms involving seeond derivatives are of principal
signifieanee. In faet, a very important role is played by the sign of the discrimi-
nant b 2 - 4ac. From analogy with the terminology associated with conic sec-
tions, we say that (49) or (50) is of hyperbolie type when b 2 > 4ae, of elliptic
type when b 2 < 4ac, and of parabolic type in the intermediate case, when
b 2 = 4ac. The importance of this classification will be indicated in later sections
with reference to linear equations, to which our attention will be restricted in
most of what follows .
In the linear case, it is clear that, as in the case of ordinary differential
equations, the most general solution of (50) consists of the sum of any particular
solution of (50) and the most general solution of the equation obtained by
replacing the right-hand member of (50) by zero. This second solution is some-
times called the eomplementary solution of (50).
It is seen also that any linear combination of two complementary solutions
will also be a com~lementary solution. However, such a combination cannot
be expected to be tfe most general complementary solution, since the general
solution should involve arbitrary jimctions. It might be expected that the most
general solution of (50) would be a generalization of the form (36),
z = s¡(x, y)f[S2(X, y)j + s,(x, y)g[s.(x, y)j + s,(x, y), (51)
8.5. Special Linear Equations of Second Order, with Constant Coefficients 397
where s" ... ,s, are specific functions and f and g are arbitrary functions.
However, this condition exists only in special cases. That is, the most general
complementary solution of(50) cannot always be written as a sum of terms involv-
ing arbitrary functions. (For an example establishing this assertion, see Problem
22.)
In Section 8.5 we consider an important special class of linear equations
in which the general solution is of simple form, and in Section 8.6 we illustrate
other special types of linear equations.
8.5. Special Linear Equations oC Second Order, with Constant Coefficients.

We here consider equations of the very special form
a 2z a 2z a 2z
a ax2 + b ax ay + e ay2 = O, (52)
where a, b, and e are constants.t To obtain the general solution of (52), we
assume a solution of the form
z = f(y + mx), (53)
where f is an arbitrary twice-differentiable function and m is a constant, and
attempt to determine values of m for which (53) is a solution of (52). By differen-
tiation, we obtain from (53) the expressions
2 2
a z2 = mJ2.f""( y
ax + )
mx, a ay
ax z = mJ,.f""( y + )
mx, aay2z =
2
f "y( 'T mx,
) (54)
and the introduction of (54) into (52) gives the condition

am 2 + bm + e = O (55)
which must be satisfied by m.
In general, (55) will determine two distinct values of m, say m, and m 2.
Then, beca use of the linearity of (52), it follows that any express ion of the form
z = f(y +
+ g(y + m 2x)m,x) (56)
is a solution of (52). Further, if the new variables s = y + m,x and t
= Y + m 2x
are introduced, it is readily verified that (52) takes the form a 2z!(as at) = O,
from which it follows that (56) is indeed the most general solution of (52).
Example l. For the equation

a 2z _ 3 a 2z + 2 a2z = O (57)
ax2 ax ay ay2 '
the determinant equation becomes
m2 - 3m +2 = O,
tEquations of lhis general type, in which al! terms involve derivatives of the same order, are
often called "homogeneous equations." However, we will conlinue lo use Ihis designalion onJy
as il is defined in Seclion 5.1.
398 Partial Differential EquatiODS
and the general solution is

z = f(x + y) + g(2x + y). (58)
If a = ° and b =1= 0, only one solution of Equation (55) is obtained. How-

•
ever, in tbis case it is clear by inspection that the second term in the general
solution is then an arbitrary function of x alone. If a = b = 0, the general
solution is clearIy z = f(x) + yg(x). In these and other cases, the alternative
assumption of a solution of the form z = f(x + ny) may be convenient.
The remairung exceptional case is that in which (55) is a perfect square, so
that the two roots are equa!. The second term in the general solution can be
found by a method similar to that used in analogous cases in Chapter l. If
mi and m! are distinct roots of (55), then the expression
h(y + m 2 x) - h(y + m¡x)

m2 - mi
is clearly a solution of (52). As m 2 ~ mi, tbis solution approaches the limiting

value
Thus, if we write g = h', the general solution in the case of equal roots can be
taken in the form
z = f(y + m¡x) + xg(y + m¡x). (59)
Example 2. The general solution of
is found in this way to be

z = f(x + y) + xg(x + y).
•
It may be noticed that since we can write
xg(y + mlx) = m¡
_1 [(y + m¡x)g(y + m¡x) - yg(y + m¡x)j
if m ¡ =1= 0, the second solution can equally well be taken in the forrn yg(y + m¡x)
unless m, = O.
It is seen that the solutions of (55) will both be real if b 2 > 4ac, so that (52)
is hyperbolic, and will be conjugate complex if b 2 < 4ac, so that (52) is elliptic.
The intermediate case, when b 2 = 4ac, so that the equation is of parabo/ic type,
is that in wbich (55) is a perfect square.
8.5. Special Linear Equations of Second Order, with Constan! Coefficients 399
Thu s Laplace's equation

ao z ao z
axo + ay2 = O (60a)
is elliplic, with general solution of the form

z = f(x + iy) + g(x - iy), (60b)
the equation
-
az
2
az
2
ax2- -ay2
-O - (6Ia)
is hyperbolic, with general solution

z = f(x + y) + g(x - y), (61 b)
and the equation
a 2z a 2z a 27
ax2 - 2ax ay + ay~ = O (62a)
is parabolic, with general solution

z = f(x + y) + xg(x + y). (62b)
A procedure completely analogous to that giyen aboye can be applied in
obtaining the general solution of a linear equation of any order N, in which each
term is a constaat multiple of a derivative of order N. In s uch a case, the assump-
tion of a solution of form (53) leads to an equation analogous to (55), but of
Nth degree in m. If N di sti nct roots exist, the general solution is ofa form similar
to (56) but involying N independent terms. If a root m, is repeated r times, the
part of the solution corresponding to the r equal roots is readily shown to be
of the form
f,(y + m,x) + xfz(y + m,x) + ... + x'-'f,(y + m,x). (63)
Occasionally it is feasible to solve a problem, goveroed by a partial differ-
ential equation together with appropriate side conditions, by obtaining the
general solution of the equation and determining the relevant arbitrary func-
tion s by imposing the si de conditions, in analogy witb the usual procedure
relating to ordinary differential equations .
Perhaps the best-knowa si tuation of this type is that in which the solution
rp(x, t) of the "one-dimensional wave equation"
a 2 rp _ c zazrp (64)
~- ax 2
is required, for all real values of x and l, subject to the conditions
rp(x, O) = F(x) , rp,(x, O) = G(x), (65)
prescribed for a// real x when l = O. Here e is a positive real constant. This
problem is often called tbe Cauchy prob/em for Equation (64).
400 Partía! Differential Equations
Ir the general solution of (64) is written in the form

<p(x, t) = f(x + et) + g(x - et), (66)
the conditions (65) require thatfand g satisfy the conditions
f(x) + g(x) = F(x), e[f'(x) - g '(x)] = G(x),
for al! values of x. The elimination of g(x) yields
f'(x) = +r(x) + 2IeG(x), f(x) = 2l F(x) l IX G(';) d'; + e,

+"2e o
where e is an arbitrary constant. The substitution of this result and the corre-
sponding expression for g(x) into (66) gives the desired solution
<p(x, t) =
l
2[F(x + et) + F(x - et)] + 2el f
x-a
x
+
a
G(';) d'; . (67)
This solution is associated with the name of d'Alembéit. A physical inter-

pretation of the problem and its solution is included in Section 9.12, and a
generalization is obtained in Problem 37.
Analogous solutions of certain other problems governed by (64) are obtained
in Problems 33 and 35.
8.6. Other Linear Equations. As a simple example of other solvable linear

equations of second order, we consider the equation
a'z -
ax' z = O, (68)
where x and y are the independent variables. Since y is not involved explicitly,
we may integrate (68) as though it were an ordinary equation, holding y constant
in the process, and hence replacing the two arbitrary constants of integration
by arbitrary functions of y. Thus the general solution of (68) is of the form
(69)
However, even though the coefficients in the general second-order linear
equation (50) be constants, it is not possible in all such cases to obtain general
solutions of similar forms. The fol!owing procedure is frequently useful.
From the analogy with the ordinary equations, we are led to expect that
exponential solutions of the equation
a' z + b aa az + e-a
a-a a'z + daz- + ey-
2
az + fz = 0, (70)
x 2
x y y aX ay 2
with eonstant eoeffieients, may be of importance. If we assume a solution of

the form
(71)
where A, IX, and pare unspecified constants, the introduction of (71) into (70)
gives the condition
arx' + brxp + ep2 + drx + ep +f= O (72)
8.6. Other Linear Equations 401
to be satisfied by a and /3. This equation will, in general, determine /3 as either

of twO functions of (1., say
/3 = {i'¡(a), (73)
It then fo!lows that any expression of either of the forms
will satisfy (70) for arbitrary values of A, B, and (1., and hence wil! be aparticular
solution of (70). The same is true for any linear combination or suitably con-
vergent infinite series of such solutions, or for aD integral superposition of the
form
S.,"' (A«(1.)e ox+ .,ca) y + B«(1.)e ax +.O<a)y] da
when A«(1.) and B«(1.) are sufficiently respectable.

As is next illustrated, this procedure leads to a precise form of the general
solution if (73) expresses /3 as linear functions of (1., as is the case when (70)
reduces to (52) and, more generally, when the condition
ae 2 + fb + 2
cd 2 = bde + 4acf (74)
is satisfied. The algebraic equation (72), associated with (70), represents a conic
section in the (1./3 plane, and is a hyperbola, ellipse, or parabola, according as
b 2 - 4ac > O, < O, or = O, respectively. In the special cases just mentioned,
the conic sections degenerate into two straight Iines.
Example 1. We consicter the equation

az_ az 2 2 _ az + az = O (75)
axz ay2 ax ay ,
for which the condition (74) is satisfied. The assumption (71) leads to the equation
a2 - /32 - (1. + /3 = ((1. - /3)((1. +" /3 - J) = O
and hence (71) is a soJution of (75) if
/3 = a or /3 = -(1. + 1.
Thus, by superposition, any expression of the fonn
z = ~ A((1.)e a(x+ Y ) + ~ B((1.)e" ·c, -y) (76)
• •
is also a solution, the notation indicating that (1. may take on arbitrary values in the
summation. But since the coefficients A and B are also arbitrary, the first ter m may be
expected lO represent an arbitrary respectable function of x + y, and the second ter m
to represent the product of e' and an arbitrary respectable function of x-y. Thus
we may suspect Ihat the general solution can be written in the fonn
z = f(x + y) + eYg(x - y), (77)
where f and g are twice differentiable, and this can be verified by direct substitutioo .
•
402 Partíal Differential Equations
The sort of simplification achieved in this example clearly is always possible

when the left-hand member of (72) contains distinct linear factors, and the
special case of a repeated linear factor is easily dealt with by use of a method
similar to that which led to (59).
In other cases, whereas this procedure usually does not yield the general
solution, it may provide a useful set of particular solutions which (as will be
seen in Chapter 9) is sufficiently extensive for the treatment of a class oC specific
problems.
Example 2. For the two-dimensional Helmholtz equation,

a2
ax2
z + a z
2
ay2'
..L k 2
z
= O
, (78)
the assumption (71) leads to the equation

a2 + ft2 + k2 = O,
and hence
Any expression of the forrn
is then a solution of (78), for any choice of A, B , and a. If we write

A +B = C, i(A - B) = D,
this expression takes the real forrn
eax(C cos"¡ a 2 + k2 y + D sin ,.¡ a 2 + k2y).
Any combination of expressions of this sort, as a, C, aod D take on arbitrary values,
will also satisfy (78). Finally, aoy expression of the form
z = f~~ [C(a) cos"¡a'2 + k2y + D(a) sin "¡a 2 + k2y]e U

da, (79)
for which the iodicated integral is suitably convergent, will be a particular solution of
(78). (See also Problems 38 and 39.) •
When a partial differential equation has variable coefficients, it is rarely

possible to obtain its general solution . However if, for example, the equation
2
aa (pz) + b a2 (pz) + c a2 (pz) _ O (80)
ax2 ax ay ay2 - ,
where a, b, and e are constants and p is a function of x andjor y, the general

solution clearly expresses z as the product of Ijp and a combination of arbitrary
functions. (See Problem 42.)
Example 3. We find that the equation

a2rp +.i...~ + 2.2rp _ J...~ (81)
ar2 r ar r 2 2- a at
8.7. Characteristics of Linear First-Order Equations 403
also can be wri tten in the form

1 aZ(r lp) =
a 2(r zrp) _ _ o
ar z cx,2 al" '
so that its general solution is
I
rp = "2 [f(r
r
+ cx,1) + g(r - cx,1)] (82)
when cx, 7"= O.
•
8.7. Characteristics of Linear First-Order Equations. In Section 8.2 we have
defined a characteristic curve of the first-order linear equation
(83)
where P, Q, Rp and Rz are functions of x and y only, as a curve III space

whose tangent at any point has the direction ratios (P, Q, R,z R 2 ). Any +
such curve is the intersection of two surfaces of the form
u,(x, y) = C p u 2 (x, y, z) = c2, (84)
where u ¡(x, y) = c, is an integral of the ordinary equation
dx dy
P-Q (85)
and u 2 (x, y, z) = Cl is an independent integral of one of the associated equations

dx dy dz
(86)
P= Q = R,z+R z
Thus any characteristic curve of (83) is the intersection of the cylinder
u,(x,y)=c" (87)
with elements parallel to the z axis, and a second surface of the form u/x, y, z)
= C z . It follows that any characteristic curve can be specified by first choosing
a particular cylinder of the family (87), and so satisfying (85), and then pre-
scribing z on this cylinder in such a way that another independent equation
obtained from (86) is satisfied. We shall find it convenient to speak of any
cylinder obtained by specializing c, in (87) as a characteristic cy/inder. A1so,
we refer to the intersection of this cylinder with the xy plane as a characteristic
base curve. Such a curve is clearly the projection of a characteristic space curve
onto the xy planeo
Example 1. In the case of the equation

az az (88)
y ax - x ay = y,
two associated equations can be written in the form

dx dy
y
-- , dz = dx,
x
with solutions
ll¡ = X2 + y2 = el,
Thus the characteristic curves are ellipses in space determined as the intersections of
the right circular cylinders X2 + y2 = Cl and the planes z - x = e2' The cylinders
X2 + y2 = c , are the characteristic cylinders, and the circles X2 + y2 = el in the xy
plane are the characteristic base curves, which are the projections of the characteristic
ellipses onto the xy planeo Along this base curve we can express x and y in terms of a
single variable A by writing
x = .ve;- cos A, (89a)
y = .vel sin A. (89b)
If along this curve we define z by ao equation of the form
= x + e2 = .ve;- cos A + e2,
z (89c)
the corresponding locus in space is a characteristic curve. Each characteristic cylinder
is seen 10 include infinitely many characteristic curves. •
We now reconsider, from a somewhat different point of view, the question

discussed in Section 8.3, as to whether an equation of form (83) has as a par-
ticular solution an integral surface which ineludes an arbitrarily prescribed
curve ID space.
Along any curve in space, the coordinates x, y, and z can be considered as
functions of a single variable, say A. Thus the curve can be specified by three
equations of the form
c: x = X(A), y = Y(A), Z = Z(A). (90)
The variable A may be taken as arc length along the curve, for example, or it
may be identified with one of the coordinates x or y. The projection C o of this
curve onto the xy plane is given by
Co:x = X(A), y = Y(A) (91)
and Z = O. Thus (90) can be considered as specifying Z along the curve C o in
the xy plane in terms of a parameter A which varies along C o .
Suppose then that Z is prescribed along an arbitrary curve C o in the xy
plane, in accordance with (90), and that X(A), Y(A) have continuous first deriva-
tives and Z(A) continuous derivatives of all orders with respect to A. We inquire
whether (83) has a solution expressing Z as a function of x and y everywhere in
a region ineluding C o in such a way that Z takes on the prescribed values on C o .
This query is elearly equivalent to asking whether (83) possesses an integral sur-
face which ineludes the space curve (90).
Let us assume that such a solution does exist and that the solution
Z = f(x, y) can be expanded in a senes of powers of x and y, about a point
(x o, Yo) on C o , of the form
z = f(x o, Yo) + af(a~ Yo)(x - x o) + af(~~ Yo)(y - Yo) +
= (z)(x,.y,) + (~Z)
x (xo,Yo)
(x - x o) + (~yZ) (xo.Yo)
(y - Yo) + (92)
8.7. Characteristics of Linear First-Order Equations 405
Then the solution is determined in the neighborhood of a point on Ca if we

know aH its partial derivatives at that point.
Jt may be reealled that in Chapter 3 we made use of sueh ideas to caleulate the
series solution of an ordinary differential equation of the form
dy
d.x: = F(x, y)
for which y is preseribed as Yo when x = xo. The differential equation then gives
dy/dx when x = x o , and by successive differentiation all higher derivatives are
generally obtainable. We then have the required solution
y(x) = y(x o) + y'(xo)(x - xo) + y"r o\ ... - xo)2 +
if suitabJe eonvergenee is assumed.
On the curve Ca' the coeffieients P, Q, R" and R z , as weH as z itself, are
now known funetions of )., and henee (83) gives one equation involving the
values of the two first partial derivatives of z at points on Ca'
P()')~~ + Q().)~; = R,().)z().) + R z().), (93)
where, for example, P().) has been written for P[x().), y().)]. To obtain a second
equation, we notiee that, sinee z is known on Ca' its derivative dzjd). along C o
is also known, and it must satisfy the equation
dz().) _ az dx + dz dy .
d)' - axd). ay d)'
Thus a second equation complementing (93) is of the form
dx az + dy az _ dz. (94)
d).ax d). ay - d)'
Equations (93) and (94) can be solved uniquely for the unknown quantities
azjax and azjay everywhere on C o if and only if the determinant of their coef-
ficients is never zero,
P().) Q().)
dx dy = P().):t - Q().) ~1 =;i= o. (95)
d). d).
But this condition is equivalent to the restriction
dx().) =;i= dy().) (96)
PO) Q().) ,
and hence (95) requires that the curve Ca along which z is prescribed never be
tangenl lo a charaClerislic base curve. If (95) is satisfied at aH points on Ca' the
values of azjax and azjay then can be calculated by (93) and (94) at aH points
on Ca.
To calculate the second-order partial derivatives of z on Ca' we may differ-
entiate (83) with respect to x, and so obtain the equation
pazz +Q azz = R az + aR,z + aR z _ apaz _ aQaz, (97)
axz ax ay 'ax ax ax ax ax ax ay
where all quantltles but a 2 z/ax2 and a 2z/(ax ay) are now known functions of
A on Ca. A second equation involving these unknown quantities is obtained by
differentiating the known function az/ax along eo, with respect to A,
dxa 2z dy a 2z d(aZ)
dA ax2 + dA ax ay = dA ax . (98)
Equations (97) and (98) determine a 2z/ax2 and a 2z/(ax ay) uniquely in terms
of known functions if (95) is true, and the remaining derivative a 2z/ay2 then
can be calculated either from the equation obtained by differentiating az/ay
with respect to A or from that obtained by differentiating (83) with respect to
y. This process can be continued indefinitely, to determine the values of all
partial derivatives of z on e o in such a case and, assuming convergence of the
series (92), it follows that the function z is determined uniquely for values of x
and y in a region about e o by its prescribed values on e o if Ca is never tangent to
a characteristic base curve. This assertion is equivalent to the statement that if
e is never tangent to a characteristic cylinder of(83), there exists a unique integral
surface of (83) which includes e.
If, however, (95) is violated for al! A on eo, so that e is on a characteristic
cylinder, then (93) and (94) are incompatible unless they are equivalent, that is,
unless
dx(A) dY(A) dZ(A) (99)
P(A) = Q(A) = R 2(A) + zR ,(A)
In this case one of the two equations (93) and (94) implies the other, and hence
only one restriction on the two derivatives is presento Thus the partial deriva-
tives az/ax and az/ay are not determined uniquely in this case but may be chosen
in infinitely many ways. But this result is to be expected, since if (99) is satisfied,
e is then a characteristic curve, which does indeed lie on infinitely many integral
surfaces. Thus it follows that if e lies on a characteristic cylinder of (83), there
is no integral surface of (83) which includes e unless e is a characteristic curve,
in which case there exist infinitely many such surfaces. It follows also that if e o
is a characteristic base curve, then z cannot be prescribed arbitrarily on it but
must be taken in sllch a way that e is a characteristic curve. In such a case the
value of z at a point near e o is not determined by the values of z on eo'
Thus, in the case of Equation (88), there exists a unique integral surface
induding any space curve whose projection onto the xy plan e is not a cirde
with center at the origino The only curves whose projections are of this type
and which lie on integral surfaces are the ellipses which are characteristic
curves, and these curves each lie on infinitely many integral surfaces.
When (95) is violated only for certain isolated values of A, so that e o is
tangent to a characteristic base curve at each corresponding point, the relevant
solution z(x, y) may behave exceptionally at each such point of tangency.
We next show that if the equation (87) of the characteristic cylinders of the
(lOO)
8.7. Characteristics oC Linear First-Order Equations 407
is known, the equation can be reduced to a normal ("canonical") form by a

suitable change in variables. If P _ O, Equation (100) already is in the desired
formo Otherwise, we take as a new independent variable the expression
rp = U (x, y),
¡ (101)
where u ¡ (x, y) = c ¡ is a solu tion of (85), and retain x as the second independent
variable, since then x and u ¡ assuredly are functionally independent. There
then follows
( aZ)
ax y =
(aZ)
ax. + (aZ) au¡
arp ax ' x
(U) x = (~~) x aa~¡ ,

and hence, with z now considered as a function of x and rp, Equation (100)
becomes
paz
ax
+ (pau¡
ax
+ QaU¡) az =
ay arp
R
¡
z +R 2
, (102)
where P, Q, R¡, and R 2 now are to be expressed as functions of x and rp. But
since U¡ = C I is an integral of(85), the function U¡ satisfies the partial differential
equation
pau I + Qau¡ = O.
ax ay
Hence the coefficient of az/arp in (102) vanishes, and the equation takes the form
p
az
ax = R¡z +R (l03)
2,
when z is considered as a function of x and rp.t
Example 2. In the case of Equation (88),
y az
ax - x az
ay = y, (104)
we write
rp = X2 + y2, y=../rp-x 2 •
Then (l03) shows that (88) can be put into the simpler form
../rp--....,.az
- X2 ax =../rp - x 2 ,
with z considered as a function of x and rp. Since ../ rp - X2 does not invo1ve z, it may
be canceled, and the equation becomes
az = 1 (105)
ax '
tThe need for such a qualifying phrase (particularly here, but also elsewhere) would be avoided
by using a new symbol Z for the dependent variable when it is considered as a function of the
new independent variables, so that Z(x, rp) = Z[x, U ¡ (x, y)] = z(x, y). (See also Section 7.1.)
408 Pnrtial Oifferential Equations
with ¡he obvious general solution

z = x + [(rp) = x + [(X2 + y2).
This resul! is equivalent 10 the form l/2 = [(u¡), with the notation folJowing (88), as
is to be expected.
In the more general case of a quasi-linear equation

•
az az
P(x, y, z) ax + Q(x, y, z) ay = R(x, y, z), (106)
an in tegral ti ¡ e ¡ of the equation

dx dy
P=Q (107)
generally cannot be obtained witbout reference to anotber equationobtained

fram tbe relations
(108)
and, in any case, ti ¡ generally will not be independent of z. Tbus here tbe concept
of a charaeteristie cylinder generally is no longer meaningful. However, it still
remains tfUe tbat, if the condition (107) is never satisfied on a specified curve
C, tben there is a unique integral surface of (106) which contains C. If (107)
is satisfied everywhere on C, then tbere is no integral surface containing C
unless (108) is satisfied on C, in whicb case C is a characteristic curve and there
are infinitely many integral surfaces containing C.
8.8. Characteristics of Linear Second-Order Equations. We next investigate

in a similar way tbe problem of determining solutions of linear second-order
partial differential equations satisfying apprapriate initial conditions. The most
general linear second-order equation is of the form
a2z
+ b a-axazy + e-a
a z az az
2 2
aa x 2 y + da-
x + e a- y + [z =
2 g, (109)
where the coefficients and the right-hand member may be functions of x and y.
In the case of ordinary equations of second order, the initial conditions prescribe
a point in the plane, thraugh which the integral curve is to pass, and also
prescribe the slope of the integral curve at that point. This is equivalent to
prescribing the values of the dependent variable y and its derivative dy/dx
corresponding to a given value of the independent variable x. In the case of a
partial differential equation, the analogous initial conditions prescribe a curve
C in space which is to lie in the integral surface, and also prescribe the orien-
tation of the tangent plane to the integraL surface along that curve. These con-
ditions are equivaLent to conditions which prescribe the values of z and its two
partial derivatives az/ax and az/ay along the projection Ca of the curve C onto
the xy plane. However, the values of z, az/ax, and az/ay cannot be prescribed in
8.8. Characteristics oC Linear Second-Order Equations 409
a completely independent way if :: is to be differentiable along the curve

since if A is a parameter specifying position aloDg Co. we mus t have
dz _ a z dx ..l.. azdy.
dA - aXdA ' aydA ( 110)
Thus, for example, if C o is the x axis and if z is prescribed as f(x) along Ca,
then the derivative az/ax along C o canDot also be prescribed but must be given
as f'(x). However, the derivative az/ay normal to Ca can be independently
prescribed. This limitation is in accordance with (110), which here becomes
f'(x) = ~ . l + ~; . O
ir we identify A with the distance x along Co.
A curve C in space, together with values of az/ax and az/ ay prescribed
along e in such a way that (110) is sa t-
isfied, is called a strip, and Equation (110) z
is referred to as the strip eondition. If it is
recalled [see Equation (103), Chapter 6]
that the tangent plane to a surface
z = z(x, y) has as its normal a vector
with direction ratios (az/ax, az/ay, -1),
we can think of the prescribed val ues of
az/ax and az/ay at points along e as
determining the normal direction to dif-
I y
~
ferential elements of surface area at
these points on the required integral sur-
x
face . The strip condition (l10) requires
that ¡he elements join together in a Figure 8.1
regular way (Figure 8 .1 ).
Suppose now that a curve Ca in the xy plane is given by the equations
x = X(A), y = y(A), (111)
where X(A) and y(A) have continuous derivatives, and that z, az/ax, and az/ay
are prescribed as functions of A. having derivatives of all orders at all points of
ea, in such a way that (I lO) is satisfied. For brevity, we introduce the conven-
tioDal abbreviations
az az
p = ax' q = ay' ( 112)
Then the differential equation ( 109) becomes

ar + bs + el + dp + eq + fz = g, (113)
and the s trip condition takes the form
dz _ dx dy
dA. - p dA + q dA. . (114)
410 Partial Differential EquatiODS
The prescribed conditions along C a are then

Z = z(A), P = peA ), q = q(A) , ( 115)
where z(A), p(A), and q(A) satisfy (114).
As in Section 8.7, we now attempt to calculate the values of the higher
derivatives of z along Ca in terms of the prescribed values , If aH such deriva ti ves
can be determined and if z can be expanded in a power series in x and yabout
points on Ca, then an integral surface satisfying the prescribed conditions is
determined for values of x and y in a region including Ca, and the value of z
at a point near Ca is determined in terms of the known values on Ca by use of
this series.
The first problem, then, is to a:tempt to determine the values of the second
derivatives r, s, and 1 for points on Ca. One condition involving these unknown
quantities is given by (113),
ar + bs + el = g - (dp + eq + fz). ( 116)
If we notice that, along Ca ' there follows also
fA (~:) = Ix (~:) ~; + Jy (~:) ~ ,

fA (U) = :x(~;)~; + :Y(U)~'
we thus obtain two additional conditions involving r, s, and t, of the form
dx dy _ dp
dA r + dA s - dA' ( 117)
dx dy _ dq
dA S + dA 1 - dA' ( 118)
Equations (116), (117), and (l18) determine r, S, and t uniquely if and only if
the determinant of their coefficients is not zero,
a b e
dx dy 2 2
O _ (dy ) dx dy (dx) O
dA dA - a dA - b dA dA + e dA 0;1=. (119)
dx dy
O
dA dA
If (119) is satisfied for all points on Ca, then (116), (117), and (118) deter-
mine r, s, and 1 everywhere along Ca. If we proceed by a method similar to
that used in Section 8.7, we then find that all higher partial derivatives of z
also can be calculated at points on Ca if only (119) is satisfied. In such a case
we may say that the strip consisting of the curve C in space and the associated
values of az/ax and az/ay along C, satisfying (110), is aproper strip, in the sense
that there exists a unique integral surface of (109) whieh inc/udes Ihis strip.
8.8. Characteristics of Unear Second-Order Equations 411
Now suppose that at al! points of C o the determinant in (119) is zero,
a(~r - b~l ~ + c(~lr = O. ( 120)
Then (116), (117), and (118) cannor be sol ved for s unless the numerator de ter-
minanr in the formal solution by Cramer's rule also vanishes,
a g - (dp + eq + jz) c
dx dp
O
dA dA = O,
dq dy
O
dA dA
dp dy dq dx dx dy
or a dA dA + c dA dA + (dp + eq + jz - g) dA dA = O. (121)
In rhis case generally two of the three equations imply the other one,t and
hence only rwo restricrions on the tbree unknown quantiries r, s, and tare
present. Then, if (120) and (121) are both satisfied, Equations (116) to (118)
serve merely to express two of the second derivarives in terms of the third (or
to determine only rwo of rhose derivatives) and the remaining one is yet to be
determined (or partially derermined). Finally, if (120) is sarisfied but the left-
hand member of (121) is nol zero, no solution exists. Equation (121), or its
replacement when it is nonresrrictive, is known as rhe compatibility condition.
If (120) is satisfied at aH points on Co, then at all such points the slope
dy/dx of the curve in the xy plane must satisfy the equation
2 dy
a (d y ) _ b + c = O. (122)
dx dx
That is, we must have
dy
-=
b ± ,.Jb 2
- 4ac , ( 123)
dx 2a
and hence the quantily b 2 - 4ac musl be nonnegative. Suppose rhar b 2 - 4ac
is positive and a is not zero. Then (123) determines two families of curves in
the xy plan e, say
rp(x, y) = el> (124)
That is, if (120) is sarisfied , rhe curve Ca must be a member of one of these
families. In su·:h a case no integral SIIrjace including the prescribed strip exists
unless along lhe curve C (which projects ioto Ca) z and irs partial derivatives
satisjy a compatibility condition. When that condition ís also satisfied, at least
one integral surface including the prescribed strip exists, and rhe strip is called
a characteristic strip. The curve C in space which bears the strip ís called a
tExceptiooal cases where both a aod dx/d)' or both e and dy/d)' vaoish 00 Ca must he treated
separately. In each of these two situatioos, (121) is idenlically satisfied aod a different conditioo
relatiog p and q 00 Ca is obtained. (See Problem 53 for an example.)
412 PartiaI DifferentiaI Equations
characteristic curve, and its projection C o in the xy plane, which must be a

member of one of the families of (124), is called a characteristic base curve.
Equations (124) determine two families of cylinders in space, with generators
parallel to the z axis, which may be called the characteristic cylinders. Sorne
writers refer to the projection C o itself as a "characteristic" of the differential
equation.
If a = O, Equation (120) can be solved for dx/dy and two families of charac-
teristic cylinders again are obtained unless also c = O. If a = c = O, the cylin-
ders reduce to the planes x = c 1 and y = c 2 . If b 2 = 4ac, so that (109) is of
parabolic type, the two families coincide.
From Equation (123) we see that an elliptic linear equation of second order
has no real characteristic curves, since for such an equation the discriminant
b 2 - 4ac is negative. Thus an integral surface of an elliptic equation can always
be determined so as to inelude a prescribed regular strip in space. Equivalently,
if z and its first partial derivatives are prescribed in a sufficiently regular way
along any curve in the xy plane, the solution of an elliptic equation which takes
on these values is determined uniquely in the neighborhood of tbis curve.
The situation in the case of a hyperbolic or parabolic equation is, however,
quite different. In these cases real characteristics exist, and if a strip is built up
along a curve C which lies on a characteristic cylinder, there will be no integral
surface including tbis strip unless the strip is a characteristic strip. In this case,
when the differential equation is hyperbolic it is found that there then are
infinitely many integral surfaces including the strip, and the solution in the
neighborhood of the strip is not determined uniquely by the prescribed infor-
mation on the strip. However, when the equation is of parabolic type, a charac-
teristic strip generally determines a unique integral surface. (Problem 53
provides an example.)t
Thus, for the hyperbolic equation
(125)
(~~r - I = O, dy
dx
= ±I
'
and hence the characteristic base curves in the xy plane are the straight lines
x+y=c" x - y = c2 . (126)
Along any such line the values of z, p = az/ax, and q = az/ay cannot be
independently prescribed in an arbitrary way. The compatibility condition (121),
dpdy dq dx _ (127a)
dA dA - dA dA - O,
tProofs of these assertions are somewhat involved, but can be based on the reducibility of
(109) to one of the standard forms (132) and (134) in those cases. In the parabolic case, the
uninteresting situations in which E ~ O in (134) are exceptional.
8.8. Characteristics of Linear Second-Order Equations 413
as welJ as the usual strip condition (110),

dz _ dx..L dy
dA - p dA ' q dA ' (l27b)
must be satisfied.
If we identify A with arc length, say s, then along a line of the first family
x + Y = el we have
dx dy,J2
ds - ds =""2
and hence Equations (l27a, b) lead to the requirements
-dz
p + q = constant, p - q =,J2-
ds
(along x +y = Cl)' (l28a)
Similarly, along a line of the second family x - y = c 2 , we have the relations

dx dy ,J2
ds = ds =""2'
and hence (127a, b) give
í'f dz
p - q = constant, p+q=,y2- (along x - y = c 2 ). (128b)
ds
A comparison of these conditions shows that the equation of compatibility
requires that, at al! points of a line of one family, the derivative of z in the
direction of the intersecting lines of the other family must be constant. Since
he re the two families of characteristic lines are perpendicular to each other,
the derivative of z normal to a characteristic fine must be constant along that
fine in this particular example.
The general solution of (125) is of the form
z = f(x + y) + g(x - y). (129)
We may notice that each function in (129) is constant along one of the charac-
teristic lines. If we write (126) in the form rp(x, y) = c" rp(x, y). c 2 • the general
solution (129) becomes
z =f(rp) g(rp).+ (130)
AIso, if rp and rp are taken as new independent variables, Equation (125) becomes
(l31)
Although it usually is not possible to express the general solution of a

specific equation of type (109) in a form similar to (130), still it is not difficult
to show that if the expressions rp and rp of (124) are taken as new independent
variables, in the case where (109) is hyperbolic, then (109) is transformed to the
so-called normal (or canonical) form
2
a z = D az +E az + Fz + G, ( 132)
arp arp arp arp
where D, E, F, and G are functions of rp and rp.
414 Partíal DilferentiaJ Equations
By cboosing suitable new real independent variables a and p, any elliptic

equatioo of type (109) can be put in tbe normal form
a'z
+ aap2 = az
+ E aazp + Fz +.
2
::.
aa' D aa G, (133)
and any parabolic equation of this type can be put in the form
a2 z az az
aa2 = D aa + E ap + Fz + G. (134)
It may be noticed that the two families of cbaracteristic base curves for
(132) comprise the curves (j1 = constant and t¡I = constant , and tbat the one
farnily for (134) comprises tbe curves p = constant .
In the more general case of a quasi-linear equation of second order,
a2 z a2 z a2z
a ax2 + b ax ay + e ay2 + F = 0, (135)
tbe curves in the xy plane which satisfy (120) again are exceptional, and are
often caBed tbe characteristics of the differential equation. However, here the
situation is considerably complicated by tbe fact that, since the coefficients a, b,
and e rnay depend upon z, az/ax, and az/ay, the characteristics of the equation
(in fact, !theír very realily) rnay depend upon tbe solulion, which in turn also
depends upon the sí de conditions (initial and/or boundary condítions) tbat
supplernent tbe differential equatíon in the complete formulatíon of a problem.
Cbaracterístíc cylínders and related entíties generally do not exíst.
8.9. Singular Curves on Integral Surfaces. Let S be a particular integral

surface correspondíng to the linear first-order equation
P~~ + Q ~~ = R,z + R2, (136)
wbere P, Q, R" and R 2 are continuous functions of x aod y. Then tbe existence
of the partial derivatives az/ax aod az/ay in (136) implies that z itself be continu-
ous; tbat ¡s, the surface S must be a continuous surface. Hence tbe equal mem-
bers of (136) must both be continuous functions of x and y. Tbis condition,
however, does not exclude the possibility that
z the two terms on the left may be each discon-
tinuous , so long as their sum is continuous. In
particular, there may be a curve C on S along
wbích tbe surface possesses a "comer" or
"edge," as is indicated in Figure 8.2. We denote
by C o the projection of C onto the xy plane,
y and consider z, az/ax, and az/ay as functions
of x and y. Then, while z is an everywbere
continuous function of x and y, and wh.ile tbe
x
derivative of z in the direction of C o is continu-
Figure 8.2 üus, we suppose tbat the derivative of z normal
8.9. Singular Curves 00 Integral Surfaces 415
10 Ca has a finite jump as the curve Ca is crossed in the xy plane (Figure 8.2).
Thus here the partial derivatives az/ax and az/ay will exist (in general) only as
"one-sided" derivatives on Ca, that is, as limits in which Ca is approached from
one side or the other. Surfaces of this general nature are of frequent interest in
the study of physical phenomena where abrupt changes of sorne sort may occur.
We now investigate more explicitly the possibility of their existence, as integral
surfaces of an equation of type (136).
Let the curve Cabe specified by the parametric equations
x = x(l), y = y(l), (137)
where 1 is a convenient variable indicating position on Ca. Then 1 may also be
considered as representing position in the direction of Ca along parallel curves
in the immediate neighborhood of Ca. In particular, the derivative of z in the
direction of Ca at points on such curves is given by
(138)
Now consider any point Pon the curve Ca. Since the left-hand member of
(136) must be continuous across Ca at this point, we must have
P !J.p + Q!J.q = O, (139)
where !J.p and !J.q are the values of the jumps associated with the partial deriva-
tives p = az/ax and q = az/ay across Ca at P. Since the derivative of z in the
direction of Ca is also assumed to be continuous across Ca, the right-hand
member of (138) also must be continuous, and we obtain a second equation
dx A
- /.>.p
dl
-dy IJ.q
+ dl A
= O (140)
re1ating the jumps in p and q across Ca at P. Equations (139) and (140) are
compatible (with !J.p and !J.q not both zero) only if the coefficients of !J.p and
!J.q are proportional, that is, if
dx dy (141)
P Q
along Ca. Thus Ca must lie on a characteristic cylinder of (136), and hence,
since we have assumed that C lies on an integral surface of (136), we conc!ude
that the "singular curve" C must be a characteristic curve of (136). That is,
corners of the type considered can exist on an integral surface of (136) only
along characteristic curves.
In the case of the linear second-order equation
(142)
the existence of the second derivatives presupposes continuity of z and its first
partial derivatives, and hence also continuity of the equal members of (142) if
the coefficients and the function g are continuous. However. the possibílity that
the separare terms on the left may be individually discontinuous is not excJuded.
Let C be a curve in an integral surface S, with projection C o in the xy plane,
and denote position along Ca by the parameter A. We investigate the possibilíty
that z, p, and q be continuous on S, and that the derivatives of p and q in the
direction of Cij'
dp _ ap dx -'- ap dy = dx + . dy , (143)
dJ.. - ax dA . ay dA r dJ.. .5 dA
dq _ aq dx I aq dy = dx + t dy ,
(44)
dA - ax dJ.. "T ay dA s dA dA
be continuous across Ca' but that the derivatives of p and q normal to Ca be

discontinuous across Ca. Here the separa te second partial derivatives may exist
along Ca only as one-sided derivatives, as Ca ís approached from one síde or the
other.
The continuity of the ríght member of (142) and of the left members of
(143) and (144) ímplies contínuity of their equivalents, and hence we obtain the
three equations
a t:!.r '- b t:!.s +
c t:!.t = 0,
~~ t:!.r + ~~ t:!.s = 0,
( 145)
dx A ,dy A_O
dA uS "T dA ut - ,
relating the jumps in the three partíal derivatives r, s, and t across Ca. These
equations are consistent (with at least one nonzero jump) only if the determinant
of the coefficients vanishes. But this condition is the same as that of Equation
(120), and hence the curve Ca must líe on a characteristic cylinder of (142).
Since C lies on an integral surface and projects into Ca, it follows that C must
be a characteristic curve of(l42). In particular, since an equation of elliptic type
has no real characteristic curves, an integral surface of such an equation cannot
contain a "singular curve" of the type required.
Results of this type are of importance, for example, in the study of two-
dimensional compressible fluid flow around rigid bodies [see Equation (200),
Section 6.20]. Here it is found that for low velocities the problem is governed
by a partial differential equation of elliptic type, whereas if sufficiently high
velocities are attained the governing equation beco mes hyperbolic, and discon-
tlnuous phenomena may then be present.
Situatlons also arise in which thefirst partial derivatives of the "solution"
of a problem governed by a second-order hyperbolic differential equation nec-
essarily exhibit jumps along certain characteristic base curves, or in which the
"solution" itself has this property (se e Section 9.15). Consideration of the phys-
ical (or other) basis of the mathematícal formulation then must determine
whether the "solution" is to be acceptable.
8.11. The Characteristics of a Particular QU3sí- Linear Problem 417
8.10. Remarks on Linear Second-Order lnitial-Value Problems. In this sec-

¡ion we summaríze briefly certain preceding r~sults and present additional facts
bearing on their significance.
For the general linear second-order eq uation of the form
a':: + b d~d'
a a---;;
al:: --'-- aa---O;, da::.
I C
2
:: -L a:: --, f-~ -_
-ax --, e -aI I
g, ( 146)
x x y y- y
we define the characteristic base curves. in the xy plane, as those curves for
which
a(dy)' - b(dx)(dy) +
c(dx)' = O. (147)
lo the e//iplic case, for which b 2 < 4ac, no real characteristic base curves
can exist. ln the hyperbo/ic case, for which b 2 > 4ac, two distinct sets of such
curves are obtained, whereas in the parabo/ic case, b 2 = 4ac, the two sets
become coincident.
ln the e//iptic case, if along any curve c in the xy plane :: and its derivative
io the direction normal to that curve are prescribed as functions of position
aloog c in such a way that these prescribed values are regular everywhere along
c, then a solution to the resulting initial-value problem is determined for values
of x and y in sorne neighborhood of c. However, here two additional facts are
of importance. First, unless the prescribed values are regular along c (in par-
ticular, unless derivatives of all orders exist at all points), no solution to the
defined problem can exist. Second, even though a solution exist, in many cases
it will exist only in a restricted neighborhood of c and will not be val id over
the whole xy planeo
ln the hyperbo/ic and parabo/ic cases, if c is not a characteristic base curve,
the initial-value problem again has a solution. However, here the prescrib~d
initial values need not be regular. The higher derivatives of the prescribed values
may have finite jumps at certain points on c. These jumps are then found to be
propagated along those characteristic base curves which pass through the re le-
vant points on the initial curve, and the solution obtained is valid throughout
the xy planeo If c is a characteristic base curve, then the initial-value problem
does not have a solution unless the initial values of z and its derivative normal
to c satisfy a certain compatibility condition, in which case infinitely many
solutions then exist in the hyperbolic case and a single solution is determined
in the parabolic case, in general.
Certain of these facts are illustrated in Problems 53 and 60-62 at the end
of this chapter.
8.11. The Characteristics of a Particular Quasi-Linear Problem. To illustrate

the nature of characteristics in other types of problems, we here consider the
simultaneous equations
au av cos 2v + av
-ax + -ax - .
SIO 2v = O, (l48a)
. ay
aa yu + aa vx sin 2v - aayv cos 2v = O. ( 148b)
418 Partíal Differentíal Equatíons
These equations, which a re quasí-línear in u and v, are of basic ímportance in

certain two-dimensional problem s in the mathematical theory of plasticity. We
suppose that LI and vare prescribed at all points along a curve Ca in the xy
plane. lf l specifies position along Ca. we may define Ca by the equations
x = x(l), y = y(l). (149)
For points on Ca the dependent functions U and vare then to be considered as
given functions of A..
We now ask whether there exísts a curve Ca, or a set of such curves, with
the property tbat ti and vare not uniquely determined for points near Ca by
their prescribed values along Ca. It ís assumed here that these prescribed values
are continuously differentiable along Ca.
Since LI and vare known along Ca. as functions of l, their derivatives along
C o are calculable, and must satisfy the equations
du(l) au dx + au dy (lSOa)
~ = JX dl ¡ry dl '
dv(l) _ av dx + av dy. ( ISOb)
dl - JX dl ¡ry dl
Equations (l48a, b) and (ISOa, b), constituting four linear equations in the four
first partíal derivatives au¡ax , au/ay, av/ax, and av/ ay, may be rewritten in the
form
u., + v~cos2v + v y sín 2v = O, (ISla)
uy + v., sin 2v - v y cos 2v = O, (lSlb)
(¡SIc)
= v', (lSld)
where a prime denotes l differentiation. A unique solution IS assured unless
tbe determinant of coefficients vanishes, that is, unless
O cos 2v sin 2v
O 1 sin 2v -cos 2v
= O. (IS2)
x' y' O O
O O x' y'
By expansion, this equation can be written in the form
y'2 sin 2v + 2x'y' cos 2v - X'2 sin 2v = O,
from which we obtain the two alternatives
y' = x' tan v, (1 S3a)
y' = -x' cot V. (lS3b)
Thus two systems of curves are determined in the plane, for one of which
dy/dx = tan v, and for the second dy/ dx = -cot V. The two systems are seen
to be orthogonal, and the slope of each curve is seen to depend upon the value
of the dependent function v at the point under consideratíon.
8.11. The Characteristics of a Particular Quasi-Linear Problem 419
Along any such curve, Equations (15Ia-d) cannot have a unique solution.
In particular, the set cannot be solved for l/x and hence no solution exists unless
the numerator determinant of Cramer's rule also vanishes, that is, unless
o O cos 2v sin 2v
O l sin 2v -cos 2v
= O,
u y' O O
v' O x' y'
or, expanding, unless
u' = v'(cos 2v - ;: sin 2V)' (154)
Along the curves (I53a) we then obtain the requirement l/' = -v', or
dY=tanv: u+v=constant; (155a)

dx
and along the curves (I53b) there follows u' = v', or
dy = -cot v: u - v = constant. (l55b)
dx
If either (l55a) or (l55b) is satisfied, U x may be taken arbitrarily, and it is readily
verified that Equations (l5Ia-c) then determine u y , V x ' and v y in terms of U u', X
'
and v'. Thus in these cases infinitely many solutions exist, all corresponding to
the same prescribed values of ti and v along Co.
The curves in the xy plane for which (155a) or (155b) is true may be called
the characteristics of the simultaneous equations (148a, b).
We next show that the equations (148a, b) are simplified if we take as new
dependenl variables the combinations u + v and ti - v suggested by (155a, b),
say
.;=u+v, 11=u-v. (156)
The variable'; is then constant along the first set of characteristics, whereas 11
is constant along curves in the second set. Then since we also have
u = t(.; + 11), v = t(.; - 11), (157)
Equations (148a, b) take the form
(';x + 11x) + (';x - 11x) cos 2v + (.;, - 11,) sin 2v = O,
~.+~+~x-11x)~~-(~-~=~=O,
or, after a rearrangement,
(11x - 11, cot v) + coP v(';x + .;, tan v) = O,
(158)
-(11, - 11, cot v) + (';x + ';y tan v) = O.
These equations imply the simplified relations
';X +.;, tanv = O,
(159)
11x - 11, cot v = O,
420 Partíal Dífferential Equations
where v = te,; - '1). Equations (159) are still nonlinear, but they are more
traetable than the equivalent equations (148) (see Problem 64), and form tbe
basis for further treatment of eertain problems in plastieity. Physically, the
eharaeteristics in this case are the so-ealled shear fines (lines of maximum
shearing stress) in a plastie problem of plane strain for an ineompressible
material.t
REFERENCES
1. BATEMAN, H., Partial Differential Eq/lations 01 ¡"lathematical Physics, Cambridge

University Press, New York, 1959.
2. BERG, P. W., aod J. L. McGREGOR, E!ementary Partia! Differentia! Equations,
Holden-Day, Ine., Sao Francisco, 1966.
3. COURANT, R ., and D. HILBERT, Methods 01 l"lathematical Physics, Vol. 1, John
Wiley & Sons, Ine. (Interseienee Division), New York, 1953.
4. HOPF, L., ¡ntrodllction 10 the Differentia! Equations 01 Physics, Dover Publieations,
Ine., New York, 1948.
5. JEFFREYS, H., and B. S. JEFFREYS, !vlethods 01 /'vlathematica! Physics , 3rd ed., Cam-
bridge University Press, New York, 1956.
6. MORsE, P. M., and H . FESHBACH, Methods 01 TheoreticalPhysics, 2 pIS., MeGraw-
Hill Book Company, Ine. , New York, 1953.
7. PRAGER, W., Discontinuous So!U/ions in the Theory 01 Plasticity, Courant Aooiver-
sary Volume, pp . 289-300, John Wiley & Soos, Ine. (Interseienee Division), New
York,1948.
8. SOMMERFELD, A., Partial Differential Equa/ions in Physics, Aeademie Press, Ine.,
New York, 1949.
9. WEBSTER, A. G., Partia! Differentia! Equations 01 Mathematica! Physies, 2nd ed.,
Dover Publieations, Ine., New York, 1956.
10. WEINBERGER, H. F., A First Course in Partial Differential Equations, Xerox College
Publishing, Lexington, Mass., 1965.
PROBLEMS
Sectíon 8.1
1. Find the differeotial equation of lowest order whieh possesses eaeh of tbe foJlowing
solutions, with I and g arbitrary fuoetions:
Ca) z = (x - y)/(x + y),
(b) z = I(ax + by) + g(cx + dy) (ad - be *- O),
(e) z = I(ax + by) + xg(ax + by).
tSee Referenee 7.
problems 421
2. (a) Obtain the partial differential equation of first order satisfied by z = f(I/f),
where I/f is a given function of x and y andfis an arbitrary function, in the form
I/f y
az -
-a I/f x
az
a-y = O.
x
(b) With the new independent variables s and l, where l = 'lI(x, y) and s is any
independent function of x and y, show that, when z is considered as a function of s
and l, the differential equation of part (a) takes the form
az
(sxl/f y - Syl/f x) as = O.
Hence deduce that the masl general solution of the equation is of the form z = f(I/f),
where f is arbitrary.
3. Noticing that z = I/f is a solution of the equation considered in Problem 2(a),
deduce that, if one solution of the equation
az az
P(x, y) ax + Q(x, y) ay = o
is of the form z = I/f(x, y), then the most general solution is z = f(I/f), where f is
arbitrary.
4. Obtain the partial differential equation of first order satisfied by z = ({Jf(I/f), where
({J and I/f are given functions of x and y and f is an arbitrary function, in the form
({J(I/fy~: - I/fx ~;) = «({Jxl/fy - ({Jyl/fx)z,
Section 8.2
5. Determine ¡he general solution of each of the following equations (with a, b, and e
constan t), writing each solution in a form sol ved for z:
al: az; az az
(a) a -a + b -a = e, (b) a -a + b a- = ez,
x y x y
() az
c y ax -
az
x ay = o, (d) a z
ax
+ az , 2
ay" xz =
o,
az az az az
(e) x ax - y ay = z, (f) x' ax + y2 ay = Z2
6. (a) Show that if a particular solution of the equation
az, az
a ax ., b ay = f¡(x) + f2(Y)'
where a and b are constant, is assumed in the form z p = ({J 1 (x) + ((J2(y), there follows
zp = ~ Jf¡ (x) dx + !I f,(y) dy (ab 7= O),
so that the general solution is of the form

z = f(bx - ay) + zp'
(b) Illustrate the results of part (a) in the case of the equation
az az
ax - 2 ay = 2x - eY + 1.
422 Partíal DiJIerential Equations
7. Use ¡he results of Problem 9 of Chapter 7, with an appropriate change in notation ,

to show that if the dependent and independent variables are interchanged in a pair of
simultaneous quasi-linear differential equations in u and v, each of the form
Au x + Bu + y CV x + Dv y = O,
where A, B, C, and D depend only on 1I and v, a pair of equivalent linear equations in
x and y is obtained, each of ¡he form
Ay, - Bx, - CYu + Dxu = O.
Section 8.3
8. (a) Show that the characteristic curves of the equation
az + az = 1
ax ay
are straight lines parallel to the vector i + i·+ k, and hence that the characterisiic
curve passing through a point (xo, Yo, zo) is specified by the equations
x - Xo = y - Yo = z - Zo·
(b) Determine the characteristic curve which passes through the point (O, Yo, zo)
in the yz plane, and show that those characteristic curves which pass through points
on the curve z = y2 in the plane x = O are specified by the equations y - x = Yo,
z - x = y5. Thus deduce that the surface traced out by these curves is the parabolic
cylinder
z = x + (y - x)2.
(c) Verify directly that the surface z = x + (y - X)2 is an integral surface of the
equation (az;ax) + (az/ay) = 1 which ineludes the curve z = y2 in the yz plane.
9. (a) Obtain the general solution of the differential equation of Problem 8(a), in the
form
z = x + f(y - x).
(b) Determine the function f in such a way that this solution is consistent with
the equations z = y2, X = O. Hence rederive the result of Problem 8(c).
10. (a) Show that the solution of the differential equation of Problem 8(a), for which
z = rp(x) along the line y = 2x in the xy plane, is of the form
z = 2x - y + rp(y - x).
(b) If z is prescribed as rp(x) along the line y = x in the xy plane, show that no
solution exists unless rp(x) is prescribed in the form rp(x) = x + k, where k is a con-
stan!. In this last case, show that z = x + f(y - x) is a solution for any differentiable
f such that f(O) = k, so that infinitely many solutions then exis!.
(c) Show that the projections of the characteristic curves onto the xy plane are
the lines y = x + e, and that if z is prescribed as rp(x) along any such curve there is no
solution unless rp(x) = x + k, in which case infinitely many solutions exis!.
11. Find the solution of the equation
az az
ax = ay
for which z = (t + 1)4 when x = t2 + 1 and y = 2t.
Problems 423
12. (a-f) For each equation of Problem 5, determine the integral surface which ineludes
the straight line x = y = z, if such a surface exists.
13. (a-f) For each equation of Problem 5, determine the solution for which z = X2
aJong the straight line y = 2x in the xy plane, if such a solution exists.
14. Determine the general solution of each of the following equations, writing each
~olu tion in a fonn solved for z:
(a) (x +y)(~~ + ~;) = z - 1,

az
(b) -a = xy,
x
az az az az
(c) xz ax + yz ay = xy, (d) ax + ay = 6xyz.
15. Suppose that the soJution of the equation

az + az = !
ax ay
is to be such that z = cp(x) when y = mx + b.
(a) Determine that solution in the form
z =
cp
(Y m
- x- -l b) _ Y -m-1
mx - b
when m "'" 1.
(b) Describe the situation when m = !.
16. (a) Obtain the general solution of the equation
az+az+az=!
ax ay al
in the form
F(y - x, I - x, z - x) = o.
(b) 1f that solution for which z = cp(x, y) when x +y
= t is required, show that
there must follow
F[y - x, y, cp(x, y) - xl = O
and hence
F[I/, v, cp(v - l/, v) + U - vl = O.
Thus deduce that ir we write y - x = u, t - x = v, we must have also z - x =
cp(v - u, v) + u - v, and hence obtain the required solution in the fonn
z = cp(1 - y, I - x) + x + Y - l.
17. (a) Obtain the general solution of the equation

az az az
x-+y-+I-=!.
ax ay al
(b) Determine that solution for which
z = cp(x, y) when X2 + y = t
in the fonn
z= ~m[t
I - y'"
- Y, y(1 - y)].
X X2
Section 8.4
18. Suppose that z = l/,(x, y), Z = "2(X, y), ... each reduce the left-hand member of
(50) to zero and that = = p(x, y) reduces that member lo g(x, y).
(a) Show that, assuming appropriale convergence,
satisfies (50) wilh g replaced by zero, when the c's are constants.
(b) Again assuming appropriate convergence, show that
=
z = p(x, y) + L:
k=l
CkUk(X, y)
is a solulion of (50).
19. (a) Show that lhe result of selling Z = l/V in (50) is of the form
. ( aou..L aOll , q2 U) -¡- Fu = g,

v a ax 2 , b axa T y C a2
y
if V satisfies both of the first-order equalions
2a aa
lJ
x
+ b aa
v
y
+ dv = O, b av
ax
+ 2c av , ea
ay
= o.
(b) When the coefficients a, b, c, d, and e are constants, show that constants IX
and p can be determined in such a way that 'V = e,·npy satisfies the two equations
obtained in part (a) provided only that b 2 =F 4ac, so that the equation is not parabolie.
Verify also that, in this case, there follows
F = [f - (aIX 2 + bIXp + cf32)]V.
(e) Illustrate the preeeding transformation in the case of the equation
a z + d a-
a x ay
2 a-
a~ + e a- + fz = g.
x y
(Notiee that if f = O the elimination of the terms involving the first partial deriva-
tives generally is at the expense of introdueing a term involving the function itself.)
20. Suppose that gis replaeed by az/at in (50), where I is a third independent variable,
so that z is to satisfy the equation
2
aa z +b a z +c a2z +d az +e az +fz=az
2
dX2 dx ay ay2 ax ay al
When a, b, c, d, e, and f are conslanls, show that constants IX, p, and y can be deter-
mined in sueh a way that the substitution
requires that u satisfy the equation

2 2
a u + b a u + c a2a = au,
a ax2 ax a y ay2 at
provided only that iJ2 7"= 4ac.
Problems 425

a + 2x -=-=-----;c-
~
2Z a 2Z
aX 2 axay
is elliptie for values of x and y in the regio n X2 + y2 < 1, parabolie on the boundary,
and hyperbolie outside the region.
(b) Show that the quasi-linear equation
a 2z a2z a 2z
x~a x 2+za a
x y +Ya---z=O
y
is elliptie when Z2 < 4xy, parabolie when Z2 = 4xy, and hyperbolie otherwise. (Notiee
thal the nature of a quasi-linear equation thus may depend not only upon position in
the xy plane but also upon the solution z, whieh in turn depends upon the boundary
eonditions.)
22. (a) Show that the assumption that the equation
a 2z az
ax2 - ay
possesses a solution of the fonn z = rp(x, y)f(lf/(x, y)], where rp and If/ are speeifie fune-
tions and f is arbitrary, leads to the requirement
rplf/U "(If/) + (2rpxlf/ x + rplf/ x, - rplf/ y)f'(If/) + (rpxx - rpy)f (lf/) = O,
for arbitrary f, and henee to the tbree requirements
rplf/; = O, 2rpxlf/x + rplf/xx - rplf/y = O, rpxx - rpy = O.
tb) Show that these eonditions imply that either rp = O or If/ = eonstant, so that
there exists no solution of the assumed form whieh aetually depends upon an arbitrary
funetion ¡:
Seclion 8.5
23. Obtain the general solution of eaeh of lhe following equations:

a 2z a 2z a 2z
(a) ax2 - 2 dx d y - 3 a y2 = O,
2 2
(b) a'rp _ 2 a rp + 2 a rp = O
a x2 ax ay ay> '
. ~ a
2 rp_
(e) d x-' - d x d y - O,
a2 r a> w
(d) ar,crw ) = e' al> '
(e) (U2 - V2) ~ .o...?U a trp ,a'rp = O

a x' , - a x al ' al 2 '
fbE_~=o
(f) a x' ay' '
a·z a·z a·z
(g) - 4
ax
-?
- ax 2 ay2
+ ay'
- = o'
lh) "V 4 rp = a·p + 2 a·p + ~ = O.
ax 4 a x2 ay' ay4
24. Obtain the general solution of the simultaneous equations

au -e 2 av
- ,
ax = ay
10 the form
(First eliminate v and solve the resultant equation for u; then use the original equa-
tions to determine v . Notice, for example, that
al = kc al.
ax ay
An arbirrary constant so introduced can be absorbed into the definitions of I and g.)
25. (a) Pro ve that the general solution of Laplace's equation, in the form
~
ax2
a2rp_
+ ay2 - 0,
rp(x, y) = I (x + iy) + g(x - iy) ,
where I and g are twice-differentiable functions of the complex conjugate arguments
x + iy and x - ¡y.
(b) Deduce that the real and imagioary parts of both I(x + iy) and g(x - iy)
satisfy Laplace's equalÍon.
(c) Use the fact that tbe real and imaginary parts of a twice-differentiable func-
tion g(x - iy) are respeclÍvely the real part and the negative of the imaginary part of
the twice-differentiable conjugate function g(x + iy) to deduce that any real solution
of Laplace's equation is the real or imaginary part of a twice-differentiabte function of
the complex variable x + iy, and conversely.
26. (a) Obtain as the real and imaginary parts of the function (x + iy)", for
n = 0, 1, 2, 3, and 4, the functions 1, x, y, X2 - y2, 2xy, x 3 - 3xy2, 3x2y _ y3,
x· - 6x2y2 + y4, and 4x 3y - 4xy', and verify directly that they each satisfy Laplace's
equation.
(b) Show that the functions r n cos nO and r" sin nO, where n is integral, are the real
and imaginary parts of (x + iy)n when x and y are expressed in polar coordinates, and
verify directly that they each satisfy Laplace's equation . [See Equation (26) of Section
1.5 and Equation (168e) of Section 6.18.]
27. In each of the following cases, fust obtain a particular solution of the equation as
a function of one variable on1y, and then obtain the general so[ution by adding this
solution to the general complementary solution:
a 2z a 2z a 2z a 2z a 2z
(a) ax2 - a y 2 = x, (b) ax2 - 3 ax ay + 2 ay> = cosy.
28. Suppose that the right-hand member of a linear partial differential equation is a
homogeneous polynomial of degree k in x and y (each ter m being of degree k), whereas
the left-hand member is homogeneous of order n in z (each term iovolviog ao nth deriva-
Problems 427
Uve) with conSlant coefficients. Show that lhe assumption of a particular solution in
the form of a homogeneous polynomial of degree n + k leads to k + 1 linear equa-
tions in n + k T 1 unknown eoeffieients. (l! can be shown that this set always pos-
sesses a solution and, indeed, that a eertain set of at least n of the coefficients can be
assigned arbitrarily, in a eonvenient way,)
29. Use the proeedure outlined in Problem 28 (andjor Problem 27) to obtain any
particular solution of eaeh of the following equations:
2 + a 2z ~ 2 +
ay ~
a x + az
( a ) az 2...L.
~ x y
2
, xy, a -z
(b) -
ax2 a y2 ~ x xy,
a2z a 2z a 2z a 2z a 2z
Ce) a---¡ ~ - a' = xy + x, (d) a x 2 ~ 2 a x a y + - a + y.
x y- y 2 = X2
[In parts (e) and (d), eonsider the two terms on the right separately and use superposi-
tion .]
30. (a) If am 2 + bm + e = a(m ~ m¡)(m ~ m,), show that the equation
a 2z a 2z a 2z
a ax 2 + b ax a y + e - ya2 = f(x, y),
where a, b, and e are eonstants, can be written in the operational forrn
a a)(aZ
a ( ax~m¡ay aZ) =f(x,y)
ax~m2ay
aod deduce the general solution when f(x, y) = °

by the methods of Seetion 8.2.
(b) Show that a particular solution of the general equation of part (a) can be
obtaioed as a particular solution of the equation
az az
ax ~ m, ay = z¡(x, y),
where z I is a particular solution of the equation
aZ¡ aZ¡)
a ( ax ~ mI ay = f( x, y ) .
31. Use the method of Problem 30 10 obtain the general solution of the equation
a'z
-- -
a 2z
-- = (x-• - y2) Sin
. xy.
ax2 ay>
32. Use Equation (67) to find the solutioo of the equation
~2 = e2 í!!:!e
al ax2
for wbieh rp = X' and arp jal = eos x for aH values of x when I = O.
33. 1 t is required 10 find the solution of the equation
a2 rp2 = e' a' rp

al ax2
whieh satisfies the conditions
rp(O, 1) = 0, rp(l, 1) = ° (for aH values of 1)

428 Partial DilTerential EquatioDS
along the boundary of the strip O <: x <: I in an xt plane, and the conditions
rp(x, O) = F(x), arp(x, O) = G(x) (when O < x < 1)

al
along the hne segment t = O in that strip.
(a) Taking the general solutíon of the equatíon in the form
rp = f(x + el) + g(x - el),
show that the conditions along the boundaries x = O and x = I lead to the relations
f(u) + g(-u) = O (writing u = el),
f(u + 2/) = f(u) (writing u = et - 1),
for all values of u, while the conditions along the segment t = O for O < x < I become
f(u) + g(u) = F(u) }
(when O < U < 1).
e[r(u) - g'(u)] = G(u)
(b) Deduce that, if we write H'(u) = G(u), then the last two conditions of part
(a) give
1 1
f(u) = 2 F(u) + 2e H (u),
1 1
g(u) = -_" F(u) - ."
_e H(u),
when O < u < 1, and that also, for all values of u, f(u) must be a periodie funetion, of
period 21, and the condition
f(u) = -g( -u)
must be satisfied.
(e) Show that these eonditions are al! satisfied if we take
1 1
f(u) = -,,5'(u)
_ + -2e X(u),
1 1
g(u) = 25'(u) - 2eX(u),
for all values of U, where 5'(u) and Seu) are odd period funetions of period 21, agreeing
with F(u) and G(u), respeetively, when O < u < 1, and where X'(u) = Seu). Thus, with
this notation, obtain the required solution in the form
rp(x, t) =
1
2[5'(x + el) + 5'(x - et)] + 1
2e
JX+O< S(.;) d';.
x-o<
34. To illustrate the solution of Problem 33, notÍce that if F(x) = sin 7tx/1 and
G(x) = O, there follows 5' = F and S = G = O, and obtain the solutíon in the form
rp = 2
1 (.
SlD 7t
x +I et + .
SlD 7t
x -
1
et)
= SlD
. 7tx C7tt
-1- cos -1- .
35. By using the method of Problem 33, obtain the solution of the problem
a2 rp = c 2 _
__ a 2_
rp ,
al 2 ax2
Problems 429
rp(.~, O) = F(x), arp~~; O) = G(x) (wben x > O),
rp(O, t) = O
in tbe balf-plane x > O in tbe form
rp(x, t)
1
= "2 [5'(x + et) + 5'(x - et)] + 1
2e
fX." S(e;) de;,
x-o<
where bere 5'(u) and seu) are oddlunetions 01 u agreeing with F(l/) and G(l/), respective/y,
when u > O.
36. Suppose that a function u(x, y) satisfies the equation
a u a,u
2
ax' - a y2 = q(x, y)
in a region D which includes the triangle with vertices at the points P(xo, Yo),
PleXO - Yo, O), and p,eXO + Yo, O) in Figure 8.3.
(X o - Yo , O) x
Figure 8.3
(a) By integrating the equal members of ¡he equation over D and using Stokes's
theorem in the plane [Equation (139) of Chapter 6], show that
fe (~~ dy + ~~ dx) = JIv q(x, y) dx dy,
where e is the boundary of D traversed in the counterclockwise direction.

(b) By noticing that dy = -dx along P 2 P and dy = dx along pp¡, show that the
integrand in the line integral is -du along P,P and +du along pp¡, and hence deduce
that
= 2"l [u(xo + u(xo 1 f'o+YO

u(xo, Yo) - Yo, O) Yo, O)] +- 2" 'o-Yo lIy(X, O) dx
1 1YO f Xo
+
YO -Y
- "2 q(x, y) dx dy.
o xo -yo + y
37. A nonhomogeneolls Cauchy problem.

(a) Obtaio the solutioo of the problem
,a 2 rp
e- ax2 -
aae'rp_- h(x, 1),
2
rp(x, O) = ¡(x), rp,(x, O) = g(x)
wheo -ca < x < ca, O -< t < ca, by makiog appropriate chaoges In notatioo 10
Problem 36, jo the form
1
rp(x, t) = 2[/(x + ee) + ¡(x - et)] + 1
2e IX+" g(,;) d,;
x-o<
- de I
[Notice that this result reduces to (67) when h = O.]
(b) Use the result of part (a) to obtain the solution of the specified probJem when
¡(x) = X2, g(x) = 1, and h(x, e) = 1, and verify its correctness.
Section 8.6
38. (a) By assuming a solution of the Helmholtz equation
2 2
a z + a z + k 2z = O
ax 2 ay2
in the form z = e"x+PY, and satisfying the resultant requirement a 2 + P2 + k 2 = O
by writing a = ik cos rp and p = ik sin rp, where rp is an arbitrary real parameter,
obtain particular solutions in the forms
cos [k(x cos rp + y sin rp)], sin [k(x cos rp +y sin rp)].
(b) Deduce that the reat or imaginary pan of any expression of ¡he fonn
S"
U(x, y) =
.' A(rp)e1k(xoO, .+y,in.) drp
is a formal solution, for an arbitrarily chosen function A(rp) and arbitrary constant
limits rp 1 and rp2'
39. By changing to polar coordinates, deduce from the results of Problem 38 that the
real and imaginary parts of any expression of the form
Ver, O) =
S.,·' A(rp)e1k ' co. (.-0) drp
formaIly satisfy the Helmholtz equation in polar coordina tes,

2 2
a z + ~ az + _1 a z + k2z = O.
ar 2 r ar 2
r a0 2
40. By assuming a solution of the heat flow equation
a 2z + a 2z = k 2 az
ax2 ay2 at
in tbe form z = eU+fty+,." and satisfying the resultant requirement by setting
Problems 431
(X = ikp cos rp, fJ = ikp sin rp, y = - p2, deduce particular solutions in the forms
e- P " cos [kp(x cos rp +y sin rp)], e- P " sin [kp(x cos rp +y sin rp)]
or as the real or imaginary parts of an expression of the forro
V(x, y, t) = rr
• JeR
A(p, rp)e-p',+lkp(x co'.+nin.1 dp drp .
41. By changing to polar coordinates in Problem 40, deduce particular solutions of the
equation
in the forros
VI(r, (J, t) = f SeR A(p, rp)e- p" cos [kpr cos (rp - (J)] dp drp
and
V 2 (r, (J, t) = SS", B(p, rp)e- p" sin [kpr cos(rp - 8)] dp drp.
42. Suppose that a function p(x, y) exists such that the equation
a 2z a 2z a 2z az az
a ax2 + b ax ay + e ay2 + d ax + e ay +fz = g
J...[a
p
2 2 2
a (pz) + b a (pz) + e a (PZ)] = g.
ax2 ax ay ay2
(a) Show that p must satisfy the three equations
2a Px
p + b Py
p = d,
and apxx + bpxy + CPyy =.!p.

(b) Deduce that, with the assumption b 2 "" 4ac and the abbreviations
Al = be - 2cd, N = bd - 2ac,
b 2 - 4ac b 2 - 4ac
the function p is determined by the equations
a
ax logp = M,
a
ay logp = N,
and that M and N must satisfy the two "compatibility conditions"

aM aN
ay = ax
and
a ~~ + ; b (~~ + ~~) + e ~~ + ; (dM + eN) =.f
43. Consider the application of the resuIts of Problem 42 in the following cases:
(a) For the equation
2
a z _ J... a z +
2
az + ~z = O .±.
ax 2 0: 2 ay2 x ax X2 '
show that p = X2 and deduce the result of text Example 3.
432 Parlial Differenlia¡ Equations
(b) For Equation (75), show thar
deduee Ihal the general solution of (75) is

z = e'X+Y) 2[F(x - y) + G(x - y)],
and show that this result is equivalent to (77).

(e) When!vI and N are both eonstant, show that the only restrietive eompati-
bility equation reduees to the eondition (74). [Notiee that part (b) is a speeial ease.]
(d) For the equation
show that p does not exist.

44. Show rhat the general solution of the equation
2
a°tp _ ~(Ax
aax2rp + ay> x
+ m) ~ -
ax
~(By
y
+ n)~
ay
-'- [ m(m
I
+
XZ
1) , n(n
I
+
y2
1) + 2Am , 2Bn...L A2
X T Y "
+ BI]rn =
't'
o
ean be expressed in terms of arbitrary funetions when A, B, m, and /1 are eonstants, and
obtain that solurion. (See Problem 42.)
Section 8.7
45. The funetion z(x, y) is required to satisfy the differential equation
az az
ax - ay = Z,
and to take on the value z(l) = sin 21 along the straight line e a speeified by the equa-
tions x = y = 1 in the xy planeo
(a) By using equations eorresponding to (93) and (94), show that
~: = eos 21 + ~ sin 21, ~: = eos 21 - ~ sin 21
at points of ea.
(b) By using equations eorresponding to (97) and (98), show that
az az -~ sin 21
2 2
= eos 21 _2 sin 21
ax2 4' ax a y 4
at points of ea.
(e) Use the result of differentiating az/a y along ea to show that
az
-
2
=
,
-cos 21 - - ...::... sin 21
ay2 4
at points of ea.
(d) Cheek tbe result of part (e) by using the result of differentiating the governing
differential equation with respeet to y.
Problems 433
46 . (a) Show that the Taylor expaosioo of the solution of the problem considered In
Pr obJem 45, in the neighborhood of the poiot (O, O), can be writteo in the form
?(x, y) = ir (x -+ y) + ]\ (x' - y2) -+

in eonsequenee of the results of Problem 45.
(b) Obtain the solution in the closed form
z = e<X-Y) ,' 2 si n (x + y).
(e) Verify the eorreetness of the result of part (a) by obtaining the leading terms
in the expansion of the closed form in a Taylor series about (O, O) .
47. Verify direetly that the proeedure of Problem 45 fails If the curve C o is taken in-
stead to be the eharaeteristie base curve x = - y = A.
48. (a) lf Problem 45 is modified in s ueh a way that C o is taken as the eharaeteristie
base curve x = - y = A and z: is preseribed as Z(A) = e). along CO, verify direetl y that
az/ax andaz/ay then can be determined in infinitely many ways.
( b) Show that the curve x = A, y = -A, Z = e). is a eharaeteristie eurve of the
dift"erential equation.
49. (a) If Problem 45 is modified in sueh a way that Ca is taken as the eirele x = eos A,
y = sin A, and z is preseribed as Z(A) = feA) along Ca, show that and y az/ax az/a
ean be detennined uniquely at aJl points along Ca exeept the points
(-yl2/2, ..,,12/2) and (--yl2/2, --yl2/2).
(b) Show that Ca is tangent to eharaeteristie base curves at the exeeptional points
of part (a).
50. Consider (he quasi-linear equation
aa Z
x
+ Z aazy l.
(a) Obtain the general solution in the form
F( (z - x, y + ;2 - xz) = O.
(b) Show that the integral surfaee whieh eontains the line on whieh z = x/2 when
y = x has the equation
__ 4x - 2y - X2
(x =;t=. 2).
"- 2(2 - x)
(e) Show (hat the integral surfaee whieh contains the line 00 whieh Z = 2 when
y = x IS
z = 1 + -yI 1 + 2(y - x).
(d) Show that (here are /wo integral surfaees,
z = l ± -yl2(y - x),
whieh eontain (he line on which z = 1 when y = x.
(e) Show that the eurve on whieh z = x + A when y = x2/2 + Ax + B lies on
infini(ely many integral surfaees.
[Notiee thal the suffieient eondition stated in the text, [or the existenee of a unique
solution, he re beeomes the requirement that z'7'= dyjdx everywhere along C, that this
requirement is violaled when x = 2 in part (b) and for all x in parts (d) and (e), and
that the eonsequenees of the violation in those eases are quite dissimilar.]
Section 8.8
51. The solution of Laplaee's equation
a+
-
2z
-a =z
0
2
ax2 a y2
is required in the neighborhood of the line C o: x = y = A, subjeet to the requirement

that z = O along C o and that azjax _ p = A along Co.
(a) Show that the strip eondition requires that azjay = q = -A along Co.
(b) With the notation of Equation (112), verify that Equations (116) to (118) take
the forrn r + t = O, r + s = 1, s + t = -1, so that there must folJowr = 1, s = O,
and t = -1 along Co.
(e) Deduee that, at the origin (O, O), there follows
z = 0, az = O
ax '
a2 z
ax ay = O,
and henee show that the Taylor series expansion of z near the origin must be of the
form
1 I 2 1 2
z(x, y) = O + 0(0.'( + Oy) + TI(x + 2·0xy - y2) + ... = 2(x - y2) + ...
(d) Verify the eorreetness of the result of part (e) by showing that the funetion
z = :!:(X2 - y2) in faet satisfies the eonditions of the problem (so that ¡he remaining
terms in the expansion all vanish).
52. (a) Show that the eharaeteristie base eurves of the equation
2 2 2
a z + 3 a z + 2 a z _ az = O
ax2 ax ay ay2 ax
are the lines y = x + C I and y = 2x + C2.
(b) Show that, along the eharaeteristie base eurve C o : x = y = A, the Strip eondi-
tion takes the form z' = p + q (where z' = dzjdA) and the eompatibility eondition
beeomes p' + 2q' = p.
(e) Verify direetly that, if z, p, and q are preseribed along C o in sueh a way that
these two eonditions are satisfied, then Equations (116) to (118) permit any one of the
seeond derivatives, say s, to be ehosen arbitrarily along Co, after whieh there follows
r = p' - s and t = q' - s along Co.
[In the next step, eonsisteney of the equations governing Zxxx, Zxxy, and Zxyy on C o
provides (he additional requirernent r' + 2s' = r whieh, when eombined with the fust
result of part (e), finally determines s along C o as any solution of the equation
s' + s = p" - p', and henee as a funetion depending upon an arbitrary constan t.]
Problems 435
53. Consider lhe initial-value problem

az 2 az
ax2 = ay'
z(x, O) = F(x), =x!x , O) = G(x), Z/x, O) = H(x),

assuming that F, G, and H have deriva ti ves of al! orders.
(a) Show that the differenlial equation is parabolie, that the line y = O is a
eharacteristic base curve, and that here Equation (121) is nonrestrietive (so that a new
compatibility eondition is to be expected).
(b) Show lhat the strip condition (110) requires that
G = F ',
obtain ¡he addilional conditions
r = q, r = p ', s=q
when y = O, from (116) to (J 18) [where, for example, p ' = dp(x,O) j dx], and deduce
the compatibility condition
p' = q,
replacing (121). Thus deduce that F, G, and H must be such that
H = G' = F"
and hence also
whereas t = Zyy is as yet undetermined on the line y = O.

(e) Obtain the additional conditions
Z.Y:~;t; = s, Zx;ry = t
, ,
and Z X.TX = r , Zx:c:y = s,
when y = O, and deduce that
1 = s' = Fiu
and also
Z xxx = F "',
when y = O, whereas Zyyy(x, O) is not yet detennined.
(d) Thus obtain (formally) the expansion
z(x, y) = F(xo) + (x - xo)F'(x o) + yF"(xo) + di [(x - xo)2F"(xo)
+ 2(x - xo)yF"(xo) + y2F'"(xo)] +

and also the expansion
z(x, y) = F(x) + [1 F"(x) + r~ pi" (x) + ...
54. (a) Show that, if cp
= y - x and ljI = y - 2x are takeo as new independent
variables in Problem 52, then there follows
:x = -UCP + :cp). 2
436 Partial Oifferential Equations
and the differential equation takes the normal form
(b) Determine constants IX and P such that Ihe change of variables

z = uea.rp + ptp
transforms the normal form obtained in par! (a) to an alternative normal form
a'u
arp al{! = yu,
where y is a constant.
55. (a) By writing r = x + el and s = x - el, transform the equation
it:.P. 1
"2
a,rp
a2 = h(x, 1)
ax 2 -
e 1
to the equation
J.:.!L =
ar as
_1 h
4
(r +2 s.' r2e- S)' .
(b) Use the result of part (a) to obtain the general solution of the original equa-
tion when h(x, 1) = I and also when h(x , 1) = cos (x + el).
Section 8.9
56. Show Ihal the solution of the equation
az + az = O
ax ay ,
for which z = Ixl along the x axis, is o[ the [orm z = Iy - x l. CNotice thal Ihe "cor-
ner" at the origin is propagated along the characteristic base curve y = x which passes
through the origin.)
57. Show that, if z satisfies the equation
aoz 2z a
ax2 - ay2 = O,
and i[ a 2z¡ax 2 has a jump of one unit across the line Ca: y = x in the xy plane, then
a 2z/cax a y) must have a jump o[ -1 and a 2z/a y2 a jump of + 1 across Ca, if az¡ax
and az/ ay are to be continuously differen tiable along Ca.
58. If z = 1(x + y)2 when y :> x and z = 2xy when y <: x, show that z, az/ax, and
az/a y are conlinuous on the Jine y = x and that z salisfies the equation
a-----,z - -a
2 a z" = 2
ax· y-
0,
and al so verify that z possesses the properties specified in Problem 57.

59. lf z satisfies Laplace 's equation
a 2z , az' _
ax' , ay2 -
°,
Problems 437
show that the assumption that aOzl azx has a jump aeross the line Ca: y = x in the xy
plane leads lo a eontradietion if azl ax and azl a y are to be eontinuously differentiable
along C o.
Section 8.10
60, (a) By replaeiog ct by iy, where i 2 -1, in Equation (67), show that formally we
obtain a solution of Laplaee's equation
~:~ + ~:~ = O
for whieh z(x, O) = F(x) and az(x, o)/a y = G(x) in the form
'" =
1
2[F(x + iy) + F(x - iy)] + 2i
1 IX+;Y G(e;) dé,.
x-;Y
(b) Show that if we speeify F(x) = x, G(x) = eX, then an expression

z = x + eX sin y
is obtained, and verify that this expression satisfies Laplace's equation everywhere and
also satisfies the conditions imposed along the line y = O.
(e) Notice that the formal solution is meaningless if, for example, we specify that
F(x) be zero along the negative x axis and unity along the positive x axis. Show also
that if
F(x)
J G(x) = O
the formal solution becomes
z
1 + X2 _ y2
(1 + X2 - yZ)2 + 4x 2 y2'
and that this expression is not defined at the points (O, ± 1). In particular, show that
this formal solution becomes infinite as either of these points is approached along the
y axis, so that the differential equation is not satisfied at this point.
[ln general, the initial-value problem for Laplaee's equation (or any elliptic equa-
tion) has no solution unless the preseribed values are sufficiently regular. Also, even
though a solution be determined by the aboye method, it may be valid only l1ear the
curve along which the function and its normal derivati ve are preseribed. A deeper
insight into the problem is afforded by the theory of analytie funetions of a eomplex
variable (Chapter JO).]
61. Suppose that z(x, y) is to satisfy the equation
and the eonditions

z(x, O) = f(x), Zy(X, O) = g(x)
along lhe entire x a x is .
(a) By appropriately modifying the deri va tion of (67), show that
z(x, y) = 21 [f(x + y) + f(x - y)] + 2I J"+Yg(e;) de;,

x -y
438 Partial Differential EquatiQns
(b) In the special case when
{~X2
(x:> O),
f(x) =
(x -< O),
and g(x) O, show ¡ha! z = x" + y2 when x:> Iyl, z = 1(x + y)2 when y :> Ixl,
=
Z = 1;(x - y» when y -< Ixl, and z = O when x -< Iyl. Also verify that z, z" and Zy
are continuous everywhere but ¡hat the jump in f " (x) propaga tes into jumps in Zx .y>
Zxy and Zyy along the lines y = ±x.
62. Writing the general solution of the equation
a2 z = O
ax ay
in the form z = f(x) + g(y), show that the initial conditions
z ( x, O) = F()
x , p ( x, O) -= az(x,
ax O) = F'( x ) , q (O)
x, -= az(x,
a y O) = G( x ) ,
imposed along the infinite line y = O, lead to ¡he conditions

f(x) + g(O) = F(x), g'(O) = G(x).
Hence deduce that the problem has no solution unless G(x) is prescribed as a constant,
say G(x) = e, in which case the problem has infinitely many solutions of the form
z = F(x) + g(y) - g(O),
where g(y) is any (differentiable) function of y for which g'(O) = C. (Notice that the
line y = °
in the xy plane is a characteristic base curve in this case.)
Section 8.11
63. Show by the methods of Section 8.11 that the simultaneous equations
aU +av =0
ax ay ,
au
2 ax + (au
a y + av)
ax tan 2 V = O.
possess the same characteristics [Equations (l53a, b)) as the problem of that section.
64. Use the results ofProblem 7 to show that if the Jacobian a(.;, r¡)¡a(x, y) is not zero,
the dependent and independent variables in EquatioDs (159) may be interchanged 10
give the equations
y. - x. tan v = 0, y{ + x{ cot v = 0,
where v = 1;('; - r¡). [Notice that these equations are linear in x and y, whereas Equa-
tions (159) are quasi-linear in .; and r¡, since v depends upon <; and r¡.)
9
Solutions of Partial Differential
Equations of Mathelllatical Physics
9.1. Introduction. Many linear problems in mathematical physics involve

the solution of an equation obtained by suitably specializing the form
V2rp =
at + f.i arp
A. azrp2
at + h ' (1)
where h is a specified function of position and A. and f.i are certain specified
physical constants. Here the operator V 2 is the Laplacian operator in the space
of one, two, or three dimensions under consideration and is of the form
(2)
in rectangular coordinates of three-space. The unknown function rp is then, in

general, a function ofthe position coordinates (x, y, z) and the time coordinate t.
In particular, Laplace's equation,
V2rp = O, (3)
is satisfied, for example, by the velocity potential in an ideal incompressible
fluid without vorticity or continuously distributed sources and sinks, by gravi-
tational potential in free space, electrostatic potential in the steady flow of
electric currents in sol id conductors, and by the steady-state temperature dis-
tribution in solids.
Also, Poisson's equation,
V2rp = h, (4)
is satisfied, for example, by the velocity potential of an incompressible, irrota-
tional, ideal fluid with continuously distributed sources or sinks, by steady-
state temperature distributions due to distributed heat sources, and by a "stress
439
440 Solutions of Partial Djlferential Equations of Mathematical Physics
function" involved in the elastic torsion of prismatic bars, with a suitably

prescribed function h.
The so-called wave equation,
(5)
arises in the study of propagation of waves with velocity c, independent of the

wave length. In particular, it is satisfied by the components of the electric or
magnetic vector in electromagnetic theory, by suitably chosen components of
displacements in the theory of elastic vibrations, and by the veJocity potential
lJ1 the theory of sound (acoustics) for a perfect gas.
The equation of heat conduction,
V'rp = _1 arp, (6)

ex' at
is satisfied, for example, by the temperature at a point of a homogeneous body
and by the concentration of a diffused substance in the theory of diffusion, with
a suitably prescribed constant ex.
The telegraph equation,
(7)
which is a one-dimensional specialization of Equation (1), is satisfied by the

potential in a telegraph cable, where A = Le and !1 = Re, if leakage is neglected
(L is inductance, e capacity, and R resistance per unit length).
Differential equations of higher order, involving the operator V\ also are
rather frequently encountered. In particular, the bi-Laplacian equation in two
dimensions,
4
a 4 rp 4
V·rp = V'V2rp = - 4
ax
+ 2 axa ay-
rp
2
a rp
, + ay 4 = 0, (8)
is involved in many two-dimensional problems of the theory of elasticity.

The solution of a given problem must satisfy the proper differential equa-
tion, together with suitably prescribed boundary conditions and/or initial condi-
tions, the nature of these conditions depending upon the problem. However, the
requirement that the solution be single-valued in the relevant region <R may
substitute in part for boundary conditions when that region is not simply con-
nected. In addition, it is to be assumed always (unless the contrary is stated) that
the solution is to be bounded in every finite subregion of <R.
The process of attempting to obtain the most general solution of the govern-
ing differential equation and then specializing it so as to satisfy the given
conditions (as in d' Alembert's solution of the problem at the end of Section 8.5)
is rarely feasible. Instead, one often seeks to determine a suitable set of particular
solutions, each of which satisfies certain of the conditions, and then attempts to
combine a finite or infinite number of such solutions in such a way that the
9.2. Heat Flow 4.:11
combination sa ti sfies all the prescribed conditions. Here one mu st rely upon
knowledge lhat the problem doe s nOl in fact possess more lhan o ne solution .
The particular so lutions frequently are obtained by a certain method of
separation 01 variables. It may be mentioned that, although this method is re-
stricted in a mathematical sense to a comparatively narrow range of differential
equations, fortunately thi s range includes a large number of those equations
which arise in practice.
Methods which are somewhat more direct, but which again are applicable
only to specific classes of problems, are considered in Sections 9.1 S and 9.16
and in Section 11.8.
It should be emphasized that most problems arising in practice, which
involve partial differential equations, are in fact not amenable to exact analytical
solution and either must be dealt with by approximate numerical methods, or
must be simulated by problems which can be treated analytically.
Before illustrating the treatment of sorne analytically solvable problems,
we first consider the mathematical formulatíon of the study of heat flow in
space, as a basis for so me of the se problems.
9.2. Heat Flow. We consider the flow of heat in a region in space such that
the temperature T at a point (x, y, z) may depend upon the time t. For any
region <R bounded by a closed s urface S, the rate at whích heat fiows outward
from <R through a surface element du with unit outward normal n is given by
-dQ, = -K
aT
an du = -K(VT) • n du,
where K is the thermal eonduetivity of the material. (The negative sign clearly
corresponds to the faet lhat heat f10w is in the direetion of decreasing tempera-
ture.) Thus the net rate of heat flow into <R is given by
Q, = +~s K(VT) • n duo (9)
However, the rate at whieh heat is absorbed by a volume element do is given by
dQ2 =
aT
sp TI do,
where s is the speeifie heat and p the mass density. Hence, if there are no sources
or sinks in <R , the rate of heat flow into <R is also given by
(10)
Before equating Q, and Q 2' we first transform Equation (9) to a volume

integral by making use of the divergenee theorem (Seetion 6.13), and so obtain
(11 )
442 Solutions of Partial Dilferential Equations of iVIathematical Physics
when K does not vary with position.t The requirement Q¡ = Q2 then becomes
' l' r T - sp aT
J• J<R ( KV 2 a¡) dr = O; ( 12)
but since this resuli must be true for any region eR not containing heat Sources
or sinks, the integrand must vanish. Hence T must satisfy the equation
( 13)
where we have written
a-, = _
K.
( 14)
sp
The quantity a" is known as the ¡herma! diffusivity of the material.
In particular, in the steady-state condition when the temperature at a point
does not vary with time, the temperature satisfies Lap!ace's equation,
V 2T = O. ( 15)
We now consider two basic problems in the theory of steady-state heat fiow.
(Generalizatíons are considered in Problems 11,12, and 13.) First, it would be
expected intuitively that in the steady state the temperatures at points inside a
given region eR would be uniquely determined if the temperature were prescribed
along the bounding surface S. Second, one might prescribe the rate of (steady)
heat flow outward per unit of area,
--=
dQ - KaT
-,
da an
at al! points of the boundary S and require the temperature at internal points.
Let Trepresent a function which satisfies Equation (15) and one of these boun-
dary conditions. Then Equations (126) and (127) of Section 6.14 give the use-
fui results
ff L (VT)2 dr = # T~~s da (16)
and A aT da = o. (17)
'jJ san
Equation (I7) is readily interpreted as requiring that the net fiow through the
closed boundary S must be zero, as must obviously be the case in steady-state
fiow without sources or sinks inside eR. Thus aT/an cannot be specified in a
perfectly arbitrary way on S, but its mean value on S must be zero.
Assume now that two solutions T) and T 2 exist, both satisfying Laplace's
Equation (I5) and both taking on the same prescribed values as a point on the
boundary S is approached from the interior of eR. Then clearly the difference
tU K is a function of position, KV 2 T must be replaced by V . KV T in Equations (11) and (12),

and (13) is replaced by V • KVT ~ sp aTla/.
9.3. Steady-State Temperature Distribution in a Rectangular Plate 443
T, - TI also satisfies (15) , and hence may be substituted for T in (16). But
since T 2 - TI takes on the value ::ero at all points of S, the rígbt-hand síde of tbe
re s ultant equation vanisbes, and there follows
( 18)
Now sínce the integrand is nonnegative , it must itself vanish everywhere in <R,
and hence we ha ve
T, - TI = C = constant. ( 19)
Finally, since T 2 - TI vanishes on S, there follows c = O and consequently
T 2 = T I' That is, ¡here can be only one solution of Laplace's equation valid in <R
and taking on prescribed values on the boundary S. In a similar way we find that
two solutions of Laplace's equation valid in <R and having the same specified
value of the normal derivative on the boundary S can dijfer at most by a constant,
noticing tbat here T 2 - TI is not necessarily zero on S.
Tbe two problems considered are known respectively as the Dirichlet and
Neumann problems. The solution to the Dirichlet problem, wbere the function
itself is prescribed on tbe boundary, is unique, whereas the solution to tbe
Neumann problem, wbere the normal derivative is prescribed on the boundary,
is determined only to within an additive constant. Tbese statements clearly
apply to the so lution of any problem governed by Laplace's equation, regardless
of any pbysical interpretation of the unknown function.
Although otber types of boundary-value problems involving Laplace's
equation may occur, tbe two discussed here are of most frequent occurrence.
Next we consider the solution of several s imple problems of tbe Dirichlet
type. Although, to fix ideas, we choose to identify the quantity to be deter-
mined with s teady-s tate tempera tu re, the re sult s may be equally well interpreted
in terms of man y otber physical quantities whicb also satisfy Laplace's equation.
9.3_ Steady-State Temperature Distribution in a Rectangular Plafe_ Suppose

tbat the tbree edges x = O, x = 1, and y = O of a thin rectangular plate are
maintained at zero temperature,
T(O, y) = O, (20a)
T(l, y) = O, (20b) T=f(x)
T( x, O) = O, (20c)
and that the fourth ed ge y = d is maintained
at a temperature distribution f(x), T=O T=O
T (x, d) = f(x), (21)

until steady-state conditions are realized
(Figure 9.1). The temperature distribution
tbroughout the plate is required. Thus we Figure 9.1
444 Solutions of Partial DilTerential Equations of i'<!alhemalical Physics
must determine that solution 01' Laplace's equation in two dimensions,
(22)
which takes on the prescribed boundary values (20) and (21).

The method of separation ofvariables consists of seeking particular "product
solutions" of (22) in the form
Tp(x, y) = X(x) Y(y), (23)
where X is a function of x alone and Y is a function of y alone. If (23) is intro-
duced into (22), there follows
dOX y
dX2
+ Xd 2 y _ O
dy2 - ,
or, after separating the variables,

1 d2 y
(24)
Since, by hypothesis, the left-hand member of (24) is independent of y and the

equivalent right-hand member is independent of x, it follows that both sides
must be independent of both x and y, and hence must be equal to a constant. If
we call this arbitrary constant k2, there follows
2
d X2
dX
+ k 2 X= O' (25a)
2
d y _ k 2y = O (25b)
dy2 .
Thus we see that the product (23) will satisfy (22) if X and Y are solutions of
(25a, b), regardless of the value of k. Because of the linearity and homogeneity
of (22), it follows that any linear combination of such solutions, corresponding
to different values of k, will also satisfy (22).
We next notice that three of the boundary conditions also are homogeneous.
Thus, if each of the particular product solutions is required to satisfy (20a, b, c),
any linear combination will also satisfy the same conditions. Equations (20a, b)
will be satisfied if
X(O) = O, (26a)
X(l) = O, (26b)
whereas (20c) implies the condition
Y(O) = O. (27)
Equations (25a) and (26) constitute a previously considered Sturm-Liouville
problem for which the characteristic values are
k = =-,
kn
nTC
(n = 1, 2, 3, ... ), (28)
9.3. Steady-State Temperature Distribution in a Rectangular Plate
445
and the corresponding solutions (characteristic functions) are of [he form
X = Xn = · nnx
A n sm __
,
o
(29)
Corresponding to (28), the solution of(25b) which satisfies (27) is ofthe form
y = yn = B n sinh nlT.y.
, (30)
Thus it follows that any particular solution of the form
(n = 1, 2, 3, ... ), (31 )
where we have written 0 n = AnBn, satisfies Eguation (22) and the three boundary
conditions (20a, b , c) . The same is true for any series of the form
T = ti 0n sin n~x sinh n7Y (32)
if suitable convergence is assumed. It remains, then, to attempt to determine

the coefficients 0 n In (32) in such a way that the remaining condition (21) is
satisfied, so that
~ ( . h nnd) . nnx < x < /).
J()
x =!:-, 0n SIn -,- SIn - / - (O (33)
But, from the theory of Fourier sine series, the coefficients 0n sinh (nlrd//) in
this series must be of the form
. h nlrd
°n SIn -,- = T2 JI J( ) .
o X
nnx d
SIn - / - x,
and hence, writing en = 0n sinh (nnd//), the reguired solution (32) takes the form
_ ~ . nnx sinh (nny//)
T(x, y) -- n..:.... en SIn - / - . h ( di/)' (34)
= I Stn n1T.
I
where
assuming appropriate convergence.

en =
2
T
Jo J(x)
. nlT.X
Sin - / - dx, (35)
If J(x) is, in fact, representable by a convergent Fourier senes of the form

(33) in (0, /), it can be established that the series (34), subject to (35), truly
converges to the solution of the stated problem inside the rectangle of definition.
This situation exists, in particular, when J(x) is piecewise differentiable in (O, /).
It is clear that the solution of the more general problem where T is pre-
scribed arbitrarily along aH four edges can be obtained by superimposing four
solutions analogous to the one obtained here, each corresponding to a problem
in which zero temperatures are prescribed along three of the four edges, al-
though more convenient alternative procedures frequently suggest themselves
in such cases. (Se e Problems 15, 22, and 23.)
446 Solutions of Partíal Differentíal Equations oC Mathematical Physics
The faet of basie importanee is that permissible values of the " separation
eonstant" k 2 were determined by the eharaeteristie-value problem arising from
the presenee of homogeneous boundary eonditions along the two edges x =
eonstant. The additional faet that a homogeneous eondition also was imposed
along one of the other boundaries atforded a simplifieation, but was not neees-
sary to the sueeess of the method (see Problem 15).
By virtue of (23) and (25), any expression of the form
T p = (el cos kx +e 2 sin kx)(e J cosh ky + e. sinh ky) (36a)
is a particular solution of (22) for arbitrary values of k. It is readily verified
that, if the equal members of (24) were set equal to -k 2, rather than +k 2 , the
signs in (25a, b) would be reversed and particular solutions of the form
Tp = (e, cosh kx + e. sinh kx)(e 7 cos ky + ea sin ky) (36b)
would be obtained. Finally, if V is replaced by zero in (25a, b), the solutions
of the resultant equations lead to the further particular product solutions
(36c)
Clearly, exponential forms could be used in (36a, b) in place of the hyperbolic
functions, if this procedure were desirable . However, only the produet solutions
Usted in (31) satisfy the homogeneous boundary e onditions (20).
The choice of the sign of the separation constant assoeiated with (24) was
motivated by the knowledge that solutions of type (36a) would lead to a Fourier
series expansion in the x direetion along the edge y = d where the nonhomogene-
ous eondition is preseribed . However, it should be noted that the alternative
proeess of equating the two members of (24) to _k 2 (rather than +k2) would
not be incorrect (if nonreal values of k are admitted) since it would lcad to the
Sturm-Liouville problem
2
d X _ k 2 X= O
dX2 '
X(O) = X(l) = O
for whieh the eharaeteristic values of k 2 are negative and such that -k 2 =
n 2n 2/12. The characteristic values of k then would be imaginary, of the form
k. = inn/l, but the same produet solutions (31) would be obtained. The use of
+k2 to denote the separation constant here is preferable only because, with
.A. = k 2, the Sturm-Liouville problem determining .A. then is proper (see Section
5.6), and the permissible values of k accordingly then turn out to be real.
9.4. Steady-State Temperature Distribution in a Circular Annulus. Suppose

that the temperature distributions al9ng the inner and outer ares of a circular
annulus are maintained as f, (O) andf,cO), respectively,
T(r!> O) =fl(O), (37a)
T(r 2, O) = f2(0), (37b)
9.4. Steady-State Temperature Distribution in a Circular Annulus
447
until steady-state conditions are realized (Figure

9.2). The temperatures at internal points of the
annulus are required_
If polar coordinates are used, Laplace's equa-
tion in the plane becomes (see Section 6.18)
V2T = J2T
ar2
+ _1 JT
l' al'
+ _1 a 2
T _ O
1'2 a0 2 - -
(38)
Assuming a particular product solution of the

form
Tir, O) = R(r)0(O), (39)
Figure 9.2
we find that Equation (38) becomes
R"0 + _1 R'0
l'
+ _1 R0" = O,
1'2
where a prime denotes differentiation with respect to the argument. By sepa-
rating the variables, there follows
_1 (r2R"
R
+ rR') = _ 0" _ k2
0 - , (40)
where k 2 is the separation constant. This condition implies tbe two ordinary
equations
r 2R" + rR' - k 2R = O, ( 41)
0" + k 2 0 = O. (42)
The sign of the separation constant was chosen in such a way that, with real
values of k, sines and cosines (rather than exponential functions) wil! be intro-
duced in the O direction where expansions along the circles r = 1', and r = 1'2
presumably will be required.
Equation (41) is an equidimensional equation (see Section 1.6) with general
solution
R = Akrk + Bkr- k (k *- O),
(43)
R = Ao + Bo log l' (k = O),
whereas (42) has the solution
o = C k cos kO + Dk sin kO (k *- O),
(44)
0= C o + DoO (k = O).
Thus any express ion of tbe form
T = G o -- b o log l' ;" (c o + do log r)O
+ 2: [(Gkrk
k
~ bkr- k ) cos kO + (ckr k + dkr- k) sin kO], (45)
where k takes on arbitrary nonzero values, is a solution of Equation (38)_

448 Solutions oC Partial Differential Equations of Mathematical Physics
In order that T be sing/e-va/lIed in the annulus, we must take

Co = do = O. (46)
Also, for the same reason, the trigonometric functions must possess a common
period 2n:. In the absence of boundaries e = constant, this reqlliremenf, in the
present case, serves fo determine ¡he permissib/e va/lIes of ¡he separafion constant,
k=n (n=I,2,3, ... ). (47)
Hence we are led to assume the solution of the present problem in the form
~
T = (ao + b o log r) .:.. 1:

n- 1
[(anr n - bnr- n) cos ne
+ (cnr n (48)
The boundary conditions (37a, b) then take the form
~
f,(e) = (a o + b o log r,) + 1: [(anr~ +

n- 1
b/i n) cos ne + (cnr~ + dnr, n) sin neJ,
(49)
f,(e) = (ao + b o log r 2) + 1: n= I
[(anr, + b/,n) cos ne + (cnr, + dnrin) sin nej,
so that, according to the theory of Fourier series , the constants a o and b o are
determined by the equations
l -2n
a o + b o log 1"1 = 2n: f,(e) de, (SOa) 1
ao
,
- j- b o log r 2 =
1
2n: 1
"21t
12(e) de, (SOb)
the constants a n and b n (n > O) by the equation s
anr'í + bnr,n = *1 -h
-h
II(e) cos ne de, (SOc)
anr'i + b nr1. n = _1
n:
Jo f2(e) cos ne de, (SOd)
and the constants Cn and d n by the equation s
cnr~ + dnr,n = -
1 J-zn I,(e) SIn. ne de , (SOe)
n: o
n J-
Zn
c/; - dnr"i. n = 1 o 12(e) •

SIn ne de. (SOf)
Two limiting cases are of particular interest. First, in the Iimiting case
r, = O, the region considered becomes the interior of the circle r = r 2 · In order
that the left-hand members of (SOa, c, e) remain finite as r, ~ O we must take
(/1 = 1, 2, 3, ... ). (51)
9.4. Steady-State Temperature Distribution in a Circular Annulus 449
If we also wrile An for a/ e.

for c nrl, a for r 2, and Ice) for /2(e), we deduce
that ¡he solu/ion o/ the problem
"
V'T=· O (r-<a),
(52)
T(a, e) = Ice)
is o/ the /orm
T = Ao -t- J:, (:) n (A n cos ne ...:... e. sin ne) (r -< a), (53)
where
Jor
h
I
Ao = 2n Ice) de,
An = - 1 fh Ice) cos ne de (n = 1, 2, 3, ... ), (54)

n o
en = -1
n
f2' ICe) sin ne de
o
(n = 1, 2, 3, ... ),
from (50b, d, f).

In this case /, (e) is merely the constant value, say T o , of the temperature
at the center of the circle r = a, so that (50c, e) reduce to the trivial identity
O = O. Also, Equation (50a) becomes
Ao
I
= -2n 12n To de = To· (55)
o
Since this result must be compatible with the first equation of (54), which
states that Ao is the mean value of the temperature distribution along the circle
r = a, we conclude that ¡he temperafure at fhe center o/ the eirele is {he mean o/
(he temperalUre distribulion along the boundary o/ (he eirele. This result also
follows directly from (53) when r = O.
In the second limiting case r 2 ~ 00, the region considered becomes (he
exterior o/ the cirele r = r ,. In this case we must take
b o = a. = en = O (n = 1, 2, 3, ... ) (56)
in (50b, d, f). Equations (50d, f) then become, in the Iimit,
Jor /o(e) cos ne de Jor 2n /2(e)

2n
O= = sin ne de (n = 1, 2, 3, ... ),
(57)
where /2(e) is the prescribed temperature distribution on the circle with in-
creasing radius r2. These relations can be true only if /2 (e) is constant in the
limit. Thus the temperature must approach a uniform value, say T_, as r ~ oo.
Equations (50a, b) beco me
Jor
ln
1
Ao = a o = 2n /,(e) de = T_, (58)
450 Solutions of Partial Differential Equations of Malhematical Physics
showing that T= is the mean of the prescribed temperature distribution on the

inner boundary, and Equalions (50d, f) reduce to O = O.
If we al50 write Bn for r, "b n> D" for r,"d", a for r. , and feO) for f. (O), we
deduce that (he sohaion of ¡he problem
V'T = O (r :> a) ,
(59)
T(a, O) = feO)
is of ¡he form
(r >- a), (60)

where
Bn = -1
n
J'n feO) cos nO dO
o
(n = 1,2,3, . . . ), (61 )
Dn = -1 J2n feO) sin nO dO (n = l. 2, 3, ... ).

n o
9.5. Poisson's Integral. We now show that, if the values given by (54) are
introduced into (53), then the resultant series can be summed, and the solution
can be expressed as an integral. For this purpose, we first replace . the dummy
variable O in (54) by rp, to distinguish this variable from the current variable O
in (53). The introduction of (54) into (53) then leads to the relation
1
T(r, O) = 2 n f" f(rp) drp + ! "~ (~ r[ cos nO fn f(rp) cos nrp drp
+ sin nO J:n f(rp) sin nrp drp 1

= n1 Jor'n f(rp) [2+
I "
"~. (~)" (cos nO cos nrp + sin nO sin nrp)] drp,
or
T(r, O) = - 1
7t
J'n f(rp) [-21 +
o
I:
= ( .!..-)" cos n(rp
n: I a
- O) ] drp (r < a), (62)
assuming the legitimacy of the interchange of summation and integration in the

first form. But sincet
cos n(rp - O) = Re (elnl~-q») ,
t"Re" means " real part or" and UIm H means himagi nar y parl of>? in the sense lhal
f= Ref + i 1m'!' so Lhat Imfis rea/(see Section 10.1).
_.
9.6. Axisymmetrical Temperature Distribution in a Solid Sphere 451
there fol1ows
J; (~ r cos n(rp - () = Re Lt, [~ ei(.-s'Jl

S
= Re { (rl a)e il . - , }
I - (rl a)e' (. -" (r < a)
= Re {
[1 -
(rl a)[ei(·-s, - (r l a)]
(r l a)e i(· "][1 - (r j a)e {(o . ' ]
l
(rl a) cos (rp - () - (r 2 1é)
- _ 2(rl a) cos (rp _ B) + (r2 I aO) (r < a).
Thus the quantity in braces in (62) beco mes
-2"(r)" a -r·
2
-!-' "_,
.2: -a cos n( rp - () = - 1 ...,.,------,.---=-----;,-:...--"'--;----.;
2 a 2 - 2ar cos (rp - f) + r 2 (r < a) ,
(63)
and, noticing that
f(f) = T(a, (),
we find that Equation (62) take s the form
T( r, () =
I
271: Jr"
o a 2 _
a" - r"
2 ar cos (e _ rp) + r 2 T( a, rp) drp (r < a), (64)
Tilis remarkable integral formula, due to Poisson, expresses the value of a

harmonic function T at all points in a circular disk of radius a in terms of the
values of Ton the bounCÚlry of the disk.
It can also be derived by methods which do not presume the validity of (53),
or of the interchange of summation and integration which led to (62), and it
holds, in particular, when T(a, () = f«() is piecewise continuous in (0, 271:), The
validity of the series representation (53) can be guaranteed only when somewhat
more stringent conditions are imposed, such as the requirement that f«() be
piecewise differentiable , The corresponding integral formula for the exterior
problem is treated in Problem 35,
z
9.6. Axisymmetrical Temperature Distributioo
iD a Solid Sphere. On the surface of a solid sphere,
let the temperature distribution be prescribed in
such a way that it is symmetrical with respect to a
diameter. The temperature distribution inside the
sphere is to be determined. Using spherical coor- y
dinates (Figure 9.3), with the z axis coinciding
with the axis of symmetry, and noticing that the
temperature T is independent of f), we recall x
that Laplace's equation takes the form (see Figure 9.3
_"
4 :>- SolutíODS of Partíal Dífferential Equations of l'vIathematical PhYSICS
.
Section 6.18)
~ ( ,a T) -.L 1 a ( . a T) _ o. (65)
ar r ar 'sin rp arp sm rp arp -
The boundary condition on the surface r = a is then
T(a, rp) =f(rp). (66)
lf a product solution is assumed,
Tp(r, rp) = R(r)(rp), (67)
Equation (65) is separable into the form
I
(r2R'), = - <1> l (<1>' sin rp)' = k2, (68)
R Slll rp
where k 2 is a separation constant. This condition is equivalent to the equations
(r 2 R'), - k2R = r 2 R" + 2rR' - k2R = O (69)
and l d ( . d<l», k' <1> O (70)
sin rp drp Slll rp drp -, - = .
Equation (70) is brought into accordance with Equation (169), Section 4.12,
if we write
k 2 = n(n + 1). (71)
The general solution of (70) is then of the form
<1> = AnPn(cos rp) + BnQn(cos rp), (72)
where P n and Qn are Legendre functions. In order that the solution be finite
on the z axis rp = O, n, we must restriet n to integral values and take
(73)
This follows from the fact that the Legendre polynomials (for which n is an
integer) are the only Legendre functions of type (72) which are finite at both
rp = O and rp = n (see Section 4.12). With the notation of (71), the general
solution of (69) is obtained in the form
(74)
To avoid infinite temperature at the center of the sphere (1' = O), we must set
(75)
Hence we are led to assume the desired temperature distribution in the form
T(r, rp) =
~
~o en
(rti )n Pn(COS rp) (O < rp < n), (76)
where we have written AnCn = c.la n. The boundary condition (66) requlres
that the constants en be determined so that the representation
=
f(rp) = L: cnPn(cos rp)
n=O
(O < rp < n) (77)
9.7. Temperature Dislribution in a Rectangular Parallelepiped 453
is valido If we introduce the new variable

Ji. = cos rp, (78)
the function f(rp) becomes a new function of Ji. , say F(Ji.), such that
F(Ji.) = f(cos - I Ji.), (79)
and (77) becomes
F(Ji.) = I:o cnPn(Ji.)

n=
(- I < Ji. < 1). (80)
The results of Section 5.14 [Equation (213)] then given the result
(81 )
or alternatively, returning to the original variable rp and USIng the relation

F(cos rp) = f(rp),
Cn =
2n + l Jor
2
n
f(rp)P,(cos rp) sin rp drp. (82)
If the first n derivatives of F(Ji.) with respect to its argument are continuous
for -1 < Ji. < 1, the second alternative form given by Equation (213), Section
5.13,
(83)
may in certain cases be more useful for the determination of Cn'

The general Dirichlet problem for a sphere, when axial symmetry is not
presumed, is considered in Problem 47. In addition, certain such problems for a
right circular cylinder are treated in Problems 39,40,41, and 43.
9.7. Temperature Distribution in a Rectangular Parallelepiped. Let the

temperatures of five faces of a rectangular parallelepiped be maintained at zero,
T(O, y, z) = T(lI>Y' z) = T(x, O, z) = T(x, /2' z) = T(x, y, O) = ° (84)
and suppose that the sixth face is maintained at a prescribed temperature dis-
tribution
T(x, y, d) = f(x, y) (85)
until steady-state conditions are attained (Figure
9.4). We investigate the resultant distribution
of temperature in the interior.
If we assume a product solution of the rele-
vant equation
(86)
f---- I,---t"
in the form
Tp = X(x)Y(y)Z(z), (87) Figure 9.4
454 Solutions of Partíal Differential Equatiol1s of lVlalhematical Physics
the equation may be separated in the form

X" Y" Z"
- X = Y + L = kf, (88)
the separation here depending upon the faet that the first member is indepen_
dent of both y and z and the seeond equal rnernber is independent of x. Henee
we rnust have
X" + krX= O (89)
and, after a seeond separation,
Y" Z" .,
-y= Z - q = k:. (90)
Thus Y and Z are determined by the equations

Y" + kiY = O, (91)
Z" - (k¡ -r- ki)Z = O. (92)
The homogeneous boundary eonditions (84) are satisfied by the produet
solution if the faetors satisfy the eonditions
X(O) = X(ll) = O, (93)
Y(O) = Y(l,) = O, (94)
Z(O) = O. (95)
We thus obtain from Equations (89), (91), (93), and (94)
mn (m=I,2,3, ... ), (96)
k l
T
X = Xm =
· mnx
A m SIn -'-1- , (97)
nn (n = 1, 2, 3, ... ), (98)
T;
- . nny
Y = y. = B.sm--¡;-. (99)
If we write further
k
2
1 +k 2
2 = n
2
Cr, + n,~2) =_ km.
m~ 2
1m 2 n2
or k mn = 'lLy -'2 + f2 '
1 2
(100)
the solution of (92) satisfying (95) beeornes

(101)
Thus, writing Qm. = AmB.Cm., we are led to assume the desired solution in
the form
T( x, y, z ) -- ~ ~ . mnx . nny . h k (102)
~ ~ a mn SIn - , - SIn - , - SIn mnZ.
m- I 11= 1 1 Z.
9.7. Temperature Distribution in a Rectangular Parallelepiped 455
This expression formally satisfies (86), as well as conditions (84), for arbitrary
values of the coefficients a mn • Ir remains, then, to determine these coefficients in
such a way that the remaining condition (8S) is satisfied. If we introduce the
abbreviation
( 103)
this condition takes the form
j( x, y ) =
;¡:, ~ . mnx . nnv (104)
,:.... ,:.... C mn
m= ¡ n= L
Sin - , - Sin - , '
1 2
(O < x < '" O < Y < '2)'
Thus the coefficients c mn are the coefficients of the doub'e Fourier sine-series
expansion of j(x, y) over the indicated rectangle.
These coefficients are readily determined by a simple extension of the
methods used in earlier work. If both sides of (104) are multiplied by
s¡'n (pnx/',) sin (qnY/'2)'
where p and q are arbitrary positive integers, and if the results are integrated
over the rectangle, there follows
J,rol, J,rol, j(x, y) sin P,~x sin q~y dy dx
=
I: =
I:
m= I n - 1
C mn
JI' JI'
o o
. pnx . qny . mnx . nnY
Sin - , - Sin -,- s10 - , - Sin -,-
I 2 I
dy dx.
2
(lOS)
The double integral on the right can be written as the product
(f sin p~x sin m,~x dX) (f sin q,:y sin n~y dy ),

and hence, by virtue of Equations (140) and ( 141), Section S.10, this product
vanishes un'ess p = m and q = n, in which case it has the value
"2'22_- 4
',!?,
'
Thus the double series in the right-hand member of (lOS) reduces to a single
term, for which m = p and n = q, and there follows
C mn =
4
TT
1 2
JI' JI'
o o
. mnx
j(x, y) Sin - . -,-
, - Sin
1
nny dy dx.
2
(106)
With these values of C mn ' and with the notation of (103), the solution (102)
becomes
= = . mnx . nny sinh kmnz .
T(x,y,z) = I: I:cmnsm-,-sm-,-. hk d' (107)
m = 1 n= 1 1 2 SIO mn
where k mn is defined by (100).

Ir is useful to notice that an equivalent approach to the present problem
consists of first seeking a product solution of (86) in the form
Tp = F(x, y)Z(z), (l08)
456 Solutíons of Partíal Dífferentíal Equatíons of r>fathematícal PhYSícs
where the :: factor is treated in a distinct way since only along a boundary z =
constant is a nonhomogeneous condition imposed, and deducing the requirement
( 109)
where k' is the separation constant. Thus F is to be a nontrivial solution of the

two-dimensional characteristic-value problem
a'F a'F
axz + ay' + AF = O,
( 110)
F(O, y) = F(l1> y) = o,
F(x, O) = F(x, Iz) = O,
involving the Helmholi:: equa/ion, and Z is to be such that
d 2Z
dz 2 - AZ = O,
( I 11)
Z(O) = O.
The functions
. mnx . nny
rpmn ( x, y ) = SIn -1- SIn -1-' ( 112)
, 2
where m and n are positive integers, accordingly are characteristic functions of

the problem (110), In correspondence with the characteristic numbers
Amn -
_ 2
k mn -
_
n
2 m-
c~, T
I n~
li
') . ( 113)
These functions are se en to have the two-dimensional orthogonality property

SLrpmn(x, y)rp p.e x , y) dx dy = O lf (p, q) =1= (m, n), (1 14)
where D is the rectangular region (O <:: x <:: 1" O <:: y <:: 12 ) associated with the
characteristic-value problem (1 lO). (See also Problem 45.)
With Zmn(z) correspondingly of the form (l01), superposition again leads to
the assumption (102) and to the ensuing determination of the requíred solution.
9.8. Ideal Fluid Flow about a Sphere. The

preceding problems have all been of the
Dirichlet type. To iilustrate the solution of a
Neumann problem, we investigate the effect of
the presence of a stationary sphere of radius
a in an initially uniform flow of an ideal incom-
pressible fluid (Figure 9.5). We introduce
spherical coordina tes and require that for large
x
values of r the flow be in the negative direction
paral!el to the z axis (rp = O, n). The flow then
Figure 9.5 clearly wil! be independent of the circum-
9.8 . Ideal Fluid Flow about a Sphere 457
feren ti a l angle B. The velocity potential is here denoted by P(r, rp). According
to the re sults of Section 6.15, this function satisfies L ap la ce's equation, which
takes the form
a ( z ap). l a ( . ap)
ar r ar .- sin rp arp SIn rp arp =
o. (1 15)
Also, the velocity vector V is determined as lhe gradient of P ,

P I
V = VP = u, aa -;-- u. - aap
r 'r rp
= u,V, ';" u.V•. ( 116)
[See Equation (16ge) Section 6.18.] At the surface of the sphere (r = a) there
mu s t be no component of V normal to the sphere; that IS, V, = apl ar must
va nish,
ap
r = a: dr = O. ( 1 17)
A s r - , =, the velocity vec tor must approach - Vok, where VA IS the un-
disturbed velocity. Making use of the readily establis hed relation
k = u, cos rp - u. si n rp, (1 18)
we see that as r - , = we must ha ve V - . - Vou, cos rp + Vou. sin rp. Referring
to (/16), it follows that as r ~ =
the radial velocity must satisfy the condition
ap
r-- co : ar - -Va cos rp. (119)
According to Section 9.2, the boundary condition s (117) and (119) aresufficient
to determine P, except for an irrelevant additive constant.
As was shown in Secti o n 9.6, particular product solutions of (l15) may be
superimposed to give solutions of the form
P = I:o [(A.r' +
n=
B.r-'- I )p.(cos rp) + (e.r' + D.r - ·-l)Q.(cos rp)] (120)
for which also
+ [ne. r·- 1
- (n + I)D.r-·-1]Q.(cos rp)} . (121 )
To a vo id infinite velocities along the axis of symmetry, we must take
e. = D. = O, ( 122)
and to avoid infinite velocities as r - . 00, in accordance with (119), we mu s t
ha ve
A. = O (n = 2, 3, ... ). (123)
FinalIy, to sa ti sfy ( J 17) we then must have
nA.a'- 1 - (n + 1)B.a-·- 1 = O
B_ a 2n + 1 A .
or
..
=
n +n J r.
( 124)
458 Solutíons of Partíal Differential Equations of M s rhematica) Physics
Thus. with (122), (123), and (124), Eguations (120) a nd (121) bec om e
P = AoPo(cos rp) + A, (r + ;;") P,(cos rp) ( 125)
and ~~ = AI(I - ~J3)PI(COS rp). ( 126)
1f we recal! [see Eg ua tion (171), Section 4.12] tha t

Po(cos rp) = 1, PI(COS rp) = cos rp,
it follows that the remaining condition (119) is satisfied if
A,=-Vo ' ( 127)
Hence the velocity potential is of the form
P = - vo(r + 2~3,) cos rp + A o, ( 128)
w here Ao is an irrelevant constant which may be set egual to zero. The velocity
vector is of the form
v = -[ Vo( 1 - ;,') cos rp ]u, + [Vo( 1 + ;r',) sin rp ]U~, (129)
by vi rtue of (116).
To find the streamlines, we make use of Eguation (179), Section 6.19, with
hl=h, = I, h,=h.=r, h 3 = h. = r sin rp,
in accordance with Eguation (169b) , Section 6.18. The stream function '11 IS
then determined from the relation
d '11 =
ap .
- drp sm rp
d
r + ap, .
Fr r S in
d
rp rp
- - va[ (r + ;;,) si n ' rp dr + (r2 - ~') sin rp cos rp drp 1

By integration we then obtain
'11 = - iO(r 1- ~') sin' rp - e, (130)
where e is an arbitrary constant. The streamlines are thus the traces of the
s urface s '11 = constant, or, eguivalently,
r2 (1- ;:) sin 2 rp = constant (131)
in the diametral planes B = constant.

A considerably s horter but less direct procedure for solving the problem
specified by (l15), (117), and (119) would consist of guessing initially that the
solution will be of the form P(r, rp) = R(r) cos rp, and of then attempting to
determine R(r) in such a way that this expression satisfies (l15), and such that
R'(a) = O and R'( oo) = -Vo'
9.9. The Wave Equation. Vibration of a Circular Membrane 459
9.9. The \-Vave Equation. Vibration of a Circular iVIembrane. lf we consider

equilibrium of an element (x, x + dx; y, y + dy) of a membrane under a dis-
tributed normal load of intensity ¡(x, y), and denote the deflection in the z
direction by w(x , y) and the tension by T(x, y), the dilferential resultant vertical
component of tensile force normal to the xy plane is found to be
a ( T ax
dx aW) dx dy + ay
a (T aW)
ay dx dy,
ir only small slopes and deflections are considered. Since the distributed load
on the element is given by f dx dy, the dilferential equation satisfied by w is of
the form
( 132a)
or
or, in terms of the invariant operator V,

TV 2 w + (VT) • (Vw) + f = O. (132b)
For small deflections of an initially tightly stretched membrane we may
as sume that T is uniform in the membrane. In addition, if we consider here
only free vibration of the merobrane, and hence replace f by the inertia loading
- p a 2 w/at\ Equation (132b) reduces to tbe two-dimensional wave equation
1 2w a
V 2 W = -"-"""""2 (133)
c- at- '
where 2
c =-'
T (134)
p
In dealing with vibrations of a circular membrane, we introduce polar co-
ordinates and write (133) in the form
a 2w+ l aw l a w
a,:z r al' + -¡:r aB2
2
=
1 a 2w
cz dj2'
(135)
If a product solution is assumed in the forro

w p = R(r)e(f3)T(I), (136)
the usual process of separation leads to equations determining the factors in
the form
T U + co 2T = O, (137a)
r 2R " + rR' + (~2\2 - k 2

)R = O, (137b)
e n + k 2 e = O, (137c)
where co 2 /e 2 and k 2 are arbitrary separation constants. The general solution of
(137b) is of the form [see Equation (130a), Section 4.10]
(138)
460 Solutions of Partial Differential Equations of Mathematical Physics
and, if k ánd Ctl are not zero, (l37c) and (13 7a) gi ve
o = ecos ke + D sin ke, ( 139)
T = Ecos Wl + F sin wt. (140)
In order that the defiection w be single-valued, the funcrions in (139) must
be of period 2n, and hence k must be integral,
k=m (m=0,1,2, ... ). (141)
If k = m = O, (139) does not represent lhe general solurion of (137c). How
ever, rhe missing solurion De is nor periodic. Nexr, ro avoid infinite defiections
ar the center (r = O), we must take
B = O. (142)
Finally, since w must vanish on the boundary r = a,
w(a, e, t) = 0, (143)
we require that each of the product solutions w p' to be superimposed, vanish
when r = a. By virtue of (142) and (141), this condition reqUlres that W be a
solution of the equation
(144)
We notice that the solution W = 0, when m > 0, would reduce the factor R(,.)
to zero (identically), and hence it need nor be considered.
If the nth positive solution of this equation is denoted by W mn '
(n = 1, 2, ... ), (145)
the superposition of permissible product solutions leads 10 an expression for w

in the form
=
w(r, e, t) = m~o =
n~ J m
(W ~nr) [(a mn cos me + b mn sm. me) cos wmnt
+ (cm n cos me + d mn sin me) sin wmnt]. (146)
In addition to the boundary condition (143), we may be given two initial
conditions which prescribe the defiection w and the velocity aw¡at as functions
of r and e at an inirial time l = 0,
w(r, e, O) = I1 (r, e), (147a)
aw(r,e,o)=/( r, e) .
al .,
-
(1 47b)
Thus the constants in (146) must be determined so as to satisfy the relations

=
11(r, e) = m~o =
n~ (a mn cos me + b mn sin me)Jm (W ~n r) , (l48a)
12(r, e) = J:o n~ w"n(c mn cos me + d mn sin me)Jm(W~nr), (148b)

9.10. The H€at-Flow Equarjon. H€at Flow in a Rod 461
when O <: r <: a and O < e

< 2n. The determination follows closely the
procedure used in Section 9.7, if use is made of Equations (l6l) of Section 5.11
and of the corresponding results in Section 5.l3 (se e Problem 51).
ror simplicity, we consider further only the special case when /, is in-
e
dependent of and 1:, = O; that is, we assume that initially the membrane is
deflected into a radially symmetrical/orm and is released/rom res/o The ensuing
motion of interior points is to be determined. In this case only the terms for
which m = O are independent of e,
so that al! other terms may be suppressed.
The initial velocities of aH points wil! vanish if the e and d coefficients are set
equal to zero. Hence , writing w. for WOn> Equation (146) reduces to the form
w = f: A.Jo(w.r)
n ..... 1 e
cos w./ (O <: r <: a , / >- O), (149)
where W n is the 11th (positive) solution of
Jo(w~a) = O (n = 1,2,3, ... ). (150)
If the membrane is initially deflected to the form

w(r, O) = F(r), (151)
where F(a) = O, the coefficients AH must be so determined that
F(r) = ~
2::
ne 1
AJo (We r)
_ n (O <: r <: a). ( 152)
Reference to Equations (179) and (1 85a) of Section 5.13, with f.1.n = w)c, P = O,
and 1 = a, then leads to the determination
(153)
It may be noticed that the motion specified by (I49) is a superposition of

modes having frequencies w./2n, where w. satisfies (I50), the amplitude of
vibration in each such mode varyiog in the radial direction as a Bessel function.
The smallest solution of (150) is given by
w,a
e
= 2.405,
to three place s (see Table 1 of Section 5.13), so that , for example, the fundamen-
tal frequency /, = w, / 2n of a circular drumhead is
/ = 2.405 ~ = 0 .3828 / T, .
, 2:n a V pa-
Conversely, the tension T required lO produce a desired fundamental note IS
given b y
9_10_ The Heat-Flow Equation. Heat Flow in a Rod. We next consider the
one-dimensional problem of heat flow in a homogeneous rod of length 1 with
insulated sides, the temperature depending only on the distance x from one end
462 SolutioDS of Partial Differential Equations of MathematicaJ Physics
of the rod and the time t. The heat-flow equation (13), Section 9.2, then becomes
(154)
where (X is a constant defined by (14). It is apparent from (154) that in the steady
sta te, when aT/at = O, T is a linear function of the distance x.
Initially, the temperature is prescribed along the rod as a function of x, say
T(x, O) = f(x). (155)
In particular, if steady-state conditions exist initially, f(x) must be a linear
function, of the form
f(x) = T\O) + (nO) - T\O» ~ ,
where T\O) and T~O) are the initial temperatures of the ends.
At the instant t = O we suppose that the temperature at the end x = O is
changed to a new value, say T" and the temperature at the end x = 1 is changed
to the value T 2 , and these constant values are assumed to be maintained there-
after:
T(O, t) = T" T(I, r) = T 2 (t > O). (156)
The temperature distribution throughout the rod is required as a function of x
and t.
In a problem of this sort, lacking homogeneous boundary conditions, it is
often convenient to express the desired solution as the sum of two expressions,
the first of which is taken here to be the limiting steady-state distribution
(independent of t) after transient effects have become negligible, and the other
of which is to represent the transient distribution (which must then approach
zero as t increases indefinitely). Thus, if we write
T(x, t) = Ts(x) TT(X, t), + (157)
the function Ts(x) must be a linear function of x satisfying (156), and hence is
of the form
(158)
and TT(X, t) is a particular solution of (154). The function TT must be deter-

mined in such a way that it vanishes when t ~ 00,
(159)
and so that the sum T s + TT satisfies the initial condition (155). AIso, sínce
Ts(x) satisfies (156), it follows that TT must vanish at the ends x = O and x = 1
for al! positive values of t,
TAO, t) = TT(l, t) = O (t > O). (160)
Thus the transient distribution satisfies homogeneous end conditions. It is for
this reason that the steady-state distribution was first separated out.t
tA procedure of the same basic type is used in Problem 22 of Section 9.3, with the fust ("homo-
genizing") part of the solution determined by a quite different method.
9.11. Duhamel 's Superposition Integral 463
Product solutions of (154) satisfying (159) and (160) are readily obtained in
the form
(n=1,2,3, ... ).
Thus, by combining (158) and a superposition of solutions of this type, the

required function T(x, r) may be assumed in the form
T( X, /) -- T --L (T 2 T 1 ) Tx ,~ . -1-
nnx e -n'n',',.l'. (161 )
1 I - I n~ a n SIn
We may verify that ([61) satisfies the end conditions (1 56), and that this solution
approaches the proper steady-state solution as / - , oo. It remains, then, to
determine the coefficients a n in such a way that the initial condi/ion (J 55) is
satisfied, and hence
f(x) - TI - (T 2 - T,) .; = nt. a nsin n~x (O < x < 1). ([62)
The Fourier coefficients a n in (162) are determined in the usual way,
an = T2 JI o
f(x)
.
SIn
nnx
-1- dx
2
+ -(T
nn 2 cos nn - T,), (163)
If we write
2
1 s pI'
A=n 2-0;2- =n-2 -
K
, (164)
the parameter A has the units of time.t Since the exponential factor in (161)
is then of the form e- n "/>, we see that A. is closely related to the time required
for the transient effect to become negligible.
9.11. Duhamel's Superposition Integral. If we think of the problem treated

in Section 9.10 as being defined over a semi-
infinite strip (O -<: x -<: 1,0 -<: / < 00) in a t
distance-time plane (Figure 9.6), the non-
homogeneous conditions along the lines
x = O and x = 1 prescribe T as a constant
along each of these boundaries. The prob-
lem was solved by determining a par/icular
T= T 2
solu/ion T s of the governing differential
equation which satisfies those conditions,
and then using conventional separation-of-
variable methods to determine the "correc-
tion" T r = T - T s , which then is to satisfy
homogeneous conditions along the two
o T ~ ((x) x
boundaries x = constant. Problems 22, 23,
60, and 61 illustrate the use of this proce- Figure 9.6
tIn cgs uníts the relevant quantíties are of the foliowing dímensions:
[si = caJ/(g)CC), [pI = g/cm', [1] = cm,
[K] = cal/(cm)(sec)CC), [A] = seco
464 Solutions of Partial DifIerential Equations of Matbematical Physics
dure in other such situations, where a diffcrent diffcrential equation may be

involved, where a prcscribed function need not be constant along a boun_
dary, and where the corresponding particular solution need not be a function
of only one of the variables (neither of which need be time).
Unless the prescribed functions are of sufficiently simple form, the deter-
mination of an appropriate particular solution may present a challenge. In
order to illustrate an alternative approach to such situations, when the problem
is governed by the one-dimensional heat-flow equation
aZT l aT
axz = (X2di ' (165)
we now consider the case where a rod is initially at zero temperature,
T(x, O) = f(x) = O, ( 166)
and where when t > O the end x = O is kept at zero temperature while the
temperature of the end x = / is varied in a prescribed way with time,
T(O, t) = O, T(I, t) = F(t). ( 167)
As a first step toward the solution of this problem, we solve the problem in
the special case when F(t) is unity. Denoting this solution by T = A(x, t), we
obtain [rom (162) and (163), withf(x) = O, T, = O, T 2 = 1, the result
_ x
A( x, t) - -
+ -2 ~
~
(- I)n SIn
. -mr:x -n',,>
- e (t > O), (168)
/ 11: n~' n /
with the notation of (164).
Next, suppose instead that the temperature at the end x = / is maintained
at zero until a certain time t = '1' and at that instant is raised to temperature
unity and is maintained at unit temperature thereafter. Then it is clear from
physical considerations (and possible to show mathematically) that the resulting
temperature distribution will be zero everywhere when t < ", and will be given
by the result of replacing t by t - " in (168) when t > "; that is, in this case
T = A(x, t - ,,) (t > ,,). (169)
Here t - " is time measured from the instant of change.
Next, suppose instead that the tempera-
T ture is raised abruptly to the value F(O)
when t = O and held at this value until
F(t) t = '" then is again abruptly raised by an
amount F(,,) - F(O) to the value F(,,) at
the time " and held at that value until
t = ' " then is abruptly raised by an
amount F('2) - F(,,) at the time '2' and
so on (Figure 9.7). From the linearity of the
problem it is seen that, at the instant follow-
o ing t = 'n' the temperature distribution
Figure 9.7 corresponding to this step-function approxi-
9.11. Duhamel's Superposition Integral 465
mation to F(/) is given by the sum

T = F(O)A(x, 1) + (F(,,) - F(O)]A(x, t - ,,) + (F('2) - F(,,)]A(x, t - '2)
+ ... + [F('n) - F('".,)]A(x, t - 'n)' (170)
If we write
(171)
Equation (170) can be written in the form
T = F(O)A(x, 1) + n-'
2: A(x, t - 'k+ ,) (~F)
~ ~'k' (172)
b=O L.1.1:' k
Finally, as the number n of jumps becomes infinite, in such a way that all
jumps and intervals between successive jumps tend to zero, the definition of the
integral suggests the limiting form
T(x, t) = F(O)A(x, 1) + I A(x, t - ,)F'(,) d" ( 173)
assuming that F(t) is differentiable. This is a version of Duhamel's principie,

and it gives the desired solution in terms of the basic function A(x, t). An alter-
native form is obtained by an integration by parts,
T(x, t) = F(O)A(x, t) + [A(x, t - ,)F(,)]~:~ - I F(,) : , A(x, t - ,) d"
or, since (a¡a,)A(x, t - í) = -(a;at)A(x, t - í), we also ha ve
T(x, t) = A(x, O)F(t) + 'J o F(í)

aA(x. t -
at
í)
dí. (174)
If we notice further that in the present case A(x, O) vanishes when O <: x < 1,
this result reduces to the form
T(x, t) = I F(,) aA(X'a~ - ,) dí (O <: x < 1). (175)
The integral appearing in (173) or (175) is often known as a superposition

integral.
When (175) is used, the solution of the given problem takes the form
T(x, t) = 2
nA
f:'
n-
(_I)n+ 'n[J' F(í) en',!> dí] e- n""> sin nnx,
o I
(176)
whereas (173) leads to the equivalent form
T(x, t) = 2..F(t)
I
+2
n n~'
f: (-I)"[F(O)
n
+ J' o
F'(í) en";> d,]e-n',n. sin nnx.
I
(177)
Although the second form appears to be somewhat more complicated than the
first, it has the advantage that in many cases the convergence of the infinite
series in (177) is more rapid than the convergence of (176).
466 Solutions of Partial Differential Equations of ]\'Jathematical Physics
This follows from (he fac¡ rhar since rhe rerm (x//)Ft!) satisfies the prescribed
conditions al x ~ O and al .\' = / for all 1 _ O, the series par! of (177) repre-
sents a function which vanishes when x = O and when x = / for 1> O, and hence
i t then converges to that funclion for all x such that O :<: x -< / (when F is dif-
ferentiable). On the other hand, since Trx, t) does nOI vanish when x = , and
1 > O unless F(I) = O, lhe series in (176) does nOI converge to the solution
when x = /, but only when O -< x < /. Convergence of rhat series accordingly
will be slow, particularly when .\" is near /.
Example. Suppose rhat the temperature ar the end x = i is increased uniformly with
time from a zero value, at the rate of T o degrees per second,
F(I) = Tal. (178)
Then (176) gives
2AT
T(,v. t) = - -o
1! n-;\
-'
¿:; (-1
11 3
)n+\ in" -
:_A
(l - e- n " " ) ' 7
sinTn . ,x, ( I 79a)
and (177) gives the alrernative form

.
T( x,!)= T 0-/["
x 2A To
-7T.
..s"-' (- 1) n I -
/13
e- n ={. ;. . n1!x
Sin - / - . (] 79b)
n=l
The equivalence of these forms is "<orified by noticing the validity of the expansion
.\" :2 ~ (- I)n . !"I7T..\'
- = - - ¿:; - - Sin -- (Ü -< x < ').
/ 7T. no \ ti /
I t is clear tha! the second form (179b) is better adapted to numerical calculation. A
third form of the solurion is obtained in Problem 61 by a generalization of lhe method
of Section 9.10. •
The formulas (173) and (174) conti nue to apply when the homogeneous
conditions T(.\", O) = O and T(O,I) = O are replaced by more general homo-
geneous conditions a \ T(x, O) + a, T,(x, O) = O and /3 \ T(O, 1) + /3 2 T,(O, 1) = O,
where the a's and /3's are constants, with the appropriate modificarion of
A(.\", 1). Here (l75) obvioLlsly will continue to apply only if a2 = O.
However, the same technigues do 1701 necessarily apply when the heat-ftow
equarion (165) is replaced by anorher one. For example, if Laplace's eguation is
subsrituted for the heat-ftow eguation in (165), with 1 replaced by y, to yield the
problem
T(O, y) = O, T(l, y) F(y), ( 180)

T(.\", O) = O
in lhe strip (O <:: x <:: /, O <:: )" < =), wirh appropriate resrrictions on the
behavior of T as.1· ~ =, and if T = A(.\", y) denotes the solution when F(y) =
1, then the solution when F(y) = O if O <:: y < t] \ and F(y) = I if Y > t] \ is no/
given by T = O if O <:: y < t]\ and T = A(x, J' - t]\) if y > t]\. Physical (or
9.12. Traveling Waves 467
mathematica l¡ consideration; mak e this facr clear when it is noted that here
T( x, y) is the s teady-swte temperature in a long plate. with a portion of one edge
permanentl y maintained at unit temperature and the remainder of the boundary
held at zero temperature, and that the temperature inside the plate then certainly
will be positive everywhere.
An alternative method of dealing with the problem (180) is presented in
Section 9.14, and another technique which also can be used to cope with non-
homogeneous conditions in more general situations is described in Section 9.17.
The formulas (173) and (175) are rederived, by methods 10 be treated in Section s
9.15 and 9.16, in Problem 101 for the heat-flow problem ([65) to (167) and In
Problem 102 for a similar problem governed by the wave equation.
9.12. Tra~'eling Waves. lt was found in Section 8.5 that the general solu-
tion of the one-dimensional wave equation
a2 rp . l a2 rp
ax2 = C2 a¡;: (181)
is of the form
rp = f(x - el) + g(x + el), ( 182)
where f and g are arbitrary functions. lf we consider the graphical representa-
tion of rp as a function of x for varying values of the time l, we see that a solution
of the form rp = f(x - el) is represented at the time l = O by rp = f(x), and
is represented at any following time t by the same curve moved parallel to
itself a distance el in the positive x direction . That is , a so lution rp = f(x - el)
is represented by a curve moving in the positive x direelion wilh velocilye. Similar-
Iy, a solution rp = g(x + el) is represented by a curve moving in lhe negalive x
direelion with velocilye. We may speak of these solutions, in a general sense, as
traveling waves.
D'Alembert'sformula (67), of Section 8.5, in the form
l 1 fX+e<
rp(x, 1) = 2[f(x + el) T f(x - el)] + 2e x-e< g(c;) dI!;, (183)
defines that solution of (181) which satisfies the so-called Cauehy conditions
rp(x, O) = f(x), (184)
for aH real values of x, where f and g are any twice-differentiable functions.
Here, for example, we may interpret rp and rp, as the lateral deflection and
velocity, respectively, of a tightly stretched uniform string of infinite length
with e = --/T/ p, where T is tension and p is linear mass density, and where
small slopes and deflections are assumed. (The case of a string of finite length is
treated in Problem 33 ofChapter 8 and in Problem 72 at the end ofthis chapter.)
In many other applications we are principally interested in solutions of the
wave equation which are periodie in time. Such solutions, for (181), must be
combinations of term; of the formf¡(x) cos wt or fl(X) sin wt or, equivalently,
468 Solutions of Partial Differentia¡ Equations of Mathematical Physics
must be the real or imaginary part of a function of the form

rp = F(x) e iw ', (185)
where F(x) may be complex. By introducing (185) into (\81) and canceling the
resultant common factor e1w ' , there foUows
F" + w>=
e- O. (186)
If we write the general sollltion of (186) in the complex form

(187)
the reqllired solution (185) beco mes
(188)
This solution is of the form of (182). By taking real and imaginary parts, we
see that linear combinations of terms of the form
cos -w (x - el), sin w (x - el) ( 189)

e e
for arbitrary values of w are available for the representation of periodic plane
waves moving in the positive x direction, whereas combinations of terms of the
form
w
cos -(x + el), sin w (x + el) (190)
e e
are solutions of (181) representing periodic plane waves moving in the negalive
x direction. In certain problems it may be more convenient to retain sllch
solutions in the complex exponential form of (188).
We next investigate the existence of analogous spherieal wave sollltions of
the wave equation. In spherical coordinates, with the solution rp dependent
only on the radius r from the origin, the wave equation beco mes
1 a ( , alp)1 a'rp (191)
r' dr r ar e' al' =
and the general solution can be obtained in the form
rp = _1 [f(r -
r
el) + g(r + el)]. (192)
(See Problem 70.) If we assume a solution periodic in time, of the form

rp = F(r)e 'w ', (193)
substitution of (193) into (191) gives the equation
,
rF" + 2F' + we 2
rF = O (194)
to be satisfied by the amplitude function F. The general solution of (194) can

be expressed in terms of Bessel functions by identifying (194) with Equation
(127), Section 4.10. The solution is then given by Equations (128) and (129) of
9.12. Traveling Waves 469
tha¡ section in the form
(195)
This result is in turn expressible in terms of elementary functions, by virtue of

Equation (115), Section 4.9, in tbe form
F = / 2c [A sin (ror/c) + B cos (ror/c)]
1/ 1I:ro r r
or, finally, in the complex form
e 1wr / c e- iwr / c
F= c¡--+c , . (196)
r - r
Thus Equation (193) becomes
e¡"W ( T+r:t) ! C e - iw ( r- c r)!c
rp = C¡
r
+ c2 - - - -
r
(197)
Linear combinations of such functions, or of their real and imaginary

parts, are thus solutions of the wave equation which represent periodic spherical
waves moving inward toward or outward from the origin, respectively. It may
be noticed that the amplitude of the oscillation is inversely proportional to r.
In cylindrical coordinates, with rp dependent only on the distance r from
the z axis, the wave equation beco mes
Tbe assumption
+:r (r ffr) = c; ~:r· (198)
rp = F(r)e i o" (199)

then leads to the equation
FU + _1r F' + ro,c-2 F = O. (200)
Although the general solution can be written in the form
F = AJo (ro;) + BYo (ro;) ,

we are guided by the preceding developments in writing this solution instead
in the complex form
(201)
where, as defined in Section 4 .8, the Hankel functions are given by
Hh¡ ) (o;r) = Jo (o;r) + ¡Yo (o;r) ,

(202)
Hh 21
(o;r) = Jo (o;r) - ¡Yo (o;r).
470 Solutions of Partial Differential Equations of iYlathematical Physics
The solution (199) then beco mes
(203)
The usefulness of the definition of the Hankel functions is now seen if we

consider the asymptotic expressions for H~' ) and H~Z) as given by Equation
(106), Section 4.9. By making use of these results, the behavior ofthe expression
(203) for large values of r is found to be given by the asymptotic expression
rp _
J -2c [ (c)e- ,n.')
. e'wt, .. "u + (c 2e,n. ')--~-
'J
.. e-;w(,-" ).
(r ~ =). (204)
nw ,...j r ,...j r
Hence we see that jor large values oj r the real and imaginary parts of
é""Hótl(wr/c) represent cylindrical waves moving inward toward the z axis,
whereas the real and imaginary parts of e'W' H~2)(wr/c) represent outward-travel-
ing waves. The amplitude of the oscillation here is inversely proportionaI to --/-;:.
More generally, reference to the asymptotic expressions for HV 1 and H~21
shows that the real and imagi nary parts of e lw , H ~1)(kr) represent inward tra veling
cylindrical waves for large values of r (for any nonnegative value of p), whereas
the real and imaginary parts of e'w' H~2)(kr) represent outward traveling waves.
Although in most problems these complex forms are preferable, the explicit
real and imaginary parts may be listed as follows:
e'w'H ~t) (ú:n = [J p (~r) cos wl - Y (~r) sin WIJ

p
+ {Jp(~r) sin wt + Yp(~r) coswt} (205a)
e'w'H~2) (~r) = [Jp(W;) cos wt + Yp(~) sin cot]

+ {Jp(W;) sin Wt - Yp(o:r) cos WtJ (205b)
The preceding special results are useful principally in dealing with prob-
lems, governed by tbe appropriate form of the wave equation, in which the
supplementary conditions are such that the solution may be expected to have a
time factor of the form e''''', where w is a known real constant, or to be expres-
sible as a linear combination of terms having such time factors. Their use in
such cases is particularly effective when a boundary condition (perhaps "at
infinity") involves the specification of inward- or outward-traveling waves, as
is illustrated in the following section.
9.13. The Pulsating Cylinder. As a simple example of the usefulness of the

results obtained in Section 9.12, we consider a pro blem which is related to
certain aerodynamic investigations of nonuniform motion. The lateral boundary
of an infinite circular cylinder surrounded by ideal compressible fluid is caused
to pulsate radially with circular frequency w in such a way that the radial velocity
9.13. The Pulsating Cylinder 471
of the boundary varies periodically about a zero

value, with maximum value VD (Figure 9.8) The
periodically varying velocities imparted to the
points in the surrounding fluid are to be deter-
mined.
According to hydrodynamic theory [se e Equa-
tions (204) and (205), Section 6.20], the velocity
vector V is here the gradient of a velocity poten-
tial rp which satisfies the wave equarion
(206)
Figure 9.8
where e is the velocity of sound in the fluid. lf
we introduce cylindrical coordinates, noticing thar rp IS independent of - and
e,Equation (206) becomes
a"rp + _1 arp a rp
I 2
ar 2r ar = cr at 2 .
(207)
The magnitude V of the (radial) velocity is given by the equation

V = arp. (208)
ar
On the boundary the fluid velocity must equal the velocity of the pulsating
boundary, say VD cos wt,
whereas for large values of r the velociry must corre-
spond to outward-traveling cylindrical waves (in the absence of a reflecting
outer boundary).t
As will be seen, it is convenient to consider rp as the real part of a complex
function 'Po' and to seek to determine 'Po' since then we are lO require that Jrp,¡Jr
reduce to VDe iM when r = a. Thus rp,
is to satisfy the wave equation and the
boundary condition
r = a: (209)
Accordingly, we are led to write
(210)
and Equation (203) of Section 9.12 shows that a suitable expreSSlOn for rp,
rhen is of the. form
(211)
If we make use of the derivative formula (114), Section 4.9, there then follows
arpo _- --
-a w[ ele
,w'H(!)
1
(wr) ,
- I cze iMH(2)
1
(wr)]
- . (212)
r e e e
tThe requirement that ¡he solution of a problem involve only oUlward-traveling waves is some-
times called the Sornmerfeld radiation candilioN.
472 Solutions of Partial Differential Equations of lVlathematical Physics
To ensure ou/ward-traveling waves for large values of r, \Ve must take

c, = O. (213)
The remaining constant c 2 is determined by (209) in the form
Voe iw' __ - -w
C
c •.. e iw'H{21
1 (wa)
-,
C
c Vo
- W H\2}(wa/c) (214)
With (213) and (21.:1), the required velocity thus is quickly determined in the
remarkably compact form
V = Re{arp,}
ar
= Re{ V e o
iM H\21(wr/c)} ,
H\21(wa/c)
(215)
and only elementary algebraic computations remain. For this purpose, we write
(215) in the expanded form
V = V o R ecos
{( wt + 1.'SIn Wl ) JJ,(wr/c)
, ( wa,/)
-
c _
iy , (Wr/c)}
.y (
1 ¡ wa c
/)
and the usuaL process of rationalization and separation of real and imaginary
parts leads to the explicit form
V = Jt(wa/c) :0 Yf(wa/c) {J, (~a)[J, (~r) cos wt + Y, (W;) sin wt]
(216)
The amplitude of the velocity oscillation at a given distance r from the axis
is given by
A(r) = V [Jf(wr/c) + Yf(úJr/C)]' 2. (217)
o Jf(wa/c) + Yf(wa/c)
With this notation, Equation (216) can "be written in the form
V = A(r) cos [wt - a,(r)], (218)
where the phase shift a, at distance r from the axis is given by
a,(r) = tan-¡ [J,(wa/c)y,(wr/c) - y,(wa/c)J¡(wr/c)]. (219)
J,(wa,'c)J,(wrlc) + Y¡(wa/c)Y¡(wr/c)
For values of r and a large with respect to c/w there follows, in particular,
A(r) - V O ,\,
la
7'
W
a,(r) - -(r - a)
c '
if use is made of Equation (105), Section 4.9. Thus, as r ---+ 00, (218) beco mes
jf aw» c.
9.1-1. Examples of Ihe Use of Fourier lnlegrals 473
We may ñotice that if the bOllndary condition had been left in lhe form
arp(a,O¡ar = V o cos wt, the assllmed solution rp(r, t) necessarily would have
involved bolh cos mI and sin mI, and also the imposition of the "outward-wa ve"
condition as r -, = would have been less simply effected.
9.14. Examples oC the Use oC Fourier Integrals. When the formulation of a

problem involves a nonhomogeneous condition specified along a boundary
consistiog of a line segment of inn.nite length, and use is made of the method of
separation of variables, the superposition of particular solutions geoerally is
accomplished by integration rather than summation.
As a n.rst illustration, we require the soJution of the two-dimensional
Laplace equation
(220)
valid in the half-plane y > O, taking on prescribed values along the x axis for
all val ues of x,
rp(x, O) = f(x), (221 )
and vanishing at large distance from the origin,
(222)
in that half-plane.
Physically, we may interpret this problem (for example) as determining the
steady-state temperature distribution in a thin plane sheet of infinite extent on
one side of a straight edge , when the temperature is prescribed along that edge.
Product solutions of (220) which vanish as y ~ 00 are readily obtained in
the form
rp p = e-UY(A cos ux + B sin ux) (u > O), (223)
where L/o has been written for the arbitrary separation constaot and A and B
are arbitrary. If we think of A and B a s functions of u,
A = A(u), B = B(u), (224)
we may supenmpose solutions of form (223) for al! positive values of L/ by
integration , and so write
rp(x, y) = L- e" U'[A(u) cos l/X + B(u) sin ux] duo (225)
This express ion formally satisfies (220) and (222) for arbitrary forms of the
fuoctions A and B. The remaining condition (221) wil! then be satisfied if we
determine A and B such that
f(x) = f~ [A(u) cos ux + B(u) sin ux] du (-00 < x < 00 ). (226)
But, in accordance with Equations (235) and (236) of Section 5.15, the right-
hand member of (226) reduces to the Fourier integral representation of f(x)
if we lake
1 ,- ":<>
A(u) = nt =f(e;) cos u.; de;, (227a)
B(u) = n1 J__ f(e;) sin ue; de:,.

'=
(227b)
With lhese results, Equation (225) takes the f orm
rp(x, y) = ~ r e - Uy{[f= fe!;) cos ue; d!; ] cos ux
+ [f= fe!;) sin u!; d!; ] si n ux} du (228)
r
or, equivalently,
rp(x, y) = ~
[r~ e-UYf(!;) cos u(e; - x) d!; ] duo (229)
We remark that, when the integral f==I f(x) I dx does not exi st, the Fourier
integral representation (226) usually is not valido Still, in certain cases when
(226) is not val id, if the u integration is carried out befo re the e; integration,
Equation (229) may nevertheless lead to a specific expression for rp, which,
however, may not satisfy the condition (222). The validity of theformal solution
so obtained should be checked by direct calculation in such cases. Physical
considerations will generally indicate whether violation of the condition (222)
is permissible in exceptional cases.
\Vhen the order of integration is rever sed, the solution (229) takes the form
q.¡(x , y) = ~ f~ [[ e- UY cos u('; - x) duJ fW d';
or, after an evaluation of the inner integral,
rp(x, y) =
f-
n1 __ y2 yf(e;)
+ (e; _d!; X)2' (230)
It can be proved that this integral converges and defines a solution of Laplace's
equation for y >
real x.
°
when f(x) is bounded and piecewise continuous for aH
In the special case when

(x < O),
f(x) = {
Ol (231)
(x > O),
there follows, from (230),
q.¡(x, y) = n1 f.=o yZ + ~ _ X)2 d !;
= ~ (~ + tan - J ;)
or rp(x , y) = e
1 --, (232)
7l
9.14. Examples Di the ese of Fourier Integrals 475
e
where is lhe polar angle. In this case f ~oo I fex) I dx does nol exist and, funher,
lhe righl-hand members of (227) do nol exist! However, (232) is easily shown to
satisfy (220) and to reduce lo (231) when y = O. Since (222) is violaled by the
prescribed value of rp on the x axis (e = O), it is not surprising that it is also
violated by the solution (232) for all values of e except e = n. Here (232) is the
only solulion of (220) which reduces to (231) on the x axis and is bounded as
y~ =.t
Other one-dimensional problems of similar type are solved in an analogous
way. As a second example, we consider the problem of one-dimensional heat
fiow in a rod of infinite length with insulated sides (see Section 9.10), and so
deal with lhe differential equation
aZT 1 aT (233)
""dX'f: = IX 2 at .
We assume that an initial distribution oftemperature is prescribed along the rod,
T(x, O) = f(x), (234)
and that at the instant t = O the temperature al the end x = O 15 changed to
zero and maintained at zero thereafter,
T(O, t) = O (t> O). (235)
The temperature at a point x at time t is required.
Bounded product solutions of (233) satisfying (235) are of the form
where u is arbitrary. Considering A as a function of !I, we may supenmpose

such solutions for all positive values of u by writing
T(x, t) = [ A(u)e- u '>" sin ux du (x> O). (236)
The condition (234) then requires that A(u) be determined so that
T(x, O) = f(x) = [ A(u) sin ux du (x> O). (237)
But, according to Equation (232) of Section 5.14, we have the Fourier sine-
integral representation
f(x) = ; [[[ fe!;,) sin u!;' d!;, ] sin ux du (x> O),
t As this example illustrates, the result of a formal manipulation may be corree! even in cases
when the steps in that manipulation cannot bejustified (or are clearly incorrecl). If the validity
of thefina! resu!! can be verified, there is no need to seek an alternative derivation (except per-
haps on esthetic grounds). Accordiogly, io a case such as the present one, if the solution were
required only when f(x) is defined by (231), it would be appropriate lo replace Ihat specified
function by a literal one, to solve the resultant problem formally, and finally to replace the
literal functionf(x) by the specific function in question only after certain questionable steps
have been taken, provided !hal Ihe validity of the fina! resu!1 is verified direcl!y.
if f(x) is sufficiently well behaved, and hence the right-hand member of (237)
then reduces to f (x) if we take
A(u) = -2 f ~ fe!;,) sin u!;' d!;,. (2 38)

7t o
Thus the solution (236) becomes
T(x, t) = -
2 f~ . ~
7t o
Jo e- ' >' ,/«(,) sin u!;, sin ux d!;, duo
u
(239)
This form can be simplified if we carry out the u integration first. Thus we
obtain
T(x, t) = ; 1~ fe?;) (1~ e- U
'- " sin ti!;, sin ux dU) d!;,. (240)
The inner integral is first simpJified by writing

sin u!;, sin ux = t[cos u(!;, - x) -- cos u(!;, + x)].
Then if use is made of the known resultt
fo
~ ,.
e-a x - 2a7i e- b ' .. " a •-
cos bx d X = ..y (a > O), (241)
Equation (240) is reduced to the form
T(x, t) =
2a ~7tt
I f=
o
f(!;,)[e- I': - X )',,4> " - e- 1U x )':4."] d!;" (242)
which can be shown to be a valid solution for any bounded and piecewise con-
tinuous J, in spite of the fact that (236) , with (238), is valid only under rouch
stronger restrictions on f
When the initial temperature is uniform,
T(x, O) = f(x) = T o, (243)
Equation (242) leads, after obvious substitutions, to the solution /
T(x, t) = T¡"-
"'" n
(f'" _ e-u' du - f~ _ e-u' du )
f
-x.' 2<x ,/ r :< : 2 « .,I (
To X."2.',. 2To f X/ l" '~ •

= I e -Ll ~ d u =;:; e -u ~ d u.
v n_J{. 2"'''''"1 .y 7t o
This result takes the forro
T(x, t) = T o erf
2a
o:';t (244)
in terms of the so-called error function
erf x = ~
2 f oXe - u' du, (245)
tSee Equation (101) of Chapter 7.

9.15. Laplace Transform Methods 477
which is of frequent occurrence in the stud y of heat flow and in other field s , and
which is a tabulated function . The multiplícative constant in (245) is so chosen
that
erf ( 00 ) = l. (246)
If we integrate both sides of (241) with respect to the parameter b, from
b = O to b = p, we obtain for future reference the relation
r- e-a'x' sinxpx dx _{-ajo
Jo
ii r =
p
e- b' /4a' db
= ~ erf fa, (247)
which is frequently u seful in dealing with problems which involve the error
function. It should be noted that the result of omitting tbe factor 2/,.Jn in (245)
is often denoted by Erf x , and tbat otber usages (sorne of whicb are inconsistent
with tbe present one) sometimes appear in the literature.
The preceding problems can be solved alternatively by use of Fourier (or
Fourier sine) transforms (see Problems 79 to 81), whicb bave the additional
useful property tbat they are also applicable to either problem when a "forcing
term" h is inserted in the right-hand member of tbe relevant differential equa-
tion, and to the second problem when the homogeneous boundary condition
T(O, t) = O is replaced by a nonhomogeneous one, such as T(O, t) = F(t).
9.15. Laplace Transform Methods. When a problem governed by a partial

differential equation with constant coefficients is specified for all positive values
of the time 1, with the initial condition or conditions prescribed when 1 = O and
with no conditions prescribed for any other fixed value of 1, the use of the
Laplace transform may be conven.ient. (Clearly, ( may be replaced by a different
variable, which may or may not represent time .)
Here we may note, for example, that the Laplace transform j(x, s) of a
function f(x, l) is defined as
j(x, s) = [ rUf(x, t) dt , (248)
where x is held fixed in the integration, in accordance witb the definition in

Chapter 2. We see that the transform of the x deriva ti ve off is the x derivative
of tbe transform j,
o
.,u
{af(x,
ax
t)} _- J=e -s< af(x, l) d _ aj(x, s)
aX l- aX ' (249)
o
and that the same statement applies to x derivatives of bigher order. However,
as regards 1 differentiation, the results of Chapter 2 can be applied to give
oC {ar~, l)} = sj(x, s) - f(x, O), (250a)
o {a2f(x, ()} = 2r( ) _ ¡f( O) _ af(x, O) (250b)

.,u at' s J x, S S x, al'
478 Solutions oC Partial Differential Equations oC Mathematical Physics
and so fortb [see formulas (T3), (T4), and (T5) , page 67]. In a similar way, all
the results of Chapter 2 are again applicable here, ¡nsofar as time variation is
concerned.
In applications to the solution of problems governed by partial differential
equations, it is frequently necessary to make use of the fact that, if ¡(x, t) is
bounded as a function of x for all relevant values of x when t has a fixed positive
value, then the transform ¡(x, s) has thesame property when s is fixed. In addi-
tion, the transform must tend to zero as s - +00, so that
,
liro ¡(x, s) =
+
° (251)
for aH such values of x.
As a simple example of the usefulness of the Laplace transform, we consider
the solution of the wave equation
a2 qJ 1 a2 qJ
ax2 = CT at2 (252)
°
in the first quadrant (O -< x < 00 , -< t < 00) of the distance-time plane
(Figure 9.9), subject to the boundary condition
and to the initial conditions

qJ(O, t) = ° (253)
qJ(X , O) = 0, (254a)
aqJ(x, O) = 1 (254b)
at .
If we take the transform (with respect to t) of (252) and u se (254a, b), we obtain
(255)
to be satisfied by tp(x, s). In addition, from (253) it follows that the condition
tp(O, s) = ° (256)
must be satisfied.
.p =0
.p =t
.p = o, .p 1 = 1 x
Figure 9.9
9.15. Laplaee T.ransform I\Iethods 479
Whereas (255) is a partia! differentia! equation, the fact that only x differ-
entiation is invo!ved permits it to be so!ved as though it were an ordinary differ-
entia! equation, with s he!d constant in the process, and with the arbitrary
"constants of integration" accordingly replaced by arbitrary functions of s.
This is the principal advantage gained by use of the Laplace transform technique.
The general solution of (255) is thu5 obtained in the form
ifj(x, s) = A(s)e-'X , + B(s)e'" , + J,.

s-
(257)
The condition (256) and the requirement that the transform be bounded as
x -, -\- 00 for s > °
then givet
A(s) = B(s) = 0, (258)
50 that the transform of the required solution is
cp-( x, s ) =
¡
---::;- -
1
-., e
_,"'c
'. (259)
s- S"
The inverse transform of the right-hand member is identifiable by direct use

of (T35) and (T9), pages 67 and 68, and we thus obtain the result
(t < ;),
rp(x,t) = t _ Jo
1t - ~ (t:> ~),
or, equivalently,
rp(x, t) jtx
= (260)
e
TtUs examp!e was chosen not only to illustrate the efficiency of the use of
the Laplace transform io a favorable situation, but also to provide an example of
an occurrence mentioned at the end of Section 8.9. Specifically, if we examine the
"solution" obtained here, we notice that, whereas rp is continuous, the two
first partial derivatives are discontinuous at all points on the line x = et in the
first quadrant of the xt planeo Hence, strictly speaking, the seeond par ti al deriva-
tives do not exist on that line (even as one-sided derivatives), and the differential
equation accordingly is not satisfied along it. The prescribed boundary and
initial conditions are indeed satisfied.
The fault is not with the Laplace transform procedure (which would have
yielded a proper solution had one existed) but with the formulation of the
tHere it is implied that rp(x, /) is required to be bounded as x - + 00 for any fixed positive
value of /. In faet, this requirement here is irrelevant (see Problem 35 of Chapter 8), and the
requisite vanishing of B(s) can be shown to follow from the fact that B(s)e a /, cannO! be a
Laplace transforrn as x ~ + 00 unJess it vanishes identicalIy.
480 Solutions oC Partial Differential Equations of l'vlathernatical PhYSics
problem. It can be shown that the only continuous function which satisfies the
specified side eondition s and which satisfies the differential equation everywhere
except along the line x = el is the one obtained. Whether it is in fact acceptable
as the required description of a certain entity (physical or otherwise) depends
upon the nature of that entity . If it is not acceptable, then it must be concluded
that the mathematical simulation [(252) to (254)] of the phenomenon in question
is at fault (and probably is o versimplified, by linearization or otherwise).
It should be noticed that the use of Laplace transforms permits the solution
of certaio problems governed by nonhomogeneous partial differential equations.
In iJlu s tration, we consider the determination of rp(x, t) satisfying the equation
a'rp _ 1 a 2 rp
ax2 - e2 al2 - cos OJt (261)
and the homogeneous side conditions

'1'(0, t) = O, (262a)
rp(x, O) = O, (262b)
arp(x, O) = O (262c)
al '
when O -< x < = and O -< l < =. The assoeiated problem governing the
transform ip(x, s) is then
a'ip s' - s (263a)
ax' - CZrp = - S2 + OJ2
ip(O, s) = O, (263b)
from which there follows easily
(264)
and hence
e2
OJ20 -
.
COS OJl)
(X)
l -< e '
rp(x, l) = (265)
1~2{ eos OJ(t - ;) - cos OJtJ

Here '1' , 'Px' and 'P, are continuous wheo x = el .
In less contrived situations , the determination of the in verse transform may
present a greater challenge. Methods to be derived in Sections 11.2 and 11.3 can
be used (to supplement the use of tables) in many cases . The example treated in
the following section exploits an expansion process which is of rather frequent
usefulness.
9.16. Application of the Laplace Transform to the Telegraph Equations for a

Long Line. In dealing with the flow of eleetricity in a long insulated cable, the
potential v (volts) and the current i (amperes) are found to be related approxi-
9.16. Application of the Laplace Transform to the Telegraph Equations 481
mately by the simultaneous equatioos

av = _La¡, (266a)
ax at
Ei = _c av , (266b)
ax at
where x is distaoce (miles) from one end o f the cable, if the effects of leakage and
re sis taoce are neglected. Here L is inductance per unit leogth of cable (hendes/
mile), and C is the capacitance to ground per unit length (farads/ mile). Both L
aud C are assumed to be constants. If i is eliminated between (266a) and (266b),
we fiud that v satisfies the wave equation
azv a2 v
-a
x-' = ----;-
LC a 1- ,
(267)
and a completely analogous equation is satisfied by i.

Initially, the line is bere considered to be dead,
v(x, O) = ave;; O) = o. (268)
Theo, beginning at tbe instant t = O, a voltage is impressed at the end x = O,

and thereafter that end is maintained at a prescribed poteotial which may vary
with time,
veO, t) = f(t) (t > O). (269)
If the other end (x = 1) is open, then i = O when x = l. With the use of (266a),
this condition beco mes
av(/, 1) = O (t> O), (270a)
ax
whereas if the end x = 1 is grounded, tben v = O when x = 1,
v(l,/) = O (t> O). (270b)
If we take the Laplace transform of Equation (267), with respect to 1, there
follows
2
a v(x , s) = LC[ 2.-( ) _ v( O) _ av(x'O)J (271)
ax2 s v x, S S x, al
or, taking ioto account tbe ioitial conditions (268),

a v(x,
2
ax2 s) - LC s 2 v-( x, S) -- O. (272)
Jt will be convenient to write

c= 1 ,
(273)
..,jLC
so that (272) becomes
a ij
2
ax2 - cr V=
S2_
O, (274)
with the geñeral solution

t'(x, S) = A(s)e- >X.'c + B(s)e'" c. (275)
We consider here only the case where the [ine is open at the end x = /, so
that (270a) and (269) are to be satisfied . If we take the transforms of these two
equations we find that v(x, s) then must sa tisfy the end conditions
ü(O, s) = ](s), (276a)
Jv(/, s) =
Jx
° '
(276b)
whe re les) is the transform of f(t). Thus the functions A(s) and B(s) in (275)
are determined by the conditions
A(s) + B(s) = les), (277a)
-A(s)e -S{:C + B(s)e'{' C = 0, (277D)
and hence are of the form
A(s) = 1 +les)
e-2s{ c' (278a)
B(s) = 1 +fes)
e-t-2'{,C (278b)
In the limiting case of an infinite fine,

1= O() , (279)
tbere follows
A(s) = les), B(s) = 0, (280)
and (275) beco mes
v(x, s) ~ j(s)e- H /
C
• (281)
This is the tran sfo rm of the desired solution, and the use of (T9), page 67,
determines v(x, [) in the form
f(t - ;) when t > ~ or X < ct,

v(x, t) =
(
° when t < ~ or
c
X > ct.
(282)
c
Thu s we see that at a given time t the effect of introducing tbe vo ltage at x =
is presen t only at di stance s not greater than ct = t/../ LC miles froro that end.
°
That is, a voltage wave is propagated along the line with velocity c = 1/../ LC,
in a manner specified by Equation (282). For example, if veO, t) = f(t) varies
periodically with time as is indicated in Figure 9.IO(a), the voltage along the
hne at time t is as indicated in Figure 9.1O(b). In general, we may verify that the
representation of v(x, t) as a function of x at time t is obtained by reflecting
tbe curve for veO, t) about the axis t = 0, replacing the t scale by an x/c scale,
and then translating tbe refiected curve to the rigbt through t units.
In the more general case of afinite fine, the introduction of (278) ioto (275)
gives the result
9,16 ..-\pplication of the Laplace Transform to the Telegraph Equations 483
v( D,t)
v(x, t)
x
e
(a) (b)
Figure 9.10
-( ) = f-( ) cosh [s(l - x)jcl. (283)

vx,S S cosh(sljc)
To determine the solution having this transform, we first rewrite (283) in the form
_ - e-sx,'c + e- s (21-';I()/c
v(x, s) = fes) 1 + e o. . s,c
¡ ,
then expand lj(l+e- 2 ,¡,,) in a series ofascending poweC5 of e- 2 '¡,", and so obtain
'v(x, s) =](s)[e-·L'I:,C ¡ - e-S(:t-X)¡'c _ e- s (2i+:ci,"c _ e- s (4/-xJ:c + e- s (41+x),'c + ... ].
(284)
Again making use of formula (T9), page 67, we obtain v(x, t) in the form
v(x, t) =
when t > ~ 1
when t < ~
e
+1:('-2T+ ~) when t > 2T - ~ 1

when t < 2T - ~
e
_1:(' - 2T - ~) when t > 2T +~ 1

when t < 2T+ ~
e
1:(' 4T+ ;)
when t > 4T - ~ 1
when t < 4T - ~
e
+ 1:(' .. 4T ;) when t > 4T + ~ + 1 (285)

when t < 4T + ~e
wbere we have written

T = -cI = -
J LCI, (286)
so that T is the time required for a wave to travel the length of tbe line.
Noticing that O <: x / c <: T, we see that in the first part of the propagation
(O <: t <: T) only the first brace in (285) may differ from zero, and hence in
this part of the process the wave decribed aboye is traveling toward the end
x = 1, reacbing it at the time t = T. Then when T <: t <: 2T, only the first two
braces in (285) are not zero. In this second interval the second brace represents
a reflected wa ve returning from the open end of the line without change in
signo The actual voltage in thi s time interval thu s consists of the superposition
of an outward- and an inward-traveling wave. When t = 2T, the reflected
wave reaches the closed end (x = O) and , according to the third brace in (285),
is again reflected, but this time with reversal of signo This process is continued
indefinitely, reflections without sign cbange occurring at the open end and
reflections with change of sign occurring at the closed end. The various inward
and outward waves so generated may combine in various ways, for various
periodic impressed end voltages, the nature of the superposition depending,
in particular, upon the relationship between this period and the time interval T.
It may be noted that, with an abbreviation such as
(t > O),
(287)
(t < O),
Equation (285) can be rewritten ID the more compact form
v(x, t) = r ( t - ;) + r (t - 2T + :) - r(t - 2T - n
- r(t - 4T + :) + .... (285 ')
9.17. Nonhomogeneous ConditioDS. The Method of VariatioD or Parameters.

In each of the problems considered in this chapter, a certain linear partial differ-
ential equation is to be satisfied in a certain regio n ffi and the solution is to
satisfy a boundary condiLion along each boundary which serves to lioút ffi. If
time also is involved, these boundary conditions may be time-dependent and the
problem may also specify one or more initial conditions to be satisfied by the
unknown function rp throughout ffi at an initial time.
The method of separaLion of variables requires for its success tbat ffi be s uch
that the relevant differential equation is separable (or beco mes separable when a
nonhomogeneous "forcing term" is suppressed) in terms of a coordinate system
for which one of the coordinates is constant along each of the boundaries whicb
make up the border of ffi. In addition, however, when the differential equation
is homogeneous and time is not involved, nonhomogeneous boundary condi-
tions can be imposed only along one pair of boundaries (or, in Iimiting cases,
9.17. Noohomogeneous Conditions . Tbe Method oC Variation oC Parameters 485
along the single boundary) o n wbich one of the variables is constant. Wben tbe
equation eitber is nonhomogeneoust or is time-dependent, al! tbe boundary
conditions must be homogeneous; combined nonbomogeneity and time depen-
dence of the differential equation generally are not acceptable. In the favorable
situations noted, the presence of pairs of homogeneous boundary conditions
leads to a set of characteristic-value problems, for which the permissible values of
the separation constant or constants are tbe characteristic numbers. (In some
cases, requirements of boundedness or periodicity, which also are homogeneous
requirements, replace one or both of a pair of homogeneous boundary condi-
tions .) Tbe representation assumed for the required solution rp tben comprises
a superposition (by summation and / or integration) of product terms containing
tbese cbaracteristic functions.
Special devices also permit the treatment of certain other types of problems
by such methods. In particular, an excess of nonhomogeneity often can be
conquered by the process of superimposing the solutions of two or more related
problems, for eacb of wbich no such excess occurs. Tbis process, however,
may be tedious and may provide tbe final solution in a cumbersome form o
Alternatively, it may be possible to determine (by inspection or otherwise) a
function u which satisfies the excess nonbomogeneous requirement (or require-
ments) as well as the homogeneous ones. The difference rp - u then is to satisfy
an acceptable set of conditions. (As was noted in Section 9.11, this procedure is
used, for example, in Section 9.10, as well as in Problems 22,23, 60, and 61.)
However, except in relative1y simple cases, this process of removing one sort of
nonbomogeneity, without introducing another, usually is not feasible .
Other methods, wbich also are available in tbese and in certain less tractable
situations, inelude the Duhamel integral superposition technique which is
illustrated in Section 9.11 (and which can cope with time-dependent boundary
conditions in certain cases), the use of the Laplace transform (which also serves
this purpose, and further permits the differential equation to be nonhomogene-
ous), and the use of other transforms such as tho se of Fourier, Hankel, and
Mellin (which can deal with combined nonhomogeneities in appropriate space
regions).
In the remainder of this section we describe and illustrate a rather powerful
technique, somewhat analogous to that treated in Section 1.9 (in connection
with ordinary differential equations), which incorporates many of the features
of the methods just mentioned, and which i s capable of dealing with a variety of
reasonably troublesome problems analytically, in a compact and uniform way.
As a first illustration of the method , we seek the solution of the nonho-
mogeneous heat-flow equation
IX
, a'T aT =
ax' - TI h(x, 1), (288)
tSee Problems 45 aod 46.

486 Solutions of Partial Dilferential Equations of IVIathematical Physics
subject to the homogeneous boundary conditions

T(O, 1) = O, T(/, t) = O (289)
when t> O, and to the initial condition
T(x, O) = f(x). (290)
In the absence of the " forcing function" h(x, t), the method of separation of
variables would lead to an assumed express ion for T, satisfying (288) with h = O
as well as (289), in the form
-= -" ~ 71 ~ !1. ~1 .'1 2· n7tX
T(x, t) = .~ a.e . Slll - , - ' (291)
where the a's are constants to be determined in such a way that (290) also is
satisfied.
The method of variation of parameters here consists of replacing the con-
stants a. by unknown functions e.(t), and attempting to determine these func-
tions in such a way that both (288) and (290) are satisfied. Since the exponential
function of t can be absorbed into en(t), we thus are led to the assumption
T(x, t) = :f: en(t) sin nn, x. (292)

. " 1
Since T(x, t) satisfies the end conditions (289), the consequence of the
requirement that it also satisfy (288) can be obtained by inserting (292) into (288)
and differentiating term by term (see Section 5.12). Whereas this procedure is
to be recommended for the purpose of minimizing computation, when it is
permissible, there are other situations (as will be seen) in which its use is improper
and leads to false conclusions. Hence we here illustrate an alternative process of
more general applicability.
For the purpose of determining e.(t), we next deduce from (292) that
e.(/) =
2 r l
-, Jo T(x, t)
.
Sin - , -
nnx
dx, (293)
notlclng that T(x, t) must be a continuoLls'y differentiable function of x for

O -< x -< , when t > O, since T satisfies (288). Accordingly, there follows
e '( t ) = - 2
. 'oJI aT
at
nnx
- - Slll - -.
,
dx
and hence, by using (288) to replace T, by a. 2 T xx - h, we obtain the relation
Jor a'T Jor' h( x, t )Slll

l
e '( ) 2
2a. . nnx d 2 ' nnx d (294)
• t = -,- ax' Sin - , - x - -, - ,- X.
If the first term on the right is integrated by parts twice, with account taken of
(289), the result takes the form
Jor h(x, t)
l
, n2 2 nnx
C.(t) + yc.(t) = - T
.
Sin - , - dx, (295)
9.17. Nonhomogeneous Conditions. The J\.Iethod of Variation of Parameters 487
with [he previously used abbreviation
(296)
It should be verified that the same eguation is obtained (Wilh less effort) by
introducing (292) into (288), differentiating term by term, and then identifying
the coefficients in the resultant sine-series representation.
Thus we have a linear ordinary differential eguation determining en(l),
where h(x, 1) is known. The associated constant of integration is determined by
the reguirement that (292) satisfy (290) when t = O, so tbat
.,~ en(O) sin n~x = f(x) (O < x < ').

Hence we must have
en(O) = a n, (297)
where an = T2 1"/ f(x) 510

. n7T.X
-,- dx. (298)
• o
The solution of (295) which satisfies (297) is obtained in the form
en(l) = ane- n"/). - 7I e-n'U-T)'). [L h(x, T) sin n~x dxJ dT (299)
(see Problem 1 LO), and the introduction of (299) into (292) yields the reguired
solution.
We consider next the generalized problem
a'
a2
T aT
ax' - al = h(x, t), (300)
T(O, t) = G(I), T(', t) = F(t), (301)

T(x, O) = f(x), (302)
with O <: x <: 1 and t >- O, where both the differential eguation and the pair of
boundary conditions may be nonhornogeneous. One method of dealing with
this problem consists of again assuming the solution in the form (292) and of
determining an appropriately modified differential eguation for en(t). However,
since a better method is available in this case, we consider that method first and
then investigate the former method.
The proposed method consists of first homogenizing the boundary con-
ditions (301) by subtracting from T any (respectable) function u(x, t) which
satifies (301). Here u(x, t) need not reduce the left-hand member of (300) to
zero, since failure lo do so merely will reguire the correction to satisfy a differ-
ential egualion with a modified right-hand member, and we now can deal with
such a modification. Wilh the choice
U(x. 1) = ; F(l) + (1 - ; )G(t), (303)

488 Solutions oí Partial Differential Equations of Mathematical Physics
which is convenient since it is linear in x, we then write

T(x , t) = u(x, t) + v(x, t) (304)
and find that the correction v(x, t) mu st solve the new problem
o
IX- a
~
a v - -a
2
av = h*( x, t ) ,
x t
veO , t) = O, v(/, t) = O, (305)
v(x , O) = f*(x),
where
h*(x, t) = h(x, t) + ; F(t) + (1 - ; )G'(t) (306)
and f*(:.:) = f(:.:) - ; F(O) - (1 - ; )G(O), (307)
under the assumption that F(t) and G(t) are differentiable.

Hence we see that the correction v(x, t) can be obtained in the form
v(:.:, t) = :f: C:(t) sin n7/cx,

n- ,
(308)
where C:(t) is defined by the right-hand member of (299) with h replaced by h*

and with a n determined fromf*.
In particular, if we take
h(x, t) = G(t) = f(x) = O, (309)
and so deal with tbe problem previously treated in Section 9.11, we obtain the
solution in the form
T(x, t) = 7F(t) + t, C:(t) sin n~x, (310)
where
C:(t) = e- n {an- ¡ I en"¡{I ; F(r) sin n~x dxJ dr}
" '> (311)
and where
an = -
2 x . nnxr'
yF(O) Jo y sin -,- dx. (312)
Wben use is made of the evaluation

'J
o
x . nnx
- , Sin - , - dx =
(_I)n+'/
nn
, (313)
we may verify that the solution just obtained is identified with the form given in
Equation (177) of Section 9.11, and obtained there by use of the Duhamel
superpos ition process.
The alternative procedure suggested aboye for the treatment of (300) to
(302) has the di sadvantage that the series solution which it produces converges
less rapidly than does the series (308). However, it has the advantages that it
does not presume the differentiability of F(t) and G(t) and, perhaps more impor-
9.17. Nonhomogeneous Conditions. The Method of Variation of Parameters 489
tantly, that it can be used in other cases when it is nor feasible to determine a
homogenizing function analogous to u(x, 1). For this reason, we next outline its
Llse in the presenl example.
Accordingly, we assume the solution of (300) to (302) in the form
T(x, 1) = f: K.(t) sin

n- l
n7,T.X, (314)
and deduce that
K.(I) =
.
') J'
7
o
T(x, t) sin n~x dx (315)
and that
2 IX 2
K '• ( 1) = -,- J.' a'
o
-. n1tx
ax'T Sin - , - dx - T2 J"o h( X,l )SIn
. nltx d
- , - x, (316)
as before . But now when rhe first term on the right in (316) is integrated twice by
parts, and use is made of the nonhomogeneous conditions (301), we obtain the
K~(I) + n; Kil) = ;~ [G(I) + (-1)·+ 1 F(I)] - 7Lh(x, 1) sin n~x dx,

(317)
in place of (295), with the initial condition
K.(O) = 7Lf(x) sin n~x dx. (318)
The introduction of the function K.(I), so determined, into (314) then yields the
required solution.
In particular, in the special case (309) it is found that the solution obtained
in this way is identified with the form given by Equation (176) of Section 9.11.
The comments made in Section 9.11 apply also in the more general case
considered here, in that whereas the series (308) will converge to v(x, 1) for all
x such that O -< x - O (when F, G, andf are respectable), the same is
not true of the series (314) unless F(I) = G(I) = O. Since each term of that series
vanishes when x = O and when x = I for aH 1, and since T(O, t) and T(!, t) do
not vanish unless F(t) and G(t) do so, the series (314) will converge to the
required solution only when O < x O.t Since T is known when x = O
and when x = 1, this fact would be of no con sequen ce if it were not a corollary
that the rate of convergence of (314) will be inferior to that of (308) when
O < x < I unless F(t) = G(t) = O.
The preceding examples illustrate the use of the method of variation of
parameters in a rather extensive dass of linear problems, specifically those m
tIn this connection, it is easily verified that if expressions for Tu and T, are obtained by
formally differentiating (314) term by term, and are introduced into (300), the requirement that
the resultant equation be satisfied yields an incorrecl differential equation for Kn(/) unless
F(t) = G(t) = O. This is a eonsequenee of the faet that term-by-term differentiation of (314)
is not justifiable if the represented funetion does not vanish when x = O and when x = I
(Seetion 5. J 2).
which a certain nonhornogeneity is troublesorne, but in which the method of

separation of variables would succeed ir that particular nonhomogeneity were
absent. The procedure essentially consists of first imagining the troublesome
nonhomogeneíty to be absent and writing down the characteristic-function
representation (series or integral) which would be assumed for the solution in
that case. The next step replaces the unknown constants of combination by
unknown functions ofthe variable or variables not appearing in the characteris-
tic functions (or, when superposition is by integration, replaces unknown
functions of a parameter by unknown functions depending also upon those
variables), and deduces differential equations determining these functions by
methods analogous to those used in the preceding examples. Several additional
examples are treated in Problems 111-116.
9.18. Formulation oC Problems. From the preceding examples it is apparent

that the number and nature of the conditions to be imposed along a physical
boundary or at a given time depend upon the type of partial differential equation
which governs the problem.
Thus, for example, in the case of Laplace's equation in two dimensions,
rpxx + rpyY = 0, we have seen that the solution rp is determined everywhere
inside a region CR if the boundary values of rp are prescribed along the closed
boundary e of CR. The same can be shown to be true also for other linear elliptic
equations, that is, for any equation arpxx + brp xy + crpyy + ... = 0, where
b 2 < 4ac. However, this statement is nO( true in general for hyperbolic equa-
tions, where b 2 > 4ac. For example, we readily verify that the express ion
rp = sin knx sin kny satisfies the equation rp xx - rpyy = 0, as well as the
°
requirement that rp vanish along the closed boundary of the rectC'.ngle
(O -< x -< 1, -< y -< 1) for any integral value of k. Thus for this equation
this particular boundary-value problem has infinitely many solutions, whereas
for Laplace's equation the only solution is that for which rp =
inside the rectangle.
°
everywhere
In the case of the particular initial-value problern (the Cauchy problem) for
which along the entire x axis the function rp and its normal derivative rpy are
prescríbed, say
rp(x, O) = f(x), arp(x, O) = g(x)
ay ,
and a solution valid (say) for al! positive values of y is required, it can be shown
that, for Laplace's equation (see Problem 60 of Chapter 8) and other el!iptic
equations, the solution exists only if f(x) and g(x) satisfy very stringent restric-
tions. In particular, it is necessary that al! derivatives of f(x) and g(x) exist for
al! real values of x, but even this condition is by no means sufficient to guarantee
the existence of a solution valid for all positive values of y. On the other hand,
unle ss the x axis is a characteristic base curve, this problem possesses a proper
solution in the case of linear hyperbolic equations when f(x) and g(x) satisfy
only very mild conditions.
9.18. Formulation of Problems 491
These examples are typical of the general linear case. Thus it may be
expected that, in general, elliptic equations are associated with boundary-value
problems, whereas hJperbolic equations are associated with initial-value prob-
lems. Parabo/ic equations, where the coefficient discriminant b 2 - 4ac vanishes,
are intermediate in nature. We next ¡¡st certain general types of problems which
commonly arise in connection with such equation s, many of which ha ve been
illustrated in the present chapter. The list is not in tended to be exhaustive.
Typical problems associated with elliptic equations may be illustrated by
considering Laplace's equation in the two-dimensional form
(319)
where (for example) T may represent steady-state temperature. Along the closed
boundary of a region we may prescribe T (the temperature) 01' the normal
derivative aTjan (a quantity proportional to the rate of heat flow through the
boundary). In the second case the temperature is determined only within an
arbitrary additive constant. Altematively, a condition of the more general type
aT
aT =
+ b(Jñ e
may be prescribed along the boundary (as, for example, in the case of heat
radiation [rom the boundary according to Newton's law of cooling). In case
the region involved is not simply coonected, and in certain other cases, it may
be necessary to add requirements of single-valuedness or periodicity. The region
in which the solution is to be valid may be the interior or the exterior of the
closed boundary. However, in case the regio n extends to infinity, restrictions
concerning desirable or permissible behavior of the solution at large di s tances
from the origin generally must be added. In sueh cases it is frequently convenient
to imagine that the conditions "at infinity" are preseribed along a circle (or an
are of a eircle) of infinite radius, forming the outer boundary (or the remainder
of a boundary whieh extends to infinity) of the infinite region involved (Figure
9.11).
x aT
Tor - x
an
aT
Tor-
an
(a) (b)
Figure 9.11
492 Solutions of Partial Differenrial Equations oC ~Iathematical Physics
Typical problems associated with hyperbolie equations may be illustrated

tn the case of the one-dimensional wave equation
a'w I a2w
ax2 = el ar' , (320)
where (for example) w may represent the defiection of a vibrating string (x is
distance along the string, t time). For a string so long that end conditions are
irrelevant, the prescribed initial values
w(x, O) = f(x), aw(x, O) = v(x)

at
of the function w (defiection) and w, (velocity) at tbe time t = O determine the
deflection for aU values of x at all following times (t> O). No additional
limiting conditions can be prescribed as t ~ 00 or as Ixl - co. However, for
a string of finite length I with both ends fixed (at x = O and x = 1) these condi-
tions are prescribed only over tbe inter valO < x < 1, and the end conditions
w(O, r) = O, w(l, r) = O
must also be satisfied, for aH time such that
t t > O. In this case, if we represent x and
t as rectangular coordina tes in an artificial
xt plane, we see that the solution is
req uired in the semi-infinite strip (O < x < 1,
w w
or or O < t < 00) (Figure 9 . 12). Along the
OW boundaries x = constant, the function w
ox ox is prescribed, whereas along the "time
boundary" t = O the quantities w and
// / aw/at are prescribed [for those values of x
w and ¡¡ x in the interval (O < x < 1)]. With this
interpretation, we may think of aw/at as
Figure 9.12 the "normal derivative" of w aloog the
"time boundary."
1t can be shown (see Problem 33 of Chapter 8) tha t tbe solution of this problem,
when O < x < 1, is the same as that for ao infinitely long string extending aloog
the entire x axis, for which at the time t = O the deflection and velocity are
prescribed for all values of x as odd periodic fuoctions of x, of period 21, agree-
iog witb f(x) and v(x) in the intervalO < x < l. Tbat is, the effect of the added
semi-infinite portions of the string is then the same as the effect of the physical
restraints at the ends. Thus, in a seose, this problem can be considered as purely
an initial-value problem.
The prescribed value of w along a boundary (eod) x = constant may vary
with time; also, an end condition of the more general form aw + bw x = e
may be substituted, as for example io the case of an elastic eod support. In
many problems io which w is prescribed as varying periodicaUy with time at
one boundary, a periodic response is eveotually propagated throughout the
9.18. Formulation of Problems 493
region, and certain transient effects introduced at the beginning of the motion
become negligible at sufficiently later times. In such cases it may be that onJy.the
limiting periodic response is of interest. Here we may suppose that the motion
was initiated at t = - 00; then only end conditions need be prescribed, and if
a periodic solution is obtained it may be considered in this sense as valid for
al! time. (We remark that, if the nature of the propagation and the character of
the transient effects are of interest, the use of the Laplace transform is particu-
larly advantageous, as is illustrated by a comparison of Problems 75 and 99 at
the end of this chapter.)
Considerations of trus general nature (but obviously with entirely different
physical interpretations) apply equal!y well to the formulation of problems
governed by a hyperbolic equation analogous to the one-dimensional wave
equation but in which the time variable t is replaced by a second space variable
y. For example (see Section 6.20), in the linearized theory of steady two-
dimensional f10w of a nonviscous compressible fluid, if the flow is nearly a
uniform flow in the x direction, the deviation in velocity from uniformity is the
gradient of a potential rp which satisfies the equation
( M2 _ 1) a2 rp _ a2 rp = O (321)
ax2 ay2 '
where NI is the ratio of the uniform f10w velocity to the velocity of sound in
the fluid. When this ratio is less than unity, this equation is elliptic and, in fact,
is reducible to Laplace's equation by an obvious change in variables. The
associated problem in this case then normally is a boundary-value problem.
However, when the flow is supersonic (NI> 1), the equation is hyperbolic, and
conditions of entirely different type must be prescribed. (When j\1 = 1, the
equation is not valid, since then the linearization which leads to this equation
is not permissible.) Suitable conditions in certain problems of this sort may be
obtained by replacing the time variable t by y in the discussion of the preceding
paragraph. In such cases the semi-infinite strip inside which the solution is to
be obtained may extend to infinity in either the x or the y direction. In particular,
certain basic problems involve the region consisting of the entire half-plane
y> O. The nature of prescribed conditions which make a problem determinate
is suggested by the consideration that the half plane may be taken as the limit
of either of the semi-infinite strips indicated in Figure 9.13, as the dashed-Jine
boundaries are moved parallel to themselves indefinitely far from the origino
In the first limiting case, where the half-plane is considered as the limit of
a strip extending in the y direction, the problem tends toward the usual initial-
value problem [Figure 9.14(a)], where rp and its normal derivative are both
prescribed along the entire x axis. Since this problem is completely determined
by the conditions along the boundary y = O, it appears that the "end condi-
tions" along the sides x = constant must be omitted in the limit. That is,
generally there are DO additional conditions to be prescribed "at infinity" in
this limiting case. However, in the second limiting case [Figure 9.14(b)], where
494 SolutioDS of Partial Differential Equations of IVIathematical Physics
y y
/ / /
x <p or ~
ély
(a) (b)
Figure 9.13
y y
/ ///
a", x -~ a.,. x
cp and- <p or-
"
ay ay
(a) (b)
Figure 9.14
tbe region is considered as the limit of a strip extending in the x direction, tbe
prescribed conditions along tbe dashed-line boundaries cannot be completely
lost in tbe limit, since the single condition prescribed along tbe x axis in this
case is not sufficient to determine a unique solution. Altbougb tbe end condi-
tion prescribed along tbe upper boundary y = constant is lost in the limit, as
in the preceding case , tbe same is not true for tbe initial-value conditions
prescribed along tbe boundary x = X o as X o ~ -oo. However, tbe number of
conditions prescribable along tbis line is, in general, reduced from two to one
in tbe limit. In sucb a problem nothing can be prescribed witb reference to tbe
behavior of the solution as x -. + 00 . Obviously, the limiting boundary could
be taken instead as tbe line x = X o as Xo - +00, in which case no conditions
could be prescribed as x ~ - oo. A problem of the type just considered is
treated in the following section.
Typical problems associated witb parabolic equations in practice may be
illustrated in tbe case of the equation
aZT 1 aT
axz = IXz al ' (322)
which (for example) govern s one-dimensional beat fiow in the x direction. When
the regioo is infinite in extent (-00 < x < +00), a prescribed distribution
9.19. Supersonic Flow of Ideal Compressible Fluid Pasl an Obsta ele 495
T(x, O) at the-time t = O determines T every-

where at all following times, t > O. When
the regio n is of finite extent (O < x < 1),
as for example in tbe case of a thin rod of
length 1 with insulated si des (Figure 9.15), the ~ T
or
distribution T(x, O) may be prescribed in this oT oT
interval when t = O, and in addition at each ax ax
end we may prescribe either the temperature
T or the rate of heat f10w through that end
(KT x ); or we may prescribe a condition of the /// / /
more general form aT + bTx = e at either Tor-

aT x
at
end, as, for example, in the case of Newtonian
cooling by radiation.t As a further alternative, Figure 9.15
the temperature distribution T(x, O) may not
be given explicitly but may be determined by specified values of T at the bounda-
ries of tbe region , under the assumption that initially (at the time t = O) steady-
state conditioos prevail. In still another type of problem , a periodically va rying
so urce of heat may be present at one boundary, leading to a temperature
distribution throughout the region which varies periodically with time and tends
to zero at large distaoces from the boundary (see. Problem 63).
The problems described in thi s section are typical of those commonly occur-
ring in many diverse fields, where only two independent variables are present.
We have chosen to identify the dependent variable with, say, temperature or
deftection of a striog for the purpose of fixing ideas and making intuitively
plausible the suítabílíty of the formulation. The aboye discussion can be
extended readily, in most cases, to problems involving a greater number of
space variables, the region eH becoming three-dimensional for Laplace's equa-
tion, and the x interval becomíng a two- or three-dimensional regíon in space
for the wave and heat-flow equations and for their analogies. In the latter cases
it may happen that an initial condition (at the time t = O) itself consists of
satisfying a partial differential equation (independent of time) together with
associated boundary conditioos.
9.19. Supersonic Flow of Ideal Compressible Fluid Past an Obstacle. As a

further illustration of physical problems involving hyperbolic differential equa-
tions, we consider the determination of the two-dimensional f10w of a nonvis-
tThe [ac! that for (322) one canDot prescribe bOlh Tand aT/al when I = O is associated wilh the
fac! that the line I = O in the xl plane is a "charaeteristic base curve" for (322) (see Seetion 8.8).
Further, we may notiee that if (322) is satisfied when I = O, and if T(x, O) = f(x), then there
mus! follow also T,(x, O) = rx,2Tx xCx, O) = rx,'f "(x), so that T,( x, O) is in fae! direetly de ter-
mined ir T(x, O) is preseribed as a twiee-differentiable funetion. (Compare Problem 53 of Chapter
8.) Sinee (322) involves only thefirst time derivalive, it is sometimes said to be "of tirst order"
in t.
496 Solutions of Partial Dilferential Equations of i'vIathematical Physics
y cous compressible fluid past an obstacle

in the form of a small protuberance from
an otherwise straight wall. The flow is pre-
1 scribed as having a uniform velocity Va
1 paralIeI to the wall at large distances
--1
-i upstream from the obstacle (Figure 9.16).
Va~ The flow velocity vector at any point
--1 can be considered as the sum of the uni-
I
o x
form velocity vector Vai and a small devia_
tion u. According to Equations (195) and
Figure 9.16 (200) of Section 6.20, for an ¡rrota/lonal
flow the deviation u is the gradient of a
"potential function" rp,
u = Vrp, (323)
where, for supersonic flow, rp satisfies the equation
(M> 1). (324)
Here M, the lvfach number of the flow, is the ratio of Va to the local velocity of
sound in the fluid.
The velocity of flow at any point can thus be expressed in the form
v = Vai + Vrp. (325)
Along the boundary comprising the wall and protuberance, the normal com-
ponent of V must vanish. If we take the equation of this boundary in the form
F(x, y) = O, (326)
the normal vector is proportional to V F, and hence this condition beco mes
[(Vai + Vrp) • VFlbOuod"Y = O. (327)
By writing (326) now in the explicit form
y = H(x) (328)
and setting F(x, y) = y - H(x), we find that this condition becomes
[{ i(Va + ~rp) + j ~rp} • (-iH'(x) + n] = O

x y boundary
or [ - ( Va + -arp), arp]
ax H (x) + -ay boundary = o. (329)
The quantity arp/aX is the x component, !Ix, of the velocitydeviation; and

hence, in accordance with the linearizing approximations Ieading to (324), arp/ax
is to be neglected with respect to Va in (329). Finally, noticing that H(x) is zero
except along the protuberance, say
H(x) unless O < x < 1,
= O (330)
and assuming that H(x) is sma/l in any case, we may further simplify the analysis
9.19. Supersonic Flow of Ide~1 Compressible Fluid Past an Obstade 497
by satisfying (329) along the projeetion of the protuberance on the x axis. Thus
we re place (329) by the linearized boundary condition
arpL-c, O) = V H'( ') (331)
ay o x.
Along any line x = x o, as x o ~ - 0 0 the yelocity must tend toward uni-

formity. In particular, the x component of the velocity deyiation must tend to
zero,
lim arp(x o, y) = O. (332)
Xo--:oO ax
Thus we require the solution of (324) whieh satisfies (331) and (332) in the
upper haif-plane y :> O. This problem is seen to be a particular case of that
indicated in Figure 9.14(b).
If we write
IX = .../M' - 1, (333)
the general solution of (324) can be written in the form
rp(x, y) = f(x - IXy) +- g(x -j. IXY), (334)
where f and g are arbitrary functions (see Section 8.5). The condition (331)
then becomes
-IXf'(X) + IXg'(x) = VoH'(x),
V
f(x) = g(x) - --'2H(x)
IX
+ e, (335)
where e is an arbitrary constant. lf we use (335) to eliminate f from (334),

there follows
V
rp(x, y) = g(x + IXY) + g(x - IXY) - --'2H(x - IXY) + e, (336)
IX
and also
arp~x, y) = g'(x + ay) + g'(x - IXY) - V o H'(x - IXY). (337)

X a
Now let y - , + ca along any "characteristic line" x + ay = e. Along any
of these lines we haye, from (337),
y
arp = g'(e)
ax
+- g'(e - 2IXY) - V o H '(e -
IX
2IXY)·
I
Bu t since (see Figure 9.17) any such line intersects I x+ ay ~ const. x- ay = const.
the boundary x = x o --> - 0 0 as y --> + 00, it I
follows that arp/ax must then tend to zero along I
I
each such line as y --> + 00, in accordance with
(332). Hence we must haye, in the limit, x
g'(e) + g'(-ca) - ~OH'(-ca) = O,

Figure 9.17
498 Solutions oC Partial Differential Equations of Mathematical Physics
or, since H(x) = O when X < O and hence H'( - co) = O, we must have
g'(e) .~ -g'(-co). (338)
But since this result must be true for al! positive values o/ e, we then conclude
that g'(e) must be a fixed constant independent of e and hence also that
g'(e) = g'(-oo). This statement contradicts (338) unless g'(e) = O, so that We
conclude that /he/une/ion gis a eons/an/.
Thus, finally, the deviation potential (336) becomes merely
rp(x,y) = - VOH(x -ay)

a
+ A, (339)
where A IS an irrelevant constant. The velocity components are then given by
Vx = Va + ~~ = v{ 1 - ; H'(x - ay)}
(340)
Vy = ~; = VaH'(x - ay).
From these results we conclude that /he

y veloeity eomponents are eonstant along any fine
x - ay = eonstan/. Hence, in particular, the
flow is modified from uniformity only in the
diagonal strip between the two lines x - ay = O
/ / and x - ay = l. The streamlines are as indi-
'
/
--- cated in Figure 9.18.
It is important to notice that any irregu-
larity at a point P of the x axis would prop-
o 1 x
agate a corresponding irregularity in the
Figure 9.18 fiow along the particular "characteristic line"
x - ay = e which passes through the point P
(see Section 8.10).
REFERENCES
1. References at end of Chapters 5 and 8.

2. CARSLAW, H. S., and J. C. JAEGER, Conduction of Heat in Solids, 2nd ed., Oxford
University Press, Inc., New York, 1959.
3. COURANT, R., and K. O. FRIEDR1CHS, Supersonic Fluid Flow and Shock Waves,
John Wiley & Sons, Inc. (Interscience Division), New York, 1948.
4. KELLOGG, O. D., Foundations of Potential Theory, Dover Publications, Inc., New
York, 1953.
Problems 499
PROBLEMS
Section 9.1
1. Verify that the equation
V2rp + 2a arp + 2b ~ = O
ax ay
is transformed into the "modified Helmholtz equation "
V2U = (a 2 + b 2 )U
by the substitution
qJ = e-((3 X +bY)U,
if a and b are constan!. (See also Problem 19 of Chapter 8.)

·2. Verify that the equation
is transformed into the heat-fiow equation
voU = ~2 au
a at
by the substitution
rp = e-(aX+Q : ctll)U,
if a and a are constan!. (See also Problem 20 of Chapter 8.)

3. Verify that the equation
is transformed into the heat-fiow equation
V2U = ~ au
a' al
by the substitution
if b and a are constan!. (See al so Problem 20 of Chapter 8 .)

4. If U(x, y, z) satisfies the equation V'rp = O, show also that the function
(ax + by + cz)U(x , y, z) satisfies the equation V4rp = O for any constant values of
Q, b, and c.
5. Ir U(x, y, z) satisfies Laplace's equation, show that a u ¡ax, a u¡a y, and a u¡ az are
also solutions.
6. Ir V(r, e,
z) satisfies Laplace's equation (in circular cylindrical coordinates), show
that aV¡ae and aV¡ az are also solutions but that aVIar generally is not a solution.
7. By assuming a solution of Laplace's equation
V2rp = O
500 Solutions of Partial Differential EquatioDS of NJathematical Physics
in the form
!p(x, y, z) = f(px + qy + z),
where p and q are constants and f is an arbitrary twice-differentiable function of its
argument, deduce that !p = f(ix cos u + iy sin u + z) is such a solution and, more
gene rally , show that
rp(x, y, z) = Sal" F(ix cos u + iy sin u + z, u) du
is a solution pro vided that F is such tbat differentiatio n under the integral sign is per-
mitted . (This is a "general " solution of Laplace 's equation in the sense that any solution
which can be expanded in a tbree-dimensional Taylor series at each point of a region
can be expressed in the form gi ven in that region .)
8. By proceeding as in Problem 7, obtain a "general" solution of the wave equation
V2 __ 1 a
2 rp
rp - e 2 al 2
in the form
rp(x , y , Z, 1) - J:" SaZ" F(x cos u sin v + y sin lt sin v + z eos v + el, u, v) du dv.
Section 9.2
9. Suppose that heat is flowing in a uniform rod of eross seetion a and perimeter p,
and that it is assumed that the temperature T does not vary o ver a eross seetion, and
hence is a funetion only of time I and distanee x measured along the rod. Assume also
tba! heat escapes from the lateral boundary by radiation , in sueh a way that the rate
of heat loss per unit of area is ¡¡,K(T - T.) , where K is the eondueti vity of the rod
material, T. tbe temperature of tbe surrounding medium, and ¡¡, is a constan!.
(a) By considering differential thermal equilibrium in an element (x, x + dx) of
the rod, show that there must follow
a ( K aT)
ax ax a dx - ¡¡,K(T - T . )p dx = aT dx ,
spa TI
and henee deduce that T then must satisfy the equation
~ (K aT ) = sp aT + ¡¡,K L(T - T o).

ax ax al a
(b) For a rod of circular eross seetion, of diameter d and of uniform eonduetivity
K, show that T(x, 1) must satisfy the equation
2
a T = -.!.. aT + 4¡¡'(T _ T)
ax2 1X2 al d·'
where 1X2 = K/sp.
(e) Verify that, if T. is assumed to remain constant, the substitution
T(x,l) = T o + U(x,/)e- 4 "." /d
leads to the normal heat-flow equation

a2 u 1 au
ax2 = 1X2 al .
Problems 501
10. Suppose that the or/hogona! coordina/es "¡, "o, and UJ of Seetion 6.17 are used to
speeify position in a heat-flow problem, and that the eonductivity K of the medium is
constant.
(a) Show that the heat-flow equation (13) beeomes
( b) If F is the flu x vector, representing the rate of heat fiow per unit area normal
to F, and K is the thermal conductivity, show that
F = - KVT = -K(!!"!'h¡ au¡

aT + U2aT + u, aT).
h 2 a"2 h, a",
Also show that the streamlines ( to which F is tangent) are determined by the equations
ht du¡ _ hi dU 2 _ hi du, .
aTjau¡ - aTjau2 - aTjau ,
(e) In the case when the temperature aod flow are independent of u" and when
h" h 2 , and h, are independent of UJ, show that the s/ream function I{/(u¡, U2), deter-
mined sueh that
where K is the thermal eooduetivity, has the property that the streamlioes in a surfaee
u, = eonstant are given by I{/ = constant.
(d) Show also that the function I{/ of part (e) has the additional property that the
total rate of heat fiow through a surfaee based on the are of aoy curve in a u, surface
joining points p¡ and P" and extendiog through a unit increment of UJ, is given by
f~ h,F - dr x u, f~ dl{/
f
PI
P' F • nh,ds = p¡
= - p¡
~ I{/(p¡) -
.
I{/(P,),
and that, when also h, = constant, the function I{/ also satisfies Laplaee's equation.
(See also Seetion 6.19. Notiee that sign differenees correspond to the fact that the
veloeity potential in fluid flow has been so defined that fiow is from lower to higher
poteotial, whereas temperature is so defined that heat fiows in the direction of decreas-
ing temperature.)
11. Suppose that, in a problem of steady-state heat fiow, the temperature T is pre-
scribed as zero over a portion of the boundary S of a regio n CR, the normal derivative
aTjan is zero over another part of S, and finally a conditioo ofthe form (aTjan) + J.lT
= O is prescribed over the remainder of S, say S ' , where J.l is a positive constant or
function of position 00 S'.
(a) Show that Equation (16) takes the form
and deduce that T then must vanish throughout CR .

(b) Use this result to show that the sofution of Lap/ace's equation V2T = O is
IIniquely determined in a regio n CR ¡f ei/her T, aT/ an , or ( a Tjan) + J.lT (J.l > O) is pre-
scribed at each point of the c/osed boundary, except in ¡he case when aTjan is prescribed
over Ihe complele bOllndary (in which case an arbilrary addilive conSlanl is presenl in Ihe
SOIl/lion).
12. Suppose that the heat-flow equation
V2T = J.... aT
(X2 al
is to be satisfied throughout a region <R bounded by a c10sed surfaee S, for all positive
values of time l.
(a) Show formally that Equation (16) then is replaeed by the more general form
ff L (VT)2 d, = - 2~2 %1 ff L P d, + # s T~~ da.

eb) Suppose that either T or aT/an is maintained as zero at all points of S, for all
time I > O, and that T is ini tially zero throughou t ffi when I = O. Show that there must
follow
for all I > O. Henee deduce that T must vanish throughout ffi for all 1> O.
(e) Use the result of part (b) to show that the solution of the heat flow equation is
uniquely determined for 1> O if T is preseribed in ffi when I = O and either T or
aT/an is preseribed on the boundary S for all 1> O.
(A rigorous treatment requires for uniqueness that the solution Talso be bOllnded
in ffi for all t ::> O.)
13. Generalize the result of Problem 12 to the case when either T, aT/an, or the eom-
bination (aT/an) + J.lT (J.l > O) is preseribed at eaeh point of the boundary S for
t > O and T is preseribed throughout ffi when I = O. (See also Problem 11.)
Section 9.3
14. The temperature T is maintained at 0 along threé: edges of a square pi ate of length
0
100 cm, and the fourth edge is maintained at 1000 until steady-stateeonditions prevail.
(Small areas near two eorners must be eonsidered as exc1uded.)
(a) Find an expression for the temperature T at any point (x, y) in the plate, using
the notation of Séetion 9.3.
(b) CaJculate the approximate value of the temperature at the eenter of the plateo
15. Determine the solution of Laplaee's equation in the reetangle O -< x -< 1, O -< Y -< d
whieh satisfies the conditions
T(O, y) = TU, y) = O, T(x, O) = g(x), T(x, d) = f(x).
[SlIggestion: Notiee that a convenient form of a particular solution of the equation,
satisfying the homogeneous eonditions, is
y )]
T" -
-
[0n Sin
. hnnY-'-b
-¡- I TI
. hnn(d-
Sin 1 .
SIn
nnx
-¡-'
where n is a positive integer.]

Problems 503
16. Suppose that the plate of Section 9.3 is of infinite extent in the y direction, on one
side of the boundary y = O, so that it occupies the semi-infinite strip O -< x -< 1,
O -< y < ce.
(a) Tf the temperature is to vanish on the lateral boundaries x = O and x = "
is to tend to zero as y ~ ce, and is to reduce to I(x) along the edge y = O, obtain the
temperature distribution in the form
nnx 2 JI. . nnx
T(x, y) = 2::Y.I
cne-nny¡1 sin - - , en = T O j(x) SIn - , - dx.
n= I 1
(b) Obtain ¡he same result formally from Equations (34) and (35), by first replac-
ing y by d - y and then considering the limit as d ~ ce.
17. Let Problem 16 be modified in such a way that the rate oI heat fiow, per unit dis-
tance, into the plate through points of the boundary y = O is prescribed as g(x), where
g(x) may be measured in calories per second per centimeter length along the boundary
y = O.
(a) Show that the condition along the line y = O then is of the form
-Kh aT~~ O) = g(x),
where h is the thickness of the plate.

(b) Obtain the solution in the form
T(x, y) = L
n= 1
ene-noy!I sin nnx,
1
en =
2
nnKh
JI .
O g(x)
nTr.x
SIn - , - dx.
18. Suppose that g(x) is prescribed in the form

(O < x < Xo - E),
(xo - E < x < Xo + E),

(xo + E < x < 1).
(a) Show that the coefficients in Problem 17(b) are then given by
21 . nTr.E . nnxo
en = n 2 n 2 KhE SIn - , - sm -1-'
(b) By considering the limit of Cn as E ~ 0, so that the function g(x) tends to

the unit singularity function O(x - xo), obtain the temperature distribution due to a
permanent unit heat source, supplying one calorie of heat per second at the boundary
point x = X o , in the form
T(x, y) = -2- L~ -1 . nnxo . nnx e-nny,I

SIn - - SIn _ _
nKhn~1 n , I
[Notice that this result al so can be obtained. directly from the result of Problem 17, by
making use of the fact that
(a < Xo < b),

for any continuous function q>.]
(e) If the solution of part (b) is denoted by V(x. y; xo), show that the sOlution of
Problem 17 for any preseribed g(x) ean be written in the form
T(x, y) = J~ V(x, y; xo)g(xo) dxo·
Interpret this relation.

19. (a) lf f(x) is identified with ó(x - xo) in Problem 16, show that the solution of
that problem beeomes
T(x, y) = 7 n:tl sin ~xo sin n~x e- nn

, '.
(b) lf this solution is denoted by U(x, y; xo), show that the solution of Problem
16 for any preseribed f(x) can be written in the form
T(x, y) = S: U(x, y; xo)f(xo) dxo·

(e) Verify that, with the notations of Problems 18(e) and 19(b), there follows
'i . ') - Kh aV(x, y; xo)
Ú \x, y, Xo - - ay'
and henee that U(x, y; xo) is the distribution due to a permanent hea/ doub/e/ of
strength Kh loeated at x = xo, the doublet being formed by the eonfluenee of a heat
souree at (xo, O) and a heat sink at (xo, 1:'), eaeh of strength Khll:', as 1:' ~ O.
20. Ir Problem 16 is modified in sueh a way that the lateral edges (x = O and x = /)
of the strip are insu/o/ed [so that aTjan = ±(aTjax) vanishes along those lines], obtain
the temperature distribution in the strip in the form
~ ., nnx
Tx,
( y) = Ca + n~l c"e- nx )-", cos -/-'
where
ca = +L f(x) dx, cn =
2
1
r n7rx
Jo f(x) eos -/- dx (n=1,2, ... ).
21. By imposing the eondition V2rp = O on an expression of the form

rp = Aa + A,x + A 2y + A3X2 + A_xy + A5y2 + A.x' + "',
deduce that po/ynomio/ so/u/ions of Lop/ace's equa/ion are speeializations of the expres-
sion
rp = 00 + a,x + 02Y + aJxy + a_(x2 - y2) + 05(X 3 - 3xy2)
+ a 6 (y' - 3X2 y) + ....
(See also Problem 26 of Chapter 8.)
22. Find the solution T(x, y) of Laplaee's equation in the semi-infinite strip O <: x <: 1,
O <: y < co, which does not grow exponentially with y, and is -sueh_that
T(O, y) = IX, + fl,y, TU, y) = 1X 2 + fl2Y' T(x, Óy· f(x).
[Sugges/ion: Write T = p(x, y) + u(x, y), where p(x, y) is a polynomial solution of
Laplaee's equation, satisfying the eonditions preseribed along x = O and x = 1, and
obtained from the result of Problem 21 in the form
Problems 505
and where the correction «(x, y) accordingly also satisfies Laplace's equation, but is
5ubject to homogeneous conditions uro, y) = 0, l/(I, y) = along x =
together with the initial condition u(x, O) =f("') - cx, - (CX 2 - cx,)(x/ I).]
and x = 1, ° °
23. Find the solution of Laplace's equation, in the rectangle
which satisfies the conditions
°-< -< °-<
x 1, y -< d,
T(O, y) = 0, T(I, y) = To -+- c; ,

T(x, O) = 0 , T(x, d) = f(x).
(See Problems 22 and 15.)
24. Find the solution of the equation
aT
2 aT
2 .
(co > O) ,
ax2 + ay' = sto COx
in the strip °-< x -< 1, °-< y < co, which satisfies the conditions
T(x, O) = T(O, y) = TU, y) = °

and the requirement that T be bounded as y - co in the strip. Consider separately the
case in which CO = kn/I, where k is an integer.
a 2z ,
a ax2 T
a 2z.
b ax ay T
a 2 z,
Cay2
az
..,.. d ax +
az
e ay +.!Z -
. _
°
is separable if and only if it can be written in one of the three forms
H(Y{ A (x) ~:~ + D(x) ~:J + Glx{ C(y) ~~2 + E(y) ~:J
+ [H(y)F,Cx) + G(x)F2(y)]z = 0,
. a 2z a 2z
H(y) [ A(x) ax2 + F,(x)z ] + G(x) ax ay -+- [H(y)D,(x) + az
G(x)D 2(y)] ax = 0,
G(x{ C(y) ~:~ + F 2(Y)zJ + H(y) a~'~ y + [H(y)E,(x) + G(x)E2 (y)] ~; = 0,
and, in each case, give the ordinary differentíal equatíons whích must be satisfied by
factors in a product solution z = X(x) Y(y) . (The first form is the most important one
in practice.)
Section 9.4
26. Show that the necessity of determining constants in Equation (48) by solving suc-
cessive pairs of equations can be avoided by writing (48) in the form
T = ao log :, + b o log :2 + J\ {a n [ (:' r- (rr' rJ + b n [ (;, r- (r/ rJ} cos ne

+ nt, {Cn[(;, r - (rr' n + dn [(;J - e)"J} 2
sin neo
(Compare Problem 15.)
506 Solutions of Partial Differential Equations of IVJalhemalical Physics
27. Along the inner boundary of a circular annulus of radii lO Cm and 20 cm the
temperalure is maintained as T(10, e) = 15 cos e, and along Ihe outer boundary the
distribution T(20 , e) = 30 sin e is maintained. Find an expression for Ihe steady-state
temperalure at an arbitrary point (r, e) in the annulus.
28. Along the circumference of the unit circ1e r = 1 a solution of Laplace's equation
is required lo lake on the value unity when O < < n and the value zero when e
n < e < 2n.
(a) Determine an expression for T valid when r < 1.
(b) Determine a corresponding expression valid when r > 1.
29. (a) Determine the steady-stale temperature at points of the s'.:ctor O <: <: ex e
O <: r <: a of a circular plate if the temperature is maintained al zero along-Ih~
straight edges and al a prescribed distribution T(a, e) = /(8) when O < e < (1" along
the curved edge.
(b) [n Ihe special case when ICe) = T o = constant, show that the temperalure
at interior points is given by
T(r, e) = -4T
1C.
o :¿:
nodd
-n1 (r-a )nnl. . nne
SIn _ .
eX
30. (a) Show ¡hat ¡he solution of Problem 29(b) tends lo the form
T(r, e) = 4 T o :¿: ..!... (..':....)n/2 sin ne

1T. nodd na 2
as the opening angle (1, of the sector tends to 2n, so that the sector fans out into the
interior of the complete circ1e with a cut along the radius which coincides with the
positive x axis.
(b) Noticing that e .......
O as the cut is approached from aboye, whereas 2n e .......
as the cut is approached from below, show that the interior temperature distribution
of part (a) is continuolls across the cut.
(c) Show that the derivative of the temperature in the positive y direction tends to
_1 aT] = 2To :¿: (.!....)n I2-t = 2To (r/a)-t / 2 (O < r < a)

[ r d1J 8 _ o na. odd a n a - r
as the cut is approached from aboye, whereas it tends to the negative of that quantity
as the cut is approached from below. [Since -(aT/ay) is proportional to the rate of
heat flow in the positive y direction, this means merely that heat must be continually
drawn off from Ihe plate at all poinls of Ihe cut in order Ihat Ihe desired temperature
dislribulion may be mainlained.]
31. The function rp(r, e) is required subject lo Ihe conditions Ihal rp salisfy Laplace's
equation inside Ihe circle r = a and al al! finite poinls outside thal circle, Ihat rp be
continuous across the circle r = a, that arp/ar decrease abruplly by a prescribed func-
tion F(e) as r increases through a, and Ihal arp/ar approach zero as r - , co.
(a) Show that rp may be assumed in Ihe form
~ (r)n
:¿: -a (A n cos ne + En sin ne) (O <: r <: a)
n_ 1
rp(r, e) = Ao + a a n
e log -r + :¿: (-) + < co).
OQ
(A n cos ne En sin ne) (a <: r

\ r n- l
Prob ¡eros 507
(b) Show that the additive constant Aa is arbitrary, and that the remaining coef-
ficien ts are gi ven by
e
a
= 2n
rz.
Jo F(e) de, A. = 2~n
r
Jo
z,
F (e) cos ne de,
B. = 2:n l Z
' F(e) sin ne de (n=I,2, ... ).
32. Suppose that the outer boundary r = r2 of an annular pI ate is insulated, and that
the temperature is prescribed as I(e) along the inner boundary r = r l .
(a) Show that the expression (48) must be specialized to the form
T = ao + I: n= I
(r· + r~·r-·)(a. cos ne + c. sin ne)
when r¡ -< r -< r2·
(b) By writing A. = r~a. and en = r~c., and introduci l1 g the abbreviation
p = rz/r¡, obtain the desired temperature distribution in the annulus in the form
T= Ao + .t [(;J. +
l
pz.(rrIYJ(AncOsne + e.sinne),
where
1
l z _t'" I(e) cos ne de e =
f
Z"
o I(e) sin ne de
Ao = 2 n • I(e) de, A. - n(1 + pz.) , • n(1 + pz.) .
Section 9.5
33. By making use of the relation
rJoz, -;-----,dArp;..-_
1 - cos rp
2n
(1 Al < 1),
obtain the result
l fZ"_~__~~a~z__~r_2__~____~ drp = 1 (irl < a),

2n o a Z - 2ar cos (e - rp) + r z
and thus verify directly the validity of the Poisson integral formula (64) in the case
when T(a, e) = T o = constant. (Notice that the integrand is periodic in rp.)
34. Use Ihe Poisson integral formula (64) and the result of Problem 33 to show that,
when O < r < a, there follows
1
TCr 8) < -2
n
f'·
o a -
' 2
ar cos
a
2
- (e r'
- rp
) + r2 [T(a, rp)]m" drp = [T(a, e)]max.
and hence, by considering also the corresponding inequality for - T(r, e), deduce that,
if T satísfies Lap/ace' s equation inside a eire/e, then Tcannottake on its maximum or míni-
mum va/ue inside the boulldary unless it is constanl Ihroughoul Ihe cire/e. (The same
result can be established for an arbitrary finite region.)
35. Show Ihat when r > a. the negalive of the right-hand side of Equalion (64) repre-
sents the harmonic function T(r, e) at all points oulside the circ\e r = a in terms of
T(a, e). [Compare (53) and (60) .]
508 SolutioDS of Partial Differential Equations of Mathematical Physics
Section 9.6
36. (a) The temperature T on the suñace of a sphere of radius a is maintained at

T = T o(l - cos rp), where rp is the cone angle. Find the steady-state temperature at an
arbitrary point inside the sphere.
(b) If, instead, the temperature is maintained at a constant value T o over the
upper hemisphere (O <: rp < 11:/ 2) and at zero over the remainder of the surface, deter-
mine the first three terms in the series solution of Section 9 .6.
37. If the temperature of a spherical surface r = a is maintained at T(a, rp) = f(rp ),
where rp is the cone angle, and if the temperature tends to zero as r ~ 00 , show that
the temperature distribution ourside the sphere is given by
T(r, rp) = .~o e. ( ~ r+ I p.(cos rp),

where e. is defined by Equation (81).
38. lf a spherical surface r = a is maintained at constant temperature T o, show that
the temperature at aH internal points is also T o, whereas the temperature at points
outside the sphere is given by Toa/r (r > a).
39. Suppose that the steady-state temperature T in a sol id right circular cylinder pos-
sesses axial symmetry, and hence is of the form T = T(r, z), where r is distance from
the z axis.
(a) Show that T then must sat isfy the equation
1 a (r aT)
-;- al' ar + aaz2T _
-O
2
inside the cylinder and that, if a product solution is assumed In the form T(r, z) =
R(r )Z(z), then R and Z must satisfy equations of the form
d 2R dR d 2Z
r dr 2 + dr + IXrR = O, ----:L"'> -
u z '2
IX Z = O
'
where IX is an arbitrary constant.

(b) If only solutions which remain finite along the z axis are admissible, show
that the choic'e IX = k 2 , where k is real and positive, leads to exponential or hyperbolic
functions of z and to the function Jo(kr), whereas the choice IX = -k 2 leads to circular
functions of z and to the function lo(kr). Finally, show that the choice IX = O leads to
R = el and Z = d,z + d 2 •
40. (a) Suppose that a solid right circular cylinder of radius a is of infinite extent on
one side of the plane face z = O, and that the temperature is maintained at zero along
the lateral boundary, whereas the temperature distribution over the face z = O is
prescribed as T(r, O) = f(r) . U se results of Problem 39 to show that the steady-state
temperature at interior points is given by
~
T(r, z) = L: A.e-k"'Jo(k.r),
11==1
where k. is the nth positive root of the equation Jo(ka) = O, and where
Problems 509
(b) If the temperature over the face z = O is maintained at a constant value T o ,

show that the resultant interior distribution is given by
41. Suppose that the faces z = O and z = 1 of a s olid right circular cylinder are main-
tained at temperature zero, and that the temperature distribution along the lateral
boundary r = a is dependent only on z, and is prescribed in the form T(a, z) = fez) .
Use results of Problem 39 to obtain the resultant steady-state temperature distribution
inside the cylinder in the form
~ Io(nnr/l ) . nnz
T (r, z) = "'-' A ( / 1) -1-'
"= 1 nIo nna
Slll
where An = 7 J: fez) sin m;z dz.

Section 9.7
42. Suppose that a column with rectangular cross section can be considered to be oC
infinite height on one side of its base O -< x -< 1 O -< y -< 12 in the plane z = O. If
"
the lateral boundaries are maiotained at zero temperature and the temperature dis-
tribution over the base is prescribed a s T(x, y, O) = f(x, y), obtain the internal te m-
perature distribution in the form
~ ~
. nm:x . nny
T(x, y, z) = ~ ~ c mil e- k ",,,Z SIn -tI- Slll -1- ,
," - 1" "" 1 2
where k mn and C mn are defined by Equations (lOO) and (106).

43. (a) If the column considered in Problem 42 is oC circular cross section, of radius
a, and ir the temperature over the base is prescribed as T(r, (J, O) = f(r, (J), show that
the internal temperature distribution is given by
T(r, (J, z) = f:
"- 1
aOne-k' "'Jo(konr) + f: L
m "" [ 11 - 1
e-km·'Jm(kmnr)(amn cos m(J + b mn sin mfJ),
where k mn is the nth positive root of the equation Jm(ka) = O, and where amn and b mn
are to be determined in such a way that
n:ti [aonJo(konr) + m~, (a mn cos mO + bmn sin mO)Jm(kmnr)] = f(r, fJ)

when O -< r < a and O < O < 2n.
(b) By multiplying the equal members oC the preceding equation by
Jp(kpqr) sinpO r dr dO,
where p and q are any positive integers, and integrating the result over the base of the
colurnn, deduce formally that
b p• n; 2
[Jp + l (k pqa)]2 = ¡"la
Jo O Jp(kpqr)f(r, fJ) sinpfJ r dr dO.
(The remaining coefficients can be determined in a similar way.)

510 Solulions of Partial Differential Equations of l\Jathematical Phnics
44. Jdultidimensiol1c¡/ characteristic functiolls.

(a) Consider lhe characleristic-value problem in which A is lo be determined such
that the differential equation
V2F -L AF = ° in CR
admits a non trivia l solution which vanishes on the boundary of the region CR. lf Al and
A2 are two distinct characteristic numbers and cI>1 and cI>2 are corresponding charac-
teristic functions , show that
By transforming the right-hand member to an integral o ver lhe boundary of CR [see

Equation (123) of Chapter 6) , deduce that
Show also that the orthogonality property slill follows if, at each boundary point, the
generating characteristic-value problem requires that either F = °
or aFIan =
F + rx(aF/ an) = 0, (No rice that the preceding results apply equally to a two-dimen-
or °
sional problem, in which CR is a portio n of aplane or surface and its boundary is a
curve,)
(b) Generalize the results of part (a) 10 the case when lhe governing equation is
V • (pV F) + ArF = 0,
where p is continuously differentiable in CR and r is continuous, showing that
(See Problem 84 of Chapter 6,)

45. Poisson's equation. Suppose that ¡he function q; is required such ¡hat
in CR,
where q; is to satisfy a homogeneous condition, of the form rp
+ rx(aq;;an) = 0, on each part of the boundary of CR,
rp
= ° or arp/all = ° or
(a) If the associated characteristic-va lue problem

V2F + AF = ° in CR ,
subject 10 the same ' boundary conditions as those io part (a), has the characteristic
numbers Al , A" .. ,and corresponding characteristic functions cI>., cI>" ... , so lhat
V 2 cl>. + A.cI>k = 0, use the results of Problem 44(a) lO show formally that the solu-
tion of the original problem is of the form
~ A
rp = - 2: .2cI>.,
'_1 A.
provided ¡hat A = °is nol a characteristic l1umber, where A. is the coefficient of cl>. in
the representation of h in CR,
Problems 511
Show a lso that

IR !fI h d-r:
k
. '
J<!l !fI~ d-r:
where J<!l ( ... ) d-r: here denotes integrarion over the two- or three-dimensional region
<R. [In a two-dimensional problem, it often is desirable to label the characteristic num-
bers and functions with a double subscript (say, as Am. and !fImn) in which case the sums
on k are replaced by double sums on m and n. Similarly, in a tmee-dimensional problem
three indices may be desirable.]
(b) Specialize the results of par! (a) to obtain the solution of Poisson's equation
in a reclangle,
a2rp a rp
ax2 + ay2
2
= h(x, y)
with rp = °
along the complete boundary, in the form
~ ~ A mn . m7rX . n7r y
rp (
x, )
y = - ¿., ¿., ~2(
171=1 n - l 1" m 2jl21 + n 2j12)
2
510 - , . - SIO -1-'
tI 2
where
Amn = ¡¡
4
1 2 J'. J"
o o
. m7rx . n7ry
h(x, y) SIO -1- sm -1- dy dx,
1 2
(Notice that, if rp were prescribed otherwise on the boundary, the required solution
could be obtained by first solving that problem with h replaced by zero, and then add-
íng the aboye expression to that solution .)
46. Let the Laplace operator V2 be replaced by the Helmholtz operator V2 + A in
the differential equatioo for rp in Problern 45, so that rp is to satisfy the equation
V 2rp + Arp = h in eR,
with the boundary conditions specified as before.

(a) Show that then
·A k
rp = k~l A - Ak !fIk,
with the notation of Problem 45(a), provided that A is not identified with one of the
characteristic values of A.
(b) If A = A" show that there is no solution unless A, = 0, SO that
S<!l !fI,h d, = 0,
in which case the coefficient of!fl, in the expression for rp is arbitrary, the other coef-
ficients being determined as before.
(c) Specialize the preceding results to the rectangle of Problem 45(b).
(Compare the results of Section 5.8.)
47. The general Dirichlel problem lor a sphere. 11 is required to determine ¡he solution
of Laplace's equation, V2T = O, inside the sphere r = a, such that T = Ion the sphere
surface, where lis prescribed.
(a) Using spherical coordina tes (r, rp, e), show that a product T p = R(r)F(rp, e)
satisfies Laplace's equation if and only if
2
d R ..L 2 dR _ AR = o
dr2 ' r dr
and V}F + AF = O,
where V} is the two-dimensional Laplacian operator on the unit sphere r = 1,
_1_ ~ (aF sin rp)

2
V'sF = + _1_ a F,
sin rp arp arp sin 2 rp ae 2
and A is a separation constan!.
(b) Show that product solutions of the F equation which are periodic in e must
be of the form (rp)0(e), where
and
d 2 
drp2 + d
drp cot rp + (A - m 2 csc 2 rp) = O,
with m an integer (which can be taken to be nonnegative) and that such a product will
be finite at the poles of the sphere (rp = O, :n;) if and only if A = n(n + 1), where n is an
integer, and the rp factor is a multiple of P'::(cos rp). (Use relevant results stated in
Section 4.13.) Hence deduce that, for each nonnegative integer n, A = An = n(n 1) +
is a characteristic number with a (2n + l)-parameter characteristic function
Yn(rp, e) = anOPn(cos rp) + t

m=t
(a nm cos me + b nm sin me) p,:: (cos rp).
(Any such expression is called a surface spherical harmonic of degree n.)

(e) Verify that
(A p - Aq ) # s Y p Yq da = [ f" {:rp[ (Yp aa~q - Yq aa~p) sin rp ]

+ _l_~(y aYq _ y ayp)} ded
sin rp ae p ae q ae rp
= O (p *" q),
where S is the unit sphere r = 1 and, accordingly, da = sin rp de drp, and deduce that
the product functions
Pn(cos rp), P'::(cos rp) cos me, P'::(cos rp) sin me
(n = 0,1,2, ... ; m = 0,1,2, ... ) comprise an orthogonal set on the surface of that
sphere.
(d) Assuming that the function f(rp, 8) can be represented by a series of the
spherical harmonics on the surface of the unit sphere, and that the series can be in-
tegrated term by term over that surface, and making use of the known formula
Jor" [pm( )],. d 2(n + m)!

nCosrp -smrp rp=(2n+1)(n-m)!'
Problems S13
deduce t hat
~
f(rp, O) = ~ Yn(rp, O)
n= O
~ ~
= ~ anOPn(cos rp)
n=O
+ ~ L:
11""0 m= I
(a nm cos mO + b"m sin mO)P':,'(cos rp)
for O -< rp -< n , O -< O < 2n, where
JI: s f(rp, O)Pn(cos rp) da,

+ l 'jJ
= 2n4n
=
2n
?
+
_1l
l (n - m)!
(' )'
n.- m .
§ s
f(rp, O)Pnm (cos rp) cos m O da,
b nm =
2n +
2n
1 (n - m)!
(n + m)!
JI:
'jJ s f(rp,
O m
) p" (cos rp)
•
Sin m
O
da.
(This representation is valid, in particular, when f and its ftrst partial derivatives are
continuous functions of rp and O on the sphere S, but these conditions are by no means
necessary.)
(e) From the preceding results, deduce that the solution of Laplace's equation
V2T = O in the sphere r = a, which tends to f(q¡, O) as r - a (and is finite at r = O),
is of the form
T(r, rp, O) = ~
n~o (r)"
a Yn(rp, fJ),
with the notation of part (d). (The terms superimposed in this representation are some-
times called solid spherical harmonics, or simply spherical harmonics.)
Section 9.8
48. Let a right circular cylinder of radius a be placed in an initially uniform flow of an
ideal incompressible fluid, in such a way that the axis of the cylinder coincides with
the z axis and the flow tends to a uniform flow with velocity V o in the x direction at
large distances from the cylinder. Assume that the cylinder can be considered to be of
infinite length.
(a) If polar coordinates are used in the xy plane of flow, show that the velocity
vector V is related to the velocity potential rp(r, O) by the equation
V = u arp
, ar
+ u.~ i!.P.,
r ao
and that rp must satisfy the equation
and the boundary conditions
apea. fJ) = O ap(CXJ. O) = V cos O

ar ' ar o ,
in addition to the requirement tbat V, = arp/ar be a single-valued function of (J.

514 Solutions of Partial Dilferential Equations of l'Vlathematical Physics
(b) By imposing lhese reslrietions on an expression of form (45), obtain the

veloeity potenlial in the form
rpCr, e) = Ca + ke + VD (r + :2) eos e,
where Ca and k are arbitrary eonstants, and show that the veloeity vector then is given
by
Ce) Verify that V, vanishes when r = a and that V tends to VDi as r -> ec, re-
gardless of the value of the eonstant k.
49. (a) Show that the stream function lJI(r, e) eorresponding to the flow obtained in
Problem 48(b) can be expressed in the form
lJI(r, e) = da - k log r + Va (r - :2) sin e,

where do IS an arbitrary eonstant, and henee ¡hat the streamlines are defined by the
equation
Vay( I - X2 a;. yo) - ~ log (x' + y2) = eonstant,
in rectangular eoordinates .
(b) Show that the f10w is symmetrieal about the x axis when k = O and sketch
typieal streamlines in t his case.
(e) Show that the added velocity potential ke (whieh is not single-valued) eor-
responds to a eireulatory flow about the eylinder, in whieh the veloeity V, is cireum-
ferential, with magnitude k/r a¡ distanee r from ¡he eenter of the eylinder seetion, and
that the eirculation fe Y, • dr is given by 27T.k, around any eircle r = constan!.
(Notice that the ftow is not determined unless the circularion is prescribed, in
addition lo the values of V, at r = a and of V as r - > ec. Notice also that the flow is
indeed vortex-free, since the point r = O about which the fluid tends to rotate is not a
point in the fluid itself.)
50. The "stagnation points" of a flow are the points at which the flow veloeity is zero.
(a) Show that when no circulation is present the stagnation points in the flow
of Problem 48 are the points (a, O) and (a, 7T.) at the intersections of the x axis and the
boundary of the cylinder section.
(b) With the abbreviation (1, = k/(2 Vaa), show that when cireulation is present
the stagnation points are (a, sin- 1 (1,) and (a,7T. - sin- t (1,) when 1(1,1 < 1. Show also
that when 1(1,1 :> 1 there is only one stagnation point in the flow, and that that point
is at a distance aCI (1, I + "';(1,2 - 1) from the center of the cylinder section, on the posi-
tive or negative y axis, according to whether (1, is positive or negative.
Section 9.9
51. By multiplying both sides of Equation (148a) by
J.(w:;qr) cos pe rdr de

Problems SIS
and integrating o ve r the area of [he membrane, obtain the results
am/,~ 2 [Jm+ 1 (w~.a)r f "f Jm (W~"r ) f, (r,

= O) cos mO r dr dO (m *0)
and
ao.na2 [J,(W~na)r f" fJo (W~nr) f,(r,O)rdrdO

= (m = O) .
(The coefficients b mn, where m :> 1, are determined in a similar way.)

52. For a freel y vibrating square membrane of side " supported along the boundary
x = 0, x = " y = 0, y = 1, obtain a permissible e x pression foc the deflection w(x , y, t)
by methods analogous to those of Section 9.9 , in the form
~ ~ . -mnx . nn y( b ' )
w = ~ ~
", - 1 n= I
sin , - Sin - , - a m" CQS OJ m,,! + m" sin OJmnl ,
where w mo = n-v/ m2 + n2.j p'2

T .
53. (a) Suppose that the square membrane of Problem 52 is initially deflected in the
form
w(x, y, O) = f(x, y)
and is released from rest. Obtain an expression for the ensuing deflection in the form
00 O<> • m11:x . n7ry
w = ~ ~ amo Sin - ,- Sin - , - cos w mnt,
m= 1 ,, - 1
where amo = ;! s: s: f(x, y) sin m7x sin n~y dx dy.

(b) Obtain an expression for w in Problem 52 if initially the membrane IS un-
deflected but if the membrane is in motion such that
aw(x, y, O) ( )
at = v x , y.
54. (a) From the results of Problems 52 and 53, show that free vibrations of a square
membrane are compounded of "natural modes" of the form
. mnx . nn y ( )
wmn(x, y, t) = A mil sin - , - SIO -1- cos OJ m,,! + cx,mn ,
with circular frequencie s
Wm • = n../mZ
::::T--;--'~2 fT ==..;/ m 2 2+
+ n ...¡prz n2
n,
where n = n../(2T)J(p, 2) is the fundamental circular frequency, and with COrre-
sponding amplitudes
A . m7tx . nn v
<¡Jmn ( X , Y ) = mn SlD -1- S in _,_o ,
where m and n are positi ve integers .
( b) Deduce that in the fundamental natural mode (w = w" = n) the amplitude
. nx . ny
!p = a S in -,- Sin -,-
is zero only along the boundary, so that there are no interior nodal lines.
(c) Show that [he second natural circular frequency corresponds to two modes
in which either m = 1 and n = 2 or m = 2 and n = 1, and is given by úJ = úJ I2 = úJ 21'
= -v'TO 0./2, and that the most general motion with this circular frequency possesses
an amplitude of the form
rp = a
. 2:n:x . :n:y
sm -,- sm -,-
+ h sm
. nx . 2ny
-,- sm -,-
. nx . n y (
=
nx
2 sm -,- sm -,- a cos -,- + b cos n-,-y ) ,
where a and b are constants. Deduce that there is one interior nodal line, along the
curve a cos nx/' + b cos ny/I = O, that this curve always passes through the center
of the square, and that in the cases b = ±a the line is a diagonal of the square.
Section 9.10
55. The temperatures at the ends x = O and x = 100 of a rod 100 cm in length, with
insulated sides, are held at 0° and 100°, respectively, until steady-state conditions
prevail. Then, at the instant t = O, the temperatures of the two ends are interchanged.
Find the resultant temperature distribution as a function of x and t.
56. (a) Obtain permissible product solutions of the heat-flow equation (154) which
satisfy the conditions aTeO, t)/ax = O, aT(l, t)/ax = O, corresponding to the require-
ment that there be no heat f10w through the boundaries (ends) x = O and x = l.
(b) Use these results to solve the modification of Problem 55 in which at the time
t = O the two ends of the rod are suddenly insulated.
57. Suppose that the rod considered in Section 9.10 is such that heat escapes from the
lateral boundary according to Newton's law of cooling, so Ihat T(x, t) satisfies the
equation
a,2 aZT
ax2
= aT
at
+ [j(T - T )
o ,
where [j is a constant and T o is the temperature of the surrounding medium. (See

Problem 9.) The initial Iemperature distribution is T(x, O) = f(x) and the ends x = O
and x = , are maintained at TI and T 2 , respectively, when t > O.
(a) Show that the substitution
T(x, t) = To + U(x, t)e- P'
reduces the problem to the following one:
a u 21 aU
- - =- --,
ax2 a,2 al
U(x, O) = f(x) - T o , U(O, t) = (TI - To)eP', u(I, t) = (T z - To)eP'.
(b) In the important speciaI case when TI = T 2 = T o , obtain the solution of the
original problem in the form
~ a_ sl·n nn,xe-[p+(nJn'(X~),'12)r,
T( x, t) = To + ,,=1
""' •.
Jor f(x) sm -,-

l
2 nnx . 1 - cos nn 2T
where °n = I dx - nn o·
Problems 517
58. A rod 01' length 1, with insulated lateral boundaries, has the end x = O maintained
at T = T¡ when t > O, whereas heat escapes through the end x = l aeeording to
Newton's law of eooling in the form
[ hI aaT + (T - T o)] = O,
X x "", r
where T o is the temperature of the surrounding medium and h is a eonstant. The initial
temperature distribution along the rod is preseribed as T(x, O) = f(x).
(a) Obtain the steady-state distribution in the form
T, - T o x
T s = T, - 1 + h T'
(b) Show that the transient distribution can be assumed in the form
= .kx ,,,.
TT(X, t) = L: a n Sin -In e- k .-.-,!/-,
n== 1
where k n is the nth positive root of the equation tan k + hk = O.

(e) Obtain the required temperature distribution in the form T = Ts + TT,
where a n is determined by the relation
an Lsin 2 k,x dx = L [/(x) - Ts(x)] sin k,t dx_
59. (a) Show that the assumption of a solution of the one-dimensional heat flow
equation as a linear function of t, in the form T = tf(x) + g(x), leads to the require-
ments f" = O and (l,2g" = f- Henee obtain the particular solution
Tp(x, t) = e¡(x 3 + 6(1,2tX) + e2(x 2 + 2(1,2t) + e3X + e.,
where the e's are arbitrary eonstants.

(b) Use this result to obtain a particular solution for whieh aTeO, t)/ax = O and
aT(l, {)/ax = C in the form
Tp = ~ (x2 + 2(1,2t).
(e) In a similar way, obtain a particular solution for which T(O, t) = O and
T(l, t) = Totin the form
60. A rod with insulated sides has its end x = O insulated, whereas heat is introdueed
into the end x = l at a constant rate, so that aT(l, t)/ax = C, where C is a eonstant.
The initial temperature distribution is preseribed as T(x, '0) = f(x).
(a) Show that no steady state can exist in which T is independent of time, but
that [se e Problem 59(b)] the temperature distribution T p = C(X2 + 2(1,2t)/21 satisfies
the end eonditions (and also happen~ to speeify a state in whieh the rate of heat flow
at any point does not vary with time).
(b) Obtain the desired distribution in the form
518 SolutioDS of Partial Differential Equations oC Mathematical Physics
where
ao =, J"I o [ f(x) - e 2 dx,

2,x J - ,
a. - 2 Jol' [J()
x - e 2J cos-,-
2'x mtx d x,
when n = 1, 2, ....
61. Use the result of Problem 59(e) to obtain the solution of the one-dimensional
heat-flow equation for whieh T(O,I) = O and T(I, 1) = Tol when 1 > O and
T(x, O) = J(x), in the form
T(x 1) =
,
T
o
[3....L 1 - _1_ X('2 -
6rx2L
x2)J + "~I
~ a n sin n1tx
L
e-""'~" ¡l' )
where a n -- 2
-,
To
Jo" [ J(X) + 61X2r~( '2
- 2
X) J. n1tx dx.
Sin - , -
62. The boundary of a circular plate of radius a is maintained at a temperature whieh

varies periodieally with time, aeeording to the law T(a, 1) = T o eos COI, where T o and
co are constan!. Determine the steady periodie temperature variation at interior points
as follows:
(a) Let T(r, 1) be the real parl of a eomplex funetion U = F(r )e'ru<, and show that
F(r) then must satisfy the equation
d'F
lJi2
+ ..!.. dF
r dr
_ ¡co F
1X2
= °
and the eondition F(a) = T o, together with the requirement that F(r) be finite at r = O.
(b) By using the results of Seetion 4.11, deduce that
_ J o(i 3/ 2kr) 1M
U(r, 1) - T o J (i3 / 2ka)e ,
o
where k = ,y' CO /IX.
Ce) Deduce that
ber (kr) + i bei (kr)
+
T(r, 1) = T o Re { ber (ka) + ¡ bei (ka) (cos COI ¡sin COI)}
= T o Re{Mo(kr) eH9.(krl-9, (kal

. Mo(ka)
."',¡}
Mo(kr) ()
= T o Iv! o(ka) cos [ o(kr) - (}o(ka) + COI],
wbere
bei (kr)
Mo(kr) = ,y'ber 2 (kr) + bei' (kr), (}o(kr) = tan
-1
ber (kr)'
(d) Show that the temperature at the center r = O is given by
T(O, 1) = M~ka) eos [COI - (}o(ka)] .
63. Assume that a portion of the earth's surfaee may be considered as plane and tbat
effects of the periodic heating due to the SUD are transmitted only in the x direetion,
perpendicular to the surface. If the resultant temperature on the surfaee of the earth
is taken in the form T(O, 1) = T o eos COI, show that the temperature at depth x is given
by
T(x. 1) = Toe-"w' 2(x/~l cos (COI - J~ ~)
Problems 519
wilh lhe notation of Section 9.2. [Consider lhe boundary value T o cos W( as the real
part of Toe''''' and determine the real par! of an expression U(x, 1) = X(x)e''''' whicb
satisfies the one-dimensional heat-flow equation, reduces to Toe'M when x = O, and
va nishes when x ~ ec. ]
Section 9.11
64. The temperature at the end x' = O of a rod of length 1 is held at 0 while the tem- 0
perature at tbe other end is varied periodically according to the law

T(l, t) = T o sin Wt
when t > O. Suppose that when t = O the temperature is zero at all points in the rod.
(a) Find the temperature distribl1tion in the rod. [Use Equation (177).]
(b) Show also that for prescribed oscillation so slow that (lw)2 « 1, and for times
such that t » 1, the approximation
T = To [1 sin rot + 2~W C~I (~~ ). sin n~x) cos rotJ

is valid. [By integrating tbe Fourier sine expansion of x twice, and determining the
constants of integration, it can be shown that the series in parentheses represents the
function 7l: 3 x(x 2 - 12)/12/ 3 .]
65. Show that Equation (177) can be written in lhe form
T(x, t) = 1F(t) + ; .:tl ( n1)"[F(t) - ~2 1: F(r)e-"('-<) /' drJ sin n~x .

66. (a) If F(t) is the unit singularity function 0(1), show that the temperature distribu-
tion given by Equation (176) takes the form
T(x, t) = 2. i; (-l).+l n e-"'/' sio n7l:X (1 > O).

7l:1.~1 I
(b) If this expression is denoted by U(x, t), verify that U(x, t) = aA(x, t)/ at,
where A is defined by (168), and that the temperature distribution (176) corresponding
to a general function F(t) can be written in the form
T(x, 1) = t U(x, t - r)F(r) dr.
67. If a rod with insulated sides initially is at a temperature distribution T(x, O) =

¡(x), and if the conditions T(O, t) = O and T(I, t) = F(t) are imposed when t > O,
show that the resultant temperature distribution can be expressed as the sum of the
right-hand member of Equation (176) or (177) and the right-hand member of (161)
with TI = T 2 = O.
68. (a) Make use of the formal argument of Section 9.1J to show that the tempera-
ture distribution for which T(x, O) = O and for which T(O, t) = O and aT(I, t)/ax =
F(t) when t > O is again given by Equation (173) or (J 75) if A(x, 1) is the distribution
corresponding to F(t) = J.
(b) Show that here
A(x, t) = x - ~ ~I 2 sin n 7l: sin n 7l: x e- n 'n'«',/ 41'.
1'Iodd n 1l 2 2l
520 Solutions of Partial Differenlial Equations of Mathematical Physics
Section 9.12
69. (a) Show ¡hat ir u = Fe'~' is a solution of ¡he wave equation (5) in three dimen_
sions, where F is independent of the time 1, then F satisfies the Helmhollz equalion
2
'f¡2F + W
2
F = O,
c
where 'f¡2 is the ¡hree-dimensional Laplace operator.
(b) By seeking separable solutions of the preceding equation, show in particular
¡hat the real and imaginary parts of the functions
u = eiw(lx + my+nz~ct) e (12 + m2 +n2 = 1),
v = r- I , 2J _ ¡
- n+
2
1,)
(Wr)pm
e n
(COS m)ei(m9+w>l ,
'1"
w = eifn9+w(J=+CI), cJH~1 or 2 ) (m~r) (12 + m2 = 1)
are solutions of the wa ve equation in rectangular, spherical, and circular cylindrical

coordina tes, respectively. [Here P': is an associated Legendre function (see Section
4.12). The solutions given in Section 9.12 are special cases .]
70. (a) Verify ¡he identity
~ (r 2
2
_1
r 2 ar
i!..P.)
ar
=...!..r a (rrp).
ar 2
(b) Hence show that, if rp depends only upon distance r from the origin (in
spherical coordinates) and upon time 1, then the wave equation (5) can be written in
the form
(c) From Ihis resull obtain Ihe general solulion

rrp = f(r - el) + g(r + el).
71. Prove that ¡he standing WGveSCQS (wx/e) cos WI, cos (wx/e) sin WI, sin (wx/e) cos WI,
and sin (wx/e) sin WI each can be expressed as Ihe superposition of two waves travel-
ing with velocity e in opposite directions.
72. The equalion governing the lateral displacement w(x, 1) in small free vibrations of
a tighlly Slrelched string of uniform linear densily p, under constanl lension T, is of
the form
a2 W 1 a2 w
ax2 = e2 al"
where e' = TI p. (Se e Problem 8 of Chapter 5.)
(a) Ir the string is fixed at the ends x = O and x = 1, and if the string is initially
deflected in Ihe form w(x, O) = f(x) and released from rest, so that also aw(x, O)/ al
= O, obtain the ensuing deflection in the form
~ . n7tx n7tel
W ( x, I ) = ~ af'¡ SIn - , - cos - , - '
n-I
where Gn =
2
I Jor' f(x) . -1-
SIO
n7tx
dx.
Problems 521
(b) Show that the solution of pan (a) can be written in the form
w(x, 1) = HF(x - el) + F(x + et)],
~ . I1nX
where F()
x = ~ Gn Sin - / - '
n~ I
and hence where F(x) is an odd periodic function of x, of period 2/, which coincides
with f(x) when O < x < / and is defined for al! values of x. (Compare Problem 33 of
Chapter 8.)
73. When resistive forces proportional to the velocity are taken into account, the dif-
ferential equation of Problem 72 is replaced by the equation
2 a w_ a w
2
2
2
-'-? aw
e ax2 - al ' -y al'
where y is a constan!.
(a) Show that this equation admits solutions of the form
where k is an arbitrary nonzero constant, and that, if y 2 < k 2 e 2 , these solutions repre-
sent plane waves which are damped in time, and which move along the x axis with
velocity e' = e../1 - (y2/k 2 e 2).
(b) Show that the solution of Problem 72(a) here takes the form
w(x, t) = e-Y'
n""l
i: a n (cos CUnt + LOJ n
sin CUnt) sin I1n/x,
,
where
and where an = T 21' I1nX dx.

a f(x) sm -/- .
(Notice that the presence of resistive forces not only cauSes time damping, but also
decreases the frequency of each component oscillation, and that the higher frequencies
are no longer integral multiples of the fundamental frequency. When y> ne//, one or
more of the cu's is pure imaginary, and the corresponding time functions then become
real exponential functions.)
Section 9.13
74. When the problem solved in Section 9.13 is modified in such a way that fluid sur-
rounding the pulsating cylinder is contained in a fixed coaxial cylinder r = b (b > a),
on which the velocity V must vanish, show that the fluid velocity is given by
_ [ iw' H\l}(cub/e)H\21(CUr/e) - H\2)(CUb/e)H\l)(CUr/e)]
V - Va Re e H\')(cub/e)H\21(cua/e) _ H\21(cub/e(H\')(cua/e) ,
provided that the denominator does not vanish. (For the purpose of this problem, it
is not necessary to proceed beyond this poin!.)
75. The boundary of a sphere of radius a, surrounded by an ideal compressible fluid,
is caused to pulsa te radially with circular frequency cu, so that the radial velocity of
the boundary is given by V = Va cos CUt. By methods analogous to those of Section
9.13, determine the steady-state periodically varying velocity of surrounding points
522 Solutíons of Partíal Dílferential Equations of Mathematical Physks
in the form
2
V = ,Vo a /r2 2 2 [ (e2
C'" \ a w
+ arro2) cos ro (r e a - 1) + ero(r - a) sin ro (r e a - I) ~J'
where e is the velocity of sound in the fluid. [Use spherical coordioates with ¡he velocity
potential dependent only on r and 1, and notice that a suitable assumption for a com-
plex po¡ential rp, consists of the second terro of Equation (197), with e2 a eomp/ex
constant to be determined. The transient effects corresponding to the inilia/ion of the
pulsation are considered in Problem 99.]
76. Let rp(x, 1) satisfy the one-dimensional wave equation e 2 rpxx = rp" when O -< x -< a,
subject to the end conditions rp(O, 1) = cos rol and rp(a, 1) = O. Determine a particular
solution rp which varies periodically with time in the alternative forms
sin [roCa - xl/el
rp = sin (roa/e) cos rol
=?
1
. ( roa /e)
_ SIO
[ .
SIO ro ( I + a -e x) .' ro ( 1 - a -e
- sin X)] '
under the assumption that sin (roa/e) =;'= O, by each of the following procedures:
(a) Assume rp = f(x) cos rol + g(x) sin rol and determine f and g.
(b) Assume rp as a ·linear combination of those terms in Equations (189) and
(190) which reduce to cos rol when x = O.
[Notice that rp could represent a permissible lateral displacement of a string with
one end fixed and the other oscillating, or it could represent either velocity potential
or veloci¡y of an ideal compressible fluid in a tube, closed at X = a, and with an oscil-
latory valve at X = O. Notice also ¡hat a steady oscillatory response of the type assumed
cannot exist if ro is such that sin (roa/ e) = O, that is, if ro = n7r:e/ a where n is integral,
and that additional transient responses satisfying the prescribed end conditions may
exist. (See Problem 78.)]
77. When Problem 76 is modified in such a way Ihal the wave equation is to be satisfied
when O -< X < ca and the condition rp(a, 1) = O is replaced by the requirement that
Ihere be no inward-traveling waves as x ---- ca (and hence for all x > O), use results of
Section 9 . 12 to write down the steady-state solution. (The determination of transient
effects is considered in Problem 95.)
78. Let the initial conditions rp(x, O) = O and arp(x, O);al = O be added to the end
conditions of Problem 76.
(a) By using the result of Problem 76, show that the complete solution can be
assumed in the form
_ sin [roCa - xl/el _ ~ A . n7r:x nnel
rp - . ( wa /) cos rol .:...1 n SIO cos a (1) O)
sin e /1 - a
if ro =;'= kne/ a, where Ihe A's are lo be determined in such a way that rp(x, O) = O.
(b) Deduce that then
(ro =;'= n~e)
and hence obtain the solution in ¡he form
rp = sin [roCa - xl/el cos rol - 2 f; nn . nnx nnel
--a cos----¡¡-
sin (roa/e) n-l n 2n 2 (ro2a2/e2) SIO
when I > O, if ro =;'= kne/a (k = 1,2, ... ).

Problems 523
(If w ~ knc¡a, the kth term of the sum ean be extracted and combined witb the
particular solution, and the combination can be shown to tend to a function of x and
t wbicb involves t as a multiplicative factor, tbe remaining terms of tbe sum remaining
finite in the limil. This is a case of resonance. If small resistive forces were present in
the nonresonant ease, the portion of the solution represented by tbe series would be
damped out with increasing time. Since such forces always exist in praetice, the par-
ticular solution ean be considered as a quasi-steady-state solution, regardless of the
initial conditions, in the sense that it is the limit, as resistive forces tend to zero, of the
true steady-state solution of the problem in which resistance is present.)
Section 9.14
79. Obtain the solution rp(x, y) of Laplace's equation in the half-plane y >- 0, with
rp(x, O) = f(x), lim
x~+yL.... =
rp(x, y) = 0,
by use of the FOllrier transform, as follows:

(a) If ifJ(u, y) is the Fourier transform of rp(x, y) with respect to x, so that
ifJ(u, y) = f:~ e-'uxrp(x, y) dx,
show that the transform of aZrp¡axZ is -u 2ifJ and the transform of a'rp¡a y2 is aZifJ¡a y'
(assuming in the former case that both rp and arp¡ax tend to zero as Ixl ~ co and in
the lalter case that differentiation with respect to y can be effected under tbe integral
sign). Hence show tbat the result of taking tbe Fourier transform of the equal members
of the equation rpxx + rpyy = ° is the equation
~:~ -
2
1I ifJ = 0,
from whicb there follows

ifJ(u, y) = A(u)e- UY + B(u)e UY •
(b) By requiring that ifJ(u, O) = ](u) and using tbe faet that the Fourier transform
cannot be unbounded as y ~ ce, deduce that
ifJ(u, y) = ](u)e- luIY .
(e) Show tbat e- Iuly is the Fourier transform of the funetion
g(x, y) = ~ x, ::. y2
and use the convolution property (Problem 91 of Chapter 5) to deduce that
rp(x, y) = r~ g(x - ,;, y)f(,;) d,;
1 J~ yf(,;) d,;
= n _~ (x - ';)2 + yZ'
in accordance with (230). (Se e also Problem 81.)

80. Obtain the solution of tbe equation
(O -< x < co, ° -< t < co),
524 Solutions of Partial Differential Equations of lVIathematical Physics

T(x, O) = f(x), T(O, t) = O,
by use of the Fourier sine transform, as follows:
(a) If Ts(u, t) is the sine transform of T(x, t) with respect to x, so that
Ts(u, t) = S: T(x, 1) sin ux dx,
show that the transform of a Z T/ax 2 is uT(O, 1) - u 2 T s , assuming that T and aT/ax
tend to zero as x ~ co, and the transform of aT/at is aTs/at. Hence, noting that here
T(O, t) = O, deduce that T s must satisfy the equation
aTs + (l., 2 u2 Ts = O
at
and the condition
Ts(u, O) = fs(u),
so that
Ts(u, 1) = fs(u)e--'·".
(b) Use (241) to show that e-·'·" is the cosine transform of the function
g(x, t) = Jn;1 e- x '/4_"
and use the convolution property of Problem 92(c), Chapter 5, to deduce the solution
as given by Equation (242).
81. Rederive the solution of the problem
rpxx + rpyy = O (-co < x < co, O -< Y < co),
rp(x, O) = f(x), lim rp(x, y) = O
(see Problem 79) by the following method:

(a) If rps(x, v) denotes tbe Fourier sine transform of rp(x, y) with respect to y, so
that
rps(x, v) = S: rp(x, y) sin vy dy,
show that the transform of a 2rp/ax2 is a 2 rps/ax2 and the transform of a 2rp/a y2 is
vrp(x, O) - v 2 rps =' vf(x) - v 2 rps, assuming that both rp and arp/a y tend 10 zero as
y ~ co, so that rp s must satisfy the equation
a 2 rps
ax2 - v 2rps = -vf(x).
(b) In order 10 avoid the solution of this differential equation, take the Fourier
transform of the equal members. Thus show that the "mixed" double transform ips(u, v)
must satisfy the equation
-u 2 ips(u, v) - v 2ips(u, v) = -vj(u),
and hence is determined in the form
v -
ips(u, v) = u 2 + V 2f (U).
Problems 525
(e) Show that v/(u 2 + v2) is the sine transform, with respeet to y, of the funetion
g(u, y) = e-I"I>
[Se e Problem 86(e) of Chapter 5J so that

qi(u, y) = e-lul>j(u),
in aeeordanee with the result of Problem 79(b), after which the required solution rp(x, y)
is obtained as before.
[This problem is intended to illustrate procedures whieh are available in situa-
tions where a simpler approaeh, sueh as that of Problem 79, eannot be followed. As
Problems 80 and 81 illustrate, it may be eonvenient to take the Fourier sine transform
with respeet to a variable in a direetion normal to a line segment along whieh the un-
known functiol1 is preseribed. The cosine transform would be similarly appropriate if
the normal derivative of the function were preseribed instead on that segment. Notiee
also that the result of introdueing a "foreiog term" h(x, y) into the governing differential
equation, so that it beeomes Poisson's equation rpxx + rpyy = h, can be tTeated in the
same way.J
82. (a) If rp(x, y) satisfies Laplace's equation in the infinite strip O -< y -< b, and if
rp(x, O) = f(x), rp(x, b) = O, and rp is bounded as Ixl ---> ca, obtain rp io the form
rp(x, y) = ~ r [r~ sin~i~~ uh y) f(l;) eos u(1; - x) dI; ] du,
wheo f(x) is suffieiently respeetable.

(b) By making use of the formula
f =
o stnhqx
sinh px
. cos rx
dx
= -
n
2q eos (pn/q)
sin (pn/q)
+ eosh (rn/q) I
and interehanging the order of iotegratioo in the resuIt of part (a), express the solution
io the form
rp(x y) -
, -
J... sin ny
2b b
f=_= eosh [n(1; -
fel;)
x)/bj - eos (ny/b)
dI;
.
(When f is a polynomial, the method used in Problems 21 aod 22 is also available,

yielding a series solution. Here the boundedoess eondition must be modified.)
83. (a) Tf rp(x, y) satisfies Laplaee's equatioo io the quadrant x >- O, y >- O, and ir rp
vanishes along the positive y axis, reduces to f(x) along the positive x axis, aod is
bounded as X2 + y2 ---> ca, obtain rp in the form
rp(x, y) = -2
n
f=o f=o e-u>f(l;) sin ux sin ul; dI; du,
when f(x) is suffieiently respeetable.
(b) By formally integrating first with respeet to u, transform this expression to
the form
rp(x, y) = ~ 1= f(l;{y2 + (~ _ X)2 - y2 + (~ + X)2] dI;.
(e) Deduce the Tesult of part (b) from Equation (230) by replacing f(x) in (230)
by an odd funetion .fo (x) sueh that fo(x) = f(x) when x > O and fo(x) = -fe -x)
when x < O.
526 SolutioDS of Partíal Differential Equations of Mathematical Physics
84. (a) If ~roblem 83 is modified in sueh a way that arp/ax rather than rp vanishes
along the positive y axis, show that
rp(x, y) = n2 Jor- Jor~ e-UYf((,) eos ux eos u(' d(, du
= ! r f((,{y2 + (~ - x)2
(b) Deduce the result of part (a) fram Equation (230) by replacing f(x) in (230)
+ y2 + (~ + X)2] d(,.
by an even funetionf.(x) such that .f.,(x) = f(x) when x > O and .f.,(x) = f( -x) when
x < O.
85. A rod with insulated sides extends from x = -= to x = +=. If the initial tem-
perature distribution is given by T(x, O) = f(x), where -= < x < =, show that
T(x, r) = ! r~ f((,{l~ e-u'." eos u((, - x) du ] d(,
= 1 f~ f((,)e-(~-xl'/4"" d(,.
2a-vnr -~
86. A rad of infinite length, with insulated sides, has its end x = O insulared. If the
initial temperature distribution is given by T(x, O) = f(x), where O < x < =, show
that
=
Jor~ f((,)[e-«-xl'/4." + e-(U l'14.',] d(,.
T(x 1) 1_ x
, 2a..¡nl
87. Source funcrions.

(a) If f(x) is the unit singularity funetion O(x) (so that a heat souree of intensity
ps is present at x = O at the instant r = O), show that the solution of Problem 85 is
of the forrn
T(x 1)
,
= 1
2a-vnr
e- x '/"." = S(x r).
'
(b) Deduce that the temperature in a rod extendiog over (-=, =), due to a unir
hear source at x = Xo at the instant r = ro, is given by
T(x, r) = l..S(x - xo, r - ro).

ps
(e) Show that Equation (242) takes the form
T(x, 1) = r [S(x - (,,1) - S(x + (" r)lf((,) d("
the solution of Problem 86 takes the form
T(x,/) = r [S(x - (,,1) + S('" + (" I)]f((,) d("
and the solution of Problem 85 beco mes
T(x,/) = r- S(x - (" I)f((,) d(,.
[lo rwo dimensions, the eorresponding source funcrion is

Problems 527
(se e Problem 92), whereas the function
S( t) 1 e-(x1+y~+z~)/4a: ~ (
x, y, z, = (4a 2 nt)'lz
is the /hree-dimensional generalization.]

88. The end x = O of a rod with insulated sides is maintained at the temperature
T(O, t) = F(t) for all / > o. Initially, all points of the rod are at zero temperature,
T(x, O) = O, where O < x < oo.
(a) Show that the formal argument of Section 9.11 leads to the solution
T(x , /) = L F(,) aA(x'a// - ,) d"
where A(x, t) is the solution of the problem when F(t) = l.

(b) Use the result of Equation (244) to deduce that
A(x, t) = 1 - erf(2a:"t) = 1 - Jn i X
" . " ' ;- e-"' duo
(c) From the eesults of parts (a) and (b), obtain the solution of the stated problem
in the form
T(x t) =
,
x
2a...jn Jor F(,)e - x'/h'('-<) (t,.-;--=d:..:':v""
,)'/ 2
89. (a) If F(t) = Ó(t) in Problem 88, show that
T(x t) = x e- x ' / ' . "

, 2at...jnt '
and that, if this function is denoted by D(x, t), then the solution of Peoblem 88 takes
the form
T(x, t) = J: D(x, t - ,)F(,) d,.
(b) Verify that
D(x, t) = -2a' aS(x, t),
ax
where S(x, t) is defined in Problem 87 and is the solution of Problem 85 which coe-
responds to the initial function Ó(x). [Hence D/2a 2 is the solution of Problem 85 cor-
eesponding to the initial unit doublet lunc/ion -Ó'(x) at x = O at the instant / = O.]
90. (a) Verify tbat D(x, t), as defined in Problem 89, does indeed satisfy the heat-flow
equation when x > O and / > O, and that
lim D(x, t) = O (t > O), lim D(x, t) = O (x > O).
_,,-.0 + t-O+
(b) Deduce that il rp = U(x, /) is a solution 01 /he problem
(O < x < ca, O < / < ca),
lim rp(x, t) = F(t) (t > O), lim rp(x, t) = I(x) (x > O),
x-,O + r - 'O+
/hen to U(x, t) can be added any cons/an/ mul/iple 01 D(x, /), so that the problem as
stated does not have a unique solution.
528 Solutions of Partial Differential Equations of lVlathematical Physics
(e) Show that if x ~ 0-:- and I ~ 0+ in sueh a way that x2 = kl, where k is a
positive constant, [he funetion D(x, 1) becomes infinite. [Thus the added speeification
that cp be bounded when O <:: x <:: A and O <:: I <:: B, for all positive values 01' A and B,
will eliminate the situation deseribed in part (b). This i!lustrates the need for the
boundedness assumption made in Section 9.1.]
91. At the time I = O, the temperature in unbounded spaee is dependent only upon
radial distance r from the z axis, and is prescribed as T(r, O) = f(r). The ensuing
lemperature distribution T(r,l) is required, under the assumption that f(r) behaves
satisfactorily for large values of r.
(a) Show that T must satisfy the equation
when I
+:r (r ~~)= ~2 ~~
> O, and obtain a product solution whieh remalns finite at r = O and as
r ~ co in the form
Ae-u'a"Jo(ur).
(b) Deduce that the required distribution is given formally by the expression
T(r, 1) = 50= uA(u)e-u'a"Jo(ur) du,

where the factor u is inserted for convenience in the following step, and where A(u)
is to be determined such that
f(r) = 50= uA(u)Jo(ur) du (O < r < col,

and hence, by referring to Equations (249a, b) of Section 5.15, obtain the desired solu-
lion formally as
T(r, 1) = t~ ue- u""J o(lIr) [r ¿;f(¿;)Jo(u¿;) d¿; ] duo
(e) By assuming the validity 01' interchange of order of integration, and making
use of the relation
lo
~ ue-a'u'J (bu)J (eu) du
n n
= _1_e-(b'+"'i 4 a'l (be),
2a2 n 2a 2
express the formal solution of part (e) in the form
(1 > O).
92. (a) Suppose that the initial temperature in unbounded spaee is zero exeept through-
out an infinitely extended right circular eylinder of radius é, with its axis eoineiding
with the z axis, and has the eonstant value T o inside that eylinder. Show that the exeess
heat required per unit length of eylinder is given by
2n S: psTor dr = npsToé 2 ,
where ps is the heat eapaeity per unit volume. Henee deduce that, if unil exeess heat
is 10 be present per unit length, there must follow T o = 1/(npsé 2 ), and use Problem 91
to show that the eorresponding resultant temperature distribution in spaee at all
problems 529
folJowing times IS glVen by
T( r, t) =
e-r~/4cr.~(
2cPt
1
. 2nps'
[21'
e"
e2 o ):e-~',"a"/o 2":2t
~., d): ():)] (t > O).
(b) By applying L'Hospital's rule (or otherwise) and recalling that 10(0) = 1,
show that the quantity in brackets tends to unity as f - O, and deduce that the tem-
perature distribution due to an instantaneous line so urce along the infinite line r = O,
emitting one calorie of heat per unit length at the instant t = O, is given by
(t> O)
in unbounded space.
(c) Deduce that if the line source is instead perpendicular to the xy plane at the
point (xo, Yo), and if the impulse takes place at the time to, then the temperature at any
point in unbounded space at any following time is given by
T = 1 e-[[x-xo)'+(Y-Yo)'l/l4a'('-'o)] (t > to).
, 4npsct 2(t to)
(See also Problem 87 for the analogs in one and three dimensions,)
Section 9.15
93. Use the Laplace transform to solve the problem
az + az _
ax at - Z,
z(O, t) = 1, z(x, O) = 1,
when O <: x < ca and O <: t < ca.
94. Use the Laplace transform to solve tbe problem
az + az = z
ax at '
z(x, O) = 1,
when -ca < x < ca and O <: t < ca,
95. Use the Laplace transform to solve the problemt
a2rp _ 1 iJ.!..rl
ax2 - e 2 at 2 '
rp(x, O) = O, rp,(x, O) = O, rp(O, t) = cos 0t,
when O <: x < ca and O <: t < ca, and verify that as t - ca the solution tends to that
obtained in Problem 77.
96. Use the Laplace transforrn to solve the problemt
a 2rp a 2rp
ax2 - at2 = 1,
rp(x, O) = 1, rp,(x, O) = 1, rp(O, t) = 1,
when O <: x < ca and O <: t < ca.
tUse the fact thal C(s)e'x " cannot be a Laplace transform as x _ 00 unless C(s) = o.
530 Solutions of Partial Differential Equations of l\<Iathematical Pbysics
97. (a) Derive the formula
oC[f(x + IXI)} = ~
1 [,gOx
~
e-'(U-X)I~!(u) du,
when IX *O, where ca sgn IX = + ca when IX > O and - 00 when IX < O. (Replace
x + IX/ by u in the definition of the left-hand member.)
(b) Deduce that , if e > O, there follows
oC[f(x - el)} = e1 IX_~ e'(U-X» )C!(U) du

and oC[f(x + el)} = e1 f~
x e-"U-,Y)/C!(II) dil.
(e) Also deduce Chat

i:-(eP'!(x + IXI)} = 1
~ f~"o~
x e- (, -P)(U-X) /d!(u) du
when IX * O. [Use formula (T8), page 67.]

98. U se the Laplace trans form and the results of Problem 97 to solve the problem
az + az = z
ax al '
z(x, O) = rp(x),
when -00 < x < 00 and O < / < oo. [Show that
;;(x , s) = J:= e" -J )(u -x)rp(u) dll + A(s)e-( , -J)x,
where the lower limit of the integral is so ehosen that the eoeffieient of s is never posi-
tive in the exponential function oeeurring in the integrand.]
99. In Problem 75 suppose that until the instant t = O the sphere and surrounding
medium are at rest in ' complete equilibrium, and that at that instant the pulsation
begins and eontinues indefinitely. Determine the veloeity of surrounding points by
using the Laplace transform as indicated below.
(a) Make use of the result of Problem 70 [see also Equations (203) to (206), See-
tion 6.20] to obtain the relevant relations involving the velocity potential rp and radial
veloeity V in the form
a (rrp)
2
= .!.. a2 (rrp).
ar 2 c2 al 2 '
r = a: V = i!.P. = Va 00: ~finite'
ar COS COI; r -
ar '
1=0: rp = ~ = o.
al
(b) By ealeulating the Laplaee transforms of these relations, obtain the equations
a2 (dp)
2 _ S2 (rrp-) = O. V = aip.
ar c2 ' ar '
r = a: r -- co: V finite.
problems 531
(e) Show [hat rip must be of the form Ae-"ic and henee also
V - = -- A(· s + -e)
er r
e-sr/c.
(d) Determine A, as a funetion of s, so that the eondition at r = a is satisfied.

Thus oblain the result
where
(e) Expand ](s) in partial fraetions and use pairs T9, 12, 15, and 16 of Table 1,
Chapter 2, to determine the inverse transform of V. The solution is of the form
v=O when r - a > et,
v = e 2v-::~~~2 [(e 2 + arco 2) eos co (t - r e a) - eco(r - a) sin co (t - r e a)
when r - a < et.
(Notiee that the disturbance travels radially outward with the speed e and at time t
is present only at distanees less than e/ from the sphere surfaee. As t --> ca aH the
surrounding fluid beeomes disturbed, the exponential term tends to zero, and the
solution approaehes that of Problem 75.)
100. The eonvohllion and Duhamel superposi/ion. Suppose that, for a certain problem,
the Laplaee transform of the required solution rp(x, t) is determined in the form
ip(x, s) = P(s,)G(x, s) + P(x, s),
where P, G, and ¡; are the transforms of identifiable funetions F(t), G(x, /), andP(x, t),
respeetively.
(a) Show that then
rp(x, t) = J: G(x, / - -r:)F(-r:) d-r: + P(x, t).
(b) Show that if

1 - -
--.sG(x, s) = A(x, s)
is the Laplaee transform of A(x, t), and if F(/) is differentiable, theo also
rp(x, t) = F(O)A(x, /) + J: A(x, / - -r:)F'(-r:) d-r: + P(x, /).
[Notice that sP(s) = oC[F'(t)} + F(O).]

(e) Show that G(x, 1) is the solution of the relevant problem whenP(x, /) = O and
F(T) is the unit impulse fune/ion ("delta funetion") "(/), whereas A(x, t) is the solution
when P(x, /) = O and F(/) is a unit s/ep fune/ion sueh that F(/) = O wheo / < O and
F(t) = 1 when t > O. [The fuoetion A(x, /) is sometimes ealled the indicial admittanee.]
532 Solutions of Parlial Differential Equations of Mathematical Physics
101. 1f 7(x, t) is the solution of the p roblem specified by Equations (165), (166), and
(167), show that
- - sinh(s'/2 x /a)
7(x, s) = F( s) sinh (s"21/a) .
Hence deduce from Problem 100 that the function A(x, t) defined by (168) must be the
inverse of the transform
- 1 sinh (sli2 x /a)
A (x, s) = s
sinh (s 1 ' 21/a) ,
and also rederive (173) and (175). (See also Problem 108, below, and Problem 70f
Chapter 11.)
102. lf rp(x, t) is the solution of the problem
a
-
2rp 1 2rp
= - -- ,
a
ax2 e' at'
rp(O, t) = 0, rp(l, t) = f(t), rp(x, O) = 0, rp,(x, O) = 0,
when °< x < 1 and °< t < =, show that
- - sinh (sx/c)
rp(x, s) = fes) sinh (si/e)
and deduce that if
A(
x, I
) = 012-1 {J..s sinh (sx/c)l
sinh (si/e) J
(se e Problem 106, below, and Problem 6 of Chapter 11), there follows
rp(x, 1) = I aA(x'a~ - T) f(T) dT
and also
rp(x, 1) = f(O)o4(x, t) + s: A(x, I - T)/,(T) dT,
if f(1) is differentiable. (Thus the Duhamel principIe of Section 9.11 also applies here.)
Seclion 9.16
103. (a) When f(l) = J, show that the voltage given by Equation (285), for any fixed
value of x, is given by v = A(x, 1), where
0, ° <1< x /e ,
1, x/e < t < (21 - xl/e,
A(x, t) = 2, (21 - x)/e < t < (21 + xl/e,
1, (21 + x)/e < I < (41 - xl/e,
0, (41 - x)/e < I < 4lje,
°
when < I < 41/e, and A is periodic, with period 41/e = 47.
(b) At the midpoint, x = 1/2, show that v = °
for 7/2 seconds, then v = J for
7 seconds, Ihen v = 2 for T seconds, then v = 1 for 7 seeonds, and then v = for °
the remaining 7 / 2 seconds of the period 4 T.
(e) At Ihe end x = 1, show that v = °
for 7 seeonds, then v = 2 for 27seeonds,
and then v = O for the remaining T seeonds of the period 4T.
Problems 533
(If resislance effeets were taken into aceount, the amplitude of the voltage oseil-
lation about the mean value 'V = 1 would damp out with inereasing time.)
104. With the notation of Problem 103, show that the solution (285) can be expressed
in the alternative form
v(x, t) = f(O)A(x, 1) + s: A(x, I - 1:)/'(1:) d1:.
Also, by making use of this result, present v(x , 1) graphieally as a funetion of I for
°< t < 41jc when f(t) = 1 and when f(t) = vol. [Notice that the equivalenee
s: A(x, I - 1:)/'(1:) d1: = S: A(x, 1:)/'(1 - 1:) d1:
is useful here.]
lOS. (a) If the end of the line considered in Section 9 . 16 is grounded, so that v(l, 1) = 0,
show that
_ - sinh [s(l - x)jc] - e-sx; e - e- s (21-x)! e
v(x , s) = fes) . h
sin s eIj = fes) 1 - e ' s/fe
= j(s)[e-Sx1c _ e- s {21 -xl!c + e- s<2f+x) ! c _ e - s( 4t-x )/ c + ... ].
(Notiee that refleeted waves are reversed at each end of the line.)
(b) When f(l) = 1, show that, for any fixed value of x, this expression takes a
form similar to that given in Problem 103(a), with the suecessive values 0, 1,0,1,
°
replaeing the values 0, 1,2, 1, given in that expression, and that, in faet, the voltage
°
here is of period 2T.
106. Show that the funetion A(x, t) in Problem 102 is the solution of Problem 105(b)
with x replaced by 1 - x. AIso use this faet in presenting v(x, 1) graphieaIly as a fune-
°
tion of I for 0 and for
O -< x < 00, subjeet 10 the initial eondition T(x, O) = 1 and to che end eondition
°
T(O, 1) = O.
(a) Show that the Laplace transform T(x, s) then must satisfy the equation
° °
rxzTxx - sT = -1 and the end condition T(O, s) = and , making use of the fact that
T must be bounded as x ~ +00 for s > [or of the more relevant (but less evident)
fact that C(s) exp (S' /'x jrx) cannot be a Laplace transform as x - +co unless it
vanishes identically], deduce that
T(x, s) = J..(1 - e-s"'x.' e).
s
(b) By noticing that the solution of this problem is in fact given by T(x, 1) =
erf [x j(2rx,,/¡)] , according to Equation (244), deduce that
+(1 - e-
qx
) = oC {erf 2a~ I }
and 1..e-qX = oC{1 _ erf x },

s 2a"j I
with the abbreviation
S1l2
q=---a.
(A direcl derivation of these results is given in Problem 8 of Chapter 11.)
108. Let T(x;t) satisfy the heat-flow eguation rXZT.u - T, =

subjeet to the initial eondition T(x, O) =
for t > and -< x::S!
°
and to the end eonditions T(O, 1) = and
° ° ° °
TU, 1) = 1 when 1 > 0, so that T = A(x, 1), where A(x, t) is the funetion of whieh the
series (168) is an expansiono
(a) Show that the Laplaee transform T(x, s) must satisfy the eguation
rXZT xx - sT = °
and the end eonditions TeO, s) = 0, T(I, s) = l/s, and deduce that
= -
1 sinh qx =
1
_e- qU - x )
1 - e- 2qx ,
T(x, s)
s sinh q! s 1 - e 2q/
where q = sl/2/1X.
(b) By expanding 1/(1 - e- 2q /) in a series of aseending powers of e- 2q / (when
q > O), express this transform in the form
-
T(x s) =
's
I
_e-qU-;d -
1
-e-q(/+x)
s
+ 1
_ e - qOJ - x ) -
s
.. '.
(e) By evaluating inverse transforms term by term (as can be justified here), and
using the results of Problem \07, deduce that T = A(x, t), where
A(x 1) =
,
(1 - erf 21X.v
I - ::) _ (1 _ erf I + X) + (1 _ erf 31 - X) _ ....
12a.v 21X.v 1 1
[Notice that the series form (168) converges rapidly when 1X.v1/1 is large. Sinee
erf u approaehes unity very rapidly as u --+ +co, the terms in parentheses in the form
of part (e) are small when a.vr/I is small. Henee the latter form is partieularlyeon-
venient when the former one is no!. The abbreviation
erfe X = 1 - erf X = .v2 f~

x e-u' du,
n
defining the so-ealled complementary error function, is a eonventional one.]
Section 9.17
109. Obtain the differential eguation (295) by the alternative method suggested in the
tex!.
110. (a) If C~'(t) + ¡.li;C.(t) = hn(t), with ¡.l. 7"= 0, show that
C.(t) = ~ ]' sin ¡.ln(t - r)hn(r) dr + An eos ¡.lnt + Bn sin ¡.lnt,

J.ln lo
where A. and B. are arbitrary eonstants. (See Section 1.9.)

(b) If the eonditions Cn(O) = a., C~(O) = b. are preseribed, show that A. = a.
and B. = b n/ ¡.l•.
(Notice that ¡.ln may be replaeed by i ¡.l•. )
111. For the problem
a'-qJ a2 rp
ax2 + a y2 = h(x, y),
rp(O, y) = 0, rp(l, y) = O, rp(x, O) = f(x),

subjeet to the requirement that rp(x, y) be bounded as y --+ ca, obtain the solutioll
Problems 535
(when it exists ) in the form

,." . I11lx
q;(x, y) = n~' c.(y) Sin - , - '
n!n 2
where C ' (y) - [ 2 Cn(y) = hn(y)
with Cn(O) = a n,
where Cn(y) also is to be bounded as y ~ 00 and where
h.(y) = I2 J'"o h(x , y) Sin

. -mtx
, - dx, an = -,
2
Jor' f(x) . nttx dx.
SIn - , -
AIso determine Cn(y) explicitly when h(x, y) = l andf(x) = l. [If h(x, y) is unbounded
as y ~ 00, but does not grow exponentially, then the boundedness requirement on
q; and on C n generally must be replaced by a restriction against exponential growth, if
a s()lution is to exist.]
a 2 q; aZq;
ax2 + a y2 = h(x, y),
q;(O, y) = G(y), q;(i, y) = F(y) , q;(x, O) = f(x),

with q;(x, y) bounded as y ~ 00, obtain the solution (when it exists) in the form
q;(x, y) = l/(x , y) + v(x , y),
where u(x, y) = ~ F ( y) + (1 - ~ )G( y)
and where v (x, y) is the solution of Problem 111 when f(x) and h(x, y) are replaced by
f"(x) and h*(x, y), with
f*(x) = f(x) - -; F(O) - (1 - ~) GCO) ,

h"(x, y) = h(x, y) - ~ F"(y) - (1 - -; ) G " (y),
when F " and G" exist. A1so determine the solution explicitly when h(x, y) = 1,
f(x) = O, and F(y) = G(y) = l.
113. Suppose that Cn(y) is to satisfy the equation
C'(y) - fJ.?;Cn(y) = hn(y) (fJ.n > O),
the condition C(O) = a n, and the requirement tha! Cn(y) be bounded as y ---+ 00 (when
hn is sufficiently respectable).
(a) Show that
for sorne choice of the constants An and Bn. (Compare Problem 110).
(b) For large positive values of y, show that
536 SolutíODS oC Partíal Dílferential Equations oC Malhematical Physics
and henee deduee that
A• = 1
- 2f1,. 1-
o e-~""h.(11) d11,
(e) Show that the result of inrrodueing these values inro the express ion for C.(y)
ean be written in the eompaet form
C.(y) = t- G(y, 11)h.(11) d11 + a.e-~·',

where G(y, 11) is the Green's funetion
(y <: 11),
(11 <: y).
Also use this result lo redetermine C.(y) In Problem 111 when h(x, y) = 1 and
f(x) = l.
a 2rp a'rp
ax2 + ay2 = h(x, y),
rp(x, O) = f(x),
in the half-plane -co < x < co, O <: y < ca, with rp(x, y) required to be bounded as
y - + co, obtain the solution in the form
!p(x, y) = t~ [C(u, y) eos ux + D(u, y) sin ux] du,

where
Cn - u2C = n1 5°0__ h(x, y) cos ux dx, D" - u 2D = 1<

1 I-
__ h(x, y) sin ux dx,
with
C(u, O) = 1< J-
1 __ f(x) eos l/X dx, D(l/, O) =1<
1 J-
__ f(x) sin l/X dx,
and with C and D lo be bounded as y - + oo. [Here, in addition to assuming that h

behaves appropriately as y -+ co, we must assume f(x) to be sueh Ihal the indieated
integrals exis!.]
a,rp 1 arp arp
-a
r
2 + --a
r r
= -a
t
+ h(r,t),
!p(0, t) = O, !p(r, O) = f(r),

when O <: r <: a, O <: t < co, obtain the solution in the form
rp(r, t) = i:
n= I
C.(t)Jo (IXn~)'
a
where Jo(IX.) = O, with Cil) detennined by the differential equation
C~(t) + ~~ e.(l) = a2[JI~IX.))2 J: rh(r, t)J o (IX. ~) dr

Problems 537
and the condilion
116. For lhe problem

a 2 rp , a 2 rp _ arp
ax 2 '- ay" - ar'
rp(x, o, t) = f(x, t),
rp(x, n, t) = rp(O, y, 1) = rp(n, y, 1) = o,
rp(x, y, O) = O,
when O < x < n, O < y < n, O < l < ca, obtain the solution in the form
'0 ~
rp(x, y, t) = 2: 2: Cmn(l) sin mx sin ny,

m==l n==l
where
C;"n(l) + (m" + n 2 )Cmn (t) = 4~

n"
1 n
o
f(x, 1) sin mx dx,
with
Seclion 9.19
117. In a one-dimensional irrotational flow of an ideal compressible fluid in the posi-
tive ., direction. in which [he velocity V x depends upon position x and time t but
differs by only a small amount from a constant U, the velocity V x can be expressed
(approximately) in the form
where rp satisfies the equation

2 2 2
(U' _ VD a rp + 2U a rp + a rp = o.
ax2 ax at al Z
Here V s is an effective mean value of the sonic velocity in the fluid. The density p is
then expressible (approximately) in the form
p = Po[l - _1
V}
(uarp + arp)J,
ax al
where po is the constant value associated with a uniform flow. (See Section 6.20).
(a) Obtain the general solution for rp(x, 1) in the form
rp(x,l) = f[x - (Vs + U)t] + g[x + (Vs - U)t],
wherefand g are arbitrary twice-differentiable functions.
(b) Show that V x and p then are defined by the equations
V" = U + F[x - (Vs + U)t] + G[x + (Vs - U)I],
P = Po[i + ts(F[X - (Vs + U)¡] - G[x + (Vs - U)lll}

where F = j ' and G = g:.
538 Solutions of Partial Dilferential Equations of iVlathematical Physics
118. (a) Show that the funetions Vx and P in Problem 117 are determined al the
point P with eoordinate x = x, at any time 1 = I[ by inilial values of V, and P (at
the time 1 = O) at the two points Q, and Q2 with eoordinates
x=x,-(Vs+U)I, and x=x,+(Vs-U)II'
(b) Show that Q, and Q2 are both upstream from P when the How is supersonic
(U> V s ), whereas they are on opposite sides of P when the flow is subsonic.
(e) Sketch typieal pairs of "eharaeteristie Jines" x - (Vs + U)I = eonstant and
x - (Vs - U)I = eonstant in the upper half (1) O) of a fietitious xl plane (compare
with Figure 9.17) in both the supersonie and subsonie cases, and show that the values
of Vx and P at any point P(XI, 1 1 ) of this fietitious plane are determined by values of
V x and p at the points where the two eharaeteristie lines passing through P interseet
the x axis.
119. (a) If the density is preseribed as p = [1 + 17(x)]po when 1 = O in Problem 117,
where Po is a eonstant and where I 17(x) I <K 1, and where 17(x) ~ O as x ~ -00, show
that there must follow
Vx = U + C[x - (Vs -+ U)I] + C[x + (Vs - U)I] + V s17[x - (Vs + U)I],
.E. = 1 + vI [C[x - (Vs + U)I] - C[x + (Vs - U)I]] + 17[x - (Vs + U)I],
Po s
where C is an arbitrary funetion.
(b) In addition, let it be preseribed that the veloeity must tend to the uniform
veloeity U at an infinite distanee upstream (as x ~ -(0) for all positive values of l.
Show that, at any positive time 1, the veloeity at any point for whieh x + (Vs - U)I
= e, where e is a positive eonstant, is given by
V, = U + C(c - 2 Vsl) -i- Cee) + V s 17(e - 2 Vsl).

Notieing that x ~ - co as 1 ~ + co, when e is held fixed, if Ihe flow is subsonic
(U < V s ), show that there then must follow C(e) = -C(-oo) for all positive values
of e, and deduce that the funetion C must be zero. Henee show that, when the flow is
subsonie, the preseribed eonditions uniquely determine V, and p in the forms
p = Po(l + 17[x - (VS + U)llJ
when 1> O. [Notiee that here we have made 1 ~ +co along a eharaeteristie line
x + (Vs - U)I = e in the xl plane, and that this line interseets the ¡ine on whieh
x = Xo - > - co in the upper half-plane 1 > O if and only if U < Vs .]
(e) Show that the same result would follow in the supersonie case if it were pre-
seribed that V x - > U as x - > - co for all negalive values of 1, or that V x - > U as
x - > +00 for all posilive time, if 17(x) - > O as x - > +00.
120. Determine V x and p in Problem 117, subjeet to the eonditions that p = po and
V x = [1 + f(X)] U when 1 = O, where Po is a eonstant and where I f(X) I <K l.
10
Functions of a COIllplex Variable
10.1. Introduction. The Complex Variable. A comp!ex number IX is an

expression of the form IX = a + ib, where a and b are re a! numbers and i is
the imaginary unit, satisfying the equation
i2 = - 1. (1)
We follow convention by speaking of a as the real part of IX and of b as the
imaginary part of IX, and write
lX=a+ib: a = Re (IX), b = 1m (IX), (2)
noting that accordingly the "imaginary part" of a complex number is in fact
real. A complex number a + ib is said to be real when b = O and to be imaginary
when b 0;1= O. In particular, when b 0;1= O and a = O the number is said to be
pure imaginary. In the same way we may define a complex variable z = x iy, +
where x and y are real variables.
Geometrically, jt is convenient to represent
a complex quantity x + iy by the point (x, y) in y
a rectangular coordinate system known as the

complex plane (Figure lO.!). Thus the points
representing real numbers are aIllocated on the
x axis in this plane, whereas those representing x 7Z=X+iY
the pure imaginary numbers are located on the y r y
axis. For some purposes it is convenient to think
of the complex number z as being represented by
a vector in the complex plane from the origin x
(x = y = O) to the point (x, y). We speak of Figure 10.1
539
540 Functions of a Complex Variable
the length of this vector as the absolute va!ue of Z, o r as the modu!us of z, and
denote (his number by 1:: 1,
Izl = Ix + iyl = ,.Jx 2 + y 2. (3)
The number x - iy is called the conjugate of the number z = x + iy and is
denoted by the symbol i,
i=x-~. ~
We say that two complex numbers are equa! if and only if their real and
imaginary parts are respectively equal; that is,
x, + iy, = x2 + iY2 implies x, = x 2 , y, = Y2'
In particular, a complex number is zero if and only if its real and imaginary
parts are both 2ero.
Addition, subtraction, multiplication, and division of complex numbers are
defined as being accomplished according to the rules governing real numbers
when one writes i 2 = -1 , in accordance with Equation (1). (See also Problem
l.) Thus, if
Z, = x, + iy"
there follows
Z, + Z2 = (x, + x 2) + i(y, + y,), (5a)
z, - Z2 = (x, - x 2) + i(y, - Y 2)' (5b)
and
Z,Z2 = (x, +
iy ,)(x2 + iY2) = (x,x 2 - y,y,) + i(X'Y2 + x 2y,)· (6)
In particular, we notice tha t
zZ = (x + iy)(x - iy) = X2 - Py2 = X2 + y2 = Izl 2 = lil 2 , (7)
That is, the product 01 a comp!ex number and its conjugate is a nonnegative real
number equa! to the square 01 the abso!ute va!ue 01 the comp!ex number. We then
use this fact to ded uce that
Z2 X2+iY2 (x 2 + iy.)(x, - iy,) _ (x,x2 + y'Y2) + i(X'Y2 - XzY ,),
2, = x, +
iy, xf + y¡ - xi + y¡ XI + YI
(8)
when z, *
O.
If we introduce polar coordinates (r, ()), such thatt
x = r cos (), y = r sin () (r >- O),
the complex number z can be written in tbe polar lorm
z = x + iy = r(cos () + i sin ()), (9)
where r is the modulus of z. It should be noticed that for a given complex
number the ang!e () can be taken in infinite!y many ways. With the convention
tThe restriction r >- O(which is imposed throughout this text) is of particular significance here.
10.2. Elementary Functions oC a Complex Variable 541
that the modulus r be nonnegative, the various possible values of e differ by

integral multiples of 2n unless z = O, in which case e is arbitrary. Any one of
these angles is known as an argument or amplitude of z, and the abbreviations
e= arg z = amp == 4 =
y
are all used.
lf we notice that addition or subtraction of
complex numbers follows the paralIelogram law of
vector combination, the truth of the useful
~
--- ~~
/
//
/
/
ineq uali ties
1=,I-l z21<lz,+z21<lz,I+l z21 (lO)
follows directly from elementary geometrical con-
siderations (Figure 10.2). An analytical proof is x
supplied in Problem 4. Figure 10.2
10.2. Elementary Functions of a Complex Variable. We now proceed to

define functions such as ez, sin z, log z, and so on, taking care that these defini-
tions reduce to the conventional ones when z becomes the real variable x. t
The simplest such function is the integral power function
fez) = zn, ( 11)
where n is a positive integer or zero. This function naturally is defined recursively
by repeated multiplication, according to the law =k+' = Zk Z (k = O, 1, 2, ... ),
with ZO = 1, and there follows
zn = (x + iy)n = rn(cos e + i sin e)n (n=0,1,2, ... ). (12)
Apolynomial in z is then defined as a linear combination of a finite number
of these functions, where the constants of combination may be imaginary,
N
fez) = L: Anzn. (13)
n= o
A rational function of z is defined as the ratio of two polynomials.
By considering the limit of expressions of form (13) as N - 00 or, more
generally, limits of the form
fez) = lim
N-oc
N
L:
11=0
An(z - a)n = L:
n=O
~
An(z - a)n, (14)
where a may be real or imaginary, we are led to definitions of functions of a

complex variable in terms of power series. The convergence of such series may
tIt should be nolieed Ihal il is impossible lO "visualize Ihe graphs" of slleh funclions geomel-
rically when z is a complex variable, as one does for funelions of a real variable, sinee if (say)
w = sin Z one would require four-dimensional "visualization" lo relale Ihe real and imaginary
parts (say, Ii and v) of w lo Ihe real and imaginary parls (x and y) of z. A melhod of exhibiling
such a relalionship geometrically, which uses Iwo planes lO supply Ihe necessary four dimen-
sions, is considered in Chapler 11 (Seclion 11.4).
be investigated by the ratio test (see Section 4.1), just as in the case of series of
real terms. Thus, if we write
L = lim I An+ ¡ I (lS)

n--= AfI
when that limit exists, the series converges when
Iz-al<_1. (16)
L
Geometrically, this restriction is seen to require that z lie inside a circle of
radius R = 1/L with center at the point z = a in the complex plane.
Inside this circle of convergence the series can be integrated or differentiated
term by term, and the resultant series will represent the integral or derivative,
respectively, of the represented function.t This fact will be of. considerable
importance in much ·of what follows.
The exponential fimction e' is defined by the power series
ztl Z2 Z3
er =
O<>
L; -, = 1
n~O n.
+ z + -,
2.
+ -,
3.
+ . . .. (17)
This definition is acceptable, since the series converges and hence defines a
differentiable function of z for all real or imaginary values of z, and since this
series reduces to the proper one when z is real. If we multiply the series defining
e" and e" together term by term (as is permissible for convergent power series),
the resultant series is found to be that defining ez,+r, (see Problem 11); that is,
the relation
(18)
is true for complex values of z¡ and Z2. Consequently, if n is a positive integer,
we ha ve also the relation
(n = 1, 2, 3, ... ) (19)
for all complex values of z.
The circular fimctions may be defined, III terms of the already defined
exponential functions, by the relations
e iz _ e- iz
. sin z = (20a)
2i
cos z
e!Z + e- iz
(20b)
2
together with the relations tan z = sin z/cos z, and so on. Consequently, we
have, from Equations (20) and (17), the corresponding series definitions
. Z3 Z5 C'>O z2n+l
slllz=z--+-+ ... =L;(-I)n 1)! (2Ia)
3! S! n- O (2n +
tThis is a consequence of the fact that, if the series (14) converges when I z - al < R, then for
any p such that O < p < R it is tme that the series converges liniformly when I z - al <: p,
and the same is true of the result of integrating or differentiating the series term by termo
10.2. Elemenlary Funclions of a Complex Variable 543
_2 Z" 00 _!n
and cos Z = 1- ;
_
f
• .
+ -4f + ... = ~ ( -1)"(;
n-O _n .) .f'
(2Ib)
which reduce to the proper forms wben z is real. From these series, or from
the definitions (20a, b), it can be shown that the circular functions satisfy the
same identities for imaginary values of zas for real values.
Equations (20a, b) imply the important relation
e" = cos z + ¡sin z, (22)
whicb is known as Euler's formula.
With this relation Equation (9) takes the form
z = x + iy = r(cos (J + i sin (J) = re'·. (23)
In consequence of (23) and (I9) we then have
z" = r"(cos (J + i sin (J)" = r"e"e (n=I,2,3, ... ). (24)
But since Equation (22) also implies tbe relation
el". = cos n(J + i sin n(J,
we deduce DeMoivre's theorem,
(cos O + i sin (J)" = cos n(J + i sin n(J , (25)
and hence may rewrite (24) in the form
z" = r"(cos n(J + ¡sin n(J) (n = 1, 2, 3, ... ). (26)
Geometrically, Equation (26) shows that if z has the absolute value r and the
angle (J, then z" has the absoll/te value r" and the angle n(J, if n is a positive integer.
In a similar way we find that if z, = r,e'·' and Z2 = r2el." then
(27a)
and Z2
z1
= r2
r1
ei«(h-8d
(z, * O). (27b)
That is, if z, and Z2 have absolute values r, and r2 and angles (J, and (Jz, respec-
tively, tben Z,Z2 has the absolute value r,r 2 and the angle (J, + (J2' and Z2!Z,
has tbe absolute value r 2!r, and the angle (J2 - (J,.
In particular, we notice that since
I e'~ I = 1 (a real), (28)
the multiplication of any complex number z by a number of the form el., where
a is real, is equivalent to rotating the vector representing the number z through an
angle a in the comp lex planeo
The hyperbolic functions are defined, as for functions of a real variable, by
the equations
. e~ - e- r
smh z = 2 ' (29a)
cosh z =
er +2 e- r ' (29b)
and by the equations tanh z = sinh z/cosh z, and so on. ConsequentJy, we have
also
_ . Z3 ZS ::.o z2n +!
sinh z - z -t- TI ,. 5!
1
+ =I:o (2n +
n" l)! (30a)
Z4 .2 n
Z2
- I: _""_o
<XI
and cosb z = 1 + .2! + 4! + ns o(2n)!

(30b)
From these definitions it can be shown tbat hyperbolic functions of a complex

variable satisfy the same identities as the corresponding functions of a real
variable. By comparing Equations (20) and (29), we obtain in particular the
relations
sinh iz = i sin z , cosh iz = cos z,
(31)
sin iz = í sinh z, cos iz = cosh z,
relating the circular and hyperbolic functions.
The results so far obtained permit us to obtain expressions for the functions
considered In terms of their real and imaginary parts, including the following
ones:
e' = e x + iy = ee iy = (eX cos y) i(e x sin y), +
Sin z = sin (x + iy) = sin x cos iy + cos x sin iy
= (sin x cosh y) + i(cos x sinh y),
cos z = cos (x + iy) = cos x cos iy - sin x sin iy
= (cos x cosh y) - i(sin x sinh y), (32)
sinh z = sinh (x + iy) = + cosh x sinh iy
sinh x cosh iy
= (sinh x cos y) + i(cosh x sin y),
cosh z = cosh (x + iy) = cosh x cosh iy + sinh x sinh iy
= (cosh x cos y) + i(sinh x sin y).
10.3. Other Elementary Fu.nctions. We next define the complex logarithmic

function as the inverse of the exponential function. Denoting this function
temporarily by Log z, we see that the equation
w = Log z (33)
then must be equivalent to
(34)
To express Log z in terms of its real and imaginary parts, we write
w = u + iv, (35)
after which Equation (34) gives
z = x + iy = e U +/ V = e"e'v = e" cos v + ie" sin v.
Hence, equating real and imaginary parts, we obtain
e cos v
U
= x, e" sin v = y. (36)
JO.3. Grher Elementary Functions 545
There then follo\Vs
cos v = -x = cos , e sinv = Lr = si n e,

r
and hence, if r * 0,
u = log r = log I z 1, v = e, (37)
wbere r represents, as usual, the absolute value of z and log r is the ordinary
real logarithm , whereas e is any particular choice of the infinitely many angles
(differing by integral multiples of 2n) which may be associated with z. Thu s
we have obtained the result
Logz = loglzl + ie, (38)
for any z * e
O. To emphas ize the fact that is determinate only within an integral
multiple of 2n, we may here denote that particular value of e Which lies in the
e e
range O -< < 2n by p , and may s peak of this value as the principal va/ue of
e for the logarithm,
(39)
Then any other permissible value of is of the form e e= e p + 2kn, when k is
integral, and Equation (38) becomes
Logz=loglzl+i(e p +2kn) (k=0,±I,±2, ... ).· (40)
Thus it follows that the function Log z, defined as the inverse of e r , is an
infinitely many valued funetion. For example, if z = i, there follows I z I = 1
e
and p = n/2, and henee
Log i = log I + I (~ + 2kn) = 4k i I ni (k = 0, ± 1, ± 2, ... ).

(41)
The value corresponding to k = °
may be called the principal value of the
logarithm.
If, in a particular diseussion, z is restrieted 10 real positive values, say
e
z = x, there follows I z I = x and p = 0, and henee, from (40),
Log x = log x + 2kni (x real and positive). (42)
Thus the complex logarithm of a positi ve real number may differ from the
usual reallogarithm by an arbitrary integral multiple of 217.i. In order to conform
with conventional usage, henceforth we will identify the complex logarithm
Log::: with the real logarithm log z when, throughout a gil'en discL/ssion, z is
real and positive, by taking k = °
in (42) in sueh a case. rn the more general
case it is also conventional to write log z in plaee of Log z, with the understand-
ing that uoless z is to take on on/y real positive values, log z is to be considered
as multiply valued. Thus we \V iII write
log z = log I z I + i(e p + 2kn) (k = 0 , ± 1, ±2, ... ), (43)
in place of (40), and avoid resultant contradiction in the ease of (42) by taking
k = ° when z is a positive real variable.
546 Func!ions of a Complex Variable
Suppose now that at a given point z, on the positive real axis we choos e a
particular value of k in (43), say k = 0, and hence determine a particular value
of log Z,. If a point ;: moves continuously along a path originating at z" the
value of log z then varies continuously from the initial value log z ,. In particu-
lar, if z traverses a simple closed path surrounding the origin in the positive
(counterclockwise) direction and returns toward the initial point, it may be Seen
that the angle O increases by an amount approaching 2n; and hence as the circuit
is completed the logarithm is increased by 2ni, the real part of the logarithm
returning to its original value. This statement is tme , however, only for a path
enclosing the origin z = O. If now the point z continues to retrace its first path,
the logarithm is now given by a different (one-valued)function or by a different
"branch" of the same (multiple-valued) function. That is, if we write
(log Z)k = log I z I+ i(Op + 2kn) (O <: Op < 2n) ,
and if on the first circuit log z is determined (with k = O) by the branch (Iog z)o,
then, if log z ¡s to vary continuously, on the second circuit log z must be deter-
mined by the branch (log z)" corresponding to k = 1. The point z = 0, wruch
must be enclosed by the circuit if transition from one branch to another is to
be necessary, is known as a branch point. We may say that the function log z
has infinitely many branches, with a single branch point at z = O.
For any multivalued function, when transition from one branch to another
is to be prohibited, so that attention is to be restricted to a single branch as a
single-valued function, one imagines the complex plane to be "cut" along the
line (or curve) along which transition otherwise would take place. The branch
in question then wiU be discontinuous (in general) along the "branch cut" so
defined.
Thus, for the function log z, when Op is arbitrarily defined as that value of e
°
for which <: Op < 2n, each branch of log z is discontinuous along the positive
real axis. Clearly, in dealing with a problem in which this situation is undesirable,
one might instead adopt the convention -n < Op <: n, so that the discontinuity
along the branch cut occurs along the negative real axis, or one could make the
transition take place along any other ray, curve, or composite curve which
extends from the origin to infinity (without crossing itself).
To fix ideas, we will assume the definition (39) henceforth except when an
explicit modified definition is given .
The generalizedpower function fez) = z', where a may be real or imaginary,
is now defined in terms of the logarithm by the equation
z" = ea loa; , (44)
When z is a positive real variable and a is real, this definition is clearly in
accord with the usual definition if the logarithm of a positive real number is
taken to be real. If a is a positive integer, this definition must be consistent with
(24). To see that this is so, we write a = n, where n is any integer, and obtain
from (44)
(k = 0, ±I, ±2, ... ).
10.3. Other Elementary Functions 547
But since k and n are integers, we have

e Oknn ' = cos 2knn + i sin 2knn 1,
and hence there follows, in this case,
(n an integer). (45)
This result is in accordance with (24), since e'n8 = e' n I8 p +2kn) = e in8p when n is
an integer, and if n is a negative integer, say n = -p, it is identical with the
result of the recursive definitíon ,,-<p+11 = ,,-p!z (p = O, 1,2, ... ) when z O, *
again with Zo = 1.
More gene rally, if a is a real rationa/ number, we can write
m
a=-,
n
where m and n are integers having no common factor. Then Equation (44)
takes the form
(k = O, ± 1, ±2, ... )
or (k=0,±I,±2, ... ).
If we remember that m and n are given integers, whereas k is an arbitrary
integer, we can easily see that as k takes on any n successive integral values,
say k = O, 1, 2, ... , n - 1, the factor e2klmnlni will take on n corresponding
different values, but that if k continues to increase through the integers, the n
values so obtained are merely repeated periodically. Hence it follows that for
any nonzero value of z the function zm'n has exact/y n different va/ues given byt
(k=0,1,2, ... , n - I ) . (46)
It may be seen also that the function can be considered as having
zm/n
exactly n branches, such that if we again restrict {}p so that O <:: {}p < 2n, and
if a point traverses a simple closed contour inc/uding the origin (so that {} changes
by 2n), then a continuous variation of zm.'n is obtained only through transition
from one branch to another. We may verify further that if such a cJosed contour
is traversed exactly n times, starting initially with a certain branch, the transi-
tion from the nth branch is back to the initial branch. The point z = O is again
a branch point.
As an example, suppose that m/II = -je, and that a point z traverses the unit circle
in the positive (counterclockwise) direction, starting at the point z = 1. If we
arbitrarily start out on the branch k = O and note that for z = 1 there follows
r = 1, {}p = O, the initial value of Z2/3 at z = 1 is given by l. As the end of a
circuit is approached, the angle (}p approaches 2n, and as z ~ 1 the power Z2/3
approaches the value e'12/3)12n+O) = e l •.'3)n'. In order that Z2/3 vary continuously
as the point z = 1 is passed, and hence {}p drops abruptly to zero and then again
tThe notallonvw is sometimes used to denote the multivalued function w',n when w is
complexo In this text, however, the former notation will not be used at all unless w is real and
positive, in which case the radical here will denote the real and positive nth root. Thus, for
example, 3 I /2 ~ ± ...;3:
increases, we must now determine Z2/3 from the second branch, for which k = 1,
since this branch assumes at () p = O the value approached by the first branch
when () p ~ 2n. As the end of the second circuit is approached and again z ~ 1,
the power Z2/3 approaches the value e'i2'"3)iZ.+Z.l = e iB /3),i; and as this point
is passed, the transition to the /hird branch (k = 2) must occur. Finally, as the
end of the /hird circuit is approached and once more z ~ 1, the power Z2/3
approaches the value é iZ /3)iZ.+ 4n ) = e 4n ' = 1, and hence (for continuity) a
transition back to thefirs! branch (k = O) must take place. The three values taken
on by Z2!3 at z = 1 are 1, (-1 - i.../3)/2, and (-1 + i.../3)/2.
In the general case when a is complex, of the form
a=a l +ia 2 ,
where al and a 2 are real, Equation (44) becomes
= ra'e-a,i8p+2kn)[cos [a 2 log r + al«(}p + 2kn)]

+ isin[a 2 10gr +
a,«(}p + 2kn)]} (k = O, ±I, ±2, ... ). (47)
lt is apparent that, in general, the function za is infinitely many valued, with
Z = O as a branch point. If a 2 = O and a, is rational, then only a finite number
of branches exist; in particular, if a 2 = O and al is integral the function is
single-valued and z = O is no/ a branch point.
The generalized exponen/ial ¡une/ion fez) = aZ, where a may be real or
imaginary, but a =1= O, 1, is defined similarly by the equation
a = e,'oga. Z (48)
If we denote by (1,p the principal value of the angle corresponding to a, such
that, say, O <: (1,p < 2n, there follows
aZ = e(X+iy)[loa!al+i(cr.p+2kn:)) = e[xloa¡o\-y(o;p+2kJt)]e{[Ylo a1 a I +x(a p +2kn)]
la IXe-y(ap+2kn)[cos [y log la I + x«(1,p + 2kn)]

+ i sinly log I al + x«(1,p + 2kn)]} (k = O, ±l, ±2, ... ). (49)
Although this function apparently also is infinitely many valued, it is seen that
here the ambiguity arises only in the choice of the angle to be associated with
the constant a, and not upon the specification of the angular position of the
point z in the planeo That is, here there is no possibility of a continuous transi-
tion from an expression corresponding to a given value of k to one corresponding
to a second value, as the result of motion of a point z around a curve in the
complex planeo Thus, in this sen se, each choice of k can be considered as deter-
mining a separa/e ¡unc/ion, rather than a particular braneh of a single multi-
valued function. In any specific situation, only one such definition is needed.
It is usually convenient to take k = O, and hence write
a =
Z
la IXe-yap[cos (y log I al + X(1,p) + i sin (y log la I + xCtp)]. (50)
10.3. Other Elementary Functions 549
We notice in particular that if a = e (and hence (Xp = O), the definition (50)
reduces to that given in (32). This is the reason for choosing the particular
value k = O.
FinaUy, to conclude the ¡ist of elementary functions, we consider the inverse
circular and hyperbolic functions. In the case of the inverse sine function, the
equation
w = sin- 1 z (51 )
implies the equation
Z = SIn W . (52)
lf we make u se of the definition (20a), this equation takes the form
or, equivalently,
e 2iw - 2ize iw - 1 = O. (53)
Equation (53) is quadratic in e/ w , with the solution
e'W = iz + (l - Z2)1 / 2
When this equation IS sol ved for w = sin - 1 z, there follows finally
sin -1:: = J..

1
log [iz + (1 - Z2)1 / 2). (54)
It is important to notice that when z =-= ± 1 the quantity (l - Z2)1 / 2 has two
possible values. Then corresponding to each such value the logarithm has
injinitely many va/L/es. Hence, for any given value oT Z =-= ± 1 ¡he function
sin - ¡ z has two infinite sets of values; for z = ± 1, the two sets coincide. This
is, of course, already known to be the case when z is real and numerically less
than unity so tbat sin - ¡ z is real. For example, we have the values sin- 1 t =
n/6 + 2kn or 5n/ 6 + 2kn, wbere in eitber case k may take on arbitrary integral
values. In a later section it will be shown that sin - ' z has branch points at
z = ±l.
We may verify that Equation (54) gives the known values for sin- ' by t
making the calculations
•
SlO
- 1
21 = T1 1og [i2 ± J3]
2
= j+[log 1 + i(~ + 2kn)]

+[IOg 1+ ie6n + 2kn)]
_ ~ + 2kn or 5n
6
+ 2kn (k = O, ± 1, ± 2, .. . ).
In an entirely analogous way, express ion s may be obtained for the other
inverse functions, including the following ones:
sin- I z = ~ log [iz + (1 - Z2)1 ( 2] , (55a)

1
cos- I Z = ~ log [z + (Z2 - 1)1: 2], (55b)

1
t an -1 z = 1 I i - Z
2i og i + z' (55c)
co
t- 1
z =
I I z+ i
2i og z _ i ' (55d)
sinh- I z = log [z + (l + Z2)1 / 2], (56a)

cosh -1 z = log [z + (Z2 - 1) 112], (56b)
tanh- I z = -} log : ~ ~, (56c)
coth - 1 z = _1 log z + l. (56d)

2 z - I
The functions so far considered in this chapter are the basic elementary
fimctions. Any linear combination of such functions or any composite function
defined in terms of a finite number of such functions is al so known as an ele-
mentary function.
The derivative of a function of a complex variable is defined, as in the real
case, by the equation
df(z) = F(z) = lim fez + dZ) - fez) (57)
~ A.-O dz
when the indicated limit exists. It is readily verified that the derivative formulas
established for elementary functions of a real variable are also valid for the
corresponding functions of a complex variable, as defined in this section.
Thus, for the function fez) = e''', with a constant, there follows
fez + dZ) -
6z
fez) = e.' (e. t!J.z 1)
A
' -
= ae"'[ 1 + a:!z + (a~~)2 + .. -}

for any dz "" O, and hence F(z) = ae"'. Since the other elementary functions
are simply related to the exponential function, expressions for their derivatives
then are easily derived from this result.
10.4 Analytic Functions of a Complex Variable. A function fez) is said to

be analytic in a regio n eR of the complex plane if fez) has a finite derivative at
each point of CA and if fez) is single-valued in eR. It is said to be analytic at a
10.4. Analytic Funclions of a Complex Variable 551
poin t z if ;;; is an interior point of sorne region where fez) is analytic.t Thus a
function cannot be analytic at a point without al so being anal y tic throughout
so me cirele with center at that point.
It can be s hown that, if a functioo fez) is analytic at a point z, then the
derivative f '(z) is conlinuous at z. The truth of this theorem (due to Goursat ,
and relatively difficult to prove) is assumed here. It then follow s (Section 10.7)
that, in fact, fez) has continuous derivatives of all orders at the point z.
To require that a function w = fez) have a finite derivative at a point z is
equivalent to requiring that the limit
dw = lim fez + ~z) - fez) = lim ~w (58)
dz 0.,-0 ~z 0.'':'0 TI
exist uniquely as ~z - > O from any direction in the complex plane o That is, the
value approached by the ratio ~w(~z must exist for any direction of approach
and must not depend upon the direction.
As an example of a simple function which is not analytic anywh ere, we
consider the relation
w = x - iy = z. (59)
Here w can be considered as a function of z = x + iy, since, if z is given, the
real and imaginary parts (x and y) of ::: are determined and hence w is deter-
mined. However, if we examine the ratio
~w ~x - i~y
(60)
~z = ~x + i~y '
we see that if ~::: approaches zero along a line parallel to the real (x) axis, so
that ~y = O throughout the limiting process, the limit of (60) is + 1, whereas
if ~z approaches zero along a line parallel to the imaginary (y) axis, so that
~x = O throughout the process, the limit is -1. More generally, if ~z
approaches zero along a curve with slope dy(dx = m at the point considered
In the complex plane, there follows
lim ~w = lim 1 - i(~y(~x) = 1 - mi = (1 - m 2 ) - 2m;,

0.,-0 ~z 0.,-0 1 + i(~Y/~x)
ay-O
I + mi l + m2
and hence a different limit is approached for each value of m. Thus fez) = i
does not have a derivative anywhere, and hence certainly is not analytic any-
where.
Suppose now that the limit (58) does exist uniquely, independently of the
manner in which ~z --' O. We indicate the real and imaginary parts of w = fez)
b y ti and v, respectively, where u and v are certain functions of x and y, and so
write
w = ~I(X, y) + iv(x, y). (61)
tA point P is an interior poínl of a region <R if and only ir every sufficiently smaJl circle with
center at P contaios only points of <R . A regíon, by definition, musl pos ses s interior poiolS and
i t also may possess boundary poínts.
Then rhere follows

dw = lim Liu + iLiv . (62)
dI. 6,-0
óy-o
Lix + iLiy
For an approaeh parallel ro rhe real axis we set Liy = O first and so obtain
dw = lim Liu + ¡Liv = lim (Liu + ¡LiV) = au + i av , (63a)

dz 6x-0 Lix 6x-·0 Lix Lix ax ax
where the partial derivatives are formed with y held eonstant. For an approaeh
parallel to the imaginary axis, we obtain similarly
dw = lim Liu + iLiv = lim (Liv _ iLiu) = av _ ¡au. (63b)

dI. 6''''':0 iLiy 6,-0 Liy Liy ay ay
But if fez) has a derivative at the point eonsidered, the two values (63a) and
(63b) must be equal,
dw _ d(u +
iv) = 1'(7) = au + iav _ av _ iau. (64)
dz - dz - ax ax - ay ay
Henee the real and imaginary parts of the last two expressions must be respee-
tively equal, so that ti and v must satisfy the equations
au av
-=- , (65a)
ax ay
au a·v
ay = - ax· (65b)
These equations are known as the Catlchy-Riemann equations. In order

that w = L/(x, y) +
iv(x, y) be analytie in a region CR of the eomplex plane, it
is thus necessary that these equations be satisfied for values of x and y eorre-
sponding to all points in that region. Conversely, it is true that, if these equa-
tions are satisfied and if the four partial derivatives involved are eontinuous in
CR, then the derivative exists uniquely at eaeh poinr in CR for any method of
approaeh to the limit.
The proof follows from the faet that, if rp is a rea)-valued funetion of x and y
with eontinuous fírst partia) derivatives in Ihe neighborhood of a poinl, Ihen
ehanges in x, y, ancl rp are relaled by Ihe equalion
Lirp = (~~ + l' 1)Lix + (~~ + 1'2) Liy,
where 1'1 and 1'2 tend lo zero as Lix and Liy both tend lo zero. Thus, if fez) =
1I + iv is single-valued, and if Ihe real funetions u and v have eontinuous fírst
parti'll derivalives in Ihe neighborhood of a poinl P in CR, then

Af
'-" =
(aU,
ax,1'1,) '-"x
A
+ (aU,
ay , 1 ' :,) ,-"y+¡
A .(av
ax+l'¡,,) '-"x+¡
A .(av
ay+1'2,,) ,-"y,
A
where I"¡, é~, 1"(, and I'~ aIJ lend lo zero wilh L!..x and Liy. If also (65a, b) hold al
10.4..-\.nalytic Functions of a Complex Variable 553
the point P, this relalion can be rewritten in the form
ól = (aaUx - iaav)(ÓX
x
+ iÓy) + (f', + if't')t"J.x + (f'z + if'{)t"J.y,
from which 1here follows

ó~ _
Ó¿
(aU
ax -l-
'
¡.a/})
ax -_ (f' ...L.
,,1
·f")t"J.x . (f'
1 Óz T 2
+ 'f")ÓY
1 Z Óz'
Finally, since IÓx/Ózl <: 1 and IÓy/Ózl <: 1, the right-hand members tend to
zero as Óx and Óy both tend to zero, and hence as Óz - , O in any way, so that
the existence of the derivative
lim ÓI = au + iav = di
",,-o Óz ax ax dz
at the point P is established. Since this result is guaranteed for aH points in <R, it
follows that I(z) is analytic in <R.
Thus a single-valued complex function w = fe:::) = u + iv, for which the
first partial derivatives of u and vare continuous in a region <R, is analytic in
<R if and only if the Cauchy-Riemann equations (65a, b) are satisfied. It is clear
that the function w = x - iy
does not satisfy this condition anywhere.
As a further example, we may verify that for the function
w = fe:::) = (x - y)2 + 2i(x + y)
the Cauchy-Riemann equations are satisfied only along the fine x - y = 1 in
the complex plane. Thus, since I'(z) accordingly exists only on this line, and
not throughout any regio n <R, fez) is nowhere analytic. A t all points on the line,
direct calculation shows that I'(z) has the constant value 2 + 2i. (See Problem
22.)
From Equations (65a, b) we obtain the additional relations
au 2
av
2 azu av 2
ax2 = axay' ay2 = -ayax'

av2
av au
2 2
-a
X
2 = ay2 = ayax
But if the second partial derivatives of u and vare continuous, the order of
differentiation is immaterial and it then follows that u and v satisfy the equations
V2 u = aax2u ..,, aay2u -- O,
2 2
(66a)
vzv = a v , a v - O.
2
"'a~ -
2
(66b)
ax" y-
The previously mentioned fact that an analytic function has derivatives of all
orders (see Section 10.7) implies, in particular, that the partial derivatives
involved are indeed continuous. Hence it follows that the real and imaginar y
parts 01 an analytic lunction are solutions 01 Laplace's equation.
554 Functions oC a Complex Variable
The Cauchy-Riemann equations (65a, b) permit the determination of either

U or V if the other is known, apart from an arbitrary real additive constant ,
since, by making use of these equations, we may write
au au av av
du = ax dx + aydY = aydx - axdy, (67a)
av av au au
dv = -dx
ax
+ -dy
ay
= - -dx
~
+ -dy.
ax
(67b)
Thus, for example, if we have

V = y2 _ x 2 ,
then (67a) gives
du = 2y dx + 2x dy
and hence
LI = 2xy + e,
where e is an arbitrary real constant. t It follows that y2 - X2 is the imaginary
part of the function
2xy + i(y2 - x 2) +e = -iz 2 + C.
We deduce that if one part of an analytic function is prescribed, the other
part, and hence the function itself, is determinate within an arbitrary additive
constant. A complication, which wil! be considered later, is that when the region
IR is not simply connected the determined part may not be single-valued.
Two functions II(X, y) and v(x, y), which are respecti vely the real and imagi-
nary parts of an analytic function, are known as harmonic fllnclions, and V is
often caBed the harmonic con;ugate of u.
10.5. Line Integrals of Complex Functions. If e is a curve in the complex

plane joining the points Zo and z l' the line integral of a function fez) = LI + iv
along e is defined by the equation
fe fez) dz = fe (u + iv)(dx + i dy)

= fe [(u dx - v dy) + i(v dx + u dy)]
or fe fez) dz fe (u dx -
= v dy) i fe (v dx - u dy). (68)
lt is supposed here, and in what follows, that any curve e along which such
tCare should be taken to avoid the errors frequently introduced by the generally improper
procedure of determining u(x, y) from dll ~ P(x, y) dx + Q(x, y) dy by integration, holding
y constant in the first term and x constant in the second one . ISee Secl ion 1.12 (Example 3) or
Section 6. 1 J .)
10.5. Une Integrals of Complex Functions 555
an integration is to be effected is at least piecewise smooth (se e Section 6.10).

Such a curve is sometimes called a contOl/r, although this term also is often
used in a less restrictive sense.
The real and imaginary parts of (68) are thus ordinary real line integrals
in the plane, of the type considered in Section 6.10. We see that the requirement
that the two integrands be exact differentials is precisely the requirement tbat
the Cauchy-Riemann equations (65a, b) be satisfied. The results of Sections
6.10 (or 6.16) and 10.4 then show that the fine integral fe fez) dz is independent
o/ the path C joining the end points Zo and ZI if C can be enclosed in a simple
(simply connected) region eR in which fez) is analytic.
The results cited state that if P x , P y , Qx, and Qy are continuous and P y = Qx,
in a simple region eR of the xy plane, then the real integral fe (P dx + Q dy)
between two points in eR is independent of the path e joining those points,
provided only Ihal e is piecewise smoolh and lies in eR, and it Ihen follows
also Ihal P dx + Q dy = drp, where rp is single-valued in eR. The two line
integrals in (68) correspond lo P = l/, Q = -v and lO P = v, Q = u, respec-
tively. In addition lO using (65), il is al Ihis poinl Ihal we must exploil the
continuity of I'(z), and hence of u " u y , "x, and V y , in eR.
In such a case the curve need not be prescribed and we rnay indicate the
integral by the notation J::
fez) dz. It follows also that here fez) dz is the exact
differential of a function F(z),
fez) dz = dF(z), (69)
and that the integral can tnen be evaluated in the usual way, according to tne
formula
f ~' fez) dz =
Zo
F(z¡} - F(zo), (70)
where F(z) is a function whose derivative is fez).

By considering the case where tbe end points Zo and z 1 coincide, we con-
elude that, if C is any closed curve lying in a simple region eR wbere fez) is
analytic, then the line integral of fez) around C will vanish. If we recall further
that, when fez) is analytic at all points on a closed curve C it is al so analytic
in sorne neighborhood (on both sides) of the curve, we deduce finally that if
fez) is analytic inside and on a closed curve C, then
fe fez) dz = O. (71)
This result is known as Cauchy's integral Iheorem.

In other cases the line integral of a function fez) around a closed contour
may or rnay not vanish. The positive direclion around a closed contour is
defined, as before, as the direction along which an observer would proceed in
keeping the enclosed area to his lefl.
556 functions of a Complex Variable
We notice tbat, if ti and vare expressed in polar coordinates, Equation (68)

is replaced by the equation
Lfez) dz L + iv)(e" dr + ire" de)

= (u
L[(u cos e ~ v sin e) dr ~ (v cos e + u sin e)r de]

+ L cos e + u sin e) dr + (u cos e ~ sin e)r de].
i [(v v (72)
Tbe requirement that the quantities In brackets be exact differentials can be

reduced to tbe conditions
au 1 av
ar =rae'
(73)
1 au av
r ae = ~ ar·
These equations are the forms assumed by the Cauchy-Riemann equations (65)
in polar coordinates (see also Problem 26).
To illustrate these results, we notice that for the single-valued function
1
fez) =-
z
the derivative F(z) = _1/Z2 exists at al! points except z = O. Hence fez) is
analytic in any simple region <R not including the origin in tbe complex plane.
If we are to evaluate the integral
r dz,
Jc Z
where e is, say, the straight line from Zo = 1 to z 1 = i, we can avoid the explicit
evaluation of a form such as (68) or (72) for trus path by choosing a more
convenient "equivalent" path joining those points (s ay, a quadrant of acirele)
which, together with e, lies in a simple region <R which excludes the origino
That is, e here could be replaced by any curve which also joins Zo and z 1 and
which hes on the same side of the origin as does C. Even more conveniently, we
could notice that I/ z = d(log z)/ dz and write
fe d: = [IOg z 1= log i - log 1.
But here care must be taken to note that both log j and log 1 are infinitely ambi-
guous, and that the correct answer will be assured only when use is made of
values of both which correspond to a specific branch of log z which is analytic
in a simple region <R which ineludes tbe prescribed are C. For trus purpose, tbe
10.5. Line Integrals of Complex Functions 557
e
branch for which -n < p <:: n will serve,t for example, and with this choice
there follows log i = ni/2 and log 1 = O, and hence
1~z = ~ i.
For any other permissible branch, both log i and log I would be modified by
the addition of the same multiple of 2ni, and the same final answer would be
obtained.
Here the value of the integral is the change experienced by any branch of
log z in correspondence with a transition from Zo '-- 1 to Z I = i along C (or
aJong any equivalent curve, as defined aboye) provided onJy that the branch
used is analytic along C.
Any c10sed curve C not surrounding the origin satisfies the conditions of
Cauchy's theorem with f(z) = l / z, and hence for any such curve there follows
i dz = o.
Je Z
However, if C endoses the origin, the integral need not vanish. If C is taken
as the unit circle I z I = 1, with center at the origin, then on C we may write
dz = ¡e" de,
and hence for this closed curve we obtain
f d =
~
el Z
Sh
o
e-"(ie" de) = S2.
o
i de = 2ni, (74)
again without any need for actually splitting the integral into real and imagi-
nary parts, where fe. indicates integration around the positive direction of the
unit circle. The value given by (74) is merely the
limit of the increase experienced by the imagi- y
nary part ie of any branch of 10g z as z
describes a positive circuit about the onglO
which tends to become dosed.
Now consider any other simple closed e
curve C which surrounds the origin and does x
not intersect el. If we make a "crosscut" from
el to C, and so determine the simply connected
region <R shown in Figure 10.3, we see that,
since f(z) = l/z is analytic in <R, the integral
around the complete boundary of <R must Figure 10.3
tThis example illustrates the necessity of abandoning the arbitrary definition O os; (Jp < 2"
when the branch is to be analytic on any pan of the positive real axis. In other situations 10 be
encountered, the earlier definition often will be the preferred one .
558 Funclions of a Complex Variable
vanish, by Cauchy's theorem. As the width of the cut is decreased to w ard zero
the integrals along the edges of the cut are taken in opposite directions and
hence cancel. Since the part of the complete integration which is carried Out
along e, is taken in the negative direction, there follows in the limit
i dz _ i dz = O or
Je z J el Z
It is easy to remove the restriction that el and e not intersect, by choosing

a third closed curve e' which encloses both el and e and noticing that then tbe
integrals along el and e are each equal to the integral along e', and hence are
equal to each other. A more direct approach, when e and el have a finite
number of intersections, consists of considering the sum of the results of inte-
grating around each of tbe simple regions which líe inside one curve and out-
side the other (Figure 10.4). By either tine of reasoning, we may deduce that
Figure 10.4
for any closed curve e enclosing the origin once III the positive direction, we
have
f -dz = 2'
e z
nI. (75)
More generally, if fez) = z·, where n is an integer, there follows
f el
z· dz = fo2' e· i9 (ie i9 de) = i f2' e'·+
o
!)i9 de
2'
= i
f
o [cos (n + l)e + i sin (n + l)e] de
= O (n * -1), (76)
where el agalll represents the unit circle I z I = 1. lf n is a positive integer or
zero, this result is in accordance with Cauchy's theorem, since then fez) = z·
is analytic for al! finite values of z. However, if n is a negative integer, z· is no!
10.5. Line Inlegra]s of Complex Funclions 559
analytic at z = O. Still, Equation (76) states that if n =;t=. - ¡, the integral fe, zn dz
vaníshes even in this case. The precedíng argument shows that this condition
is true also for any elosed curve e surrounding the origin, and hence, SlOce
zn is ar.alytic except at the origin, we conelude that, when n is an integer,
~e zn dz = O (n=;t=. -1) (77)
for any elosed curve not passing through the origino If n = - 1, the integral is
zero unless e ineludes the origin, in which case the value is 2ni if e ís simple
(and positively oriented).
If we replace z by z - a, where a is any complex number, we deduce from
(76) and (77) the additional results
fe (z - a)" dz = O (n =;t=. -1), (78a)
f e
-dz
-- =
Z - a
2 '
ni,
where e now is a simple elosed curve enclosing the point z = a in the positíve
(78b)
direction and n is an integer. These results wíll be of use in Section 10.12.

Fundamentally, the mathematical definition of the complex line integral
fe f(z) dz along a curve C between the points a and b in the complex plane
consists of the limit of the sum
as the number n + 2 of the division points Zo (= a), z 1> . . • , Zn' Zn+ 1 (= b)

along e increases and the chord lengths 1 dZ k 1 alI tend to zero. This definition
can be shown to lead to the evaluatíon (68), in lerms of two realline integrals,
whenever f(z) is continuous on e and e is piecewise smooth, whether or not
f(z) is analytic.
Repeated use of the inequality (lO) leads to the inequality
I kto f(Zk) dZ k I < kt 1 f(Zk)

o
11 dZ k 1,
and henee, in the limit as n - , 00, there folIows
Ifef(z)dzl < fe If(z)lldz 1 = fe If(z)lldx + i dyl = fe 1 f(z) 1 ds,
where s is are length along C. If now 1 f(z) 1 is bounded 00 e, say

1 f(z) 1 < M 00 e, (79)
the last member of this expression is numerieally not greater than
r M ds = M r ds
~C ~C
= ML,
where L is the length of the curve e between the speeified eod points.
Hence we obtain the very useful " NfL inequ a lity" which states that
ILf(z)dzl <: ML, (80)
where NI is an upper bound for I fez) I Oll C, and L is the length of C.

In addition, for future use we generalize preceding considerations by n o tino<>
that if C I and C, are two nonintersecting simple closed curves, one enclosing
the other, and if fez) is analytic on c l and c l and in the region CR between
C I and C 2 , then by introducing a crosscut as in Figure 10.3 and proceeding as
III the development leading to (75), we show that
fe, f(z)dz fe, fez) dz.

=
More generally, by also considering situations in which two simple closed curves
C I and C 2 intersect in two or more points, as in Figure lOA, with fez) analytic
on c l and C l and in the regions which lie inside only one of the two curves,
we deduce that in dealing with an integral of the form fe, fez) dz, where C I is
a simple c!osed curve, we may deform C 1 in a continuous way into any other
simple c!osed curve C 2 without changing the value of the integral, so long as fez)
is analy tic on C l and c 2 and at al! points passed over by the curve in (he process
of /he deforma/ion. In such cases, we say that el and C , are equivalen/ con/ourS
fo r the relevant integral.
10.6. Cauchy's Integral Formula. Let C be a simple closed con tour inside
which and along which fez) is analytic, and let ct be a point inside C. Finally,
let C, be a small circle of radius é, with
y center at the point Ct, inside C (Figure 10.5).
Then, clearly, the function
fez)
z-fX
is analytic between C and C,' and so there
folIows
fe /~~dZ _ ji e , Z
fez) dz
- ct
(81)
x
for all positive values of é. In particular,
Figure 10.5 Equation (81) mu s t remain true in the limit
as é ~ O. But for points on C, we have
(O <: f) < 211:), (82)
i fez) dz =
j c~ z-(X Jo
rz. f(ct + ée i O) i df}. (83)
10.7. Taylor Series 561
In the limit, as f -40, we then obtain from (81) and (83) the result
J.i e ¿
/(z) dz =
- ct
lim
F.-O
fh o
f(1X -i- fe'") i de = 2nifCa), (84)
since fez) is continuous at z = a. Equation (84) expresses the value of fez) at

z = a in terms of the values of fe;;) along an enclosing curve C. This result
can be written in more convenient form if we interchange z and a in (84),
fez) = _1. I fea) da, (85)

2m jea - z
where C is now a closed contour enclosing the point z. This is Cauchy's integral
formula for a function fez) analytic inside and along a simple closed curve C.
If we recall that the real and imaginary parts of fez) are solutions of
Laplace's equation inside C, and hence (see Section 9.2) are determined at
points inside C by their values on C, we see that (85) affords exactly this deter-
mination, if bOlh parls are given on C.
10.7. TayJor Series. We next use this result to show that, if fez) is analytic
at a point z = a, then fez) can be expanded in a Taylor series of powers of
z - a. For this purpose we first notice that the function I/(a - z) can be
expanded in the geometric series
1 1 1
a - z - (a - a) - (z a) - a - al z-a
IX-a
_
ct -
1
a
:t (z - a)"
71=0 a IX -
(1 z - al < IIX - al). (86)
We now take for C any circle, with center at z = a, inside which and on which
fez) is analytic (Figure 10.6). Then for points inside C we have
I z - al < I a - a I = p,
where p is the radius of the circle, and hence for such points the power series
(86) converges and also can be multiplied by fea) and integrated around the
y
Ci
Figure 10.6
curve e term by termo Thus (86) can be introduced into (85) to give the result
fez) = fo- 2niI [!J e(IXf(lX) dlX ]

_ ar" (z - a)
n
(1 z - al < p),
or, equivalently,
=
fez) = 1: An(z
n- O
- a)n (1 z - al < p), (87a)
where A =_1_ i f(lX) dlX .

(87b)
n 2ni Je (IX - a)n+'
Thus, if fez) is analytic throughout the circular disk z - al -< p, it can be I
expanded in a Taylor series (87a) converging inside this disk, with coefficients
given by (87b). Since any Taylor series has derivatives of aH orders (which can
be obtained by successive term-by-term differentiations) inside its circle of
convergence, and since it is now established that fez) can be represented by such
a series when I z - al < p, it follows in particular that fez) indeed has deriva-
tives of al! orders at any point z = a at which it is analytic.
By differentiating (87a) k times, setting z = a in the result, and using (87b),
we obtain the additional formula
f 'kJ(a) = k'
•
A
k
= k!
2ni
i
Je
f(lX) dlX
(IX - a)"+'
or, with a change in notation,
f'nJ( 7) = n! i f(lX) dlX , (88)
- 2ni Je (IX - z)n+'
where e is now the circle with radius p about z, or any equivalent contour
(Figure 10.7). We notice that (88) is obtained
y
formally by differentiating (85) n times under
the integral sign.
If R is the distance from z = a to the
nearest point at which fez) is not analytic, the
series (87a) converges to fez) when I z - al < p
for every p such that p < R, and hence indeed
the convergence is assured when I z - al < R.
From these results we conclude that if fez)
is analytic at a point z = a (and hence in sorne
x
circle around this point), derivatives of fez) of
Figure 10.7 al! orders exist at that point, and fez) can be
expanded in the Taylor series
= f'nJ(a)
fez) = 1:o
n==
,(z
n.
- a)n, (89)
the circle of convergence of the series coinciding with (or including) the largest
circle with center at z = a inside which fez) is analytic.
10.8. Laurent Series 563
Conversely, we see also that, if I(z) can be represenled by a Tay/or s eries in

some eire/e aboLlt z = a, Ihen I(z) rnust be ana/Ylic at z = a, since then deriva-
tives of all orders exist at and near z = a.
General/y, the radius of convergence Re of the Taylor series (89) will equa!
the distance R from z = a to the nearest point where I(z) ceases to be analytic .
However, there exist somewhat unusual situations in which Re> R . For
example, suppose that 1(::) has the definition
{~
inside Ca,
I(z) = (90)
outside Ca,
where Ca is a simple closed curve enclosing the point z = a. Then R is the dis-
tance from z = ato the nearest point on Ca. But, since I(a) = l and I ' n'(a) = O
for n::> 1, the series representation (89) reduces to the form
I(z) = l CI z - al < R).
In this special case, the representation is valid, more general/y, everywhere
inside Ca and, furthermore, the one-term series itself c1early converges for all
values of z, so that Re = oo. In fact, the series defines a function I*(z) = 1
for all z, which is analytic everywhere and which is identical with I(z) inside
Ca. Whereas the definition (90) may seem to be somewhat artificial, it may be
noted that the more compact definition
l idex (90')
I(z) = 27ti j c, ex - z
is equivalent to it.
More generally, the only situations in which Re > R are those in which
I(z) is so defined that the neighborhood CR of z = a inside which I(z) is analytic
is limited, but in which it is possible to define or redefine I(z) outside that
region in such a way that I(z) is analytic in an extended regio n CR" which
includes CR. Such a process of extending the regio n of definition of an analytic
function is called ana/ytie continuation.t
10.8. Laurent Series. A generalization ofthe Taylor series expansion of I(z),

which may involve both positive and negative integral powers of z - a, exists
whenever there is a circular ring or annu/us CR, for which PI < Iz - al < P2'
such that I(z) is analytic (and hence single-valued) in CR and on its inner and
outer circular boundaries C I and C, (Figure 10.8).
As a first step toward its derivation, we introduce a crosscut between C I
and C 2 and apply Cauchy's integral formula to the resultant simply connected
region CR ' bounded by the curve C' (Figure 10.9). By considering the limit as
the width of the cut tends to zero. we thus deduce that, for any point z in the
Un situations less contri ved than the one considered here, there generally will be points on the
boundary of (!\ which cannot be interior points of the extended region (!\ *.
y y
x x
annulus CH, there follows

fez) = _1. i fea) da _ _1. i
fea) da, (91)
211:/ J e, a - z 211:1 J e, IX - z
where both el and e 2 are to be traversed in the positive (counterclockwise)
direction.
The first integral in (91) now is dealt with exactly as in Section 10.7, by
expanding l j(a - z) in powers of (z - a)j(a - a) and integrating term by term,
to yield the relation
_1.
27tl
i
J c~ (J.
fea) da
- Z
= i: An(z -
IJ=O
a)n (1z - al < P2), (92)
where again
A = _1_ i fea) da (n = 0, 1, 2, ... ). (93)
n 2ni J e, (a - a)n+ 1
An important point here is that the equal members of (92) now generally do
not also equal fez), nor does n! An generally equal f'n >(a), since now fez) is
not necessarily analytic everywhere ioside e 2 •
In the second integral of Equation (91), we iostead expand l j (a - z) io
powers of (a - a) j(z - a) ,
I I 1 I
( t - z = (z -a) - (a - a) = z - a 1 a-a
z-a
= i; (a - a)n-l
n ~1 (z - a)n
-1
(z _ a)n
= n~_ (a - a)n+l (Iz - al> PI)'
so that, after multiplying this convergent power series by f(a) j(2ni) and inte-
10.8. Laurent Series S65
grating term by term around el, there follow s
__1_. J
f(a) da =
2m J e, a - Z
1:
n_ _ ~
Bn(z _ a)n (94)
where
B = _1_ I f(a) da (n=-I,-2, ... ). (95)
n 2ni J e, (a - a)n+ 1
Now, s ince Equation (92) is true for z in si de e 2 and (94) is true for z
outside el' both are true for z in the ring eR, and hence then can be introduced
into (91). Further, since f(a)/(a - a)n+ 1 is an analytic function of a when a is
in eR, when n is zero or any positive or negative integer, it follows that , in (93)
and (95), el and e 2 can be deformed into any simple closed curve e which
su rroundsz = a and lies in eR.t
Hence, finally, if we write
_ 1 i f(a) da (n = O, ± 1, ±2, ... ), (96)
en - 2ni J e (a - a)n+l
there follows An = en when n >- O and En = en when n <: -1, and the result
of introducing (92) and (95) into (91) thus becomes
(p, < Iz - a I < p,). (97)

n= -~'"
This expansion is known as the Laurent series expansion of f(z), in the specified
annulus eR, and is seen to involve both positive and negative powers of z - a,
in the general case.
If R I is the smallest radius and R 2 the largest radius for wruch it is true that
fez) is analytic when R I < I z - al < R 2 , so that also f(z) fails to be analytic
somewhere on the circle I z - al = R, and somewhere on the circle I z - al = R 2 ,
then since PI and P2 can be taken to be any radii such that R, < P, < P2 < R 2 ,
it follows that the series (97) converges to f(z) when R, < I z - al < R 2 •
However, if fez) is analytic everywhere inside the circle I z - al = R 2 , (96)
shows that en = O when n < O, and (97) then reduces to the Tayfor series (87),
valid when I z - al < R 2 • On the other hand, if fez) is analytic everywhere
outside the circle I z - al = R" the series involves only a constant term and
negative integral powers of z - a and can, in fact, be considered as an ordinary
power series in (z - a)-I, va lid when I z - al > R l ' Here (as will be seen in
Section 10,10) the function f(z) must, in particular, be analytic " at infinity."
Just as in the case of ordinary power series , it can be proved that, within
its annular region of convergence, a Laurent series can be integrated (or differen-
tiated) term by term, and the result will converge to the integral (or derivative)
of the function represented by the original series .
tI! should be noted (hat this slatemenl generally does no! apply to (he integrals in Equation
(91), sinee f(a.) f(a. - z) generally is not an analytie funetion of a. w hen a. = z.
Further, it is known that the expansion (97) is unique, in the sense that ir
an expansion of ¡he form
fez) = I: bn(z - a)n
can be obtained in any way, and is valid in ¡he specific annulus
R¡ < Iz - al < Rz,
then necessarily b n = en· (See Problem 56.) This fact frequently permits one to
obtain a desired Laurent series by elementary methods, without attempting to
evaluate the coefficients by using (96).
Example l. From the expansion e' = ¿:; ¡n/n!, which is valid for al1 ¡, we may deduce
other relations, such as
zn
+ (n + J) ! + (z 7"= O)
and
= J 1 1
el,';:
1 + 1! z + 2! Z2 + ... + n ! zn + (z 7"= O).
Both these series are Laurent expansions with a = 0, R I = 0, and Rz = co. •
Example 2. The geometric series expansions
_1_ = 1
1 - z
+ z + Zz + ... + zn + (Izl <1)
and
_1_
1 - z
= -(l/z)
1 - (l/z)
= -...l(1
z
+ J... +
z
... )
1 1 1 1
(Izl> 1)
have been assumed to be familiar. The first of these expansions is the Taylor series
va lid inside the circle I z I = 1, the second the Laurent series valid outside that circle .
In the general case, if a Laurent expansion of fez) is required in powers

•
of z - a, in the annulus R¡ < I z - al < R 2 , we may attempt to express fez)
as the sum of two functionsfJz) andf2(z), such thatf¡(z) is analytic everywhere
outside the inner circle I z - al = R¡ and f2(Z) is analytic everywhere inside the
outer circle I z - al = R 2 • The desired representation then can be obtained by
expandingf2(z) in an ordinary power series (TayJor series) in z - a, expanding
f¡ (z) in an ordinary power series in (z - a)- ¡, and adding together the results.
Occasionally, it is preferabJe to write fez) instead as the product fl (Z)f2(Z),
wheref¡ and f2 have the properties defined abo ve, and to multiply their expan-
slons.

2
fez) = z(z - l)(z - 2)
10.9. Singularities of Analytic Functions 567
cJearJy is analytic excepl when z = O. 1. and 2. To expand fez) in a Laurent series

of powers of z in ¡he annulus CR: 1 < 1 z; < 2, we may firs¡ write
ez) = _1 _ _2_...:... _1_.
f z z-I':;-2
Since the first two terms are analytic when z > 1, whereas the third term is analytic
I 1
when Izl < 2, we may expand the first two terms in powers of l/z and the third in
powers of z, to obtain
2"z
fez) = zI - 1 -(I/z) -
1/2
(z/2)
Thus there follows

~
fez) = ~ enz n (1 < Iz 1< 2),

n- -<XI
where en = _1/2n+1 when n > 0, e-l = -1, and en = -2 when n <: -2. The same
°
function also possesses two other Laurent series in powers of z, valid when < 1 z 1 < 1
and when 2 < I z I < ca, as well as two, three or four Laurent (or Taylor) series in
powers of z - a. for any O/he,. value of a. •
10.9. Singularities of Analytic Functions. If any region CR exists such that

fez) or a branch of fez) is analytic in CR, then we speak (somewhat loosely)
of fez) as "an analytic function." Points at which a single-valued function fez)
is not analytic are called singular points or singularities of the function. AIso, a
point at which a branch of a rnu/tivalued function fez) is not analytic is called
a singular point of fez). From the definition, we see that any point at which
C{jjdz does not exist is a singular point. AIso, when fez) is multivalued, any
point which cannot be an interior point of the region of definition of a single-
valued branch of fez) is a singular point.
Points of the latter type are known as branch points. At such points the
first derivative may or may not exist, but it can be shown that a derivative of
sorne order will fail to exist. However, the characteristic feature of a branch
point is the property that if a circuit is described around a small simple closed
curve enclosing that point, but not enclosing any other singular points, then
the value assumed by the function after the circuit differs from the initial value.
In Section 10.3 we have seen that the function log z and functions of the
form Zk, where k is not an integer, each have a branch point at z = O. Since
(d/dz) log z = l/z is analytic elsewhere, this is the only singularity of log z in
the finite part of the plane. Similarly, since (d/dz)Zk = kZk-l, we see that a1so
Zk has no other finite singularity. Unless the real part of k is smaller than unity,
reference to Equation (47) shows that the first derivative of Zk is finite at z = O.
However, it ís clear that if successive derivatives are calculated, eventually a
power of z with a negative real part will be obtained (when k is not a positive
integer or zero-), and hence the corresponding derivative will no¡ exist at the
branch point. Ir is seen that ¡he functions log (z - a) and (z - ay both have a
branch point at z = a, with the aboye restriction on k.
Reference to Section 10.2 and 10.3 then shows that the elementaryfunctions
can have branch points only for those values of z for which a quantity raised lo a
nonintegral power vanishes, or for which a quantity whose logarithm is taken
vanishes or becomes injinite. It is assumed, for this purpose, that inverse trigo-
nometric and hyperbolic functions are expressed in terms of the logarithm, as
In Equations (55) and (56).
Thus, for example, it is to be expected that the function
fez) = (1 - Z2)li2 = (1 + z)1/2(1 - Z)I/2
has branch points at z = l and at z = -l. This is readily confirmed if we

notice, for example, that a circuit about a simple c10sed curve enclosing z = 1,
but excluding z = -1, will change the second factor but not the first one, and
hence will change the product. Similarly, the function fez) = log (z - Z2) can
be written in the form
fez) = log [z(l - z)] = log z + log (1 - z),
and we may conclude that this function has branch points at z = O and at
z = l. In the cases of the inverse circular and hyperbolic functions, reference
to Equations (55) and (56) shows that the functions sin- I z, cos- I z, cosh- I z,
tanh -1 z, and coth -1 z ha ve branch points at z = ± 1, whereas the functions
tan- I z, cot- 1 z, and sinh- 1 z have branch points at z = ±i. The expressions
for the first derivatives of these functions show that there are no other singular
points (in the finite part of the plane).
It should be noted that the criterion mentioned aboye applies only to the
elementary functions, and also that it expresses a necessary but not sufficient
condition. For example, the functions e lo", and (Z1l2)2 c1early do not have a
branch point at z = O, and expansion in a Taylor series about z = O shows
that the same is true of cos (ZI/2).
In working with a multivalued function (possessing one or more branch
points) it is usually desirable to specify one particular branch of the function
and to artificially prevent the possibility of
y transition from that branch to another branch.
Clearly, this result may be accomplished by for-
w = log z bidding curves from surrounding the branch
points.
x In the case of the function w = log z, we
have seen that we may imagi'~ the complex
plane to be "cut" along the entl. , positive real
axis, so that transition across this part of the
Figure 10.10 axis is impossible (Figure 10.10). In the resultant
10.9. Singularities oí Analytic Functioos 569
"cut plane" we then choose an appropriate branch of log z, say the branch
such that
logz = logr + iO p (O <:: Op < 2n),
which is a single-valued function of z, and which is ana/ylie everywhere exeepl
on Ihe eul. The "cut" or "barrier" could equally well be introduced along any
other ray or curve extending from Ihe braneh poinl z = O lO infinily, with a
correspondingly modified definition of the selected branch of the function.
Such a modification would be neeessary if, say,
that branch were to be analytic on any part of
the positive real axis; a more elaborate modifi- y
cation would be necessary to define a branch
which is analytic in the region indicated in w = zk
Figure 6.10 (page 294).
For the function w = zk, where k is non- x
integral, the principal value of O is rather fre-
quently defined by -n < Op <:: n, so that the
cut is taken along the negalive real axis (Figure
10.11). Figure 10.11
For such a function as
l i - z
w = tan - I Z = - log - -,
2i +
i z
with branch points at z = ± i, two cuts may be made along the imaginary axis,
from the two branch points outward to infinity, thus avoiding contours which
surround eilher of the branch points [Figure 1O.12(a)]. In this case it is readily
verified that a circuit which encloses bOlh branch points does not introduce a
transition from branch to branch. Thus, in place of cutting the plane as des-
cribed, we could merely introduce afinile cut joining Ihe Iwo braneh poinls, and
hence avoid contours which enclose only one of the branch points [Figure
10. 12(b)]. The method of cutting to be preferred depends upon the use to which
the corresponding branch is to be put. Further facts bearing on the nature of
the cuts are brought out in Section 10.10.
y y
w =tan- 1 z w=tan-1z
x x
-i -i
(a) (h)
Figure 10.12
Thus, in place of dealing with multivalued functions, we may always intro_

duce a suitably cut plane such that only one particular branch of the function
need be considered, that branch then being single-valued in the cut planeo
If it is oecessary to retain the possibility of transition from branch to branch ,
we may think of constructing a sequence of superimposed planes cut andjoined
along the cuts in such a way that, if a poiot starts on a given plane, moves
around a branch point, aod approaches its original positioo after such a circuit,
it moves across a cut fram the initial plane onto another superimposed plane
corresponding to a second branch of the fuoctioI1. For the function w = log z
an infinite number of such planes would be required, and the resultant configu-
ration would resemble an endless helicoidal surface. For the function w = Z'/2
only two such planes would be needed, the junctions being so arranged that
as a point describes a continuous contour surrounding the origin it moves
from the first "sheet" to the second after the first circui t, and returns from the
second sheet to the first after a second circuit. Such arrangements are known
as Riemann surfaces.
In the remainder of this work, however, we wiU think of the plane as cut in
such cases, so that all branch points are removed and the fuoctioos considered
can be taken to be single-valued. The remaining singular points of a fuoction
fez) are then points at which the derivative df/dz does not exist.
In particular, if fez) is analytic everywhere throughout sorne neighborhood
of a point z = a, say inside the circle C: I z - al = R, except al the point
z = a itself, then z = a is called an isolated singular point of fez). The results
of Section 10.8 show that a Laurent expansioo of the form
~
fez) = 2:::
n"'" _00
cn(z - a)" (O < Iz - a I < R) (98)
then is valido
Now if fez) were of bounded magnitude as z ~ a, it would necessarily
follow that all c's with negative subscripts in (98) would be zero. Thus, every-
where inside the circle C except at z = a, fez) would be expressible in the form
fez) = Ca + c,(z - a) + c 2(z - a)2 + ....
If it were also true that fea) = Ca, then fez) would be analytic everywhere
inside C. That is, fez) would become analytic at z = a if it were suitably defined
(or redefined) there. Such a point is often called a "removable singular poiot."
For example, the functions fJz) = z/z and f2(Z) = (sio z)/z are undefined at
z = O, but both functioos approach 1 as z ~ O. If we define f,(O) = 1 and
f2(0) = 1, then fJ and f2 are aoalytic at z = O (and elsewhere). If we were to
artificially define fJ (O) or f2(0) to be O, say, or to refuse to define f, (O) or f2(0)
at al!, thenfJ(z) or f2(Z) would have a "removable singularity" at z = o.
Henceforth, we will assume that all such "removable singularities" have
indeed been removed by suitable definition.
10.9. Singularities oC Analytic Functions 571
Apart from such artificialities, it thus follows that J(z) cannot be oJ bounded
magnitude near an isolated singular point. To illustrate such points, we may
notice that the function w = I/z has a singular point at z = O, the function
w = 1/(z2 + 1) has singular points at z = + i, and the function w = cot z has
singular points where z = kn (k = O, ± 1, ±2, ... ). These singularities are
examples of points known as poles. Specifically, we say that if J(z) is not finite
at z = a but if for sorne integer m the product
(z - a)m fez)
is analytic at ::. = a, then fez) has a pole at z = a.l If m is the smallest integer
for which this is so, the pole is said to be of order m.
Ir follows that any rational function of the form
(99)
where the numerator and denominator are polynomials of degree M and N,

respectively, without cornmon factors, can have only poles, such poles occur-
ring at points where the denominator D(z) vanishes. Thus, if z - a, is a simple
factor of the denominator, the product (z - a ,)w is analytic at z. a, and the
point z = a, is a pole of order one, or a so-called simple poleo If z = a, is an
m-fold zero of the denorninator, so that (z - a,)m is a factor of D(z) but
(z - a ,)m+' is not, then z = a, is a pole of order m.
Example. The function

Z3 - 2z + 1 Z3 - 2z +1
fez) = z' + 2z 3 + z = z(z + iP(z - i)2
has a simple pole al z = O and double poles at z = ±i because zf(z) is analytic at
z = O, whereas (z - i)2 fez) is analytic at z = i and (z + i)2 fez) is analytic at z = -i.
That is, each product listed possesses a derivative at and near the point noled. •
Functions other than rational functions may also have poles. Thus the
function
fez) = ese z
is not finite at the points z = O, ±n, ±2n, .... Considering first the point
z = O, we see that
zf(z) = ....,!-
sm z
is finite at z = O, since
lim~=lim
,-osmz ,-oz-z
3/3 z
.+z'/5!-···
=lim(l_z2+ ...
,-o 3!
)-'=1.
'
tIn accordance with the conventionjust iotroduced, it is meant he re that (z - a)m fez) becomes
analytic al z = a wheo it is defined to take on its limiting value at that poio!.
[The same result follows more easily by use of L'HospitaI's rule (Problem 49).]
Also, zf(z) has a derivative at z = O, since
d sin z - z cos z (z - z'/3! + ... ) - z(l - z2/2! + ... )
dz [zf(z)] = sin 2 z = (z _ z'j3! ~ ... )2
= z'/3+ ... = ~ + ...
Z2 + ... 3
near z = O, and hence (d/dz)[zf(z)] = O when z = o. Thus fez) has a simple
pole at z = o. For the purpose of studying a singular point z = kn we may
make the substitution
t = z - kn.
There then follows
(z - kn)f(z) = t = (_ly_t_.
SID ~l + kn) Slll t
Thus, since t - O as z - kn, we see that, except possibly for sign, (z - kn)f(z)
behaves near z = kn just as zf(z) behaves near z = O, and hence fez) has
simple poles at all the points z = kn (k an integer). From the expression for
I'(z),
I'(z) = _ cos z
S1n 2 z
we see that I'(z) exists at all other points, and hence these points are the only
singular points of the function in the finite part of the plane.
From Equation (98) we may deduce that fez) has a pole olorder m at z = a
if and only if fez) has a Laurent expansion (98) in the neighhorhood 01 z = a,
with e" = O when n < -m and with e_m =7= o.
In accordance with the convention adopted at the beginning of this section,
we say that "the function
fez) = ;o¡ zl
has a simple pole at z = -1" when we mean to assert that any braneh of fez)
has this property (presuming that the relevant branch cut does not pass through
the point z = -1). However, the function
fez) = log z
z
has a branch point (not a simple pole) at z = O, SlDce no branch of z/(z) is
analytic at z = o.
AH isolated singular points which are not poles (and are not removable)
are called essential singular points. Thus, when fez) has an isolated singularity
at a point z = a, that singularity is essential if the derivative dfldz does not
exist at z = a and if there is no integer m for which
d
dz [(z - a)mf(z)]
10.9. Singularities oC Analytic Functions 573
exists at z = a. It should be noted that a branch point is not an isolated sin-

gularíty, sínce any circle with center at that point must also contain points on
sorne branch cut passing thr o ugh that point, if the branch considered is to be
single-valued.
As an example of an essential singular point, we notice that for the function
fez) = e l / '
the derivative
/,(z) = _ _
1 el;'
z'
exists everywhere except at the origino Thus the orIgm is an isolated singular
point. For real values of z, the function f(x) = e l X is seen to approach zero
as x ~ O from negative values, and to become infinite as x ~ O from positive
values. Thus we see that el ;' approaches zer o if z ~ O along the negative real
axis and that the absolute value of el /: increases without limit as z ~ O along
the positive real axis. Since the same is true for z m e 1/:, for any value of m, we
see that zmf(z) is not analytic at z = O for any value of m, and hence the
function ha s an essential singularity at z = O.
To investiga te in more detail the behavior of fez) near the origin, we write
z = re iO and so express f(z) in the form
fez) = e ( l / de- i8 = e(Iír")(co s8 -J $ia8 1
= e (Co,.) /,[cos ei~ e) - i sin (Si~ e) J.

In particular, we have
I fez) 1= I el /:I = e(CO'.)," .
It follows that if z ~ O along any curve for which cos > O as r ~ O, the e
absolute value of fez) increases without limit, whereas if z ~ O along any
curve for which cos e < O as r ~ O, then fez) ~ O. Now suppose that z ~ O
along a curve r = c- l cos e,
that is, along a circle of diameter 1/1 e I tangent
to the imaginary axis. For such an approach we have always I fez) I = e and C
,
hence I fez) I can be made to take on any positive vall/e as z ~ O along a properly
chosen circle of this kind. On such a circle we have also
f(z) = e'[cos (e tan e) - i sin (e tan e)] .
Now as z ~ O along this circle the angle e tends toward n/2, and hence the
argument of the circular functions increases without limit. Consequently, it is
seen that fez) takes on al! complex values with absolute value e' infinitely
many times. Thus it follows that in any neighborhood of z = O the function
el !' takes on aH values except zero an infinite number of times!t Further, the
tThis situation is also evideneed by the raet that the equation el ! : = pe 'v has infinitely many
solutions, of the form
1
z -- log
~~-.~~,-~~
p + i(rpp + 2k1t)
(k = O, ±l, ±2, ... ),
if p :,l= O.
574 Fuoctions of a Complex Variable
exceptioDal zero value is approached as z ~ O aloDg the negative real axis, or

aloDg any curve which passes through the origin along a tangeDt in the second
or third quadrant.
Picard has shown that any analytic function having an isolated essential
singularity behaves iD this remarkable way near the singularity, all real aDd
imagiDary values being approached and all values except one beiDg actually
taken on infinite/y often iD any arbitrarily small regioD iDcluding the singularity.
From Equation (98) we may deduce that fez) has an iso/ated essentia/
singular point at z = a if and on/y if fez) has a Laurent expansion in the neigh-
borhood of z = a with infinite/y many negative powers of z - a.
In addition to the isolated siDgularities and braDch points, still other types
of siDgular POiDts may exist. For exarnple, iD the case of the functioD
fez) =
1i
2ni J e, IX -
dlX
z =
{IO iDside Ca,
(00)
outside Ca,
mentioDed iD Section 10.8, all POiDtS on the closed curve Ca are singular poiDtS.
ODe rnight speak of such POiDtS as "artificial boundary points" siDce they
comprise a curve across which a transition occurs from one dornain, inside
w hich a first aDalytic fUDctioD f, (z) is defiDed, to another domaiD iD which a
second analytic fUDction f2(z) is defined. In this example there is no difficulty
in extending the domain of definitioD of either f, (z) or f2(z) across Ca, so that
POiDtS on Ca are DOt singularities for either of the two aDalytic functions f,
and f2.
However, there exist fUDctions, defined on one side of a curve C, such that
their definitions cannot be extended aDalytically across C. As aD example, ir
fez) is defined by the series
+ Z2 + Z4 + z· +
~
fez) = L: Z2"
11"00
= Z
wheD I z I < 1, it can be showD that the series diverges everywhere on the circle
I z I = 1 and that I fez) I is unbounded in the neighborhood of each point on
that circle. Thus it is impossible to extend the definition of fez) in such a way
that the result is analytic at aDy points OD the unit circle. (See Problem 66.)
Such a curve is called a natural boundary for the relevant function, and points
on such a curve usually are referred to as (Donisolated) esseDtial siDgularities.
When branches of a rnultivalued function are defined, the pOiDtS OD branch
cuts are also singular points, but are rather artificially so since (except for the
branch points themselves) they would no longer be exceptioDal ir the branches
of the parent fUDction were redefined.
In addition, it may happen that isolated siDgular points cluster about a
POiDt z = a, in such a way that there are iDfinitely many such points in every
circle with ceDter at z = a, no matter how small the radius. TheD z = a also
is a singular point and is iDcluded in the category of nonisolated "essential"
singularities. The function fez) = tan (J / z) has such a singularity at the origiD .
10.10. Singularities al Infinily 575
10.10. Singularities at Influity. To study the behavior of a function fez)

for large values of 1::: l. we may make the substitution
=_1.
I (101)
(
The function fez) is then transformed to a new function gel),
g(t)=f(+)· ( 102)
Now as a point z moves indefinitely far from the origin in any direetion, so that
Izl ~ co, the corresponding point t approaches t = O. It is convenient to say
that the point l = °
corresponds ro a "point at infinity" in the complex plane
which may be denoted by z = z-. such that this point is approached by any
point z receding from the origin indefinitely far in any direction. Thus we
"close" the complex plane by adding the point z~ "at infinity." The result is
often known as the extended (or c/osed) eomp/ex planeo lt is then natural to say
that fez) is analytic at z_ if g(l) = f(lll) is analytic at ( = 0, and sirnilarJy that
if g(t) has a particular type of singularity at / = 0, then fez) has the same type
of singularity at Zw
If g(t) = f(l/t) is analytic at l = 0, then it can be expanded in a series
f(-I) =
t
Aa + Al/ + Azt Z + ... = f:o A"/"
n=
in sorne circle about / = 0, and conversely. Hence, by replacing / by I/z, it

follows that fez) is ana/y/ie a/ z_ if and on/y if it ean be represen/ed by a series
of /hcform
fez) = Aa + .:iJ.
z
+ Az
Z2
+ ... = :t An
11= o Z"
(103)
for values of z ou/side a sufficiently large circle with center at z = O. In particu-

lar, fez) must approach a limit f( co) as z ~ z-. that is, as I z I ~ co in any way,
and Equation (103) shows that
Aa = f(co). (104)
We notice also that if f(1ll) is analytic at / = 0, then
~f(+) = -/~f'(+)
must exist at l = 0, and hence z2(dfldz) muSI exiSI a/ z_.
If fOil) has a pole of order m at t = 0, then I mf(1I/) can be expanded in
a power series in / converging near / = 0, and conversely, and hence near / = °
we may write
where Aa "'" O. Thus we conclude that fez) has a pote of order m at z_ if and
onJy if we can write

fez) = A ozm + A,zm-' + ... + Am + A m+,
z
+
-L; (A o T'= O), (105)

n= o
for values of z outside a sufficiently large circle with center at z = O. In particu_
lar, f(z) jzm must approach a nonzero limit as z ~ z_, and this limit is identified
with the coefficient Ao in (105),
Aa = lim
%-:00
-k fez).
Z
( 106)
Also, sínce (djdt)[tmf(ljl)] must exist when t = O, it follows that the expression
-z2~[f(z)J = ~f(z) _ _ I_df(z)

dz zm zm-¡ zm-2 d::
must tend to a limit as z ~ Zw

From (105) we conclude that any poJynomiaJ of degree m has a paje of arder
m al Zw lf we consider the general rationaJ function
fe -") = a o "T'I a,z +.

.. + aNz, N
(107)
bo T' b lZ '
--t- .. . . .. b MZ W 1
aüt N + a¡t N - 1
+ ... + a N tl\<J-N
bot M + b,t M , + ... + b M •
Thus it follows that f(ljt) is analytic at t = O if M >- N and that f(lft) has a
pole of order N - M at t = O if N > NI. The same is thus true for fez) at z~.
Since fez) clearly is anal y tic at aU finite points except at zeros of the
denominator, where poles exist, it follows that any rationaJ function possesses
no singuJarities other than pajes. If we count a pole of order k as equivalent to k
simple poles, we see that since the denominator has M linear factors, the
multiplicity of finite pajes is M. Similarly, the multipbcity of finite zeros is N.
If M > N, the function f(l ft) has a zero of order M - N at t = O, and hence
fez) has a zero of order M - N at Zw In this case there are exactly M poles
and M zeros (counting multiple poles and zeros separately). Similarly, if M < N,
there are N poles and N zeros. In any case, for a rationaJ jimcl,ion th e totaJ
muJtiplicity of pajes is equal to the totaJ muJtiplicity of zeros in the extended
pJane, that is, when the point z~ is taken into account.
Since the expansions of e', sin z, cos z, sinh z, and cosh z involve infiniteJy
many positive powers of z, and converge for all finite va lues of z, reference to
the criterion (105) shows that, although these functions have no finite singu-
Jarities, they each have an essentiaJ singuJarity at z~ . In the case of the exponen-
tial function this situation may also be seen by notjcing that e' behaves for
large Izl as e' !: behaves for small Izl.
10.10. Singularities al Infinity 577
For the function fez) = log z there foIlow s f (l l t) = log (l I t ) = -Iog t.

Hence it follows that fez) = log z has a branch point at z ~ as well as at z = O.
A similar statement applies to fez) = Zk, where k is n o nintegral. The " cuts"
introduced in Section 10.9 for these functions can be con s idered as j o ining the
two branch points.
We notice that a small circle surrounding the point t = O correspond s ro
a /arg e circle with center at z = O in the complex plane. Thus, just as we m a y
consider a "circle of zero radius" as enclosing only the origin, we may in a
similar sense consider the exterior of a circle C_ "of infinite radius" as consist-
ing only of the point z=- Consequently, the exterior of a closed con tour C may
be considered as composed of the region between C and the infinite circle C _
rogether with an added exterior point z_ at infinity.
AIso, a positive circuit around thefinite area el1closed by any simple closed
contour C can be considered also as a negative circuit aroul1d the infinite area
outside C, in c /uding ¡he point Z w
If f ez) is aoalytic on C and at all finite points outside C , it follows that C
aod C~ are equi valent contours, and hence
f e
fez) dz = f c~
fez) dz. ( 108)
The right-hand integral may now be considered as taken along a negative

contour enclos ing the exterior of C _ , that is, enclosing only the immediate
neighborhood of the point z =' However, even though fez) be ana/y tic at z_ ,
this integra/ may not vanish . It can be shown to vanish if ::;2f(z) is analytic at
z_ (see Problem 84). Thjs fact leads to the useful result that
fef(z)dz =0 (109)
in the case when z2f(z) is ana/y tic on and outside C and at z =, as well as in the
case when fez) is an a /y tic on and inside C.
If outside a closed contour C no points, including the point z=, are branch
points , and if we think of a closed circuit around C as surrounding the exterior
of C, we may expec t that a funct ion fez) wiU return to its initial value after
s uch a clo sed circuit, wbether o r not there are finite branch points inside C.
Tbus, for the functioo
t an - - z
l 1og -.i -
f()
z = 1
z = -.
21 / + z
-
considered in Section 10.9, we obtain
f(_I)
t
= J.,.¡oo t
2/
+ ~.
'" t - /
Hence, since t = O is nO! a branch point for f(l / t), it follows that z_ is not a
branch point for fez). In the first method of cutting the plane for tbis function
[Figure 1O.12(a)], the cut can be considered as joining the two branch points
z = + i along an infinite segment passing through z=. However, since z~ is not

a branch point, closed circuits surrounding z~ need not be prohibited, and a
finite cut joining the branch points [Figure 10.12(b)] is equally satisfactory.
That is, we need only prohibit closed circuits which enclose only one of the
branch points. It must foUow that a closed circuit surrounding both branch
points will return tan -, z to its initial value. As was stated earlier, this sta te-
ment can be directly verified (see Problem 57).
It may happen that z~ is a branch point on certain branches of a multi-
valued function but not on other branches, as for the function
.
sin-' z + i 10<>
b
z
(see Problem 72).
10.11. Significance oC Singularities. An important theorem, due to Liouville,

states that a function which is analytic at al! finite points and at z~ must be
constant. To establish this theorem, we consider any two points z, and z, in
the complex plane and apply Cauchy's theorem (Section 10.6) to express fez ,)
and f(z2) in the form
f(z¡) = _1_. i fea) da, (110)

2m Je a - z,
where e is a circle of radius R, with center at the origin, including the points
z, and Z2' By subtraction we obtain also
(111 )
For points on e we have

lal =R. (112)
Sincef(z) is assumed to be analytic everywhere (including z~), it must be bounded
everywhere. Denoting this bound by K and making use of the inequality (lO),
we obtain, for points on e,
fea) I< K - K .
I(a - z,)(a - Z2) = (Ial - Iz,I)(lal - IZ21) - (R - Iz,I)(R -IZ2)
Hence, making use of the A1L inequality (80), we obtain from (111)
1 f(z2) - fez,) < 1 1 Z2 - z, 1 K . 2nR

211: (R - 1 z, I)(R - 1 z 21)
<I Z 2 - Z ,1 K (113)
R
Since fez) is analytic everywhere, we are permitted to let the radius R of e

increase without limit. But since the right member then tends to zero, it follows
that the left-hand member must be zero for any two points ZI and Z2' Hence
fez) is constant, as was to be shown.
10.11. Signi6cance of Singularities 579
This theorem has many important consequences. In ¡Ilus tration, we have

seen that any polynomial of degree m has no singularities except a pole of
order m at z~. We can now show that, conversely, a funetion having no s ingu-
larities other than a pole of arder m at :!~ is neeessarily a poly nomial of degree
m. For if fez) has a pole of order m at z~ . then for large I z I it must be repre-
sentable by a series of form (105). Hence fez) must be expressible as the sum
P m(z) + g(z ) of a polynomial
Pm(z) = Aozm + A ,zm - ' + ... + Am ( 114)
and a function whose expression, for large I z 1, is of the form
g(z) = Am+ ,
z
+ A m,+2
z-
+ .. ' . (115)
But since the function g(z) must then be analytic e verywhere, and since it
vanishes as I z I ~ 00, it follows from the preceding theorem that g(z) must be
zero and hence fez) must be a polynomial, as stated .
Further, we can show that if a fillletion is analy tie in the extended plane
except for afinite number of poles it must b e a rational funetion; that is, it must
be the ratio of two polynomials. For suppose that fez) has poles of order k n
at the N finite points z = Zn (n = 1,2, ... , N). Then, from the definirion of a
pole, it follows that the function
F(z) = (z - z,Y'(z - ZZ)k , .•• (z - z,vYNf(z) (116)
is analytic at all finite points. The coefficient of fez) in (116) IS a polynomial
of degree K, where
K = k, + k 2 + ... + k N ,
and hence it has a pole of order K at Z w But since we have supposed that fez)
is either analytic at z~ or, at worst, has a poi e at z ~ , it follows that F(z) has, at
worst, a pole at Zw Now since F(z) is analytic at all finite points, the preceding
theorem shows that it must be a polynomial. Thus fez) must be the ratio of
this polynomial to tbe polynomial which multiplies it in (1]6), as was to be
shown .
These examples ill ustrate the fact that analytic functions are, to a certain
extent, essentially characterize d by their singularities . Those functions which
are single-valued, and analytic at a/lfinite points, are known as integralfunctions
or as en tire functions. Such a function is either analytic also at z~ , in which
case it must be a constant; or it may have a pole at z~ , in which case it must be
a poly nomial; or, otherwise, it must ha ve an essential singularity at z~ , in which
case it is known as an integral transcendental function. The functions er, sin z,
and sinh z are examples of such functions . We have seen that if an analytic
function is expanded in powers of z - a, the circle of convergence extends at
least to the nearest singularity. It follows that the Taylor series representation
of any integral function converges for all finite vall/es of z. Conversely, if a power
series converges for all finite values of z, it defines an integral function.
580 Functions oí a Complex Variable
10.12. Residues. Suppose that the analytic function f(z) has a pole of order
m at the point z = a. Then (::: - a)mf(:::) is analytic and hence can be expanded
in the Taylor series
(z - a)mf(z) = Ao + A 1(::: - a) + ... + Am_l(z - a)m-l + Am(z - a)m + "',

(117)
where ( 118)
in accordance with the results of Section 10.7. This series will converge within
any circle about z = a which does not include another singularity. If z 0;>'= a,
Equation (117) can be rewritten in the form
( 119)
Now let Ca be any simple closed contour surrounding z = a which lies inside
the circle of convergence of (117) and which is such that f(z) is analytic inside
and on Ca' except at z = a. If we integrate (I19) around this contour and
review Equations (78a, b), we obtain merely
fCo
f(z) dz = 2niAm_.,
the only term contributing to the integration being Am_1!(z - a). The same
relation then continues to hold when Ca is deformed into any equivalent con tour,
which need not lie inside the circle of convergence of (117).
We call the coefficient A m- 1 tbe residue of f(z) at z = a and denote its
value by Res (a), the function f(z) being understood. If a more explicit notation
is desirable, we will denote the residue of f(z) at z = a by Res {f(z); a}. Thus,
if f(z) has a pole of order m at z = a, then
f Co
f(z) dz = 2n1 Res (a) (120)
with Res (a) = (m _I I)! dzm-I

- [d
{(z -
m I
a)mf(z)} ] z~a' ( 121)
where Ca is a simple closed contour enclosing z = a but excluding al! other sin-
gularities of f(z).
We notice that Res (a) is the coefficient of 1!(z - a) in the Laurent ex pan-
sion of f(z), in powers of z - a, which is valid near z = a. The value of the
residue can be determined from this fact or can be determined directJy from
(121).
In the case of a simple pole (m = 1), Equation (121) gives
m = 1: Res (a) = [(z - a)f(z)lz~a = lim (z - a)f(z). (122)
10.12. Residues 581
In this case, if fez) is expressed as the ratio

__ N(z)
fe") - D(z)' (123)
where N(a) is finite and nonzero, then it follows that D(z) must vanish at z = a
in s uch a way that D(z)/(z - a) approaches a finite limit a s z - > a. Tbus, if we
write (z - a)f(z) = N(z)[(z - a)/ D(z)], we may use L'Hospital's rule (Problem
49) to evaluate the limit indicated in (122) in the form
_ N(a)
m = 1: Res (a) - D ' (a)' (124)
Wben D(z) is a polynomial, z - a must be a factor, and bence (122) is readily

eva1uated by merely deleting tbe factor z - a in the denominator and setting
z = a in the remaining ratio.
This procedure is applicable, however, only in the case when z = a is a
simple pole. Otberwise, use sbould be made of Equation (121) un1ess tbe series
(119) is easily obtained, in which case the residue is read directly from tbe
series.
Example l. We consider first the function

1 1
fez) = Z2 + 1 = (z + i)(z - i)' (125)
There are simple poJes at z = ±i with residues

Res (i) = Jim (z - i)f(z)
;:-.1
= (_+1.)
Z 1 : - j
= 21.
1
and
Res (-i) = (_1_.)
z - 1 , ~ _,
= -J.." 21
[rom (122). 1f we use (124), we have, with N = 1 and D = Z2 + 1,
Res(i) = (21z ) :.:,. = 21 ,., Res(-i) = (J.-)

2z ::~-(
= -l."
2z
as befo re .
Example 2. For the function

•
1 1 (126)
fez) = Z3 _ Z2 = Z2(Z - 1)'
there is a double pole at z = O and a simple pole at z = 1. At z = 1 we obtain, froro

(122),
Res (1) = (-\-)

z := 1
= 1.
At z = O we use (121) with a = O, m = 2, and find
[,1zC ~ 1)1-0
Res (O) = = -1.
•
582 FunctioDS of a Complex Variable
Example 3. For the function

fez) = Sin Z - z ( 127)
Z6
a singularity in the finite part of the plane can occur only at z = O. By expanding the
numerator in powers of z, we may write
fez) = (z - z3/3 ! + z' /5 ! - z7/7 ! + ... ) - Z
Z6
or fez) = - 6~3 + 110z - 5~40 + ... (128)
Thus fez) has a pole of order three at z = o. The residue is merely the coefficient of
ljz in (128),
1.
Res (O) = _
120
It is easily shown that if fez) has a pole of order m at z = a, the value of

•
the right-hand member of (121) is unchanged when m is replaced by any integer
A1 such that Jl¡1 > m. (See Problem 79.) That is, if the order of the pole is over-
estimated when (121) is used, the correct value of the residue will still result.
This fortunate state of affairs can be verified in the case of Example 3 by using
(121) to calculate Res (O) with m = 6, in place of the correct value m = 3
which is not readily determined by inspection, and which would lead to a more
involved calculation.
If fez) has an isolated essential singularity at z = a, but is single-valued
in the neighborhood of that point, then reference to (97) shows that (120) again
applies, with
Res(a)=c_l' (129)
where e _ 1 is the coefficient of (z - a)-I in the Laurent expansion of fez) which
is valid in the immediate neighborhood of z = a. It has been seen that this state-
ment is true al so when z = a is a pole, so that, in fact, it applies when z = a
is any isolated singularity of a single-valued function. However, the alternative
formula (121) applies only when z = a is a poleo

fez) = ze 1/, (130)
has an isolated essential singularity at z = O, but has no other singularity at a finite
poin!. From the expansion
fez) = z ( 1 + -z1 + -
2! -
1
Z2
+ .)
.. = z + 1 + -2!1 z + (z *- O),
there follows
Res (O) = i-.
Suppose now that fez) is analytic on a simple closed curve e which bounds
•
a finite region inside which fez) is single-valued and has only isolated
10.13. Evaluation of Real Definite Integrals 583
singularities, at a finite number of points y

z = al' a 2, ... , Gn · We endose these points
by small nonintersecting simple closed curves
CI' C 2 , • . • , Cn' each of which lies inside C
and endoses only one singularity. Then, by
introducing a crosscut from each curve C k
to C, a simply connected region cR is obtained
inside which fez) is analytic (Figure 10.13).
Thus the ¡ine integral of fez) around the
complete boundary of this region vanisbes. x
Noticing that the integrals along the crosscuts Figure 10.13
cancel, since fez) is single-valued, and that
the integrations taken around the smaU contours are in the negative sense, we
deduce that
t fez) dz - [fe. fez) dz + t, fez) dz + + p fez) dz]

c"
= O. ( i31)
But since the integral taken in the positive sense around C k is 2¡¡;i times the
residue of fez) at z = a k , Equation (131) leads to the result
¡ fez) d:: =
re 2¡¡;¡, i:
k=l
Res (a k ). ( 132)
Thus, if fez) is analytic inside and on a simple e/osed curve C, except at a

finite number of interior isolated singularities, then fc fez) dz is given by 2¡¡;i times
the sum of the residues of fez) at those points. This result is known as Cauchy's
residue theorem.
10.13. Evaluation oí Real Definite Integrals. The residue theorem is useful

in evaluating certain types of real definite integrals. In this section a few examples
are presented.
Example 1. Any real integral of the form

2.
1 =
fo R(sin e, cos e) de, (133)
where R is a ratiana! function of sin e and cos e which is finite for all (real) va!ues of
e, can be eva!uated by residue theory as follows. If we make the substitution
z = ed}, dz = ¡e'· de (134)
there follows a!so
. Z2 - 1 Z2 ...!... 1
de = dz sme= 2'IZ , cose= )' . (135)
IZ ' _Z
Thus R(sin e, cos e) de takes the form F(z) dz, where F(z) is a rational function of z.
Since z describes a positive circuit around the unit circ!e el in the complex plane as
S84 Functions of a Coruplex Variable
e varies from o to 2n, the integral (133) takes the forro
1 =!Tel F(z) dz = 2ni ~ Res (ak),

(136)
k
where the points a. are the poles of F( z ) inside the unit circle.
For example, with (134) and (135) the integral
1 -
,-
'1· o A +
de
B cos e (Az > B', A > O) Cl37)
takes the forro
1, = I F(z) dz, F(z) = Bz ' +

2/ i
2Az + B
Je ,
The poJes of F(z) occur when
_.d.. +../Al - B'
- B B
and z = a2 =
We see that, since a,az = 1, one poJe (that at a"

y since A is assuroed to be positive) is inside the unit
circle e" whereas the second pole is outside e,
(Figure 10.14). Using (124), we obtain
2/ i 1
Res (a,) = 2Ba, + 2A - i,../ AZ _ B2
1 x
1 , = 2ni Res (a,) =,../ (138) 2n
B' A2-
The restriction A2 > B2 is necessary in order tbat
Figure 10.14 the integral (137) ex ist. •
Examp/e 2. We next indicate how con tour integration may be used to evaluate an
integral of the forro
r~ f(x ) dx, (139)
where f(x) is a rational fimction, whose denoroinator is of degree at least two greater
than the numerator, and which is finite for all (real) values of x . To illustrate the pro-
ceduce, we consider the in tegral
I = L-_ '1-+.:;x:..2-x"74 dx. (140)
We fust write
Z2
fez) = 1 + Z4 (141)
and consider the result of integratíng fez) aTOund the con tour indicated in Figure
10.15, consistíng of [he segment of the real axis extending from -R to R and the
10.13. Evaluation of Real Defioite Integrals 585
-R R x
Figure 10.15
sernicirc1e C R in the upper half-plane. For any value of R there follows
J R 1
-R
~2 x 4 dx + 1 eR,
fez) dz = 2ni ~ Res (ak), (142)
where the points ak are the poles of fez) ioside the con tour. As R increases without
limit the fust integral on the left approaches the required integral. Also, eventuaUy all
poles of fez) in the upper half-plane are absorbed into the contour and hence contri-
bute residues to the right-hand member.
We show next that the integral taken along C R tends lo zero as R - ca. On C R
we ha ve 1 z 1 = R and hence a1so, using (lO),
Z2 I 1Z 1 R2_ 2
1f(z) 1 = I 1 + Z4 -< 1214 _ 1 = R4 _ 1 = M (R> 1).
The length of C R is L = nR, and hence by Equation (80) there follows
11. f(Z)dZ\ -< ML = R~~ 1 (R> 1). (143)
Hence the integral along C R tends to zero as the radius R increases indefinitely.
Thus, proceeding to the limit as R --> ca, we find that Equation (142) gives
l = [~l .:;.2;4 dx = 2ni ~ Res (ak), (144)
where the points ak are the poles of fez) in the upper half-plane, namely, the two
vaJues of (_1)1/4 which have positive imaginary parts,
al =e ni/4 , (145)
Also, making use of (124), we obtain
Thus (144) is evaluated in the form

(1
J
~ 2
dx = 2ni
__ 1 +x x4
_e-nI/4
4 (146)
It should be noticed that the crucial step in this procedure consists of showing that
the integral along C R tends to zero as R --> ca. Considering the more general case
described in connection with (139), we see that, if the denominator of f(x) is of degree
at leas! two greater than the numerator, then along C R [he maximum value JH of
I fez) I is at worst of the order of 1/1 ,.1 2 ~ 11R2. Since the lenglh o f C R is L ~ nR, it
follows that ¡VIL is at worst of the order of 1I R and hence, as in the example, the in-
tegral tends to zero as R ~ =. Thus in such cases we have
t=- f(x) d" ~ 2ni t: Res (ak), (47)
where the points ak are the poles of fez) in the upper half-plane. lt is easily seen that if
f(x) is a rational function, the conditions specified are necessary in order that f:_ f(x) d:.:
existo Thus (147) is true if f(x) is a rational function and if the integral exists.
We note that if f(x) is an even function, there follows
Jor- f(x) d:.: ~ 21 ¡-__ f(x) dx.
Thus , for example, Equation (146) also gives
fl X2
+
n
x4 dx ~ 2..y"2· (148)
Example 3. If we attempt to apply the method of Example 2 to the evaluation of an

•
integral of the form
f:- f(x) cos mx dx or f:- f(x) sin mx dx, (149)
where f(x) is a rational function of the type described, we encounter difficulty when
we attempt to show that the integral of fez) cos mz or fez) sin mz along C R in Figure
10.15 tends to zero as R ~ =. Tbus, if we notice first that on C R we have z ~ Re'o
and hence
lefm:1 = I e'''' Rff l8 I = !e,mR(COS(J+lsin9) 1 = e-mRsio9, (150a)
le-Im:l = [e-imRflf91 = [e-imR{cos9+I.llln8JI = emRslo9, (150b)
\ it follows that on C R the functions

el,": + e- 1m : eimr _ e- imz
cos mz ~ 2 sin mz ~ - - ' 2 " i - - (151)
increase exponentially in magnitude as R ~ co because of the presence of the tenn

e- imr when m > O and beca use of the tenn e'mz when m < o. Hence, the maximum
value of the integrand is unbounded as R --> co, and the integral along C R cannot be
shown to tend to zero.
However, ifwe notice from (150a) that, when m > O, I e'mr I decreases exponentialfy
on C R as R --> oó in the upper half-plane (sin (J > O), we may avoid the difficulty noted
by considering the integrals (149) as the real and imaginary parts of the integral
f- e'm·'f(x) d:.: (m > O). (152)
Then, if the maximum value of I fez) I on C R is, at worst, KI R2, we have, for points on
Ie'm, fez) I ~ e-mR'ia81 fez) I - O).

10.13. Evalualjon of Real Definjte Integrals 587
Thus the integrand is bounded by !vI = K/R' and we have
IIR e'm, f(z) dz I-< ML = n ~

so that the integral along C R again lends to zero as R - =.
Hence we conelude that if f(x) is a rationa/ function whose denominator is ofdegree
at leas t two greater than the numerator, and which is finite for a/l (real) values of x, then
L~~ e'mx f(x) dx = 2ni ¿: Res [e'mz f(z); ak} (m > O), (153)
where the points ak are the po/es of f(z) in the upper half-plane. The inlegrals (149) are
obtained as the real and imaginary parts of this result.
The theory lo be developed in the following seelion shows in fael Ihat, if m > 0,
the degree of the denominator need be only one greater than that of the numerator.
In illustration, 10 evaluate the integral
e/mx
J_~
~
a2 + X2 dx
(m > 0, a> O),
where a and m are real, we notice that f(x) = 1/(a 2 + X2) is a rational funetion of the
required type. The only pole of f(z) = 1/(a 2 + Z2) in the upper half-plane is at z = ia.
Sinee the residue of e'mz f(z) at z = ia is given by
r e/+mz Z2; la. } eimr ) 1 ma

Res la 2 = (z + ia ;r - {a = 2ia e - ,
we obtain the result
~ -::>::.e'.,.·m_.,= dx
J_= a
= _n e-m. (m > 0, a > O). (154)
2
+ X2 a
By taking real and imaginary parts, we obtain
J
~ eos mx d - !E..e-m. (l55a)
_00 a 2 '+ X2 x - a '
sin mx dx =
[ _ooa 2 +x2
°' (155b)
when m > 0, a> O. JI is se en from the fonn of the result that the restrietion m > °
ean be removed in this ease if we replaee e-m. by e- Iml •. •
Example 4. When we are to evaluate an integral of the form
L=~ f(x)e 'mx dx,
where again f(x) is a rational funetion whose denominator is of higher degree than its
numerator and whose zeros are nonreal, but where now m < 0, the melhod of Example
3 may be modified by dosing Ihe eonlour in the lower half-plane. In lrus ease, a posi-
tive progress atong the real axis implies a negalive (clockwise) eireuil about the closed
588 Functioos of a Complex Variable
path and there follows accordingly
Cm < O), (56)
where ¡he poinls ak are Ihe po/es of fez) in ¡he lower half-plane.
Whereas this modification in procedure couJd be avoided here by merely replacing
x by - x in the original integral to yield the equivalent form
t-- f( -x) e- 'm .. dx,
where now -m is positive, the use of the modified approach sometimes is more COn-
venienL
In illustration, we consider the evaluation of the real integral
1 = [ sin ¡x sin ax dx (a >- 0, b > O), (157)

. o X2 + b 2
which can be interpreted as the Fourier sine Iransform of (sin ax)/(x 2 + b 2 ), expressed
as a function of I (see Section 5.15). Since the integrand is an even function of x, we
have also
1 = ~
2
f- sin Ix sin ax dx
X2 + b 2
and, since also
sin Ix sin ax = ~ [cos X(I - a) - cos X(I + a)]
-- 1.-
2 Re [e{.-.:(r-a) _ eix(r+o)],
we can write
1 = 4 Re
1 [f- __ x'
elx"-Q)
+ b2 dx -
f -
__ x'
e'x"+Q'
+ b2 dx
]
1
= 4Re(I, - 1,). (158)
Dealing first with 1" we see that when [ > a the formula (153) corresponding to
a con tour closed in the upper half-plane applies and there follows
lZÜ O
1} = 2ni [ e - '] = n
_e-b(r-a) (t> a).
2z , - lb b
However, when I < a, c10sure in the /ower half-plane is needed and hence we use (156)
to obtain
11 = -2ni [ e1r"-Q']
-- -:- -n eb(r-o) (1 < a).
2z r--Ib b
When I = a, eilher method of c10sure is permissible, and hence
1, = 2ni[~J
2z ,-lb
= -2ni[~]
2z ,--lb
= !!.-
b
(1 = a).
10.14. Theorems on Limiting Contours 589
These three results can be combined into (he form
Consequently, there follows also
and hence we have the result
1 = J~ sinX,Ix+sinb2ax d x = !!...(
4b e
-'I<-al
(b > O), (159)
o
¡he condition a > O now being unnecessary since both members of (159) are odd func-
lions of a . •
10.14. Theorems on Limiting Contours. In many applications of contour

integration it is necessary to evaluate the Iimit of the result of integrating a
function of a complex variable along an arc of a circle as the radius of that
circle either increases without limit or tends to zero. In this section we collect
and establish certain general results of frequent application. First, however, it
is convenient to introduce a useful definition.
Ir, along a circular arc C, of radius r, we have 1 fez) 1 <: K" where K, is a
bound depending only 00 r and hence independent of angular position on C"
and if K, ~ O as ,. ~ 00 (or r --+ O), then we will say that fez) tends to zero
uniformly on C, as r ~ 00 (or r ~ O). Thus, for example, if C, is a circular
arc with center at the origin and fez) = z/(z' +
1), we have
1f(z) 1= Izl <: Izl _ r (r:> 1)

IZ' + 11 == Izl' - l - r2 - l
on C,' if use is made of the basic inequality (lO). Hence, if we then take
K, = r/(r 2 - 1), we conclude that here fez) teods to zero uniformly on C, as
r --+ oo. AIso, we may take K, = r/( l . - r') when r < l to show that the same
is true when r --+ O.
In particular, any ratfonal fimction (ratio of polynomials) whose denomi-
nator is of higher degree than the numerator tends uniformly to zero on any
C, as r ~ oo. This follows from the fact that then 1 z 11 fez) I tends to a limit
(which may be zero) as I z I = r --+ 00 , and hence is bounded by some constant
k when r is large (say, r :> ro), so that we may take K, = k /r (when r :> ro)'
The following theorems now rnay be stated :
Theorem 1. Ir, on a circular arc C¡.. with radius R and center at the onglD,
zf(z) ~ O uniformly as R --+ 00 , then
tiro
R-oo
J fez) dz
CR
= O.
590 Fonctions of a Complex Variable
Theorem 11. Suppose that, on a circular arc e R with radius R and center at
the origin, fez) ~ O uniforrnly as R ~ =. Then:
1. (m> O)
if el< is in the first and/or second quadrants.t
2. lim
R-->oo
feR
e- Imz fez) dz = O (m > O)
if e R is in the third and/or fourth quadrants.
3. lim r- e mz fez) dz = O (m> O)

R-oo JCR
if e R is in the second and/or third quadrants.
4. lim
R-oo
feR
e- mz fez) dz = O (m > O)
if eR is in the first and/or fourth quadrants.
Theorem 1IJ. If, on a circular arc e p with radius p and center at z = a,

(z -a)f(z) ~ O uniforrnly as p ~ O, then
lim
p-+Q
fep
fez) d= = O.
Theorem IV. Suppose that fez) has a simple pole at z = a, with residue Res (a).
Then, if e p is a circular are with radius p and center at z = a, intercepting an
angle rx at z = a, there follows
lim
p-o
fCp
fez) dz = rxi Res (a),
where rx is pOSltIve if the integration is carried out ID the counterclockwise

direction, and negative otherwise.
The proof of Theorem I follows from the fact that if 1 zf(z) 1 <: K R , then
If(z) I <: KRIR. Sinee the length of e R is lalR, where a is the subtended angle,
Equation (80) gives
11R fez) dz 1 <: ~R • IrxlR = lixlKR ~O (R ~ =). •
The proof of Theorem II is somewhat more complicated. To prove part 1,

we use the relation
It e,m z fez) dz <: I ti e lmz

11 fez) 11 dz l·
tThis result is known as Jordan's lemma.
10.14. Theorems 00 Limitiog Cootours 591
But on C R we have I dz I = R de, I fez) I <: K R , and I e 'm : I = e-mR'in., according

to (lSOa). Hence there follows
I IR I -1 fe. e'm: fez) 1

dz <: RKR ( ' e-mR'in. de,
where o <: eo < el <: n. Since the last integrand is positive, the right-hand
member is not decreased if we take o = o and e
= n. Hence we have e,
(160)
This integral cannot be evaluated in terms of elementary functions of R. How-

e
ever, in the range O <: <: n/2 the truth of the relation
sin e:>
-n
l:...e
is easily realized by comparing the graphs of y = sin x and y = 2x/n. Thus
we ha ve also, from (160),
(161)
and hence, if ni > O, IR tends to zero with K R as R ~ =, as was to be shown.

The other three parts of Theorem II are established by completely anal-
ogous methods. •
To prove Theorem III we notice that the integrand is not greater than
Kplp in absolute value and the length of the path is lalp, where a is the sub-
tended angle. Hence the integral tends to zero with K p • •
To establish Theorem IV, we notice that if fez) has a simple pole at z = a

we can write
f( ...-'" -- Res (a) + qJ () Z ,

z-a
where rp(z) is analytic, and hence bounded, in the neighborhood of z = a.
Hence we have
r
Jc p
fez) dz =
Jc
r p
~e~5~ dz +- r
Jc p
rp(z) dz.
On c we can write z = a + pe'·, where e varies from

p
an initial value eo to
eo + IX. Hence the first integral on the right becomes
Res (a)
1··+"
~
. ,.
pIe e'.de
p
= i Res (a) 1·'+"
~
de = lXi Res (a).
The second integral on the right tends to zero with p, in consequence of

Theorem IlI, establishing the desired result.
Several applications of these theorems are presented in the following
sections.
•
10.15. Indented Contours. In many cases the presence of a pole or branch

point may lead to the necessity of indenting a contour by introducing an arc of
a circle of small radius p, to avoid integration through a singularity. The desired
result is then obtained by considering a limit as p ---> o.
In such cases we frequently encounter a new difficulty which can be
explained best by considering what is meant by the Cauchy principal value of
an improper real integral. To introduce this concept, we consider first the
function y = l/x. It is a familiar fact that the integral S:¡ x-¡ dx does not
exist in the strict sense beca use of the strong infinity of the integrand at x = O.
We recall that this integral would exist, according to the conventional definition
of an improper integral, only if the limit
1"
6¡,l"r::.. o
(f-" dxX + J6~r dX)
_ 1 X
2
were to exist, and have a unique value, as 0 1 and O2 independently approach

zero through positive values (Figure 10.16). But since this limit is of the form
lim (log 0 1 + log 2 - log O2 ) = ¡im (log 2 - log ~2),

61,6,-0 6 1. 6 2- 0 U 1
y it c1early does not exist unless ti and O2

tend to zero in such a way that O2 /0 1 tends
toward a nonzero limit, in which case the
value to be assigned to the integral depends
-1 -0,
upon this limit. However, if 0 1 and O2 are
taken to be equal, so that the gap around the
infinity in Figure 10.16 is symmetrical about
x = O, the limit is seen to be log 2. Inciden-
tally, the sarne value would be obtained
by formal substitution in the formula
S!x-l dx = [Iog Ixll!. This limit is defined
Figure 10.16 to be the Cauchy principal value of the
improper integral, and we write
dx
f
2
P - = log 2,
-1 X
the symbol P denoting a principal value.

More generally, if f(x) is not finite at a point x = A inside the interval of
integration, we ha ve the definition
P lb f(x) dx
a
= lim
p-O
(l"-P f(x) dx + fb
Q A+p
f(x) dX) , (162)
if that limit exists. If the integral exists in the conventional sense, the true value
necessarily agrees with the principal value so defined, and the symbol P then
may be omitted.
10.15. lndented Contours 593
The consideration of principal values of this kind is frequently necessary

in the process of evaluating proper integrals, as is seen in the folJowing examp le s.
Funher. principal values of improper integral s are not infre qu entl y of physical
significance in applications.t Clearly, some care should be exercised in dealing
with them in such cases.
Example l. In order ro evaluare the integral

y
1 ~ f.o
-=
sin x dx =
x
21=
o
sin x dx,
x
( 163)
we consider rhe result of integra[ing the function

F(z) = e l,/z around the closed contour of Figure
10.17. As before, we replace the sine by a complex
exponential [O obtain satisfactory behavior on -R -p p R x
C R. In so doing, however, we obtain a function
F(z) which then has a pole at the origin and hence Figure 10.17
mUSl inlroduce the indentation Cp- Since F(z) is
analylie inside the eontour, Cauehy's residue theorem glves
J--
P
-R
e/X
- ' dx
x
+ l'-
Cp
el::
z
dz + IR -x. dx + l'-
P
el."C
CIi
e'Z
Z
dz O. (164)
Sinee 11/21 = 1/ R on CR, the fourth integral tends to zero as R ~ =, by Theorem

n.1 of Seelion 10.14. A Iso, sinee F(z) has a simple pole al z = O, with Res (O) = 1,
Theorem 1V sta tes that lhe seeond integral tends to -ni as p - . O, the negative sign
eorresponding lO the negalive sense of C p' Thus, proeeeding to the limil as p ._-, O
and R - =, we have
lim
p _. O (J
-p
-é·' dx
x
+ IR - é·< dx ) -
X
ni = O. ( 165)
R- -R p
Sorne eare mUSl be laken at this stage, sinee the integral r: ~ x-1é x dx does not
exist in the striet sense beeause of the faet that the inlegrand behaves like l /x near
x = O. However, we may notiee that the limit in (165) is in faet the Cauehy principal
value of this divergenr integral, and henee (165) takes the form
p L<O_ e: dx = ni. (166)
By taking imaginary parts of both sides, we obtain the desired re sult
f =
- =
s in x d
-x- x=n '
the symbol P now being omitted si nee this is a convergen! integral, the integrand being
(167)
jinile at x = O. H oweve r, the result of equating real parts of (166) should be written
tFor example, they arise frequently in the aerodynamic theory of airfoils. In technical work,
the sym bol P is orten ornitted before suc h integrals. Also, variolls alternative notations are
used.
in tbe form
(68)
It may be noted tha! !he principal value in (168) truly is taken in two senses, since
in (165) we not only have taken the gap (- p, p) about the origin to be syrnmetrical
but also have taken the upper and lower cut-off radii to be both equal to R, befor~
proceeding to the limit as the gap closes and the extreme Iimits become infinite in
magnitude. •
Example 2. We next consider the integral
J
~ cos x
1 = _~ n 2 _ 4x 2 dx, (169)
noticing that since the integrand is finite at x = ±n/2, the question of principal values
does not arise in the definition of J. To evaluate the integral, we integrate F(z) =
e'z/(n 2 - 4Z2) around the contour of Figure 10.18, taking into account the poles of
71" 71"
2 2
Figure 10.18
F(z) on the real axis. Again making use of the results of Section 10.14, we obtain the
result
p r~ n 2 ~x 4X2 dx - ni [ Res ( - ~) + Res (~ ) ] = O,
where
ni[Res( - ~) + Res (~) ] = ni[e~::2 _ e;~2J = ~,

1
f
O<> elx
p _~ "'"n'"'2""::'---C4"'x'2 dx = -2-' (170)
By equating real parts of the two sides of this equation, we obtain the desired result
1...
f =
n2
cos x
- 4x 2
dx =
2
(171)
•
-~
10.16. Integuls lovolving Branch Poinls 595
10.16. Integrals Involving Brancb Points. When contour integration leads

to the consideration of a complex function with one or more branch points,
care must be taken in selecting an appropriate branch of that function and
using that branch con s is tently throughout the calculation. Two rather typical
examples follow and a number of otbers are considered in Problems 121 to 131.
Example 1. In many cases a real integral can be transformed by contour integration

into a more tractable formo To i!lustrate this proce-
dure and also to consider the treatment of a branch
point, we evaluate the integral. y
1 = l ~ cos
O
xl-m
dxX (O < m < 1). (172)
For this purpose we integrate F(z) = e"/zl-m around

the contour of Figure 10.19. The indentation is
necessary to avoid the branch point at the origino
We must, of course, choose a branch of the multi-
valued function z I-m which is real on the positive
real axis and analytic in the first quadrant, so that
p x
we may define
Zl-m = rl-mei(1-mJ9 (-n < () < n) Figure 10.19
when z = re'·. Cauchy's integral theorem gives
I p
R e'x
xl-m dx + 1
en
e"
zl-m dz + 1P (iy)l
R
e-Y .
m(¡ dy) +
l
Cp
e"
zl-m dz = O.
The integral along C R vanishes as R --> 00, by Theorem n.l, and that along C p
r
vanishes as p --> O, by Theorem III. Hence, proceeding to the limits, we ha ve
x~'~m dx + f ie Im:::yl mi dy = O
or
1~ e-yym-l dy
1
=
O
-eix
- cL'C = é
xl-m
mn / 2
The integral on the right in (173) may be recognized as that defining r(m). Thus, by
(O < m < 1). (173)
equating real and imaginary parts, we obtain the results

r (m) COS-,
1
~ cos x mn
- - dx
xl-m
=
2
O
(174)
r ~ sin x r . mn
Jo xl-m dx = (m) Sin "2'
when O < m < 1. We may notice that the right-hand member of (173) is obtained
formally if we replace x by a new (complex) variable iy in the left-hand member and
do not concern ourselves with distinguishing between 00 and oo/i (or, more precisely,
between the oJd and new paths of integration). The danger of using such a formal sub-
stitution without establishing its validity by contour integration (or otherwise) is

illustrated by the consideration thal by the same substitution in (172) we should deduce
that 1 = i m f~ ym-¡ cosh y dy. However, this integral does nol exist! •
Example 2. We attempt lo evaluate the integral
1 = i~:: 11 dx (O < m < 1) (175)
by integrating fez) = zm-¡/(z + 1) around the

y con tour of Figure 10.20 and proceeding lO the
limit as the small radius p tends to zero and the
large radius R tends to infinity. Here we must
choose that branch of zm-¡ for which
Cut (O < e< 21l)

R x when z = re'··, with lhe underslanding that f) may
approach O from aboye and 21l from below.
Accordingly, along the upper bank of a cut along
the positive real axis we write z = r and along
the Jower bank we write z = re'"', when we
calculate zm-¡. In the limit, the integrals along
Figure 10.20 the two circles vanish (with the given restrictions
on m) and we ha ve
JO (re2.')m-¡ r zm-¡ . -l}
l
~ rm-l - .
r + 1 dr + r + 1 dr - 2111 Res lz + l'
o =
or, equivalently,
(1 - e 2mn')1 = 21li(e",)m-¡ _ -21lie mn'.
Hence we obtain the result
-
1+
o x
xm-¡ dx: -
1 -
1l
sin m1l
(O < m < 1). (176)
(See al so Proólem 119.)

•
REFERENCES
J. CARRlER, G. F., M. KROOK, and C. E. PEARSON, Functions of a Complex Variable:

Theory and Technique, McGraw-Hill Book Company, Inc., New York, J966.
2. CHURCHILL, R. V., Camplex Variables and Applications, 2nd ed., McGraw-HiIl
Book Company, Inc., New York, 1960.
3. CURTlSS, D. R., Analytic Functions of a Complex Variable, Second Carus Mathe-
matical Monograph, Open Court Publishing Company, La Salle, m., 1926.
4. FRANKLlN, P., Functions of Complex Variables, Prentice-Hall, lnc., Englewood
Cliffs, N.J., 1958.
5. KAPLAN, W., A First Course in Functions of a Complex Variable, Addison-Wesley
Publishing Company, lnc., Reading, Mass., 1953.
Problems 597
6. K "o pp , K ., Theory 01 Fun clions, Dover Publieations, lne., New York, 2 pts., 1945,
1947.
7. TIT C HMARSH , E . c., The Theory 01 Funclions , 2nd ed., Oxford University Press, Ine.,
New Y o rk, 1950.
8. WHITTAKER, E . J ., and G. N. WATSON, lvfodern Analysis, Cambridge Univers ity
Press , New York , 1958.
PROBLEMS
Section ]0.1
+ ib is written as a number pair (a, b), then for

1. Show that, if a eomplex number a
two sueh eomplex numbers the laws of eombination take the form
(a" b,) + (a 2, b 2 ) = (a¡ + a2, b¡ + b 2),
(a" b¡)(a2' b 2 ) = (a,a2 - b,b 2, a,b, + a2b,),
(a2 ' b 2) = (a¡al + bl2, a,b1¡ - aªb,).
(u, , 6,) a, + b, al + b,
2. (a) Show that, if the real and imaginary parts of a =
a + ib are the eomponents
of the vector v = ai + bj, then the real and imaginary parts of a, ± a 2 are the eom-
ponents of the vector v, ± V2, but that no sueh statement applies to multiplieation
(with referenee to either the dot or the eross produet of veetors) or to division.
(b) Show, however, ¡hat aa =
la 12 is equal to the dot produet v·v Iv 12, and =
that a,a 2 + a,a z = 2v, 'V2' (Notiee that these qua ntities are real sealar numbers).
3. Establish the following results:
(a) Re (z, + Z2) = Re (z ,) + Re ( z,), but Re (z ¡z,) ,,: Re (z,) Re (Z2) in general;
(b) 1m (z, + Z2) = 1m (z,) + 1m (z,) , but 1m (Z¡Z2)": 1m (z,) 1m (Z2) in general;
(e) IZ,Z21 = Iz,llz21, but Iz , + z21,,: Iz,1 + IZ21 in general;
(d) z, + Z2 = ZI + z, and Z'Z2 = ZlZ,.
4. Establish the following resuIts:
(a) z + z = 2 Re ( z),
( b) z - z = 2i 1m (z),
(e) Z ,Z2 + Z;-Z2 = 2 Re (Z¡Z2 ) = 2 Re (ZlZ,),
(d) Re ( z) < I z 1,
(e) 1m ( z) < I z l,
(f) Iz,z2 + Zlz21 < 21z, Z21,
(g) (izll -IZ21)2 < Iz ¡ + z21 2 < (lz,1 + IZ21)2 . [Use part (f).]
SecLiOD 10.2
5. Express the following qllantities in the form a + ib, where a and b a re real :
(a) (J + i)3, (b) 11 +
-
i,
I
(e) e"'/l,
(d) e'h1 / 4, (e) sin (~ + 2i), (f) c o sh (2 + ~).

598 FunCtiODS oC a Complex Variable
6. Use the definitions (20a, b) to establish the following identities:

(a) sin ' z + cos' z = 1,
(b) sin (z, + z,) = sin z, cos Z2 + COS ", sin =2,
(e) eos 2z = 2 cos' z - l.
7. Prove tha! the funetions sin z and eos z are periodie, with real period 2n, whereas
e r , sinh z, and cosh z are periodic, with pure imaginary period 2ni. What are the periOds
of the other circular and hyperbolie funetions ?
8. Deduce expressions for the derivative of sin z, cos z, sinh z, and cosh z from the
established result deo, /dz = ae o r • Also obtain tbese express ions by exploiting the per-
missibility of term-by-term differentiation of the Taylor series representations.
9, Prove that e' possesses no zeros, tbat the zeros of sin z and cos z all lie on the real
axis, and that those of sinh z and cosh z aH lie on the imaginary axis.
10. If fez) = el:, show that fez) = e- I :, fU) = el!, and fU) = f.(z) = e-IZ.
11. Establish (18) by obtaining the relations
e"e" = - -
........ z~
1 Zk
.LJ .4J ., k I
¡ - Ok = O)"
=.... ....-;
-
.4J.4J. I (
['
. _ O¡_o).n-).
Z{ 7'-¡
.),'
]
where n was written for j + k, and reviewing the binomial expansion of (z, + Z2)'.
SectioD 10.3
12. Show that the nth roots of unity are of the fo rm co~ (k = O, 1, ... , n - 1), where
CO. = cos (2n/n) + ; sin (2n/n).
13. Determine all pcssible values of the folJ o wing quantities in the form a + ib, and
in each case give also the principal value, assuming the definition (39):
(a) lag (1 + 0, (b) (i)' 14, (e) (1 + i) 1/2.
14. Express the roots of the equation z' + 2z' + 2 = O in the form a + ib.
15. Express the fuoction z' in the form given b y Equation (47) and also find the
principal value of tbis function [assuming (39)] when z = (1 + ;)/../2, in the form
a + ib.
16. If fez) = Z i and g(z) = ¡Z, with the convention of Equation (50), distinguish be-
tween f(i) and gU) .
17. Derive Equations (55c) and (56a) .
18. Determine aH possible values of the quantities
(a) sin-' 2, (b) tan-' (2i).
19. (a) Verify that, if a is a positive real constant,
coth- ' ~ = ..!.. log z + aa = ..l [Iog Iz + al + i arg (zz+- aa)] .

a 2 z - 2 z-a
(b) Verify that
z + a _ (z + a)(Z - a) 1 z 12 - a' - 2a; 1m (z)
z=a - (z - a)(E - a) = 1 z 12 + a' - 2a Re (z) .
Problems 599
(e) Benee deduce that
eoth _ I -a
z = -4 (x +
I log (x a)2
- a)2
+
+
y2
y2
+ -2i [ tan- I ( a2 - 2ay
x2 _ y2
) + 2kn ]
when a is real and positive, where k is any integer.

20. Bipolar coordinates.
(a) If (p 1, W 1) are polar eoordinates relative to the point (a, O) and (P2, (2) are
polar eoordinates relative to the point (-a, O), show that
(b) Use the result of Problem 19(a) to show that
coth- l .3...
a
= 1.-logP2
2 PI
+ 2-(W2 -
2
W,),
where W 1 and W2 are eaeh defined only within an arbitrary additive integral multiple
of 2n.
21. (a) Suppose that the principal values of w , and W2 are such that O <: WIP < 2n
and O <: w 2P < 2n in Problem 20. Verify that, as the segment of the real axis
between -a and a is erossed from aboye, the angle w 2 - w l ehanges abruptly by 2n,
whereas no sueh jump oceurs at a erossing outside this segment, so that transition from
one braneh to another then can and must take place only aeross the finite segment
joining the "branch points" z = ±a.
(b) Suppose that the definitions O <: wlP < 2n and -n < w 2 P <: +n are
adopted. Verify that w 2 - w I then is eontinuous aeross the fiuite segment joining the
braneh points, but that it ehanges abruptly as the real axis is erossed at any point out-
side this segment.
8(,ction 10.4
22. Let fez) = (x - y)2 + 2i(x + y).
(a) Show that the Cauehy-Riemann equations are satisfied only along the curve
x - y = ], and hence deduce that fez) has a derivative along that curve, but is nowhere
analytie.
(b) Verify direetly that fez) has a derivative along the curve x - y = ] by show-
ing first that
AI 2(x - y)(Ax - Ay) + (Lh - Ay)2 + 2i(Ax + Ay)
Az = ..ó.x + i..ó.y
in the general case, and that when x - y = ] there follows
-AI -_ 2 + 2·1 + ux
A - uy -
1• A 2 Ax Ay .
Az Az
(e) Simplify the expression for AfI Az when x - y = - ] as mueh as possible.
(d) Verify that
F(z) = aax [(x - y)2 + 2i(x + y)] = 1.-

1
aay [(x - y)2 + 2i(x + y)]
when F(z) exists, in aeeordanee with Equations (63a, b).
23. Ca) lf 3x2y - y3 is the real part of an analytie funetion of z, determine the
imaginary par!.
(b) Prove that xy2 eannot be the real part of an analytie funetion of z .
(e) Determine whether 2xy + ¡(x 2 - y2) is an analytie funetion of z.
24. Suppose that fez) = u(x, y) + ¡v(x, y) is analytie in a region eR incIuding part of
the real axis.
(a) Show that f(x) = u(x, O) + ¡v(x, O) and henee that
fez) = u(z, O) + iv(z, O)
when z is in eR. [Henee we then obtain w = fez) by merely replaeing y by zero and x
by z in w = u(x, y) + iv(x, y).]
(b) I1lustrate the result of part (a) by deterrnining the funetionfsueh that w = fez)
when
w= Y(X2 + (X2
y2' - 1) + ¡x(x 2 + y2 + 1)
y2 + 1)2 + 4X2 y2
.
(e) Why is the same result obtained in part (b) when x is replaeed by zero and y
by -jz?
25. Let s represent distanee in the eountercIockwise direetion around a cIosed curve
e in the xy planeo
(a) If, at any point P on e, t represents the unít tangent vector in this direetion,
and n represents the unit outward normal vector, show that
t = i cos rp + j sin q¡ and n = i sin q¡ - j cos qJ,
where q¡ is the slope angle .
(b) ]f u(x, y) and v(x, y) are the real and imaginary parts of an analytie funetion
of z in a regíon eR incIuding e, show that
au av
as=-an'
at any point of C. [Recall that alas = t . V and a/an = n . V, and use Equations
(65a, b).]
26. By applying the result of Problem 25 to a circIe r = eonstant, obtaio the Cauehy-
Riemann equations in polar coordioates, in the fonn
au
- = -1- av
-, -1 -au
=-_av
.
ar r ae r ae ar
27. If fez) = u + iv is analytic, detennine tbe following:
(a) v when u = r- 2 sin 2e (r O), *
(b) u when v = r J (1 - 4 cos 2 B) sin e.
28. If fez) = ver, e) + iV(r, e) is an analytic funetion of z = re'· in a region ineluding
part of the positive real ax.is, how can f be expressed in terms of U and V? (Compare
Problem 24.)
29. If fez) is analytic in a regio n Gl and if
Re [fez)] = log r - r sin e (r * O)
in that region, sbow tbat fez) must be defined as a single-valued branch of a multivalued
funetion.
Problems 601
30. Show that the real and imaginary parts of any twice-differentiable funetion of the
form f(i) satisfy Laplace's equation, but that sueh a function is nowhere an analytie
funetion of z unless it is a constan!. [Compare the values of af/ax and af/aUy).]
31. Show that f(l z 1) is nowhere an analytie funetion of z unless it is a constan!. (Con-
e
sider the derivative of f in the direetion.)
Section 10.5
32. Suppose that fez) = ti + iv is analytie in a simple region <R and that
ti dx - v dy = dU, 1) dx + ti dy = dV
in <R (where U and Vare single-valued). If F(z) is the complex funetion F = U + iV,
eontaining an arbitrary additive eomplex eonstant, show that F(z) is analytie in <R and
that it possesses the properties deseribed by Equations (69) and (70) in <R. (Show first
that U and V satisfy the Cauehy-Riemann equations in <R.)
33. (a) Use the definition (68) to ea1culate direetly the integral 1e z dz, where e is
the unit circJe x = eos t, y = sin t.
(b) Use the definition (72) to ealeulate direetly the integral 1e log z dz, where e
is the unit eircJe r = 1, taking the principal value of the logarithrn.
(e) Obtain the results of parts (a) and (b) by appropriately dealing with the fune-
tions FI(z) = z2/2 and F 2(z) = z log z - z.
34. (a) Show that the value of the integral
+
J I
-1
z
Z2
1 dz
is -2 - ni if the path is the tlpper half of the eircle r = 1. [Write z = el", where e
varies from n to O, or from (2k + l)n to 2kn, where k is any integer.]
(b) Show (al so by direet integration) that the value is -2 + ni if the path is the
lower half of the sarne eirele.
(e) Obtain the results of parts (a) and (b) by appropriately dealing with the fune-
(¡on F(z) = log z - Z-I.
35. (a) Evaluate the integral
Ji e z + Z2
1d
z,
where e is the eircJe r = 1, first by using the results of Problem 34(a, b), seeond by
eonsidering the fllnetion log z - Z-I, and third by using Eqllations (75) and (77).
(b) Evaluate the integral in part (a) (by any method) when e is the circJe r = a
(a> O).
36. Evaluate the integral
fez dz,
when e is the unit eircJe r = 1 and also, more generally, when e is the eircJe r = a
(a> O).
37. Proceed as in Problern 36 with the integral
fe (lzl - e r sin Z2) dz.

38. (a) Prove that the integral
J o dz
- 1
Z2
is independent of the path, so long as that path does not pass through the origino By
integrating along any convenient path (say, around a semicircJe and thence along the
real axis) show that the value of the integral is -i.
(b) Show that the real integral
dx
J
2
X2
-1
does not exist, but that rhe value given by formal substitution of limits in the indefinite
integral agrees with that obtained in part (a). (Notice that, in spite of this fact, the
integrand is never negative 1)
39. Show that
Ji e dz
Z =
4'
ltl
when the integration is once around the closed curve e defined by the polar equation
. , ()
r = ?- - sm- 4
and explain why the result differs from Equation (75). (Suggestion: Sketch the curve.)
40. Let e represent a semicircle of radius R, with center at the origin, where R > 1,
and consider the funcrions
1
fl(z) = Z2 - 1, f2(z) = Z2 + I
(a) Use Equation (lO) to show that on e there fol1ows
R2 - 1 -< Ifl(z)1 -< R2 + 1,
(b) Deduce from (80) that
(c) Show also that
I f /,(z)f,(z) dzl -< nR~~ + ~.

41. (a) Suppose thal el and e, are two simple closed curves which intersect at exactly
two points. If fez) is analytic on both el and e 2 and also in the two regions which lie
inside one of the curves but outside the other, use Cauchy's integral theorem to prove
Ihat §C, fez) dz = fe, fez) dz.
(b) Generalize the result of part (a) to deduce the "equivalent contour" property
stated at Ihe end of Section 10.5 when el and e o intersect at a finite number of points.
Problems 603
Section 10.6
42. (a) Use the results of Equations (78a, b) to verify Cauchy's integral formula (85)
when I(z) = Z2. [Express I(rx.) in the form (rx. - Z) 2 + 2z(rx. - z) + Z2.]
(b) Verify also the derivative formula (88) in this case.
43. If F(z) (z + 6)/(Z2 - 4), show that the integral fe F(z) dz is O if C is the eirele
=
X 2 + y2 = 1, is 4ni iJ C is the circle (x - 2)2 + y2 = 1, and is -2ni if C is the eircle
(x + 2)2 + y2 = 1. [Use Equation (84) in the seeond and third cases, with the fune-
tions I(z) = (z + 6)/(z + 2) and I(z) = (z + 6)/ (z - 2), respeetively.]
44. A mean-value Iheorem. Let Zo denote a point in a region <R where I(z) is analytie,
and let C denote any eircle, with center at zo, whieh lies inside <R. By writing
IX = Zo + ae/~ in Cauchy's integral formula (85), show that
1
I(zo) = 2n 5.o2
•
I(zo + ae'~) dfP,
and deduce that Ihe value 01 an analYlic lunclion al any poinl Zo is Ihe average 01 ils
values on any circle, wilh Zo as ils cenler, which lies inside the region 01 analYlicily.
45. A maximum-modulus Iheorem. By applying the inequality (80) to the resuJt of
ProbJem 44, deduce that the absoll/te vall/e 01 an analytic lunclion at a point Zo cannot
exceed the maximllm absolute value 01 that lunction along any circle, with center at zo,
which lies inside the region 01 analyticity.
Section 10.7
46. Cauchy's inequality. If I(z) is analytie inside and on a cirele C with eenter a and
radius R, and if I/(z) 1 <: 1\1/ on C, use Equations (88) and (80) to show that
1/ "' (a) 1 <: n ~::w.
47. Uniqueness ol Taylor series. Suppose that, by any method, we have obtained a
representation
I(z) = L- an(z -
n- O
a)n
whieh converges to I(z) when 1z - al < R . By making use of the fact t hat such a
series can be differentiated term by term any number of times inside the circle of con-
vergenee, show that there must folJow
I 'n' (a) = n ! a. (n = O, 1, 2, .. .),
so that this series necessarily is the Taylor series of I(z) in that circle .
48. Obtain each of the following series expansions by any eonvenient method:
(a) SIO z = 1 - 31 + 5! - ... = .~o (_1)n (2n + 1) !
Z2 Z4 z2n 00
z Clzl < =),

cosh z - 1 1 z2 Z 4 z2n OC>
(b)
Z2 = 2 '. + 4 '. + 6 '. + ~ (2 n + 2) .,
= n~O
(Izl < =),
e' 5 8
(e) 1 _ z = 1 + 2z + 2 Z2 + 3 Z3 + (1 z 1 < 1),
2
(d) a
Z2
= 1 + 2z +a a + 3(z -1- a)2 -1- .. .
a2 Clz -1- al < lal)·
1
49. L'Hospital's rule. Suppose that fez) and g(z) are analytie at z = a and that
fea) = r(a) = ... = f'k)(a) = O
and g(a) = g'(a) = ... = g(k'(a) = O
but not both f(k+lJ(a) and g(k+¡'(a) vanish. Prove that
lim fez) = f(k+ ¡)(a)
,-a g(z) g'k+ "(a)
when g(k+ ¡'(a) "1'= O. and that the limit fails to exist when g(k+ !J(a) = O. (Suggestion:
Consider the Taylor series expansions of f and g near z = a.)
50. Use L'Hospital's rule (Problem 49) to evaluate the following lim.its:
(a) lim sin z, (b) lim 1 - ;os z,
=~O z z--->o z
(e) lim sinh z , (d) lim (z - ~) tan z.
r-1I1" e
Z
+ 1 z-1I/2 _
Section 10.8
51. Expand the funetion fez) = 1/(1 - z) in eaeh of the following series:
(a) a Taylor series o[ powers of z for 1z 1 < 1;
(b) a Laurent series of powers of z for 1z 1 > 1;
(e) a Taylor series of powers of z + 1 for 1z + 11 < 2, by first writing
fez) = [2 - (z + 1)]-¡ = i.[1 - (z + 1)/2]-¡;
(d) a Laurent series of powers of z + 1 for 1z + 11 > 2, by first writing
fez) = -[1/(z + 1)]/[1 - 2/(z + 1)].
52. Expand the funetion fez) = 1/[z(1 - z)] In a Laurent (or Taylor) series whieh
converges in eaeh of the following regions:
(a) O < 1z 1 < 1, (b) 1z 1 > 1,
(e)O<lz-11<1, (d)lz-ll>l,
(e)lz+ll<l, (f) 1<lz+ll<2,
(g) 1z + 11 > 2.
53. Determine aH Laurent (or Taylor) expansions in powers of z of the funetion
fez) = (z - 2)/(Z3 - 1), speeifying the region o[ eonvergenee for eaeh series.
54. Without determining the series, speeify the region of eonvergenee for the Laurent
series of the funetion fez) = 1/(z4 + 4), in powers of z - 1, whieh converges when
z = i.
55. Determine the first three nonzero terms in the Laurent expansion of fez) = ese z
whieh is valid when O < 1z 1 < n by first showing that the expansion must be of the
form fez) = c¡z-¡ + e¡z + e3z3 + " ' , then determining C¡, el, and e3 from the
eondition
1 = (e;¡ + e¡z + e3z3 + ... )(z - ¿z3 + I~Oz5 - ... ).
56. Uniqueness 01 Laurent series. Assume that the expansion
fez) = ~ bnCz - a)n

n= -QO
somehow is known to be valid when R¡ < 1z - al < R2•

Problems 605
(a) Let e be the cirele I z - al = p, where R, < P < R 2 , and deduce the rela-
tion
1 i f(lX)dlX _ 1 i ~ n-m-!
2ni J e (IX - a)m+' - 2ni J e n-"'=_ bn(1X - a) dlX,
wbere the left-hand merober is the Laurent coefficient Cm, by (96).

(b) By making use of the perrnissibility of integrating terro by terrn around e,
and by reviewing Equations (7Sa, b), show that
(m = 0, ±1, ±2, ... ),
so tbat the expansion must be the Laurent expansion in the specified annulus.
Section 10.9
57. (a) Show that the function w = A log (z - a) increases by 2niA after a simple
positive circuit around the point z = a, and is unchanged if the circuit does not inelude
that poin!.
(b) Show that the expression for w = tan-' z can be written in tbe forrn
w = tan-' z = ~ log (z + i) - ~ log (z - i) + ~.

If the point z describes a simple positive circuit, obtain the increase in w when the circuit
ineludes z = i but exeludes z = -i, and when the circuit ineludes z = - i but exeludes
z = i. Thus verify that w is unchanged if the circuit surrounds both branch points.
58. With the bipolar notation of Problem 20, sbow that
coth-' 3.. = ..llog P2

a 2 p,
+ ~«(¡J2
2
- (¡J,)
defines a single-valued function in the plane cut along the real axis from -a to a if (¡J,
° °
and COz are restricted such tbat -< (¡J, < 271. and -< co 2 < 271., and a single-valued
function in the plane cut by jnfinjte rays from the points z = ±a a10ng tbe real axis if
° -< COI < 271. and -71. < co 2 -< +71..
59. (a) With the notation of Problem 20, show that we may write
and deduce tbat transition from one branch to the other can take place only if
(co! + co 2 )/2 changes abruptly by an odd integral multiple of n .
°
(b) Show tbat, with the restrictions -< (¡J, < 271. and °
-< co 2 < 271., the angle
(co, + co 2 )/2 has a jump of 71. over the finite segrnent between -a and a, is continuous
across the real axis to the left of -a, and has a jump of 271. across the real axis 10 the
right of +a, if a is real and positive. Deduce that a cut between -a and +a then is
necessary and sufficient to make the function so defined single-valued. Show also that
the branch so defined takes on values which are real and of the same sign as z when z
is real and ZZ > a 2 .
(c) In a similar way, show that cuts are needed along infin.ite rays from the branch
°
points z = ±a if tbe restrictions -< COI < 271. and -71. < C02 -< +71. are imposed.
60. Pro ve that the funetions

sinh az 1/2
f,(z) = eos Z12 f2(Z) = sinh bz'i2'
with f2(0) = alb, are analytie at z = O.
61. Loeate and elassify the singularities of the following funetions:
z 1
(a) z2 + l' (b) zJ + l' (e) log (1 + Z2),
(d) (Z2 - 3z + 2)2/" (e) tan z, (f) tan- 1 (z - 1).
62. Show that the funetion
eosh z - 1
fez) - sinh z - z
has a simple poI e at the origino
63. Show that the funetion fez) = ese (llz) has po1es at the points z = 1/(nn), where
n is any integer other than zero, and deduce that fez) has a nonisolated essential
singularity at z = O.
64. Show that the funetion w = 1/(1 + Z 112) has a braneh point at z = O, and that
if the principal braneh of z 1/2 is ehosen, with -n < p -< n, there is no other finitee
singularity, whereas if the seeond braneh is taken, with = p + 2n, there is also a e e
pole at z = 1. [Notiee that we can write w = (1 - z'/2)/(1 - z); also that zl/2 is
+ 1 when z = 1 on the principal braneh and is -1 when z = 1 on the seeond
braneh.]
65. An example of analytic continuation. Suppose that a funetion fez) is defined by a
speeifie series of the form
fez) = Aa + A,z + A 2Z 2 + ...
when 1z 1 < .v2 and that fez) is known to be analytie exeept at the points z = ±.v2i,
but that nothing more is known about fez).
(a) If values of fez), F(z), f"(z), ... were ealculated from the series when z = 1,
and were used lO determine a new Taylor series in powers of z - 1, inside what circle
wciuld it converge?
(b) If the value of f(5) were required, show that one additional series expansion
(launehed, say, from z = 2.5) wOuld permit its eomputation.
(e) If, instead, the value of f(1.5i) were required, how eould it be obtained from
the result of part (a)?
(A sketch disp1aying the circular regions of eonvergenee for the sueeessive series
may be helpful.)
66. Suppose that fez) is defined by the series
=
fez) = ~ Z2"
,,=0
= Z + Z2 + Z4 + Z8 + ...
when Izl < 1.
(a) Show that fez) has a singularity when z = 1.
(b) Show that f(z2) = fez) - z and henee deduce that fez) has singularities when
Z2 = 1.
(e) Show, more generally, that fez) has singularities at the m points (on the unit
eirele) for whieh zm = 1, where m can be assigned any of the values m = 1,2,2 2 ,
Problems 607
2" .... [From this faet it can be dedueed that fez) is not analytie at any point on the
eirele I z I = l.]
67. Using the eonvergenee theorem stated on page 131, determine inside what real
interval an infinite Frobenius series of the form
~
y(x) = L: Akx k +"

k=O
satisfying eaeh of the following differential equations, would converge [excluding the
point x = O when Re (s) < O]:
(a) (1 - X2)y" - 2xy' +
p(p +
I)y = O,
(b) x(1 +
X2)y" y'+ +
xy = O,
(e) x2y" xy'+ +
(X2 - p2)y = O,
(d) y" sin x +
y' eos x +
p(p +
l)y sin x = O,
(e) X(X2 + +
2x 2)y" - y' (x + +
l)y = O,
(f) xeXy" +
y tan x = O.
Section 10.10
68. (a) Ir fe z ) is analytie at z=, show that the real and imaginary parts of fez) must
eaeh tend to eonstant limits (whieh may be zero) as VX2 + y2 ~ 00 in any way, and
that these limits must be independent of the manner in whieh this limiting proeess takes
place.
(b) lf fez) has a simple pole at z~, show that the preeeding statement applies
instead 10 the real and imaginary parts of F(z).
69. (a) Show that le'l = eX. Henee deduce that e' ~ O if z ~ z~ on anyeurve along
which x ~ - 00 , that le' I ~ 00 along a curve for which x ~ + 00, and that along a
curve with an asymptote parallel to the y axis le' I tends to a fioite limit whereas e' does
noto
(b) Obtain a corresponding result for each of the functioos e-', el', and e- I ,.
70. (a) Show that I sinh z I = vsiuh 2 X + sio 2 y. Hence deduce that I sinh z I ~ 00 as
z - z_ in sueh a way that Ixl- 00, but that, if z - z_ such that Iyl ~ 00 while x
is bounded (for example, along a line parallel to the imaginary axis), I sinh z I is bounded
but does not tend to a limil.
(b) Obtain a corresponding result for eaeh of the funetions eosh z, sin z, and
cos z.
71. Determine the nature of the point z_ for each of the following funetions:
.!..,
~
(a) Z2 ,
(b) z + l' (e) z sin
z
(d) (e) (1 + Z2) 1/\ (f) log (1 + z).
72. Show that for the function w = sin- 1 z + i log z [see Equation (55a)] the point z_
is a braneh poiot only on those branehes for whieh I wl is not bounded as z ~ z=.
73. By eonsidering the funetion fez) = Ilz, show that the integral §c= fez) dz may
not vanish eveo though fez) is analytic at z_. (Notiee that in this case C= is equivalent
to any other eontour enclosing the origin.)
608 Funclions of a Complex Variable
Seclion 10.11
74. Prove that any polynomial of degree N, fez) = ao + a¡z + ... + aNz N , has at
least one zero unless it is eonstant. [Assume the eontrary and apply Liouville's Iheorem
lO F(z) = Ilf(z). This result is known as thefundamenral rheorem ofalgebra and is
assumed in elemenlary eourses.]
75. (a) If fez) has a pote of order n¡ at z = al, show Ihat

fez) = (z - a¡)-n'g¡(z),
where g¡(z) is analylie at z = a¡ and al all points where fez) is analytie.

(b) If fez) has a zero of order m¡ at z = p ¡, show that
fez) = (z - p ¡)m'h¡(z),
where h¡(z) is analytie at all points where fez) is analytie, and h,(P¡) 7= O.
(e) Deduee·that a funcrion fez) which is analyric in the extended plane except for a
finite number of poles is determined except for a multiplicative cOllstant by the position
and order of its poles alld zeros ill Ihe finite part of Ihe plane.
76. The argument principIe. Let fez) be analytie inside and on a simple c10sed curve
e exeept for a finite number of poles inside e, and suppose that fez) 7= o on C.
(a) Make use of the results of Problem 75 to show that
1 i F(z)
2ni Jc nz) dz = M - N,
where M is the number of zeros of fez) inside e and N is the number of poles, poles
or zeros of order k being eounted k times.
(b) Show that this result also can be expressed in the fonu
1
M - N = 2n[.ó. c arg fez)],
where the braeketed expression denotes the ehange in the imaginary parl of log fez)
eorresponding lO a eountercloekwise eireuit of C.
77. Use Cauehy's inequality (Problem 46) to give an altemative proof of Liouville's
theorem. [Show thal fez) = ~;: anz n and that I a n I < MI Rn, where M is independent
of R and where R may be inereased without limit.]
SectioD 10.12
78. Ca!culate the residues of the following funetions al eaeh of the poles in Ihe finite
part of the planeo
eZ 1 sin z
(a) Z2 + a 2 ' (b) Z4 _ a 4 ' (e) -Z2-,
sin z 1 + Z2 1
(d) ----z-3 (e) z(z - 1)2' (f)
J
(Z2 + a 2 )2'
e az ez- t - 1 1 - eos az
(g) (h) (i)
2z 2 5z + 2' 1 - Z2 Z9
sinh z z (1 - eos z)2

(j)
1'
(k) (1)
eosh z sin 2 z' Z7
Problems 609
79. If f(z) has a pole of order m at z = a, prove that

1 ,- d M -
1
]
Res (a) = (¡VI _ 1) ! LdzM 1 ((z - a)·>f fez)} :~a
for any positive integer ¡'vi sueh that Ñ!:> m. [Show that Res (a) is the eoeffieient of
(z - a)M- 1 in the Taylor series expansion of (z - a)M fez) about z = a, when M >- m.l
80. (a) 1f f(z) is that braneh of log z for whieh O -<: Bp < 2rr., determine the sum of
the residues of f(z)/(z2 + 1) at its poles.
(b) Proeeed as in part (a) when the restriction on Bp is -rr. < Bp -<: n.
81. (a) If f(z) is that braneh of the funetion e"'" [or whieh ZI/2 = r L2 e "p!2 with
O -<: Bp < 2rr., determine the Silln of (he residues of f(z)/(z2 + 1) at its poles.
(b) Proeeed as in part (a) when -rr. < Bp -<: n.
tz/~ 1
when e is the curve sketehed in Figure 10.21.
Figure 10.21
83. Show that the substitution z = 1/1 transforms the eircle I z I = R into the eircle
I t I = l/R in sueh a way that a positive eireuit around one circle eorresponds to a nega-
tive eireuit around the other. [Write z = Re", 1 = pe" and determine p, rp in terms
of R, B.l
84. (a) Use the result of Problem 83 to show that
! f(z) dz = ! f( +) ~; .
J ¡:¡~R J ¡'¡ - l/R
By letting R ~ co, deduce that fc~ f(z) dz is given by 2ni times ¡he residue of
(1/t 2 ) fO/1) at t = O.
(b) Show that this result implies the result of Problem 73.
(e) Use the result of part (a) to show that if Z2 f(z) is analytic at z~, then
fc~ f(z) dz = O.
85. The residue al z~. It is eonventional to define the residue o[ f(z) at z~, when z~ is
an isolated singular point, by the equation
iJc_ f(z) dz = -2ni Res (z=),

the negative sign corresponding to the fact that a positive circuit around C= is des-
cribed in a negative sense with respect to the exterior of the curve.
(a) Use the result of Problem 84 to show that then
(b) If fez) has only isolated singularities, use Equation (108) to show that, with
the given definition of Res (z=), the sum of the residues of fez) at all singularites in the
finite part of the plane and at z= is zero. [Notice that fez) may ha ve a nonzero residue
at z~ even though fez) is analytic at z_, if Z2 fez) has a pole at z_.)
86. Use the result of Problem 84(a) (or Problem 85) to evaluate the integral
f e
a2 -
a2 +
Z2
Z2
dz,
Z
where C is any simple closed contour enclosing the points z = O, ±ia, and check the
result by calculating the residues at those poles.
87. (a) If fez) can be represented by a Laurent series
fez) = 2: =
11 _ _
Cn(Z - a)n
when R < I z - al < co, for sorne R, show that C-L is the sum of the residues of fez)
at all singularities in the finite part of the plane, so that the number -C-I is the residue
of fez) at z~. (See Problem 85.)
(b) By expanding the integrand in increasing powers of l/z, and identifying the
coefficient of l/z, show that
if C is any simple closed curve enclosing the points z = O and z = ±(ú. (Compare
Problem 86.)
88. Determine the residue of each of the following functions at each singularity:
(a) e l /:, (b) e L /:',
n , (d) (1 + z2)e L /:.
(c) cos
z-n
89. Show that for the function
fez) = e"e L/:
it is true that
1 -
Res (O) = -Res (z~) = ¡-11(2./I),
v t
where JI is a modified Bessel function. [Identify the coefficient of Z-I in
00 O<l fizj-k
j~ k"fa j! k !
and see Equation (95) of Chapter 4.)
Problems 611
Seclion 10.13
90. Use residue ea1culus lo evaluate the following integrals:
Jor
2
(a) de
•
A + Bsine v A .2n:.... B2 (
A>IBI),
(b)
r 2
• de r 2
de• n
(a > O),
Jo a 2
sin 2 e = Jo a + eos 2 2 e = aVa 2 + 1
fo·n eos e de i~,

2
(e) fo·/ sin 4 e de = 4
=
. 2
• sin2 e de = lE..
(d)
J o 5 + 4 eos e
91. Use residue ealculus to evaluate the following integrals:
4
~ n
f
dx
(a) _= (x + b)2 a2 = (a> O), + a
(b)
rf= (X2 +
dx
dx
a 2)(x2 +
n
b 2)
n
2ab(a + b)
(a> 0, b > O),
(e) (a> O),

o x4 + 4a 4 = Sa 3
(d)
r
92. Use residue ealeulus
(X2
dx
+ a 2)2
10
4a 3
n (a> O).
evaluate the foUowing integrals:

(a)
fo a
= x sin mx d
2 r
,x
2 X
n -am
= le (a> 0, m > O),
n (e- e- bm
f
am
(b)
= eos mx e/x -
,.-c,---,---"",.,-'-,---.,---,..,,-;- -- - -- )
o (X2 +2 a )(x + b )
2 - 2(b 2 - a 2 )
2 a b
(a > 0, b > 0, m >- O, b += a),
(a> 0, m::> O),
(e)
(a> 0, m::> O),
(d) Jor- (Xeos mx

2 + a 2)2 dx
_ n -am
- 4a3e (1 + am). (a> 0, m::> O),
(e) Jor- x eos+mx

4a 4 4
_ n -am
dx - Sa3e (eos am +
.
sIn am) (a> 0, m::> O),
Jor= x4
3
x sin mx _ n -am
(f) + 4a4 dx - 'Te eos am (a> O, m > O).
93. (a) By differentiating the equal members in Ihe fusl result of Problem 92(e) with
respeel lO m and b, deduee the evalualions
f-=
_= (x
x sin mx n e am (
+ b)2 + a2 d x = 2a -
b
a eos m + b'
SIO
b )
m (a> O, m > O),
f_= [(x +
eos rn.,.y
b)2 + a2J2 dx =
n
4a3 (1 + am)e
_am
eos bm (a> o, m:> O).
( b) Obtain a formula for the in tegral
l
~ cos mx
o (x' + 4a4)2 dx
from the result of Problem 92(e).
94. U se residue ca\culus to s how that
.o2n Zn
e de = J.zn
o sin Zn 2 (2)
e de = 2;:' : = 2n
1 ·3·5 ... (2n - 1)
,
J CO S
2·4·6 ... (2n)
where e:) is the coefficient of xn in the expansion of (1 + x)z,.
95. U se residue ca1culus to show that
r~ (xZ + 1)(x
xZ dx n
Z - 2x cos ro + 1) = 21 sin ro I
if ro is real and sin ro 7"= O.
96. The Fourier transform of an unknown function f(t) is known to be
. 1
f(u) = UZ + aZ (a> O).
Use residue ca1culus to determine the function f by use of the formula

itx
1 [ e < < ce).
f(t) = 2n _~ xZ + aZ dx (-co t
(See Section 5.15.)

97. Use residue ca1culus to show that
1~~ (~0~t~)~0~C:2 dx = ;a(e-a[,-c[ cos [b(t - c)]+ e -a[,+ c [ cos [b(t + c)]}
when a > O.
98. It is required to show that
-
-
f. ~ sin rt d - 1
o
r+lr=z (t > O) .
(r :-x' r
(a) Show that
1, = L i dx -
e
x -;.' i dX ) .
(b) By considering the limit of the integral of the function

e -U
f(z) = - -o
z-¡
around the sector O <: r <: R, - n / 2 <: e <: O, as R - ce, show that
e - X'
f.
~ e'"
f.
~
- -. dx = r + 1 dr
x - ¡
o o
when t > O. (Use Theorem IT.4 .)
Problems 613
Ce) In a similar way, show that
1- e
x -.;' i dx = 1-/:"} dr,
when t > o, and henee complete the proof.

99. By writing u = Cr + l)t in the definition of 1 2 , show that the result of Problem 98
r r
can be written, in terros of tabulated funetions, in the forro
x:-~' 1 dx = :i:r~ dr = eos t [~ - SiC!)] + sin t Ci(t),
where
SiCt) = I si~ u du, CíCt) = _feos u du

, li
are the sine-integral and eosine-integraI funetions.

100. It is required to evaluate
J'::' e-x' eos 2ax dx,

by making use of the faet that
J: r x
' dx = y'n.
Show that, if x is formally replaeed by x + ia in the known integral, there follows
J~- r(x+io)' dx = ea' J'::' e-x' Ceos 2ax - i sin 2ax) dx = y'n,
and henee, sinee e-x' sin 2ax is an odd funetion of x,
J~- e-x' eos 2ax dx = y'ne- o'.

(Notiee, however, that the validity of sueh eomplex substitutions is not established by
the familiar rules for real substitutions.)
101. Investigate the validity of the proeedure of Problem JOO, by eonsidering the
integral of fez) = e- Z ' over a c\osed rectangular path C inc\uding the segment of the
real axis y = O from x = -A to x = +A and the segment of the line y = a from
x = +A to x = -A.
(a) Notieing that fez) is analytie inside and on C, show that
J A e-x' dx
-A
- JA
-A
e-(x+ia)' dx + JSI
e-" dz + J
5'!,
e- Z
' dz = O,
where .S, is the line segment from CA, O) to (A, a) and S2 is the line segment from
(-A, a) to (-A, O).
(b) Show that, on both S, and S2, there follows
and, by notieing that the lengths of S, and S2 are equal to a, deduce that the integrals
along S, and S2 tend to zero as A ~ ca for any fixed value of a. Thus deduce that the
relation
f~ _'>O e--'C~ dx = J'=

_00 e-(:c+ia)2 dx
is valid for any real value of a, and hence that the result obtained formalIy in Problem
100 is indeed correct.
j --
'~
e PX
(O < p < 1)
I -L
, e x dx
by the folJowing method:

(a) Show that the res ult of integrating fez) = eP' / (l + e') around a closed
rectangular contour e, including the real axis from x = -A to x = +A and the Jine
y = 2n from x = +A to x = - A, can be written in the form
A ,..:e7-p_,,~ dx - e 2p•• J A
J -A 1 + eX
-A
1
eP'~
+ eX
dx + l f ( Z ) dz
s.
+ L, fez) dz
_
-
.
2nl Res
fti e+P ' e" • .}
m ,
where SI and S2 are the closing segments of the rectangle. (Notice that e' has the
period 2ni, and that fez) has poles at the points ni ± 2kni.)
(b) Show that the integrals along SI and S2 tend to zeTO as A ~ 00 (if O < p < 1),
)C -epnl)
J
~ ePX dx - ( 2ni _ n (O < p < 1).
_~ I + eX - I - e 2pni - sin pn
103. Show that the formula obtained in Problem 102 also is valid when p is complex,
provided that O < Re (p) < ]. (Only the integrals along SI aod S2 need be reexamioed.)
Section 10.14
104, By making the substitution x = ¡y, and noticing that then y ~ 00 as x ~ 00, we
formally transform the integral
-r
into the integral
y':4 dy ,
which is the negative of the original integral, and heoce are led to the conclusion that
the value of the integral is zero. But the true value is fouod by elementary methods to
be n / B. What is the fa/lacy? [Integrate fez) = z/ (z' + 4) a round a con tour coosistiog
of the portions of the x and y axes for which O -< x -< R and O :s; Y -< R, and a quad-
rant of ¡he circle I z I = R, let R ~ 00 , and ioterpret the result.]
105. Establish part 3 ofTheorem n.
106. If fCz) is analytic everywhere on the imaginary axis and has no singularities to
the left of ¡hat axis except for a finite number of poles at the points al, a2, ... ,an ,
Problems 615
and if f (~) te nd s t o zero u niforrn ly on an arc of the circJe I? I - R in the second and
th ird quadran ts as R ~ ca . show th at
lim
R-~
J'iR
-IR
e m , fU) d z = 2ni f
k _ 1
Re s [e m' fez); akl (m > O),
w he re the integration is carried o ut a lo n g the imaginary axis.

107. Suppose that, on a circular arc C R with radius R and center at the origin, Iying
in the sector -n/4 -< () -< + n /4, a function fez) is such that
limMR =0,
R--
where M R is independent of () on C R . Pro ve that
lim
R ~<>o ~
r CR
e-m"f(z) dz = O (m > O).
[lf the integral along C R is denoted by h , s ho w that

' /4
liRI os:; 2R2¡1.1R
S o e- mR'co' 29 d(},
a nd se t () = (n - 291)/4,]
108, Use the result of Problem ]07 to prove that if I P I -< n /4 aod if fez) is analytic
throughout the sector bounded by () = O, () = P, and r = R, except fo r a finite num-
ber of interior poles, and if II(z)1 -< R lvl R on the curved boundary, where M R is in-
dependent of () and l.IR ~ O as R ~ 00, then
J'o= e-mx' f(x) dx = S=e - m"f(z) d= + 2ni ~ Res [e- m"I(z); akl
o k
(m > O),
(r)
where r is the radial line () = P, a n d where the p o int s ak are the poles o f fez) inside
the sector bounded by 8 = O and 8 = p. [Notice tha t hence the fo rma l complex sub-
stitut io n x = ,e'P in the first (real) in tegral generally would modify the value of the
integral if fez) were not analytic in the relevant sector.]
.109, (a) Show that the integrals
C = S; cos ,2 d" s= S; sin ,2 d,

can be combined in the form
e-,2~ ,nl l
DO
e- iS = So"'" e - ir 2 di = f o=-O dI = e- n1 / 4 fo e- r 7. dz,

(n
where r is the radial li ne () = n /4.

(b) From the res ult of Problem 108, deduce that
C - iS = e -· i; 4
r= e - x' dx
Jo = e-. i / - ~n , -
and hence tha t
fo- cos /2 dI = fo- sin 1

2
dI = -} J!f
and also
r- ~ dx
Jo,vx
= r- sin :..< dx
Jo ,vx
=
V
I n.
2
Section 10.15
J lO. By making use of integration around suitably indented contours in the complex
plane, evaluate the following integrals:
= .
(a)
J- .
_00
SIll X
x (X 2 T
'
a-?)dx
(a > O),
(b)
J--
111. Show that
(2
xn
SIll X
x 2) dx.
Tri > O),

p 1_ e.~
00 'r:c
dx =
{O
-Tri
.
(1
(1 = O),
(1 < O),
and hence also that
p 1~, co~ Ix dx = O
and
-
J-- sin
x
IX d
x-1 _ roTr
(1
(1
> O),
= O),
-Tr (1 < O).
112. The Fourier transform of an unknown function [(1) is known to be
- 1 - e-(au
[(u) = -'--~i"-
u- (a> O),
and hence (see Section 5.15) lhere follows

1 J<>O 1 - e-/ ax
[(t) = 2Tri __ e"X -'----:ex=--- dx (-= a),
1 (1 = a),
r
[(1) = 1 (O < I < a),
:l: (1 = O),
O (1 < O).
[Write
[(1) = 2ni J- x
1 [ p __ eI'x dx - J-
P __ ei(,-·)x
x dx ]
and use ¡he result of Problem 111. See a1so Equalions (243) and (246) in Chapler 5.]
Problems 617
113. The Fourier sine transform of an unknown function f(l) is known to be

- cos au
fs(u) - (a> O),
u
r
and hence (see Section 5.15)
f(l) = ; .::.I_---=-~,,-=-o-=--s-'-a:.:. :x sin Ix dx (1 ) O).
Show that
1
f(/)=-Tm [J
P
~ 2e"x
- - d x - P J~ e" ,+a)x dx-P [e",-a)x d.x ]
2n .x -00
x .x _ 00 _00
and consequently that

(O < / < a),
(/ = a),
(/ > a).
[Use the result of Problem I1 l. See also Equations (243) and (246) in Chapter 5.]
114. The Fourier cosine transform of an unknown function f(/) is known to be
r( ) -
JC 1I
_ 510
--
au (a > O).
11
By proceeding as in Problem 113, show that f(/) has the same definition here as in that
problem. [See also Equations (243) and (246) in Chapter S.]
115. (a) Suppose that fez) has a simple pole at ao on a simple elosed curve e, but is
analytic elsewhere inside and on e except for poles at a finite number of interior points
a" a2, ... , ano If the con tour e is indented at ao by a circular are with center at ao,
show that the limiting form of the integral of fez) around the indented contour is
p'¡; fez) dz
Te = ni Res (ao) i; Res (ak)
+ 2ni k=1
as the radius of the indentation tends to zero, regardless of whether the indentation
exeludes or ineludes the point ao, provided that e is smoo/h at ao.
(b) What form results in part (a) when e is the boundary of the sector O <: r <: 1,
O <:e <: ex and ao = O?
116. Obtain the evaluation
cos ax - 2 cos b.x d x=7t (b - a ).

[ x
-~
[Nolice Iha! fez) = (e,a, - é b')/ Z2 has a simple poi e at the origin.] By taking a = O
and b = 2, also deduce the formula
J
~ sin2x d
-x-.- x = n.
-~
117. Obtain the evatuation
~ sin (x + a) sin (x
f
- a) d.x n..,
x
2
- a
2 = -2a s In _a.
-~
[Notiee that sin (x + a) sin (x - a) = !<eos 2a - eos 2x).]

118. (a) Obtain the evaluation
p f-~ 1 ~.Tex dx = n eot pn (O <> eP;C -
__ 1 _ eX dx = n(cot pn - cot qn)
if O < p < 1 and O < q < 1. (Notiee that this integral is eonvergent, so that principal
values are not involved.)
(e) By replacing x by 2x and writing p = (eo + 1)/2 in the result of part (a),
deduce that
= eWX neo
p
f -=
sinh x dx
(d) By replacing x by bx, where b

= n tan 2
> O, and writing eo

(-1 < eo <
=
1).
alb, obtain the result
J
~- e"X n na
p _= sinh bx d.x = b tan 2b (b>lal).
(e) Deduce that
f = sinh ax d.x = -n
sinh bx b
t na
an-
2b (b > I al)·
Section 10.16
119. Use the result of Problem 58, Chapter 2, 10 evaluate the integral considered in
text Example 2 in the alternative form
1= xx'".;. 1 dx = r(m)r(1 -
1
m) (O < m < 1),
and henee, by eomparing the results of the two ealculations, deduce the relation
r(m)r(l - m) = .
slnmn
n (O < m < 1).
[This relation was stated without proof in Equation (59), Chapter 2. Although the
presen t proof is valid only when O < m < 1, the relation is generalized to all non-
integral values of m by making use of the recurrence formula for the Gamma function.)
120. Use the method of text Example 2 to show that
p fo= xm(ld<: x) = -n cot mn (O < m < 1).

Problems 619
[Jndent the upper and lower banks of the cut to exclude the pole at z = and deduce
that
(1 - e- 2m ")1 = ni[Res (e Oi ) + Res (e 2 .,)],
where 1 is the required integral.
121. Suppose that zm fez) tends uniformly to zero on the circle I z I = R as R ~ ce ,
where m > O, and that fez) is analytie exeept for a finite number of poles al, a2, ... ,
a n , none of whieh is on the posilire real axis or at the origino By proeeeding as in text
Example 2, show that
ne- i( m- I )n "
=. I;Res(zm-If(z);n.},
Sin mn ' -1
where zm-l = ym-l e ; ( m-I )8 , with 0 < (J < 21t, when z = re i8 .
122. Use the result of Problem 121 to obtain the following evaluations:
a
r-
( ) Jo x m- 1 d
X2 + 1 x = n
sin [( 1 - m)n/ 2]
sin mn
n
= 2 sin mn/ 2
(O
< m < ,
2)
r~
m
(b) x - 1 dx = (1.- m)n (O < m < 2).
Jo (x + 1)2 sm mn
123. (a) Verify that the relation

bx + a x=--
I-Q
1= x+l' b-t
transforms the finite interval a -< 1 -< b into the semi-infinite intervalO -< x < ca.
(b) Use thd result of part (al to show that
1
( 1
+ I)m-I g(t)dt= r-
1 1. x=-If(x)dx,
_o
where f(x) = (x +2 1)2 g X

(X -+ 1)
1 '
when the integrals exi s t.

(e) Show similarly that
f~, (J _12)m- ' h(l)dl = 1- xm-'f(x)dx,
where f(x) = 2
1(x +2 ] )2m h (x-l)
+ 1 ' X
when the integrals exist.

124. Use ¡he results of Problems 122 and 123(b) to deduce the following formulas:
(a) r e+ I)m-
I
_ - 1 - f
I di = 2(J -
Sin m71
m)n (O < m < 2),
(b) r G+ t)"'-
--1
I
- I
1
12
dI
+] =
n
2 sin mn/ 2 (O < m < 2).
125. Use the result of Problem l23(c) lo oblain the evaluation

l
f-1
,vl-t 2
t 2 +l
-
dt=n(,v2-1).
126. Consider the integral
fe fez) log z dz,
where C is the contour indicated in Figure 10.20, fez) is analytic except for a finite
nurnber of poles, none of which is on the positive real axis or at the origin, and the
branch of log z for which
log z = log r + i () (O < () < 2n)
is to be employed. If zl+cf(z) - > O uniformly on C R as R - > ca, for sorne positive
constant e, show that in the limit as R - > ce and p - > O the integral becomes
S: f(x) log x dx - S: f(x) (log x + 2ni) dx,
J.-o f(x) dx = - L:
k
Res (f(z) log z; ak},
where the points ak are the poles of fez). (Note that p log p - > O as p - > O and that
R-C log R -> O as R -> ca if e > O. Care must be taken in using the specified branch
of log z when the residlles are evaluated.)
127. Use the result of Problern 126 lo evaluate the following in te gral s :
dx na + 2b log (b/a)
(a) [ o (x + a)(x2 + b2) = 2b(a2 + b2) (a> O, b > O),
(b) 1-
o
x 3
dx
+ a3
= 2n:,v3
9a 2
(a> O).
128. Proceed as in Problem 126 with the integral
fe f(z)(log z)2 dz,
showing that in the limit this integral becomes
S: f(x)(log X)2 dx - S: f(x)(log x + 2ni)2 dx,
and hence deducing that
1- f(x) log x dr: = - ; Re [ ~ Res (f(z)(log z)2 ; ak:l]
and
Jor- f(x) dx =
1
-2n 1m [~ Res (f(z) (log Z)2; ak:l].
where the points ak are the poles of fez). (The second result generally is les s convenient
than the formula obtained in Problem 126.)
Problems 621
129. Use the first result of Problem 128 to evaluate the following integrals:
(a) Jor= X2log+ xa2 dx =

n
2a log a (a> O),
(b) r= (x + log
Jo
x dI: _ log (b/a) log (ab)
a)(x + b) , - 2(b - a)
(b> a> O),
(e) f= log x dx log a= (a> O).

(x + a)2 a
o
130. Use the first result of Problem 128 to show that
Jor ~ (x +
log x -!L
b)2 + a 2 dx - a log p (a> O),
where p = ";0 2 + b 2 and O. Then take eare when writing -b ± ia in eor-
responding forms for the purpose of ealculaling lhe residues at lhose points.)
131. (a) Show that
J e+ ~)g(/)
I
-1
log dI = r~ ¡(x) log x
Jo
dx,
where ¡(x) = (x +2 1)2 g (xX -+ 1)

1 .
[See Problem 123(a).]

(b) Use the results of pan (a) and Problem 129(b) to obtain the evaluation
I lo g(1I+l1) 1 (11 +- ee)J2

J - 1
dI
el
= J....[IO
2e g (Iel < 1).
11
Applications of Analytic
Function Theory
11.1. Introduction. This chapter relates sorne of the results obtained in

Chapter 10, relative to properties of analytic functions of a complex variable, to
considerations of earlier chapters, and is principally focused on methods for
dealing with problems governed by partial differential equations.
The first two sections apply contour integration and residue calculus to the
determination of inverse Laplace transforms, and hence complement the
treatments of Chapters 2 and 9 dealing with the use of Laplace transform meth-
ods in the solution of appropriate types of initial-value problems.
In the five following sections the possibility of considering an analytic
function of z = x + iy as effecting a mapping of a regíon in the xy plane onto
a region in a new plane, and the distinctive properties of such a l1)apping, are
exploited for the purpose of solving a variety of problems whose formulation
involves the Laplacian operator.
The three final sections define the so-called Green's function associated with
a two-dimensional region and with a cJass of problems specified over that region,
relates that function to a certain analytic mapping function in important special
cases, and illustrates the determination and application of the Green's function
in sorne specific instances.
11.2. Ioversioo of Laplace Transforms. Whereas Equations (240a, b) of

Chapter 5 permit the determination of both the Fourier transform of a given
function and the inverse of a given transform, when the relevant functions are
sufficiently well behaved, no direct method was given in Chapter 2 for the
determination of the inverse Laplace transformo We now use the former rela-
tions to derive a formula for this purpose.
622
11.2. Inversion of Laplace Transforms 623
Equation (238a) of Chapter 5, which is equivalent to (240a, b) of that

chapter, is valid if f is piecewise differentiable and if f:~ I f(t) I dt ex.ists, but
usually is not valid if tbat integral fails to exist. In order to obtain a relation
which is more usefuJ for present purposes, we first suppose tbat f(x) vanishes
for all negative x,
f(x) = O when x < 0, (1)
and then replace f(x) by e-exf(x) in (238a), writing tbe result in tbe form
e-exf(x) = _1 lim fR e iUX [ r~ e-iu'e-c<f(l) dt] du, (2)

2n R-~ - R Jo
and correspondingly assuming now tbat e is a real constant such tbat
r e -c 0, of any piecewise differentiable function f(x) for wbicb tbe real
constant e can be chosen so tbat (3) bolds, and witb the usual convention tbat,
at a finite jump, f(x) is to be assigned the average of its rigbt- and left-hand
limits.
This representation is more general than the Fourier integral representation,
for a function f(x) which vanishes when x < 0, since it permits I f(x) I to grow
exponentially as x ~ 00 , whereas the latter representation generally requires
tbat I f(x) I tend fairly rapidly to zero.
The inner integral in (4) becomes the Laplace transform of f(l) if iu e is +
replaced by s. Whereas this replacement is merely a cbange of notation for the
inner integral, it corresponds to a change in the variable of integration for the
outer one, leading to the equivalent representation
f(x) = 2 1 . lim JeT iR e sx [ r~ e-S( f(l) dE] ds. (5)

1tl R-- e-iR Jo
Thus if we bere denote the Laplace transform of f(l) by F(s),
F(s) = .BU(t)} = r e-sif(l) dt, (6)

then (5) states tbat
f(x) = r 'TU1 lim

R--oo
Je+iR
c - iR
es F(s) ds,
where, since the real variable u increases from -R to R in (4), tbe complex
variable s = e + iu is such that its real part is constantly e wbile its imaginary
part increases from -R to R. If, for convenience, we replace the free variable
624 Applications oC Analytic Function Theory
x by 1 and ¡he complex dummy va riable s by z in this relation, it takes the fOrm
1 .
I(t)=.e- ' (F(s)} = - 2. 11m
fC+Ri e"F(::)dz, (7)
nI R - oo c- RI
where now the integration is along the line x = c in the z plane, indicated in
Figure 11. J (a), and where e is assumed to be sufficiently large to ensure the
existence of the real integral S:: e-e< I I(t) I dt.
4
y
c+ioo B
e ~c
Y':
BR
x
D(;, x
~
v«
c- Loo v-~ C/.
E A
(a) (b)
Figure 11.1
The right-hand member of (7), specifying the inverse Laplace transform of

F(s), is known as the Bromwich integral, and often is abbreviated in the form
1 . fC+'~ e"F(z)dz=")"'
I(t) = - 2
ni c-{~
1
_ni
f
B
e'~F(z)dz, (7')
where B is the infinite Bromwich lineo Generally, it must be evaluated by use of

residue calcul us, by methods next to be iIlustrated.
First, \Ve suppose that F(z) is analytic except for afinite number 01 poles, all
of which lie to the left 01 the fine x = c. In this case we form the closed contour
indicated in Figure 11.1 (b), consisting of the part B R of the Bromwich line froro
y = -R to y = R, segments BC and E A paral!el to the real axis, and the
semicirc1e CDE. In addition we suppose that F(z) tends uniformly to zero along
the e/asure BCDEA as R ~ 00, so that I F( z ) I <: K R on al! oE BCDEA , where
K R depends only on R and where K R ~ O as R ~ oo.
When we consider the integral
/=2
1
ni
.f ABCDEA
e"F(z)dz,
we see tbat the co ntribution OD AB teods to f(t) as R ~ 00, uDder the assump-
tion that c is also so large that (3) is true. Further, the contribution 00 CDE
tends to zero by virtue of Theorem 11.3 of Section 10.14. Along BC we may
write z = x +
Ri, as x varies from e to O; we then have
11.3. Jnversion of Laplace Transforms with Branch Points. The Loop Integral 625
and since I F(::) I <: K R on Be (wbere K R ~ o as R - 00), and al so the length of

tbe patb Be is constantly e, the ML inequality (80) of Chapter 10 gives
I Le e" F(z) dz I <: ce" K R ,
so tbat tbe contribution on Be tends to zero as R - ca. A similar argument

applies to EA.
Finally, since ultimately all poles of F(z) are inside the contour ABeDEA,
as R ~ 00, we deduce that when the preceding conditions are satisfied by F(z),
there follows
f(t) = ~-l{F(s)} = t= Res (e"F(z); a k }, (8)
where the points a k are the poles of F(z).

Since this function c1early is of exponential order, there assuredly exisls
a value of e for which (3) is true. We have assumedthat the value of e used in the
derivation of (8) is also sufficiently large for this purpose. If this were not
the case, then resolving the contradiction by increasing e would not change the
result, since no additional residues exist to be introduced into (8). Hence the
assumption is irre1evant to the outcome and (8) is valid.
Example. As a simple illustration, we make the determination

~- 1{ S
s 1 -t- 1
} _
-
{ze". ·1
Res z 2 -t- 1 ' 1 J+ r ze"
Res l z 2
+ 1; - iJ
l
=
1 ,'"
Ze T
l
Ze-' .,
= COS l.
in accordance with formula (TI6) in the table of transforms, page 67.

•
11.3. Inversion of Laplace Transforms with Branch Points. The Loop Inte-
gral. In order to illustrate the evaluation of the Bromwich integral when F(z)
has one or more branch points, we suppose next that F(z) satisfies Ihe conditions
imposed in the preceding seclion, except Ihat il also possesses a single branch
point, at z = O, and has no po/es on the negative real axis.
We tben cut the z plane along the negative real axis and form the c10sed
contour of Figure 11.2(a), containing the excursion DOE from tbe contour of
Figure ll.l(b), inward along the upper bank of the cut, around a circle e, of
radius E about the origin, and outward along the lower bank of the cut. Here E
is taken sufficiently small to exclude all poles from e,. Correspondingly, we
restrict attention to that branch of the multivalued function F(z) for which the
angle e associated with z is such that -n < e
< n, so that tends to ±n as e
the upper and lower banks of the cut are approached, and we write F(re 'n )
e
and F(re-;') for the limits of F(re i9 ) as tends to n fram smaller values and to
-n from larger values, respectively.
Y
e
Cut
t«//D~
e ,
E-- ~ x
'~
~v.
V.v~
A
F
(a) (b)
Figure 11.2
Finally, we suppose that zF(z) tends uniformly to a constant A (which may

be zero) on the circle I z I = E, as E --+ O.
Then, if we consider the integral
1
-2.
ni
i
Je
e"F(z)dz,
where C is the c10sed contour ABCDOEFA, and first let R --+ 00, so that tbe
contribution on AB again tends to J(I) and the contributions on BCD and EFA
again tend to zero, we deduce that
k
1
J(I) = :2: Res (e"F(z); a k } + -2. e"F(z) dz , (9)
nI
r
JL
where the "loop integral" is taken along the " loop" L indicated in Figure 11.2(b).
Here the orientation of L has been taken as the negative of that o f DOE, in
order that the sign prefixed to the transposed integral in (9) be positi ve.
On the lower bank of the cut we write z = re - i . , on the E-circ1e z = Ee'9,
and on the upper bank of the cut z = re'", so that
Le" F(z) dz = [ e,,·-i. F(re-i')(e -" dr)
+ f.~ e"'" F(re")( e i• dr ) + i f. e'U i9( Ee i9) F( Eei9 ) dO.

Only with reference to the multivalued function F (z) is there need t o di stingui sh
bet ween el" and e- in (as indicato rs o f limiting processes invo lving e'" as
0--+ n- or -n + ). Hence we can write
Le" F ( z ) dz = t~ e -" [F(re -'") - F( re i . ) ] dr + i f .ef«'9( a ") F(Ee'O) dO ,

i
(10)
w here E is an y sufficientl y s mall pos iti ve number.

lo the general ca se , it may happen thal neither of the two integrals on the
rig ht in (10) tend s to a finite limit a s E - O, but only that their sum does so.
However, if it is tme that z F(z) tend s uniformly t o a certain cons tant A on C,
11.3. Inversion of Laplace Transforms with Branch Points. The Loop Integral 627
as é ~ 0, we see that the integrand of the () integral itself tends to A and also
that the integral then tends to 2niA. Accordingly, the r integral also must tend to
a limit as é ~ °in this case. The introduction of the limiting form of (10) into
(9) is then the desired formula
f(t) oC-1{F(s)} = 2: Res {e<zF(z);

k
ak } + 2 1 . Jor= e-"[F(re-
nl
i
") - F(re i ")] dr + A,
(11)
where A = lim zF(z) (12)
,-o
and where again the points a k are the poles of F(z) [or, more specifically, of the
branch of F(z) under consideration].
Example. We seek the inverse Laplace transform of the function

F(s) = s-m (O < m < 1). (J 3)
In this case we verify easily that F(z) satisfies a11 the specified conditions, so that (11) is
indeed applicable. Here the inverse transform is given entirely by the loop integral.
That is, the infinite Bromwich line and the loop L of Figure 11.2(b) are equivalent
conlOurs in this case. With A = O, there then foIlows
= sinnnm fo= e-"r- m dr (O < m < 1). (14)
Frequently, in similar situations, it is not possible to proceed beyond this stage, in which
the inverse transforrn is expressed in terrns of a certain real integral, because of the
fact that the integral cannot be expressed in closed form in terms of tabulated func-
tions. (In such cases the integral must be evaluated approximately, in practice, by
numerical methods.) Here, however, the substitution tr = u yields the evaluation
so that also
(15)
and, indeed, the use of Equation (59) of Chapter 2 simplifies this result to the final
form
tm-1 tm - 1
oC-1(s-m} = r(m) - (m - 1)! (16)
in accordance with formula (T35) in Table 1, page 68.

Whereas we have established this resull here only when O < m < 1, it is valid
also for a11 m such that m :'> l. When m = 1, or when m is any other positive integer,
the result is very easily derived by the method of Section 11.2, since then F(z) possesses
only a poleo However, the situation when m :'> 1 and m is nonintegral is one of those
noted previously, in which both integrals on the right in (10) diverge as é - O but
their sum converges, and the derivation is correspondingly much less simple. •
628 Applications of Analytic Function Theory
Problems 15 and 19 illustrate ¡he rreatment of situations in which the

transform function F(e) possesses two or more branch points. In such cases the
"Ioop integral" appears in different forms, as an integral which is equivalent to
the Bromwich integral apart, perhaps, from the contributions of certain po/es
of F(e).
11.4. Conformal Mapping. In order to represent geometrically the inter_

pretation of a functional relationship w = !(z), it is conventional to employ two
planes, in one of which (the z plane) the real and imaginary parts (x and y) of the
independent variable z are plotted as the point (x, y), and in the second of which
(the w plane) the real and imaginary parts (11 and v) ofthe dependent variable ware
plotted as the point (11, o). In this way a correspondence is set up, in general,
between points, curves, and regions in one plane and their images in the other
·plane. We speak of such a correspondence as a mapping between the two planes.
If the function fez) is sing/e-va/ued, then corresponding to each point z
where f(e) is defined there exists one and only one value of w = u iv, and +
hence one and only one point in the w plane. Otherwise, lO a given point z
there will, in general, correspond two or more points in the w plane. Usually,
if f(e) is multiple-valued, we introduce suitable cuts in the z plane in such a
way that a given branch of f(z) is single-valued in the cut plane. Then u and v
are single-valued functions of x and y. However, the reverse may not be true,
in the sense that two or more values of z = x +
iy may correspond to the same
point w = u ~ iv. To investigate this possibility, we notice that if we write
w =f(z) = f,(x,y) + if,(x,y) = u + ¡?J, (17)
then there follows
v = f2(X, y). ( 18)
Thus, lO determine x and y in terms of u and ?J, we must solve Equations (18)
for x and y. According to the result of Section 7.2, these equations can be
sol ved un iquely for x and y in some region abou t any point where (L 7) holds
and where the Jacobian determinant
af, af, au au
J _ aU,J2) - ax ay ax ay
- a(x, y) - af, af2
-
av av
ax ay ax ay
is nOl zero. If f(z) is ana/ylic at Zo = X o -+- ¡Yo (and hence in a region including
7 0 ), the Cauchy-Riemann equations are satisfied at zo, and hence there then
follows, from Equations (64) and (65) of Chapler 10,
[J L, = [au av _ au aV'J = [(aU)2 + (aV)'l = 11'(zo)1 2 • (19)

ax ay ayax" ax ax .J"
Thus we conclude that, if fez) is ana/y tic al a point Zo and if 1'(zo) =7= O, then
there exists a region inc/uding Zo in the z p/ane and a region including W o = f(zo)
llA. Conformal Mappiog 629
in the w plane sueh that the mapping w = fez) gives a one-to-one eorrespondenee
between points in the two regions.
To illustrate such a mapping, we consider the mapping function
w =f(z) = Z 1l2. (20)
Since fez) is double-valued, it follows that to each point z = x + iy except the
origin there correspond two points in the w plane. However, since the inverse
funetion
z = F(w) = w 2 (21)
is single-valued, we see that to each point w = u + iv there will correspond a
unique point in the original z plane. To make the mapping one-to-one, we may
cut the z plane along the negative real axis, as in Section 10.9, and consider
the principal braneh of z 1/2, for which z 1/2 is real and positive when z is real and
positive,t
w = fez) = zl/2 = ,JiZT e'·p/2 (-n < ep < n). (22)
To investigate the nature of the mapping, we may introduce polar co-
ordinates in the two planes by writing
z = x + iy = re'·, w = u + iv = pe'~. (23)
Then (20) beco mes pe'. = fie'·'" from which there follows
rp ="2'
e (24)
We see that as e
varies from -n to n, the angle rp varies from -n/2 to n/2.
Hence the cut z plane is mapped, by the branch of w = Z1l2 which we have
chosen, anta that half of the w plane for which u > O. Equation (24) is con-
venient for plotting the image in one plane of a given point in the other plane, or
for roapping curves expressed in polar coordinates.
If (21) is exptessed in terros of real and imaginary parts, there follows
x + iy = (u + iv)2 = (u 2 - v 2) + i(2uv),
and hence we have the relations
y = 2uv. (25)
Thus it is seen that any straight line x = e, is mapped onto that portion of the
hyperbola u 2 - v 2 = e, for which u > O, whereas the straight line y = e 2 is
mapped onto one branch of the hyperbola 2uv = e 2 • The nature of correspond-
ing regions is indicated in Figure 11.3.
tln applications of conformal mapping (and in related considerations) it usually is desirable to

define a branch in an open region. Thus, with the definition (22). the cut along the negative
real axis in the z plane is excluded. In correspondence with the possibility of appraaching the
cut arbitrarily closely from aboye or below, we continue to speak of the " upper bank" and
"Iower ban.k" of ¡he cut.
630 Applications oC Analytic FunctiOD Tbeory
y
~ 0
CD 1, 2)
® 0
® (j) ® ®
x u
/91
\.':J @ @ @
@ @ @ @
Figure 11.3
If we imagine the z plane to consist of a sort of compressible material and

think of the z plane as cut along the negative real axis and then pulled apart
along the cut in such a way that the edges of the cut are each rotated through
90°, so that all the material is compressed into the regio n to the right of the y
axis, then straight lines drawn on the plane may be expected to become distorted
into the shapes assumed by the corresponding curves sketched in the w planeo
We notice that the upper and lower banks of the cut in the z plane are
mapped onto the positive and negative parts of the v axis, respectively, and that
neighboring regions in the z plane are mapped into neighboring regions in the
w plane except for Ihose regions which are separaled by the cut, across which
transition is prohibited. Any continuous curve in the z plane not crossing the cut
maps onto a continuous curve in the w planeo
If neither the function fez) nor its inverse is multivalued, it may happen
that each plane maps as a whole into the other planeo If, as in the present case,
fez) is multivalued and its inverse is single-valued, a branch of fez) may map
the z plane onto only a portion of the w planeo The mapping w = Z2 would be
described by interchanging the z and w planes in the aboye sketche·s. In this
case fez) is single-valued but its inverse is multivalued, and a portion of the z
plane maps onto the entire w planeo
Since in the mapping (20) the derivativ~f'(z) is not zero for any finite value
of z, the theorem cited aboye states that any point in the cut plane (that is, any
point not on the cut) can be inc\uded in a region which is mapped onto a region
in the w plane under a one-to-one correspondence. It is seen, in fact, that any
region not inc\uding points on the cut can be so mapped in this case.
If we indicate the inverse of w = fez) by z = F(w), we notice that since
dF dz 1 1 (26)
dw = dw = dwjdz = j'(z)'
the condition /'(zo) =;t= O ensures that z = F(w) is an analytic function of w
when w = W o if F(w) is single-valued.
11.4. Conformal Mapping 631
If I(z) is analytic at z = zo, then, in particular, to any curve e passing

through Zo in (he z plane there corresponds a curve e' passing through the
point W o = I(:!a) in the w plane. If we consider a second point z, on C, and its
image w, on e', and write
~w w, - W n ~I (27)
~z = z, - Zo = ~z '
we see that this ratio is a complex number whose modulu s is the ratio of the
lengths of the chords (wow,) and (zoz,) and whose argument is the angle
between the directions of these chords. Thus in the limit as ~z and ~ w approach
zero, the limiting argument is the angle b etween the tangents to e and e ' at
corresponding points Zo and w o• and the limiting modulus represents a local magni-
fication lactar in the neighborhood 01 zo. Sut since for an analytic function the
ratio (27) tends to rezo), independently of the direction of the chord (zoz,), it
follows that if I(z) is analytic at z o, and if rezo) *
O, aU curves passing through
Zo are mapped onto new curves passing tbrough w o, all of wbicb are (approxi-
mately) rotated through the same angle arg I'(zo) and magnified in the same
ratio I I'(zo) I in the neighborbood of w o. Thus, in the neighborhood of a point
where I(z) is analytic and r(z) * O, relative angle and shape are preserved in
sucb a mapping, and the mapping is said to be conforma!. That is, a small closed
figure wiU map onto a similar closed figure with a certain nearly uniform rota-
tion and magnification.
In particular, if I'(zo) = koe i . , and ka *O, tben tbe tangent at Zo to any
mapped curve passing tbrough that point is rotated through exactly tbe angle
(xo into the tangent at W o to the image curve. Consequently, the angle between
two mapped curves intersecting at z a is the same as the angle between the two
image curves at W o if I(z) is analytic at Za and I'(zo) * O. It is important to
notice that this statement applies not only to the magnitude, but also to the
sense of the angle.
At points where I'(z) = O the magnification factor is zero and the angle of
rotation is indeterminate. Such points are often known as critical points of the
mapping. It should be noticed that a critical point Zo corresponds to a point W o
at which the inverse function z = F(w) is not analytic.
If we write
w = I(z) = u + iv = I,(x, y) + i/2 (x, y) (28)
and notice that the straight lines u = constant and v = constant are orthog-
onal in the w plane, we see that the curves l/eX, y) = 1, (x, y) = constant and
v(x, y) = 12(x, y) = constant in the z plane (of which these lines are the images)
also must be orthogonal at points where r(z) *
O. This is, if u(x, y) and v(x, y)
are the real and imaginary parts 01 an analytic function I(z), then the curves
u = constant and v = constant are orthogonal at points of analyticity where
I'(z) * O. At singular points of I(z) the u and v curves may or may not exist
and, in any case, need llOt be orthogonal. Similarly, at points for wmch r(z) *O
632 Applications of Analylic FunclioD Tbeory
the inverse function z = F(w) is analytic (if single-valued), and the curve sets
x = constant and y = constant in the z plane correspond to orthogonal sets
of curves in the w plane. Thus, in general, small rectangles bounded by coordi-
nate lines in either plane correspond to small "curvilinear rectangles" in the
other plane except near points wbere j '(z) is zero or nonexistent.
11.5. Application to Two-Dimensional Fluid Flow. Since tbe real part of an

analytic function of z satisfies Laplace's equation (Section 1004), it represents tbe
velocity potential of an ideal fluid flow in tbe xy plane (Section 6.19). Thus, if
we write
(z) = <)?(x, y) + iljl(x, y), (29)
where <1> is an analytic function of z, and consider <)?(x, y) as a velocity potential
corresponding to a flow velocity with components V., and V, in the x and y
directions at a point, we then have
V _ arpo (30)
y - ay
Since the curves ljI = constant are ortbogonal to the equipotential lines
rp = constant, tbey must be identified with the streamlines of tbe flow. In fact,
since the Cauchy-Riemann equations give
(31)
it follows that ljI(x, y) can be taken as the stream fimction of the flow. That is,
Slnce
pdljl = p( - V y dx +
V x dy) = PV • n ds = dj, (32)
where p is the fluid density, n the unit normal vector to a curve e, andjthe flux
across e, the difference between the values of ljI at two points in the xy plane is
numerically equal to the rate of mass flow of fluid with unit density across a
curve joining these points.
The complex function (z) is sometimes called the' complex potential; its
real part is tbe velocity potential and its imaginary part is tbe stream function.
We notice that, from Equation (64) of Chapter 10, we bave also
"""( ) _ d<l> _ a<l> _ a<)?
..,. z - dz - ax - ax
+ ¡.aljl
ax
_ V _ ·V
- x ¡ y. (33)
The function '(z) is frequently called the complex velocity; its real part is V x
and its imaginary part - V,. It may be seen that the conjugate function
'(z) = V x iVy +
can be considered as specifying tbe actual velocity vector.
In addition, we find that, since
 '(z) dz = (VX - iVy)(dx + i dy)
= + V y dy) + i(V
(V x dx x dy - V y dx)
= (V • u + iV • n) ds,
11.5. Application lO Two-Dimensional Fluid Flow 633
where u and n are unit tangent and normal vectors on a curve e, there follows
fe <1>'(z) dI. = fe V • uds + i fe n V • ds.
Thus we deduce that

~c<1> = (circulation) + i (flux), (34)
that is, that the change in <1> corresponding to a circuit about a simple closed
curve e has as its real part the circulation of V around e and as its imaginary
part theflux of V across e (when p = 1).
Now suppose that a second complex variable w = u + iv is defined as an
analytic function of the complex variable z = x + iy by the relationship
w = fez) (35)
and suppose that this mapping gives a one-to-one correspondence between
points in a region (R in the xy plane and a region (R' in the uv planeo Then the
equipotential lines and streamlines corresponding to a fiow in the region (R
will be mapped onto a corresponding configuration in the regio n (R'. If we
write the inverse of (35) in the form
z = F(w), (36)
the original potential function <1>(z) is then expressible as a function of w in the
form <1>[F(w)]. Since we have
d<1> d<1> dz <1>'(z )
(37)
f'(z)
and since <1>(z) and fez) are analytic, it follows that <1> is an analytic function of
w except at points in the w plane which correspond to points in the z plane
where f'(z) = O. Hence, if we think of ti and v as rectangular coordinates in
the w plane, we see that the real and imaginary parts of <1>[F(w)] satisfy Laplace's
equation in rectangular (u, v) coordinates, and hence the new configuration in
the w plane represents a new fiow pattern of an ideal fluid.
The complex velocity is of the form
(38)
or, equivalently,
v _ "V = d<1> dI. = <1>'(z) = V x - iVy (39)
u I v dI. dw f'(z) I'(z)
From (39) it follows that
,y
IV2
u
+ V2 _ "'¡V;
v -
+
II'(z) 1 '
V¡ (40)
so that the absolute velocity at a point o in the w plane is obtained by dividing

W
the absolute velocity al the corresponding point Zo in the z plane by 1 I'(zo) l.
Moreover, the actual (vector) velocities at the corresponding points will be
equal if and only if I'(zo) = l.
634 Applicatíons of Analytíc Function Tbeory
The streamlines r¡;(x, y) = constant and equipotential lines rp(x, y) = con-

stant map onto corresponding curves in the w plane, with equations obtained by
replacing x and y by their equivalent expressions in terms of u and v, or deter-
mined from the transformation (35) or (36).
We rnay verify directly that the velocity potential and strearn function
satisfy Laplace's equation in rectangular (u, v) coordinates as follows. Since ti
and v satisfy the Cauchy-Riemann equations, we have
and hence
arp _ arp arp
-ax - u'-ati
- u y -a V '
a rp _ u ~ _ u ~
2
I U .!L. arp _ u .!L.arp
ax2 - xx au Xyav T x ax au y ax av
_ arp arp...l... 2 a 2rp 2 a 2rp I

2
2 a rp
- Uxx Tu - U'yav I u, au2 - u,u y au av T u y av2 .
Similarly, we obtain
iJ!:!l _ ~ ~ 2
2 a rp a 2rp 2
a rp
ay- ? - Uyy a u + UXy a V + y
u -ua2 + y
2u,u a u a V T
I 2
Ux =.
aV-
When the last two results are added, there follows

2 2 2rp 2rp
a rp a rp arp 2 (a
V'rp = ax2 + ay2 = (u xx + uyy ) au + (u 2
x ..
I
u y ) au 2 + aav2 ) .
The coefficient of arp / au vaoishes, since l/eX, y) satisfies Laplace's equation
in xy coordinates. Also, since l/x - iu y = l/ x + iv x = f·(z) there follows
u; + 11; = IF(z) 1', and we have
2rp
iE!e ...l... ~ = I F(zW(a'rp2 + av
a ). 2
(41)
ax' I ay' au
(See also Problem lO 1, Chapter 6.) Thus, unless F(z) = O, the vanisrung of the
left-hand member implies the vanishing of the parentheses on the right. The
sarne result naturally holds also when rp is replaced by r¡;.
11.6. Basic Flows. In trus section we investigate the f10ws corresponding

to a few elementary complex potential functions.
Examp/e 1. Tbe potential

el> = Voz, (42)
with Va real, corresponds to the complex velocity
11 .6. Basic Flows 635
and hence represe n ts u nifonn flow with velocit y Vo in the positive x directioo (Fig ure
)) .4) . Simi larly , if IX is a real angle, the potential
(43)
corresponds to flow with velocity componen ts
V x = Vo cos IX, Vy = Vo sin IX,
and hence represents uniform fiow with velocity Vo in a direction making an angle IX
with the positive x axis (Figure 11.5). •
y
y
x
"
Example 2. For the potential

el> = k log z = k(log r + i8), (44)
with k real , the streamlines are
r¡t = kfJ = constant.
Hence the fiow is radial, with radial velocity
y
V, = arp / ar = k /r and with zero circUflÚerential veloc-
ity V. = r - I arp /afJ = 0 , so that there is a point source
at the origin if k > ° (Figure 11.6), or a point sink if
k < O. Its strengrh m, defined as the outward rate of
mass flow of fluid with unit density across an arbitrary
simple closed curve surrounding the origin, is gi ven by "
the change in r¡t corresponding to a positive circuit
around such a curve, and hence has the value m = 2nk
[see also (34)]. Similar/y, el> = k log (z - a) corresponds
to the presence of an isolated source (or sink) at z = a
if k is real. • Figure 11.6
EXfJ,mple 3. For the potential

el> = -ik log z = k(fJ - i log r ), (45)
with k real , the streamJines are

r¡t - -k log r = constant.
636 Applications of Analytic Functioo Theory
y Hence the ftow is circurnferential, with zero radial

velocity and with V o ~ r-1aq¡ / a{} = k / r. Tbe flow
may be called a circulalOry fiow, about a point vortex
at rhe origin, and is counterclockwise iI k > O
(Figure 11.7). It is conventional to define the
x strengrh m of tbe vortex to be m = 2nk. Reference
10 (34) (or direct calculation) shows al so that 21tk
is the value of the circulation fe V • dr around an
arbitrary simple c10sed curve e surrounding the
origin.t SimiJarly, it follows tbat the complex poten-
tial = -ik log (z - a) corresponds to a circula-
Figure 11.7 tory ftow about a vortex at z = a if k is real. •
Example 4. If a sink (negative source) of strength -m is located at the origin aod

a source of strength m is located at a neighboriog point z = a, the potential corre-
sponding to the combination is given, according to Example 2, by
m I m z-a
n -log z
.(z) = 2-[ + og (z - a)] = ?_n log z .
As the sink is brought into coiocidence with the source , the two will cancel unless at
the same time the streogth m is continuously increased in inverse proportion to the
separation I a l. If we write
K
m = TCiT' a = la I el',
the potential can be put in the form

 ( ) _ _ K log z - log (z - a)..!!.... _ - -Ke la log z - log (z - a)
.
• z- 2n a lal 2n a
Proceeding to the limil as a - . 0, we obtain
Ke'a 1
(z) = - - _ . (46)
2n z
The Iimitiog combination of source and siok is called a doublet (or dipole) of
strength K. The angle OG, measured from the positive real axis, specifies th.e direction
along which the coincidence was effected, and is caUed the orientation of the double!.
If we write z = re fO , Equation (46) can be expressed in the form
(z) = - ~ +e- ICO -,, ) = - 2~r [cos ({) - OG) - ¡sin ({) - OG)].
Heoce the stream1ines are the circles
111
T
= ..Ji..
2nr
sin ({) - OG) = constan!.
The velocity at any poiot is of magoitude K/2nr 2 , where r is distaoce from the doublet
(Figure 11.8). •
tThe term "circulation" is often used in the Iiterature to refer to the flow (here caBed a "cir-
culatory fiow"), as well as to the quantity fe V . dr associated with this f10w (or with any other
fiow).
11.6. Basic Flows 637
Figure 11.8
Examp/e 5. By superimposing two or more of the basic flows described, various other
ftows of iDteres! may be obtained. Thus, for example, if we superimpose upon a uru-
form flow wilh velocity VD in the posilive x direction Ihe flow correspondiDg lO a
doublel with orientation (J. = 7l and slrength 271C, we oblain the potential
el> = Voz + ..E...z .

In terms of real and imaginary parts we then have
. Vor 2 + e () + I. Vor 2 - e Sin
. () .
rp + 1'1/ =
r
COS
r
y
The streamJiDes
sin ()
'1/ = --(Vor 2
r
- e) = constant
are seeD lo include Ihe circle r = a as weIl as

the real axis sin () = O ir e = VD a 2 , so that x
(47)
In fact, it is easily seen Ihal we have obtaiDed in

(47) a represenlatioD of the ftow of an iDitially
uniform stream around a circular barrier (Figure Figure 11.9
11.9). The velocily is determiDed in Ihe form
a2 •
Vy = - V o -r 2 Sto 2B·,
V=-v'V; + V; = [el> '(z)[ =

r
V~"¡r4 - 2a 2 r 2 cos2B + a 4.
At the points (r = a; () = 0, 7l) on Ihe surface of the barrier, the velocity is zero.
These POiDtS are known as stagnation points . By superimposing also muItiples of the
circulatory f10w (45), we may obtain furlher flows in wruch the stream fUDction '1/ is
again constanl along the circular boundary and in which the f10w tends toward uni-
forrnity wilh large distances from Ihe boundary, but in which the stagnation points
are displaced aDd the ftow is no 10nger symmetrical with respect to the x axis. (See also
Problems 48-50 of Chapler 9 aDd Problem 30 of this chapter.) •
It follows from the preceding examples that the presence of a logarithmic

singular!I)) in the complex potential function corresponds to the presence of a
source (or sink) in the flow if the coefficient of the logarithmic term is real, and
to the presence of a vOrlex if the coefficient is purely imaginary. Also, a simple
pole at a point corresponds to a doublet at that point. If we recal! (see Section
10.10) that a function <1>(z) behaves near z~ just as <1>(1/1) behaves near 1 = O,
we see that the potential Voz, which represents uniform flow ("streaming") in
the finite plane, has a simple pole at Z~, and hence this flow ineludes a doublet
at Z~. Similarly, the potential k log z represents a flow with a source at the
origin and a sin k of opposite strength at z=; the potential ik log z indicates equal
and opposite vortices at the origin and at z~; and the potential c/z defines a
flow with a doublet at the origin but with no singularity at Zw
It should be noted 'that, since the "two-dimensional" flows considered here
are in fact only plane sections of flows imagined to not vary in the direction
normal to that plane, it follows that the "point" sources, sinks, vortices, doublets,
and curves (or fines) considered he re in the xy plane truly are sections of infinite
"line" sources, sinks, and so forth, and cy/inders (or planes) in the three-dimen-
sional flow.
11.7. Other Applications of Conformal l\lapping. As has been seen in

Chapter 9, it is frequently necessary in various fields to determine a function
of x and y which satisfies Laplace's equation and which takes on prescribed
values at points of a given curve e in the xy planeo Suppose that we ha ve some-
how obtained an analytic function w = fez) = u + iv which maps tbe curve e
onto the real axis (v = O) of the uv planeo The same mapping relation will trans-
form the prescribed values of rp along e to corresponding values at points along
tbe ti axis in the w planeo If now we can find in the w plane a solution of Laplace's
equation which takes on these values, this solution can be transformed back into
xy coordinates to give the solution of the original problem. Tbus, once the prop-
er transformation is known, the problem is reduced from the determination
of a function taking on prescribed values at points of a curve to the determina-
tion of one which takes on prescribed values along a straight lineo This latter
problem is considerably less difficult and, as a matter of fact, has been sol ved in
Section 9.14 [Equation (230)].
In the case of ideal fluid flow, the problem just discussed can be considered
as essentially reducing to tbe determination of a stream function If/(x, y) which
satisfies Laplace's equation, takes on a constant value at points along a pre-
scribed streamline, and behaves suitably at infinity. This is true since, once If/ is
known, its conj ugate rp(x, y), the velocity potential, is determined by the Caucby-
Riemann equations (31), and the flow is completely determined.
Similar problems involving steady-state temperature distributions, electro-
static fields, and so on, are of frequent occurrence.
In general, the direct analytical determination of a suitable mapping func-
tion is not easily accomplished. However, if the curve e is polygonal, that is,
11.7. Other Applications of Con formal Mapping 639
made up of straight line segments, such a mapping can be obtained by methods

to be presented in t he following section.
In certain cases it may be more desirable to map the curve e instead onto
the unÍl circle in the w plane and to solve the transformed problem by the method
of Section 9.4 or 9.5. As a single illustration we consider the mapping represented
by the relation
z = +(w + ~). (48)
Here the mapping function is conveniently expressed in a form solved for z.

If we introduce coordinates (p, rp) in the w plane by writing
w = u + iv = pe " ,
(49)
u = p cos rp, v = p sin rp,
x + ly. ="2l (1
pe' + l
/ie-".),
so that we have
x = +(p + ~) cos rp, y l (p

="2 - ti1).SIn rp. (50)
By eliminating p and rp successively from these equations, we obtain the rela-

tions
(5Ia)
X2 y2
---=..:.__
, - 2 = 1. ( 5 lb)
cos - rp sin rp
Thus it follows that the circles p = constant in the w plane correspond to the
ellipses (5Ia) and parts of the radiallines rp = constant correspond to the hyper-
bolas (5Ib), as represented in Figure 11.10. In particular, we verify that the unit
circle p = l in the w plane is tlattened into a two-sided "slit" in the z plane
y u
x u
Figure 11.10
640 Applications oC Analytic FunctioD Tbeory
between z = -1 and z = +
l. The other circles in the w plane correspond to
ellipses which decrease in flatness with increasing size. "Ve notice that the entire
cut z plane maps onto tbe exterior of the unit circle.
Now suppose that it is required to determine a harmonic function T(x, y),
say steady-state temperature, whicb takes on prescribed values along the upper
and lower edges of the slit, say
T(x, 0+) =f,(x), T(x, 0-) = f2(X) (-1 < x < 1).
For points on the slit (50) glves
p = 1, x = cos rp,
where O < rp < 71, for the upper edge of tbe slit and 71, < rp < 2n: for the 10wer
edge. Thus in the region p :> 1 of the w plane we must determine a harmonic
function T(p, rp) which reduces when p = 1 to fl(cos rp) wben O < rp < 71, and
to f2(COS rp) when n: < rp < 271,. This problem is solved in Section 9.4. The value
of T determined at any point (p, rp) is also the required value of T at the cor-
responding point (x, y), as determined by (50).
Example. As another illustration of the use of the mapping (48), we notice that a
circulatory fiow with circulation m in the w plane, and specified by the complex poten-
tial el> = - (im/2:n) log w, is mapped onto a flow with complex potential
el> = - im log [z
271,
+ (Z2 - 1) 1/2]
im
= - 271, cosh- I Z (52)
about Ihe slit in Ihe z planeo The streamlines and equipolenlial lines of the two flows
accordingly are the curves appearing in the z plane in Figure 11.10 and (he velocity
compone nI S are derivable from the relation
. del> -(im/2n:)
V x -¡Vy = dz = (Z2 _ 1)U2' (53)
Here, in order to specify a branch of (Z2 - 1)1/2 having a cut on the real axis between
z = -1 and z = 1, we may write
(54)
with the bipolar notation of Figure 11.11, where PI >- O and P2 >- O and where
O <: COI < 271, and O <: C02 < 271,. Then, in correspondence with a crossing of the real
axis, the angle (co I + co2)/2 is continuous when x < -1, has a jump of 71, when
-1 < x < 1, and has a jump of 271, when x> 1, so Ihat exp [i(co I + co2)/2] is dis-
continuous along the cut but continuous elsewhere. With this definition of the inter-
pretation of (Z2 - 1) 112, we verify easily thal a positive circuit around any simple closed
curve enclosing z = -1 and z = 1 increases el> by m, so that, according lo Equation
(34), the circulalion around any such curve is indeed m, the flow accordingly being
counterclockwise when m > O. AIso, along Ihe real axis V x = -(m/2n)(1 - X 2 )-1/2
on the upper bank of the cut, with the sign reversed on Ihe lower bank, while
V y = (m/2n)(x 2 - 1)-1/2 when x > 1, with tbe sigo reversed when x < -1.
11.8. The Schwarz-ChristoJIel Transformation 641
-1 x
Figure 11.11
More generally, the complex potential

im z
el>..:.. - - cosh-¡- (55)
2n: R
would specify the result of expanding this flow pattern in a ratio R to 1, so that the
points z = ±l, in particular, are replaced by the points z = ±R. •
11.8. The Schwarz-Christoffel Transformation. In this sectioo we describe
aod il!ustrate a mapping which sets up a one-to-one correspondeoce between
points on a polygonal boundary io the z plane and points 00 the real axis (v = O)
in the w plane. It will be convenient first to think of the mapping as from the w
plane to the z planeo
For this purpose, we consider the mapping z = F(w) for which
(56)
where U¡, U 2 , • •• , Un are aoy n points arraoged in order along the real axis in
the w plane (Figure 11.12), tI¡ < tl 2 < ... < u,,, and where the k's are real
constants and e is.a real or complex constant. By taking the logarithm of both
sides, we may rewrite Equation (56) in the form
log ;~ = log e+ k¡ log (w - u¡) + k 2 log (w - uz) + ... + k n log (w - un),

(57)
it being agreed that principal values are taken of al! logarithms on the right (so
that each associated angle is between O and 2n). We then consider the mapping
of the u axis onto its image in the xy plane, remembering (see Section 10.9) that
the magnitude of dz/dw at a point in the w plane is the local magnification factor
in the mappiog from the w plane to the z plane, and that the angle (or argument)
of dz/dw, which can be expressed in the form
4 ;~ = 1m (Iog ;~) ,
642 Applications of Analytic Function Tbeory
z*
x
gives the angle through which the tangent to a mapped curve passing through that
point is rotated in the mapping. Thus we may obtain from (57) the relation
4:~ = 4C + k¡ 4(w - u¡) +k 2 4(w - u 2) + ... + k n 4(w - un) (58)
by taking imaginary parts of both sides.

Let the point at infinity in the w plane be mapped onto the point z* in the z
plane, where z* may be z~. Then, if we consider the image of a point w = u
moving from an infinite distance toward the right along the negative real axis
in the w plane, we notice that so long as w = u < U ¡ the numbers w - u 10
W - u 2 , ••• , w - u" are aH real and negative. Hence their angles are aH con-
stantly equal to 70 in Equation (58), and so there follows
dz
4
dw
= 4C + (k , + k2 + ... -t- kn)n (w = u < u , ). (59)
Thus the portion of the u axis to the left of the point u, is mapped onto a straight
fine segment, making the angle defined by (59) with the real axis in the z plane,
and extending from z* to z" the image of w = u ¡ (Figure 11.13).
Now as the point w crosses the ,point u, on the real axis, the real number
w - u, becomes positive, so that its angle abruptly changes from 70 to zero.
Hence 4(dzj dw) abruptly decreases by an amount k, 70 and then remains constant
as w travels from u¡ toward fl 2 • It follows that the image of the segment (u¡u 2 )
is a line segment (Z¡Z2) in the z plan e making an angle of -k,n with the segment
(z*z ,).
Proceeding in this way, we see that each segment (u pU p +,) is mapped onto
a line segment (z pZ p+ ,) in the z plane, making an angle of - k pn with the seg-
ment previously mapped, Thus, if the interior angle of the resultant polygonal
contour at the point z pis to have the magnitude IX p , with O -<: IX p -<: 270, t we must
set 70 - IX p = -kpn or
k = IX p _ 1 (60)
p 70
tThe identification (J;p = O can apply only to a verte x at Z_, as in Example 3 of this section.
U.S. The Schwacz-Christoffel Transformalion 643
in Equation (56). After an integration, we then conclude that Ihe mapping
z = e f (w - UI)"'(w - U2)"' ... (w - u,,)""dw + K, (61)
where e and K are arbitrary (complex) conslants, wi!! map the real axis (v = O)
o/ the w plane onto a polygonal boundary in the ¡; plane in such a way that the
verliees Z l' ::2' ••. , z"' with interior angles <x" <x 2 , ..• , <X"' are the images o/ the
points u l' U2' . . . , Un'
We notice that for the final segment (w = u > u") the numbers w - u,
are al! real and positive and hence possess zero angles, so that this segment is
rotated through the angle
4 dz = 4c (62)
dw
For a closed polygon the sum of the interior angles is
<XI + <X 2 + ... + <X" = (n - 2)n,
and hence also
k l
+ k2 + ... + k" = (n - 2)n - n = -2.
n
Thus, according to Equations (59) and (62), the two infinite segments of the line
v = O are rotated through the angles 4C - 2n and 4C, as is clearly necessary
for a closed figure.
If the constant C is written in the form C = ce'P, we see that it comprises
an arbitrary magnification factor e and a rotation p and, in fact, that according
to (62), the angle P is the angle through which the infinite segment to the right
of w = Un is rotated when it is mapped (onto a finite or infinite segment). If we
write K = Uo + iv o, we see that tbe additive constant K represents an arbitrary
translation of the polygon, without rotation or distortioil, through the vector
Uo + iv o '
lt can be shown that the numbers u" U z' ... , Un and the complex constants
e and K can always be chosen so that any prescribed polygon in the z plane is
made to correspond point by point to the real axis (v = O) in the w plane and,
in fact, that the correspondence can be set up in infinitely many ways, in that
three o/ the numbers u 1> u 2 ' . . . , Un ean be determined arbitrarily (provided that
they are appropriately ordered).t
Further, the mapping can be shown to establish a one-to-one correspond-
ence between points in the interior of the polygon in the z plane and points in the
upper ha/f of the w plane o In this connection, one must be careful to label the
vertices on the polygonal boundary as ZI' Z" . . . , Zn in the order in which they
would be passed by an observer traversing the boundary wilh the interior o/
the polygon to his le/t. The corresponding points u l ' U2' ... , u" on the ti axis in
tWhen the polygon has a vertex at Z . , the mapping of this vertex onlo u~ amounls lo
specifying one of the three assignable u·s. (See Problem 56.)
[he w plane then are to be labeled in order from left to right, as has already been
specified.
The propriety of this rule follows from the fact that the conformal property
of lhe mapping guarantees that, if an observer traveling along the polygonal
boundary turns to his left from the boundary (into the polygon), his image in the
w plane will be traveling forward along the II axis and also will turn to his left
(into the upper half-plane) .
Thus, for example, if the open region above (and to the left of) the polygonal
boundary in the z plane in Figure 11.14 is to be mapped onto the upper half of
y v
1 1
o x u, u
Figure 11.14
the w plane, with the two finite vertices mapped (say) onto u 1 = -1 and u 2 = 1,
then the proper identification is Zl = O and Z2 = i, in order that the vertex
sequen ce z = 0, z = i, z = z~ map onto the image sequence w = -1, w = 1,
w = U w Consequently, there follows also al = n / 2, a2 = 3n/ 2 and k 1 = -1/ 2,
k = 1/ 2, and hence
dz _
-
dw
- c (WIV +
_
1)1 /2
I
.
After the integration, C and the constant of integration K wou1d be determined
so that z = O f---> W = -1 and z =i f---> IV = l. In order to map instead the
region belolV (and to the right of) the polygonal boundary onto the upper
half-plane, one would take Zl = i, Zz = °
and a l = n / 2, a, = 3n /2. In this
special case it happens that the same differential equation for the mapping
function would be obtained, but different values
IY for C and K would result from then making
I
I
°
z = i f---> IV = - I and z = f---> IV = 1.
Certain artifices may also be used in obtain-
ing mappings related to problems involving the
exterior of a cIosed polygon . As an example, sup-
x pose that we wish lO investigate the effect of the
barrier ofquadrilateral section, sketched in
Figure 11.15, on an initially uniform flow of an
• ideal fluid. From the physical symmetry it is cIear
that, when no circulation is present, one stream-
Figure 11.15 line will consist of the part of the x axis exterior
11.8. The Schwarz-Christoffel Transformation 645
IY
v
U, u
Figure 11.16
to the polygonal section and, say, the upper half of the boundary. Thus we are
led to consider a mapping which tran sforms trus streamline into the u axi s of the
w plane, as is indicated in Figure 11.16. Here we map the region above the poly-
gonal contour onto the region v > 0, taking as the "interior angles" those indi-
cated in the figure. The mapped region may be considered as an infinite polygon
having the con tour as its finite boundary. Once the fiow is determined in the
upper half of the z plane, by mapping an appropriate uniform fiow in the upper
half of the w plane, the fiow is determined in the lower half-plane by symmetry.
A superimposed fiow, with circulation m, and with a velocity vector which also
is tangential to the barrier but which tends to zero with increasing distance
from it, would be obtained by effecting the same mapping on the complex
potential
im
<1>= --cos h- 1 -
W
, (63)
2n: R
provided that the upper boundary of the barrier is mapped onto the segment
-R -< u -< R of the real axis in the w plane. This fact follows immediately from
the considerations at the end of Section 11.7.
A similar procedure clearly can be applied in all analogou s cases in which
there is an axis of symmetry parallel to the direction of ftow .
A generalization of the Schwarz-Christoffel transformatíon permits the
mapping of the complete exterior of a closed polygon in the z plane onto the
upper half of the w plane.t In this connection, it should be noted that only a
simply connected region can be mapped conformally onto another simply con-
nected region, so that, for example, an annular region could not be mapped onto
a half-plane. However, when the point z~ is included, the regíon outsíde a simple
closed curve e is in fact simply connected since any simple closed curve enclos-
ing e can be "shrunk to a point" at Z=.
Many illustrations of the Schwarz-ChristoffeI transformatíon may be found
in the Iiterature. For closed polygons, even in the simplest cases of a triangle and
a rectangle the use of elliptie funetions is necessary, and in other cases resort
usually must be had to numerical analysis. We here consider three applications
to open polygons, in increasing order of complexity.
tSee, for example, Reference 1 of Chapter 10.

Example 1. As a simple example, we consider first the case in which the polygon is
merely the interior of the sector of the ::: pJane bounded by the positive x axis (e = O)
and the line e = tx (Figure 11.17). If we take the poiot ti" corresponding ro the single
finite vertex at z = O, at the origin (Ul = O), Equation (61) gives
z = e J w k dw,
where k = (tx/n) - 1 and e is an arbitrary constant, and hence z = [CJ(k + l)JwH ¡ + K.

v
y
o
T=O
/
Z¡ =0 1 x U,=O 1 u
Figure 11.17
For z = O and w = O to correspondo the constant of integration must vanish, but

e remains arbitrary. If, for convenience, we require that the point z = 1 map onto the
point w = 1, we find that e = k + 1, and there then foUows
Z = wO:;/7f, (64)
subject to the obvious branch specifications O <: 4z <: tx and O <: 4w <: n.
Suppose now that we require the solution T of Laplace's equation in the infinite
sector O <: e <: tx for wruch T = 1 when O < x < 1 and T = O when x> 1, along
the edge (] = O, and for which T = O everywhere along the edge = tx (Figure 11.17). e
The mapping (64) transforms this problem to that of determining the solution of
Laplace's equation in the half-plane v >- O which reduces, when v = O, to 1 when
O < u < 1 and to O otherwise. But the solution of this problem is given by Equation
(230) of Section 9.14 in the formT
1 r¡ v d¿; 1(tan- 1-u U)
+ tan-¡-
T = n Jo v 2 + (¿; - u)2
= -
n
1 ---
v v
= J.. cot-¡ u
2
+ v2 - u (65)
n v
or, in poJar coordina tes (p, rp),
T = J.. cot-¡ p - cos rp (66)

n sin rp
tI! is useful to notice that the conventions -n/2 -< tan-¡ x -< n/2 and O -< cot-¡ x -< n,
when x is real, imply the relations
-cot- l [(AB - 1l/(A + Bl] if A + B S O
tao-¡ A + tan- 1 B = tan- l [(A + Bl/(l - ABl] if AB:5: 1
{
cot-¡ [(1 - ABl/(A + B)] if A + B >- O
when A and B are real.
From Equation (64) there follows

rp
ne
IX
and hence the required so luti o n is o btained in the convenient dosed form
T - c o t- I [rO " - co, (n e /IX)] .

- -1 (67)
n sin (n e / IX)
The more general case in which T is prescribed arbitrarily along the two radial
boundaries clearly can be treated io the same way. •
Example 2. A s a second exarople , we investigate the

effect of a plaoe barrier of finite length 2a 00 ao ini- _ IY
tially uniform f10w of ao ideal fluid perpeodicular ro I
it (Figure 11.18), when no circulation is preseo!. This • ,
is the limiting case of the problem discussed io con- a
nection with the barrier of Figure 11. I 5, when the •
interior angles of the quadrilateral teod to O and 10 2n. x
á
Thus we are led 10 consider the mapping of Figure •
11.19. Here we wish to map the negative x axis onto
Ihe segment u < l/I, the left-hand edge of the barrier
onto the segrneot t/, < t/ < "2, the right-hand edge of
the barrier 0010 the segment l/o < " < t/3 , and the Figure 11.18
positive x axis onlO ,, > t/3. For purposes of sym-
metry, we take ", = -b, "2 = O, "3 = b, where b is as yel unspecified. With
IX, = IX 3 = n/2, IX 2 = 2n, there follows , from (60), k, = k3 = -1, k 2 1, and
hence (61) gives
z = c f (w2
w dw2
b )"2 -- C(W o - b 2 )' / 2 + K, (68)
where K, C, and b are yet to be determined. The desired correspoodence

w = -b <---> z = O, w = 0<---> z = ia, W = + b <---> z = O (69)
theo is obtained if we set
K=O , Cb = a, (70)
so tha! the mapping becomes

w = _b ( Z2 + a2 ) 1,' 2. ( 71)
a
IY v
b·
Ul u
Figure 11.19
Finally, we notice that =~ ....,

w~ , and hence if a unifonn fl ow in the w plane is to be
undi s torted in the mapping at large dislances from the barrier, we mus t have
dw/ dz ---+ I as z ---+ z~, or
Izl-~
·
l 1m dw =
-d
z
l 1m
I,, _~
' [b
-
a (z
2
z
+ a 2 ) 1/ 2
] - ..t
a
= l '
'
b = a. (72)
Thus the desired mapping takes the fonn

z = (w 2 - a 2 ) 1.' 2, w = (Z2 + a 2)1 / 2 = f(z), (73)
where the angle associated with each of the roots is 10 be between O and 7t.
H is readily verified that this mapping gives the specified correspondence. In
particular, points in the interior of a quadrant in one half-plane correspond to points
inside the same quadrant in the other half-plane.
A uniform flow cll(w) = Vow in the upper half of the w plane then maps onto the
desired flow in the z plane,t with complex potential cll[f(z)], or
cll = V O(Z2 + a 2)1 /2 = rp(x, y) + iljl(x, y). (74)
The streamlines and equipotential lines are obtained by equating to constants the real
and imaginary parts rp and 1jI.
The complex velocity is given by
dcll Va z Vo(x + iy)
Vx - iV, = dz = (Z2 + a2)1 / 2 = [( X2 _ y2 a2) + +
i(2xy)]'/2' (75)
where the sign of the denominator is most conveniently determined (without analytical
specification of the relevant branch) by noticing that, for a given point z, the deno-
minator represents the image point w. In particular, for points on the y axis there
follows
. _ iVo Y
Vx - IV, - (a 2 y2)l / 2' (76)
with the proper interpretation. At any point on the right-hand edge of the barrier the
denominator is real and positive, since the image point is on the positive u axis, and
there follows
v = _ Voy ,
, ,.,¡ a2 _ y2
whereas along the left-hand edge the sign is reversed. Along the remainder of the posi-
tive y axis (y > a) we have
and hence he re
V y = O.
The flow velocity is seen to be infinite at the ends of the barrier.

A superimposable fiow with circulation m, in which the flow is tangential to the
barrier and the velocity tends 10 zero with increasing distance from the barrier, is
tAllernativeIy, b could have been assigned any convenient positive value (s ay, b = 1 or b = a)
in advance, in which case the condition (dw/dz)z~ = 1 could not have beeo imposed, unIess
b = o. Here one would wrile <l> = A IV and determine A in such a way that (d<l>/dz)z_ = Va,
and hence A = Voa/b, and would again obtain (74).
obtained from (63), with R = a and w = (ZO + a 2 )1.·2, in the form
1m cosh - , (az22
<ll = - 271: + 1)1/2
or, equivalently,
'"
w = -2n Sin -
im·hlz
Q' (77)
Example 3. To illustrate additional features and techniques of the mapping, we con-

•
sider the determination of the flow of an ideal fluid from a long channel with parallel
walls. The width of the channel is denoted by 2nl, and the coordinate axes in the z
- =
----.-------,,
---,..
---f
-- -'"
x
A~
---- y/B
Tri C~«««««(
t D
-=
(a) (b)
Figure 11.20
plane are located (for later convenience) as is indicated in Figure l1.20(a). We make
use of the syrnmetry with respect to the x axis and consider only the flow in the upper
half-plane. The polygonal boundary to be mapped then can be considered as the limit
of the boundary indicated in Figure 11.20(b) as the points A and C tend to infinity and
y o u
- ~ :~~\"CE\~~
1 .
(-1 + iTrI)
Figure 11.21
the indicated angles tend to 2n and O. We next attempt to accomplish the mapping of
Figure 11.21, in which the point Bis mapped (for convenience) onto the point -1
in the w plane and the point (el C 2) at infinity is mapped onto the origino We should
notice, however, for future reference, that the desired fiow around the barrier is in the
direction elBA, whereas that along the axis of syrnmetry is in the direction e 2 D. This
condition suggests that we shall need to map a' flow from a source at the origin in the
w plane, rather than a uniform flow as in Example 2, onto the desired flow.
With l/, = -1, l/2 = O and k l = 1, k 2 = -1, Equation (61) gives
dz CW + 1 (78)
dw w
650 Applications of Annlytic Func!ion Theory
and hence there fotlows

z = C(w + log w) + K, (79)
where K is a constan! of integration . Since log w is multivalued, we must se lect a
specific branch of that function. Sioce Ihat branch must be analytic in the upper half
of the w plane (and since only that half-plaoe is involved here), we adopt the definition
log z = log I z I + ¡{}p (O <: (}p <: n).
We next determine the constants C and K.

It frequently is convenient to obtain conditioos determioing unknown coostaots
by essentially integrating the equal members of the basic differential equation betweeo
corresponding limits. In illustration, we notice that transition from CItO C 2 in the z
plane must correspond to traosition from C'l 10 C~ in the w plane. The transition io
the w plane must be considered as takiog place along a semicircle of vanishing radius
around the braoch point at w = O. U the radius is denoted by p, then near w = O we
have, from (78),
dz = C w + ] dw _ C dw as Iwl = p ~ O
'v w
and hence, on the small circle,
dz _ C ¡pe'; dq; = iC dq;. (80)

P ~
Thus, in the limit as p ~ O, we have
z lc,
c,
= icfo dq; =
•
-inC, -inl = -inC,
and hence
C = l. (81)
lo order that the point B (z = -1 + inl) correspond to B' (w -1), we must
have, from (79),
-1 + inl = I[ -] + log ( -])] + K = -1 + inl + K,
and heoce
K=O. (82)
(The position of the origin in the z plane was chosen, after a prelirninary analysis, in
suc h a way that this convenieot result would be obtained.)
Heoce the desired mapping is given by
z = I(w + log w) . (83)
To verify the correctness of this result, we may write w = p e l •• Then if q; = €, a simple
calculation shows that z =
I[(p + log p) + i(1 + p)€] except for higher powers of €;
and if q; = n - €, there follows similarly z = Ir
-[p - log p] + i[n + (p - l)€]).
From these resuIts it is easy to verify the correspoodeoce between the upper boundary
of the u axis (as € ~ 0+) and the path ABC I C 2 D.
Next it is necessa r y to determine a f10w in the w plane, for which C 'I B ' A ' aod
CíD' are slreamlioes. Also it is necessa ry thal lhe associated complex potential fuoc-
tion <l> have the property that the velocity d<l>/dz shall reduce to a uniform velocity
V o in the x direction at the point (C I C 2 ) in the z plane . These, however, are ¡he only
requiremenls 10 be sOlisfied. Although <l> can be determined from these cooditions by
purely direct methods, it appears that in the preseat case a fiow from a source at the
11.8. The Schwarz-Christolfel Transformation 651
origin in the w plane will have the desired properties. Hence we write
<l> = k log w
and attempt to determine k in such a way that the velocity d<l>/dz in the z plane will
tend to VD as z --' (C I C z ) and, correspondingly, w --' O. Making use of (78) and (81),
we obtain the general result
d<l> dw d<l> 1 w d<l>
dz = dz dw = T w + 1 dw'
and hence, with <l> = k log w, there follows
k 1
dz T w + 1
Setting w = O, we obtain the condition
k = IVo
so that thedesired potential function takes the form
<l> = IVo log w. (84)
The corresponding potential function in the z plane then is to be obtained by
eliminating w between Equations (84) and (83), and hence is the solution of the equa-
tion
z = l(e<t>/lVo + <l».
IVo
(85)
Whereas <l> cannot be expressed as a function of z in explicit closed form, we may

write
<l> = !p + ir¡!, (86)
where, in consequence of (84) and the limitation of permissible values of the logarithm,
-ce < !p < ce, O -<: r¡! -<: nlVo, (87)
and Equation (85) thus is equivalent to the two real equations
~ = e" cos p + a,
(88)
j = e" sin p + p,
with the abbreviations
(-ce < a < co, O -<: f3 -<: n). (89)
These parametric equations are convenient in plotting the streamlines f3 = constant

and the equipotential lines a = constan!. It can be verified that the streamlines for
which O -<: p -<: n/2 extend indefinitely in the x direction, whereas on the remaining
strearnlines a maximum value of x exists, so that these streamlines turn back around
the channel walls. •
In each of the preceding examples, the regio n enclosed by the polygonal

boundary could instead be considered as a homogeneous conducting sheet
carrying an electric current, or as aplane section of an electrostatic field
between long plane or cylindrical conductors. The stream j'unction r¡! and the
652 ApplicatioDs of Analytic FUDCtiOD Theory
stream/ines 1f/ = constant in the problem of .fluid .flow then would correspond
to the potentia/ funetion and to the equipotential fines in the problem of eleetro-
staties, whereas the potentia/ funetion rp and the equipotential fines in the flow
problem would correspond to the stream funetion and to the fines of force in
the electrostatic problem.
In two-dimensional electrostatics, the negative gradient of the potential
function is the electric field intensity E, whose tangential component must
vanish on the surface of a perfect conductor. The normal component of E at
such a surface is proportional to the surface charge density (J. The stream
function then has the property that the difference between its values at two
points on the conducting boundary is proportional to the total charge q, per
unit height of the conductor, on the portio n of the conductor joining those
points. (See Problems 57-62 and Reference 11.)
11.9. Green's Functions and the Dirichlet Problem. In correspondence with

a linear differential operator L, a related region CR, and a specification of the
nature of appropriate information to be prescribed along the boundary of CR,
there often exists a certain associated function, called a Green's funetion, which
plays an important role in the solution (or formulation) of problems governed
in CR by differential equations involving the operator L.
In this section and in Section 11.10 we identify L with the two-dimensional
Laplacian operator V 2 and suppose that it is to be associated with a related prob-
lem of Dirichlet type, so that the solution is specified in advance along the
boundary e of CR, in which case the determination of the Green's function can
be related to the problem of mapping the interior of CR conformally onto an
upper half-plane, when CR is simple. Section 11.11 deals with sorne other two-
dimensional examples, one of which is the Neumann problem associated also
with L = V2, and with conformal rnapping.
Green's functions also are defined over intervals for one-dimensional prob-
lems (see Problems 55-59 of Chapter 1 and Problems 65, 66, and 79-82 of this
chapter) as well as over regions in three or more dimensions, in correspondence
with appropriate linear differential operators and boundary conditions.
Each Green's function can be defined in a variety of equivalent ways, as
will be illustrated. In addition, it should be noted that such definitions in the
literature lead frequently to a function which is the negative of the one to be
defined here, or occasionally to another constant multiple of that function.
For the two-dimensional operator L = V2, our starting point is the two-
dimensional specialization of Green's theorem, Equation (123) of Chapter 6,
as applied to a region CR of the xy plane, bounded by a curve C. Here it is con-
venient to think of the point P(x, y) as a temporarily fixed point in CR and to
use'; and 11 to denote the "dummy variables" of integration, so that Green's
theorem can be written in the forro
11.9. Green's Functions and tbe Diricblet Problem 653
ep(x.y)
------+----------------------------------~
Figure 11.22
where here V = ¡aja!; + j aja17 in rectangular coordinates and ajan = n • V,

with the unit normal vector n directed outward from <R at a point of e (Figure
11.22).
Suppose now that rp satisfies the Poisson equation
V2rp=h inCR (91)
and is to be such that
rp=f on e, (92)
where h andf are prescribed functions. If a function G can be determined so that
V2G = b(!; - x)b(17 - y) in <R, (93)
where J(t - e) is the deltafunction of Section 2.6, located at t = e, and having
the basic property that
<
~ 1:(,)
(e a),
s: F(t¡.5U - ,) d' (a<e<b),
(e> b),
when b > a [see Problem 31(a) of Chapter 2], then since b(!; - x)b(17 - y),
considered as a function of !; and 17, is a delta function located at the fixed
point P(x, y), the double integral on the right in (90) will reduce to rp(x, y). If
also G is such that
G=O on e, (94)
Equation (90) wiU take the forrn
II 6\ Gh d!; d17 = - f/ ~~ ds + rp(x, y),

where both integrands involve only known functions.
The function G satisfying (93) and (94), which will depend upon the current
variables (x, y) as well as the dummy variables (!;, 17), is in fact the Green' s
funetion associated with V2 in <R when the related boundary conditions prescribe
654 ApplicatioDS of Analytic Function Theory
the so/uríon of a problem along the boundary of eR, so that the problem is a
Dírich/el prob/em. Ir the Green's funetion is denoted by G(x, y; C;, r¡), the aboYe
result transposes into the form
rp(x, y) = fe f ~ ds + ff <H G(x, y; C;, r¡)h(c;, r¡) de dr¡. (95)
It should be noted that, in addition to the faet that formal operations involv-
ing the delta function need to be jusúfied,t it is also true that sioce the function
G is unbounded when the point P(x, y) and the point Q(e;, r¡) approach each
other (as is next shown), the rigorous mathematical proof that (95) does indeed
satisfy (91) and (92) when h andfare sufficiently well behaved is rather formida-
ble, particularly when the region eR also is unbounded, and is omitted here. In
addition, in the examples which follow it will be tacitly assumed that the behavior
of h and f is in fact satisfactory.
We notice that when f = O, so that rp = O on e, the solution rp is given
completely by the integral over eR, and when h = O, so that V2rp = O in eR, the
solution is provided by the integral along C. In addition, it is seen that if the
differential equation (91) is generalized to the equation
V2rp + Arrp = h, (96)
where A is a constant and r is a prescribed function, and if the boundary con-

dition (92) is unchanged, then the additional term
- A ff <H G(x, y; C;, r¡)r(c;, r¡)rp(c;, r¡) de; dr¡ (97)
is introduced into the right-hand member of (95). The result is an integral

equation to be solved for rp, with rp appearing in the integral (97) as well as
outside it. Whereas such an equation usually cannot be sol ved analytically, in
closed form, it often is preferable to an alternative formulation of a certain
problem when an approximate solution is to be obtained, by numerical methods
or otherwise.t (See, for example, Reference 7.)
In order to exhibit the behavior of the present Green's function as the
point Q(C;, r¡) approaches the point P(x, y), where P is an interior point of eR, we
consider the result of integrating V2G over a circular disk D interior to eR, with
center at the point P (Figure 11.23). By making use of the two-dimensional
tThe juslificalion of Ihe use of (90) when V 2G is identified with Ihe delta function could con-
sisl of fírsl defining V 2G 10 be a con tinuo us fllnction which (say) vanishes except in the circular
disk (.< - ' ';)2 + (y - ,,)2 -< €, and which has the properly Ihat ils integral over thal disk is
unily for aH posilive vallles oC €, Ihen applying (90), and aftenvard laking Ihe limit of the result
as € --> O. Similar delayed limiling processes also are lO be imagined in connection with sub-
sequent operations involving the delta functions.
tTreatments in which the principal interest in Green's functions centers on their use in inlegral
equalions generally define G with an algebraic sign Opposile to that used here, so that Ihe rninus
sign Ihen is nOI present in (97).
11.9. Creen's Functions and the Dirichlet Problem 655
Figure 11.23
specialization of Equation (127), Chapter 6, we obtain the relation
Jlo V2G dI; dll = L~~ ds, (98)
where r is the circular boundary of D. Since G satisfies (93), it follows that the
left-hand rnernber of (98) has the value 1, regardless of the radius of r, so that
(98) gives
L~~ ds = 1. (99)
If we denote by R the distance PQ,

R = ,y"(x - 1;)2 + (y - 11)', (100)
we see that on r we rnay write
ds = R dO,
Hence (99) becornes
r
Jo
2
• R aG dO =
aR
1
and we deduce that Ihe mean value of R aG/aR on r is 1/(2n), so that for srnall
values of R we rnust have aG¡aR 1/(2nR), and henceN
G ~
I
-100" R (R ~ O). (101)
2n o
Thus, in the present case, the Green's function rnust becorne logarithrnically
infinite as P(x, y) and Q(I;,lI) approach coincidence inside CR. Consequently,
an allernalive definition of G consists of the requirernent that
V2G = O in CR when P(x, y) * Q(I;, 11), (102)
thal (10 1) hold as P and Q approach each other inside <R, and that (94) be
satisfied, so that G vanishes on C. The term (2n)-1 log R is sometimes called the
principal part of G. Since V'(log R) = O when R 7"= O, as is easily verified, it
follows al s o that we may write
G(x, y; .;, r¡) = J-_n log R + g(x, y; .;, r¡) (103)

where
I
and g = - - lo!! R on C. ( 104)
2n -
It is found in Problem 63 that
G(x, y; .;, '1) = G(';, r¡; x, y) (105)
in the present case, so that the Green's function is syrnmetric In (x, y) and
(.;, r¡). Thus the operator VZ in (93), (102), and (104) can be taken to be either
a'/ a';2+ a' / ar¡z or az / a .. ' +
az / ay' (or to be either of two equivalent alterna-
tive forms in terms of another coordina te system in the plane).
Example 1. When <R is the upper hall-plane, the definition of (103) and (104) permits
the determination 01' G by inspection. For since the function
,f- log R
_Te
=-21 10gv'C~ -
7r
';)2 -r (y - r¡)2
satisfies Laplace's equation in Ix , y) for fixed (.;, r¡l, when (x, y) * (.;, r¡), the function
,f-
_n log v' (x - ';)2 + (y + '1)"
also satisfies Laplace's equation in (x, y) for any (x, y) and (.;, r¡) in <R, since then
y> O and r¡ > O, and clearly the differel7ce between these functions vanishes when
y = O or r¡ = O. Since also that difference, which can be written conveniently in the
form
G=-Io1
4n
[(x -
';)2 + (y - '1)2 ] '
g (x - ';)2 + (y + '1)2
(106)
tends to zero as x' + y' ._ - =

or ';2 + r¡2 - co in the upper half-plane, it satisfies
(103) and (104) in that region and is in fac! the required Green's function.+
A simple calculation shows that along the boundary r¡ = O there follows
~~ = [ - ~~I= o = ~ (x _ ';~2 + y2 (107)

and hence the solution of the problem
a 2rp , a1rp_
ax2 a y2 - h(x, y),
( 108)
rp(x, O) = /(x), lim rp(x, y) = O
x f +y:_·O
tA procedure of this type, in which superposi,ion is used to achieve antisymmetry of a con-

tinuous runction about a line when the function is to vanish on the line, or symmetry when the
normal derivative of the funclion is to vanish there, is orten called the merhod o[ images. (Com-
pare Problems 83 and 84 of Chapter 9 and Problems 32 and 33 ofChapter 11, as well as Prob-
lem 33 of Chapter 8.)
11.9. GreeD ' s FunctioDs and Ihe Dirichlet Problem 657
in the half-plane - co < x < co, O <: y < co can be written in the form
q¡(x, y) = _ ~ (x ~1n
7t1 f~ ~)2 de;
+ y2 + __f- j'- o G(x, y; e;, 1/)h(e;, 1/) de; d1/, (109)
where G is defined by (106), by virtue of (95). In the special case h = O. this result prop-
erly reduces to Equation (230) of Section 9.14. Here, as in other examples to follow,
it is assumed that f(x) and h(x, y) are such that the integrals in (109) exist and tend
to zero as X2 + y2 _ 00 when y > O. •
Examp/e 2. In order to determine the Green's function ofExample 1 by a more direct

method, we can write G = (2n)-1 log R + g, where g(x, y; e;, 1/) is the solution q¡ of
the problem
a2 q¡ ~ _
ax2 + ay2 - O,
1
q¡(x, O) = - 2n log R(x, 0;';,1/), lim q¡ = O
;c~ + y2_0
in the upper half-plane. Here .; and 1/ are to be treated as constants while x and y vary.
The solution can be obtained (for example) by use of the Fourier integral and is, in
fact, given by Equation (230) of Section 9.14 in the forro
• J'
g(x, y, ." 1/) = - 4n2
1 J-
__
y log [(u - ';)2 + 1/2]
y2 + (u _ X)2 duo
This integral can be evaluated by first differentiating with respect to 1/, then using
residue calculus (Section 10.13), and finally integrating with respect to 1/. A somewhat
involved calculation thus yields the anticipated result
g(x, y; e;, 1/) = - 4~ log [(x - e;)Z + (y + 1/)2].

•
Example 3. Another direct determination of the Green's function in Examples 1 and
2 can be based on the identificationG(x, y;';, 1/) = q¡(x, y), with'; and 1/ again treated
as constants, where rp is the solution of the problem
~ aZq¡ _
axz + a yZ - O(x - ';)O(y - 1/),
q¡(x, O) = O, lim q¡(x, y) = O

;cZ+y2_0
in the upper half-plane. Perhaps the simplest method of solving trus problem is by use
of the Fourier transform (see Problem 64). •
In physically motivated problems, it is often possible to in terpret h and rp

in (91) (and in other sucb equations) as cause and e/lect, or as input and OUlpUI,
respectively. In such cases, the definition (93) s tate s tha t G(x, y; e;, 1/) can be
interpreted as lhe effeCI al P(x, y) due lO a unil cause dislribulioll concenlraled al
Q(';, 1/), so tbat Gi s sometimes referred to as an "i nfluence function," and the
syrnmetry property (l05) then corresponds to what is often called a " reciproc-
ity relation" in applications. Tbe condition (94) requires here that no causes
(or etfects) be pre sent along the boundary of the relevant regio n <R when G is
observed.
658 Applications oC Analytic FUOCtiOD Theory
For example, if cp is interpreted in (91) as veloeity potencial in a two-dimen_

sional fluid fiow, then h(x, y) measures the intensity of a distribution of line
sources and sinks and, as was sbown in Section 11.6, the term
l
2 n log R = }n ,J(x - ';)2 + (y - 17)2
is tbe potential at a point P(x, JI) due to a unit fine souree (that is, a line Source of
unit strengtb per unit length normal to the plane of tbe fiow) at a point Q(';, 17).
The Green's function (106) accordingly corresponds to tbe introduction al so of a
unit fine sink at the external point Q'(';, -17) which is the "image" of Q relative
to tbe boundary 17 = O.
If cp is steady-state temperature, then the term -(2n)-1 log R, witb a prefixed
negative sign, is the temperature at P(x, y) due to a unit line heat source at
Q(';,17) wben the medium has unit conductivity. Analogous interpretations are
possible, for example, in terms of gravitational potenrial (without a negative sign)
and electrostatic potential (with a negative sign).
It sbould be noted tbat the properties exhibited for the Green's function in
this section bave been with reference to the specific operator L = V 2 , subject to
the specification of the solution of the associated problel)l. along the boundary of
ffi, and tbey do not necessarily persist in other cases. (See Section 11.11.)
11.10. The Use of Conformal Mapping. When the region ffi is simply con-
nected, we show next that tbe determination of the Green's function of Section
11.9 can be reduced to the problem of determining an analytic function wbicb
specifies a mapping of tbe region ffi onto the upper half of tbe w plane.
For this purpose, we begin by noticing that the relevant Green's function
for the upper half of the xy plane, as obtained in the preceding Example 1,
can be written in the formo
G(x, y;';, 17) =

1
2 n log 1: =Él = 2~ Re [IOg (; =?)]
where z = x + iy, , = .; + i17.
Accordingly, we may verify that ifw = F(z) is an analyticfimction which maps the
region ffi in the z plane onto the upper half of the w plane, with F '(z) O in ffi so *
that the mapping is one-to-one, then the Green's fimction which satisfies (102),
(101), and (94) is given by
1
G(x, y; .;, 17) = 2 n log I~~:~ ~~g I =
= _1 Re {IOg [F(z) - F(t;)]} (110)
2n F(z) - F«()
where again
z = x + iy, t; = .; + i17· (111)
11.10. The Use of Conforma) Mapping 659
To eonfirm this assertion, we first observe that, sinee log I rx I = ±log (rxli),
Equation (110) also can be written in the form
G . - _1 lo {[F(Z) - F(OHF(z) - F(O]} ( 112)
(x, y, .;, r¡) - 4n: g [F(z) - F(OHF(z) - Feo] ,
from whieh it is evident that the right-hand member is symmetric in z and "
and henee in (x, y) and (.;, r¡). Seeond, sinee the real part of an analytie funetion
of z satisfies Laplaee's equation exeept at singular points of that funetion, and
sinee F(z) = F(O only when z = " and also F(z) 7"= F(O, when z is inside <R,t it
follows that the right-hand member of (110) satisfies Laplaee's equation when
z is in <R and z 7"= ,. Third, sinee F(O is real when , is on e, it folIows that the
right-hand member of (110) vanishes when , is on e and henee also, from the
symmetry, when z is on C. Finally, sinee we have
F(z) - F(O = (z - OF'(O + (z ;-! 0 2

FU(O + ... ,
when z and , are inside <R, where F(O 7"= 0, we may deduce that the right-hand
member of (110) behaves like (2n:)-1 log I z -1;1 = (2n:)-1 log R when z - , - + 0,
where R is given by (100). Thus the right-hand member of (110) satisfies (102),
(101), and (94), and henee it is indeed identified with G,:j:
We see that the Sehwarz-Christoffel transformation (Seetion 11.8) is avail-
able for the determination of a mapping funetion F when e is polygonal. An
additional useful faet is that the funetion
w= F(z) - F(O , (113)

F(z) - F(O
whieh appears in (110), is an analytie funetion whieh maps <R eonformalIy onto
the unit circular disk I w I -< 1, in sueh a way that the point z = t;, maps onto the
origin w = O. (See Problem 69.)
Example 1. If <R is the circular disk of unit radius, X2 + y2 -< 1, the result of Prob-
lem 23(e) gives F(z) = i(l - z)/O + z), in aeeordanee with whieh (110) determines
the Green's funetion in the form
Il - ' z I
1 log z - , ,
G = -2
n:
(114)
with z = x + iy and , = .; + ir¡. If polar coordina tes are introdueed, sueh tha!
z = re/·, , = ye'P, (115)
a simple eomputation yields the expanded forrn
1 [r 2 -2ryeOS(e-[3)+y2]
G(r, e; y, [3) = 4n: log 1 _ 2ry eos (e - [3) + r2y2 . (116)
tNotice that the imaginary part of Fez) is positive and the imaginary part of F(O is negative
or zero.
:j:The consequence that two different identifications of the mapping function Fez) must produce
the same G is independently verified in Problem 68.
In addition, we find that on the boundary y = 1 there follows
~~ [~f]'~1
1 - r2
= 21t 1 - 2r cos (O - P) + r2 (I17)
Thus the solution of the problem

V2rp = h(r, e),
(118)
rp(1, e) =f(O)
in the unit disk Ir I <: I is given by
- 1 r2
f" 1-2rcOs(0-p)+r
)f(p) (1 - f2" fl G(r, e,.y,P)h(y,P) dy dp.
rp(r,0)-21t o 2dP + o o
(119)
When h = O, this result reduces to the Poisson integral formula of Section 9.5, with
a = 1. •
Example 2. 1f al. is the infinite strip -ca < x < ca, O <: y <: b, the result of Problem
23(a) combines with (110) to yield the expression
G -- 1 I
e", ! b
2n lo g e1f. zl b
-
-
e"e/ b
ent:Jb
I (120)
or, after sorne manipulation,
.
G(x, y, e, r¡) -
_ 1
411: log
[COSh ~ (e; -
11:
x) - cos ~
11:'
(r¡ - y)] (121)
cosh b(e; - x) - cos b (r¡ + y)
Thus, for example, the solution of the problem
V2rp = h(x, y) ,
(122)
rp(x, O) = rp(x, b) = O,
for - ca < x < ca , o <: y <: b is given by
rp(x, y) = J: roo G(x, y; e, r¡)h(e;, r¡) de; dr¡, (123)
with G defined by (121) . To solve the problem

V2rp = O,
(124)
rp(x, O) = f(x), rp(x, b) = O
in that strip, we first make the computation
. 1ty
a~ 1 SIn b
[- (125)
ar¡ J.=o = 2b cosb ~(e
b
_ x) _ cos 1ty
b
and deduce that
rp(x, y) = l...
2b
sin 1ty
b
f .o 1t
f(e)
1ty
de
,
(126)
- ~ cosh b(e - x) - cos b
in accordance with the result of Problem 82 of Chapter 9.
•
11.11. Other Two-Dimensional Green's Functions 661
11.11. Other Two-Dimensional Green's Functions. For a more general two-

dimensional linear partial differential operator L and an associated region <R,
one method of defining a Green ' s function G starts with the derivation of a
related "Green's theorem" of the form
f L. G Lrp di; d11 = t (... ) ds +f L rp L *G di; d11, (127)
where L * is a new differential operator, said to be the adjoint operator relative to

L, and where ( ... ) denotes an expression to be integrated along the boundary
C. When L* = L, as has been seen in (90) to be the case when L = V', the
operator L is said to be seif-adjoint. In the special case L = V2, reference to (90)
shows that the expression ( . .. ) is of the form G(arp!an) - rp(aG/ an).
The relevant Green's function, ifil exisls, then is defined to be that function
G = G(x, y; e,
11) which satisfies the differential equation
L*G = 0('; - x)O(11 - y) in <R, (128)
together with certain specified homogeneous boundary conditions on C. In par-
ticular, if rp is to satisfy the equation
üp = h(x,y) in <R (129)
and if, along each part of e, rp and/or arp/an is prescribed, then along each part
of e one requires that the corresponding G be such that the only nonvanishing
terms in ( ... ) are those which involve prescribed information relative to rp
and/or arp/an. With G so defined, Equation (127) takes the form
rp(x, y) = -fe(...)ds + ff'" G(x, y; e, 11)h(e, 11) di; d11, (130)
where ( ... ) now involves only known functions.

If L is self-adjoinl, and if also it happens that the expression ( ... ) in (130)
vanishes everywhere on e whenever rp(e, 11) satisfies the same homogeneous
conditions on e as does G, it is found that G is symmelric in (x, y) and (e, 11),
so that G(x , y ; '; , 11) = G(i;, 11; x, y). (See Problem 85.) In thi s case the operator
L * in (128) is identified with L and, because of the symmetry, the differentiations
in LG may be either with respect"to the coordinates of Q(e, 11), as was assumed
in the derivation, or with respect to the coordinates of P(x, y).t Except in the
special cases when syrnmetry of G can be anticipated, the operator L* in (128)
mUSI differentiate with respect to the coordina tes of Q if (i 30) is to be valido
In illustration, the symmetry of G in (95) could be deduced from the facts
that, first, L * = L (= V2) and, second, G is to vanish on e and the boundary
integrand in (95) will vanish if f (the boundary value of rp) also vanishes. Ex-
tIn sorne referenees, the differential operator L is said to be "formally self-adjoint" when
L' = L. When also the homogeneous conditions which must be satisfied by an associated
Green's funetion G on the boundary of the relevant region cR are such that G is in faet sym-
metrie, the term " self-adjoint" (without a modifier) then is used to describe the "domain
operator" specified by L and by those homogeneous conditions.
662 Applications oC Analytic FW1CtioD Theory
ample 4 of this section will illustrate a situation in which L * = L but G is nOI

symmetric.
The transformation of the left-hand member of (127) into the right-hand
member can be effected by use of the two-dimensional analogies to integration
by parts, yielded by the two-dimen sional specialization of the divergence the_
orem, in the forms
or
JJ'" u ~~ de; dr¡ = fe uv cos ex ds - JJ", V ~~ dI; dr¡ (l31a)
and, similarly,
JJ", U ~~ dI; dr¡ = fe uv cos f3 ds - JJ", V ~~ dI; dy, (l3Ib)
where n = i cos ex +j cos f3 (132)

is the unit outward normal vector at a point of C. One of the variables may
represent time, in which case the region <R is a space-time domain.
Example 1. For the modified Helmholtz operator

L = V2 - k 2 ,
with k real and positive, there follows
JL. G(V2cp - k2cp) dI; dr¡ = f) G ~~ - cp ~~) ds + JJ", cp(V2G - k 2 G) dI; dr¡.
Hence, for the self-adjoint operator L = V2 - k 2 , an associated Green's function

must be such that
V2G - k 2 G = O in <R when (x, y) * (e;, r¡) (133)
and the requirement
1 = l~ JJrR (V2G - 2
k G) de; dr¡ = li.?;: (f" R ~~ de - k2 JJrR G de; dr¡),
where rR is a circular disk of radius R, with center at P(x, y), leads to the condition
1
G ~ -logR (R - O) (134)
2:n
as before, the term Jf rR G de; dr¡ then tending to zero with iR2 lag R as R ---> o.
In the special case when <R is the entire plane, it is convenient to write
e; - x = R cos Ij/, r¡ - y = R sin Ij/
and to introduce R and Ij/ as new (polar) coordinates . Since G clearly will not depend
upon Ij/, Equation (133) becomes
2
d G
dR 2
+ _1 dG
R dR
_ k2G = O (R * O)
11.11. Other Two-Dimensional Green 's Functions 663
and ¡here follows

G = eJoCkR) + e2KO(kR) (R "* O), (135)
where lo and Ko are modified Bessel func¡ions (Section 4 .8). Since l o(x ) - > 1 and
KoCx) ~ -log x as x O, the condition (134) requires that e 2 = -1 /(2n). In addi-
tion, siDce IoCx) ~ ex/.../2nx and Ko(x) ~ e- x /.../2x/n as x - > ca , we must take el = O
in order that G tend to zero as R - ca. HeDce there follows
(136)
Thus, for example, tbe solution of the problem

V2rp - k 2 rp = h(x, y),
(137)
lim rp = 0,
for -ca < x < ca , -ca < y< ca, is given by
rp(x, y) = - 1
2n f~
_ ~ f~
_~ Ko(k.../(x - ';)2 + (y - '1)2) he';, '1) d,; dr¡. (138)
Clearly, the corresponding Green's function in the upper half-plane is
(139a)
when rp(x , O) is to be prescribed, and is
G = - 2~[Ko(k.../ (x - ';)2 + (y - r¡ F ) + Ko(k..! (x - ';)2 + (y + '1)2)], (139b)
when rp,(x, O) is to be prescribed.

11 is important to notice that conformal mapping, as used in Section 11.10 when
k = 0 , cannot be used when k "*
0, to determine the Green's function in a region <R
by mapping that region onto (say) a half-plane . Tbis negative result follows from the
fact that if u + iv = F(x + iy), where F is an analytic function of z = x + iy, then
[See Equation (41) of Chapter 11.) Thus, whereas rp xx + rpyy = implies rpuu + ({Jvv = O °
when I F(z) I "* O, the equation ({J xx + rpyy + Árp = O does not imply the equation
rpuu + rpvv + Árp = O unless Á = O; that is, the Helmholtz (or modified Helmholtz)
operator is not invaríant under a conformal mapping unless k = O.
In addition, we see that, whereas tbe Green's fUDction (136) properly tends to
zero as x2 + y2 - > ca , for any positive constant k, the function (2n)-1 log R + e
(e = constaD!) which results when k = °
(see Problem 86) does not have this property.
Correspondíogly, it is found that lhe problem
V2rp = h(x, y),
(137')
lim rp = O,
x '+)"- oo
for - ca < x < ca , - ca < y < ca , generally does not possess a solution. When h is
664 Applications of Analylic Funclion Theory
respectable, the function
(138 ')
will satisfy V2q¡ = h, but it also wiU tend to zero as X2 + y2 ~ 00 only if also h is
such that
r~ r~ h(~, 1/) d~ d1/ = O, (140)
in wrueh case the addition of a constant c to (2n)-1 log R will not modify the definition
(138 '). •
Examp/e 2. For the Helmholtz operator

L = V2 + k 2,
with k real and positive, we find that (135) is replaeed by
(R": O). (141)
Sinee Jo(x) ~1 and YoC-c) ~ (2jn) log x as x - > O, we must have C, = t. But since
Jo(x) ~ ./2/(nx) cos (x - nj4) and Yo(x) ~ ./2j(nx) sin (x - n/4) as x - > 00, the
requirement that G (and/or aG/aR) tend to zero as R - > 00 pennits C l to be arbitrary!
FrequentIy the Helmholtz operator appears in a formulation as a result of requiring
that the equation
a-
V2 U - - 1 - u = h(x y)e'w'
2
(142)
c'
2 at '
admit a solution of the form

U(x, y, t) = q¡(x, y)e'w',
in whieh case q¡ must satisfy the equation
V2rp + k'rp = h(x, y), (143)
with k = ro/ c. If the relevant regíon <R eomprises the entire plane, the usual require-
ment imposed by the underlying problem is the specification that U satisfy the so-caUed
radia/ion condition, that is, that it involve only "outward-traveling waves" as R - > 00
(see Section 9.13). The corresponding requirement that Ge'w' satisfy trus condition
demands that the right-hand member of (141) be a multiple of the Haokel funetion
H&')(kR) = Jo(kR) - iYo(kR)
(Seetion 9.12), and hence that C2 = -icl in (141) . Thus we must have C, = ¡e, = i/4
and hence
([44)
so that the solution of (143) for which U = q¡e'w, satisfies the radiation condition as
X2 + y2 -----+ CX) is
(145)
•
11.11. Other Two-Dimensionnl Green's Functions 665
Example 3. For the Laplace operator

L = V2,
when the associated boundary condition specifies the normal derivative of the solution
on the boundary e of a region <R, so that the related problems are Neumann problems,
it would appear from the related Green's theorem (90) that the relevant Green's func-
tion is the solution of (93) for which aGjan = O on C. However, we notice that, if trus
were assumed to be the case, then the relation
(146)
would provide a contradiction since the left-hand member would be 1 and the right-
band member would be O. Thus G in fact does not exist in this case. This situation is a
consequence of two related facts. First, if rp is to satisfy
V2rp = h in R,
(147)
~: =g on C,
then the relation
ff al V2rp dx dy = fe ~: ds
requires that
ff al h dx dy = fe g ds. (148)
Unless g and h satisfy tbis compatibility condition, the problem (147) has no solution.
In addition, as has been noted, if (147) does have a solution, then to that solution can
be added any constant, so that the solution is not unique.
When (148) is satisfied (and g and h are respectable), the problem (147) is known
to have a solution which is unique apart from this additive constant. (See Section 9.2
relative to the case h = O.) In addition, a modified Green's jimction can be defined in
either of two ways, to serve in place of the nonexistent G.
Neumann'sfunction, denoted here by N(x, y; e;, '1), is required to satisfy the equa-
tion
VbN = O(e; - x)O(r¡ - y) in <R (149)
and the boundary condition
aN = a(s) on C, (150)
anQ
where o:(s) is any function for which the condition replacing (146),
ff al Vb N de; dr¡ = fe ;~ ds,
is satisfied, and hence for wruch
fe a(s) ds = 1. (151)
HeTe, since (for the fust time) symmetry in (x, y) and (e;, '1) is not necessarily present,
ambiguity has been avoided by use of the notations Vb and alana to specify that the
relevant differentiations are with respect to .; and 1], with x and y he Id fixed.
In terms of the function N, the associated Green's theorem
then provides the formula
rp(x, y) = A - fe N(x, y; ';(s), '1(s))g(s ) ds + f J!R N(x , y;';, '1)h(';, '1) d'; d'1, (152)
where A =fc IXrp ds is the anticipated arbitrary constan!.

Ir is apparent that to any spec-ific N can be added any function of x and y Only,
since the satisfaction of (149), ( 150), and (90') ",ill not be affected. Moreover, by virtue
of (148), this modification of N will not modify the right-hand member of (152). The
additive function of x and y can be so chosen that N satisfies the suppIementary condi-
tion
fe lX(s)N(x, y; ';(s), '1(s)) ds = constant, (153)
in which case Problem 88 shows that N becomes symmetrie in (x, y) and (.;, '1). Problem
89 deals with the Neurnann function when <R is a unit disk and hence also provides the
basis for the determination of N in correspondence with other (simple) regions by use
of conformaI mapping.
Hilbert's ¡unetion, denoted he re by H(x, y; .;, '1), is required to satisfy the equa-
lion
VbH = 0('; - x)O('1 - y) - P(';, '1) in <R (154)
and the boundary condition
on C. (155)
where p(.;, '1) is any function for which the condition replacing (146) is satisfied, so
that
f f!R VbH d'; d'1 = fe ~~ ds,

and hence for which
(156)
In terms of the function H, the relation (90), with G replaced by H, takes the form
rp(x, y) = B - t H(x, y; ';(s), '1(s)) g(s) ds +f t H(x, y; .;, '1)h(';, '1) d'; d'1, (157)
where B = j j!R prp d'; d'1 now is the arbitrary constan!. An arbitrary additive function
of x and y, which clearly is present in H, but which does not affect (157) when (148) is
satisfied, can be so determined that
f f!R p(.;, '1)H(x, y; e;, '1) de; d'1 = constant, (158)

11.11. Other Two-Dimensional Green's Funetions 667
in which case H becomes symmetric in (x, y) and (.;, r¡). (See Problem 91.) The Hilbert
functioo is treated in Problem 92 when CR is the unit disk and in Problem 94 in other
cases. 11
Example 4. For the operator

L = e2 a2
--
a z
axz - atZ'
associated with the wave equation, there follows
-r [Gq¡, - G,'P[o d'; + s: s: q¡(ezGt;t; - Gn ) d'; d., (159)
when CR is the space-time domain a -< x -< b, O -< t -< T, where T is any positive
constant. Thus L * = L and the Green's function must satisfy the equation
eZ aZG _ aZG = 0('; _ x)O(. - t) in CR.
a.;z a. z (160)
We suppose here that (a, b) = (-co, co) and that the associated problem is that
of determining q¡(x, t) such that
aZq¡ aZq¡_
e Z ax2 - at 2 - h(x, t), (161)
for -co < x < co, O -< t -< T, subject to the initial conditions
q¡(x, O) = [(x), 'P,(x, O) = g(x), (162)
Accordingly, in order to suppress the boundary tenns in (159) which involve unspe-
cified information, we require that G satisfy the conditions
lim G = lim G. = O (O -< .. -< T) (163)
{_±oo ':_±oo
and
G=G,=O when.=T (-co<x<co). (164)
In terms of lhis function, the required solution 'P is given by (159) in the forro
q¡(x, t) = L~~ [Glx, t;';, 0)[(';) - G(x, t;';, O)g(';)] d';
+ s: r~ G(x, t;';, .)h(';,.) d'; d.. (165)
For the purpose of determining G,t it is convenient to fust take the Fourier trans-
form, with respect to';, of the equal members of (160). If we write G for the transfonn
of G, so that
G(x, t; u,.) = S'::' e-i"<G(x, t;';,.) d';, (166)
tHere, in spite of the fae! that L* = L, the Green's funetion will not be symmetrie in (x, t) and
(.;, -r). lt should be notieed that the eondition required for symrnetry in Problem 85 is not
satisfied here, sinee G and its -r derivative are to vanish when -r = T, whereas the boundary
integrand in (165) will vanish identieally only if f and g vanish, so that 9>(';, 't") and ils -r deriva-
tive vanish when 't" = O.
then the tra nsforms of a'G jaf,°, alG j aT ' , and 0('; - x) are -ulG, a 2G jaT2 , and
r ' u X , respectively, and from (160) and (164) there follows
aG+
--
2
c'u'G
- = -e-luxO(T - t),
aT2 (67)
- aG
where G = -;n=O when T = T, (168)
The solution of (167) satisfying (168) ¡s obtainable by the elementary method of varia-
tion of parameters [see Section 1,9 or Problem 110 of Chapter 9] in the form
G= _e- iux S: sin CU~Tu - T,) O(T[ - t) dT,
= e- Iu.• [sin CU(T - T ,) O(T _ t) dT

cu 1 1
r
e- ,UX sin Cl/(T - t)

(t> T),
= cu (169)
{
O (t < T).
Here the lower ¡imit T in the first right-hand member is introduced to make G = O
when T = T and the second right-hand member is written in such a way that the
variable of integration is increasing, so that the contribution of the delta function be-
comes evident. When the transform ( 169) is inverted, there follows
1 J~ e -iu (x -~) sin cu(t - T)

G = -- du
2n cu
- ~
= __1_ J~ cos u(x - .;) sin cu(t - T) du

2nc _ ~ u
when T < t only. This integral can be evalua led by use of residue calculus [or by refer-
ence to Equation (247) of Chapter 5] to give the result
-2c when Ix - el < c(t - T),

G(x,t;';,T) = 1 (170)
- 4c when Ix - .; I = c(t - T),
O otherwise.
We see Ihat G is indevenctenl of T, and hence can be considered lo be associated with
the domain - 00 < x < 00 , O -< t < oo.
Thus G = - l j (2c) inside the sector
x - c(t - T) < .; < x + c(t - T)
of the';T plane and G = O outside that sector. With the Heaviside unit step function
ootation
{~
(u > O),
H(u) = (171)
(u < O),
11.11 . Otber Two-Dimensional Green's Functions 669
Equation (170) al so can be written in the forrnt
G =
- _1 [H(x -
2c
.; + c(t - ,» - H(x -.; - c(t - ,»] (, < t),
(I72)
{
O (, > t),
and hence there follows further (see Problern 32 of Chapter 2)
a,
aG
=
1
Z[O(x - .; + c(t - ,» + O(x - .; - e(t - ,»] (173)
when, < t . Accordingly, Equation (165) takes the forro
rp(x, t) = Z[f(x
1
+ et) + f(x - et)] +
1
2e
r+«
JX-CI g(';) d';
- 1
2c l 'f
o
X
+
x-c«(-t')
d l
-
d
h(';, ,) d'; d,. (174)
This relation is equivalent to the result of Problem 37 of Chapter 8, and it reduces

when h = O to D'Alembert's formula [Equation (67) of Chapter 8]. •
Example 5. For the operator

L = 1X2 -
a
--,
2 a
ax2 at
associated with the heat-fiow (or dijJusion) equation, there follows
SoTfb G(1X2rp{{
a
- rp,) d'; d, = 1X2 ST [Grp{
O
- G{rpJb
~-a
d,
(175)
when <R is the space-time dorna in a -< x -< b, O -< t -< T. In particular, we see that
L* = 1X2 -
a +-,
a 2
ax2 at
in terms of x- and t-differentiation, so that L bere is not self-adjoint. An associated
Green's function correspondingJy must satisfy the equation
in <R, (176~
together with appropriate homogeneous conditions along tbe space and time bounda-
ries of <R.
In illustration, in correspondence with the problern
1X2 a'rp = arp + h(x t)

ax2 at " (177)
rp(x, O) = f(x),
for -= < x < 00, O -< t -< T, we require that G satisfy the conditions
Jim G = lirn GI; = O (O -< , -< T) (178)
~_±oo {_±:_
tlt is easily verified that, jf [(u) = 1 when I u I < ex and f(u) = O when I u I > ex, then there
follows [(u) = H(" + ex) - H(u - ex).
670 ApplicatioDS oC Aoalytic Function Theory
and
G = O when't" = T (- ce < C; < 00 ), (179)
so that (175) gives
rp(x , t) = - f_ G(x, t; C;, O)[(e,) de, + s: r~ G(x , t; e" 't")h(!" 't") de, d't". (180)
In order to determine G in this case, we again introduce the Fourier transform

with respect to !" and find that G must be such that
aG (X 2 u 2 G
- = e-'· ' O(, - t)
a,
- -
' (181)
G=O when, = T,
so that
G = r'·· S: e- ·.'<"-' )0(,
4
I - t) d, I
_ e -iU. x J·T e-(l2u. 2

t' { q - ';" ) c:5(r 1 - t) dr I
(, < t),
(182)
Hence there follows
-
- _1_
2/t
f~ e-· '·'('-') cos u(!, - x) du
- ~
(183)
when, < t onJy, by virtue of Equation (241) of Chapter 9. The result of introducing
(183) into (180) reduces to the result of Problem 85 of Chapter 9 when h = O. A s in the
preceding example, Gis indepeodent of T, which accordingly can be taken to be + ec .
It is easily seen that the corresponding Green's function for the regio n O -< x < ec ,
° -< 1 < 00 is given by
(184)
when 't" < 1 only , with the upper sigo applying if rp(O, t) is prescribed and the lower
sigo if rp ,(O, 1) is known . The appropriate specialization of (175) then solves the relevant
modification of the problem (177). [See also Equation (242) of Chapter 9, as well as
Problems 86 and 87 of that chapter, when h = O.] •
REFERENCES
l. References at end of Chapters 8 and 10.

2. GREENBERG, M . D., Application o[ Green's Funclions in Science and Engineering,
Prentice-Hall, Ioc., Englewood Cliffs, N.J., 1971.
Problems 671
3. KOBER, H., Dielionary ofConformal Represenlalions, Dover Publieations, lne., New

York, 1957.
4. LE PAGE, W. R., Comp/ex Variables and Ihe Lap/aee Transform for Engineers,
MeGraw-Hill Book Company, Ine., New York, 196J.
5. MACKIE, A. G., Boundary Va/ue Prob/ems , Hafner Publishing Company, Ine., New
York, 1965.
6. McLACHLAN, N . W ., Modern Opera/iona/ Ca/eulus, Dover Publieations, Ine., New
York, 1962.
7. MIKHUN, S. G., In/egra/ Eqlla/ions and Appliealions, Pergamon Press, Ine., Elms-
ford, N.Y., 1957.
8. NEHARI, Z., Conforma/ Mapping, MeGraw-HiII Book Company, lne., New York.
1952.
9. ROTHE, R., F. OLLENDORF, and K. POHLHAUSEN, Theory of Func/ions as Applied 10
Engineering Problems, Dover Publieations, lne., New York, 1961.
10. SOMMERFELD, A., Par/ia/ Differen/ial Equalions in Physies, Aeademie Press, Ine.,
New York, 1949.
11. WALKER, M., Sehwarz-ChrislOffel Transforma/ion and l/s Appliea/ions: A Simple
Exposi/ion (original title: Conjugale Funelions for Engineers), Dover Publieations,
Ine., New York, 1964.
PROBLEMS
Section 11.2
1. Use Equation (8) to determine the inverse Laplaee transform of eaeh of the fol-
lowing funetions, and compare with the appropriate formula given in Table 1, pages
67 and 68:
1 1 1
(a) s + a' (b) (e)
(d) ;4' (e)

s
(f) S4 + 4a4
2. Use Equation (8) to determine the inverse Laplaee transform of eaeh of the follow-
ing funetions (with n a positive integer):
1 s 1
(a) (s + aY' (b) (s + aY' (e) (s + b)2 + a2'
s 1 1
(d) (s + b)2 + a2 ' (e) [(s + b)2 + a 2 )2' (f) (s Z - a 2)2
3. Establish the following result : Jf F(z) is ana/y/ie exeepl for a finile number o/ po/es,
and if e'" F(z) lends uniformly 10 zerO on Ihe semicire/e C R : I z I = R in Ihe firsl and fourlh
quadranls as R --+ 00, for sorne posilive va/ue of lo, Ihen
.,e-' ( F(s )) = O when 1 < lo.
[Suppose that 1 < lo, close the relevant Bromwieh line from e - Ri to e + Ri to the
righl by an are of C R ·, with R ' = ,,¡ R2 + e 2, write
e" F(z) = e- u, -"' [e'" F(z)],
and use Theorem HA of Seetion 10.1 4.]
672 Applications oC Analytic Function Tbeory
4. Use residue calculus 10 derive the formula

< a). (1
(1 > a),
when a> O. (Use the result of Problem 3 when f < a.)
5. Use residue calculus to obtain the result
j3-1 _1 1 - e- ,'} = _
n 4 ¿:; -smkl
l.
{ s 1 + e ns 7r k odd k '
where k takes on the values 1, 3, 5, ... in the infinite series, assuming tha! !he fonnula
(8) is applicable. [The series obtained is the Fourier expansion of a "square-wave
function" f(t) such that f(l) = I when 0< f < n, f(t) = -1 when n < 1< 2n, and
f(t + 2n) = f(t) (see Problem 10 of Chapter 2). Rere, in addilion to justifying !he
presence of an infinite series, one mus! take into account !he fac! !ha! F(z) has its infinite
set of poles z = ki (k = ± 1, ±3, ... ) on the imaginary axis, so that I F(z) I cenainly
is not even bounded on any semicircle e R for which R is an odd integer. This difficulty
can be avoided, in the present case, by leHing R ~ co through an appropriately chosen
discre!e sequenee of values R" R 2 , • • • • ]
6. Use residue calculus to obtain the result
j3-1{~ sinh (sx/e)) = ~ -L 2 ~ (-1)" sin knel sin knx,
s smh (s//e) J / ' n k~ 1 k , ,
assuming that (8) is applicable. (See Problems 102 and 106 of Chapter 9.)
7. Use residue calculus 10 obtain the result
j3 _, (_1 sinh (s ¡(2 x / a )} _ ~ 2 ~ (-I)k o-k'n'a',/l' s' knx
1 s sinh (sl/2//a) - , n + k~' k ~ m -,-'
assuming tha! (8) is applicable. [See Problem 60 ofChapter 10. Also see Equation (168)
and Problems 101 and 108 of Chapter 9.]
Section 11.3
8. Determine !he inverse transfonn of
(P> O)
in the fonn
f(l) = 1 - ~ l~ e-" sin PV; dr
n o
r
and, by referring to Equation (247) of Section 9.14, deduce that
j3 1 _{e- p
,,,,} P ).
= 1 - erf ( --_
s 2VI
9. Obtain the inverse transforms
(a) j3-1
{es-P""}
.
1!2 nv2 I
= ___ l-o
e-u' cos P"_ du
VI
= ~e-P'/4,
1
vm '
(b) j3-1(e-P""} = 21-
m o
e-u' sin P~l/ du
VI
= P
2V m J
e- P'/4"
where P > O. [Use Equation (241) of Section 9.14 to evaluate the integral in part (a)
and differentiate that equation with respect to b to deal with part (b).]
Problems 673
10. Solve. by l/se 01 Laplace tral1slorms, the heat-flow problem considered In Section
9.14, in which T(x, 1) satisfies the differential equatioo
a2 T 1 aT
ax2 = (Xz al '
the initial condition T(x, O) = T o = constant, and the end condition T(O, t) = O for
O <: x < =, O <: t < co. [StlOw that the transform T then satisfies the equation
aZT _ ~T = _To
ax 2 (X2 (X2
and hence must be of the form
Determine A and B so that the end condition is satisfied, and so that t is bounded as
x---> co, and show that
Thus, using the result of Problem 8, obtain the result
T = T o erf (2(X~ 1 ) ,
in accordance with the result of Sectioo 9.14.]
11. Use (11) to show that
oC- 1 { log
52 +
S}
1
and deduce from the result of Problem 98, Chapter 10, that this result can be writteo
L'
O x
-¡
Figure 11.24
in the form
oC-I{s~0!S1} = eos I Si(l) - sin t 0(1).
(Notiee also tha! in this case the infinite Bromwieh line would be equivalent to the
modified loop con tour L' indieated in Figure 11.24, in whieh the eontributions of the
segments along the imaginary axis would cancel in pairs, leaving the eontributions of
the integral on L and of the two poles.)
12. Use (11) to show that
1
oC -1 { S1/2(S _ a)
} ea<
=..j a -
1
Ti Jo
r= r !l2(r
e-U
+ a) dr (a> O).
13. Express the result of Problem 12 io terms of tabulated funetions as follows:

(a) Show that
roo e-teu2+al
where g(l) = 2 Jo v2 +a dv.
(Write r = v 2 .)
(b) By differentiating under the integral sign (see Seetion 7.9), show that
g'(I) = -2e- a, 1= e-'V' dv = -,¡~ e-ato
(e) Noticing that g(co) = 0, obtain the result
=
n
7a(l - erf val),
/-
where erf X is the error funetioo.

(d) Henee deduce from Problem 12 that
1 } ea, -
oC -1 { S 1/2(S _ a) = ..jli erf ..jal (a> O).
14. Generalize the results of Problems 12 and 13 as follows:

(a) Use Equation (11) to show that, when °
< m < 1 and a > 0, there follows
oC- 1 { 1 } _ ea, _ sin mn ea, (t)
sm(s - a) - am n g,
roo e-1(r+ a )
where g(l) = Jo rm(r + a) dr.
(b) Show that
g'(I) = - r(1 - m) e-a'l m - 1
and henee that

g(l) = r(1 - m) s.= e- '-r
a m- 1 d-r
=
ro - m
m)
[r(m) - y(m, al)],
a
Problems 675
where (m> O)
is the so-ealled ineomple/e Gamma ¡une/ion (and is rather well tabulated).

(e) Use Equation (59) of Chapter 2 to express the eonsequenee of the preceding
results in the form
.c-¡{ I
sm(s - a)
} = ea, y(m, al)
am r(m)
(O < m < 1, a > O).
(d) Verify that

y(1, x) =..;-n erf";x
and aeeordingly that the result of part (e) properly reduces 10 the result of Problem 13
when m = ~. [Use Equation (57) of Chapter 2.]
15. Let
F(s) = log s + 1
s
and eoosider the integral 1c e"F(z) dz arouod the c10sed eontour of Figure 11.25(a),
where F(z) is defined to be that branch of log [(z + 1)/z] whieh is analytie exeept along
a cut joining tbe braneh points at z = O and z = -1, and whieh is real on the positive
real axis.
y y
e + iR
x x
-1 o
(a) (b)
Figure 11.25
(a) Show that tben

F(z) = log pz
p¡
+ i(Wz - W ¡),
where z - a = PIe/a> , and z + a = pze'a>·, with -7r. < W¡ < 7r. and -7r. < W2 < 7r..
and that W z - w¡ = O on tbe re a) axis except wheo -1 < x < O, whereas
W z - w¡ = -7r. on the upper bank of the cut and Wz - w¡ = 7r. on the lower branch.
(b) By letting R - ex> and )etting the radii of tbe indentations tend 10 zero, sbow
that the "loop integral" reduces to an integratioo along the cJosed loop L of Figure
676 Applications oC Analytic FunctioD Tbeory
1 1.25(b) where, in fact, the integral s around the small circ1es tend to zero and there
follows
[Se e also Problem 13(b) of Chapter 2.]

16. Jf F(z) satisfies the conditions presumed in Section 11.2, except that it al so has
poJes at a finite number of points z = b j on the negative real axis, show that the
expression
-1 l: [Res (e" F(z); I b¡ I e- l' } + Res (e" F(z); lb ji el"}]
¡
must be added to the formula for f(t) and that the Cauchy principal value of the
integral generally must be taken in that formula. [Indent the upper and lower segments
of the loop L outward about each such pole (Figure 11.26) and use Theorem IV of
Section 10.14. An alternative procedure (which may or may not be preferable) c1early
consists of rotating the cut so that it avoids all poles.]
L
x
o
Figure 11.26
17. Use the method of Problem 16 to obtain the result
.,e-l{ sm(s 1+ l=e-ICOsm7t

a)J am
+ sin JYl7t pI~ e-N
7t
dr
rm(a - r)
o
when O < m < 1 and a > O.
18. When m = -1 in Problem 17, use the procedure suggested in Problem 13 to show
that
(a > O),
where
and hence also
g(l) = ..j7i Jo
r' :/7: d-r
QT
Problems 677
when use is made of Equation (22) of Chapter 2. Thus deduce that
Ll/2e: +
.e - I a)} = ~~a e-a, 1"";;; e"' du (a> O).
[The function (2/~7i) J~ e"' du is sometimes called the modified error function and
denoted by erfi x, and is tabulated.]
19. For the purpose of identifying the function
f(t) = .e- l {(S2 +la 2)1I2} (a > O),
show that we may consider the limit of the integral fe e" F(z) dz around the con tour of
Figure 11.27(a), where F(z) is that branch of the function 1/(z2 + a 2 )1/2 for which
y y
e + iR
r~ ia
L
ia
0L/U¿¿/L/.
W//7--7 --7/C x x
~0- -ia - ia
ú
0'"
«<~~ :2L/L::: e - iR
(a) (b)
Figure 11.27
with z - ia = Pie'"'' and z + ia = P2eiw" where -7t < COI < 7t and -7t < C02 < n,
a:1d where F(z) is analytic except on the finite cut joining the branch cuts. Also show
that (co ¡ + co 2 )/2 is O on the right bank of the cut and 7t on the left bank and that
~ P¡P2 = ~ a 2 - y2 on the cut. Hence, after proceeding to the limit, deduce that
f(t) = 2~i f L e"F(z) dz,
where Lis the c10sed loop of Figure 11.27(b), and where the integrals around the circ1es
tend to zero in tbe limit, so that
f(t) = 2ni
1 (fa ~ a~ _
-a
i<y
y2 i dy T
1-a ~ a
a
i",
2 e_ y2 e l .i dy
)
1
=-
n
Ja
-a ~ a 2
el"
- y2
dy
.
Finally, express this result in the form
f ( l ) -- -
n
7JI .¡
-
o
eos a/u 2 d U -- J o( al ) )
1-u
where Jo is a BesseJ funetion. [See ProbJem 65(e) of Chapter 5 .]
Section 11.4
20. Consider the mappmg w = az + b, where a and b are complex constants and
a,.: O.
(a) Show that the mapping is one-to-one everywhere.
(b) Verify that the mapping consists of a rOlalion of the z plane about the origin
through the angle arg a and a homogeneous magnification or s/re/ching by a factor
1a 1, followed by a Iramla/ion through the vector whose components are the real and
imaginary parts of b.
(e) Deduce, in particular, that circles map onto circles, and straight Iines onto
straight lines.
21. Consider the mapping w = l/z.
(a) Show that the mapping is one-to-one everywhere when the points z~ and w~
are included.
(b) With the notations w = pe l. and z = ré", show that p = l/r and ({J = -O,
and deduce that the mapping consists of an inversion with respeet to the unit circle
combined with a reflection about the real axis. [Notice that the poiot (r, O) maps onto
the point (l/r, - O).]
(c) Show that the general straight line in the z plane has the polar equation
r cos (e - eo) = p, where e o is the angle made by its normal with the positive x axis
and p is its distance from the origin, and deduce that its image in the w plane is the
curve pp = cos «({J + eo), where p and ({J are polar coordina tes in the w plane o
(d) Show that the general circle in the z plane, with center at the point with polar
coordinates (ro, ( 0 ) and radius R, has the polar equation
r2 - 2ror cos (e - ( 0) + (r5 - R2) = 0,
and deduce that its image in the w plane is the curve
(r5 - R2)p2 - 2roP cos «({J + eo) + 1 = O.
(e) Deduce from the results of parts Ce) and (d) that, under the mapping w = l /z,
eireles and straight lines in the z plane not passing through the origin map onto
eire les in the w plane, whereas eircles and straight lines in the z plane passing through
the origin map onto straight lines in the w plane o
22. Bilinear /ransforma/ions. Coosider the mappiog eorresponding to the general
bilinear (or linear frac/ional) transforma/ion
w = (1,z +P (A = (1,0 - py ,.: O),

yz +O
where (1" p, y , and ó are eomplex eonstaots.
(a) Show that the mapping is one-to-one everywhere (when z~ and w~ are in-
cluded) but that it would not be so if A = O. (Notice that the z plane maps onto a single
poiot if A = O.)
Problems 679
(b) Verify that the mapping can be eonsidered as lhe result of the three sueeessive
mappmgs
z' = y.:: + 15, w = - ,1, z"
y
+ !!:...
y
when y * O, and is of the form
when y = O.
(e) From the results of Problems 20 and 21, and the results of part (b), deduce
that any linear Iractionaltranslormation always maps circles anta circles, with the eon-
vention that straight lines are to be eonsidered as limiting forms of eircles.
23. l\1appings anta a half-plane. In eaeh of the following mappings, a speeified region
in the z plane is mapped onto the upper half (v >- O) of lhe w plane, with an indieated
eorrespondenee of eertaio bouodary poiots. Verify lhe eorreetness of eaeh mapping
and show also that dwjdz exists aod is nonzero everywhere ioside the mapped region.
(As is shown in Problem 55, in eaeh case the given mapping is only one of a triple
infinity of avaiJable ones, ehoseo beeause of the eonveoienee of the boundary-point
eorrespondenee.)
(a) Infinite strip:
(-co < x < co, O -< Y -< b).
Mapping: w = e 1tz / b .
Boundary poiols:
(+co, b) --> (-co, O), (O, b) --> (-1, O), (-co, y) --> (O, O),
(O, O) --> (1, O), (+co, O) --> (+co, O).
(b) Semi-infinite strip:
(O -< x -< a, O -< y < co).
11:z
Mappiog: w = -cos-'
a
Boundary points:
(O, +co) --> (-co, O), (O, O) --> (-1, O), Cta, O) --> (O, O),
(a, O) --> (1, O), (a, +co) --> (+co, O).
(e) Interior 01 circle:
(1 zl -< R) .
.R - z
Mapping: w=IR+z'
Bouodary poiots:
(-R, 0-) --> (-co, O), (O, -R) --> (-1, O), (R, O) --> (O, O),
(O, R) --> (1, O), (-R, 0+) --> (+co, O).
(d) Inferior 01 semicircle:
(\zl -< R, Y >- O).
Mapping: _~(~
2 R
+ B).
z
Boundary points:
(O +, O) - , (- ca, O), (R, O) - , (- J, O), (O, R) ~ (O, O),
(-R, O) ~ (1, O), (0-, O) ~ (+co, O).
(e) Half-plane indented by semicircle:
<1 z I >- R, Y >- O).
Mapping: w = ~(~
2 R
+ R).
z
Boundary points:
( - co, O) ~ ( - co, O), ( - R, O) ~ ( -1, O), (O, R) ~ (O, O),
(R, O) ~ (1, O), (+co, O) ~ (+=, O).
24. Afappings O] sectors. Verify eaeh of the following mappings, showing also that
dw/dz exists and is nonzero everywhere inside eaeh speeified mapped region.
(a) Circular sector anta unit semicircle:
(1 z I < R, O< e< a) ~ (1 IV I < 1, O < !p < n).
Mapping: w-
_ ( Z
R )n:< .
Boundary points:
(b) Circular sector anta half-plane:

(1 z I < R, O < e< a) ~ (v >- O).
1 '( z )n.'a , (R )n'"l

Mapping: _
w - -2.-L R -, z ~.
Boundary points:
(e) ¡nfinile sector onlo half-plane:

(O < e< a) ~ (v >- O).
Mapping: IV = zn.·'a..
Boundary points:
25. Mappings onlo a unil disk.

(a) Ha if-plane. By inverting the mapping of Problem 23(e) with R = 1, and inter-
ehanging z and w, deduce that [he mapping
z-z
w =-.--
1 Z
transforms the half-plane y .> O into the unit disk I w I < 1 with the following boundary-
point eorrespondenee:
(-co, O) - , (-1,0-), (-1, O) ~ (O, -1), (O, O) ~ (1, O),
(I,O)~(O,I), (+co,O)~(-I,O+),
Problems 681
(b) Other regions. If the analytic function G is such that the mapping IV = G(z)
maps a region CR onto the upper half of the IV plane, show that the mapping
i - G(z)
IV = i + G(z)
maps CR onto the unit disk 1 IV 1 -< l. (First map CR onto the upper half of a complex t
plane, then map that half-plane onto the unit disk in the IV plane.)
26. Examples of mappings onto a unit disk. Use the rnappings of Problerns 23(a) and
24(c) to illustrate the result of Problem 25 in the two following cases:
(a) Infinite strip:
(-ca <x < co,O-<y-<b).
- e 1tz / b
Mapping: w = ii + e =Jb
.1t
Boundary poin ts:

(+ca,b)~(-I,O-), (O,b)~(O,-l), (-ca,y)~(1,O),
(O, O) ~ (1, O), (+ca, O) ~ (-1, 0+).

(b) Infinite sector:
(O -< (] -< (1" O -< r < ca).
i - zn/a.
Mapping: W = i + z1[,o; l'
Boundary points:
=e'·~(-I,O-), e'·~(O,-I), (O,O)~(I,O),
(l,O)~(O,l), cae'°~(-l,O+).
[Other mappings onto the unit disk follow similarly, for example, from the mappings
of Problems 23(b, d, e) and 24(c).]
27. Detennine a mapping of the semi-injinite strip (O -< x -< a, O -< Y < =) in the z
rp
plane onto the infinite sector (O -< -< (1" O -< P < co) of the IV plane, by first mapping
the semi-infinite strip onto the upper half of a t plane [problern 23(b)] and then map-
ping that half-plane onto the sector [Problern 24(c), inverted), in the forrn
W =
nz)./' .
( -cosa
Also verify the correspondences (O, +co) ~ =e'·, (O, O) ~ e'., (aI2, O) ~ (O, O),
(a, O) ~ (1, O), (a, +co) ~ (+co, O).
Section 11.5
28. vVith the notation of Section 6.17, show that the scale factors associated with the
curvilinear coordinates defined by the relation u + iv = f(x + iy) are given by
hu = hv = 1dfldz 1- 1 . Hence, by using the results of that section, deduce that
ds 2 = 11z 1- 2
(duz + dv 2 )
and
Vrp = Idf
dz au u+ av
I(arp arp v),
where u and vare unit vectors in the u and v directions, and also that
'\12!p = I d/12
dz
(aau!p + aav!p),
2
2
2
2
where '\12 is the two-dimensional Laplacian opera tor, in accordance with the results
of Section 11.5.
29. B ernoulli 's equation [Equation (212) of Section 6.20] states that the pressure at a
poiot in an incompressible fluid is given by
P =Po -1PV2 ,
where P o is the equilibrium pressure, p is the density, and V = IVI is the absolute
velocity at the point.
(a) In a flow past a cylindrical surface whose cross section is bounded by a simple
c10sed curve e in the z plane, show that the normallorce per unit length of cylinder,
exerted by the fluid, has components
-
f e
dy
p-ds
ds
and
in the x and y directions, respectively, and show that these components are the real and
imaginary pans of the complex quantity i fcp dz.
(b) By making use of Bernoulli's equation, deduce the relation
-D + iL = - ~ ip tl~~12 dz,
where <l> is the complex potential function, D is the drag component of the force due
to pressure in the negative x direction, and L is the "jt component io the positive y
direction, Ihis terminology imply ing that the preponderant flow is in the positive x
direction .
30. (a) Show that the complex potential <l> = Aw (A real and positive) corresponds to
a steady flow, with velocity A in the positive u direction, in the w planeo
(b) Use the result of Problem 23(e) to map the upper half of the w plane onto the
upper half of the z plane, indented by a semicirc1e of radius a with center at the origino
(c) Show that the complex potential takes the form
<l> = .:i(~
2 a
+ ~).
z
By noticing tha t the real axis in the w plane corresponds to the portion of the x axis
outside the circ1e X 2 + y2 = a 2 , and to the upper half of the boundary of that circ1e,
deduce that the flow has this composite con tour as a strearnlioe.
(d) Determine A in such a way that the velocity tends to V o at large distaoces
from the origin, aod heoce obtain
<l> = V o( z + :2)
as the complex potential in the half-plane y > O of a flow, about a circular cylinder,
which tends to a uniform flow in the pos itive x direction at large distances from (he
origin (see Figure 11.9).
Problems 683
31. For the flow obtained in Problem 30, show that the veloeity eomponents are of the
form
a2(x2 - y2)]
Vx = Va [ 1 - (x2 y2)2 , +
and verify that, beeause of appropriate sy rnmetry in these expressions, the flow is eor-
reetly specified by <Il over the entire z planeo
Section 11.6
32. Suppose that two sourees of equal strength m are present at the points z = ±a,
where a is real and positive.
Ca) Show that the eomplex potential is given by
el> = ; . log (Z2 - a 2 ),
and the eomplex veloeity by

V
.x -
·vy =
1 nm Z2
Z
a2
(b) Show (by syrnmetry or otherwise) that the stream funetion is eonstant along
the y axis, and deduce that el> is also the eomplex potential in the half-plane x > O
eorresponding to a souree at z = a and a fixed boundary along the y axis.
(e) Show that the veloeity at a point on the boundary x = O is given by
V, = (m!n)[y/(y2 + a 2»), and at a point on the line y = O by V x = (m!n)[x!(x 2 - a 2)].
33. Suppose that a vortex of strength m is loeated at z = a, and that an equal and op-
posite vortex is at z = -a, where a is real and positive.
(a) Show that the eomplex potential is given by
el> m
= -.log - - ,
z-a
2nl z +a
and the eomplex veloei ty by
V 'V m a a2
x - 1 y = ni Z2 _
(b) Show that the vortex at z = -a may be replaeed by a fixed boundary along
the y axis in part (a), without affeeting the flow .
(e) Show that the veloeity at a point of the boundary x = O is given by
V y = -(m!n)(a/(y2 + a 2)] , and that the velocity at a point on the line y = O is given
by V y = (m / n)(a/ (x 2 - a 2 »).
34. Fluid is introdueed at a steady rate m through a small opening at the origin into
an infinite ehannel oeeupying the region ( - 00 < x < + = , O -< Y -< b) of the z planeo
The flow pattern in the ehannel is required.
(a) Use the mapping of Problem 23(a) to obtain a transformed problem in whieh
fluid is introdueed into the upper half of the w plane at the rate m at the point w = 1,
and is wíthdrawn at lhe rate m /2 at the points w = O and w=. Notieing that a souree of
strength 2m then must be present at w = l, and a sink of strength m at w = O (as well
as at w=), sinee only half the fluid enters the upper half-plane from a souree on its
boundary, obtain the complex potential in the w plane in the form
lI> = ;[IOg (w - 1) - ~ log w] + constant

= m log
n
(WJ / 2 _ W-I.' 2) + constant.
[The availability of an arbitrary additive constant is noted for the purpose of part
(b).]
(b) Obtain the required complex potential in the fonn
~ m i · h nz
'W' = Ti: og Sin 2b'
neglecting an arbitrary additive constant.

35. (a) Show that the velocity vector in Problem 34 is of the fonn
. m n(x - iy)
Vx + 1V y = 2b coth 2b '
and that the velocity along the boundary y = bis given by V x = (m/2b) tanh (nx/2b),
whereas the velocity along the boundary y = °
is given by V x = (m/2b) coth (nx/2b).
(b) Show that the stream function is of the form
I¡J -
m co t- I tanh nx/2b ,
-
- n tan ny/2b
and verify that I¡J increases by m as a small semicircle is traversed about the origin in
the counterclockwise direction. Deduce also that the streamlines are the curves
ny nx
tan 2b = e tanh 2b .
36. (a) Show that the complex potential
lI> = V a( z + :2) + i;'~ log z

defines a flow past the circle I z I = a, in which the velocity tends to Va as I z I ~ co, for
any constant real value of m'. (See also Problem 48 of Chapter 9.) (Notice tl1at a posi-
tive value of m' corresponds to the presence of a clockwise circulation.)
(b) By making use of the result of Problem 29(b), show that the drag D and lift
L, per unit length, exerted on a cylinder with I zl = a as a cross section is then given by
-D + iL = - ~ ip fJVo(1 - ~~) + iZ';'J dz,

where e is the circle I z I = a.
(c) Show that the integral involved in par! (b) can be e xpressed in the fonn
Jor'" [Vo( 1 - e- 2i") + i m ' e- l "] [ VoO -

2na
e"") - i m '
2na
el"] iae l" dO.
By noticing that al! product tenns of the form el'" will integra te to zero when r is an
integer other than zero, and to 2n when r = 0 , obtain the value of the integral in the
form -2Vom ' . Hence deduce that
D = 0, L = pVom' ,
where m ' is the clockwise circulation.
Problems 685
Section 11.7
37. Use the mapping (48) to obtain the solution T(x, y) of Laplace's equation in the xy
plane for which
T(x, 0+) = -vil - x 2, T(x, 0-) = --vil - X2
along a slit Ixl -< 1, io the form

T = sin rp,
p
where p and rp are to be determined in terms of x and y by use of the relatioos (50).
In particular, show that T(x, O) = O when Ixl ~ 1 and that
T(O, y)
38. (a) If rp solves the problem

¡ Iyl
Iyl
+ +
1
-vI;2
+ -vly2 +
1
1
(y> O),
(y < O).
V2rp = O in <R, rp = g on e,
where e is the boundary of <R, and if I{I is the harmonie eonjugate of rp, so that rp + il{l
is aoalytic in <R, show that I{I solves the problem -
on e,
where ajas and ajan differentiate with respect to distance along e and in the direction
of the outward normal, respectively.
(b) Verify the result of part (a) when
sin x sinh y
rp(x, y) = sio a sinh b
in the rectangle (O -< x -<: a, O -<: y -< b).
39. (a) Mapping 01 exterior 01 ellipse onto exterior 01 circle_ Verify that the transfor-
mation
e2
z= w+-
4w
maps the exterior of the ellipse
(a> b)
ooto the exterior of the circle I w I = i(a + b), where e is the distance -vi a 2 b 2 from
the center of the ellipse to each focus, in such a way that the mapping is not distorted
or rotated at large distances from the ellipse. [Notice that the point w = 1:Ca + b)e i •
corresponds to the poiot z = a cos rp + ib sio rp-l
(b) Mapping 01 plane with finite cut onto exterior 01 eircle. As a limiting case of
part (a) as b ---> O, show that the transformation
a2
z = w + 4w
maps the z plaoe with a cut along the reaL axis from x = -a to x = +a ooto the
exterior of the circle I w I = aj2, with no distortioo or rotation at large distances from
the cut. (Notice that the point w = -±ael • corresponds to the point z = a cos rp.)
40. Rota/ion o/ mappings o/ Problem 39.

(a) Suppose that the ellipse defined in Problem 39(a) is rotated about its center
through [he angle IX, where IX is positive when the rotation is counterclockwise. By
considering the result of replacing z by e- i ' z and w by e-"w in the transformation of
that problem, show that the exterior of the resultant ellipse is mapped onto the exterior
of the circle Iw I = ~(a + b) by the mapping
where again e' = a 2 - b 2 , without distortion or rotation at z~.

(b) In the limiting case b = O, show that the transformation
maps the z plane, with a cut of length 2a extending from the point -ael' to the point
ae i ., onto the exterior of the circle I wl
= a/2, without distortion or rotation at z~.
+ b)e'('+P) corresponds to the boundary point
(c) Verify that the point w = 1(a
z = eI'(a cos P + ib sin p) in the transformation of part (a), and that the point
w = 1aei('+P) corresponds to the point z = ae" cos P on the cut in part (b).
41. Flow about an elliptic cylinder. Show that the complex potential for an ideal fluid
f10w in the xy plane about a section of an elliptic cylinder, with center at the origin,
with semimajor axis of length a along the line fJ = IX, and with semiminor axis of length
b, which tends to a uniform f10w with velocity V o in the positive x direction at large
distances from the ellipse, is obtained by eliminating w between the two equations
'"
"" =
[
Vo w + (a +
4w
b)2] -
.2n
I
m
log w, z = w +
(a2 - b2)e2i~
4w '
where m is the circulation. (Notice that the first equation defines the complex potential
for the f10w of Problem 36 in the w plane about a circle of radius 1(a + b), and use the
result of Problem 40.)
42. By specializing the result of Problem 41 when b = O and IX = n/2, deduce that the
complex potential for a steady f10w about aplane barrier of width 2a, at right angles
to the flow and with center at the origin, is of the form
<1> = V o(Z2 + a 2)(/2 _ im sinh-¡~,

2n a
where V o is the Iimiting uniform velocity in the x direction. (An alternative derivation
of this result is given in Section 11.8).
43. By using the result of Equation (230) of Section 9.14, show that the steady-state
temperature distribution in the upper half of the w plane, which reduces to unity along
the 11 axis when I u I < 1 and to zero elsewhere along the u axis, is given by
T = -1 ( tan-¡ -
1 - II
- - +. tan- J -
1 ..L,-U) = -I u2 + v2 1
cot-¡ =-----'---,i----'-
n v v n 2v
(See footnote, page 646.)
44. The steady-state temperature distribution is required in the semi-infinite strip
(O -< x -< a, O -< y < co), subject to the requirements that it reduce to unity along
the end y = O and to zero along the sides x = O and x = a. Use the results of Prob-
Problems 687
lems 23(b) and 43 10 oblain this distribution in the form
T = ~ cot- I [ ( sinh 2 n; _ sin2 n~~ ) / 2 sin n;" sinh n;J = ~ tan- I (sin n~'(/sinh n;).
(Notice tha! the methods of Chapter 9 would give this solution in the forro of an innnite
series.)
45. The sieady-state temperature is required in the semicircle bounded by the upper
half of the circle X2 + y2 = R2 and the x axis, subject to the requirements that it
reduce to unity along the curved boundary and to zero along the straight boundary.
Use the results of Problems 23(d) and 43 to obtain the distribution in the form
-
L-21 (R-r - -Rr ) ] - sm
. 2 1=2
j
2 ()
T=..!...cot- I tan -¡2rRsin(}.
n (R r). ()
r - R sm
n R2 -
2
r
46. The steady-state temperature T(x, y) is required in the upper half-plane y :> O
T(x, O) = O (-co < X < -1), T(x, O) = 1 (1 < x < col
and aT(x, O) = O (Ixl < 1)

ay
along the x axis.
(a) Use the mverse of the transformation of Problem 23(b), m the forrn
z = -cos nw
(taking a = 1), to map this problem into the problem of deterroining the solution
T of Laplaee's equation in the semi-infinite strip (O -< l/ -< 1, O -< v < col for whieh
T = O along the edge II = O, T = 1 along the edge II = 1, and aT/av = O along the
end v = O.
(b) Notieing that the solution of the transforrned problem is T = l/, by inspeetion,
deduce that the required funetion is
T = ~ Re [eos-¡ (-z»).
(e) Verify tha!

.y'(x + 1)2 + y2 = eosh n v - eos nl/, ,y'(x - 1)2 + y2 = eosh nv + cos nu
anct deduce that T can be expressed in the explieit form
T = ...!..
n eos- I +[,y'(x
_ - 1)2 + y2 - ,y'(x + 1)2 + y2).
(d) Verify the satisfaetion of the prescribed eonditions when y = O.
Section 11.8
47. Derive the mapping of Problem 23(a).
48. Derive the mapping of Problem 23(b).
49. (a) Show that the mapping of Figure 11.14, in whieh the region aboye the indicated
polygonal boundary is mapped onto the upper half of the IV plane, is specified by the
relation
z = ni [(1 - W 2 )1/2 - COS- 1 W + n].
(b) When the regio n be/ow the polygonal boundary in Figure 1 l. l 4 is mapped onto
the upper half of the w plane, obtain the mapping relation
i 1
z = -[cos- W - (1 - W 2 )1/2].
n
(In both cases, the multiyalued functions must be properly interpreted. The fact
that neither relation can be inyerted, to express w as an explicit function of Z, un-
fortunately represents a situation of frequent occurrence. Text Example 3 illustrates
the feasibility of using such relations in applications in spite of trus inconvenience.)
50. Obtain a transformation which maps the regio n in the z plane, aboye the line
y = b when x < O and aboye the x axis when x > O, onto the upper half of the w
plane, in the form
Z =
b
n
[(w 2 - 1)1/2 + cosh- 1 w],
by making the point z = bi correspond to w = -1 and the point z = O correspond

to w = + 1.
51. The infinite strip (-= < x < +=, O <y < b) has a semi-infinite cut extending
in the positiye x direction from the point (O, b/2). It is required to map this cut region
onto the upper half of the w plane, in such a way that the x axis maps onto the segment
u < -1, the lower boundary of the cut onto -1 1.
(a) Show that the mapping must be such that
dz Cw C
dw - w2 _ l' z = '2 log (w 2 - 1) + K.
(b) Show that the point z = ib/2 maps onto w = O if and only if
!!.....
2 l -
- nC .
2 l
..L.
I
K
,
and that the line z = x + ib maps onto w = u, where u > 1, if and only if K = ib.
Hence obtain the mapping in the form
z·= ib - i'n log (w 2 - 1),
52. Suppose that the temperature is maintained at zero along the boundaries y = O
and y = b of the strip of Problem 51, and at unity along the semi-infinite cut. Use the
results of Problems 51 and 43 to show that the steady-state temperature distribution
T(x, y) in the strip is obtained by eliminating u and v from the relations
T = ~ cot-1 ,,2 + v 2
1
n 2v
u2 _ 'v2 = 1 + e- 2 ".:x/b cos 21ty 2 llV = -e- 2 n.XI'b sin
. 2ny
- -.
b ' b
53. Mapping 01 half-plane onto half-plane. Show that the bilinear transformation
w = az ""\' b
cz +d
Problems 689
defines a conformal one-lo-one mapping of the upper half of the z plane on to the upper
half of the w plane if and only if a, b, e, and d are real and ad - be > O. [Note that w
must be re a l when z is real. AIso notiee that dw,ldz must be real and po s itive when z
is real, in order that as an observer travels in the positive direction along the x axis,
with the mapped region to his left, his image will travel in the positive direction along
the u a x is, with the image regio n (to his left) accordingly the upper half of the w plane.]
54. Show that only the transformation defined in Problem 53 defines a conformal one-
to-one mapping of the entire upper half of the z plane onto the entire upper half of the
w plane. [Notice that the mapping funetion must have exactly One zero and one pole
in the extended upper half-plane (including the point at infini ty) and tlse results of
Section 10. 11.]
55. Deduce from Problems 53 and 54 that, if w = F(z) maps the interior of a regio n <R
conformally onto the upper half of the w plane, then the same is true of the mapping
.w = "-a",,F07-C--¿z):----,:
+----,-¡b
eF( z ) +d
when a , b , e , and d are real and ad > be, and only of that mapping. [Let t = F(z ) map
the interior of R onto the upper half of a eomplex t plane and then introduce the map-
ping w = (al + b)/ (et + d) . Notiee that, since the numerator aod denomioator can
be divided by one of the four con s tants, there are three independent parameters. This
accounts, in particular, for the assignabiJil y of Ihree u ' s in the Schwarz-Christeffel
trans formation (56) .)
56. (a) In the mapping of text Example 1, show that the three-parameter mapping
funetion corresponding to the re sult of Problem 55 is of the form
w =
az rr ! « + b.
cztt /at. + d
where a, b, e, and d are real and ad > be , and Ihal il yields the relation
:~ = [ca d - be) ~J(b - dw) (· ¡n)-I(ew - a)-(· /, )-I.
(b) Ir the requirements that z = °

map ooto w = and z = 1 map onto w = 1
again are impo sed, show that there results the one-parameter mapping funetioo
°
from whieh
:: = ~ (1 - k) a/'w'·/, )-l(1 - kw)-(· /·)-l,
where k = e/a is to be a real constant such that k < 1.

(e) Verify the eorreetness of the mapping of part (b), showiog in particular that the
point z,o maps onto the paiot w = l / k.
[Notiee that aeeordingly the third degree of freedom promised in the text and in
Problem 55 here eorresponds to the choice of k. The derivation of the relation (64),
whieh eorresponds to k = 0, in effect ignores the presenc e of a vertex at z~ and causes
it to be mapped onto w _. It can be seen also that when k ,: O the deri vative formula in
par! (b) would eorrespond te mapping "a vertex al z_ with an interior angle a _ = -a"
onto w = l / k. Whereas this phrase is not readily subject to a geometric interpretation,
it is in faet in aeeordance with the requiremenl thal the sum of the interior angles of a
closed polygon be (n - 2)n, where n is the number of ver tices, if we here consider n
to be two! lf, more plausibly, we assert that there are two vertices at Z~, so that n = 3,
then the sum of the "interior angles at those vertices" must be n-IX; but the Schwarz_
Christoffel transformation cannot accept this assertion. (A similar analysis of the
mapping of text Example 2 defines in the same way "the interior angle IX~ = -n" at a
single vertex at z~, in accordance with the requirement that the sum of the lour angles
IX l = IX) = n/2, 1X 2 = 2n, and IX~ be equal to (n - 2)n when n = 4.)]
57. Electrostatic fields about cylindrical conductors. In a region free of charges, in
which an elecirostatic field is independent of position normal to the xy plan e, the
electric field intensity vector E is the negative gradient of an electrostafic potential
ep(x, y) which satisfies Laplace's equation,
E = -Vep where V2ep = O.
(a) Show that the f8.ct that the component of E tangential to a perfect conductor
vanishes leads lo the conclusion that ep is constant along the boundary 01 a perlect con-
ductor.
eb) If 'I/(x, y) is the function conjugate to ep(x, y), so that ep + i'l/ is an analytic
function of x + iy, show that the fines of force then are the curves for which
'1/ = cons tan t.
(c) The surface charge density a on a conductor is given by -Keaep/an), where
K is the electric inductive capacity. Show that the total charge q on a cylindrical con-
ductor, per unit length normal to the xy plane, is given by
q =
:ri B a ds = -K
:ri B an
aep ds = -K
:ri B a'l/
as
ds = -K.6.B 'I/,
where .6. B'I/ is the increment in '1/ corresponding to a counterclockwise circuit of the
boundary B of the section of the conductor in the xy plane. [Se e Problem 25(b) of
Chapter lO.]
58. Capacity 01 a condenser. The capacity C, per unit length, of a condenser formed by
two cylindrical conductors Bl and B 2 , is defined as the absolute value of the ratio of
the total charge per unit length on either of the conductors to the potential difference
between them. lf el>(z) = ep + i'l/ is analytic in the region between the conductors, and
if ep is constant on each conductor, show that
C -1 epB,
q
epB,
\- K IepB,
B
.6. 'I/
epB,
\.
59. By making appropriate use of the function

el> = ep + i'l/ = A log z + B,
show that the electrostatic potential outside a circular cylindrical conductor is given
by
rp = - 2:K log r + constant,
where r is distance from the axis of the cylinder and q is the charge per unit length on
the cylinder, whereas the function '1/ is given by
'1/ = - 2:K e + constant,
where e is an angle measured around the axis of the cylinder.

Problems 691
60. Show that the capacity, per unít length, of a condenser formed by concentric
circular cylinders of radii R 1 and R 2 is given by
e = 2nK
log (R 2 /R , )
61. (a) By making use of the results of Problems 39 and 59, show that the result of
eliminating w between the relations
c2
el> = rp + il¡! = - 2::K log }ti + constant, Z = W + 4w
gives the potential rp and the function l¡! for the electrostatic field outside an elliptical
cylinder wi th cross section
X2
a2 + y2
b2 = 1 where a 2 - b2 = c2 ,
with charge q per unit length.

(b) Show that the elimination described in part (a) leads 10 the result
rp + il¡! = - 2::K cosh- 1 ~ + constant.

62. By combining the results of Problems 61 and 60, show that the capacity, per unit
length, of a cylindrical condenser whose cross section consists of confocal ellipses with
focal length e and semi-major axes a, and a2 is given by
e= 2nk = 2nK
10g(a,
al
+
+ vaf - c-
va! c~) - cosh-1 a2 _ cosh- l al
e e
[Notice that the ellipse x2/a 2 + y2/(a' - e 2) = 1 corresponds to the circIe 1}tI I
-Ha + va' - c 2 ) in the mapping z = }ti + c 2/(4w).]
Section 11.9
63. Sbow that G(x, y; .;, r¡) = G(';, r¡; x, y) if G(x, y; .;, r¡) satisfies (93) and (94), by
taking G = G(a, b; .;, r¡) and rp = G(c, d; .;, r¡) in (90) and deducing Ihat G(a, b; e, d)
= G(e, d; a, b).
64. Determine rp such that
a'rp +~ = O(x - C;)o(y - r¡),

ax' ay'
rp(x, O) = o, lim
Xl+y~_oo
rp(x, y) = o,
in the upper half-plane, by tbe following steps:
(a) If ifJ is the Fourier transform of rp with respect lo x, show Ibal
a,-
~ 1I 2 ifJ = e- iu { o(y - r¡).
ay2 -
(b) If ifJs is the Fourier sine transform of ifJ with respect to y, show that
-
rpS<u, v) -
_
-
e-fuI!
u' +sinv'V" .
692 Applications oC Analytic Function Tbeory
(c) Deduce that

sin v 11 J~ e-i.lol{~-x) du
rps(x, v ) = - ~?-"-'-'-
_7r. u 2
+v 2
e-· I ,:- xl sin v11

2v
and hence that
rp(x, y) =
_ -.l f~ e- ·I ~-xl sin v 11 d v .
7r. v
o
(d) Evaluate this integral by first differentiating with respeet to 11, then evaluating
the resu1t and integrating with respect to 11, noting that rp =
obtain (106).
when 11 = 0, and so °
65. Develop the one-dimensional analogy to the treatment of Section 11.9 for the
operator L = d 2 /d.x 2 in the interval (0, 1), as follows:
(a) Derive the "Green's formula"
f~ G Ly d¿; = [Gy' - G'y J~ + f>LG d¿;.
(b) Deduce that the solution of the problem

d2y
2 = h(x),
dx
y(O) = A, y(l) = B,
for °-< x -< 1, is of the form
y(x) = B aGJ~ 1) - A aG~~ O) + f: G(x, ¿;)h(¿;) d¿;
if G(x, ¿;) is such that

aa¿;2G
2
= ó(x - ';),
for ° -< ¿; -< l.

G(x, O) = 0, G(x, 1) = °
(e) Show that G(x, ¿;) = G(¿;, x), so that x and ¿; may be interchanged 10 the
definition of G in part (b). (Use the method of Problem 63.)
66. Let G(x, ¿;) denote the Green's function of Problem 65.
(a) Use the results of parts (b) and (e) of that problem to deduce that
a2G
l
{+<
lim
f'-O ~-f'
-a dx
X
2 = 1
and hence that
a~!r
[ axJ.:_ = 1
'
so that aG/ ax, considered as a function of x for a fixed value of ¿;, must have a uní!
jump as x increases through ¿;.
(b) Use the formulation
(x * ¿;),
Problems 693
G(O, ,;) = O, G(l, c;) = O,

a~{+
[ axJ{- = 1,
G(x, .;) =
¡~
~ (,;
T (x -
-
with G defined to be continuous in (O, 1), to determine G in the form
l)
1) (.; <: x <: 1),
and deduce that the solution of the problem

d'y
dx' = h(x),
y(O) = A, y(l) = B
is
y(x) = B~ + A(l - ~) + {G(X,C;)h(';)d';.

[Compare Problem 59(a) of Chapter 1. Since, to a first approximation, a small deflec-
tion y(x) of a flexible string under uniform tension T, and subject to a load distribution
of intensity /, satisfies the differential equation T d 2y/dx2 = fwhen fis positive in the
negative y direction (see Section 5.2), this Green's function can be interpreted as the
deflection of a string fixed at x = O and at x = 1, and subject to a unit concentrated
load at x =.; when T = 1.]
67. Obtain relations analogous to those of Section 11.9 associated with L = V2 in a
three-dimensional region <R. In particular, show that the relevant Green's function G
is to be such that
and G=O on S,
where S bounds <R, and that
1
G = G(x, y, z;.;, 11, O ~ - 4nR (R -> O)
where R = -V(x -';)2 + (y - r¡)2 + (z - 0 2•
Section 11.10
68. Pro ve that if F l (z) and F 2 (z) are both analytic functions in <R, such that w = FI (z)
and w = F 2 (z) each map the region <R onto the upper half of the w plane, then both
F 1 (z) and F 2 (z) yield the same Green 's function (110). [Use the result of Problem 55 to
deduce that F 2 = (aFI + b)/(cF, + d), where a, b, e , and d are real and ad - be > O.]
69. If w = F(z) maps the interior of a region <R conformally onto the upper half of the
w plane, wíth F(z) "" O in <R, show that
w = :::Fc'.:(z",,)_~F"",(,~)
F(z) - F(O
maps the interior of <R conformally onto the interior of the unít disk I w I ",; 1, with
dw/ dz;: O in CR and w ith z = ( mapping onto w = O. [Wrire F(z) = r + is and

F(O = P + i (j and s h ow that O <: 1 w 1 <: 1 when s :> O and (j > O.]
70. Suppose thal Ihe analy lic funclion w = K ( z) maps Ihe region CR onto Ihe unit disk
1 w 1 <: l. Sho w that the associated Green 's fUDction is
G( .)!) -
x, y, ." r¡ -
I I K (z) - K(OI
2n og 1 _ K(OK(z) ,
I
where z = x + iy and ( = .; + ir¡. [Verify Ihal Ihe conditions (96), (97), and (9 8) are
salisfied.]
71. Derive the Green's function for Ihe unil circular disk 1 z 1 <: I from the result of
Problem 70. [Take K(z) = z.]
72. Oblain the Green's function for Ihe first quadrant O <: x < ca , O <: y < co in the
form
G = -.L log 1z - (11 z + (1
2n 1z - ( 11 z + ( 1
= -.L lo [(x - ';)2 + (y - r¡J2][(x + ';)2 + (y + r¡) 2].
4n g [(x - ';)2 + (y + r¡J2][(x + ';)2 + (y - r¡)2]
Also show that
[aG] "~o -_ Ji4

an [y2 + (x -
xy,;
';J2l[y2 + (x + ';)2]
and
aQ'l _ 4 xyr¡
[ an J{~O - Ji [X2 + (y - r¡)2][X 2 + (y + r¡)2]
and verify thal the firsl of Ihese Iwo results is in accordance wilh the res ult of Problem
83, Chapler 9. [Notice rhat here F(z) can be taken to be Z2.]
73. Interpret Ihe result of Problem 72 in terms of "images" of a basic singularily at
z = (. Also write down the correspooding Green's function s associated with tbe
right half-plane and wilh the rhird quadrant, by use of the "melhod of images."
74. Obtain Ihe Green's function for the infinite sector O <: 8 <: 0:, O <: r < co in the
polar form
where'; + ir¡ = ye'p. Also show that

r ,
11, a. y"I « .
Sin _
n8
[~f]p~o = ~y -r-2-'!-d--2-r-'!-d-y-'!-d-C-O-S--::~""'e'--+-y-2-'-!d
o:
and
rlf ! «yll! (l .
sin _n8
o:
ay r 21f/ 4 + 2rn/ ayl'l/ CIt cos nO + y'Z.n/«
o:
75. (a) Specialize Ihe results of Problem 74 to the case o: = n/2 and verify that they
then are equiv alent to the results of Problem 7 2 .
Problems 695
(b) Specialize the results ofProblem 74 to the case e!: = n and verify that they then
are equivalent to the relations (106) and (l07).
76. Use the results of Problem 74 to show that the solution of the problem
V2T= O,
T(" O) = f(r), T(r, e!:) = g (r),
for O < r < 00, O < 8 s e!:, can be written in the form
T(r 8)
,
= 1. [J ~
n
.
f(ua !n) p sin q¡ du
o p2 510 2 q¡ + (u P cos tp)2
+ J~
o p'
al n
g(u ) p sin q¡ du ]
sin' tp + (1/ + p cos tp) 2 '
where p = r n1a and q¡ = n8/e!:, so that zn/ a = pelo.
77. (a) Verify that the mapping

w = -1--
.1 - z
1 + z
transforms the exterior of the unit disk (1 z I > 1) into the upper half of the w plane.
[Note the rever sed sigo relative to the mapping of Problem 23(c).]
(b) Show that the Green's function for the exterior problem is the same as that
given by Equation (116) for the interior problem.
(c) Account for the negative sigo in the result of Problem 35, Chapter 9.
78. Use the result of Problem 45(b), Chapter 9, 10 obtain the Green's function for the
rectangle O < x < a, O < Y < b in the form
. mnx . nny . mn'; . n7UZ
4ab ~ ~ SIO --¿¡- SlO b SIO --¿¡- SIO b
G(x, y; .;, r¡) = - n' m~¡ n~¡ m2b2 + n'a'
(This result can be expressed in c10sed form only in terms of elliptic functions.)
Section 11.11
79. (a) Obtain the analogy to Equation (127) in the one-dimensional case when
Ly = py" + sy' + qy
in an interval (a, b) and show that then
L*y = py" + (2p' - s)y' + (p" - s' + q)y.
Thus deduce that L is self-adjoint [or "formally self-adjoint" (see footnote, page 661)]
if and only if
s = p',
in which case there follows
Ly = (py')' + qy.
[Since any equation of the form aoy" + a¡y' + a2Y =
fcan be written in an equivalent
form (py' )' + qy = h (see Section 5.6), there is no 1055 in generality if this form is pre-
sumed, so that the associated operator L is self-adjoint.]
(b) When
Ly = (py')' qy, +
696 Applications oC Analylic Function Theory
s how Ihal Ihe Green'g formula of par! (a) lakes Ihe form
f CLydc; = [p( Cy' - ~gYH -1. f yLCdc;.
80. (a) Use Ihe resull of Problem 79 lo show thal the solulion of Ihe problem
fx [p(x) ~~J + q (x)y = h(x),

y(a) = A, y(b) = B,
for a <: x <: b, is of Ihe form
y(x) = Bp(b) ac~~ b) _ Ap(a) aG~~. a) + f C (x, e )h(c; ) de
if C = C(x, c;) is such Ihal
:.;[P(C;) ~gJ + q(';)C = 0('; - x),
C(x, a) = 0, C(x, b) = 0,
for a <: .; <: b.
(b) Use Ihe method of Problem 63 lO show Ihal here C(x, = C(e, x), so that x e)
e
and can be inlerehanged in the preeeding definilion of C.
(e) Use Ihe melhod of Problem 66(a) lo show Ihat C must be sueh Ihat
aQlu = _1 ,
[ a:d~ - p(';)
under the assumption Ihal p(x) * °for a <: x <: b, so Ihal C can be specified by Ihe
alternalive formulalion
J...(p
dx
aC) + qC = °
dx
(x * C;),
C(a, .;) = 0, C(b, C;) = 0,
[~~:: = pf.;) ,
wilh C defined lo be continllolls in (a, b). (Compare Problem 57 of Chapler l.)
81. Use Ihe resull of Problem 80 lo oblain the solution of Ihe problem
d 2y
- - k2 y = h(x) t
dXl
y(±co) = °
in Ihe form
when k> O. (Compare Problem 93 of Chapler 5.)

82. (a) Show Ihal, if Ihe eondilion y(b) = B is replaeed by y'(b) = e in Problem 80,
Ihen Ihe eondition C(b,';) = °
musl be replaeed by aC(b, i;)¡ax = O.
Problems 697
(b) Use the result of part (a) to obtain the solution of ¡he problem
d 2y
dX2 = h(x),
y(O) = A, y'(l) = e
in the form
y(x) = A + ex + S: G(x, ,;)h(c;) d,;,

(O -< x -< c;),
where G(x, c;) = {=~ (e; -< x -< 1).
[The results of Problem 80 generalize analogously to admit end conditions of the
form
k,y(a) + k 2y'(a) = A, k,y(b) + k 4 y'(b) = B.
In this more general case (but generaIly not when a condition is of two-point type) the
Green's function again is symmetric.]
83. For the operator
a2 az a a
L = A ax2 +e a yZ + D ax + E ay + F,
show thal
JS (f{ G Lrp dx dy = t (P cos el + Q cos {J) ds + SS (f{ rp L *G dx dy,
where
P = AGrp, - rp(AGlx + DGrp,
Q = eGrpy - rp(eG)y + EGrp,
and
az a2 a a
L* = A -a
X2
+ e-a
y2
+ (2A, - D)-a
X
+ (2ey - E)-a
Y
+ (A xx + e yy - Dx - E y + F).
Deduce that L then is self-adjoint if and onJy if

D = Ax, E = e"
in which case
Lrp = (Arp.,)x + (erpy)y + Frp
and
Q = C(Grpy - Gyrp).
[When B aZjax a y appears in L, the self-adjointness conditions become D = Ax + -!B y,
E = e y + -tB" in which case Lrp = (Arpxlx + '.<Brpx)y + .z(Brpylx + (erpy)y + Frp.]
84. Show that
a a a a a a
-
ax
= COS el -
an
- cos {J - ,
as
-
ay
= cos el -
as
+ cos {J -
an
,
on a curve e with unit normal vector n = i cos el +j cos {J, and use this result to
expre ss the bo undary integra l in Pro blem 83 in the form
fe [ CA cos' CG + C cos 2 /3)( G ~ - ~~ rp ) - CA - C) cos (X cos f3 ( G ~ - ~~ rp ) ] ds
when L is self-adjoint.
85. Use the method of Problem 63 to prove that
G(x, y ;'; , 7]) = G(';, 7]; x, y)
if it is true that both L * = L and al so the boundary integral in (130) va ni shes when
rp(';, 7]) satisfies ¡he same homogeneous conditions on ¡he curve C in the ';'1 plane as
does G. [Notice that then it is true that
ff <Jt G Lrp d'; dr¡ = fJ<Jt rp LG d'; dr¡
when G and rp satisfy the same homogeneous conditions on C.l

86. Show that , if G is to be a Green 's function for L = V 2 a ssociated with the entire
plane, and accordingl y must be independent of ang le about the point C';, 1]) , then it
must satisfy the equation
d ( dG) (R =;1= O)
dR R dR = O
and hence cannot be made to approach zero as X2 + y 2 --> co .

87. Use the Fourier transform to determine the Green's function for the modified
Helmholtz operator L = V2 - k 2 in the entire plane, as follows:
Ca) Starting with ¡he formuJation
a2 G + a2 G _ k2G = oCx - .;)o(y - 1] ) ,

ax 2 ay2
lim G = O,
take ¡he repeated Fourier trans form, relative to bo th x and y , and deduce that
and hence that
G= - 2n2 _ _
1 f-f- __ u 2
e ;(" (x -~ )+v(, -. )l
+ v2 + k 2 du d v .
(b) By introducing polar coordinates (y,8 ) such that II = Y cos 8, v = Y sin (J,
and writing
x -.; = Rcos '1/, y - 7] = R s in '1/,
show that the preceding result can be written in the form
G = - _1_
4n 2
f- [f1. e" R co, (S- IV) d8]
y2
Y
+dy k2
o o
and, by referring to ProbJem 65 of Chapter 5, deduce that
G = - _1
2n
f- yJo(rR) dr
o r
2 k2 +
Problems 699
[The integral obtained here is known to have the value K o(kr) when k > O, in ac-
cordance with the result of texl Example l.]
88. By replacing G by N(a, b;';, '1) and rp by N(e, d;';, '1) in (90), and using (149) and
(150), deduce that
N(a, b; e, d) = t N(a, b; ,;(s), r¡(s»IX(s) ds
- f_ e N(e, d; ';(s), r¡(s»a(s) ds + N(e, d; a, b).

Hence show that the satisfaction of (153) ensures that N(x, y;';, '1) = N(';, '1; x, y).
89. Verify that the function
1 -
N = 2n log I (z - ')(1 - zO I
serves as a Neumann function for the unit disk x2 + y2 -< 1, with a(s) = 1/(2n) in
Equations (150) and (151). [Writing z = re'" and , = ye'/1, show that
1
N = 4 n (log [r 2 - 2ry cos (e - {3) + y2] + log [1 - 2ry cos (e - {3) + r 2y2]}
and verify, in particular, that the condition [aN/ay],_¡ = 1/(2n) is satisfied.]
90. (a) Verify that, if w = K(z) maps a region ffi conformally onto the unit disk
I w I -< 1, then
1 --
N = 2n log I [K(z) - K(')][l - K(z)K(')] I
serves as a Neumann funcUon for ffi. [The fact that
i aN
re an ds =
1
2n ri cIK'(')lds = 1
2nJo
i 2n
d{3 = 1,
where e is the boundary of ffi, follows from results of Problems 89 and 28.]
(b) By interchanging z and w in Problem 23(c) and taking R = 1, deduce from
part (a) with K(z) = Ci - z)/Ci + z) that the function
N =
1 1 (z -
2n log (z
')(z -
+
i)2(, +
üi)2 1
is a Neumann function for the upper half-plane. (Here an irrelevant additive constant
was discarded.) Also verify directly that
[ - aNJ
ar¡ FO = n(';2
1
+ 1) = a(';)
and hence also that (150) and (151) are indeed satisfied.
(c) Deduce that, if w = F(z) maps a region ffi conformal1y onto the upper half-
plane, then the function
N = _1 lo 1 [F(z) - F(O][F(z) - F(')] 1
2n g [F(z) i]2[F(') + + iJ2

is a Neumann function for ffi.
91. Modify the treatment of Problem 88 to show that tbe satisfaction of (158) ensures
that H(x, y;';, '1) = H(';, '1; x, y).
92. Verify !hat the function
H = .,J-
_n log 1(z - ,)( I - z() 1- -41
n (1 z 12 + "12)
is a Hilbert function for the unit disk XO + y2 -< 1, with p«(" r¡) = l/n
in (154) and
(156). [Notice that H = N -c- (r 2 + y2)/(4n), where x + iy = re'· and (, + ir¡ = ye'P,
and where N is the Neumann function of Problem 89, and recall that V2N = O when
R 7= O and aNjan = 1/(2n) on the circular boundary.]
93. Le! w = fez) map a region CR conformally onto a region CR *. With the abbrevia-
lions V" = a 2 /ax Z + a Z /ay2 and v*z = a 2/au Z + a Z /av 2 , where w = u + iv, show
tha!
V*2 _ V2rp
rp - 1 /'(z) IZ
and that
ff eR' V*2rp da dv = ff eR V"rp dx dy.
[Use Equations (19) and (41), together with (46) of Chapter 7.]
94. Use the result of Problem 93 to justify the use of con formal mapping with refer-
ence to !he Hilbert function. In particular, proceed as in Problem 90(b) to deduce that
2
in 1(~z ; i~~i, ~ ~;21- ¿n (1; + : 1 + 1~ ~ : 1)
2
H = log
is a Hilbert function for the upper half-plane and that, if w = F(z) maps CR conformally
onto the upper half-plane, then the result of replacing z by F(z) and , by F(O in this
function is a Hilbert function for CR.
95. Develop the analogy N(x, (,) to Neumann's function of text Example 3 in cor-
respondence with the problem
dOy
dX2 = h(x),
y'(O) = e, y'(l) = D,
shoWihg that the condition (148) corresponds to the compatibility requirement
D - e = f: h(x) dx,
that the conditions (149), (150), and (51) correspond to the formulation
a'N
a(," = 0«(, - x),
with IX, + IX , = 1,
and that, when the compatibility condition is satisfied, the problem solution is
y(x) = A + eN(x, O) - DN(x, 1) + f: N(x, (,)h«(,) d("
where A = IX,y(O) + IX 2 y(l) is an arbitrary constant. Show also that the syrnrnetry-
ensuring condition (153) corresponds to the requirement
IX, N(x, O) + IX 2 N (x, 1) = constan!.
Problems 701
FinaJly, with the special choice et¡ = et 2 = ±, show that

N(x, .;) = i: Ix - .; I + constant.
[Notice that an arbitrary funcrion of x can be added to N(x,,;) if symmetry is not
required.]
96. Proceed as in Problem 95 with the analogy H(x, .;) to Hilbert's function in cor-
respondence with the specified problem, showing that H is to be determined by
aa';2H
2
= 15('; - x) - P(';),
H,(x, O) = 0, H~(x, l) = 0,
where f~ p(.;) d'; = 1,
and that then, if the compatibility condition is satisfied, tbere follows
y(x) = B + CH(x, O) - DH(x, l) + f~ H(x, ';)h(';) d';,

where B = J~ p(';)y(';) d( is an arbitrary constant. Show also that the symmetry-
ensuring condition (158) corresponds to the requirement
f~ H(x, .;)p(.;) d'; = constant.
Finally, with the special choice p(x) = 1/1, show that

X (.; -< X)} X2 + ';2
H =.;
{ (.; -< x) - 21 + constant.
[Notice that an arbitrary function of x can be added to H(x,,;) if symmetry is not
required.]
97. Deduce Equation (174) from (165), (172), and (173).
98. Derive the Green's function (184) (with the upper ambiguous sign) for the heat-
flow operator in the region -< x 00, ° -< °
-< I < 00, with ({I(O, 1) assumed to be pre-
scribed, by use of tbe method of variation of parameters (Section 9.17) as follows:
(a) For fixed x and 1, write
G = f: C(u, .. ) sin u'; du

and show that
ac
- -
2
et 2 u 2 C = - 15(.. - 1) sin ux.
a.. n
(b) Deduce that
C(u, .. ) - - 2 e-rx-u
. ,&-T
( ) SIn
. ux (.. < 1)
n
and hence that
G= 1 [e-[({-xP/4a.2(r-'.T)] _ e-[(<!+x)2/4a.2(t-.dJ]
2et..j n(t - .. )
when .. < 1, by virtue of Equation (241) of Chapter 9.

99. Use the method of variation of parameters (Section 9.17) to determine the Green ' s
function for the heat-f1ow operator L = a'(a'tax 2 ) - aja, when CR is the domain
o < x < /, o - T).
, n- I "
[F or fixed values of x and 1, assume
G = ~ C('r) sin n1T.,';

n_ I
in Equation (176) and deduce that

, n2 1T. 2 a 2 2 s: • n1T.x
Cn - '2 C n = T u(T - 1) SIn - , -
- and accordingly
when T < l.]

- 1 e- (C-x)~/4a2 (r-1')
S ( x - C;;. I - T) = 2a"; 1T.(t T) ,
for the negalive of the Green's function associated with the heat-flow operator in the
region - 0 0 < x < 00, O ~ I < 00, use the method of images to show that the cor-
responding Green's function in the region O ~ x -< '. O -< t < can be expressed as =
the sum
G = - ~ [S(x - .; + 2n', t - T) - S(x +.; + 2n/. t - T)]
n'""'-oo
if G is lO vanish when'; = O and when'; = /, as Ihe sum
G = - ~ [S (x - .; + 2n'. I - T) + S(x + .; + 2nl. I - T)]

, , - -OQ
if aGja.; is to vanish when'; = O and when .; = 1, and as the sum
G = - ~ [-S(x - .; + 21 + 4nl.' I - T) + S(x + .; + 4nl. t - T)

+ S(x - .; + 4nl, t - T) - S(x + .; - 21 + 4nl, t -- T)]
if aGja.; = O when .; = O and G = O when .; = l. (Nolice Ihat S is symmetric in x and
.; (but not in I and T), These expansions may be preferred lO series of the type obtained
in Problem 99 when t is smalL]
Answers to Problellls
Chapter 1
1. (a) y' - y = O.
(b) y' - 2xy = 2 - 4x2.
(e) y' - y2 = O.
(d) xy' - y lagy = O.
(e) yy' + x = O.
(f) X2y'2 = (4y +
l)(xy' - y).
2. (a) y. - y = O.
(b) y" - 2y' + Y = O.
(e) y" + y = O.
(d) yy" - y'2 = O.
(e) ( y - 1)y" - 2y'2 = O.
(f) y"2 = (1 + y'2)3.
3. (a) y = ee x '.
(b) x - e = --! 1 - y2, Y = ± 1.
(e) (y - I)(x - 1) = e(y + I )(x + 1).
(d) --! 1 - y 2---! I - x 2 = e, x = ±I,y= ±1.
6. (b) Dependent.
7 . (a) Dependent.
(b) Dependent.
(e) Independent.
(d) Dependent.
13. (a) y = x 2/ (2 - k) + ex· i.f k ,,: 2 ; Y = x2 log x + ex2 ir k = 2.
(b) y = x tan x + 1 + e see x.
(e) y = sin x + e eas x.
(d) y = 2 sin x - 2 + ee-,IQx.
(e) y = (x + l)(x + e) /(x - 1) .
(f) y = (ex - ¡ )/(x lag x).
(g) y2 = x(x + e).
(b) x = y(y + e).
703
704 Answers to Problems
18. (a) y = e ¡r x + e2e h .

(b) y = e¡e- x + (e2 + eJx)e x .
(e) y = ex(e ¡ eos x + el sin x).
(d) y = ex(e¡ + e2X + e, cos x + e4 sin x).
(e) y = e,e x + e- x / 2[e2 eos (./3x/2) + e3 sin (./3x/2)].
(f) y = c¡.e(l+¡)x + C2e-( l+f).~.
19. (a) y = e¡e kx + ele- kx + C3 cos kx + e4 sin kx.

(b) y = ekx(e, eos kx .,.. e2 sin kx) + ekx(eJ eos kx + e4 sin kx).
(e) y = ekx(e¡ e2x) + e- kx (e3+ e4x). +
20. (a) y = e eos kx + e2 sin kx + (sin x)/(k l - 1) ir k 2 "'" 0, l.
¡
(b) y = e ¡ eos x + e2 sin x - ,Ix eos x.
(e) y = cle x + e2e- x - t sin x.
(d) y = e,e" + e2e-x + ixex.
(e) y = e,e x + e2ex + HX2 - x)e x .
(f) y = eX(e ¡ eos x + e2 sin x) + !(2 eos x + sin x).
(g) y = ex(e ¡ eos x + e2 sin x) - ,Ixe x eos x.
(h) y = e ,e 4x + e2e'x + (2x 2 + 6x + 7)e h .
(i) Y = e¡e 2x +
e2e -x_ xe-" - 3.sinx eas x 2x - l. + +
21. (a) y = e,x k + e2x-k ir k"", O; Y = e, + e2 log x ir k = O.
(b) Y = x(e, eos log x + e2 sin log x).
(e) y = e¡x 2 + e2x-'.
(d) y = x(e, + e2 log x).
(e) y = e,x-¡ + x(e2 + eJ log x).
(f) y = c¡x n + C2X- n - 1 •
(g) y = e¡x + e2.e' + 1x2.
(h) y = e,x + e2x-' + ,);x log x.
(i) y = e,x J -'- e2x2 + 3x + 2.
22. (a) y=e,(a + x)2 +e2(a+x)-¡.
(b) y = e,(a -+ X)2 + e2(a + x)-¡ + (a + X)2 log (a + x) - t.
(e) y = e,(a + x)2 + e2(a + x)-' + :l-(a + x)2lag (a + x) - 2ax - a 2.
28. y = Cle X + C2e- X + !xe x .
30. y = c¡x 2 + C1.X-1 + !x3.
32. x = e,1 + e2,y = -,Ie,12 - (e¡ + e2)1 + ej.
33. Ca) x = ele 21 + C2e!, Y = - e l e l / + Cle r •
(b) x = ele! + e2 cos t + C3 sin 1,
y = 2(e¡ - l)e' + (e2 - eJ) eos 1 + (e2 + eJ) sin l.
0
(e) x = e'(e, + e21) + eJe- 3'/2 - 11,

y = e'(6e2 - 2e ¡ - 2e21) - l¡eJe- J'¡2 - l¡.
(d) x = e, eos cú,kl + e2 sin úJ¡kl + eJ eos úJ2kl + e4 sin úJ2kl,
y = ~2(Cl cos úJ¡kt + C3 sin úJ¡kt - C3 cos úJ2kt - C4 sin úJzkt),
where úJ¡ =./2 -+ ./2 and úJl =./2 - ./20
(e) x = e 2, + 1, Y = e"' - l.
(f) x = ele-lE, Y = 5c¡e- 2r + eze-r.
(g) x = Cle 2t , Y = ~Cle2r + e2e-21. z = -~Cle2r - ~C2e-2( + e)e 3f •
34. (a) x = e,1 -¡- e212, Y = e,1 - e212.
(b) x = Clt 2 , Y = ~Clt2 + e2/- 2 , z = -!c¡t 2 - ~C2t-2 + C3t 3 ,
35. (a) py = S ph dx + e, where p = ela,dx.

(b) y = el eos x + e, sin x + (sin x) log (tan }x).
(e) y = el eos x + e2 sin x + (e os x) log (e os x) + x sin x o
(d) y = el eos x -r- e2 sin x + Sx sin (x - e) log e de.

(e) y = ex[el eos x + el sin x - (e os x) log (se e x + tan x)].
Answers lO Problems 765
(f) y ~ e¡x' + e2xz + x' sin ----xzx dx - X2 sin

X-x
f f dx.
(g) y = e¡x + e2x 2 - x log x - :l:xOog x)2.
39. y ~ elX + e2e" + 1.
40. y = el (! + x tan x) + e2 tan x.
41. y = e¡x + e2 (~ log 1: + ~ 1- 1) + ~ (x 2 + 1) .
49. Uo ~ .!.(
2 a
x + E..)
x'
u 1 ~ '!'(x
2
_ ax 2 ) (a? O).
SI. Uo = 1, Ul = sin x, U2 = 1- cosx.

52. Uo = :l:(eosh x + eos x), Ul ~ :l:(sinh x +
sin x),
U2 = :l:(eosh x - eos x), u, ~ :l:(sinh x - sio x).
54. (a) k ~ ntt/a, n a noozero iDteger.
(b) k ? m7t/a, m an odd integer.
60. (a) x2 y 3 + xy - y2 ~ e.
(b) X3 y - xy' ~ e.
(e) ye x + xeY ~ e.
61. (a) y2 + 2xy - X2 ~ e.
(b) y2 + xy ~ ex.
(e) y + "¡x 2 + y2 = e.
(d) y ~ x sin- l (ex).
62. (a) xy = e'X.
(b) x 3 - 3xy ~ e.
(e) y3 - 3xy ~ e.
(d) (y - 2)2 +
2(x - l)(y - 2) - (x - 1)2 ~ e.
(e) xy(x +
e) = 1.
(f) (x 2 +
1)2(2y 2 - 1) ~ e.
64. (a) X2 - y3 ~ ex 4 •
(b) xm + Y" = eyp.
65. (a) y = log (x + el) + e2.

(b) y = el tan- 1 (elx) + e2.
(e) y2 = elx + e2.
(d) y = e2 - [1 - (x - el)2jl / 2.
(e) y = el eos x+ C2 sio x.
(f) x = 2p' + p + Cl, Y = !p4 + :l:p2 + e2.
Chapter 2
3. (a) (s - a) / [(s - a)2 + k 2 ].
(b) n! /(s + a)·+l.
(e) (1 + e- u )/(s2 + 1).
(d) (e-,Q - e-'b)/s.
4. (a) 6/s 4 . (b) 2/(s + 3)'. (e) a(s2 - 2a 2)/(s4 + 40 4 ).
(d) 4(s - 1)/(s2 - 2s + 5)2. (e) (6os 2 - 2a')/(s2 + a 2)'.
(f) (s - a)/[(s - a)2 - b 2 ].
5. (a) s'j(s) - s2/(0) - sf'(O) - 1"(0). (b) - i'{s - 1).
N N
(e) L (n! a.)/s·+l. (d) L a.s/(s2 + n 2 ).
n- o 11= o
11. (tanh !as) /s2.
706 Answers lo Problems
(2 (4 (6
20. (a) 1 - 2 ! 3 ! -1- 4 ! 5 ! - 6! 71 -+- . ..
(b) 1 _ (1 / 2)0 -"- (1 / 2)4 _ (1 / 2)6 _L.. •••
(I!)2 '(21)' (3 1) 2 '
26. (a) (1 - eos a/)/ a. (b) (ea, - J - al )/a 2 (e) (e"' - eb')/(a - b).
(d) (b sio al - a sin bl)/(b ' - a 2 ).
28. (a) t e-U f(t - l/) du = t e-"-u' f(u) du.
(b) -1
a J' o
f(1 - u) sin au du = - J
a J'
o
f(u) sin a(t - u) duo
(e) t ue- au f(1 - u) du ~ t (1 - lI)e-ar,-u' f(lI) du.
(d) - -1b
Q-
J'
o
(e-bu - e-au')f(1 - u) du = - -1b
a- J' o
[ r bU - u ' - e-arh"]f(u) du.
33. (a) e 21 - el. (b) e' cos 21 + ter sin 2t.

(e) sin Tsinh T + (sin Teosh T - eos Tsinh T)/./2, where T ~ 1/./2.
(d) (1 + 2r' - 3r Z')/2. (e) 1 - sin l.
(f) e- U -" when 1 > 1 ; O when O <: 1 < l.
34. (a) (sin al + al eos al)/ (2a). (b) (sinh al + al eosh al)/ (2a).
(e) -e-' + 5e- l ' - 4e- 3 ,.
(d) [e-a, - e",n (eos ~./3 al - ./3 sin i'/} al)] /(3a 2 ).
(e) J when O <: 1 < J ; O when 1 > 1.
(f) (b sin al - a sin bl)/ [ab(b 2 - a 2 )] if b 2 "" a 2 ; (sin al - al eos al)/ (2a 3 ) if b = a.
35. (a) e-' eos l. (b) (sin al eosh al -+- eos al sinh al) /(2a).
(e) O wheo O <: I:S; 11:; -sio / wheo 1:> 11:. (d) ~(2 - 4a/ + a 2 12)e- a ,.
(e) H(3 - /2) sio I - 31 eos 1]. (f) He3 + (2) sinh 1 - 31 eosh 1].
36. (a) Stairease funetion: b (O < 1 .- a), 2b (a < / < 2a) , 3b (2a < I < 3a), ....
(b) Square-wave fuoetion: 1 (O < 1 < a/2), - J (a/2 < I < a), J (a < I < 3a/2), ....
(e) Reetified sine: sin ro/ (O < 1 < 11:/ro), O (11:/ro < / < 211:/ro),
sin rol (27</ro < I < 37</ro), ....
(d) I sin rol l·
39. O (O 1.
(d) Y = e- k ' [YO + f(u)e ku dl/]. t
41. (a) Y = e-' cos l. (b) Y = 1 - e-' eos l.
(e) Y = r' eos / when O <: / <: J;
Y = e-' eos 1+ e-"-" sin (/ - J) when 1 ;;": 1.
(d) Y = e-' [YO eos I + (Yo + y~) sin / + J~ f(II)e" sin (/ - u) du].
42. (a) Y = !(sin I eosh I - eos I sinh 1).
(b) Y = J - eos I eosh l.
43. Y = -[sin I sin (b - a)]!(sin b) when O <: 1 <: a;
y = -[sin a sin (b - I)]/(sin b) when a <: I <: b.
44. (b) x= {O[l - e-",(eosPI + p sin PI)] (PZ = rocr -1X 2 > O),
X = {o [1 - (J + 1Xt)e-·'] (IX = roo),
x = fO[1 _ IX + Ye-(.-y" + IX - Y e-(.+r),] (y2 = 1X2 - rol > O).
k 2y 2y v
(e) x = ae- a , ( eos PI + P sin (JI) (P2 = C05 - ex 2 > O),

x = ae-a'(l +
exl) (ex = coo),
x = a[ex + ~e- ('-Y), - ex - Ye- (<<+Yl'] (y' = ex 2 - C05 > O).
2~ 2y
45. x = -e-'(eos [ + sin [), y = e-' (l + sin t).
50. (a) ./7i/2. (b) 1.54. (e) 3.33. (d) 1.43. (e) -4.33. (f) 3./7i/4.
54. (a) O when ex > O; 1 when ex = O; no finite ¡imit when ex < O.
(b) 1/./7i.
Chapter 3
15. y(O.I) = 0.9949, y(0.2) = 0 .9785.

19. Rounded true values: (a) 1.5527; (b) 1.5841; (e) 0.07842; (d) 1.60Si.
21. Rounded true values: (a) 0 .3152; (b) 0.3121.
26. Rounded true values: (a) 1.4049; (b) 1.4333; (e) 0.07096; (d) 0.07024; (e) 2.8394;
(f) 5.6856.
Chapter 4
1. (a) Al! values of x. (b) -3 < x < 1. (e) 1 - ./3 <: x <: 1 + ./3.
(d) -1 < x < 1; also x = -1 if k < 1 and x = + 1 if k < O.
(e) Al! values of x. (f) x = O.
(g) a - 1 < x < a + 1 ; al so x = a + 1 if ex > 1 and x = a - 1 jf ex > O.
(h) -4 < x < 4. (i) x > 1 and x <: -1. (j) x > -1.
5. (a) y = e, ( 1 + xJ
2.3 +x6
2.3.5.6 +
x4
... ) + e2 ( x + 3.4 x'
+ 3.4.6.7 + ... ).
1·3x 6
(b) Y = e, ( 1 +
X2 X.
- 1·3·5x'
2! - 4! + ~ 8! + ... ) + e2X.
(e) y =e,(1 + ~~ + ~~ + ... ) + e2 (x2 + ~~ + "'s1!O + ... )
= el cosh X2 + e2 sinh x 2 ,
6. (a) y = Ao (1 - ~~ + ~~ - ... ) + A, (x - ~~ + ~~ - ... ).
(b)Y=Ao(l- ~X2+ ~xJ+ ... )+A,(x- ~x3+/2x'+"').
(e) y = Ao(1 + ~ x2) + A,x.
(d)Y=Ao(l+x + ~X2+ ... ).
(e) y=Ao(1 + t! + ~~ + ... ).

(O y = Ao .
7. (a) No singular points.
(b) No singular points.
(e) Regular singular points at x = ± ./2.
(d) Irregular singular point at x = O.
(e) Regular singular points at x = O and x = -1.
(f) Irregular singular point at x = O.
8. (a) Irregular singular point at x = O, regular singular point at x = 1.
(b) Irregular singular point at x = O.
708 Answers lo Problcms
(e) No singular points.

(d) lrregular singular points at x = ± 1.
(e) No singular points.
11. (a) y = c, (1 + + X2
~_!
x
+ -3 )
! + ... +
J , 2
C2X .
(1 + 2x + (2X)2 +- (2x)J , ... )
, x ]·3 1·3·5 , 1·3·5·7-'-
(eoeffieient of c, is ex).
(b) y = c, X- l i 2 (1 ~ ~~ + :~ ~ ... ) + C2XI ' Z(l ~ ~~ + ~~ ... )
= x- 1 / 2 (c,+ C2 sin x).
eos x
(e) y = c ,.e' (J - ~~ + :~ + ... ) + C2 (1 ~ ~~ + ~~ - ... )
= x-'(c, eos x + C2 sin x).
(d) y = e, (1 + x + x') + e2x J (J + x + x2 + x J + ... )
= e,(J + x + X2) + e2xJ/(l ~ x) (Ix 1< 1)
= (e, + C 2 x J )/(J ~ x) (lxl < 1).
12. (a) y = e,x(1 + ~~ + :~ + ... ) + C2X2(1 + ~~ + ~~ + ... ).

(b) y = e, (1 + ~ x2 .;- ~ x' + ... ) + e,x.
(e) y = e,(l ~ ~~ + :~ - ... ) + e2(x 2 ~ ~~ + x ,!a - ... ).

5
(d) y - ' e, (x - ~ x' + .,.) + e2(x2 - /2X4 + ... ).
13. (a) (-00,00). (b) ( - 1,1). (e) (-oo,co). (d) (-27<, 27<).
14. M(a, c; x) = J
+ i:; a(a+ 1 )(a + 2) ... (a + k - 1) Xk
k~' e(e+ l)(e + 2) ... (e + k 1) k!
=1 + .E... x + a(a + J) x2 + .. .
e e(e+l)2!
15. IX = O; y = Aa + A ?- (X2 ~ 2J!

X
3 + 3x' _
4!
.... )
= Aa + 2A 2 [l - (x + 1)e- x ] = e, + C2(X + l)e- x •.
. x )
3
x4 X
S
IX = 1; y=Aa(l-x)+A2 ( x2~3+3.4-3.4.5+···
= Ao{l ~ x) + 2A2(e- x - 1 + x) = c,(1 - x) + c2e- x •
~ k
16. (b) u, (x) = ¿; (t')2
k- o .
(d) v(x) = - (2x + ! x2 + /dsx3 + ... )
1
-2 [ (I !)2 x + !...±.J 2
(2 !)2'~ +
+(3 ~!)2+ j .]
x
00 k+l
17. (b) l/¡(x) = k~O x k ! = xe X
•
x = 1 ~ x ( T!x + 2! x + +1+,. )
1!....±l2]
(d) v () 3! x 3 + ....
+ x6! -,- 71
X2 Xl x 6 7
+ 31 + .. .
I
19. (a) y = 2!
X2 3x' 3·6x 8 3·6·9x"
(b) y = 2 ! ~ TI + ---s-! - 11 ! + ....
X2 x3 x4 XS
(e) y = 2(TTj2 + 4(3 !)2 + 5(4 !)2 + ....
+ 3(2 !)2
2 J
(d) y
4
= x- 1/2 ( 7' 4x
+ 3' + 2x
"7 + 2x 57 + ....
)
Answers lo Problems 709
20. See Problem 16.

21. See Problem 17.
22. Y = e ¡1I¡ (x) + C2[U ¡ (x) log x - v(x) ],
xl. x3
where u ¡ (x) = 1 + x + 2! + 3! + ... = eX,
~ (I+~+···+k) 3 11
v(x) = ~
.~¡ k .
I x, = X + -x2
4
+ -x 3
36
+
23. y = e¡u¡(x) + C2[U¡(X) log x + 1 - v(x)],
where u,(x) = x ( 1 +.~ + 2! + x3
TI + ... ) =
X2
xe x ,
~ (1 + ~ + ... + k ~ 1) 3 11
v(x) = .'f. 2
(k 1)! Xk = X2 + -¡-x
3
+ 36 x ' +
27. (a) 0.1483. (b) -1.081. (e) 0 .1924. (d) 0.1357.
(e) 0.9900 - 1.08li. (f) 0.4539.
28. x = c,Ja(t) + C2 Ya(/), y = e¡tJ,(t) + e2/Y,(t).
29. y = c,Ja(Ax + B) + e2 Y a(Ax + B).

30. (a) ,J'z/rG). (b) -2/ 1<. (e) 2. (d) 1/(2 n n !). (e) 2/ 1< .
46. (a) y = x 2[c,Jz(x) +

e2 Y 2(x)].
(b) y = x[e,JI/,(x 2) +
C,J-'/2(X')] = eos x2 e, + e 2 sin x2.
(e) y = c,II/2(x) +
e 21-1I2(X) = x-'/2(e, eosh x + e 2 sinh x).
(d) y = e-x[c,Ja(x) +
e2 Ya(x)].
(e) y = x[e,I,(x) +
e2K¡(x)].
(f) y = c 'x eos (a/x) +
e2X sin (a /x ).
(g) y = e,x' / 2I, / .(x 2 /2) +
e2 x ' / 2I-, / .(x 2/2).
(h) y = c,e-xla(x) +
c2e- x K a(x).
(i) y = e,e-x'Ja(x) +
e2e-x' ya(x).
47. (a) y = X' / 2Z(2m+ ')/20",x).
(b) y = x{2m+ I)/2Z(2m+ 1) /2 (i",x) .
48. y = x-¡J,(2x) .
49. y = e, J,(x) - iPY,(x).
50. (a) y = x'Z .(x/l), where p = (l - n)/ 2.
(b) y = x"Z .(x'/ Is), where p = (l - n)/(2 - n) and s = (2 - n)/ 2 , n "'" 2.
(e) y = x-' / 2(C, + e2 log x).
51. (a) 0 .9844. (b) 0 .2496. (e) 1.016. (d) 0.2483.
55. U = rn[e IPn(eos rp) + e2Qn(eos rp)].
71. (a) y = el + C2(1 + 1 ~x + 2 !IX 2 + ... ) = el + e2e l / x .

(b) y = el (1 - _1_ + _1_ _ ... ) +c 2 (.!. __ 1_ + 5!lx _ ••• )
2 ! X2 4 !x4 X 3 !x3 5
1 . 1
= el eos-
x
+ e2SIn-.
x
72. Y = e, ( 1 -
1 1
3 ! X2 + 5! x. - ... ) + C2 (x __1_ + 4!
_ 1_ _
x2!.~ 3
... )
. 11
= CIX StO -
x
+ Czx cos -::c .
Chapter 5
2. k = nn /' , where n is any inleger alher !han zero.

4. y = CJo(A I · 'X), where Jo(J. Ii!a) = O.
JO. (Un - P 1':'

- '\fIEI /2' w here cos JJ..n -- sec h J.ln,
.
sioh (Pn X/') - sin (pnx/') eosh (Pnx/l) - eos (Pnxl')

<f1n = sinh p" sin ¡J" - cosh J-L" cos J1." .
11 Wn = JI} j; ;
S ymmetrical modes: tan (Pn / 2) = - tanh (Pn/2),
eosh (Pnx/f) eos (Pnx/l)
!pn = eosh (Pn/2) - eos (J.l.n/2) .
Antisymmetrica' modes: tan (Pn/2) = +tanh (Pn/2),
sinh (Pnxll) sin (PnX/l)
!pn = sinh (Pn/2) - sin (Pn/2) .
19. Exaet value rounds lo 6.55.
22. (a) (X2y,), + (x 2 + ).x)y = O.
(b) (y' sin x)' + Ay sin x = O.
(e) (e ax y')' + eax(b + A)y = O.
(d) (x'e- x y')' + X,-I e-xC -a + J.)y = O.
31. (a) An = 2( 1 - eos nn)/(nrc).
(b) An = ( - l)n+12/1l.
(e) An = (1 - eos n rc) /;rc.
2
~
34. 1 = L: An sin Pn ~ (O < x < 1), where J.l.n + k lan J.l.n = O;

n-I
2(1 - eos Pn) 2(1 - eos Pn)
A Pn i SIO 2pn ~ Pn(l + k I eos 2 Pn) '
2
35. (e) f(x) = 1 : An = -(1
n1t
- eas n1t).
f(x) = x: An = n 2 rc 2 +2n1t(Iog b)2 (1 - b e Os nrc ) .

'
J
pTrX
37. o h(x) sin -,- dx = O.
y = ~
~
rf""l
. nrcx
a,.,sm-
/
with an = (p 2 2' n 2 )rc S: x
h(x) sin n7 dx (n '"" p), a p arbitrary.
. nx 3nx 5rcx
49. (a) -8'2 (1
rc' -l ' SIO - , + -3l3 .SIO -
,
-l .SIO
+ 5' -
I
+ ." ) .
(b) 4' (1 . 1tX
1t2 f2 SIO T -
l.
32 slo - , -
3nx
+ 1 . 51tx
52 SlO - , - - ....
)
(e) n2 (.SIO -rcx, . 21tx + 1 . 3rcx +

, .,.. SIO - , - '3 SlO - , - • • •
+ 1 - eos (nrc/2) . nrcx
n SIO - , -
+ ... ).
8 ~ (- 1 )n+ J n . nrcx
(d) - ¿",.¿ sln-'
n n~l 4n 2 t - 1
(e) 1 . nx 2 ~ n eas (nrc/2) . n1tx l . rc x 4 ~ (-1 )m+ J m . 2mrcx.

-2 SIO - , - - ¿",
n n"='2 n
2 _ 1 SlO - , - = -2 sin - , + -1t ¿",
m= 1
4
m
2 _ 1 SlO - , -
(f) 3. ~ 1 - eos (nnE//) . nrcx

1l 11= 1 nE SIn -1-'
,
50. ( a ) 12
6 -
4/ 2
,,2
( 1 2"x
2' eas -,-
+ 1
41 eas -,-
4"",
-¡-
1 6"x
6 2 eas -,-
)
+ ....
( b )4' - 8' ( 2'
,,2
1 eas -1-
2".'1: + 1 eas -,-
62 6"x + 1 eas -
la' 1O"X
, - T, .. ,
) .
1 "x 3"x 51tx )

(e) 2 + "2 ( 1
Teas T - 3I eas -,- + "'5I eas -,- - .. , '
(d) 2(1 - 2 i: n =: ¡4n

2
1 1 eas n1t,x).
nt
Jr
(e) ~ (J + eas 1t( + 1 n Sinn~n':.!.2] nrrx)

1 eas -,- .
(f) ..L, (1 + 21 ,,=~ sin (n"f/') eas n1tx).

7r 1 n€ I
54. Y ~
sin ~
2-'-2-LJ
w, n= 1
( - l)n+ I n1t sro n1tx
1
55. A ~ p 2 1t 2 / ' 2 : na salutian unless alsa w = rn /' (r a pasilive inleger, bul r 01= p), in whieh
case infinitely many salutians.
ro ~ pn/': lInique sallltian unless alsa A = p 2 n 2 / , ' , in which case no salutian,
12 21 2 ( 1 2nx 1 41tx, 1 6nx )

57. (a) 11 - -;¡r 22 cas -,- + 4' eas ~ ., 6 2 eas -,- + ...
4'2 (1 ,nx 1 , 3nx 1 , 51tx, )
+ 1t' J3 SIn T + 33 SIn - , - + 5' SIn - , - . , ....
, 4' ( 1 21tx, I 61tx 1 10nx
(b) '8 - nO 22 eas -,- ., 62 eas -,- + lUZ eas - , - + ... )
..L
,
2' (1
1t2 11 ,nx
SIn -,- -
1, 3nx
32 SIn - , -
+ I ,5nx
52 SIn - , - - ....
)
31(1 1tX 1 31tx 1 5nx )

(e) '4 + "
Teas -, - 3 eas -,- + "'5 eos ~ - ...
1 (, nx ,2nx 1, 3nx
- " SIn T - SIn-,- + 3 SIn - , - + ...
eos (nn / 2) - eos (nn) " nnx + ... ).
+ n SIn - , -
8 ~ (_I)n+l n ' nnx

(d) - 4J Sin _ _ o
1t ~ =l 4n 2 - 1 I
()e 2n1 (1 + nx
eas -,-
n , nx)
+ '2 SIn T
+l..2:: I ,nn
n Sin 1) cos n1tx
- - - n cos n1t
- . n1lx]
1 [(
- - Sin - - •
1tn=2n2- 2 1 2 1
1 ~ 1 [. n1t€
(f) 2'1 ", nf n~l nnnx
SIn -,- eas -,- + ( nn€) . n1tx]
1 - cas -,- Stn -1- .
bP) bP (1
58. (a) ( a +T - n T '
SIn
2nx..L I ,4nx
P l . 61tx
, '2 SIn P + '3 SIn P + ... ) .
(b) ~ + ~ (+ sin 1tX + ~ sin 3x + + sin 5n.~ + ... ).
(e) ~ - ~ (-teas2x + /5eas4x + ···+4n21_leas2nx+ ... )-
(d) 32 - "I (1
T '
SIn 2 1tX ' 4 nx + 31 SIn
+ '2I SIn ' 6' nx + ... ) .
, nn
sin 2 cos nx + - CQS nn) , nx
T sin (1
1
60. y = 41\ + nL:
= I
nn(A - n2)
63. (a) a o = 1, a n = O (n": O) .

( h) a"
= I sin leos n7T. .
/ 2 _ n2 n 2
(e)
sinh al eos n7T.
Q" = al - nni
(d) 00 = O, 0" = 1 - CO.S nn (n ,.: O).
n1<1
71.
72 p - 21p ~ IJnlJ p+ ¡ (IJ./) J ( )

• X - k< ( 212 _ 2)[J ( /»)2 p IJn X
" """ 1 j.J" P p j.J"
= 2/p i; '//
,, - I j.J" P
2 ~<IJ.~?-
p j.J"
where J'p(IJn/) = O, when p ,.: O.
When p = O, expansion is I = l.
76 . f(x) = ~Po(x) + íP, (x) + "T\PZ (x) + ....
77. fe'!') = iPo(eos '1') + tP¡(eos '1') + r\P2(eos '1') +
85. O (x < -a), -1</4 (x = -a) , -1</ 2 (-a < x < O), O (x = O), 1</ 2 (O < x < a),
1</ 4 (x = a), O (x > a).
86. (h) A(u) = B(u) = e- au .
87. e-U eos bx

f n
[ = o
~ 2
(02 _ b2
u(a 2
+ bu2)2+ +u 4a
-
2 2
) J'
2b 2 Sin ux du (O < x < = ),
e-U Sin
. = f~ [n
bx o 4 (aZ _ b2 +abu
U2) 2 + 40 2 b 2 J'
Sin UX
d It (O < x < = ).
99. A(u) = -11:2 f~ f(x) sin /IX dx.

o
Chapter 6
l. 3; G, -" -j).
2. e = (na + mb)/(m + n) .
6. (a) ±(-h " - ,J. (h) ±(ji + H- ik).
8. (a) j. (h) cos-¡ -?S.
9. i.
12. 45 °.
13. (a) ±(v'2/ 10)(4i - 3j - 5k). (h) 5v'2/2.
14. (a) ±(2i - j - k)/ v'6. (h) v'6/ 2.
15. v'Tw.
16. - 5.
20. el.
24. ± (i - k)/ v'2.
28 . F • dF x d1F.
dI --¡¡j1
29. (a) v = -wa sin wl + wbeoswJ ; r X v = w(a X b).

(h) a = -w 2 r.
Answers lo Problems 713
35. (e) u = (-ai sin 1 + aj eos 1 + c k)/ ../ a 2 + e',

n = - (i eos 1 + j sin 1),
b = (ci sin t - cj cos t +
ak)/ ../ a' + e 2 •
42. I/p = ../266/98, * =-¡\.

44. (a) ±(l, 1, 1)/../3; 2../3. (b)-l
45. grad r = i eos e + j sin e; grad e = - i (sin e)/r + j (eos e)/r.
46. (/, m, n) = (rpx, rpy, '1")/../'1'1 + '1'; + ",r.
47. ±(2i - j + 2k)/3.
48. div V = O, eurl V = O.
52. (a) 2/x + 2my + 2nz . (b) 2/xi + 2myj + 2nzk.
60 (a) No solution. (b) '" = x2 y + y2z + z + C.
61. (a) -T"-J· (b) 3.
62. (a) O. (b) O.
65. (a) 1../4n 2 + 1.
(b) 2n.
(e) 270.
7
67. ¡.
68. n/2.
69. 24.
70. O.
71. (a) ../3n/4.
(b) (2 + .../2)/3.
73. (a) 4"'.
(b) -n/ 2.
(e) n /2.
(d) O.
88. Both sides of Equation (133) reduce to n.
89. -n.
90. (a) n.
(b) !.
(e) O.
104. (a) u./r . (b) nrn-1u,. (e) u,/r. (d) O. (e) 3 eos (J. (f) O.
106. (a) u,/r. (b) u./(r sin '1'). (e) O. (d) O.

110. O.
111. 4n.
113. (b) V = 3(x 2 - y2)j - 6xyj; V = 3(x2 + y2).
(e) t¡I = 3x' y - y3; 3x2 y - y3 = eonstant.
(d) 14p .
117. (b) V = -2ru, + 4zk; V = 2../r 2 + 4z 2 •

(e) t¡I = 2r 2 z + c.
Chapler 7
1. (a) eos (J . (b) -r sin (J. (e) sin (J. (d) reos (J.
(e) eos (J. (r) sin (J. (g) -r-' sin (J. (h) r-' eos (J.
2 . (a) -eot (J . (b) -tan (J. (e) tan (J. (d) eot (J.
(e) r tan (J. (O r-' ca! (J. (g) -reot (J. (h) -r-' tan (J .
10. (a) iJ(f, g)"* O h

a(u , v) ,:
"* O.
(e) z.< = 4x' - 4 xy J - :!x, Zy = 6y' - 6X2 y2 - 3y2.
16. Yeso
17. Yeso
29. 1 + x + y -+ j;x2 + xy + :ty2 + ... .
30. 1 - x +y + x2 - 2xy + y2 + ... .
32. 0 .906 .
33. (a) Min at (O, O) ir b 2 < 1, saddle point at (O, O) ir b 2 > 1 ; min aloog line y = - x ir
b = 1; min along line y = x ir b = -l.
(b) Mio at (1, O).
(e) Re/. mal' at (O, O); saddle poin!s at (1, 1), (1, -1), (-J, 1), (-1, -1).
(d) Saddle point at (1, 1).
34. (a) Min at (O, O).
(b) Min at (1, -1).
(e) Mio along lines y = ±x.
(d) Max at (1, -1); min along eircle (x - 1)0 + (y + 1)2 = 2.
35. (a) Min at (O, O).
(b) Min along lines y = ±x.
(e) Saddle point at (O, O).
(d) Min at (O, O).
36. Max = 3 at (-1, O); min = -i al (j;, ±1./3).
37. Min at (1, 1), (-1, -1).
38. Max at (1, 1), (-J, -1); min at (1, -1), (-1, 1).
39. (a) Min = 12 when X2 = 4, yO = 1, z2 = ~.
(b) Min ce 4 wheo X2 = 2, y2 = i, z = O.
(e) Min = 1 at (1, O, O).
40. Furthest at (1, O, O); nearest at (j., j, O).
41. (a) (py')' + qy = f. (b) (Ay')' + (B' - C)y = E - D'.
42. (a) (Au',)' + (Bu:')' = O, (BII',)' + (Cuí)' = O.
(b) (a,.u',)' + (a'2uí), = b.IUI + b12U2,
(a. 211'.)' + (a22ui), = b.2"1 + b 22 u2.
43. (a) z = c.(J + C2. (b) (J = c)z + c.. z = e,/ + e" (J = c,l + ca.
(e)
[Here a geodesie is either a helix, a circular are, or a straight line. Part (a) ignores the
slraight lines (J = eonstant and par! (b) ignores the eireles z = eonstant, while parl (e)
ineludes 011 possibilities.]
44. (J = sin-' (el eot rp) + e2.
48. (a) Value or integral always i.
(b) Only eandidate (y = l / x) inadmissible.
50. n 2 , eorresponding to y = ../2 sin nnx with n = 1.
53. y = 1 + sin x - eos x, Z = l - sin x - eos x; 1= 4.
58. y'" = h, y(a) = y'(o) = y " (a) = O.
60. 2.0945515.
Answers lo Problems
715
61. 4.493.
62. 3.927.
63. l.3l.
64. x . 0.883, Y . 0.469.
65. x...:.. l.01941, y ...:.. 1.0306Z.
Chapter 8
1 (a) (x - y)(zx - zy) - Zz = O.

(b) bdz xx - (ad + be)2 xy + aezyy = O.
(e) b 2 z xx - 2abz xy + a 2 z n = O.
5. (a) 2 = eyjb + f(bx - ay) (b "* O).
(b) 2 = eexla f(bx - ay) (a "* O).
(e) 2 = f(x2 + y2).
(d) 2 = e-x' f(y - x).
(e) 2 = xf(xy).
(f) 2- 1 =x- I +f(x- I -y-I).
6. (b) z = f(2x + y) + X2 + X + ice Y •

11.2 = (x +y)2.
12. (a) 2 = [(b -e)x + (e - a)y)/ (b - a) (b "" a);
2 = X +
f(x - y), wheeef(O) = O ir a = b = e .
(b) 2 = e<Cy-x).'(b-a)(bx - ay)/ (b - a) (b "" a).
(e) 2 = ""(x 2 + y2)j Z.
(d) No solution.
(e) 2 = X.
(f) Z = x/[1 + xf(Y ;;, x)]. wheee feO) = O.
13. (a) 2 = [(bx - ay)j(b - 2a)j2 + e(y - Zx)j(b - 2a) (b "" Za ).

(b) 2 = e dy - 2x '!Cb-2·'[(bx - ay) / (b - 2a)J2.
(e) 2 = %(x2 + y2).
(d) 2 = (y - x)2eY(Y-2x,.
(e) z = (tx'y)li2.
(f) 2 = (x 2 y2)j(4x2 - 8xy + 4y2 - xy2 + 2x2 y).
14. (a) 2 = "" X + Y f(x - y) + 1.

(b) z = iX2 Y + f(y)·
(e) 2 2 = xy + f(yjx).
(d) z = e 3XY ' - Y 'f(x - y).
15. (b) No solutioo unless 9'(x) = x + a (a eonstant), in whieb case z = x + f(y - x),
wheee f is any differentiable funetion foe whieh f(b) = a.
23. (a) z = f(3x + y) + g(x - y).
(b) 9' = f(x + y + ¡x) + g(x + y - ix).
(e) 9' = f(x + y) + g(y).
(d) rw = f(r + el) + g(r - el) (e =1= O).
(e) 9' =f[x - (V+ U)I) + g[x + (V- U)t) (V=I=O).
(f) 9' =fl(x + y) +h(x - y) +h(x + iy) +f4(X - iy).
(g) 2 =fl(X + y) + xh(x + y) +h(x - y) + Xf4(X - y).
(h) 9' = f. (x + iy) + h(x - iy) + xh(x + iy) + xf4(X - iy).
27. (a) 2 =f(y + x) + g(y - x) + !;X'.
(b) 2 = f(y + x) + g(y + 2x) - i eos y.
29. (a) + ix2 ye.

z = f(y - x)
(b) + ¡y) + g(x - iy) + T[ZX 4 + ¡\x 3 y.
z = f(x
(e) + x) + g(y - x) + !x3(y + 1).
z = f(y
(d) + x) + xg(y + x) ¡'zx 4 + ¡\y3.
z = f(y
31. Z = f(y + x) + g(y - x) + sin xy.
32. q; = X2 + C 2 t 2 + c- cos x sin el. l
37. (b) rp = X2 + (e 2 - iJ/' + l.

44. rp = xm yneAx+BY[f(x + iy) + g(x - iy)].
54. (1, = 2, /1 = 1, Y = 2.
55. (b) h = 1: '" = f(x + el) + g(x - el) + 1(X2 - e 2 /2).
h = cos (x + el): '" = f(x + el) + g(x - el) + 1(x - el) sin (x + el).
Chapter 9
14. (a) T = 400 ~ sinh (mry/100l sin mrx.

re lIodd n slnh n7t 100
(b) Exact value is 2SO.
~ [ . n1<y . n1«d - y)] . n1<x
15. T = ~ On slnh -,- + b n slnh , Sin - , - ,
n==1
w h ere 'hmrd
al! sm -,- = T2S'f()·n1<x
o x SIn -,- d x
an d · h-
b "Sin n1<d
,- = T2 S'og () . n1<X d x.
x sln-,-
22. T = (1,[ + «(1,2 - (1,1) 1 + /1IY + (/12 - f3¡)X( + ~

n- I
onrn,,!' sin n~x,
where G n = 7 S~f(X) sin n~x dx - },,«(1,1 - (1,2 cos /11<).
23. T -- Tx j -'- ~
(T o ,' c (y) ~
[ ansln
. h -,-
n"y I + . h n,,(d / - Yl] sln-,-'
b "Sin . n1<X
n=\
. h n1<d
w h ere o" Sin -,- = T2 J"o f( x ) Sin
. n1<x d +
-,- x
2(T o +n1<e) cos liT<
and b
"SIn
. h n1<d
-,- = 2To n7t
cos mr
.
24. T = sin
- ro' -x - sin rox
- + ~~ '1
a e- mry , SIn - - ,
. n1<x
w2 I W2
n- [
1) I
h
w ere Un -
_ 2mr cos n1< sin col (1
ro2
+ 1
ro2/2 _ n 2 rc 2
) if co * kr (k = 1, 2, ... ).
T = - (1 - e- roy ) sinco cox 'f

1 ro = k1<
-.
2 ,
25. AX" + DX' + (F[ + AG)X = O, CY" + EY' + (F2 - AH) Y = O;

AX" + (DI + AG)X' + FIX = O, Y' + (D 2 - AH)Y = O;
X' + (El - AG)X = O, CY" + (E2 + AH) Y' + F 2 Y = O.
27. T = ,
1..(r2 - 100) sin e + 21, (400 - ,2) cos e.
1
+ -7t2
n
28. (a) T = -2 ~ -r
nodd n
sin ne (, <: 1).
(b) T
1
= "2 ,. -
I 2
~
,-n.Sin ne (, >-
- 1).
1t 11 odd n
29. (a) T = I;
n= 1
An (.!...a r nl
" sin nn;O , where An
(t
= 1..
ex
S"o feO) sin nn;O dO.
Gt
36. (a) T = To (1 - ~ eos </,).
(b) T = T o [ "2
1
+ 43 tir p ¡ (cos rp) -
7(r))
16 ti p) (cos </,) + .. -}
~ ~
Amn . mnx . nny
46. (e) </' = ~ ~ A
n;Z(m2 / 1f + n2 //'D Sin --¡;- SIO"T;'
17'1 _ L n= 1
ir A"'" 1[2(pl/ ,y + q'/'i), where p and q are integral, with Amn as defined in Problem
45(b).
~ ~ b . rnn;x . mey .
53 • (b) w = ..... ..... mn 510 - , - 510 - / - 510 Wmnl,
'2 J'o SIo

m:o. 1 n=- 1
4
where b mn = w mn v(x, y)
. m1tx
510 - , - SIO - , -
. nny
dx dy.
SS. T = 100 - x - 200 ~ .!... sin ~e-m'n'""/2'oo.

n m~ ¡ m 50
56. (a) TT , P = b" cos n~x e- nJ -:f.1«l r/ P (n = O. 1,2, .. . ).
(b) T = 50 - 40.? L -; cos ~ e- n 'ln ' «'I/lO'.

1[- ti odd n 100
64. ( a ) T = T O [ TX sIn
. 2AW ~ ( - I )nn ( AW . ' 1') . nn;x]
ro! + ---¡r- lJ~l n4 + 1 2m2 cos rol + n 2 sm ro/ - e-n t sin -,- .
77. rp =
W
cos - ( x - el).
e
93. z = {eeeX (1 <

(1 ;z:
x),
x).
94. z = e'.
~)
(1 < ;),
95.
rp = {:OSW(I - (1 > ;).
_{1+1-'¡'12 (t < x),
96. rp - 1 + x - Ix + !X2 (t >- x).
98. z = e'rp(x - 1).
111. Cn(y) = 2(1 -n~OS nn)[C~:2 + 1)e-nnYII - n~:2l
112. rp = 1 + n~¡ Cn(y) sin n~x, where
e (Y ) =
11
2(1 -
n7t
cos nn)[( '2 - l)e-nnYil -
\Ji21l2
~J.
n21[2
113. (e) Cn(y) = h n S~

o G(y,
. '1) d'1 + a"e-nnYII, where
hn 2
= On = n1t
-(1 - cos me), f~ G(y, '1) d'1
o
= - '2
22(1
n 1t
120. V;; = 1 + ~ (dx - (Vs + U)ll + f[X + (Vs - U)llj,
fa = 1 + ~(f[X - (Vs + U)lldx + (Vs - U)llj.

718 Answers fo Problems
Chapter 10
5. (a) -2 + 2i. (b) O + i. (c) O + i.

. h'
h ' -'- ~.
()
e
~
-./2 ' -./2 l.
7. sin z, cos Z, sec z, and ese z have period 2n; tan z and cot z have period 7t; sinh z. cosh z
sech z, and csch z have period 2ni; tanh z and coth z have period ni. J
13. (a) log ./2: + i( ~ + 2k7<), where k is any integer;

priocipal value is log -./2: + (7<i/4).
(b) cos 3 7< (1 + 4k) + i sin 3 " (1 + 41<), where k ~ O, 1, 2, 3;
8 8
. . l l· 37< . . 37<
prmclpa va ue IS COS "8 + 1 sIn "8'
(c) ,vZ[cos ~ (1 -i- 8k) + i sin ~ (1 + 8k) J, where k = O, 1;
principal value is ,v2 (cos ~ + i sin ~).

14. z = ~' 2(cos" + isin,,), where" = 37< / 8, 57</ 8, 11,,/ 8, 13,,/ 8.
15. zn = r"[co s 1«Op + 2k7<) + isin 7«Op + 2k11)], where z = ré 8 ; cos (7<' /4) + isin (7<1 /4).
16. f(i ) = r ( n / 2+2h ) (k = O, ± I, ± 2, ... ); g(i) = e-n ' l.
18. (a) (~ + 2k7<) ± i log (2 + -./3), where k is any integer.
(b) (2k + 1) ~ + i log -./3, where k is any ioteger.
23. (a) 3 xy o - x + C.
3 (c) Jt is no! an aoal y tic function.
24. (b) fez) = iZ/ (Z' + 1).
(b) Replace iy by z in f(iy) = u(O , y) + iv(O , y) when fez) IS analytic 00 part of the
imaginar)' axis.
27. (a) v = r- l cos 20 .
(b) " = - r 3 cos 30 .
28. fez) = U(z, O) + ;V(z, O) .
29. fez) must be a branch of log z + iz o
33. (a) O.
(b) 21<i.
35. (a) 27<;.
(b) 21<;.
36. 21<i; 27<a ' i.
37. O; O.
50. (a) l. (b) í. (c) l. (d) -1.
51. (a) I + z + zl + ... + zn + . ...
I I I I
(b) --Z - Z2 - Z3 - .•• - zn -
(c) -l
2
+ z +
22
I + (z +23 1)2 + ... + (z +
2-+ I
1)" +
I 2 2 2-- 1
(d) - z + I - (x + 1)2 - (z +' 1)3 - ... (z + 1)"
I
52. (a) -
z
+ I+ z + Z2 + ... + z" . >-
I 1 1
(b) - Z 2 - Z 3 - Z 4 - ••• - ZM -
I
(e) - - - 1
z-
+ I - (z - 1) + (2 - 1)2 - . .. + (-1 )"(z - 1)" + ....
(d) (z - 1 1) 2 + (z - l 1)' (z - 11)4+ ... + (-I)"+'(Z~1)"+""
(e) -~ - i(z + 1) - ¡(z + 1)2 - - (1 - 2-"-')(z + 1)"-
( f) ...
+ (z+l l)n + ... + (z
l
·_ 1)2
+ z+11 + 2'~
l , z + l
..l. (z + 1) 2 ..l. + (z + 1)n +
, 23 ' 2 n +)
1 3 7 2n -' - 1
(g) - (z + 1)2 (z + 1)' (z + 1)4 (z + 1)"
53. 2 - z + 2z 3+ 2z 6 - Z7 + ...
- Z4 (1 z I < 1),
l
zZ -
2+..!...._2
Z'"J z!i Z 6
+ ....!.._2+
Z 8 Z 5'
(1 z I > 1).
54. 1 < I z - 1 I < ../,.

l l 7
55. z +6 z + 360 z ' +
61. (a) Simple poles al z ~ ±i.
(b) Simple poles al z ~ -1, z ~ i{1 ± i ../3).
(e) Braneh points al z ~ ±i.
(d) Braneh points al z ~ 1, z = 2.
(e) Simple poles al z ~ (2k + 1),,/2, where k is any inleger.
(f) Braneh points at z = 1 ± i.
65. (a) Inside the circle I z - 1 I = ../3.

(e) Launeh anolher series expansion from (say) z = 1 + 1.7i.
67. (a) Ixl < 1. (b) Ixl < 1. (e) Ixl > O.
(d) Ixl <: ". (e) lxl <: .J2 . (f) I x l < ,, / 2.
71. (a) Pole of arder two . (b) Regular poin!. (e) Regular poin!.
(d) Braneh poio!. (e) S imple poleo (f) Braneh poio!.
78. (a) Res ( ± aj) ~ ±e±a'/2ai.
(b) Res (±a) = ± 1/4a' , Res (±ai) = ±i/4a J •
(e) Res (O) = l.
(d) Res (O) = O.
(e) Res (O) = 1, Res (1) = O.
(f) Res ( ± ai) = ±1/4a'i.
(g) Res (2) = je 2a , Res (í) = -~eaI 2.
(h) Res (-1) = ±(e- 2 - 1) .
(i) Res (O) = - as /(S!).
(j) Res (2hi) = 2 (k = O, ±1, ±2, ... ).
(k) Res (h) = 1 (k = O, ±1, ±2 , ... ).
(1) Res (O) = - z'•.
80. (a) -,,/2.
(b) +,,/2.
81. (a) -ieail"! z sioh (a/../?').
(b) eaNZ sio (a/../?').
82. 2"i[Res (1) - Res (-1») ~ 2"i.
86. -2"i.
720 Answers lo Problerns
88. (a) Res (O) = l.

(b) Res (O) = O.
(e) Res (n) = O.
(d) Res (O) = -Res (z~) = t.
93. (b) (n/l28a 7 )e- am [(3 + 2am) sin am + 3 eos am] (a> O, m ~ O).
96. f(l) = (J / 2a)e- a1rl (a> O).
110. (a) n(l - e- a)/a 2 (a > O).
(b) 2/n.
115. (b) p'¡: fez) dz

re = ",i Res (ao) +
.2n¡ ±
k - l
Res (ak).
Chapler 11
l. (a) e-o'. (b) J..

a
sin al. (e) J..
a
sinh al.
(d) 1- 1 ,
6 (e) ia 1 sin al. (f) 2;2 sin al sinh al.
I,,-le- al
2. (a) (n _ l)!
(b) (n - 1) - al ln - 2 e-"' (n > 1) .
(n l) !
(e) (l/a)c b ' sin al .
(d) (l / a) e-·'(a eos al - b sin al) .
(e) (l / 2a')e- b '(sin al - al eos al).
(f) (1/2a')(al eosh al - sinh al).
73. Right half plane:

1 1z - ' 1
G = 2" lag Iz + 'l '
Third quadrant: G same as for tirst quadranl.
86. G = (1 / 2n) lag R + constan!.
Index
Ilalicized figures in parenlheses refer lo problem numbers.
Approximate solution of difI"erential

A
equations:
Abel's formula, 28, 29, 40 (8), 45 (38), by graphical methods, 93
46 (45) by numerical methods, 93
Absolute value, 540 Arc length. differential or, 279, 308
Acceleration, 279 on surface, 376 (26)
normal, 280 Area, differential of, 308. 353, 376 (25)
in polar coordina tes, 325 (34) Argument, of complex number, 541
in spherical coordinates, 338 (108) Argument principIe, 608 (76)
tangential, 280 Ascending differences , 111
Acoustics, 440 Associated Legendre functions, 161,
Adams' method, 96 512 (47)
modified, 100 Associative law, 270
Adjoint operator, 661, 697 (83) Asymptotic approximation:
Admissible function, 360 for Bessel functions, 148
A1gebra, fundamental theorem of, 608 (74) for Gamma function, 80
Alternating series, 119 Asymptotic expansion, 166
Amplitude, of compIex number, 541 of Bessel functions, 166, 185 (79)
Analytic continuation, 563, 606 (65) Average value, 215
Analytic function, 550 Axial load, 195
residues of, 580, 609 (85)
singularities of, 567,575 B
Angle, between vectors; 272 Backward differences, J07
Angular velocity vector, 274, 302 Base curve, characteristic, 403,412
721
722 Inde"
Beam: Boundary-value prob!em:

bending of, 192 for ordinary difl'erential equation, 31,
buckling of. 195 186
rotating, 19~ nonhomogeneous, 211
vibrating, 245 (12) ror partia! differential equation, 491
Bending moment, 192, 195, 327 (4]) Branch, 546, 567
Bending s t iffness, 328 (43) Branch cut, 546, 568
ber and bei funclions , 153 Branch poim, 131, 546, 595
asymptotic approximations foro Bromwich integral, 624
179 (53) Bromwich line. 624
deri va tive formula s for, 179 (52), Buckling, of long column, 195, 246 (J 7)
180 (54) Buckling mode , 197
polar notation for , 153
Bernoulli ' s dilTerential equation, 35 , e
41 (15) CaJcu!us of variations, 360
Bernoulli's equalion , 321,682 (29) Capacity, of condenser, 690 (58)
Bessel functions. 141 Cauchy principal value, 592
asymptotic approximation s for, 148 Cauchy problem, 399,430 (37), 467, 490
asymptotic expansions of, 166, Cauchy-Riemann equations, 552, 556,
185 (79) 600 (25, 26)
derivative formulas for, 148 Cauchy's differential equation (see
differential equations satisñed by, Equidimensional different,al
151 equation)
expansions in, 226, 461 Cauchy's inequality, 603 (46)
of ñrst kind, 142 Cauchy's integral formula, 560
generating function for, 177 (42) Cauchy's integral theorem, 555
graphs of, 151 Cauchy's residue theorem, 583
integral formulas for, 176 (40), Central differences , 111
177 (41), 257 (65) Central force, 324 (30)
integrals of, 176 (]8, 39) Chain rules, 342
modiñed, 147 Characterislic base curve, 403, 412
of order 11 + 1, 70, 149 Characteristic curve, 388, 403, 408, 412
recurrence formulas for, 149, 177 Characteristic cylinder, 403 , 412
(43) Characteristic equalion, 8
of second kind , 144 Characteristic functions, 187
of third kind , 146 expansions in series of, 207
Wronskian of, 177 (44) multidimensional, 510 (44)
zeros of, 150, 175 (35), 176 (36, 37), onhogonality of, 205
181 (61) Characteristic numbers, 187,510 (44)
table, 230 positivity of, 249 (29)
Beta function, 91 (56) reality of, 249 (28)
Bi-Laplacian equation, 333 (86), 440, Characteristic strip, 411
499 (4) Characteristics, 414,419
Bilinear transformation , 678 (22) Charge density, 652, 690 (57)
Binormal vector, 280 Chebyshev pOlynomials, 140, 174 (26)
Bipolar coordinates, 599 (20, 21), of second kind, 174 (26)
605 (58, 59), 640 CircJe of convergence, 542, 562
Body force, in fluid ftow , 341 (J 18) Circular cylindrical coordina tes, 307, 311
Jode" 723
Circular disk , 305 n Coordinates (cont .):

Circular functions, 542 elliptical cylindrical, 335 (96, 97)
Circulation, 290, 514 (49).633 , 636 n parabolic cylindrical, 336 (98)
Circulatory flow , 636, 640 paraboloidal, 336 (99)
CIosed complex plane , 575 spherical, 312
COlumn, bending of, 195,246 ( 18) Cosine-integral fun c tion , 30, 83 (16) ,
Commutative law, 16, 270,272,273 613 (99), 674 (/1)
Compatibility, equation of, 411 Cramer's rule , 19
Complementary error function. 534 (108) Critical length, of column, 246 (18)
Complementary solution, 6, 396 Critical load , for columo, 196
Complete solution, 4 Critical point, of mapping, 631
Completeness, 208 Critical speed:
Complex number, 539 of rotating shaft, 195,2[2
as number pair, 597 (1) of rotating string, [90, 2 [2
Complex plane, 539 Cross product, 273
closed, 575 Crosscut, 557
extended, 575 Cur[, 283, 302, 306
Complex potential, 632 in curvi[inear coordina tes, 311
Complex variable, functions of, 539 Curl theorem, 332 (79)
Complex velocity, 632 Curvature, 195, 280 , 326 (37)
Components, of vector, 270 of beam, 193
Compressible fluid, flow of, 316,416, radius of, 280, 326 (37)
470,493,495 Curves, geometry of, 278, 326 (36-39)
Condenser, capacity of, 690 (58) Curvilinear coordina tes, 306, 352,
Conductivity, thermal, 441 681 (28)
Conductor, cylindrical, 650, 690 (57-59), arc length, 308
691 (60-62) curl, 311
circular, 690 (59), 691 (60) del , 309
ellíptícal, 691 (61,62) divergence, 310
Confluent hypergeometric function, 140, gradient, 309
165,171 (14) Laplacian operator, 310
Con formal mapping, 628, 658 nonorthogonal, 352
Conjugate complex numbers , 540 orthogonal, 306, 681 (28)
Conjugate flows, 315 surface area element, 308, 376 (25)
Conservative force, 291 volume elernent, 308, 352
Constraints, 357, 364 Cut, 546, 568
Continuity, equation of, 318
Contour, 555 D
Convergen ce, of series, 118, 208, 541,
562 D' Alembert's so[ution, 400, 467, 669
Convolution , 63, 263 (91), 264 (92), Damping, 88 (44), 521 (73)
531 (100) Dashpot, 88 (44)
Coordina tes: De Moivre's theorem, 543
bipolar, 599 (20, 21), 605 (58, 59), 640 Definite integrals, evaluation of:
circular cylindrical, 311 by approximate methods, 110,
curvilinear, 306, 352, 681 (28) 11 5 (20), 116 (25, 27, 29)
nonorthogonal , 352 by calculus of residues, 583
orthogonal, 306 by Lap[ace transforms, 82 (14)
724 Index
Defiection: Differential equation (cont.) :

of beam, 192 quasi-linear, 362, 384, 387, 396, 417
of membrane, 459 Riccati, 52 (66)
of string, 188 separable, 31, 505 (25)
Defiection mode, 190, 195 series solutions of, 118
Del, 283 Differential surface area vector, 295,
in curvilinear coordinates, 309, 308, 376 (25)
681 (28) Differentiation:
Delta function, 66, 653 of F ourier series., 223
Dependence: of integral containing parameter, 364
functional, 350 n umerical, 110
linear, 3 of series of orthogonal functions, 209
Derivative: of Taylor series, 121,562
of complex function, 550 of vector function, 277, 284
of vector function, 277, 284 Dilfusion, 440
Descending dilferences, 111 Diffusivity, 442
Dilferences: Digamma function , 91 (55)
ascending, 111 Dipole function, 66
backward , 107 Dirac delta fuoction, 66, 653
central, 111 Direction cosines, 271
descending, 111 Direction ratios, 271
forward, 97, \07 Directional derivative, 282
Dilferential distance vector, 287, 376 (26) Dirichlet problem, 443, 652
Dilferential equation: Discontinuity, finite, 55
Bernoulli, 35, 41 (15) Distributions, theory of, 66
Bessel (see Bessel functions) Distributive law, 16, 272,273,285
complementary solution of, 6, 396 Divergence, 283, 297
complete solution of, 4 in curvilinear coordinates, 310
confluent hypergeometric, 140, 165, theorem, 297, 331 (76)
171 (14) two-dimensional, 301, 305, 332 (78)
equidimensional, 12 Division, of series, 121
exact, 32 Domain operator, 661 n
Hermite, 142, 173 (25) Dot product, 271
homogeneous, 5, 12,34 n, 186 Double Fourier series, 455
homogeoeous solution of, 6 Doublet, 66, 504 (19), 636, 638
of homogeneous type, 34, 35 orientation of, 636
hypergeometric, 162 strength of, 636
Jacobi, 140, 174 (26), 183 (68) Doub1et function, 66
Laguerre, 140, 173 (24) Drag force , 682 (29), 684 (36)
Legendre (see Legendre functions) Duhamel's principIe, 465, 531 (lOO)
linear, 2, 6, 8, 392, 396 Duplication formula, for Gamma
nonlinear, 2, 31 function, 92 (61)
numerical solution of, 93
order of, 2 E
ordinary, 2
partial, 2, 384 Eigenfunction (see Characteristic
c1assification of, 396 function)
normal form of, 413 Eigenvalue (see Characteristic Dumber)
Index 72S
Elasric end support, ¡ 93 Extremal, 362

Electric field intensity. 652, 690 (57) Extrapolation, 107
Electromagnetic theory , 440
Electrostatic field, 651, 690 (57) F
Electrostatic potential, 439, 652, 690 (57) Factorial function (see Gamma function)
Elementary functions, 541 Factorization, of operarors, 16
Elliptie functions, 645 Faltung (see Convolurion)
Elliptic integrals, 190,243 (9) Family, of a function, lO
ElIiptic partial differential equation, Field intensity vector, 652, 690 (57)
396, '412, 417, 490 Fixed end support, 193
normal form of, 414 Fluid flow, 313, 632
Elliptical cylindrical coordinares, from channel, 649
335 (96, 97) compressible, 316
Entire funcrion, 579 about circular c y linder, 513 (48),
Envelope, of set of curves, 32 514 (49, 50), 637, 682 (30)
Equations of motion, for fluids, 317 about elliptie cylinder, 686 (41)
Equidimensional differential equation, about plate, 647, 686 (42)
12 about sphere, 456
Equipotential line, 314 supersonie, 319, 495
Equipotential surface, 292 two-dimensional, 313, 632
Equivalent contours , 560 Fluid pressure, 316, 682 (29)
Error function, 476 , 673 (JO), 674 (13) Flux, 300, 501 (JO), 633
complemenrary, 534 (J08) Force, 287 , 291
modified, 677 (18) conservative, 291
Essential singularity, 572 Force poten tial, 291
nonisolated, 574 Forced mode , of vibration, 73
Euler equatíon, 361 Forced vibration, 72
first integral of, 362, 379 (45) Formally self-adjoint operator, 661 n,
tran s formation 0[, 379 (45) 695 (79)
Euler load, of column, 197 Forward differences, 97
Euler' s constant, 91 (55), 144 Fourier conditions, for Newton 's
Euler's differential equation (see method, 368
Eq uidimensional differen tial Fourier cosine integral, 237
equation) Fourier eosine series, 214
Euler's equations, of fluid motion, 317 Fourier integral, 234
Euler's formula, 10, 543 applications of, 473
Even function, 215 complex form of, 238
Exact differenrial equation, 32 Fourier series, 214, 219
Expansion: complex form of, 256 (62)
asymptotic, 166 differen tia tion of, 223
in series of characteristic functions, double, 455
207 integration of, 258 (70)
Exponent, of differential equation, 130 Fourier sine integral, 236
Exponential function, 542 Fourier sine series, 214
generalized, 548 double, 455
Exponential order, function of, 56 Fourier transforms, 238, 265 (97)
Exponential-integral function, 30, 83 (16) applications of, 477,523 (79,80),
Extended complex plane, 575 524 (81)
726 Iodex
Fourier transforms (cont.) Functions (cont.):

convolution of, 263 (91), 264 (92) logari tlunic, 544
of delta function, 263 (90) modified error, 677 (18)
of derivatives, 263 (89) multivalued, 546
Fourier-Bessel integral, 240 odd, 215
Fourier-Bessel series, 226 potential, 291, 302
Fourier-Bessel transforms, 240 power, 541, 546
Free end condition, 189, 193 rational, 122, 541, 571, 576, 579
Free vibration: regular, 122
of beam, 245 (J 2) sine-integral, 30, 83 (16), 613 (99),
of membrane, 459 674 (11)
of string, 243 (8) single-valued, 289, 546
Frenet formulas, 281 singularity, 65
Frequency, natural, 73, 243 (8), square-wave, 82 (/0)
245 (J 2), 461 stream, 315, 50J (10) , 632
Fresnel integrals, 615 (109) triangular-wave, 82 (/1)
Frobenius, method of, 128, 170 (10) unit impulse, 65
convergence, 131 unit step, 86 (32), 668
exceptional cases, 130, 134 Fundamental set, of solutions, 47 (48)
for nonhomogeneous equations, Fundamental theorem, of algebra,
172 (18) 608 (74)
Functional dependence, 350
Functions: G
analytic, 550 Gamma function, 76, 618 (J 19)
Bessel (see Bessel functions) asymptotic approximation for, 80
Beta, 91 (56) duplication formula for, 92 (61)
circular, 542 incomplete , 675 (14)
complementaryerror, 534 (108) table of, 78
conftuent hypergeometric, 140, 165, Gauss's interpolation formula, 116 (24)
171 (14) Gauss's test, for infinite series, 1 J 9 n
cosine-integral, 30, 83 (16), 613 (99), Gauss's theorem , 299
674 (11) Generalized factorial function (see
delta, 66, 653 Gamma function)
Digamma, 91 (55) Generating function:
doublet, 66, 527 (89) for J.(x), 177 (42)
elementary, 541 for P.(x), 182 (63)
error, 476,534 (108),673 (10), 674 (13) Geodesics , 379 (43)
even, 215 Geometric series, 162, 561, 566
exponential , 542, 548 Goursat's theorem, 551
exponential-integral, 30, 83 (16) Gradient, 281
Gamma, 76 in curvilinear coordinates, 309
harmonic conjuga te, 554, 685 (38) Gradient theorem, 332 (79)
homogeneous, 34 GraphicaJ integration, of differential
hyperbolic, 41 (16), 543 equations, 93
hypergeometric, 162 Grav itational potential, 439, 658
incomplete Gamma, 675 (14) Green's functions:
Laguerre, 81 (6), 140,173 (24) one-dimensional, 49 (57), 251 (39),
Legendre, 155 692 (65, 66), 695 (79), 696 (80-82)
lndex 727
Green functions (con/.): Homogeneous solution, 6

three dimensional, 693 (67) Hyperbolic functions, 41 (16),543
two-dimensional, 652 Hyperbolic partial differential equation,
Green's theorem, 301,652,661,697 (83) 396, 41 2, 417, 490
first form, 30 I normal form of, 413
generalized, 333 (84) Hypergeometric function, 162
in the plane, 305 confluent, 140, 165, 171 (/4)
second form, 301 Hypergeometric series, 162
Gregory's backward difference formula,
109 1
Grounded line, 481 Jmages, method of, 656 n, 694 (73),
702 (100)
H lmaginary number, 539
Hankel functions, 146,469,664 Imaginary part, of complex quanity, 539
Hankel transforms, 240 lmaginary unit, 539
Harmonic conjuga te, 554, 685 (38) Implicit functions, 347
Harmonic function, 303, 554 Improper integral, principal value of,
Harmonics, 215 592
Heat flow, 441 Impulse function, 65, 73
in column, 509 (42, 43) Jncomplete Gamma function, 675 (14)
in cylinder, 508 (]9, 40), 509 (41) Incompressible fluid, 298, 313
and orthogonal coordinates, 501 (/0) lndented con tour, 592
in plate, 443, 446, 473 Independence:
in rectangular parallelepiped, 453 functional, 350
in rod, 461,464,475 linear, 3
with radiation, 500 (9) Indicial admittance, 531 (100)
in sphere, 451,511 (47) Indicial eguation, 129
Heat-flowequation, 430 (40), 431 (41), lnfinite series (see Series)
440,442,461,464,475,494, Infinity:
502 (I2), 669 point at, 575
nonhomogeneous, 485, 536 (115), 669 residue at, 609 (85)
Heaviside unit step function, 86 (J2), 668 lnfluence function, 657
Helix, 326 (J5), 327 (40) lnitial conditions, 30, 384, 396, 401
Helmholtz equation, 402, 430 (38, 39), lnitial-value problem, 30, 392, 399, 404,
456, 511 (46), 520 (69), 664 408,417,490
modified, 499 (1), 662 lnner product, 271
Helmholtz operator, 664 Insulated boundary, 504 (20), 5I6 (56)
modified, 662 Integral:
Hermite polynomials, 140, 173 (25) elliptic, 190, 243 (9)
Hilbert's function, 666, 699 (91), improper, 592
700 (92-94) principal value of, 592
one-dimensional, 701 (96) line, 287, 554
symmetry of, 699 (91) surface, 294, 296 n
Hinged end support, 193 Integral equation, 85 (29), 654
Homogeneous condition, 186 Integral function, 579
Homogeneous differen tial equation, 5, Integral power function, 541
12, 34 n, 186 Integral surface, 384
Homogeneous function, 34 Integral transcendental function, 579
728 Index
lnlegraling faclor , 2,7,35 , 51 (63) Laplace's equalion , 303, 439, 442, 473,
Imegralion, numerical , 110, 115 UO), 491 , 501 (l1), 553
116 (25, 2 7, 29) polynomial solution s of, 426 (26),
Inlerior poinl, 551 n 504 (21)
Inlermediate variable, 343 Laplacian operator, 286, 652, 665
Inlerpolation, 109, 116 (24) in curvilinear coordina les, 310,
lnterval 01' convergence, 119 608 (lOO)
Inverse circular functions, 549 Laurent series , 563
Jnverse hyperbolic functions, 550 differentiation of, 565
Inverse Laplace transform , 62, 622 uniqueness of, 566, 604 (56)
Inverse tangenl, addition formula for, Least squares, method of, 255 (56),
646 n 260 (78)
Inversion, of complex plane, 678 (21) Legendre functions, 155
Irregular singular poin!. of differenlial associated, 161
equation, 127 recurrence formula for, 158
lrrotational f1ow, 302 of second kind, 158
Isoclines, 93 Legendre polynomials, 157
Jsolated singular poin!, 570 generating function for, 182 (63)
lteration, methods of: graphs of, 160
Newton-Raphson method, 367 recurrence formula for, 158, 182 (64)
Picard's method, 105 zeros of, 160
in solving transcendental equations, Legendre series, 230, 452, 511 (47)
191 n Leibnitz's rule, 365
Stodola-Vianello method, 197,212 Lerch's theorem, 62
L'Hospital's rule, 604 (49)
J
Lift force, 682 (29), 684 (36)
Jacobi polynomials, 140, 174 (26), Limiting contours, 589
183 (69),184 (70,71) Line integral, 287, 554
Jacobian determinants, 347,350,352, of a complex function, 554
370, 628 independent of path, 290, 306, 555
Jacobi's equation, 140, 174 (26), 183 (68) Linear combination, 3
Jordan's lemma, 590 n nontrivial, 3
K Linear dependence, 3
ker and kei functions, 154 Linear differential equations (see
asymptotic approximations for, Differential equations)
180 (53) Linear fractional transformation,
derivative formulas for, 180 (54) 678 (22)
Kutta (see Runge-Kutta methods) Lines of f1ow, 314
Lines of force, 652
L Liouville's theorem, 578, 608 (77)
Lagrange, identity of, 277 Logarithm, complex, 544
Lagrange mullipliers, 357, 364 Long line, equations for, 480
Laguerre polynomials, 81 (6), 140, Loop integral, 626, 628
173 (24.)
associated, 140 M
Laplace transforms, 53, 477,480 ML inequality, 560
inversion of, by contour integration, Mach number, 319,496
622 Mac1aurin series, 122 n
table of, 67 Mapping, conformal, 628, 658
Index
729
Maxima and minima: Newton's method, 367

absoJute, 356 convergence of, 368
of functions, 356 Fourier conditions, 368
of integrals, 360 error in, 368
relati ve, 356 for simul taneous equations, 369
Maximum modulus theorem, 603 (45) Nodal Iines, of membrane, 515 (54)
Maxwell's equations, 334 (93) Nonhomogeneous conditions, 484
Mean value (see Average value) Norm, of function, 206
Mean-value theorem, 603 (44)' Normal acceleration, 280
Membrane: Normal derivative, 301
deflection of, 459 Normal form, of partial differential
vibration of, 459 equation, 413
Minimax, 356 Normal plane, 281
Mixed boundary conditions, 205 n, Normal vector, 280, 294
248 (26) Normalization, of characteristic function,
Modified Bessel functions, 147 206
Modified error function, 677 (l8) Normalizing factor, 206
Modified Helmholtz operator, 662 Numerical methods:
Modulus, of complex number, 540 of differentiation, 110
Moment: of integration, 110, 115 (20),
bending, 192, 195, 327 (43) 116 (25, 27, 29)
of momentum vector, 324 (30) of interpolation, 109, 116 (24)
vector, 274 of solving ordinary differential
Momentum vector, 324 (30) equations, 93
Multiple products, of vectors, 275, characteristic-value problems, 197
324 (27) of solving nonlinear equations,
Multiplication, of series, 121 191 n, 367
Multivalued function, 546
N o
Natural boundary, 574, 606 (66) Odd function, 215
Natural frequencies: Odd-harmonic function, 223, 255 (59)
of vibrating beam, 245 (l2) Open line, 481
of vibrating membrane, 461,515 (54) Operator, linear differential , 4, 15,
of vibrating string, 243 (8) 346, 652
Natural mode , of vibration, 73, 243 (8), Operational methods, 17, 42 (24, 25),
245 (l2), 515 (54) 43 (26, 27), 427 (30)
Neumann problem, 443 , 456, 665 Order:
Neumann's form, of a Bessel function, of differential equation, 2
144 of pole, 571
Neumann's function, 665, 699 (88-90) Ordinary differential equation, 2
one-dimensional, 700 (95) Ordinary point, of differential equation,
symmetry of, 699 (88) 126, 164
Newton-Gauss formula, 116 (24) Orthogonal coordinates, 306
Newton-Raphson method, 367 and fluid flow, 313
Newton's backward-difference formula, and heat flow, 501 (lO)
109 Orthogonal functions, 203
Newton's law, of cooling, 491,516 (57), expansions in series of, 207
517(58) Orthogonal set, of functions, 203
730 lndex
Orthonormal set, of function; , 206 Polynomials (con!.):

Osculating plane, 28 I Jacobi, 140, 174 (26), 183 (69),
Outer product, 273 184 (70, 71)
Laguerre, 81 (6), 140, 173 (24)
p Legendre, 157
Parabolic cylindrical coordinates, Position vector, 278
336 (98) deriva ti ve of, 279
Parabolic partial differential equation, Potential, 291, 302
396,412,417,491 complex, 632
normal form of, 414 electrostatic, 439, 652, 690 (57)
Parabolic rule, 117 (29) existence of, 291
Paraboloidal coordinates, 336 (99) gravitational, 439, 658
Parseval equalities, 256 (61) velocity, 313,439, 632, 658
Panial deri va ti ve, 342 velocity deviation, 319
Panial differential equation, 384,439 Potential energy, 291
classification of, 396 Power function:
normal form of, 413 generalized, 546
Partial fractions, 71, 83 (! 7) integral, 541
Panial integration, 33 n Power series, 118, 541, 561
Panicular solution, 6, 396 convergence of, 118, 541, 562
Periodic extension, 22 I opera tions wi t h, 121 , 542
Periodic function, Laplace transform of, solutions of differential equations, 118
82 (9), 87 (37) Pressure, fluid, 316, 682 (29)
Picard's method, 105 Principal normal vector, 280
Picard's theorem, 574 Principal value:
Piecewise continuity, 55 of improper integral, 592
Piecewise differentiability, 208 of logarithm, 545
Piecewise smooth curve, 287 of power function, 569
Piecewise smooth surface, 295 Proper strip, 4[0
Plane, equation of, 322 (JO) Pulsating cylinder, 470
Plasticity, 420 Pulsating sphere :
Point at infinity, 575 steady-state solution , 521 (75)
Poisson's equation, 439, 510 (45), transient sOlution, 530 (99)
534 (111), 535 (J 12), 536 (J 14), Pure imaginary number, 539
653, 660, 665
Poisson's integral , 450, 660 Q
Pole, 131, 571 Quasi-linear equations, 362, 384, 387,
order of, 571 396, 417
residue at, 580 characteristic curves of, 388
simple, 571 c1assification of, 396
Polygonal mapping ~see Schwarz-
Christoffel transformation) R
Polynomial, 541 Raabe's test, 119
ex is tence of zeros of, 608 ~7 4) Radiation, from rod, 500 (9)
Polynomials : Radiation condition, 471
Chebyshev, 140, 174 (26) Radical, use of, 547 n
of second kind, 174 (26) Radius :
Hermite, 140, 173 (25) of convergence, 119,542,563
fndex
731
Radius (cont.): Sea lar, 269

of eurvature, 280 Sealar produet, 271
of torsion, 281 Seale faetor, 308
Ratio test, 118, 542 Sehwarz-Christoffel transformation, 641
Rational funetion, 122,541,571,576, Self-adjoint operator, 661, 695 (79),
579 697 (83)
Real number, 539 Semieonvergent series, 167
Real part, if eomplex quantity, 539 Separable differential equation, 31,
Reeiproeity relation, 657 505 (25)
Reetifying plane, 281 Separation eonstant, 446
Reeurrenee formula: Separation of variables, 1,444,484,
for Bessel funetions, 149 505 (25)
for Frobenius method, 130 Series:
for Gamma funetion, 77 asymptotie, 166
for Legendre funetions, 158 Fourier (see.F ourier series)
Redueible eurve, 293 Fourier-Bessel, 226
Reduetion of order, 28 Frobenius, 128
Refleetion: geome trie, 162
of eomplex plane, 678 (21) hypergeometrie, 162
of waves, 484 Maclaurin, 122 n
Regular funetion, 122 power, 118,541
Regular singular point, of differential semieonvergen t, 167
equation, 127,164 Taylor, 122, 354, 561
Removable singularity, 570 Shaft, rotating, 192
Residue, 580 Shear lines, 420
at infinity, 609 (85) Shearing foree, 192, 327 (43)
at isolated essential singular point, Simple eurve, 287
582 Simple pole, 571
at pole, 580 Simple region, 290, 555
Resistive foree, 88 (44), 521 (73) Simple surfaee, 295
Resonanee, 74, 523 (78) Simply eonneeted region, 290, 555
Rieeati's equation, 52 (66) Simpson's rule, 116 (27, 28)
Riemann surfaee, 570 Simultaneous differential equations, 18,
Right-handed eoordinate system, 308 75
Rod, deformation of, 327 (43) Sine-integral funetion, 30, 83 (16),
Rodrigues' formula, 160, 181 (60), 231 613 (99), 674 (11)
Rotating axes, 325 (32) Single-valued funetion, 289, 546
Rotating shaft, 192 Singular eurve, on integral surfaee, 414
Rotating string, 188, 243 (9) Singular point, of differential equation,
Rotation: 126, 164
ol'axes, 336 (100) Singular solution, 5,32, 36
of fluid, 302 Singularities, of analytie funetions:
of eomplex plane, 678 (20) braneh point, 131, 546, 567
Rouehé's theorem, 608 (76) essential, 572
Runge-Kutta method, 102 at infinity, 575
isolated, 570
S pole, 571
Saddle point, 356 removable, 570
732 Index
Singularities, of anal y tic functions (cont.): Stretching, of complex plane, 678 (20)
significance 01', 578 String:
Singularity functions, 65 deflection of, 188
Sink, 298, 635 rotating, 188, 243 (9)
Sliding clamped end support, J 93 vibrating, 243 (8), 492, 520 (72),
Smooth curve, 287 521 (73)
Smooth surface, 295 Strip, 409
Solid sphericaJ harmonic, 513 (47) characteris tic, 411
Solution, of differential equation, 1, 384 proper, 410
complementary, 6, 396 Strip condition, 409
complete, 4 Sturm-Liouville problem, 205, 380 (49)
general, J, 5 proper, 205
homogeneous, 6 Subsonic fluid flow, 319
particular, 6, 396 Substantial derivative, 317
singular, 5 Superposition integral, 463
Sommerfeld radiation condition, 471, Supersonic fluid flow, 319,495
664 Surface:
Sonic velocity, 318 conical, 392
Source, 298, 635, 658 cylindrical, 391
strength of, 635 equation of, 294
Source function, in heat flow, 526 (87), equipotential, 292
528 (92) normal to, 294
Space curve, 278 Riemann, 570
Specific heat, 441 Surface area, element of, 295, 308,
Speed, 279 376 (25)
Spherical coordina tes, 312 Surface charge, 652, 690 (57)
Spherical harmonics, 513 (47) Surface integral, 294, 296 n
Spherical waves, 468 Surface spherical harmonic, 512 (47)
Spring constant, 72
Square-wave function, 82 (10), 672 (5) T
Sta bili ty, 196 Table:
Stagnation point, 514 (50), 637 of Bessel functions of order n + ±, 70
Staircase function, 82 (12) of Gamma function, 78
Standard form, of linear ordinary of Laplace transforms, 67
differential equation, 4 of zeros of Bessel functions, 230
Standing wave, 520 (71) Tangent vector, 279
Stationary value, 362 Tangential acceleration, 280
Steady state, 442 Taylor series, 94, 122, 354, 561
Stirling formula, 80, 90 (52) differentiation of, 121, 562
Stodola-Vianello method, 197, 212 with remainder, 122,354
error in, 253 (44) uniqueness of, 356, 563, 603 (47)
Stokes's theorem, 290, 294 n, 303 Telegraph equations, 440, 480
two-dimensional, 305 Temperature, steady-state:
Straight line, equations of, 322 (5) inside circle, 448
Stream function, 315, 501 (10), 632 outside circle, 449
Streaming, 634, 638 in circular annulus, 446
Streamline, 314, 501 (10), 632 in circular cylinder, 508 (39, 40),
Stress function, 439 509 (41)
Index 733
Temperature, s teady-state (cont.): Variation of parameters (cont.):

in half-plane, 473 for partial differential equation s, 484
in rectangle, 443 Vector product, 273
jn rectangular parallelepjped, 453 Vectors , 269
in rod, 462 algebra or, 269
in sphere, 451 , 511 (47) differentiation or, 277, 284
Tension, 188, 327 (43), 459 products or, 271, 273, 275, 324 (27)
Thermal conducti vi ty, 441 Velocity, 279
Thermal diffusivity, 442 Velocity deviation potential, 319, 496
Torsion , 281, 327 (39) Velocity potential , 313, 439, 632, 658
radius of, 281, 327 (39) Vianello (see Stodola-Vianello method)
Torsional s tiffne ss, 328 (43) Vibrating beam, 245 (12)
Total differential, 342 Vibrating membrane , 459
Transform: Vibrating string, 243 (8)
Fourier , 238 Vibration, 72, 88 (44)
Fourier-Bessel, 240 natural modes or, .73
Hankel, 240 Viscosity, coefficient or, 341 (119)
Laplace, 53, 477, 480 Viscous fluid, 341 (119)
Mellin , 4 85 Volume, element or, 308, 352
Transient solutjon, in heat flow, 462 Vortex, 313, 636, 638
Translation : strength of, 636
ofaxes, 336 (100)
of complex plane, 678 (20) W
Translation properties, of Laplace Wa ve equation , 334 (94), 399,427 (33),
transform, 60 428 (35), 430 (37), 440, 459 , 467,
Traveling wa ves, 467 478, 492, 667
Triangul a r-wave function, 82 (11) nonhomogeneous , 667
Trigonometric function s , 542 Waves, 467
Triple scalar product, 275 standing, 520 t7 J)
Twisted curve, 281 Weber's function, 144
Twisting moment, 327 (43) Weighting function, 203
Work, 287
U Wronskian determinant, 4, 25 n , 40 (8),
Undetermined coefficients, method of, 46 (45), 177 (44), 375 (21)
lO , 18
Unit doublet function , 66, 527 (89) y
Unit impulse function, 65
Young's modulus , 192
Unit s tep function , 86 (32), 668
Unit vector, 270
z
V Zero vector, 270
Variation, 360 Zeros , of Bessel runctions, 150, 175 (35),
Variatíon of parameters: 176 (36, 37), 181 (61)
for ordinary differential equations, 24 ra bIe or, 230

Hildebrand CA

Enviado por

Dados do documento

Direitos autorais

Formatos disponíveis

Compartilhar este documento

Compartilhar ou incorporar documento

Opções de compartilhamento

Você considera este documento útil?

Este conteúdo é inapropriado?

Direitos autorais:

Formatos disponíveis

Hildebrand CA

Enviado por

Direitos autorais:

Formatos disponíveis

..

© 1976 by Prentiee-Hall, Ine.

AlI rights reserved. No part of this book

Printed· in the Uoited States of Amer:ea

PRENTICE-HALL lNTERNATIONAL, IN e. , Londan

2.1 An introductory Example 53

2.3 Properties of Laplace Transforms 58

4.1 Properties of Power Series 118

5.1 Introduction 186

6.1 Elementary Properties of Vectors 269

6 .18 Special Coordinate Systems 311

9.5 Poisson's Integral 450

10.1 Introduction. The Complex Variable 539

11.1 Introduction 622

11.3 Inversion of Laplace Transforms wilh Branch Points . The Loop

Answers to Problems 703

The purpose of thi s text is to present an integrated treatment of a number of

for extended diseussions or for the introduetion of additional (or aJternative)

1.1. Introduction. A differential equation is an equation relating two or

where h(x) is a given function of the independent variable x. The solution is

where e is an arbitrary constant. Whether or not it happens that the integral

if suitable account is taken of situations in which a divisor may vanish.

is said to be a linear differential equation of order n. The distinguishing charac-

ao(x):%: + a,(x)y =f(x),

1.2. Linear Dependence. By a linear combination of n functions u, (x),

e,u, + e,u + ...

Thus the n constants must satisfy n homogeneous linear equations. However,

Iinearly dependent aYer an interval J, then the determinant

W(u¡, u 2 , ••• , un) = (7)

The problem of solving Equation (8) consists of determining the most

IS of first order. A solution containing one arbitrary constant IS of the form

YH(X) = c,u,(x) + c, u,(x) + .. . + e,u.(x) = kL:-• 1 ekuk(x), (14)

In this connection, it should be explained that we refer to a function as a solu-

since trus expression contains n independent arbitrary constants and satisfies

mine an intf'grating factor p(x) such that the standard form

is equivalent to the equation

where p is the integrating factor defined by Equation (20), and e is an arbitrary

where the dummy variable io the iotegraod is denoted by ¿; in order to distio-

Example. To solve the differential equation

we· first rewrite lhe equation in the standard form,

An inlegraling factor is then

1.5. Linear Differential Equations with Constant Coefficients. The simplest

in which the coefficients a k are constants.

However, if one or more of the roots is repeated, les s than n independeot

when we make use of the fact that

Example 1. For the equation

Example 2. For the differential equation

By making use of Eu{er'sformu{a,t

Example 3. The equation

lioear combioations, the following families may be listed:

The family of a function consisting of a product of n terms of this type is readily

It will be found in all cases that the number of coefficients to be determined

Example 5. Consider the differential equation

The general homogeneous solution is

1.6. The Equidimensional Linear Differential Equation. An equation of the

There then follows

Thus, in particular, we obtain

and, in general, it is found that

xm ~:~ = ;z(;z - l)(;z - 2) ... (~ - m + l)Y. (29)

Example 1. To solve the differential equation

The solution is found, by the methods of Section 1.5, to be

) corresponding to the assumption YH = e" III the transformed equation. By

making use of the relationship

there follows, with the notation of Equation (27),