Escolar Documentos
Profissional Documentos
Cultura Documentos
HANDBOOKS
HANDBOOK OF
Avlleme and Baumeisler MARKS' STANDARD HANDBOOK FOR
MKCHANICAL ENGINEERS
Brady andClauser MATERIALS HANDBOOK
Callender TIME-SAVER STANDARDS FOR ARCHITECTURAL DESIGN
DATA *&
NUMERICAL
Chopey andHicks HANDBOOK OF CHEMICAL ENGINEERING
CAI.CULATIONS
Dean tANGE'S HANDBOOK OF CHEMISTRY
Fink andBeaty STANDARD HANDBOOK FOR ELCTRICA!. ENGINEERS
Fink and Chrisliansen ELECTRONICS ENGINEERS' HANDBOOK
CALCULATIONS
Gaylord and Gaylord STRUCTURAL ENGINEERING HANDBOOK
Gieck ENGINEERING FORMULAS
larris SHOCK AND VIBRATION HANDBOOK
Juran and Gryna JURAN'S QUAI.ITY CONTROL HANDBOOK
IN ENGINEERING
Maynard INDUSTRIAL ENGINEERING HANDBOOK
Merrill STANDARD HANDBOOK FOR CIVIL ENGINEERS
Perry and Green PERRY'S CHEMICAL ENGINEERS' HANDBOOK
Rohsenow, Hartnell, and Ganic HANDBOOK OF HF.AT TRANSFER
FUNDAMENTIS DEFINITIONS THEOREMS COMPUTER
Rohsenow, Hartnell, and Ganic HANDBOOK OF HEAT TRANSFER
APPLICATIONS MODELS NUMERICAL EXAMPLES
Rosaler and Rice -STANDARD HANDBOOK OF PLANT ENGINEERING
Seidman, Mahrous, and Hicks HANDBOOK OF ELECTRONIC POWER TABLES OF FORMULAS
CALCULATIONS
Shigley and Misckhe -STANDARD HANDBOOK OF MACHINE DESIGN TABLES OF FUNCTIONS
Turna ENGINEERING MATHEMATICS HANDBOOK
Turna HANDBOOK OF STRUCTURAL AND MECHANICAI. MATRICES
ENCYCLOPEDIAS
Contents
About the Author
Contents
vi
This handbook prcscnts in onc volumc a concisc summary of the major tools
of numerical calculations cncountcrcd in the preparation of micro-, mini-,
and mainframc computer programs in enginccring and applied scicnccs. It
was prepared to serve as aprofessional, users-oriented, desklop referente book for
engincers, architcets, and scicntists.
Since the publication of his Engineering Mathematics Handbook, which now
appcars in its third edition, this author has rcccivcd a large number of
rcqucsis to supplcmcnt his mathematieal handbook by a companion volumc
dcaling cxclusivcly with numerical methods, and to summarize thcse
methods in the telcscopic form which was found so uscful by many users.
The naturc ofthe subjccl ofnumerical calculations discouraged any prior
cfTorts to write a handbook such as this onc. Numerical calculations, as the
ame indcales, can be cxplaincd only by numerical cxamplcs, and defy the
ctiort to reduce thcm to simple formulas.
In early stages of planning of this handbook, it became apparent that the
material would require tripart presentation consisting of
(1) Goveming relalions, supportcd by theorctical rcasoning
(2) Particular cases, expressed in symbolic form and serving as micro-
numcrical modcls
(3) Numerical examples, providing the illustrations of the analytical back-
ground and serving as flowcharts of calculations
This effort required derivations of many ncw formulas, large numbers of
which appcar in the litcraturc for the first time, and a very carcful selection
of cxamplcs, which would providc the insidc understanding, but would not
extend in lcngth over onc page of the book.
The rcsults of this cTort, which spans a period of five years, are 1400
analytical relalions, 7000 micronumcrical modcls, and over 400 numerical
cxamplcs, kept in a volumc of rcasonable sizc ihat will allow dcsktop
handling.
* -1
Proface
(a) Methods Of calculation fall into two major cateogries, designated as analytical methods and
I Numerical Calculations 1.03 APPROXIMATIONS OF FUNCTIONS
(a) Algebraic polynomials are the only functions which can be evaluatcd at a given point and/or in
3
numerical methods. Analytical methods use algebraic and transcendental functions in the solution of a given interval in a direct and closed form.
problems, whereas the numerical methods use arithmelic operations only.
(b) All Other functions can only be evaluatcd for a given argument and/or in a given interval by
somc approximations. The most important approximations of functions are:
(b) Numerical methods which form the major part of this book are classifed as:
(1) Collocation polynomials (Chap. 3)
(1) Approximations of constants (6) Operations with seres and producs (2) Power series expansions (Chaps. 9. 10)
(2) Approximations of functions (7) Solutions of algebraic equations (3) Continued" fractions (Chap. 9)
(3) Approximations of derivatives (8) Solutions of transcendental equations (4) Rational fractions (Chap. 10)
(4) Approximations of integris (9) Solutions of systems of equations (5) Orthogonal polynomials (Chap. 11)
(5) Summations ofseries and producs (10) Solutions of diflerential equations (6) Lcast-squares polynomials (Chap. 12)
(7) Trigonometric series expansions (Chap. 13)
Two frequently used abbreviations are CA = computer application and TF = Ulescopic form.
(c) Example. The evalation of/(x) = 5*" + 3* 2 for x = 0.5 is obtained by a direct substitution
as
The argument that the same vale can be obtained by the calculator does not change the Tact that
(b) Fundamental physicalconstants are produets of natural laws. and as such their vales can only the calculator evaluation is based on some kind of approximation built into the calculator.
be obtained by measurcments. A complete list of thcse constants is given in Appendix B, and in
the back endpapcrs.
(a) Algebraic polynomials are the only functions which can be intcgrated in given limits in a direct (a) Series and producs are fmite or infinite, and they are classified as:
and closed form. Integris of all other functions must be evaluated by somc approximations. They
(1) Series and produets of constants (Chap. 5)
are classified as:
(2) Series and produets of functions (Chap. 9)
(1) Closed-form integris evaluated by functional approximations (1.03)
(2) Numerical quadratures in tcrms ofdiflcrcncc polynomials (4.05) to (4.09) (b) Sums ofseries and produets ofconstants are calculated by:
(3) Numericalquadratures in tcrms ofasymptotic series (4.10)
(4) Numerical quadratures in tcrmsoforthogonal polynomials (4.11) to (4.16) (1) Fundamental theorem ofsum calculus (5.03)
(5) Integration by powcr series expansin (11.01) (2) Diferencc series formula (5.04)
(6) Integration by trigonometric expansin (13.14)
(3) Powcr series formula (5.04)
(7) Goursat's formulas (4.17)
(4) Euler-MacLaurin formula (5.05)
(8) Filon's formulas (4.18) (5) General and special transformations (5.06), (5.07)
(9) Multiplc-integral reduction formulas (14.03)
(d) Euler-MacLaurin formula adds correction tcrms ofasymptotic series to the trapezoidal rule and (d) Sums of series of produets of special constants are available for:
provides the numerical analysis with one of the most important rclationships, which is uscd
invcrsely in the summation of series (4.10). (1) Series of produets of numbers (5.25). (5.26)
(2) Series of produets of fractions (5.27). (5.28)
(e) Gauss' Integration formulas, known as Gauss-Chebyshev, Gauss-Legendre, Radau and (e) Series Of functions are the power series and the transcendental series. The most important series
Lobatto formulas, use orthogonal polynomials as the basis for the cvaluation (4.11) to (4.16).
in this group are:
They gained rece popularity in connection with the finitc-element methods.
(1) Algebraic powcr series (Chaps. 9, 10)
(f) Power series expansin of the integrand allows the cvaluation of integris of functions which (2) Finite-diflcrence powcr series (Chap. 9)
cannot be evaluatcd in terms of elementary functions. The advantage of this method is its (3) Fourier series (Chap. 13)
application in the cvaluation of mltiple integris and itssimplicity in handling in general.
(f) Algbrale power series are either the result ofcvaluation ofa function in tcrms of
(g) Trigonometric series expansin of the integrand ofers the same advantages as the powcr series
method. In addition the trigonometric expansin of the integrand gives a symbolic expression for (1) Taylor's series expansin (9.03)
the integral cvaluation. (2) MacLaurin's series expansin (9.03)
or they are the solutions of diflerential equations dcsignated by special ames such as Bcssel
(h) Goursat's formulas allow the evaluation of product functions by a finitc series and are functions (10.07) to (10.17), Legendre functions (11.10), Chcbyshev functions (11.14), Laguerre
particularly uscful in producs ofalgebraic and elementan- transcendental functions. functions (11.16), Hcrmite functions (11.17), and many more.
(g) Finite-difference power series (9.03) may also be used for the cvaluation offunctions, but their
(I) Filon's formulas ofler the approximatc evaluation o( definite integris of(x)costx and applications yield only a particular valu of the function at a point.
f(x) sin tx, where / is a constant.
(h) Fourier series are used for the same purpose as the algebraic powcr series. They can be used in
(j) Multiple-integral reduction formulas reduce the mltiple integral to a single integral by mcans of the continuous range and also over equally spaced vales in that range. They are particularly
the convolution theorem (14.03). useful in the solution of partial diflerential equations (13.01), (13.17).
i l i 1) ' 11 r*ll '"Til f~^ll 1) r~3-|j] r^-^|
Numerical Calculations
1.08 SOLUTIONS OF EQUATIONS 7
6 1.07 OPERATIONS WITH SERIES Numerical Calculations
AND PRODUCTS (a) Algbrale equations of nth degree have closed-form solution if= 4. If>4. only special
cases can be solved in closed form. In general, the higher-order algebraic equations can be solved
(a) Nesting Of power series for the purpose of summation is the most commonly uscd operation, in only by approximations.
which the given series is rcplaccd by a nested product (Rcf. 1.20, p. 102) as
(b) General Closed-form SOlutionS are available for the following algebraic equations:
(i)V = <J0 fl,X a^x* a,Jt3 + (\)'a.x"
(1) Quadratic equations (6.02a) (4) Cubic equations (6.03a)
bkx\ (2) Binomial equations (6.02) (5) Quarticequations (6.04a)
= a(l b x(\ b3x(\ b%.r(l *)))) ==A li
4-1
(3) Trinomial equations (6.02/)
where ak = constant, bk = at/ak_%, x = real orcomplex variable, n + 1 = number of tcrms, and
(C) Special Closed-form SOlutionS are available for the following algebraic equations:
\ [ ] = symbol ofnested sum (1) Symmetrical equations offifth degree (6.06a*)
(2) Antisymmetrical equations offifth degrec (6.060
(b) Alternative form of nesting is (d) General methods used in the approximatc solution ofalgbrale and transcendental equations are
based on the concept of iteration and intcrpolation.
(i)V =(. *K +*( x(a* *< .*))) =ok *l (6) Iteration methods are applicable in solutions of equations of all types, and the most important
where the last expression in the product is a_, a,x. methods in this group are:
(C) Example. In (1.03c) with x = 0.5, the truncated series of n = 6 is (1) Bisection method (6.08) (3) Tangcnt method (6.10)
(2) Sccant method (6.09) (4) General iteration method (6.11), (6.12)
xcosx = x - *72! + *74! - *7/6! + *78! - *"/10! = 0.438 791 2809
The same series in nested form is (f) Polynomial methods of practical mporunce uscd in the scarch o real and complex roots of
algebraic equations of higher degrec arer
XCOSX =*(' "(IM2) (' "<5<4) (' "(5)(6) (' "7^8) 0"(9HI0j))))) ;l) Newton-Raphson's method (6.16) (2) Bairstow's method (6.17)
= 0.438 791280 9
Methods dcveloped in the era of hand calculations, such as Horner's method, GraeVs method,
As apparent from the nested structurc, x'2 is a constant used recurrcntly. Bcrnoulli's method, and Lagucrrc's method, are oriimitcd valu in computer apphcations and are
not covered in this book.
(d) Algbrale operations with standard power series are perfonned by:
(1) Summation theorem (9.14a) (3) Divisin theorem (9.14r) 1.09 SOLUTIONS 0FSYSTEMS 0FEQUATIONS
(2) Multiplication theorem (9.14A) (4) Powcr theorem (9.18a)
(a) Systems Of linear algebraic equations are classified as nonhomogencous and homogeneous.
All thcse operations can be performed in matrix form. Methods of solution of nonhomogencous equations fall into two categones:
(1) Direct methods, producing an exact solution by using afinite number ofarithmctic operations
(e) Algbrale Operations With Fourier series are possible and are given in series form in (13.14). The (2) Iterative methods, producing an approximatc solution of dcsired aecuracy by y.cld.ng a
rcsulting series are again a Fourier series. sequence o solutions which converges to the exact solution as the number of itcrations
tends to infinity
(f) Derivativos and Integris Of Standard power series are obtained by diflerentiating and integrat-
ing the series term by the term as an algebraic polynomial. Thesc operations can be performed in (b) Direct methods introduced in this book are:
a closed interval fand only if this interval liesentirelywithin the interval of uniform convergence
of the series. Derivatives and integris of Fourier series are given by relationships described in (1) Cramer's rule method (7.03) (5) Cholesky method (7.09)
(13.14). (2) Matrix inversin method (7.04). (7.05) (6) Square-root method (7.11)
(3) Gauss elimination method (7.06) (7) Inversin by partitioning (7.12)
(g) Derivatives and integris Of nested ser es introduced in <a) above are given symbolically as (4) Succcssive transformation method (7.07)
(a) Sequence is a set of n numbers ai,ai,ai,... ,a. arranged in a prescribcd order and formcd
according to a definite rule. Each number ak {k = l,2,3,...,n) is called a term, and the (a) CompariSOn test. Whcn ak is the *th term oaseries to be tested, bk is the *th term oanother
sequence isdefincd by the number ofterms as finite (n < ) or infinite (n = >). series which is known to be absolutely convergent, and cis a constant independen! ofA, then il
(b) Serles is the sum of a sequence. kl < c kl
the series of a terms is also absolutely convergent.
at =a, +a2 +a, 4- +a. =Sn (- < 5. < )
i-i (b) Example. If two series aregiven as
1
a, =a, +a2 +a, 4- +a.=5 (-,s) a. + a2 + flj + * " - 7^2
77o
1
2-3 3-4 k{k 4- 1)
*-i
General term ak (th term) defines the law of formation of the sequence. 2 2
, + 62 + A, 4- c\bk
2 3*"11
(c) Monotonic and alternating series are, rcspectively,
then \ak\ < c\bk\ for any valu of Aand the a series is absolutely convergent.
y. ak - a, 4- a2 4- a3 + + a. , Bak = a - a2 + a3 - aam
*-i '-' (c) Cauchy'S test. Ifa, is the Ath term of aseries to be tested and
where 6 = (-l)t+] ando- = (-l)"+l lim yfak = L (test fails for L = 1)
*
(d) Convergent series approaches the limit S as the number of terms approaches infinity (n -* ); then the series is absolutely convergent for L< 1, divergent for L> 1.
that is, if
(d) RatlO test. la, and ak+x are two successive tcrms of aseries to be tested and
2 ak = S and lim S = S
lim Ii I = i (test fails for L= 1)
exists, the series is said to be convergent and 5 is the sum (|5| = ). 4_ | ak I
(e) Dlvergent series increases without bound as the number of tcrms increases.
then the given series is absolutely convergent for L< 1, divergent for L> I.
(f) Examples
(e) Example. If the given series is
l 11 then
,.hm |I !/(*+')' II - m = 0
Al III c 3
> = l + - + - + + is convergent, since o = - i + a + a + " *= Ti + 2! 3 1/*! \k + I
S*-' 3 9 27 8 2
and since L < 1, the series is absolutely convergent.
2(l+A) = 2 + 34-4+--- isdivergent, since 5 =
(g) Absolutely convergent series isan infinite series whosc absolute terms form a convergent series
2.03 OPERATIONS WITH SERIES OF NUMBERS
(a) Grouping Of terms of an absolutely or conditionally convergent series may be changed without
X kl = \s\ aflecting the sum. ^
-i
(h) Conditonlly convergent series is a series which is not absolutely convergent. (b) Example
_L +JL+... = _ =ln2
(i) Examples 1-1 *
1-2 3-4 &k(k+\)
111 . . , ,
1 is absolutely convergent. since
1-3 3-5 5-7 (C) Order Of terms of an absolutely convergent series can be rcarranged without aflecting the sum.
Howevcr, a particular change in order of tcrms of aconditionally convergent senes may change
J- +-L +-L +
1-3 3-5 5-7
is a convergent senes
('-i)- the sum.
III , .. .
(d) Example. The conditionally convergent series 1- J4- \ - \ +=In2, but the rearranged
1
2
h
3 4
+ = In 2 is conditionally convenrent, since
7
series, l+i-* + i + *-i + "- = lnV5.
. ' l
(e) Sum, dlfference, and product of two or more absolutely convergent series is an absolutely
1 +^ + 3o +7 + is a divergent series <|i = ao) convergent series.
2 4
T ~~J "I ' I ~~1 "1 '1 '" "1 1 1 ""1 1
2.05 PRODUCTS OF NUMBERS, CONVERGENCE 13
Numerical Constants
12 2.04 PRODUCTS OF NUMBERS, BASIC CONCEPTS Numerical Constants
(a) Product of a set of n factors (1 4- a,), (I 4- a2),.... (1 + a.) is defined by the numberof factors
(a) Convergent prodUCt is an infinite product that approaches the limit Uas the number of factors
as a finite product ( < )
approaches infinity ( - ); that is, if
fl (i +ak) =U. and TfaU.-U
i-i
fl (I +a*) =(I +i)(l +i)(l +) (1 +O =/. i-i
1
f (i +kl) =l"l
fl 0"FT) =(Z)(Z)(Z)"' ='
12 2
i-i
2 3 4
(f) Example. Infinite product
fl (1 +ak) =(1 +a,)(l 4- a2)(l 4- a3) (1 +a.) (g) Conditionally convergent prodUCt is an infinite product which is not absolutely convergent.
i-i
T
"H 1 ~w rS-j "i "~1 1
208~ CONTINUED FRACTIONS, is
Numerical Constants
14 2.07 CONTINUED FRACTIONS, BASIC CONCEPTS Numerical Constants NUMERICAL PROCEDURES
(a) Continuad fraction is of the form (a) Conversin Of COntlnued fraction to a simple fraction is obtained by performing all the
operations indicated by the continued fraction.
(b) Example
i ii i 1 43 931
5 +
where the right-side expression is a symbolic reprcsentation. 6+ *
(b) Number Of terms ay,bx,a2,b2,... ,aK,bm defines the continued fraction as terminating (n < )
(C) Any pOSltive rational number may be converted to afini.e continued fraction by areversed
process.
or infinite (n = ).
(d) Example
(c) Simple COntlnued fraction has all partial numerators cqual to I.
f, = 4,+
I 0 r j ! l i
1 L"'b. + 'b, + ',+ * "J i + 5
b, +
, + (e) Any positive Irrationa. number may be converted to an infinite continued fraction by the same
*,+ process. or it may be approximated by the convergent Fk (2.07rf).
(d) Successive convergents for ak > 0, bk > 0 are defined as
(1) Fk of odd order is greater than F. but decreasing.
(2) Fk of even order is iess than Fm but increasing.
/=o = 77
Qo
= *o ^. = 77
Ci
= *o + 7*
b,
= :b.
*-2 X - 1 k
*i + -1 0 1 2 3
*2 *
"l-l i
0 fl2
i
4 *4
F _ P* _ a*P*-* +M1-.
* a* iai-2 + *iGi-, laJ tai-, cJUJ 1 P, P,
A-, /V-i
a-i di-i \'
0 1 a.
and
1
2.10 BINOMIAL COEFFICIENTS i?
16 2.09 FACTORIALS Numerical Constants Numerical Constants
1 (* = 0)
and by dcfinition, 0! = 1. Numerical vales ofn! for n = 1,2, 3,..., 100 are tabulatcd in (A.03). (k = I)
(b) Factorial of proper fraction (/(O < U< 1) is calculated by the procedure described in (2.29), and and for k > 0,
numerical vales of u! for u = 0.005. 0.010 0.995 are tabulatcd in (A.04). (x+ 1 x + 1
_N^__N^ A^ _ A^ (k = 1 or k = n - 1)
12 360 1260 1680 (A = n or A- = 0)
1 / S / 2S i 3SU\ (k < 0)
I,, nested form, A= (l -~(l -y(l - j)l) with S = A'
id for 0 < k: S n,
(f) Example. For A' = 7.3. with 5 = (7.3)-', * - n- 1
CK-,> ! 4-1/ A+ lU/ ( l) '
A =
12(7.3) V 30V 7V 1 'n-\\_{2-\)\= ,_.(-
7.3
By (e). A! = /""' *'"( Vl4.6,T = 9281.39 J (!)-'
Numerical vales of. (/"\1 -for n = 0.1,2,...
,iio 9<-, and
,25 and A
/ = 0u. 1 2 25 are shown in (A.06).
By (c), A! = (7.3)(6.3)(5.3)(4.3)(3.3)(2.3)(1.3)(0.3>! = 9281.39
where rrom (A.04), (0.3)! = 0.897 471. (e) Examples
(g) Factorial of N > 100 can be approximatcd by Stirlinv;s formula '8\ (8)(7)(6) &\ _ (8)(7)(6)(5) 8! = 7Q
0- (3)(2)(1) = 56
:) (4)(3)(2)(1) (4!)
A! =(-) \2 +e \e\ <5x lO""'
'A\A'-,
/3\ _ (3)(2)(l)(0)(-D(-2)(-3)(-4) - -
but W (8)(7)(6)(5)(4)(3)(2)(1)
This formula applics particularly well in computing ratios of two factorials.
1 r ~1 '""~ r'"-Ll
' 1 r"~1 1 "I ' 1 r" "1
Numrica! Constants
2.12 BINOMIAL FUNCTIONS, OPERATIONS 19
18 2.11 BINOMIAL FUNCTIONS, BASIC CONCEPTS Numrica! Constants
(a) Factorial polynomials X?> in real variable x and * = l, 2,3,... ,/> are, by definition, (a) Sum Of binomia! functions. If Q and (*) are binomial functions in areal variable xand p, a
A-(') =(*)n =x=SPj' are positive integers (p > q), then
A-<J =(^3! =x(x - l)(x - 2) =y*3'* +y'x2 +y^'x3 (b) Derivativos of (*) are
(b) First recursion formula. If 5^"' and ytkm_\ are known. then for m 2 * > 0, ! i-. W
P
PC.+I) = yJ-)( _ m9Xml (OT = 0, 1,2, ...) 4C)]= I (n=p)
Lo (n>p)
where for k = m, S^"0 = I, and for k s 0 or for k 2: m + 1, ff* = 0.
where Z)" = d"/dx" and n = 1,2,3,
(C) Example. If y(,2> = -I and Sff1 = +1, then by definition, S^02 = 0 and
(c) Indefinito integris ofrj are
^," = ^o2) - 2^2 = +2 y*3 = y*2' - 2y,-' = -3 y*3 = I
(f) Example. If 9*? = 4-1 and S/f* = 4-1, then by definition, 9f> = 0 and
i f /xx 1A^/>x f^aV-i^nx-^W^^^+C
&i* = 9f*-+ &f>^ 4-1 9?> **-&?>+-2&' =^1-3 ^J-^+i^ wherejr, b arejigned numbers and n + l ^ p.
1
20 2.13 BINOMIAL SERIES, GENERAL RELATIONS Numrica! Constants Numerical Constants 2.14 BINOMIAL SERIES, SPECIAL RELATIONS 21
(a) General case. The nth power of a x can be expanded by Newton's formula in a power series (a) Exponent n= -m(m = 1,2,3,4; u= x/a)
called a binomial series.
o
=i(lT(IT u(l =F u(\ 4= (1 =F (1 =F (1 =F )))))))
a x a
(.*)=(. .)=[. *(;).+ (2>- *...*(>]=.-^y
^=?(-H'TH-H-H-H'^<' >))))))
where n = signed number, I = binomial coefficient, a = (1)", B = (l), and u = x/a.
(c) CA model. The most convenient model of this series is the nested series introduced in (1.07),
y/a x=Val (("H'^-O'H"^'!<" '))))))
which for = x/a is
(a x)" =a" (1 cku) =a"(l nu(\ -^ a(l ^.(l )))) ^a x=V^(l (('H"I"("!-("t"("> *-'))))))
Five distinct forms of this model are given below in (d), and selected particular cases are
tabulatcd in (2.14).
V"a x=Va(l ((.H-(""("7-("T-("> *->))))))
(d) OistinctCA models. Uu'2 < I and X = u/p, (c) Exponent n = -1/m (m = 2, 3,4; u = x/am)
P
,m = 1 ^,,^(,,^-(,=^.,(,,3^.,, ,..,))) 1 1 /
WTx=Ta\i:fu[('*H"H"7-{"T-("7'" *->))))))
n =
m
P
{/(I r - 1mA'(l T>2H'.V( i;23 "A'(l T*4".V(l T)))) (d) Exponent / = p/fl (p = 2,3,4;a = 3,4,5; = x/a?)
22 2.15 GENERAL SERIES AND PRODUCTS Numrica! Constants Numrica! Constants 2.16 SPECIAL SERIES OF BINOMIAL COEFFICIENTS 23
OF BINOMIAL COEFFICIENTS
(a) Particular sums ofcomplete, monotonic sequences (n = 1,2,3,...)
(a) General sums of complete, monotonic sequences (n,r,N = 1,2,3,...)
.4 -'^o-o-(;)-<-<)-
.t<-o=()-2(;)+o- (-<)-<-r
-*C)--K0*<)---O-- (b) Particular sums of complete, alternating sequences (n = 1,2,3,...)
In 1882 C. L. P. Lindermann proved that n is both an irrational and a transcendental number The very slow convergence ofthis series is shown in (A) (for n= 32, < 5 X 10"\ or n- 1001,
and thus showed that the problcm of rectification and squaring a circle with rulcr and compass f<5X I0"7, and for n = 31 623, 6 < 5 X lO"'0).
alone is a mathematical impossibility.
355
113
- = 3.141 592920 - e (e < 2.7 x 10-7) ,r"4i?1L2it- 1s2*"' 239M-')J
_ v rziLi (JL- + _? 5\]
(e) Vieta'S product given in explicit form as * , L2A - 1V1821"' 573-1 239*-'W
2 The rapid convergence of these series is shown in (A) (in Machin's series of n - 7,
VV +T^V +V +V- e < 5 X I0"'; in Gauss' series of n= 4, < 5 x 10"*). Considering that cach term o
Machin's series involves only two fractions, this series is the more convenient of the two modcls
can be expressed as given above.
4*o =40 n i
R, R2; R3- i-i Rk
T II + R0 /l + R /l + R.-t
where R0=y^,R, =V-2~~'*2 =V-^'*" =V 2 (h) Vales of x based on n-term approximations
The convergence of this product is shown in (A) (for n = 15, e < 5 X 10"l0).
Wallis Lcibnilz Machn Gauss
Vicia
can be expressed in a limit form as 3.023 170 192 3.058 402 766 3.141592 652
6 3.141513 801
3.141572 940 3.038673 629 3.070254618 3.141592 654
7
"is{Ki-V43'''a.-i)r} 8
9
3.141587 725
3.141 591 421
3.050 588 266
3.059063837
. 3.079153 394
3.086079801
3.091623807
-
3.066 730687
The very slow convergence of this product is shown in (h) (for n = 40, 6 < 5 x 10"\ for 10 3.141592 345
n = 1253, < 5 X 10"7, and n = 35633, e < 5 x 10"'). 11 3.141592 576 3.073 080023 3.096 161 529
12 3.141592634 3.078 424 310 3.099944033
13 3.141592 649 3.082 985 139 3.103 145 314
(e) Standard Gregory-Leibnitz' series based on 14 3.141592652 3.086922 340 3.105889 726
15 3.141592 654 3.090 356 269 3.108 268 557
x3 x5 x7
16 3.093 377 145 3.110 360274
(a) Definition. The symbol edenotes the limit, (a) Definition. The symbol ydenotes the limit,
t = lim (1 + -)" = lim (1 4- n)"" = 2.718281 828459045 235 360 287 v- Hm fl + - + - + 4- - - Inn) = 0.577 215 664901 532 860 606512 -
._- \ mi o y ilIV 2 3 n I
and is thebase of the natural system oflogarithms. In 1873, C. Hermite proved thate is both an No proofis yet available whether yis an irrational number.
irrational and a transcendental number.
(b) Calculation Of yto adcsired degree of aecuracy can be accomplished by the cvaluation of the
given limit for large nas shown below or by the continued fraction listed in (rf).
(b) Calculation Of e to a dcsired degree ofaecuracy can be accomplished by means ofinfinite series (c) Example.
and continued fractions as shown below. For simple and rapid approximation,
.+I + I+...+-L- in(10s) - 6 = 0.577 715664 - 6 (e < 5 XI0"4)
r 4- = 2.714285 714 4- 6 ( < 4 x I0's 2 3 10J
1264
V=J l+I +4-...4--L-ln(106)- e =0.577 216165-6 (6 < 5x 10"7)
+ e = 2.718 279570+ e (f < 2.3 x 10-6) y- \ l+ 2 3 10"
1465
or directly by CA as ,+! +!+...+-L-ln (109) - 6= 0.577 215665 - 6 (6 < 5x 10"')
( 11 \\ io5
I + ) + = 2.716923932 + e (e < 1.4 x 10"') (d) Euler's continued fraction (e) Vales of y based on
n-term approximation
/ 1 \106 l
e =< O +~) + e = 2.718280469+ e (e < 1.4 X10"6) y = 0 + n Fraction
1 +
/ 1 \l(fl 1 1.000 000 000
1
(I H =1 +6 = 2.718 281828+6 (6 < 4.6 x 10_') 1 +
0.500000000
k\ 10/ 2
2 + 3 0.600000000
1 +- 4 0.571428 571
5 0.578947 368
2 +
(c) Euler's series based on 1 + 6 0.576923 077
4 + 7 0.577 235 772
1 1 I
8 0.577 215 190
1-1*! 3 +
9 0.577 215671
13 + 10 0.577 215664
can be expressed in TF as 5 +
II 0.577 215665
converges very rapidly as shown in {e) (for n- 11, \e\ <
i(2*- 1)! 5 X 10~') and is also a good CA model.
The rapid convergence ofthis series isshown in (e) (for n = 7,e < 4.6 x 10~10). (f) Simple and rapid approximations by simple fractions are the convergents of {d).
fu
- 6 = 0.578 947 368-6 (6 < 1.8 x lO"3)
(d) Euler's continued fraction (e) Vales of e based onn-term approximation 19
I Fraction y =<
- 6 = 0.577 235 772-6 (6 < 2.1 x 10"5)
m Seres 123
e = 2 +
I + 2.000000000 2.000 000 000
1
2.666666667
+ 6 = 0.577 215 190 +6 (6 < 4.8 x 10"7)
2 + 2 2.666666667
.395
3 2.716666667 2.727 272 727
3 +
4 2.718253968 2.718981 132 and for routine calculations,
4 + 2.718281526 2.718263 332
5 +
5
y = y/I _ = 0.577 350269 - e (e < 1.3 x 10"4) i
6+ 6 2.718281826 2.718283 694
7 2.718281828 2.718 281658 (g) FibonaCCi'S formula in terms OR =1(1 +V5) and , = 2.718 281 828 (2.18.).
converges rapidly as shown in (e) (for 8 2.718 281843
n = 11,6 < 4.6 X 10"') and is alsoa good 9 27718281827
CA model. 10 2.718 281829
2.718 281828
11
is the best approximation of all.
"~1 "^ " 1 1
Numrica! Constants 2.21 FIB0NACCI NUMBERS 29
20 2.20 NATURAL LOGARITHM OF 2 Numerical Constants
(a) FibonaCCi numbers Fr (r= 0, 1,2, 3,...) are defined by their gencrating rclatton
(a) Definition. The symbol ln 2 denotes the natural logarithm of 2, defined by the relationship
Fr = /%-, + /=",-!
e,a2 =flim ('+-)] whcrc m2 =-693 147 180559 945 309417 232 where F0 = 0, Ft = 1, F2 = 1, Fs = 2, F4 = 3, Numerical vales of the first 50 Fibonacci
numbers are given below in (*)
This constant is an important component in the construction ofCA modcls of natural logarithms
of large numbers (9.06). (b) Closed form of this number is
(b) Calculation Of ln2 to a desired degree of aecuracy can be accomplished by means of continued
fractions and infinite series as shown below. For simple and rapid calculations,
ln2 = Vi - 6 = 0.693 375 245 - 6 (6 < 2.3 x I0"4)
r.-u^'-^y-^-"
where
(c) Napier's continued fraction, a = 1.618 033 988 749 894 848 204 B = -0.618 033 988 749 894 948 204
1
ln 2 = 0 + are the gotden mean numbers.
1
1 +
(c) Golden mean numbers are the roots of
2 +
x* - x - 1 = 0
3 +
1 + which is the condition of the divisin of a line segment AB in
6 + the adjacent figure by the point C in the mean and exteme
3 +
ratio, so that the greater segment AC is the geometric mean of
AB and CB,
1 +
1 + * = V(i)(* + i)
2 +
and the golden rectangle is the rcctangle whose sides have the ratio of(l +V) and is supposcd to
converges as shown in (/) (for n = 11, \e\ < 2.8 x 10~a). Simple and rapid approximations by have the most pleasing cffect on the cyc.
simple fractions are the convergents of this fraction constructed by (2.08/).
(d) Series representation of the u + l)th (e) Table of Fibonacci numbers
(d) Gauss' series based on (f) Vales of ln 2 based on n-term approximation Fibonacci number is
r F, r Fr r F,
III I
- = - + + 7 + i Fraction Gauss Goursat
1 x 3*J 5*'
1 1.000000000 0.600000000 0.666666667
'--(,1*) ?
1 1
1
18
19
2 584
4 181
35
36
9227 465
14930 352
for x = g yields 2 0.666666667 0.672000000 0.691358025 where m is the nearest higher integer 3 2 20 6 765 37 24157817
ln2 =T+ _ +_ + ... j 0.700000000 0.687 552000 0.693 004 115 above jr or |r. 4 3 21 10946 38
39
39088 169
63245986
4 0.692 307 692 0.691 551 016 0.693 134 757 i 5 22 17711
5 0.693 181 818 0.692670830 0.693 146047 6 8 23 28657 40 102 334 155
where a = 5. The relatively slow con (f) Ratio of two consecutive Fibonacci 7 13 24 46368 41 165 580141
vergence of this series is shown in (/) 6 0.693 140 794 0.693000646 0.693 147 074
numbers is 8- ' 21 25 75025 42 267914 296
; 0.693150685 0.693 011 129 0.693 147 170 433 494437
(forn = II, 6 < 4.1 X 10-7). 9 34 26 121 393 43
8 0.693146417 0.693132 459 0.693 147 180 701 408 733
K = FJF,. 10 55 27 196418 44
9 0.693 147 362 0.693 142 154 317811 45 1 134903 170
28
(e) Goursat's series based on 10 0.693147097 0.693 146226
11 89
1836 311 903
12 144 29 514229 46
where R, =
N+ I= f l 1 11 0.693147 208 0.693 146672
13 233 30 832040 47 2971215073
4807 742049
1 + 14 377 31 1 346 269 48
N 12N + I 3(2N + l)3 2178 309 49 7 778 742049
I + 15 610 32
16 987 33 3 524578 50 12586269025
1 +
for N = yields
is a continued fraction with r divisions.
r t>* f 1
(g) Examples
where b ^-4^-The better convergence^ of
this serie;* is shown in (f) (for n = 8,
6 < 5.1 :; lO"'0).
By (b), FB ^yjr"8 - 0") =21- -tw *~t>*GVQ-
"~ ~ J ~J " 1 ""H '"" H ' 1 I
30 2.22 BERNOULLI NUMBERS Numerical Constants Numerical Constants 2.23 BERNOULLI POLYNOMIALS di
(a) Generatlng function (a) Definition. Bemoulli polynomials B,(x) oforder r = 0, 1,2, 3,... are defined as
(c) Relatonships. Bcrnoulli numbers Br and B, of order r can be always computed in terms of their and in general,
lower countcrparts as
r - 1 1 A r+ l\-
'Qnl BK.m(x) (n <r)
B'=i(7T0-rTT.?,(2*K (-4,6,8...., Dm[B,{x)) =-
( = r)
2rr-
- 1 l , 2r + l\
/2r l l>0 (>0
5'=^ =ai(i
where Dn = and =1.2,3,
where j = r - 1, / = r - I, a = (-1)'+', and 8 = (-l)+\ dx
,5 /rU -_ <.(*) +C
H*MJ + <, (n *"_ i), +
h- c,
<"{H *"_ 2)! + +C. for n r
0=1 2=-l 4 = 5 6=-61 8=1385 l0=-50521 where B,,B2,BS,.. . ,,+l are Bemoulli numbers introduced in (2.22fl). Particular polynomials for
, = 0 3 = 0 s = 0 , = 0 9 = 0 ,, = 0 r = 0,1,2,3,...,25 are tabulatcd in (A.13).
are Euler numbers of order r = 0, 1,2,... (A. 12). (b) Properties. In general,
,(x + 1) + ,(x) = 2x' ,(x) = -ar,(l - x)
(b) Auxiliary generating function
Forx = {, r = 1,2,3,..., the polynomials define the respective Euler numbers as
-S*i-S*+++- (l")<w
where
2M) = 2"2'2, = -2"2'ar
where a = (-I)'4"'.
2r+l(3) = 2r+l = 0
(c) Relationships. Euler numbers , and E, of order r can be always computed in terms of their and in general,
lowcr countcrparts as
(n = r)
,+.(*)
6 =-I - Q^ - (J)A =-61 E3 =-(-1)46 =+61 H fjr^l =
'lfi'WJ /r +\, + C'(- O! + C,(-2)! + + C for n r
(e) Seriesrepresentation. For r = 2,3,4,...,a = (-l)r+\ where /"[ ] = /// (<&)". " = 1, 2,3,... , C and C,,C2 are the constants ofintegration.
(a) Definition. Complementary Rieman zeta functions Z(r) of order r = 1,2.3,... are defined by
A I 1 I I
the series as
zw =.?,P =P+? +? +
B _ 1 _l_ + 1
_....
or by the product Z(r)
1-1 (2k- I)' ~ 1' 3'
where the product is taken over all prime numbers p = 2, 3,5, 7, 11, 13, (b) Special vales of Z(r) are
n - jis - 5n5
Z(l) = -,Z(3) = ,Z(5) =
y' 4' K' 32' W 1536
(b) Alternativeseries defining Z(r) is and in general in terms of Euler numbers and polynomials (2.24), (2.25),
2' A I 2' / I 1 1 \
z(r) = y = (- + - + - + 31' /',
r' tiW
2' - iiT,(2* - i)' 2' - i \v y y / dx\ ifriseven
B r- I)! Jo 4c
Z(r) = S
The series converges rapidly for r > 5. if, (2* - 1)'
if r is odd
L 12'*' (r - l)!l
where r & I and B = (-1)**'. Numerical vales of Z(r) for r = 1, 2, 3,... , 20 are tabulated in
(c) Special vales of Z(r) are (c) below. Large tabics of Z(r). including decimal vales of argument, can be found in Ref. 2.03.
A particular case, Z(2) = 0.915965 594 177219 015 -, is called the Cataln constant.
Z(,) =oo,Z(2)=^,Z(4)=^,Z(6) =^....
and in general, in tcrms of Bemoulli numbers and polynomials (2.22), (2.23), (c) Numerical vales*
{2n)'\B,\ 'Z(r)
+ o- if r is even Z(r) fl - ()']Z(r) [1 - (i)'-]Z(r)
Z(0 = 2 T, =
i-i *
-.fifi/'
(2*)' f B,(x)
r! Jo
(cot-T.v r ifrisodd
r
i-
-i* T,(2*- 1)' i-i* t (2* - d'
1 03 00 0.693 147 181 = In2 0.785 398 163 = a/4
where r > 1 and a = (-l)'+l. Numerical vales of Z(r) for r = 1,2,3,... ,20 are tabulated in 2 1.644934 067 1.233 700550 0.822 467033 0.915965594
(2.27r). Large tables ofZ(r), including decimal vales of argument, can be found in Ref. 2.03. 3 1.202056903 1.051799 790 0.901542677 0.968946146
4 1.082323 234 1.014 678032 0.947 032829 0.988944 552
5 1.036927 755 1.004523 763 0.972 119 770 0.996157 828
|.i_i+i.t....[l.(ir>w (r = 2,3,4,...
17
18
1.000007 637
1.000003817
1.000000008
1.000000003
0.999923 782
0.9999% 188
0.999999992
0.999999997
19 1.000001908 1.000 000 001 0.999998094 0.999999999
20 1.000000956 1.000000000 0.999999047 1.000000000
r) (r =-2rS-,-4,.
Ref. 1.20, p. 461.
~ "1 1 ~i "H ' 1 1 " ""1
36 2.28 GAMMA FUNCTION, GENERAL RELATIONS Numerical Constants 2.29 GAMMA FUNCTION, NUMERICAL EVALUATION 37
Numerical Constants
The graph shows that T(x + 1) is single-valued except at x = -I, -2, -3,... , and r = A,u(l - i4,*(l - .4s-(l - A,u2(l A,3u2))))
its alternating extremes are T(l.462) = 0.886, T(-0.504) = -3.545, T(-1.573) = 2.302, The general expressions for ak and Ak are a, = I - y, a:l = Z(3) I, a5 = Z(5) I,... and
r(-2.6ll) = -0.888, T(-3.635) = 0.245. T(-4.653) = -0.053, T(-5.667) = 0.009, A, = a,, A3 = a3/a,, i45 = as/a3,..., where y = Euler's constant (2.19) and Z(r) = Riemann
T(-6.678) = -0.001 (Re. 2.08). zeta function (2.26).
(C) Example. If T(1.25), T(0.75), T(3.25), and T(3.75) are dcsired, eT is calculated first with
(C) Functional equations. With the restrictions placed uponx in (b), u = 0.25.
r(x + n) = (x + n - I)(* + n - 2) u + 2)(x + l)r(* + 1) axu = 0.105 696083 8 -asir
,.* =
-0.000007 2124 -a.,u" = -0.0000000008
rt.-.)-7(* - 0iJl
)(* - n + I) (x - 2)(* - 1)*
-a33= -0.001032 379 7 -a77 = -0.000 000 072 7
the sum of which is r = 0.104 636418 2 and eT = 1.110306 850. Then by relalions (a),
wherc n = positive integer and x = signed number (positive or negative, integer or fraction).
n, , nA r /(0.75)(0.25)tt /(L25)(0.25)g
(d) Integer and fraction arguments (n,p, q = l. 2.3....)
r(,+0-25)-V,.25sin0.25^ T(l - 0.25) - , V0.75sin0.25^
0.906 402 477 3 + e 1.225 416 702 2 + 6
r( + i) where = -2 X 10"'" based on comparison with the tabulated vales in (A.04). The remaining
= n(n - !)( - 2) 3 - 2 - I = ni = (n - 2)(n - 3) 3 2 1 = (n -2)! vales are calculated by (2.28a*).
r(n + p + 1) = (n + p)\ r(n-p+ !) = ( -p)\ (d) Large argument. If x > 4 is an integer or decimal number, the Stirling's expansin (2.09),
n- +) - (2n!!4^
n!4
r(- + A) = (_4)--'^
(2n)!
10.00
20.00
' 3.628800000
1.216451004
(+05)
(+17)
3.628800000
1.216451004
(+05)
(+17)
50.00 6.082 818 640 (+62) 6.082818 640 (+62)
__^^
Numerical vales ofT(n + 1) and r(u) are tabulatcd in (A.03) and (A.04), rcspcctively.
' i " "M 1 '"""J "" ~1 '"'1 ' l "l "Ti
Numrica! Constants 2.31 Pl AND BETA FUNCTIONS, 39
38 2.30 OPERATIONS WITH GAMMA FUNCTIONS Numerical Constants
GENERAL RELATIONS
(a) Binomial coefficient ( 1can be expressed in terms ofgammafunctions (a) Pi function l(x) is the gencralization othe factorial (2.09a) and is defined as
(*\ = *! = H* 4- I) n(x) =T(x+ 1) =[ e-'t'dt =x! (x>0)
W (x-*)!*! r(x-k + \)r(k+ 1)
where k = 0, 1,2,... and x = signed number exccpt 1, 2, 3,.... where x! is the general factorial. All relalions introudccd in (2.28) to (2.30) are applicable in
operations with this function.
(b) Factorial polynomialof step 1 (2.1la) of order min x is
T(x + 1) (b) Betafunction H(x,y) is defined as
A'(|,, = x(x - l)(x - 2) (x - m + 1) =
ru - m + i)
1 r(* + i)
B(*,.r) ={/-'(i - 0"' dt =jf(l+<r^f <* >>J >0)
x\- (x + l)(x + 2) - (x + m - l)(x + m) T(x + m + 1) and is related to the gamma function and pi function as
where m = positive but m = I, and x is the signed number restricted by (2.28).
Hx)r(^) n(x - \)U{y - 1)
B(x,jr) B(x,y) =
(c) Factorial polynomial of step h of order m in x is r(x,j) n(x + y - I)
0 0.577 2156649
where T'(x + 1) is the derivative of T(x + 11 with respect to x. V()-,+2u (I +)(!-) 2tana* 2 0.2020569031
4 0.036927 7551
(b) Natural logarithm of Gauss'limit in (2.296) can be diQcrcntiatcd 6 0.008 349 277 4
term by term so that V(-) - ' 2 (I +00 " ) 2tani/JT 8 0.002008 3928
10 0.000494 1886
12 0.000122 713 3
M =lim (in. - ij^) =i (1 - --^) - y where in terms of A* given in (A), 14 0.0000305882
0.000007 637 2
o, = A0 + b2u* + b4u* + + Al6'6 + e 16
where y = Euler's constant (2.19a). In some literatures ip{x) is called the psifunction. The graph and |c| = 5 X 10"".
in (a) shows that \p(x) is singlc-valued except at x = 1, 2, 3,..., and its zero points are at
x = 0.462, -1.504, -2.573, -3.611, -4.635. -5.653, -6.667, -7.678,....
(d)
Large argument. If x>5is an integer or decimal number, the Stirling's expansin
(e) Reflection and duplication formulas
t//(x) = lnx + +
V(x + 1- n) - lfi(x + I) = .i x k
tp(x) - y>(x + l) =
yields good results as shown in (e) below, where in terms of A(2.09),
2y(2x - 2) = y(* - O + V(* - 5) + 1"4 y() - V(-)
u tan uJl
a"A B2t B2 B_B_B J_ _L_ _L- +^_
where x = signed number, n = 1,2, 3,... . and |a| < 1.
<P==-S^=-i-47-6x6 8x8 12x2+120x* 252x6 240x8
(f) Special vales (e) Errors of Stirling's expansin oftp(x)
V(-n) = V(0) = -y V(D = l ~ 7 V(2) = 1 + 5 - Y VO) = l+| + !"y Corree! valu of V>(*)
Approximatc vale of rp(x)
Numerical Constants
2.35 POLYGAMMA FUNCTIONS, NUMERICAL EVALUATION 43
42 2.34 POLYGAMMA FUNCTIONS, GENERAL RELATIONS Numerical Constants
(a) Higher derivatives of the digamma function ip(x) in (2.32a), called the polygamma functions
(a) Small argument. For 0<u* l the polygamma functions can be evaluatcd in tcrms o the
respective derivative ofthe formulas in (2.33a) as
V(,)(*)> V<2,(*) V("'(x) arc defined symbolically as
1 1 aVo
V<m,W ==[*(*)] =^TCnin* +I)]} 4g5i =V("() =-^ +^H)5" 2(I _.,. T2sin** ;+
The series representing thcse functions converge very slowly and arc of a very limited practica! * =2A2 + 12u'A4 + 30u4A6 +56^ + + ISOa'V = 2 undn
valu. If, however, thcse series occur in the solution of a probiem, they can be approximated by du2 *'
the rapidly converging asymptotic series introduced in (2.35a). and no error oceurs in 10-digit display if At's are taken up to k= 10. In nested form,
(c) Functional equations. With the restrictions placed upon x in (2.32A), = c2(i + cy2(\ + eyu + cyn + + c*,2))))
du
(-ir+,m! (-l)"+,m!
V(""W =-v ^,"" +Vlm,(x - 1) Vlmi(*) =x+irZ\ +vim)(* + i) - = D2{\ + Dy(\ + D6u2(\ + 0.a(l + 4- D.Mu2))))
du
which are the mth derivatives of the functional equations in (2.32c). where 2i, a, C*. D2k are the factors tabulated in (c) below.
(c) Factors c2, d2*, C2k, 02*
(d) Integerand fraction arguments (m,n = 1.2.3.. D u
d.a
0.404 1138064
V(,,() =Z(2) - ./"() =om! [z(m +1) - t^Tt] 0.4O4II38064
0.147 1100206
0.404 1138064
0.443 132665 2
0.404 1138064
0.365 5183719 1.0965551160
0.565244 0386
0.250478 322 0 0.3391464232
0.050095664 3
jp<(i) = 3Z(2) - 4 y""'^) = aTO!l(2",+, - l)Z(m + 1) - 2"*+lJ 0.016067 142 6 0.1124699968 0.320729 205 5 0.449020887 7
0.307 577 1568 0.395456 3444
0.0049418860 0.044 769 7400
0.297975 3396 0.364I92O8I7
where Z(2), Z(3),..., Z(ra + 1) are the Riemann zeta functions (2.26a). 0.001 472 560 2 0.0161981556
0.2908100578 0.3468461381
0.000428235 3 0.005 567 0524
0.285 3458364 0.329245 1958
0.000 122195 2 0.0018329280
0.281089924 2 0.318 5685808
0.000034 347 0 0.000 583 909 2
(e) Reflection formulas 0.0001812220 0.277 707 792 5 0.310379297 4
0.000009 538 7
v,(.)W_v?<n(x+,) =-^i- sin jzx (d) Example. Ifthe valu of y"^) is dcsired, dw/du is calculated first with u- i
y>w(u) a,2 =0.202056903 1 u\ =0.000 1255245 *, =0.0000000522
r v ' + V(,)(-)
ir v / = --
aj + sin2
-r^ Jtu u du tan Jtu
u\ = 0.018 463 877 6 u9c>0 = 0.000 009 652 1 ,5*, =0.000 000 003 7
where x = signed number and \u\ < I. u% = 0.001 5654895 \i"cn = 0.0000007190 .*,. = 0.0000000002
= 2 h2*"'^ = 0.222 2l:2 222 1, and by the relation (a),
(f) Special vales d A-| " i
' _+ * 0.222 222 222 I = 0.934 802 200 7
y>l,)(-n) = V(0) = Z(2) tf(,,(l) = Z(2) - 1 v(,,a) = -^r7T5
2(0.5)2 + 2(1.5 2J0.5)2 2sin2^
l2 o *.:.
y<->(-) = 0O ^"'(O) = Z(m + 1) V("'(l) = am![Z(m + 1) - 1] ... . /o oijv . ,i,(D/i\ V7C)\ 4 = 09348022005, and the error is
The correct valu is given by (2.34rf) as V (2) - i'\) * U,3J
due to round-off only.
Numerical vales of ^""'(x u) are calculated by the procedures described in (2.35).
i J
EVALUATION (Continued)
(e) Intermedate argument. For < x < 5. the recurrcnt formulas in (2.34c) yield good resulta
(f) Examples. If the vales of V("(3.5) and of yM>(-0.5) arc dcsired, then by (e),
*<(-0.5) = -U
(0.5)
+tt-
sar 2Jt
- V -5) =4.934 802 200 7 DIFFERENCES
(g) Large argument If.v > 5 is an integer or decimal number, the derivatives ofStirling's expansin
in (2.33) yield
1 I <i->
where
0 =J L +J L +^ rf-> _ 1 _1_ J_ 3 5
</x 6x3 30xs 42a-7 30a" dx2 ~ 2*4 + 6*,; 6.v" + 10x'" 6x12
66x"
dm<p _ a_ r(w + 1)! _ (w + 3)! . m + 5)! _ (m + 7)! 5(w + 9)M
<**""*"[. 12.V" I2P3!)*4 232t5!)* 240(7!)a" 660(9!)a-,uJ
Iffivc terms of the asymptotic series are included, no error occurs on 10-digil display. Tablc (A)
below shows the range ofapplications of this formula for T//"(.v).
(h) Errors of Stirling's expansin of t/,(1,(*)
Approximatc vale
X Corren valuof y("(*) ofV(,,(x)
5 jrs/6 - (1 + 1/2 + 1/3' + l.'4; + 1/5-) = 1.813229557 (-01) 1.813 229 557 (-01)
Kl .t76 - (1 + 1/2' + 1/3" - - 1/10'-)= 9.516633 556 (-02) 9.516 633 556 (-02)
m 31 j a i
y
-/<> where m,n are signed numbers.
is the slope of the chord k, k + I,
*7 ' '
tan q> = yY - 1
l
1
_l
(C) Classification Of operators. In this chapter three classes ofopcrators occur: transformation operators
Eand, difference operators A, V, 6, U, and derivative operator D.
A, Ai
where Jt = 0, 1, 2, 3 i-v
i. - ,i
* *-l
(d) Tableof applications ( = 1,2,3,...;k =0,1,2,3,...; r=0,1,2,3,...)
(c) Forward difference of first order, defined as *..
4+1
/ i
Central difference operator tyi = .n +i/2-St-in >, = 2 (-1 >'("),.,-,
A**) -A**-i)x -J*-' - i t
denoted as -i -i
Dcrivalive operator D Djk~ (dy/dx)k />>i = (o/tni
1 A
v>* = yk -jk-i
where the superscript ndefines the order ofthc operation.
(e) Central difference offirst order, defined as (e) Examples. Uyk =f{xk), then
ir /W/
A*.+in) -/(**-i/a) =yk+\n ~ Jk-m i'Vk =yk l\ =yk E> = j*+2 E^i=.Vi+:,
4+ I
is the difference of the vales of the function at xk+m, A>* =^+3 - 3^4+2 + 3y+I -jr4
AS =^+2- 2y4., +.v,
xk-in> dcnoted as *+i A
V->4=^ - 2y4_, +^,_, V>* =^i ~ 3.yi_, + 3y4_a -.Vi-:.
fyk =7*+1 ~ yk-
5">i =.v+1 - 2j +>,_,
where yk.m, yk+t/2 are the vertcal coordnales of
midpoints betwccn k - 1, kand *,k + 1. respectively. O = iU-, + 2yk +.v,+l) M^ = Jk-vi +Jk-fi + 3-'* +> +->'*+:)
)> = (d'y/dx2),. />>* = {d"y/dx')k
(f) Central means of first order, defined as
f[xt-in) +f(xk+,n) yk-xn +yk +m (f) Relations of differences (m, n = 1,2,3,...)
2 2
t\yk = v><+, = <5>1+,2 V>4 = A>i_. = >4_.
is the arithmcc means ofthe vales ofthe function at xk_xl2 a id x+l/2, denoted as
A" v>i = A"*>i_. = 6"+>i+,/2-.- >, = A>t_. = V>4,
m = Cn-i/ +y*+m)
i I
48 3.03 LINEAR OPERATORS, GENERAL RELATIONS Numrica! Differences Numerical Differences 3.04 DIFFERENCE TABLES 49
(a) Notation. n,r = positive integers E,A, V, , u = operators (3.02) (a) Types Of tables. Thc difference tabics shown n (d) to (/) below are the flowcharts for the
calculation of finitc differences. In general, two formis of these tables are available: the diagonal
(b) Relatlonships between operators formal and the horizontal format.
A V D
E
E E A + 1 (1 - V)- 1 + h- + Vi + i<5'-' e
(b) Construction Of tables. First, thc numerical vales ofxk and yk are recorded in thc lirst two
A E - 1 A (1 - V)-' - 1 ?,- + 6Vi + J- /" - 1 columns of the respective tablc. Ncxt, thc respective diflerenccs are succcssivcly calculated from
1 - (1 + A)" V - ?.- + Vl + \6'2 1 --" the nearest column on the left. This process is repeated to the desired order.
V 1 --'
6 E"--E-"' A(l + &)-'-' 7(1 - V)-'" 2 sinh KhD
i(E"- + E-"v) ^(2 + A)(l + A)1 -' \(2 - V)(l - V)-"-' Vi + iy 2 cosh \hD (c) Three rules must be observed in thc construction ofthe difference tablc:
D A"'lnE A'1 ln (1 + A) -A ' ln (1 - V) 2A~'sinnh~' (/2) 1) (1) Thc computation of thc nth-order difference requires n + 1 vales of xk and.v4.
(2) The nth-order differences of an nth-order polynomial in x are thc same vales, and the
(jl + l)th-orderdilfercnccs of thc same polynomial are zcros.
(c) E serles in terms of A, V, and D (3) The sum of numbers in onccolumn cquals the diferencc of the topand bottom numbers in
thc preceding column
E'=(l + A)" = E A' E-= (1 + A)'" = S(-l)'( A'
,-o \ r /
(d) Forward differences
E- = (i-v)-" - j n ? E-= (1 -V)- = (-l)'('V Tablc Example
,_o \ r i
y~i
-1 200
*-,
(e) Vseries n terms of E and D v>-, -180
*-> _2 20 162
>-! y-\
V" = (1 -E-)" = (-l)'E-' V = 1 - e' -St-ir? > y
0 2
-18
816
654
-1068
nr
v>, / ^ Vy3 798 -411
<02
V-- = (| -,-*) = 2 (-D'(2r - 2) l j\ ^ ^ v>-' 2 800
-^200
1 2000
V- = (1 - e-"")" = r-J
(-l)'+,[3' - (3)2' +31-!
r-
>, y-
(a) Derivative operator. By definition, the first-order derivative ot~yk = f(xk) is (a) Notation. a, A,c = constants A = spacing m,n,r = positive integers
D[c] 0 Me) = 0
(c) Transcendental functions
D[x] 1 A[x) = h A[a**] = (a** - l)a** A[sin (a + Ax)J = 2 sin(^AA) sin [Ax + a + (AA + jt)]
Z>[J = cDu A[cu) = c Au A"(a**] = (a** - 1)V* A"[sin (a + Ax)J = 2" sin" (5AA) sin [Ax + a + (nAA + nn)\
>[u + ] = Du + )p A[u + .- ] = Au + Av A"[cos (a + Ax)J = 2"sin* (^AA) eos [Ax + a + r2(nbh + nn)\
A"[f**J = (/* - l)V*
>M u Dv + v Du A[ur] = u Av + v Au + Ai/ Ap
v Du u Dv i' Au u Av (d) Leibnitz' rule. If u0 =/(x), a, = f(x + A),... ,a, = /(x + kh), v0 = (x). r, = g(x + Al.
4;] v2 4;] =
vw "* = g(x 4- A), then
A["uij _ "kt A
Aat -~ "*
u A* 0 6
aT--1 = - -=*- -2 0 0 c -6c
0 c -5e
-1 0 c -4e 15c
where uk # 0 and p4 = 0. C -3c 10f
0 c -2c 6e -20c
(e) Examples -c 3c -le
1 0 C -4 15c
(x + A + 1) - (x + 1) 0 -c 5c
c
-6c
fx + 11 _ (x - l)A - (x + l)A 2A
2A 3 0 c
52 3.07 APPROXIMATION OF FUNCTIONS Numerical Differences Numerical Differences 3.09 NEWTON'S INTERPOLATION FORMULA, 53
BY INTERPOLARON GENERAL SPACING
(a) Process Of interpolation is uscd for thc development ofa substitute function y(x) which closely (a) Newton's divided-difference formula in terms ofx ,yk defined in (3.08a) is
approximates amore complicated function^x) in agiven interval and coincides (collocates) with
y(x) at certain specific points. Thc substitute function of a polynomial form in terms of finite r(*) =7o 4- (x - x0)A + (x - x0)(x - x.JA^ + [(x - x0)(x - x,) + (x - *,-,)]A I
differences is called the finite-difference collocation polynomial. Collocation polynomials of finite-
difference type are ofgeneral orequidistant spacing. All formulas in (3.08) to(3.13) are exact and u a - 7i ~yp , Al2
a _J[2 2s. ,...,*-
a -^y. -yo
yield the same vales if the function to be represented is a polynomial. If the function is not a where A - _
X| Xq X2 X0 X, X,
'o
polynomial, the remainder R,+, must be added.
A,2 - AM Al3 - A,, A,r - AM
A. = , Aj, . A. =
(b) Newton-Gregory formulas (3.10) are used mainly for interpolations atthe near or far cnd ofthc x2 x, x, X,
interval.
AM - A.. A24 - AM
A = ,Am = ,...,A3, -
(C) Central difference formulas (3.11) to o. 13) are ingeneral preferrcd over those nvolving forward xs x2 x4 - x2 x, X.,
and backward differences and are particularly uscful for interpolations at intermediate points of
thc interval.
The flowchart of the calculation of divided differences is shown in (A) below.
* .V| *i *0 y\ -yo A
y(x) =yM*) +yM*) + + yX{x) + /c,+l =,(x) + R,+} X, y *2 *0 yi -y Al2 x. - X, A,,. - A A
A X, - X| A,., - A,, A21 X., A, - A.w A,3
x - x, (x - x, <x - xt) (x - x,.,)(x - xk+l) (x - xr) *i *i *S ~ *0 ->3 -y **
where A,(x) = fl
-0Xt - Xj (xk - .v v, - x,) {xk - xk_,){xk - xk+x) (x, - xr) 4 . * xO . -.0 Al4 * - x, A - A,, A * X-, Aa - A22 A
l+k
The sum of Lagrange's multipliers is cqual to I for all vales of x in the range of interpolation, 0 -12
and n is the argument in [x,x,], which makes thc valu of the derivative a mximum. 1 -12 1 0 0
2 -24 2 -12 -6 1 -6 -6
2 2
(b) Example. If 4 -60 4 -4 -12 3 -12 -4 :- 1
yk -12 -12 -24 -60
54 3.10 NEWTON-GREGORY INTERPOLATION Numerical Differences Numerical Differences 3.11 EVERETT'S INTERPOLATION FORMULA, 55
EQUAL SPACING
FORMULAS, EQUAL SPACING
(a) Argument. If the increment of x is a constant vale Aso that xk = a + kh, where a = constant (a) Notation. a, A, k,yk are defined in (3.10a)
and it = 0, 1, 2, 3,.... r, then the substitute function may be taken in a polynomial form a = (x a)/A v = 1 u 6 = central difference operator (3.02rf)
(collocation polynomial) whosc coefficicnts can be expressed in forward, backward, or central
differences of ascending order as shown below and in (3.11) to (3.13), where u = (x a)/A, (*> = tt(u - !)( - 2) ( - * + 1) vw = v(v- l){v- 2)--{v-k+ 1)
<> = ttuw = U(H _ !),...,<'> = u(u - 1)( - 2)- ( + 1 - r)
(b) Open-form formula (r = 2,4,6,...)
and y0 =f(o),yl =f(a + A),... ,yk = f(a + kh),... ,y, =f(a + rh)
(b) Newton-Gregory forward difference interpolation formula n standard form in terms of and j*
defined in (a) as
where_>0, 62y0, 6*y,... ,yx, 2!*,, ^y,,... are the numerical vales taken from thc difference
y(x) =y0 + A^0 + A2y0 + + A>0 + *,+ 1=y(x) + Rr+i table (3.04/) according to the flowchart (e) shown below.
where r; is an argument in [a, a + rh] which if inserted in the (r + l)st derivative of/"(x) makes ( 4- l)(v + 1) (a + 2)(u + 2)
C, =
the valu of this derivative a mximum. Co~2-3 C' 4-5 6-7
(a + 3)(t> + 3) (u + k){v + A)
y(x) =y0 +u(Ay0 - ^ (*% ~^ (a>0 r 1 C, =
8-9 C* =(2/t +2)(2* +3)
For CA,
r^))) A, = ( 4- 2)<5V3 + (a + 2)>,
A, = (v + i)6^o + ( + i)*!ri
where Ay0, A'2y0, A3y0,..., A'y0 are the forward differences of first, second, third,..., rth order A3 = (v + 3)0% + (a + 3)6*>, A, = (* + *)<52> + (a + k)62tyt
at x0 [table (3.04</)].
(d) Remainder in (A) and (r) is
(c) Newton-Gregory backward difference interpolation formula in standard form nterms of uandyk
\R,\ |("+H2r)(,>A>ax /?/(!?)]|
defined in (a) is where r; is defined in (3.106).
<> ( + n<2> ,
y{x) -* 4- UVy0 +{U+2ll) V% +-+^(a + rJ-^^o
- l)("
+* =Ax) 4- /?,+ l (e) Flowchart
yk *bk >yk Kn
wherey0 = /(a) and thc remainder Rr+l can be cstimated as
-2
-1
where r/ and max/)''*0/^) have the same meaningas in (A).
0 yo V>
whereVy0,VVj, VVj,... , Vy0 are thc backward differences of first, second,..., rth order at x0
[table (3.04<)].
-HL
' 1 13 3 1 J
Numerical Differences 3.13 STIRLING'S INTERPOLATION FORMULA, 57
se 3.12 BESSEL'S INTERPOLATION FORMULA, Numerical Differences
EQUAL SPACING
EQUALSPACING
(a) Notation. a, A, A, jt are defined in (3.10a)
(a) Notation. a,A, k,yk are defined in (3.10a) = central difference operator (3.02a*)
a = (x - a)/A
_ (x _ a)fh v= - o = central difference operator (3.02a1)
<" = a(u - l)(a - 2) ( - * + O
(1> = u(a - l)(u - 2) (-* 4- I
(b) Open-form formula (r = 2,4,6,...)
(b) Open-form formula (r = 2,4,6,...)
< ,<> (a + 1)(S) (a + l)w i4 _,_ (a + 2)(M
um ut2) {u + i)<4>, (+ ir' o y(x) =y0 4- "-<< +^o +Si "' +Sr";" +-iT"0'
j(*) =foo +*> +^' +1TA-' ~T^"2 + 4! 4 + 5! ^
(a + r-5)<-2) ,_, , (u + r-4)"-"
+ +
(a + r- 5)"-2) ,
A,_- +
(a + r - 5)(-2) ftf-,
: "0 ym + Kt
^ +- + (,-2)1 " * + (r-i), - +*
(r - 2)! " (r ~ 1)
where *,% o>1/2,s7I/a. ^,,2,. ,>_l/2, >.,/2 5>.1/2 are the
where.,,62y0,6%,... ,yt, 6% ,\ .... and 6yU2, fym,.. .are the numerical vales numerical vales taken from the difference tablc (3.04/) according to thc flowchart {e)
taken from thc difference table (3.04/1 according to thc flowchart (e) and A2, A4 ,... arc the and o>,, t3, a>5.... are the equivalcnts defined in (r) below.
equivalcnts defined in (c) below.
(c) Nested-form formula
(c) Nested-form formula
y(x) =4U +,.) 4.(*,* +C,(>, -5(<*. +C,(A4 +^(6tv,, +)))) +* A*) -*4- .(, 4- i(S +C2(o,3 +l (*,. +C4(, +"- (6% + )))))) +*,
where
where a2- I a2 - 22 a' - 32
(2 + a)(3 - a) C = C = lu
a(l ~u) 1 + u)(2 - u) 3a 5a
C, = - C, = - C, = -
4v 6v
2v
a2 - k1
(3 + a)(4 - a) (k - 1 + u)(* - a) 2* (2* - l)u
Q=-
2*
81/
<Wi = 2(^-1/2 + ^Tin) <o3 = ^-i* + *^i>
A2 = (2jo + *7i) A4 = (*>o + >i)
">-, = 5(*5>-i/s + ^1/2)
A6 = (67o + *!*) A* = (o'% 4- "jr.)
(d) Remainder in (A) and (c) is
(d) Remainder in (A) and (c) is
s |< +ttM +iHt'-'rytMMi)yW1|
l^l^jiJLil^A'jrnaxZ)/^)]!
1 " I r! I where r/ is defined in (3.106).
where 7} is defined in (3.106).
(e) Flowchart
(e) Flowchart
fyk V 6\ 6\ 6syt
k yk 6yk 6\ V 6*yk V
-2
-2
-I
-1 6y. V *V2
0 yo 6\ y0
0 tro **7i
yo ' 1 J 1 i ' &y\n
^1/2
A*. 11 <5V
..
+ 1
r- 1
V. 1^ + 1
+2
+2
7
i i 4
k X yk = sin xk 6yt V fh v
where thc remainder by the formula given in (3.08c) is (f) SubtabulatlOn is thc process of computation of vales corrcsponding to smallcr intcrvals than
|/c4| < |0.065(0.000 795)| = |5.2 x 10^1 thosc listed in a tablc.
1 | 1 1 J J
J. (c) Intermediate-value interpolation in yk.2, yk-,, .n* *+i *- at xk < v, < *1+l is given
(b) Remainder of this approximation can be cstimated as convcniently by the five-point Lagrange's formula (3.08c) as
= b_ / yk-2 _ 4^., + fin + 4yk+x _ yk+, \ + ^
R, =J, -Sf ="-^-^^[m^s D2f(n)\
'1
1 * X
where n is so selected as to makc thc second derivative where 1< k< n- 1 [figure (3.15f)J, a = (*, - **)/(*, +, - xt) and A= (2 - a)(l + a) X
a mximum. If/(x) is not known, its second derivative a(l + a)(2 + a).
can be approximated as
(d) Far-end-values interpolation in y0,yltyt,... ,y.-*,y.-i ,J. at xk.t < xr < xk is given con
^ImaxZ)2/*!,)]*^ vcniently by the Newton-Gregoy formula (3.10c) as
(C) n-place tables. Ifthc vales y. ,y, are listed in n-dccimal-placc aecuracy, then thc condition of
yf =*-. =yk + *.(*>* + B9{Vyt + B3(V3yk + BXyk))) + *, *-e;>
permissibility of linear interplation is
where kis nor n- I (figure (3.15rf)], = (**- *,)/(** - **_,) and Bx = u/\.B2 - (1 + a)/2,
, = (r-l+ u)/r.
^"|maxZ)2/(r,)|<^x 10" (e) Examples. The numerical vales of, = sin 0 are given for 0 =8 9,... ,13"(the difference
tablc shown below. The end valu sin 8.345 and the intermedate valu sin 10.345 are desired.
wherc thc second derivative can be approximated as in (A).
62 3.18 INVERSE INTERPOLATION Numerical Differences Numrica! Differences 3.19 BIVARIATE INTERPOLATION 63
(a) Inverse process. Interpolation is the process of finding/(x) for a particular x located between two (a) Functions Of tWO variables can be interpolated in two ways. The simplest method is to first
known vales. Inverse interpolation is the process of finding the argument x if the valu of f{x) is interplate with respect to one variable and then with respect to the other, using any onc of the
standard formulas of (3.08) to (3.13). A more involved but more direct approach is by thc
known.
bivariate interpolation formulas presented below.
(b) Inverse linear interpolation. If in (3.16a) a,A and ym,y,jt are given, then the desired p is
calculated as (b) Newton's bivariate interpolation. With constant spacing p and q in the direction of x and y,
rcspcctively, the forward differences of z f{x,y) are
p = 6- R= a +%^s(b -a) - R 2(A - a)(yt -y.)
A,4 = Ax + P,y) - Ax.y) = (A - i )z
yk - y .
Ao, =Ax,y + q) -Ax,y) = (, - 0*
(C) Example. From (3.16a*), y. = sin a = 0.139 173, yf = sinp = 0.145 133, ^ = sinA = A* = Axo[f(x,y + q) -f(x,y)\ = (, - \){E, - \)z
0.156434, and a = 8, A = 9. By (A) above,
= Aoil/X* + P,y) -Ax.y)) = (E, - D(, - \U
0.145 133 - 0.139 173
p = 8 + (9 - 8) - R = 8.345 280 - R Aao = A10[/(x + p,y) - f(x,y)\ = (E, - l)2z
0.156434-0.139 173
A** = Aox[f(x,y + p) -f(x,y)\ = (, - l)h
(8.345 280-8)(8.345280-9) r_/M2s.n90-| = x ]Q.
2(0.156434-0.139 173) L \I80/ J
where EMz = f{x + p,y), E\z = /(* + 2p,y),..., ET = f(x + np,y)
(d) Inverse Lagrange's interpolation. If in (3.08a) x andy are interchanged, then
Eyz =Ax,y + q),.E$z = f{x,y + 2?) E",z =Ax,y + ?)
P = xoX^y,,) + xMyfi) = + xMyfi) With a = (x - x0)/p,.... uw = a(a - l)(a - 2) (a - k + 1)
where
9 = (y -y0)/q, ,w = < - )(" - 2) ( - * + 0
k \_y*-x_<*-r-'{-v"-y o~.y*-')U-.>w.) u-*)
1Uo M* -Ji U -.oH.n -*) U -*-,)U -*-.) ' U - *) the substitute function becomes
where v = {y, -yk)/(yk+i ~yk) and r = (2 + a)(l + v)v{l - v)(2 - )/24. +-,t (?)<-'VA1I.wl/(*^o)
n!*-o\*/
(f) Example. If_y_2 = sin6j., = sin 7,_y0 = 8,j>, = sin9,^2 = sin I0,.v, = sin/> = 0.145 133, where x = x0 + up &ndy =y0 + "?
v = (0.145 133 - 0.139 173)/(0.I56434 - 0.139 173) = 0.345 274, and
(c) Multipoint formulas. If the vales of z are known
c = (2.345 274) (1.345 274) (0.345 274) (0.654 726) (1.654 726)/24 = 0.049 175 f**l
at certain points of the grid shown in the adjacent >
64 3.20 ORDINARY DERIVATIVES IN Numrica! Differences Numerical Differences 3.21 ORDINARY DERIVATIVES IN 65
TERMS OF DIFFERENCES TERMS OF 0RDINATES
(a) Approxlmate derivatives of a tabulated function can be found at any point of its interval of (a) Approxlmate derivatives ofa tabulated function can be also found by expressing cach difference
continuity by representing this function by a selected interpolation polynomial y and in (3.20) intcrms ofthe tabulated valesy0 ,yx ,y2,... ,yk-, ,yk ,.T*+ O"* *.-i O*. * shown
differentiating it with respect to a. Thus the first dervave of y = /(*) tabulated in constant
below in (A).
intcrvals of length A is
dy dy 1 dy n_
(b) Examples. From (3.20A) in terms of(3.02a*), (3.02), thc first-order dcrivativc approximations
-- = Dy as = - = Dy at 0 are
dx J ax hdu J
and in general, its nth derivative is [^]o =(A,o)/A =-(7o-j-1)/A
= /; = = = D v a (Av - l2A2y0)/h = -(3^ - 4yx +y2)/h
dx" J dx" h"du"
= (Ay0 - {A2y0 + A>0)/A = -(-II*, + I8r, - 9y2 + 2y.,)/A
where n = 1,2,3,...,/) = derivative operator (3.02a*), a = (x x0)/k, and j> =/(*) is one of
the interpolation formulas (3.10) to 3.13). where the error is 0(A), 0(A2), and 0(A3), rcspcctively.
(c) Forward difference approximations in tcrms ofthc first difference only arc
(b) Doperator in terms of A series
hDyQ -1 1% 0 0 0 0" yo
hD 'ln (1 - A)" "l _i
'N~:>
i
5
_i
5
i
5 A
h2D2y0 1 -2 ^1 0 0 0 y\
h2D ln2(l + A) o Xi,-i J -I A2
-3^1 0 0
h3D\
h3D ln3(l + A) = o oxi,-| l A3
h*D%
=
-1
1
3
-4 6 -4^1. 0
y-i
+ 0(A)
y*
h*D ln4(l + A) 0 0 0 ^L -2 A4
h5Dsy0 -1 5 -10 10 -5 1 y*
h5D Jn5(l + A)_ .0 0 0 oN A5
h3D
-ln3(l -V) o oxi | | 12
8
22
IS
V3 h4D*ym 0 1' -4 6 -4 1 y*-*
where V = backward difference operator (3.02a*). (e) Central difference approximations in terms ofthc first difference only are
h*D\ -4 6 -4 1 ^ yk + 0(AJ)
where A = j and 6 = central difference operator (3.02a*). The series in terms of are then
=
\
Ni
hbD*yk -Y 2 -2.5 0 2.5 -2 2 jk+i
32.
h2D (2 sinh"'A)2 0 4. 0 -i 0 >k + S
h3D = (2 sinh-'A)3 = 0 0 ^8 0 -4
h*D (2 sinh-'A)4 0 0 0 16 0
(f) Errors ofthcse approximations are oforder Ain (c), (d) and oforder A* in (/). Results should be
uscd with caution. and even when very accurate input data arc available, the aecuracy diminishes
hsD - -
(a) Forward differences of>i =Ax + t:h] mtcrms of D- D'*> D*> arc
~1
"aJ4 1 1 I 1 1 1 hD/V.
2 6 14 30 62 h2D2/2\
A'!)'. 0
0
0
0
0
0
24
0
240
120
1560
2440
h4D*/4\
ksD$/5\
yk
4
0 0 0 0 720 h"iy/6\
AS 0
(b) Backward differences ofthc same function in tcrms ofD, D2, D* are INTEGRALS
"Vv," 1 -1 1 -1 1 -1 AD/1!
0 2 -6 14 -30 62 h2D2/2\
v>
v>4 0 0 6 -36 182 -540 A3/>3/3!
J*
VS 0 0 0 24 -240 1560 A4Z)4/4!
% 0 4 8 16 32 64 (|*)*2>721
6% 0 0 24 78 240 726 (U)3Z)3/3!
yk
s\ (1 0 0 192 896 3840 (A)4Z)4/4!
f(x) dx = lim {f(a) +/(a +A) +/(a + 2A) + +/[a + (n - 1)//]} ff(x)dx =nF(T) +F(a) - (n +I)F(O)
= }'[n,] = [F(x] + CJI= FA) - F(a)
where dF(x)/dx = f(x).
where n = positive integer and h = ib a)/n.
(d) Antiperiodic function. Ifn, Tare thc same asin (c) and
(d) LimitS. Thc signed numbers a.b which may be 0 or ) are called thc lowcr and upper limits g(x) is anantiperiodic function so that[(2fl - 1)7" + x]
ofintegration, and thc closed interval [a,b] is called the range of integration. = -g(x) and g(2nT + x) = g{x), then for A=
(2n - \)T + a and c = 2n7" + a,
(e) reas bounded by f(x) may takc onc of the shapes shown below. /W
g(x)dx = GiT) - G(a)
i
i .'
i
1
a
i
-Ah
'g(x) dx =G{a) - C(0)
Jo
\ /I) where dG(x)/dx = g{x).
""'' IT
v4
X
/i)
X
1^1lSrt (e) Change in limits and variables. Uf{x) is integrable in
/le)
/|6l [a,b] and x = git) and its derivative dg(t)/dt are
,
-< >-
\ continuous in q < / S /3, then with a = g(a:), b =
(f) Geometric interpretation. Thc definite integral introduced in (c) above cquals numerically the ,4g(t]
rea bounded by the graph of/(x). thc A" axis, and thc ordinales/(a),/(A) as shown in (e) above. |* f(x)dx =j AgC)]- f/ dt
The sign of thc rea is governed by the sign of/(x). If thc graph interseets thc A' axis onc or
scvcral times inside thc interval [a.b], then thc definite integral equais the algebraic sum of all Particular cases o( this transformation are:
arcas above the axis and of all arcas below thc axis. If the sum of arcas above the axis (+ arcas)
and thc sum of reas below thc axis (- reas) arc numerically equal, then the definite integral
equais zero. ff(x) dx = ff(b - x) dx
Ja 'o
//(*) dx =f/(a +b- x) dx
* *
(g) Example. If> = sin* and he closed interval ofintegration is 10, n], then
I sinxx = [-cos.*]o = ii
Jo ff(x) dx = f f(x +a) dx =(A - a) I /[(A - a)x +a] dx
since thc arcas above and bi low the X axis arc +1 and I, rcspcctively.
J J I i 1 I " | 'I ~|
70 4.03 EVALUATION OF DEFINITE INTEGRALS Numerical Integris Numerical Integris 4.05 NEWT0N-C0TES FORMULAS 71
(a) Numerical integration. Whenever the closed-form integration is too involved or is not feasible, (a) ClOSed-end formulas based on (4.04A) are in symbolic form
thc numerical valu of a definite integral can be found (to any degree of aecuracy) by means of
any of severa!quadrature formulas which evalate the definite integral as a linear combinaon of
a selected set of ntegrands in the range of integration. ff{x)dx =2k'yk +R,
Ja k-0
where x0 = a, l = A- a, h= f/r, x, = a + rA. The particular brms ofj\'} and R, are shown in
(b) Quadrature formulas fall into four basic categories: numerical integration in terms of ordinates (A) below. Ifr + I is odd andj(x) is a polynomial or + 1or less degree, the result is exact. Bul
(4.05), (4.06), in terms of finitc differences (4.07) to (4.09), in terms ofasymptotic series (4.10), if r + 1 is even, only r-or-lcss-dcgrcc polynomials can be exactly evaluated. Consequently,
and in terms of orthogonal polynomials (4.11) to (4.16). odd-point formulas yield a higher degree of aecuracy.
1 k
/
2
/
2
-&*<,)
2 2A
/ 4/ /
->/(/)
4.04 LAGRANGE'S FORMULA 6 6 6
75/ 19/
where^0 fyx ,y2 > - JV arc ^c vales otjk =f(xk) at k and 75/ 50/ 50/
5 5A
288 288 288 288 288 288 12096 M'
? =\\k{x)dx
}.
=*(n *^u =k
i. \j-0Xk - XjJ
j+k (C) Open-end formulas. If the end ordinales y and y, are undefined, unknown, or arc singular
points, the numerical evaluation ofthe definite integral can be expressed as
where k,j' = 0, 1, 2,..., r, kk(x) = Lagrange's multiplicr function defined in (3.08a), and
sdkr) = weight which is symmetrical with respect to the center of the sequence.
f(x) a** = '>, + R,
i-i
(b) Equally spaced ordinates. If the spacing ofy0 ,yx,y2,... ,y, is constant so that A= (A a)/r where a, A, A, / are the same as in (a) above and the weight factors <" and thc remainders Rr are
and u = (x a)/h, Lagrange's factor becomes
tabulated in (d) below.
f/(*) dx =f /(*> dx +"/(*>*+ +jf+(._1(/W*=.?, f*l<,'a +**] +*" Cf(x) dx =l{y0 +4yx +2y2 +4y3 +2y4 + +2ym_2 +4ym_x +ym) +/?.
wher =integer' A=(A - a)fn =l/n, Y[ ]=approximation ofthc respective definite integral
based on onc ofthc formulas in (4.05), and
nA*
R, s R[a,a +A] +R[a +A.a + 2A] + + R[a + ( - I)A,A] where R. a -]^5Z>V'(?) (si)
,*
is the sum of aii remainders.
i A practical estmate of/?, is Rm * ^(fj. - fj s - (A>0 + V>)
(b) Trapezoidal rule approximates the arca under/(x) by aseries ofn
trapezoids formed by ordinates yk,yk+, and thc constant width 1 , .-1 where the symbols have analogical meaning to thosc in (A) on the opposite page. This formula
A= l/n as shown in the adjacent figure.
yields exact results f/"(x) s a polynomial of al most third degree.
><
*l * A A
(e) Example. The application of (d) is llustrated as a comparison with (c) by
Cf(x) dx =A(iy +_y, +.r +" +J.-i +b.) +R- | k k
1 /-i--nA
k k k X
Apractical estmate oTR. is * =5(^ ~ ) ="24 (A^ +^^ ?B + * s 1.000008 295 + ^(, - f4) = 0.999 999864
where =quadrature based on nsubdivisions of/, f2 =quadrature based on 2a subdivisions of where f4, Ka ^4- #a> ^- arc tabulatcd below.
/, and 2y0, V2yK = second-order differences at 0,, respcctively. k A4 = */8 For Y k A = */l6 Fori,
cosO 1.000000000
(C) Example. Thc application of(At is llustrated by 0 co$0 1.000000000
0
}8 +RB s 0.996 785 172 +^A8 =1.001 831 767 3 4 eos 3AB 3.325 878449
5 4 eos 5A 2.222280932
where ?4, ?8, *4, *a. *, are tabulatcd below. 3 4 eos 3A4 1.530 733 729 6 2cos6AH 0.765 366865
Numerical Integris
4.08 NEWTON-GREGORY INTEGRATION 75
Numrica! Integris
74 4.07 INTEGRATION IN TERMS OF DIFFERENCES FORMULAS (Continued)
(al General case. As notcd bcforc in (4.036), any interpolation difTcrencc polynomial can be used for
(Ihc dcvljmcm of numrica, integration formulas. Symbo.ica.lv, onc of hese polynomials , (b) Numerical constants (k = 1,2,3,.... n)
taken as
d =it!.,m
-^ ,*+,[,<> +A] =A'+,^r+1Z)'+l/(r) (a < n < a + A)
v = g<lY. + .(")>>. + sA)"y. + + *<)>. + 1
where 0<u< l, a'y, =th-order dTcrcncc oCy such as A<y V% or 6% &() =Tunction > - A,a] a -A'+,8r+I/),+,/(l?) (a-A<n?Sa)
factor in u, and i is a subscript specifie for a particular point.
(b) Definite integral 0> =/<*) u, [..*] " ms of =(* - )/(! - ) can be approximated as %= sk + % R%M ~h,a + h] mA" "g,tl/r+l/(i/) (a - A* if =s a + A)
where S**? = Stirling's number ofthe first kind (2.11a), (A.07).
ff{x) dx ={b-a) fi() rfu +ft,"+l =(b - a) Xka>y, +**h
Jo J" '
where thc change in limits is based on the relation (4.02) , (c) Backward difference polynomial. Ifjr() in (4.07A) is expressed by (3.10c), then thc dcBnite
integris of/(.v) in [a, a + A], [a - h.a], and [a - A, a + A] becomc, rcspcctively,
Xt = Jo gk()du
is a specifie constant for cach formula, and f+I is the remainder shown in he subsequent j"*/M dSf-*[i+| +^ +*Y+-" +^ v']-v" +|/??+,Ifl*a +
sections in this chapter.
(C) Example. Ifthe Newton-Gregory forward diUerence interpolation polynomial (3.10) is used, /(,) ^-^i-l--2 Kl V'], - \RU,ia ~h,a
'ala - .)( - 2)---(u - k + \] du a - A,a + A]
cy<*i
i /y;" y," where \rfk\, |S94|. |%| for k> 1 and the remainders \R?+t[ || arc the absolute vales of their
and = 1. siy = 5. ^ = ^"2" + 3 +'-+A + . counterparts in (a) listed in (A) above.
, ,,*> *> ^> are the Stirling's numbers of the first kind (2.11a), (A.07) and the
pScuS ^uo'rkrollow in ,4.08a). Similarly, ifthe Newton-Gregory backward difference (d) CompOSite formulas for [x0,X0 + nh] are Ibrmally identieal, but for k> 1, their constants take
polynomial (3.10c) is uscd, on the following general forms:
a(a - l)(a - 2) (-* + 1) du 1 * nm+x&" 1 fn2^4 n*Sfi? nk +i<nl)'
3*; = *= I QrfH = k\
*-f0*-5S + 1 A!L 2
+
3
+ +
k + 1
Cfik
i /y1,41 y," ii+i_"* <*<*) i fH-'y1/1
r,.-V1> 's*^
i ' (-vir^__j_rryr_ ,,W (-n)*+l9*tk)~>
and 38,,= 1, , = "5,- -. =j (Y" ~^ + "' ' +( ' k+ 1. ,)du = --
J_V*/" 2.
*! ~, I 1
m+ , -
+
A! L 2
*!L 2 3
A+ 1 J
where the symbols in k are the same as in thc calculation of k. and %= &* + A- The remainders must be expressed in tcrms of j4,+x, S3r+1. and %+l of this
Jrkmdx
J\x)ax =h(y*+yi
ny ^ _^ +- h. +...%x\
144Q ,20960 ^ *) +Rl+2[a,a +a]
(b) Stirling's difference polynomial. Ifj(a) in (4.07A) is expressed by (3.13a), the definite integral of
f(x) in [a h,a + AJ becomes Thc tcrms in /_. first decrease and then begin to incrcasc. Thus thc most accurate rcsult is
obtained by stopping with thc term G-,, preceding the smallest one G.,,+., which is Icfl for thc
remainder.
j Ax)dx =h(2 +j--
-' *
+^-
"
'.-..Vi-, i A';. .:,
h2B.i> Irw, A4,(u:1 h4(03
vhere ror k 2: 1, p 2A - 1. G-2 2! 12 4 " 4! 720
(d) Corrected Simpson's rule in terms of (A) above is " oosx =0.987 115 801 +- (l +- (l +^)) +R\ =0.999 999 999 6+ff|
f* A&' "<o / <54 "
where 0.987 115801 is thc rcsult ofthc trapezoidal rule given in table (4.06c). Since
A-< -1
: G.2 = = -0.018.151 047 4 (? = -=-0.000 000 1212
12 30 240
where 4^, =^-1 _ 4-a + ^',-1 ~ \Yi +.Vi A4- + l) ah(+i: = -0.000 000 000 4
= -0.000 C33 029 8 C= -
!y, =J>/-< - -V:> + 'QV2 - 10->Vi + %& -.>>i c- = --S 1 209 600
is monotonic,
and n = even integer. R* is a better remainder than Rm in (4.06a"). The spacing Ais thc same as in
R* <\-GJ = 0.000 000 0(0 4
(c) above.
J f l i l i l
(b) General Gauss formula introduced below in symbolic form is thc model of this type of 0.555555 555 556
0.888888 888 889
-0.183 434 642 496
0.183 434 642 496
0.362 683 783 378
(weight), xk = coordinatc, both given and tabulatcd in the following sections of this chapter, and 0.236 926 885 056 -0.324 253 423 404
0.000 000 000 000
0.312 347 077 040
0.478628 670050
C = scaling constant |for illustration refer to (4.12a)|. 0.568 888 888 889 0.324 253 423 404 0.3I2 347 077O4O
more direct and simpler approach is to compare thc results of scvcral approximations, each 0.968 160 239 508 0.081274 388 362
involving diflerent number of tcrms (r = 3,4,5,...). If this comparison leads to w.dcly d.flerent 0.171 324 492 379
0.360 761 573 048 10 -0.973 906 528 517
results, thc singularities or steep oscillation may be present. 0.467913 934 573 -0.865 063 366 689
0.467913 934 573 -0.679 409 568 299 0.219086 362 516
-0.433 295 394 129 0.269 266 719 310
0.661 209 386 466 0,360 761 573 048
0.171324 492 379 -0.148 874 338 982
4.12 GAUSS-LEGENDRE FORMULA 0.932 469 514 203
O.I48 874 338 082
0.129 484 966 169 0.433395 394 129 0.269266919310
(a) Integration formula. With thc transformation .
where
where u4 are thc zeros of the following polynomials:
(2k - \)n
G2{u) = u2 - - Gs() =s-f34-^ Ga() - 6 "* +i"* *T5 2r
(A- = 1.2,3,...,r)
r = 6 r = 7
r = 5 U| = -0.965 925 826 u, = -0.974 927 912
u, = -0.951 056516
-0.866 216 818 -0.883 861 701 u., = -0.707 106 781 u., = -0.781831 483
, = -0.832 497 487 "i = ., = -0.587 785 252
. =
-0.422 518 654 u2 = -0.529 656 775 u] = 0.000 000 000 u] = -0.258 819 045 u] = -0.433 883 739
., = -0.374 541 410 ". =
-0.323 911811 u, = 0.258 810045 u4 = 0.000 000 (KH)
u~ = 0.(KK) 000 000 i = -0.266 635 402 "i = < = 0.587 785252
0.000 000 (KH) u5 = 0.707 106 781 u5 = 0.433883 739
u, = 0.374 541410 "4 = 0.266 635 402 "< = u5 = 0.951056516
0.422 518 654 0.323 911811 u,; = 0.965 025 826 u6= 0.781831483
"i =
BS = 0.832 497 487 U., =
0.529656 775 u-= 0.974 927 912
"l. =
0.866 246 818 "6 =
"7 = 0.883 861 701
= 8
r = 9
u, = -0.195 090 322 u- = 0.831 469 612
U| = -0.980 785280
ut = -0.167 906 184 "7 - 0.528 761 783 u = -0.831 469 612 B5= 0.195090322 ul = 0.980 785 280
u, = -0.911 589 308
u., = -0.601 018 655 "i = 0.000000 000 "B =
0.601018 655
u\ = -0.555 570 023 Ufc = 0.555 570 023
0.167 906 184 u9 = 0.911 589 308
~ = -0.528 761 783
(c) Example. If<i = 0, b= \.t. r= 4,f(x) = acosa-, xk - }jr(l + ut), )TWO auxiliary cases in which thc argumcnts uk and the weight factors s, are given in closed
x, =0.161 278 x, = 0.638 063 x, = 0.! 32 733 x = 1.409 519 form are
/() =0.159 185 ./ *.,) =0.512 524 /(x3) =0. 55 524 /(x4) =0.226 340
[VT^^^-T^rrSsm _^tj + 2,(2r)
and by (a) above,
f xcc-sx* =?2/.xt) +R=0.570 836 -0.000 027 =0.570809 f' /T=l 4,-r A . ., kn I 2kn \ ng{2'\n)
Jo 8*"' 1, VTT;^* =27+-T1?,s,,r27T74cos27T7) +22>w
and the correct valu is 0.370 796, which again shows the ch ieney of the formula even when a
The last term in each formula is the remainder R with -1 S 17 < I.
very small number ofpoints is uscd.
1 r 1 "I "H "1 "1 ~~J ' 1 1 J ' 1 ~1 i i i
82 4.15 RADAU FORMULAS Numerical Integris Numerical Integris 4.16 L0BATT0 FORMULAS 83
(a) Notation, r + I = number of terms k = 1,2,.... r (a) Notation. r + 2 = number of terms k= 1,2,... ,r
P,{u) = Legendre polynomial of order r (11.10o*) P,(u) = Legendre polynomial oforder r (\l.\0d)
,,*,,...,* = roots of[/r() + Pr+,(u)\ = 0 , = (1 + uk) ux,u2t... ,u, = rootsof polynomial [dPr+x(u)/du = 0] "* = 5(1 + *)
(b) Integration formula A. When in (4.116) the interval of integration is transformed to [-1, 1], (b) Integration formula A. When in (4.11 A) the interval of integration is transformed to [-1.1],
w(u) = I, and the lower limit is preassigned as 0 = 1, then w(u) = 1, and the lower limit and the upper limit are preassigned as u0 = 1, ur+x = 1, then
/> du
(r + 1)
!(-') + dig(i) + *a*(a) + + st,g{ur) + R, g(u) du =
(r + l)(r + 2)
L?(-1) + (l)J + ^iK) + ^(a) + + J^(,) + *,
where where
(c) Integration formula B. When the same integral is transformed to [0, 1], w(u) = 1, and the lower (c) Integration formula B. When the same integral is transformed to [0, 1], w(v) = I, and thc lower
limit is preassigned as u0 = 0. then limit and the upper limit are preassigned as v0 = 0, vrJhX = 1, then
1 1
h(v) dv = ;A(0) + 9xh(vx) + ^(i/j) + + &rh(r) + R* h{v) dv = (A(0) + A(l)] + SS,h(ox) + SM(2) + + S8,A(pf) + R?
(r+ 1) (r + l)(r + 2)
where where
(d) Numerical vales of uk, sk and vk, S84 with respective remainders are tabulated below. (d) Numerical vales of uk, sk and vk, 9Sk with the respective remainders are tabulatcd below.
r = 1 b, = 0.333 333 six = 1.500000 p, = 0.666667 38, = 0. 0000 r = 1 b, = 0.000000 si, = 1.333333 , = 0.500000 3B, =0.666667
r = 3 b, = -0.575 319 si, = 0.657 689 p, = 0.212 341 38, = 0.328835 t = 3 b, = -0.654654 si, = 0.544444 v, =0.172673 38, =0.272 222
b2 = 0.181 066 si2 = 0.776387 p2= 0.590 533 3&V = 0.388 194 a2 = 0.000000 J2 = 0.711 111 p2= 0.500000 SBj = 0.355 556
b, = 0.822834 J, = 0.440925 p, = 0.911 417 38, = 0.220 463 a, = 0.654654 j, = 0.544444 p, = 0.357 384 38, = 0.272 222
Numerical Integris
84 4.17 GOURSAT'S FORMULAS
(a) Notation. u= u(x), v= r(*) = diflerentiable functions ofx
h',",..., 1 _ succcssve derivatives ofuand v, rcspcctively
4(sinfl - Ocosfl)
0 = 03
where/(x) is called the summand and A-' = Eis the sum operator so that AE - 1. (c) Binomial coefficients (A = l,x * 0)
(C) Limit theorem. Thc relationship ofthc indefinite sum to thc indefinite integral is given by
a-q-s'<'-'k-v'"^',-c:.)
lim [*/(*) +C(x)] = \f(x)dx +C
kO J
w=g(x + A) *""' = x(x - A)(x - 2A) [x - (m - 1)A] (e) Logarithmic functions (x > 0)
A-n
A ' In x+h: = ln
i x A 'iog -= log
and C(x) = periodic constant, which is omitted but implied. x A x A x A x A
i x A log = log x
a dx - ax A-a = - x
(.+)
dx = A" W =
(m + 1)A (f) Trigonometric functions
m + 1
(m+l) Arl sinAx = -/Veos [A(x - A)] A-' eosAx = TV sin[A(x - j-A)]
(a + bx)m* , , (O + **)
(a + Ax)" dx = A~'xsin Ax = -Nx eos[A(x - A)] + tfh eos[A(x + $h))
{m + 1)A
(WH). For thc'cvalua.ion of this form,* ,a, ., and Amust be g.vcn. = 100(49 + f (32 + f (8))) = 1456 900
49
32
81
where yt = 49, A* = 32, A2* = 8, and A3yx = 0 are 40 1 8
(d) Example. The sum ofthc sequel in (b), taken from the diflerence table given by the first four terms 121
48
8 0
where <p'(x), <p"(x),... are the derivatives of<p(x) and the diflerence series terminates with A'+Iu,
(e) Change in limits. As in (4.02), for r < n. For CA,
A-1 sinx ~
eos 33 eos
= -3.460 586
<'+'>H,+7 +7+f>'
2 sin 3
1 r~T
90 5.05 APPROXIMATE SUMMATION OF SERIES Series and Products of Constants Series and Products of Constants 5.06 GENERAL TRANSF0RMATI0NS 91
(a) SuperpOSitlon. Since for many series their sums are known or are available in collections of
(a) Notation formulas, the rclationships
a, A=constants k, m,n,r =integers A=(A - )/. =spacing
B2, = Bemoulli number (2.22a), (A.10) Rm =remainder tx = 2* - 2 yk =s. - 2 -v,
k-0 *-0 *-+! -.+l
(b) Sum Of finlte series. If.) =/(*) is single-valued and continuous in [a, A] and* =/(a +AA),
2(-i)** = 2(-DS- 2 (-DS--- 2 (-ib.
^ SI /(x)a-x +\{y0 +y.) +*. +*. where 5 and S. are the known sums.
B 2m+2
- = n D2"+2An) (a<T}<b) then the sum ofthe first 100 terms is
(2m + 2)!
99 - 1 A100 I - ,0
Ifflx) is apolynomial of mor m+1degree, <!>., terminates with i, the remainder Ris zero and 2*4 =2*'- 2 *'=t^-
!n{:umlexyact. If/(x) is not a polynomial, he series is asymptotic and the sum ,s an *-o *-o -ioo I o 1 1 I - b
(C) Sum Of monotonic infinite series. If A=and all terms of thc series are positive, then I -A'
I'-"v -1'-"' - ..(-,)V" T "TTi TTT
2\ 6 360 15 120 ^ (c) Resolution in partial sums. I the th term ofseries Ak can be expressed as a sum oftwo orany
I,-*A
.2\ 6
+lA2--^A>
12 360
+---k
r
number of tcrms, so that
Ak = Blk + Bu + + Brk
where the remainder can be estmated by the second formula in (4.03) and (4.04), respectively. x> a
y. =2 2 B2k + + 2 B
(d) Sum Of alternating infinite Series. If A=- and the signs of terms in the series altrnate, then then
k-0
Ak
*-0
Bn +
1-0 *-0
where r = 2,3,... and thc partial sums of Bxk ,B2k,...,Brk are either known vales or may be
f'2\(i - H-D
2 +-D3
24 - &+-^-rD1
240 40320 - U
^ evaluated by the rdations of the preceding sections.
A-o i(i --A+ -A3 --A5-A7 - -bo (d) Examples. Ifn < ,
L2\ 2 4 8 16 y
where the remainder is cstimated as in (c) above. 2 (fl + kbf = 2 (a3 + 3a2AA + 3ab2k2 + bV)
-1 *-i
f -n = 0.616666667 + 2 (_,)rTT
2* = 2*+ 2 yk
1 , 1 3! 5!
= 0.693 147 125
= 0.616 666 667 + +
(2)7 (4)72 (48)74 (480)76 where the first series in A = 0, 1,2,. .., s can be summed term by term and thc second series can
be evaluated by a sclcctecJ.
Since the correct vale is In2 = 0.693 147781, thc error is * |6(I0)-7|. formula.
1 ' i i "11 ' "'""-I ' """1 73 -f j ~i i i ---i -i --i -!
Series and Products ofConstants 5.08 ARITHMETIC AND GEOMETRIC SERIES 93
92 5.07 SPECIAL TRANSFORMATIONS Series and Products of Constants
(a) Notation. a,A, c,A= signed numbers n = number oftcrms
(a) Kummer's transformatlon. Given two convergent series, A = 1,2,3, ...,n Ak , Bk, Ck,Dk = Ath terms m = 1.2,3,...
0 = (-l)1+l _ 0
ro ifn iseven
ifn i
A. = 2 ak B=2bk
"ti ifn ii s odd
at
where the sum Ax is desired and the sum Bis known, and if lim f - A* 0 (b) Monotonic arlthmatic series (n < )
then , =AZ? +2(l -A% =Afi.+ fr, 2 Ak = a + (a + A) + (a + 2A) + + la + (n - 1)A] = nyt, + A.)
where the series of ck may be more convenient for numerical evaluation. This transformatlon is Ak = a + (A - 1)A 4*+i - i4 = A n = I +
Am - A,
quite useful in thc cvaluation ofslowly convergent senes.
(b) Example. If (c) Alternating arithmetic series (n < )
A+ I , nh
lim = 1
* A 2 A, = a - (a + A) + (a + 2A) o-la + (n - l)A] = -- + 0{Vm + AA)
|fl.l - B,
then A
-"+4?IFTT i) 04 = B[a + (A - 1)A] |Zf1+l| - [Bk[ = A n = 1 +
(c) Abel's transformatlon. When uk=f(k), vk =g(x), and A-0,1,2, then with j =
(d) Finite monotonic geometric series (n < )
1,2,3,...,
+ a'-'>* = : -1
2 *& =o" +*- *"'
i-o
2 i - *"'2 [Au* *"'5>J 2 .4 = + a** + ab2" + c 1
ln.4. - ln/1,
leads in some cases to a rapidly convergent series, .44 = ab' = A* = f n = 1 +
Ak Inr
100
e"' - A, = 1
2' = ln |/*1 - ln B,
2' = - 1 7-1
- 1 fl4 = /3aA'*-"* i = -A* = c n = I +
4-1
5, ln|f|
100 f! - <- ' - t f - e emi - e
and S^-IOI-T^-j--,-! e - 1
(f) Example. In
_,(101, "-l) _y-Q =100r'w - 10U'0' +e ln49 152 In3
3 - 12 + 48 - 192 + 49 152 c = -4
,- i {(- i)' (' - 0a " = ,+ ETJ =8
(e) Difference series transformatlon. When ais areal number (a *0), A=a- 1, A=a/A, and 3 - (-4)(-49152) nnnnt
A = 1.2 . then O ,?,*- l-i-4) -39321
y ay{k) =f! [| _AA+A2A- - A;,AJ +!/(*) (g) Infinite geometric series (n = *>, -1 < 6 < l)
*-i b "
is thc sum. If/(A) is apolynomial in Aof nth order, thc sum terminales with the term containing 2 Ck = a + aA* + aA2* + 2 /> = am + a"-'* + aaA2 +
I - c a - A
A"/(A). For CA,
he re
where u are the zerosofthe following polynomials
< . J
G.t(u) - a2 - - ftW-.*-|- + l,A ' 5 105
uk = eos (* 1. - J. . ' )
T = 3 f = 4
r = '2
U| a -0.TU7 106781 u, = -0.866025254 u, = -0.923 879 533
u, = -0 707 106 781 -0 794 654 472
u, = 0 707 106 781 u. 0.000 000 000 u. = -0.382 683 432
u, = -0.577 350 269
u, = 0.577 350 269 u. = 0 000 000 000 U.j -0.187 592 474 u, = 0.866 025 254 u] = 0.382683432
,= 0.707106781 "i = 0.187 592 474 u, = 0.923879533
u, = 0.794 654 472
= 6 = 7
t = 6 r = 7
r = 5
u, = -0 951056516 u, = -0.965 925826 u, = -0.974927912
u, = -0.866 246818 -0.883861 701 u, = -0.707 106781 u. = -0.781 831 48 1
u, = -0.832 497 4H7 u. = -0 587 785 252
u. = -0 374 541 410 u, = -0 422 518 654 -0.529 656 775
u, = O.OuOOOOOOO u, = -0.258 819045 u\ = -0.433883 739
u\ = -0 266 635 402 -0.323 911811 u, = 0.258 810045 u, = 0.000 000 IKK)
u\ = 0.000 000 000 n = u, = 0 387785252
u, = 0.433 883 739
u, = 0.374 541 410 u, = 0 266 635402 u, = 0.000 000 000
, = 0 951056 516 BS- 0 707 106 781
HS- 0 422 518 654 0.323 911811 u = 0.965 025826 u= 0.781831483
Bj = 0.832 497 487
u = 0.866 246818 ui = 0.529 656 775 u- = 0.974 927 912
i = 9
u = -0.195 090 322 u- = 0.831 469612
u = -0980785280
u, = -0 167 906 184 0.528 761 783 ,s = 0.195090322 uK = 0.980 783 280
u, = -0.911 589 308 u. = -0S31 469 612
0.601018 655 ., = 0.555 570 023
u, = -0 601018 655 s = 0.000000000 " = u\ = -0 535 570 023
,, = 0 167 906184 0.911 589 308
H, = -0.528 761 783
(c) Example. Ifa = 0. b= lx r = 4,/(x) = *cosx, *, = Jjt(1 + uj, (C) TWO auxiliary Cases in which the argumcnts u, and the weight factors SS4 are given in closed
X) = 0.161 278 ... = 0.638063 x, = 0.' 32 733 . = 1.409519 lorm are
/(x.) =0.159 185 ; i.l - 0512 524 f{xs) = O. 55 524 /(*,) = 0.226 340 (' ,
j^v^v-s^du n ^ . , kn i
=r^i2sm-7TT4coska
k* \ ng" l|7)
1/ 2""(20!
and by (a) above,
94 5.09 MONOTONIC ARITHMOGEOMETRIC SERIES Series and Products of Constants | SeriesandJWucteo/Consfans 5.10 ALTERNATING ARITHMOGEOMETRIC SERIES 95
(8) Notation. a, A, c, A= positive numbers a = number oterms (a) Notation. a, A, c, A, A, n, /44, fl, Q aredefined in (5.09a)
A= 1,2,3,..., n 4*A* Q = Ath term <r=(-l)"+' fl = (-ir'
ln C. - ln C, ln C, - ln C, ,
Ak = *,C, Bk = [a + (A - 1)A] C =c"""* -= /;|nf +1 .4, = fl,Ct fit = /5[a +(A - 1)A C4 = c" A Inr
(d) Infinite serles. When in (A), n=and c* >1, the series is divergent. If however, 0<c* <1, (d) Infinite series. When in (A), n= and c* > I, the series is divergent. Ifhowever, 0 < c* < 1.
the series converges and the series converges and
a - (a - b)c" At*
2 /! =+ ( +)f* + (fl +2)f += (l _ c*)a 2 .4 = a - (a + b)ch + (a + 2A)r* 1 + c" (1 + c*)2
where 3> = 2 A" and / = n for even nor i(n - 1) for odd n.
where **. =r*M - (7)2-"'*-.- +("J2""**--- " "" ''
(c) Particular series, all integers
(c) Particular series, all integers
A an + 0 A = ar(a3 + 2a2- 1)
2 ^ = z i-i 2
.?,'=T+2 ,* 5 2 3 30
. a0 as A5a4 n2 i re =i^ 2 A*5 =0aS ~(" +^ - 5(" +0)4 - 5(" +g>-'
+ _ 2/ =7+o
6 2 + 12 12
1-1 *
1-1 * A aa2(2n + 3) - 0
r." A a i n . R P* = 7* ^ Bk6 = (ni + 3a4 - 5n- + 3)
i-i 2
!-H' +
4
i-i
JL 4aJ . 2
A.2,,s_(^+(2^+5(2^_2i2i0j: 2 A(2A)2 = 2an(a + 1) 2 A(2A)5 = 320n5 - 16(n + 0)A
2(2A)2 = y+-"- +7 2,(2A) " ,2 + 2 6 15 i-i i-i
Series and Products of Constants Series and Products of Constants 5.14 ALTERNATING HARMONIC SERIES 99
98 5.13 MONOTONIC HARMONIC SERIES
(a) Finite seres (a = 1,2,3,...) is
(a) Finite serles (a = 1.2,3,4,...) s
f-,-i +i-i +... +f-iW-a.j-i..
where ... = sum (power 1, number of terms a), B= (1)'+', a = (-I)"*', P= number of
Lre *,.= <i >. <*>- 57"156649' =Euicr's cons',n,
(2.19), and
positive terms, q= number ofnegative terms, a = p + a,and
3... = *[*<*-!>-*<?)] +l2-e
1 1 / 1 / _ JLi/i 21 \\\ _
V() = lnn + i*,"l2a1V "TOa^V 2I2\ 20aW n which y( - 3), V(?) are thc respective digamma functions evaluatcd by thc general formula
given in (5.13a), and for n> 11, ^ l X lO"'0.
is the digamma function (2.32), and for na 5, 1XlO"'0.
(b) Special finite seres
(b) Example. To show the aecuracy of the sum in (a) thc first five terms (lower bound of the
approximation) are computed first by direct summation as y =I_I +!+...: =il
ft2k 2 4 6 2a 2'
2Jt=i+i + i+,{ + i = 2.283 333 333
i-i A A 1 1 ij*
y -= 1 - - + 7- 2-*l.
2ls = y + .Ino- +
^ i i #1 _ -Mi - -^(l - &))) = 2.283333 333
t ~ sot1 -sol MiV1 xo"> where 1,^ is the sum given in (a),
(c) Special finite series 1j. =*_ln ^ +i(A/ - A^) - *(A/2 - AT2) +(A/4 - A*4)
- SvA* - A*41) + f^/" - .VH) - <r
f~2A
=I2 +!4 +-6 + + 2a
j- -5ly
2
+ *()! = *a"
in which M = ,N = , and for a 2: 11, e = I x 10 '".
A_L_=1+I +I +.-+--4 =^y +'n4 +V(0] =^T.B 4/>-3 4a - 1
(2A-1) 3 5 2a- 1 2 (c) Example. Bydirect summation,
where U, , W, are the sums, ,-.-*. and for i>5.5 * IX10"". The digamma
Tuncln ;?/) is computed by the truncated series given ,n (a) for *(.). if, 2A - 1
A = 1_l + a_...Jr = 0.8O8 078 952
(d) Example. By direct summation, and bv the sum formula with a = 11,p = 6, a = 5, M = 57 $ = 75.
a > 104 n
a > 10'
>j(Sy + ln4/+-J = X 2,..= In2 ?f. 4
ar.-r +ln4, +-153.
(e) Superposition (r = 1,2,3,...)
(f) Superposition
f.i__!_ +_i--. = , - ,.
T*. A r r+l r + 2 r + a
A"r r+l r+2 "
~~1 "1 1 1 f**D "~ 1 ~~M 1 J 1
Series and Products of Constants 5.16 ALTERNATING SERIES OF POWERS 101
Series and Products of Constants OF RECIPROCAL INTEGERS
too 5.15 MONOTONIC SERIES OF POWERS
OF RECIPROCAL INTEGERS (a) Finite series (m,n = 1,2,3,...) s
(a) Finite series (m, n= 1.2,3,...) is
y = -
,A-
- - +--
1" 2" 3"
= a - 22_
a" "" ""
*=(-!)*' A = (-D*
T = + + -T+-+ = Z(m) ~ /, _ ni "
whcre ,,a = sums given in (5.15a) with a = number ofterms and a = number ofnegative
whcre ^ =sum (pww m, number of terms a), Z(m) =Riemann zeta function (2.26), and
y;("-",(w) 1 L
(m-l)! (m-l)n""' 2a" (b) Special finite series
/ i-+ DC +2)/, 10(m +3)(m +4)/ _ 21(ai +5)(m +6)UX _
,2^0
m
a =
i... = Z(m)
a:.. = ( - 2_,")Z(m)
Numerical vales of Z(m) and Z(ai) for ai = 1,2,3,... ,20 are tabulated in (2.27c).
Numerical vales of Zim) for ai =1.2, 3 20 are tabulated in (2.27c).
(d) Particular infinite series (d) Particular infinite series
y = -
YB it-,2*"! 4
, (2A - l)2 8 *' 12 *f, (2A)2 48
f,k2 6 i (2*)a 24
y /? _ ?*4 " 6 - *
2 2L-
, (2A - l)4 =-96 4't'l(2*)4 U520 if, (2* - I 32
(2A)4 =
*4 720
,A4 90 & 1440
A 3lg6 A __ = 3ln6 A B 5*
8 \ (2*)" 1935360 f, (2* - 1536
ftk6 945
^ =
*-,(2*)b
60480 2,(2* - l)6 =960 , i" ~ 30240
A t_ 127jtb 127** V /3
1 n"
--i" 1206600 i't'l(2*)B 308889600 i-,(2* - I 30 720
y - = 9450
,-r, (2A)B ~ 2419 200 iTi(2*-'l)" 806400
/T.A8 A 6 51Ijt"' A B 511jt' A 0_ 277,t"
* i jr' l' i'" = 47900160 i-i (2*)1 49049764 000 if, (2X - 1 8 257 536
y 1 = i?, (2A)' =95 800 320 fl(2k-\l)m "903
2903 040
, A' 93 555 , 6 a{2" - 2)n2', A /? _(/2)"(2*-2) B Q/2)a'-
A r = *(22'- \)n2'- V
(2r)!2 it*, (2*)2' (2r)!2 * if, (2* " I (2r)!2
V v,_3> =
1 1,1, . ! = I/,i__i_\=
(C) Unit Step binomial coefficient series. Since the Ath term in (a) above can be expressed as 1-2-3 + 2-3-4 3-4-5 8-9 10 2 \2! 10-9/ 45
m + k\ (m + A)! -3_ = !
sai!--0 r (-;*)- Alai! and for a = , thc sum is ;
3!(2) 4'
thc sum formula in <ti) gives also thc sum formula (C) Unit Step seres Of reciprocal binomial COefficientS. Since the Ath term othe series in (a) above
can be expressed as
i(;>>(">("> -+(":>("::?) A! I
r + A\
I
(r + A)!
A!r!
irn- o+r V)+r d +(-:*)-(':::') VA'i"' =Um {773^ Lia +(r +DAJi'-" "iTT^FJ] =(' ~0*1 +<r +'"fiF^
- i "i '--"j I """"i l ' ~j i --i 1 | 1 "1
2M 5.19 MONOTONIC SERIES OF POWERS Series and Products of Constants ' Series and Products ofConstants 5.20 ALTERNATING SERIES OF POWERS 105
OF DECIMAL NUMBERS OF DECIMAL NUMBERS
(a) Basle series. When a,A= signed numbers, ai = 1,2,3,...,I, and A= 1,2,3,..., a, then | (a) Basle serles. With notation of (5.19a) and with a = (-I)"+', B= (-l)i+l, thc alternating
with A = A/A, A = signed integer, fraction or decimal number,
2a(A +AAf =ab[n +(~)a*,.. +(J)aV, ++Qa".-] i M. +kkr =-[ +(7>.- +(>#.. +O"*-] '[',[
[0 a even
odd
(b) Particularcases
(b) Particular cases
abn
2(* + AA = [A(a + 1) + 2] abn
i-i 2 Ba(b + AA) = [A(or + 0) + 2]
2 a{b +khr =^[A2(a +l)(2n +1) +6A(a +1) +6]
i-i " 2 Ba(b +kh)2 =^[A2a(a +1) +2A(aa +0) +2]
.-i *
(c) General seres. Ifa. b, k, m, nare thc same as in (a) above, r = 1,2,3 s. A,, h2, h3,... ,h, |
are signed numbers. then with (C) General serles. Ifa,A, A, m, n, rand A, ,h2, A3 ,... ,Ar are thc same as in (a) above, then with
Ak = a(A + k-.. + k% + + k'h,)m = aAM(l + cxk + r,*2 + + c^k"") | Bk = Ba(b + AA, + A2A2 + -- + A'A,)" = Babm(\ + cxk + c2k2 + -- + cmkm)
and for a < x.
nd for a < ,
1r -V 1 1 A / 1 1 \ ZM
*'-(f) =*l'n7 " 2c +2* +,?, (?" Ar*) Vi +*** <--.) - --><*> =2 ^p " S^Tly
in which Bjj = Bemoulli number (2.22) and ,n which the polygamma functions can bc evaluated by the formulas given (2.34), (2 35). An
deernative and essentially the same procedure is thc application of Euler-MacLaunn formulas
in
al
l"flg.+2| - i- i ^ |fljr+a|
|-^T7| < If>.sl ^ | c2,+2 I 4.06),
ljt(c) =l[v(/. +cx) - tifie,) - V(a +c2) + x>(c2)] where c, =(a +A)/2A, c2 =a/2A, Nx =^+c, AT2 =a+c2, and ^"-( )=polygamma
function (2.34), which can be evaluated by thc formulas given in (2.34), (2.35). For a- *.
.V, = oo, N2 = oo,
iwhich c, = <c + A)/2A, c2 = a/2A, and V( ) = digamma function (2.32). For a = ,
^..(c) =^ly[V<-,^.)-V<"-,V2)}
Iri-sMiM^)]
I i i J
108 5.23 MONOTONIC SERIES OF PRODUCTS Series and Products of Constants Series and Products of Constants 5.24 ALTERNATING SERIES OF PRODUCTS wa
OF CONSTANTS
OF CONSTANTS
(a) Finite general series. When a,b are signed numbers, (a) Finite general serles. With a, Adefined in (5.23a) and oc = (-I)"*'. P= (-1) *. 0 - iffl is
(xld or 0 = 0 if a is even,
2 (a, + MKt + ,A) (,+ A,A) = -"I.. + t^i- + A,9,_ + + A,9,_, 2 P(a, + b,k)(a, + b,k) (, + b,k) = BQ + //,">,. + A-.- + "+ B,9tJ,
whcre ?. 9, , 9, are thc exact sums ofmonotonic series ofpowcrs ofintegers (5.11) and
A A A, ... .A, arc trie factors ofthc polynomial obtained as the producl ofthc binomials. where # ,&> 9, are thc exact sums of alternating series of powers of integers (5.12) and
B ,B,, l,',. '" B, are thc factors of the polynomial obtained as thc product of binomials.
(b) Example (b) Example
2 (2A - D(3A- 1) = 2 (I - 5A + 6*2) = n - 50>,. + Q,92jl 2 fl(2* - 1)(3A - 1) = 2 f(l - 5A + 6**) = 6 - 59,M + 6&,
i-i
i-i
where from (5.1la*), 9U = (a + l)/2 and 93m = n(n + 1)(" + 2)/6. where from (5.120. 9 = (an + 0)/2. &,. = an{n + l)/2.
I 2a + 2 3a + 3 4-' + + a(a + l)2 = n(n +!)("+ 2)(3a + 5)/12 1 - 2- 2 - 3J + 3 - 42 an(n + l)a \[an(n + 2)t.2a + 3 + 0|
l 32 + 2 5J + 3 7a + + a(2a + I)2 = n(n + l)(6n2 + 14a + 7)/6 1 32 - 2 5' + 3 7' o(2n + l)2 2[an{An2 + 10/, + 5 - 0|
i 23 + 2 V + 3 4* + - + (a +1) = n(n + 1)(I2;1 + 63a-' + 107a + 58)/60 1 2* - 2 33 + 3 4:' - an(n + l);i {[an{2n* + 10/r + \m + 6) - 0\
1 35 + 2 y - 3 ls + + n(2a + l)3 = a(a + l)(48as + I62n2 + 158 + 37)/30 1 3' - 2 5' + 3 7:1 on(2fl + I)' i|Q7i(8a:1 + 28- + 2te - 1) - 50)
1-^rW ^ ) . J J*^ ' 1^ r-^j
<p +MP +2?) +(/> +2?)(> +3?) " ' (A +?)(* +( +')?! (A +1)\P +< +Wl
(b) Special forms " - ^ +. + 2
23 2 -4 4-5 (n + l)(n + 2) '" n+ I n+ 2
1 1 I 2 3 14 / 1 I I \ _ 22
2 7-rTT7 =-^M-a'] r5 +3~6 +W+ "" +(n +!)( +4) =2" +b^TT +3U+2+n+3+r. +4/ 9
i-i (e + A)A
A 1
I
- = i2f - % 2af - 23... - a2j, (c) Three-lraction produets
if, (2A- 11A iT,(2A- l)A2
__!
1 +_L_ +^ + - + n(n =1(1 __L_ +_LA
y !; =-^ iva,. - a2. +a,] 1-2-3 2-3-4 + l)(n + 2) 2\2 n+ 1 n+ 2/
3-4-5
fAc + k)t rjL
1 =i/ll__! !_ + _!_ +_!_)
' +^+ ^
3-5 7+ '"+n(n-i-2)(n+ 4) 8\12 ,n-+l n+ 2 n+ 3 n+ 4/
1 = 42f - 43,, - 2<22, - Z3,* 1-3-5 2-4-6
i?,(2A- DA 1 I I _-![, ^ 1
1-3-5 3-5-7 5-7-9 (2a- l)(2a + l)(2n + 3) 12 L (2n + l)(2n + 3)J
where <&,. f,are defined in (5.13), 2nj,, 3*2, in (5.15) and <>, is defined in (a) above.
' +_L_, + _i_T+...+ J+ l)(3n4-4) -Ifi
24 L
i
(3n + l)(3n + 4) J
1
1-4-7 4-7-10 7 10 13 (3 - 2)(3n
(C) Alternating series. Although sum formulas of alternating series are y||aWc,Jtheir P^V
1 + ^ +-L^+-- + (2 - ;W--alJ-5
2n - I
m 3+ 5 5-6-7 l)(2n)(2 4- l) 2
2~3^4 +3_45 +4-5-6+ " +(i + l)(n +2)( +3) 4^ n+2 a+3/
, 2 3 n \_( 12 ( 12 60 60 \
a .-ir1 _ i r* ! ! 2~^6 +3^7 +\~67B+ '" +(n+ l)( +3)0 +5) ~96 I'7 ++2 n+3~n +3 n+4/
fr (,-+*,*><*+ M) 4A,AS L~, (*, +A)(a*2 +A) fr, (<, +A)(<2 +A)
I 2 3 f --(3- 8 I )
whi-re a*, =ir: - D/2, a*2 =(c, - l)/2. f, =f,/2, e2 =c2/2, and each sum is expressed by the r n + 4T6 + 5-6-7 " ( + 2)( + 3)( + 4) 4V n + 2 (n + l)(n + 2)/
relalions given in (a).
3 3 J J J
112 5.27 INFINITE SERIES OF PRODUCTS Series and Products of Constants Series and Products of Constants 5.28 INFINITE SERIES 0F PRODUCTS 0F 113
OF FRACTIONS, GENERAL RELATIONS FRACTIONS, PARTICULAR CASES
(a) Monotonic serles. The sum ofthc series in (5.25a) for a-+ is (a) Two-fraction produets (a = )
n~
1 l .
i _. a i m>(c2) - y{ct) 1-2 2-3 3-4
= 1
1-2
+
2-4
+ z +
3-6
i?, (a, +A,A)(a2 +A2A) " ^ frMx +A,A)(a2 +A2A) AA2(c2 - cx) _ 3 1 I 1
J- +-L +-L + .. + + + = 2 - ln 2
where c, = ,/*,, c2 = a2/A2, and y( ) = digamma function (2.32). 1-3 2-4 3-5 ' ~ 4 1-3 2-5 3-7
_ 11 1 1 _1_ \_
(b) General formulas in terms of a,A,r =positive numbers, m,p,q,r = positive integers, y = l-4 + 2-6 + 3-8+ '" 2
0.577 215 664 (2.19), _4 + 275 3-6 ~ 18
_L + J_ +_L+.. _ 25 1 + 1 + +
^ = -8 - -2iln 0
2
2,, = " a. =7 a'-' =
*T,* if.AI z(m) =? =Ricmann zeta funcdon
i-i * 1-5 2-6 3-7 ' ~48 1-5 2-7 2-9 9 3
-L +-L. + J- + .. _ 1 1 14-8
' 16-12
! 1 -**
(2.26) are 2-4 48
2-4 4-6 6-8 ~ 4
y 1
' 1
= -a, 1 1,11, 4-^
fr(p + k)k p '' 3-5 4-6 5-7 ' ~24 3-5 7-9 11 13 ' 8
-L +-L. +-L- + 1 ! + = ln -
771^ =l-/'Z(2)~i2,'J
i-,(/> + A)A
/> |(rTijP"7lffZ(2,-*(r,-yl 12 2 22 3 23
= ln 2
13 + 232 33' "2
- I _ V>() ~ V(?) y ! =-i-
.(a + AA)[a + A(A +1)] a+6 (b) Three-fraction produets (a = )
,-,(/> +A>(9 + A) />-?
J _ (y - r)3,,, +(r-j)3... +(/> ~ 9)2,., 1
1-2-3
1 2-3-4
' 1' -'4 1-2-3 ' 3-4-5
1 1 ln
"""
-
2
*-,(/> + Al? + A)(r + A) (/, -?)(a - r)(r-p)
-J-+ ' + =!(-*- 1 l ' I -'ln-
A 1 I
1-2-3 4-5-6 4\V3 1-2-3 " 5-6-7 4 ""
tT, [a + AU - l)](fl + AA)[a + A(A + l)] 2aA(a + A) _ 1 1 3 3
1 + 2 +... ~ 4 1-2-3 ' 2-3-4 4
A i _MT-(/Z(2) - />-'Z(3) + A'Z(4) (-/.r-'Z(ai) - 2,.,] 1-2-3 3-4-5
T, (/> + A)*" " _ 11 1 . ! . -112
1 =_JTicZ(2) - C2Z(3) +r3Z(4) (-e)-'Z(m) - (') " yJ l-3-5 + 2-4-6 ~96 1-3-5 3-5-7 "
jf,(c + k)k 43
1 1 4-7-10
' 1 - 24'
1 ' + ~ 1080 1-4-7
1-4-7 2-5-8
(c) Example
1 _!_... -2-h* 2-3-4
1 , l
6-7-8"
1 -*
8
V
2" "
y__L_ =_L4._L +J-+... 2-3-4 44-5-6 .
(3 +*l*J 4-13 5-23 6-33 1 1 l_
1 , ' , 1
246 +468 32
=Zll!i3Z(2) - 32Z(3) - V(3) - y] =0.307 194 775 2-3-4 3-4-5 12
1 , l , ... 1
1 I ' .------
where from .2.26), Z(2) = 1.644 934067, Z(3) = 1.202 056903, and y(3) + y = 1+ += 3-4-5 4-5-6 24 3-5-7'5-7-9 60
1.833 333 333.
7 1 2 ( 1
' + 2 +...
(d) Alternating series can bc conveniently summed by the procedurc described in (5.25c). 2-4-6 3-5-7 ~96 3-4-5 4-5-6 6
(e) Example. From (A) above, with p = 1. c = 5and by (2.32/) ic) Four-fraction produets ( = 00)
1 I 1 1 1 In65 - n
s,
2' = c - I
f,ak (a - l)a" a2(l + aa2**) y / _ a + a*
2/*** = 1 +a2 fra2k (1+aV
a- a2" - 1)
si2 y_L = '*-' *t ^ _P _ <x + alm
2 * =
i-, a".. - ,i a24 (a2-l)a2- a(l + aa2")
2 * = 1 + a2 fia"-' (l+aV"-'
A 1 _ a2'-l
s, 4_ a + o-(a + l)a"*' + ana" A Bk an + Of(a + l)a + a"'
i?," " a2-\ a2*"' (a2-l)a2"-' 2/?Aa' =
(l+)a ftak~ (1 + a)V
V * c - (a + \)a"+l + aa'+2 A A _ a - (a + l)a + a"
s44
fr,ak~ (a-l)V A /3 _ or + a"
frka " (T-7? fu* (1+aV
A 1_ a" - 1
2** =^a - s.
a'* (a--l)a'" A a' + <r(a + Qa"^" + ana'm+2) A Bk _ an +a{n + l)a' +a,("""
i -, 1
A A_ a - (a + l)a' + a""*" .?/*<* = (TT7? iT.a'*" (1+aT'a-
- , .; - (a + l)af"*"" + aa""*2'
ha'k (a'-l)'V 0(2* - 1)<i' = 2, - % 2^=29.-9,
t-i a
1-1
" 2A 1
2 (2A - 1a4 = 2s4 - s, s, X^T-L = 284-8,
i-i
i-i a
2 /3(2A - l)art = 2% - % mpi=2%-%
2 (2A - 1a" = 26 - .8S s 2A_Z =2g86-S85
i-i a
:) Derived sums. If(A) and g(k - I) are alternating functions ofAand A- 1,2, 3 a. then
(C) Derived sums. If/(A) and/(A - 1) are functions of Aand A= 1,2,3,...,n, then
2k<*)-*(*- \)\ = gW - g(0)
2 [/(*)-/<*- )]=/()-/(O)
(d) Examples
(b) Monotonic series ofsine terms (a < ) | (b) Monotonic serles of cosine terms (n < )
sin [(a + l)a/2]sin(aa/2) " , a sinnacos [(a + l)a] cos [(a + l)a/2]sin(aa/2) " a sin aacos[(a + l)a]
2 sin2 Aa = - - - 2 sin a 2 cosAa = y. eos Aa = - + :
2 sin*fl = MaTij i-i sin (a/2) frx 2 2 sin a
i-i
a sin 2aa eos [2(a + l)a] sin [(a + l)a] cos na " a sin 2aa cos [2(a + l)a]
sin [(a + l)a]sinaa 2sin22Aa=--- >. eos'' 2Aa = - + .
2 sin2Aa = 2 sin 2a 2 eos 2Aa = sin a 4f, 2 2 sin 2a
sin a t-i
a sini2aa
2aa eos
ce 2aa " a sin 2aa cos 2aa
2><2*-i>.
sin2 aa
sma
2sin2(2A- l)a=--
2 2 sin 2a 2cos(2A - l)a =
sin2aa
2^(2*-^ = ^+ 2sn2a
2 sin a
" sin[(a + l)a] (a + 1) eos [(2a + l)a/2] A cos [(a + l)a] - I (a + 1) cos [(2a + l)a/2]
2 *sin Aa - 4sn*(a/2) 2sin (a/2) fr kCOS ka ~ 4sin2 (a/2) 2sin (a/2)
sin[2(a + l)a] (a + l)cost(2a + l)a] A cos[2(a + l)a] - I t (a + 1) sin[(2a + l)a]
2 2A sin 2Aa = ^r^ sin 2 2A cos 2Aa =
f->,
V2 sin'
. 2a + T-
sin a
eosa sin (2aa) - 2a sin a eos 2aa nsin (2na) sin a sin2 aa cosa
2(2*- l)sin(2A-l)a = r^ 2 (2A - l)cos(2A - l)a
(c) Alternating series of slne terms (a < ) (c) Alternating series of cosine terms (a < )
sin2a + a sin2aa + a sin [2(a + l)a] I + cos 2a + a cos 2aa + a cos [2(a + l)a]
2 P sin 2Aa = 2 P eos 2Aa = 2(1 + cos 2a)
2(1 + eos 2a)
d "
2 20A 2 2/3Acos2Aa = - 2 P**" 2Aa*
frx H sin 2Aa = - -dakm
2 Pcos 2ka* i-, ^ </fli-i
2/5(2* - l)sin(2A - l)a =-- Pcos U2* _ 'M* 2 P(2k - 1) cos (2A - l)a = aVii
2 Ps (2* ~ l)*
1 + acos4aa 6 I + arcos4aa B
2 P sin"* (2A - I )a = + ~ 2Pcos2(2A-l)a= 4cQs2fl +-
4 cos 2a
Dervame ofehe rcsrcciivc sum formula in (5.32r). Dcrivativc of cherespectivesum formula in (5.3lr).
-a ^^ H 1 1 1 ^ 1 1 -~^ ~1 *i
y-U-i-
iT.A*
y=_L_
a-i 6
.t*.A* *2" '
y_J
y A = __*__
*_
ALGEBRAIC AND
.T.A* (A "O1
A i (1 + ) A 2A- I
f, bk "(* "O2
1 +b
TRANSCENDENTAL
(c) Alternating series of power terms (a < l,A > 1)
y _!_
EQUATIONS
i-i A* 1+ A
vi !_
l^'-rr? ,t,,*M~l+*2
V_ *_
i-i
fr,bu-1 l+A2
y/_ A
fr, *' (1 + A)2
A B(2k - 1) A- I
A1 (1 + A)2
cos a
A eos (2* - l)a y;
sin a/2
t-i
i-i *
= -ln Vi - 2acos A + a2
^> pB sinka
sir _ T A 0 sina , / / . a\
H<-
1-1 *
|a|<
A Bcvska , / a\ A B eoska _ jr - < < a
iT, 2* - I " 4 2
T 3 H '' ""1 ~ 1 ~1 i1 '~ " i l 1 '3 ~3 1 1 '" 1 1
Algbrate and Transcendental Equations 6.02 QUADRATIC, BINOMIAL, 121
120 6.01 ALGEBRAIC EQUATIONS Algebraic and Transcendental Equations AND TRINOMIAL EQUATIONS
OF nth DEGREE (a) QuadratiC equations in canonical form are
(a) Definition. An equality of two algebraic functions/(*) and g(x) in one variable x, + Ax + c = 0 x2 + px + q = 0
Ax)=g(x) .; -A W -4ac
is called he algebraic equation, which is valid br certain vales ox, ,x2 x designated as ^ 2a
*i.a= _2 v!F
the roots.
(b) Canonical form. Any algebraic equation can be reduced by algebraic operations to the canonical Thcse roots depend upon the discriminant D= A2 - 4ac or D- y-j - a.
form
If D > 0, the roots arc real and unequal
IP(x) =a0 +, +**** +' +a'x' = 1 If D = 0, the roots are real and equal
which has the same roots as the equation (a) and the exponent adefines the degree of this If D < 0, the roots arc complex conjgate
equation.
(b) Examples. (/ = V-i, i2 = -l, 3 = -/, 4 = 0
(C) Example. V3i +4x +5*'5 = .l +x2
, au i -i. &.y + s*a = 1 + 2x'2 + x*
becomes. 3+ 4* +5* 1+ and thc canonical
D <0
D >0 D =0
form is
v- - 2v + 8 = 0
/>(,, = 2 + 4x + 3x2 - x* = 0 c2 - 2x - 8 = 0 r2 - 8x + 16 = 0
c,,, = 1 Vi -8
(d)conjgate
Roots of and
the seprate,
algbrale and/or
equationcomplex
may beconjgate
rea, *^" ] ^ ^^^
and repcated. 11 x, ,x2
I Vi +8
I 3
xl2 = 4 Vl6 - 16
= 4 = 1 V7i
P{x) = 0. then
P(x\ = (x-xx)(x-x2)---(x-xm) = 0 ^c) Relationships of roots by (6.01 are
MSuperfluOUS rOOtS. In transforming (a) o acanonical form (A), it may happen that (A) has one fi + x2 = - - = -p x, -x2 = - = a
or more roots which do not satisfy (a). Two such cases may occur.
When thc roots arc real and onc is much larger than the other one, che smaller onc may be
(f) Vanishing denominator. If /(*), GW are polynomial and calculated with a higher degrec of precisin from thesecond relation above as
P{x)
= o x2 = = - (x2 < x.)
Q.(x) ax, x,
then after multiplying by <*(*), W =0. This canonical equation has all the roots ofthe origmal v,
equation. and it may have in addition some superfluous ones. ^ (d) Binomial equations (A > 0)
i B/7/ 2A.T . . 2A*\
(g) Example || x" - h = 0: xl2, .,. = VA (cos
\
+ 1sin
a
1
a /
a = 2,3,4,...
(h) irrational equation. If/(x) =g(x) involves an irrational term, then again some superfluous roots (e) Example. The roots ofx3 - 27 = 0are x, = 3and x2:, = (-l i'V5)
may occur.
(f) Trinomial equation (a = 2,3,...)
i
(i) Example. y/T+ = x +. 1. l.becomes xa +
j. x
v -- 92 =
- O
u, x.
x, = 1i, x,
x, = -2, whcre thc first root fj ax2" + Ax" + c = 0 or x2" + px" + a = 0
x, = l sacisfics the first equation and x2 is superfluous.
reduces by substitutionj = x" to
(j) Relations between rOOtS and COefficientS are given by the following equations:
a/ + by + c = 0 / +/>>+? =
2 x^x^-^2 (<j<U -A VA2 - 4ac
y, -=i 2^> = (f<J> j*-i a- and yields x'^
2a
<, - -| VT7
the 2a roots of which are calculated by (a*).
122 6.03 CUBIC EQUATION
Algebraic and Transcendental Equatons W Algbrate and TranscendentalEquations 6.03 CUBIC EQUATION {Continued) 123
(f) Transcendental SOlutionS of the original cubic equation in (a) are given below where
(a) General case. Incanonical form, r = y/\p\ and thcsign of r and qmust coincide.
Iax3 +Ax2 +cx +d=0| Case Roots Conditions
(b) Classification Of rOOtS. When a, A, c, dare real numbers and D=p3 +?2, then if, A
c, = -2rsinh---
b
D> 0, there is one real root and two conjgate complex roots r2 = rsinh ^ +rV3 cosh - /> > 0 0 SO 0 = sinh
(3)
D= 0, there are three real roots of which at least two are equal
D < 0. there are three real unequal roots
r, = rsinh ^ - rVj cosh - -
(C) First root Of the original equation is always real and occurs in one of the following forms.
(9) Example (0 > 0). x3 + x2 + x + 1= 0, where a = 1, A= 1, r = 1, a* = 1, p - 5, a - M,
Case
> = 0.148 148 148 > 0.
(d) Second and thlrd root of the original B> (7-2), 0 = eos" 7= = - and using + for r,
equation are the roots of tic quadratic
equacion <p b
x, = 2rcos = 1 2 = 3
1 3 3a
x2 + 2Bx + C = 0
(d) LargeSt real rOOt y, is then uscd for the calculation of A, B. C, Das (j) Symmetrical quartic equation
A= U, B = iv, D =_r2 - A ax* + Ax3 + cx2 + Ax + a = 0
((,bx-AB)/D
,[},bx- AB)/, ifZ)*0
C reduces after dividing by x' to
~lVil2- A., +~2B ifO = 0
(e) Auxiliary quadratic equations from the product in (A), in ccrms oA,B,C.D calculaced in (a*)
above. 1
and with x + - = y x* + _=/_ 2
..; + (A + C)x + B+ D= 0 x2 + (A - C)x + f - D= 0
vicie thc roots ofthc quartic equation in (a), liccomes a quadratic equation my,
,, ,=-i(yi +C) y/\[A +C)> - {B +D) x:M =-*(* - C) y/4(A-C)2-[B-D) ay2 + by + c - 2a = 0
ihc roots of which when inserted in
(f) Relations Of rOOtS used for a numerical check are
l 1
= x,x2 + x:tx4 .va = x,x., + x2x4 y3 = x,x4 + x2x3 x + -=yx
x
x + -=y2
x
(g) Classification Of rOOtS. If the coefficients of (a) arc real, then the following classification is yield thc four desired roots of the original equation.
posible. (k) Antisymmetrical quartic equation
126 6.05 SPECIAL QUARTIC EQUATIONS Algbrate and Transcendental Equations Algebraic and Transcendental Equations 6.06 SYMMETRICAL AND 127
ANTISYMMETRICAL EQUATIONS 0F HIGHER DEGREE
IN TWO UNKNOWNS
(a) Case I
(a) Properties of binomials (A = 1,2,3,...)
(x2* + l) :(x 1) = x2*-' =F x2*-2 + x2*-3 T 1
xy = A
(jr+i + 1) :(x + 1) = x24 - x2*- + x"-2 1
be solved as x2 + 2xy +y2 = a + 2A x2 - 2xy +/ = a - 2A (jr+i - ,) : (, - 1) = x2* + xM- + x2-2 + + 1
x +y = Va + 2A = A x -y = Va - 2A = fl (x2i- l):(x*+ 1)= x*-l (xM- l):(x*- 1) = x1 + 1
and
(b) Symmetrlcal equation ofthird degree
r, =|( +fl) x2 =( - B) x, =-|( - 5) =~x2 x4 =-&A +fl) =-x, ax3 + Ax2 + Ax + a = 0 is transformed to (x + l)|ax2 - (a - A)x + a] = 0
2A 2A
2A 2A
= -J*2 ^4
A + fl
^2 =
^ -fl
.yj
-fl A + fl
and x, = 1 x2.3 =^- [(a - A) V(a - A)2 - 4a']
a
2a -A 2a, - A2
(e) Antlsymmetrical equation of fifth degree
from which with 0, =Va2 - 4, and D2 = Va2 - 4t>2,
ax5 + Ax4 + rxJ - cx2 - Ax - a = 0
x, = |(, + 0.) x2 = *(, - />,) x3 = |(2 + 02) X4 = (3 - *>*)
transforms to
2a, 2v,
2o, 2o, ya =
^2 =
a, - 0,
73 -
a, + 0, a, - 02 a(xs - i) + Ax(x3 - 1) + cx2(x - 1) = 0
T| , + 0,
(x - l)[ax4 + (a + A)x3 + (a + A+ c)x2 + (a + A)x + a] = 0
(c) Case III
'""iT^ C fl
| a,x2 +A2xyj+cj2 +a*2 =0| i,x2 + A.xy + c,y + a*. = 0 and similarly as in (d),
can be transf-rmed by climination of dx and a*2 to -fl i Vfl2 - 4a(C - 2a)
7l.2
2a
\axd2
\' Z
- __y
*>dx)x2 + (bxd2
V
-v Mi)*r'+
'
(fi^
N
-y c^y
'
=
A \ A C and ,.- ^--a^rT .,-?* VF7"
and reduced l-v thc substitution x = uy to Au2 + Bu + C = 0.
1 1 1 'J " 1 V'H| 1 "" ~3 1 1 3 113 1
Algebraic and Transcendental Equations 6.08 BISECTION METHOD 229
128 6.07 APPROXIMATE SOLUTIONS Algebraic and Transcendental Equations
OF EQUATIONS (a) Bisection method begins with the interval [a0, A0J in which the
function changes its sign so that
(a) Numerical SOlUtion ofany algebraic or transcendenul equation/(x) =0, should bc preceded by
the investigation of the graph of f(a0) > 0 >/(A) or f(a0) < 0 </(A0)
10" 10" 10" 10" 10" 10" 10" 10" 10" 10"
10"
(d) Example
10 20 24 27 30 34 37
./' - - xs _ 4A + i = 0 in a = 0, A = I
(d) Numerical procedure uses only thecoordinates ofthe endpoints ofthe interval for thecalculation
The repetition ofthis process leads to abetter approximation (6.08a). ofthe next approximation, and it always converges to the correct valu. It requires usually 10 to
20 iterations, but this involves one function cvaluation in onc iteration steponly. In searching for
fe) Methods used in the approximatc calculations of roots of equations arc based on the concept of the interval of sign change", there is always the danger (if large steps are used) of overstcpping
interplation and/or iteration. They are classified as general methods applicable to equations o some of thc roots. The graph of/(x) helps to elimnate this ovcrstepping.
all rvpes and polynomial methods restricted to the solution ofalgebraic equations of higher degree
(bu: ipplicablc co all algebraic cquacions ofa > I).
(f) General methods introduced in the subsequent sections ofthis chapter arc: (e) Example./(ao)<0</(Ao).
11 Bisection mechod (6.08) (3) Tangent method (6.10) f(x) = tan x - x = 0 a0 = 4.4 A = 4.5 /() = -1.303 368 f(b0) = 0.137 332
(2 Secant method (6.09) (4) Iteration methods (6.11), (6.12) m = i(4.400000 + 4.500 000) = 4.450 000 f(m) = -0.726 731 < 0
(a) Polynomial methods of practical importance used in the search of real and complex roots of m = '(4.450 000 + 4.500 000) = 4.475 000 J\mx) = -0.341 933 < 0
alcebraic equations ofhigher degree and described in thc final sections ofthis chapter are: mi. = ^(4.475 000 + 4.500 000) =4.487 500 /(ai2) = -0.116078 <0
11 Newton-Raphon's method (6.17) (2) Bairstow'smethod (6.18)
Methods developcd in the era ofhand^calculations such as Horner's method (Ref. 6.9), GraetTe's ni';,, = (4.493 263 + 4.493 556) = 4.493 409 /(ml0) = -0.000 000 4 < 0
method (Ref. 6.9), and Bernoulli's method (Ref. 1.10) are of limited valu in computer and! the correct root is x = 4.493 409.
applications.
T" r~TI 1 "U ' 1 J|
(b)
Next approximation depends upon the relations of(a0),f(c0),Ab0) as shown graphically below. , <
-i
* -
/ir.i
V
-y t
i
>.
J
(b) Next approximations are
bg.fjco) ~ Cof(b0) c,=c0-f(c0)/f(c0) iAcolfico) > 0
_ cnf(aa) ~ aoAco)
Ac0)-Abo) c2 = cx-f(cx)/f(cx) if/(filT(.) > o
c,~ A<o)-A'o)
x a c, = cr.x -AC-i)/fbr-J '*Ac.-iV(C-i) > 0
f> '. where c. c2 ... ,c are the first, second,.... rth approximations ofthc root x. Thc method works
ifca is cise to xand/(x) * 0,f'(x) # 0. Thc sufficicnt condition for thc convergence is
^
. y. \ACkV"(ck) < 1
I [f(ck)]2
but it is not thc final rate ofconvergence but the initial rate which controls thc number ofsteps
1 required.
(d) Examples
Case III: fM </('o) < 0 <bo) (d) Example, /(a0) <0 < f{bQ).
Case II: /(*) >f(c0) > 0 >f{b0)
f(x) = tan x - x = 0 fl0 = 4-4 bo = 4-5 /(x) =tanx - x=0,ao =4.4,A0 =4.5,/(A0) =0.137 332,/(x) =scc2x - 1
f(x) = xi - 4i- + I =0 a0 = 0 *0 =
/() =1.000000 Ab0) = -2.000000 /(a0) = -1.303 676 /(*0) = 0.137 332 and since f(b0) = 21.504 843, c0 = 4.5
/(,) = 0.592593 fo = 4.490470 /(r0) = -0.058 543
c0 = 0.333333 /(r.) = -0.001 843 -A'k)/f{*k)
f(c,) = 0.170916 r, = 4.493318 k 'k A'k) />*)
c, =0.485 714
21.504 843 -0.006 386
/(f<) = -0.000002 0 4.500000 0.137 332
/(8) =0.000004 c4 = 4.493409 20.229 717 -0.000204
ce = 0.537 401 1 4.493 614 0.004132
- 20.190813 -0.000001
Correct root x = 4.493 409 2 4.493 410 0.000009
Correct root x = 0.537 402
"1 j ' ] "5 ~3 8 "1 "1 "1
ax ax I dv dr I
(d) Example
f(XtJ) = 3/ - 2x3 - 10xJ +1=0 g(x.y) = / - 3x2 + 5/ -5 = 0
and a = 0.5, A = 0.7, A = 0.9, i? = 1.1. a = 0.6. A = 1.0. Taking
/3/ + 1 . /3xJ +
*= ' = ^2TTTo y=+>= ^7^
che partial derivatives in terms of x = a.y = A arc
c3 = tan-'3 + * = 4.493 404 The roots a = 0.602014, v = 1.006685 must satisfy the initial equations
c, = tan"' c +* = 4.4913,2
c4 = tan-'c4 + Jr = 4.493409 3a3 - 2a3 - 10a2 + 1 = 5.32(10)"* M - 3a2 + 5a2 - 5 = 2.44(10)-' =" 0
c, = tan-'f, + * = 4.493 3,2
"1 1 1 - 1 - ~1 1
6.14 SYNTHETIC DIVISIN 135
Algbrale and Transcendental Equations
Algebraic and Transcendental Equations
134 6.13 PROPERTIES OF POLYNOMIAL
EQUATIONS
wsynthetiedivision is an ^^^^rt^X^Z^^
and n ^aluat.ng Hynomials amd *<^ malrPix form. T^c first two forms are shown
(a) General properties ofthc polynomial equation S 3^ ~ intr'oduced in (6,5,), (6,5,). (6,5,).
P(x) =a0 +a.x +a2x2 +--+a..,x"" +a.x" - 0 (b) Long form without loss ofgenera.ity can be demonstrad by the fo.lowing particular case.
are summarized in (6.01). Additional theorems related o the methods introduced in the
subsequent sections are dcscnbed below. SL = (4x3 +2x2 + I) +(x - 2) =4x2 + 10x +20 +
x-2
-4xJ + 8x2
(b) Factor theorem. Ifx - cis afactor of P(x), then , is the root of P(x) =0. l pm,qm are real,
pl < qm, and x2 +2pmx +}.isa factor of P(x), then
Cj =-pm +i'JZ^H < - -'-" V9"* " *'"
arc the conjgate roots of P(x).
(c)product
Factorization theorem. Apolynomial />(x) with rea. - ^ - ^77=" ^
of linear and quadratic factors with real coeffiecnts. and PM =4x3 +2x2 +=(V<+><>x . ^x "2) +41, where the remainder R=,(2) =
ai = 1,2,...,*, 41 as required by thc remainder theorem m(6,3rf)
P(x) =a0(x - cxr (*- cy(x2 +2p,x +a,)* (x2 +2p,x +q.) (C) ShOrt form for the same case is
where a B. =positive integers and *, +*,+ +. +W, +ft +'" ' +A> " "
2
(d) Remaindertheorem.IfP(x) *0is divided by x- r, where ris cCj or , defined in (,) so that
4.-^ 10 ' 20^ 41
| P(x) =(x - f)g() +R\ ;. f Ptx) the second row is 0. (4 + 0)(2) = 8,
then for x = c the remainder fl = P(c).
(d) Successive divisin ofW by x- ayields the factors ofTaylor's series
combination of them. In particular:
P(x) =r0+rl(x-a) +r2(x-a)2 +M(x-^-r- "+'.(* " >"
(1) /!// odd-degree equations have at least onc real root.
. , ofr rk rr
(e) Example. The calculation Px> =4x3
for P{x) - *x +2x2 + 1=0 at x =4 is performed by short
(4, A^antisyn^^^-^^^^^ta antisymmetrica, pairs and/or divisin (c).
where the roots of <2(x) in terms ofPm, qm defined in (A) above occur in real pairs 2
4 16 12-4 Tm-\
r_2 a-z -A0 -A,
1
- 5 - 20 - 15 5
6 24 18-6 A.. 0 0 -Ao -A, ri
a2 0 0
T}
r.
,(x) = 4x + 1lx4 - 2x3 + 5x2 + 23.v - 6 r, i 0 0 0 0 0 -A -A,
r a0 0 0 0 0 0 0 -A -A,J -r -1
_ -I
(c) Divisin of polynomial bybinomial
amxm + am_|X"~' + a--2x" + - + a.,x2 + a,x + a
where r =a/A2, rn.x =(._,
(ak - A0r,+2 - A,ri+1)/A2
- A,rJ/A ^ " ^-">/;-^0/*!'"
r, =(, - *r, " *.',)/*.. 'o " < ** 'r,)/*2-
, i =
A,x + A
= r.*""1 + rn.xxm-2 + r./-;' + + r,x + r, + R
where r ,r_,, rm_,,..., r, ,r, ,r are the coefficients of the matrix given below and
(f) Example
rm
where from
r*-\ am-i An
'--i
rm--i -, 0 -A0 ' 4
1 r5
8 -2
= ^ 0 0 0 0 0 -A0
'4
r> 0 -6 -2
r-, 1
0 0 0 0 0 0 -A0 -2
'. "4 16 0 -6
i
0 0 0 0 0 0 0 -A_ L'' J
-6 -2
.r - 20 0 0
ri
0 -6 -2_
where rm =ajbx rk = (ak - bnrk_x)/bx r0 =( " A0r,)/A, ro 30 0 0
where rs = 2(4) =2, r4 =418 - 6(2)) = ^2,.. . ,r0 =l - 6(40)] = 105. md the remainder tf = (39x + 5.5)/(4x2 + 2x + 6).
^^^^w w
Algbrate and Transcendental Equations 6.17 NEWTON-RAPHSON'S 139
138 6.16 ROOTS OF ALGEBRAIC EQUATIONS Algebraic and Transcendental Equations METHOD (Continued)
(a) ROOtS (zeros) Of polynomial equation are classified in (6.13*). Their existence and relalions to
thc algebraic curve y = P(x) are llustrated by the following examples. (3) Second approximation in terms ofxu by (6,4*)
(b) Example of real distinct roots. The roots of
12.257 1.000 -14.000 9.000 180.000
P(x) = x3 - 6x2 + llx - 6 = (x - l)(x - 2)(x - 3) = 0
0.000 ^rl2.257 ^,-21.364 >r-151.545
are x, = 1, x2 = 2, x3 = 3, shown in the adjacent graph.
12.257 1.000"""" -1.743"* -12.364 28.455 = P( 12.257)
(c) Example of real repeated roots. The roots of 0.000 ^,12.257 ^>128.870
P{x) = xJ - 4x2 + 5x - 2 = (x - l)2(x - 2) = 0 1.000""^ 10.514"" 116.506 = /'(12.257)
are x12 =1, x, = 2, shown in the adjacent graph, where thc
repeated roots x, and x2 coincide. 28.455
is similarly as in (2), x2., = 12.257 - jt~^ = 12.013.
(d) Example of real and complex conjgate roots. The roots of
P(x) = x3 + 3x2 + 7x - 5 = (x - l)(x* - 2x + 5) = 0 (4) Last approximation in terms of x2_, by (6.14*),
uo 6.18 BAIRSTOW'S METHOD Algebraic and Transcendental Equations A/gebmicamfntmscendenaiBqualioiis 6.18 BAIRSTOW'S METHOD (Continued) 141
(a) Divisin ofan even-degree polynomial (d) Coefficients 6, (j = l,2,...,m) in (f), required for Che calculacin of A/>, Aa and for the
deflated polynomial equation /'_,>(*) = 0, are expressiblc in closed form as
PJx) = xm + a,xm-' + a.,xm~'2 + + am_,x + am
by a quadratic factor Qfx) = x2 + px + a yields Fi i
/;. F, I
PJx) = Qix)Pm-2(x) + bm.xx + bm
/-i F> ^i 1
As
whcre Pm.t{x) = x"'2 + A,*""' + Aax""2 + + A_,x + _,
and _,,A are algebraic functions of/>,a. Thc goal is now lo find thc vahu-s op,q such that Fm. Fm- Fm-4 i
A. /*"-
A_, =0,C=0. Fm- Fm- /;-..
A_ /-
b F
(b) Approximations of p, 0which satisfy this condition are given by the following recurren! relations,
whcre a (j = 1,2 '") are the COefficientS of Pm(x) and in terms of M= a//)",
a, = a0 + Aa,
Vt, _ At.m-,rt.m-. ~ At.mrt.m--,
px = Po + AAl A/4 =
?, = ?i + Aa A( Cj.-a ~ Jlf^-S /.; = ft\ - M(5 - A#*(6 - A/')))
/>. = pi + *h /', = ~P
Vf.v _ At.mrt.m_,. - Atm_,aV.._, /.-_ = -p\] - M{6 - M2(\Q - 4.W1))
Aa, = F, = p2(\ ~ M)
p, = />,_, + A/., 7, - ,-i + Aa, A, ftj.-a ~ Vr-ic*.*.-3
F3= -p'\\ -2.U) /.; = pl\ - .1/(7 - A/-'(15 - Ml(W - M'))))
ich p0, q0 are thc initial estmales of/>, a and k - 1.2 , r. /.-4 = p\\ - .1/(3 - M2)) F = -//'(l - .\/(H - A/'(21 - M*(20 - 5M4))))
F, = -p*(] - .1/(4 - 3A/2)) Fjfl = ^1(1 _ .U(() _ m'(2H - A/3(35 - Af*(15 - A/5)))))
(C) Coefficients in thcse formulas are again related by recurren! relations (with the subscript k
omitted but implied) as or recurrcntly
Fj^pFj^-qFj.,
A = a0 = I c0 = A0 = I
A, = a, - )&Ao *, = A, - /'<
<-._, = Aa - pe, - </<
(e) Coefficient cy (7 = 1,2 ai - 1) in (O required for the calculation of A/a Aa are again
A.., = a2 - />A | - '//' expressible in closed form as
r, = A., - pe, - qcx
b3 = :. - pl>- ~ 'lb\
c = Fb
A = a_ /</'-! - qbm-2 c.x = bm_, - pc_ - qcm_} whcre c is a column matrix of1, r, . r, f,_, and F and b are thc matrices given in (d) above.
-Ar>s
(1) Initial vales of p. a in
~P 0 -p -pcx -/"-.,-. -Ara
~1 0 II -q -?'--* I\{x) = x* - tV + 39xa - 62x + 50 = 0 (m = 4)
r~, r., . , tr,
l r, c, ' r -1 *"-.- .' are taken as pt, = > 1.. . a = ir, 15.
i
I
(h) Convergence and improvement of convergence is discussed by Hildcbrand (Ref. 1.10, pp.
472-477) and Hamming (Ref. 1.09. pp. 108-111).
~1 r 1 " "1 '"-1 '--"1 ' --1 'i '1 :1
144 7.01 SYSTEM OF LINEAR ALGEBRAIC EQUATIONS Matrix Equations Matrix Equations 7.02 THREE SPECIAL CASES 145
(a) Algbrale form. A system ofa simultaneous linear equations given as (a) Diagonal system. If A in (7.0IA) is a diagonal matrix so that
n ~ir *i ~\ f b, "
a,,x, + aX2x2 + + a,.x, = A, 22 A,
a2,x, + a22x2 + + a2x. = b2
- xm _ A. J
a.,x, + a2x2 + + axm = A,
then x, = A,/a,, ,x2 = A2/a22 x A/a
where a ,a,2,... ,a and A,, A2,. .., A. are constants and x, , x2,.... x. are thc unknowns, is
said to bc consisten! if there is a solution for thc unknowns; otherwisc, neonsistent. Thc system ;b) Lower triangular system. If A in (7.01 A) is a lower triangular matrix so that
has a unique nonzero solution if the determinant
r'.. r x, [A.-j
a,, a,2 *2I '22 x A,,
=
aX2 a.,2
=0
U, / L L xm _ - A, _
(c) Upper triangular system. If A n (7.016) is an upper triangular macrix >o that
(b) Matrix form of thc same system is
>,. <ij r *" r a, -i
r xi r A.-i X-2 A.
=
L x,x. _J
CJL L_ *.
A .J
a' - A. .
. <.
thcnx, =bjtm,xn_x =(A_, - /_,.*)//_,._,....,x, =(A, - E'..v.)//.
X B
where A is a squarc matrix [a Xa], X is a column matrix [a X IJ, and B is a column matrix
[a X 1]. 7.03 CRAMER'SRULE
(c) Classification Of SOlutionS. The vales ofx, which satisfy the given system are the soluons (a) Determinant solution of a given system (7.01a) is ofthe form
(roots) of the system, which fall in the following categories:
D.
(1) Unique solution. If D * 0, B = 0, the system has a unique solution in which some. but not D
all, Xj may be zero.
where D is the determinant of A in (7.01 A) and Dx, D2,..., Z> are the augmented determinants
(2) Trivial solution. IfD =P 0, B = 0, the system has only onc solution, x, = xa = = x. = 0. defined below.
(3) Infinitely many solutions. l D = 0, B = 0, thc system is called homogeneous and it has infinitcly
many solutions, oneol which is the trivial solution.
(4) No solution. If D = 0. B = 0, the system has no solution. (b) Augmented determinantS obtained from D by removing the respective column in A and
replacing it by the column B as
A, 12 a,| A, - - a, <:.. a,2 - - - A,
(d) Methods Of SOlUtion introduced in che subsequcnc sectio -s of this chapter fall into two a2, A2 a
A., a,.,
categories: Dx , D2 = D. =
(1) Direct methods producing an exact solution by using a finit^ number ofarithmetic operations A. aB2 a, Aal
(2) Iterative methods producing an approximate solution ofidesired aecuracy by yielding a Although simple in appearancc, this method is too laborious for practical use in cases ofa > 4.
sequence ofsolutions, which converges to thc exact sol ition as the number of iterations Thc number of required operations s a(3 + n2 + a3)/3. Thc application of this method isshown
tends to nfinity ' in (7.04A).
^ 1 ~ - l
where *3 s
Aml-
*-l
I_ xm m.
Am2
A"=i> where rI2 = ax2/ax,, r2X = a2,/a^,...', r_, = -,/* and m, - A,/a ,m2 -
.A A./a.2,... ,'m. = bjam can be inverted by geometric series (Ref. 7.01). Sevcral particular
solutions are given in closed form below in terms of
is the inverse of A discussed in Ref. 1.20. Thc solution by inversin requires a anthmetic
operations and is numerically more involved ihan the determinant solution (7.03) or the other D = I - r,2r2 DX3 1 r,2r2| r23r3 1 - r34r4 D,4=\-
methods introduced in the subsequent sections. If, however, there are sevcral B matrices D,2D3
involved, the inverse A-' may be applied for each particular B wiih a mnimum amount of
numerical work as shown below.
b) Second-order system
lr0 ~
(b) Example. If three conditions are involved in (7.01 A), such as If0 < rv<\, i>
r- -1
2"1
m, X, r 4
3 6 -\ \\Z = 10 3 6 -4 *1! U -30 X,
DX2 DX2
=
3 3
then
1
52 17 2 12 4 20 1 +
>13 DX3 DX3
xt
' 2
-32 30 19 -25 10 -30
263 r2i 1
-8 30 32 22 40 x2 2 -1
-9 =
Du Dl3 Dl3
from which the particular solutions are r32r2X _ r32 r32r23
- -1
1 + x3
D,3 Dxi DX3 L J X J
1.000" *I2 ' 1.604" *I3 " 12.319"
*ll
-2.000 x22 2.243 x23 = -2.965
x2. = =
-526 1052
263 _ = -2 = 4 ^ \DX4 / D,4D3
*l ~ 263 263 263
I +
rn \D,4 ) r23\Dl4 ) DX4D>
The same procedure must bc repeated for B2 and B3.
ijj _i II 1 1 " ~i
(b) Backward procedure is starting with x. and closing with thc first equation in x,, thus (b) Particular case illustratcs this procedure.
transforming the initial set to the special case (7.02A)
|~ l/ao 0 0" "no 120 130 "l 121 131
0 221 231 = A,,
T|A,y0 -2io/uo 210 220 a23o
(C) Number Of required arithmetlC operations in Gauss' and Cramer's procedure is, respectively, _3io/no '- _3IO 320 330- LO 321 331.
NG =n(n2 +3a - l)/6 s 3/6 Nc =(a* +n2 +3)/3 =a4/3 1 -aX2X/atn 0 l 121 131 "l 0 132
a23l 0 1 = A,
which shows he advaniage of the Gauss mcihod T^, = 0 1/a.*,, 0 0 221
=
232
0 a.ilx 2.11 *2 =
A.,, x., = (A..,, - a23,x:,)/a22l 1 0 0 ,.. 4
4 -2 1
0 1 12
2 1 8 0
(e) Example introduced in (7.04A) s solved by the procedure described in (a*) above. 1 0 7.5000 -4.7500 -0.7500 1.0000 0 ,, = >( - 3l(1/4 3
-0.5 ,.
E* y. 0.0667 0 ,., 1.2
k a.u "2l "..M M 1.0000 0 -0.0667 0.2000
= '"-T5--
2 1 12 10 15 2 0 1.0000 -0.6333 -0.1000 0.1333 0 E-a = -,,/7.5 0.4
4
-25 -20 9.2
0 3 6 -4
0 0 8.7667 -0.3000 -0.2667 1.0000 ,,
32 43
2 1 8
-4.3833 -2.5
Numerical check requires t0/4 - EJ3 =F2,. FJ4 - FJ2 = Fn , -F.,/2.5 + r",,/ = F32, ^ and A"' above must be identical to A"' in (7.04*). The numerical check used in (7.0&) is also
whcre u is che sum ofelements in che rcspeccivc row. ^ applicable in this case.
wm mm
I
Matrix Equations 7.09 CHOLESKY METHOD, EXAMPLE 151
150 7.08 CHOLESKY METHOD, UNSYMMETRICAL SYSTEM Matrix Equations
(a) Unsymmetrical system uscd in (7.03c) is solved below as an illustration ofthc Cholesky method
(a) Decomposition process, dcveloped by Cholesky in algebraic form, by Banachiewicz in matrix (7.08). First, from thc given system
fnrm
form, and
and aonlicd bv Crout to machine calculations, assumes that thc initial system in (7.016)
applied by
can be reduced to the upper triangular system (7.02c), 4 -2 l" *i 12
3 6 -4 X., = -25
R la 'y ' tu" ~*i~ 2 1 8 -*3_ 32 J
1 t2 hn Xj
Che following identity is constructed according to (7.08A):
1 *3 x-s
l. |2 '.:. 4.000 0 -0.500 0 0.250 00
*/,. 1 til f|. *l A, I = ~'i, 'u = 4 arc lllc second and third elements ofthc lirst row in Adivided by a,,,
1 /._, x2 A2
'>! ! '23 and thc remaining coeflicients are calculated from the following schemes:
1 " 'in x* A3
'3. l32 k3
Ul '13
1J x _A
Ll 'ni u
122
-
L T
>
X B
'2, 1,1 1',-1
(d) Number of Significant figures los! in thcse computations is approximaiely 0.3a to 0.5n. Ior jhc
(c) Final matrix equation is then thc matrix equation given in (a) above, the elenu :us / and <:, of system oforder a = 3, with required thrce-digit aecuracy, all calculations must bc carried out
which arc now known from (A) and the unknowns ofwhich can bccalculated by b; ck substitution with at least five significant figures as shown above.
as in (7.02c).
A i i i i i
Matrix Equations
7.11 SQUARE-ROOT METHOD, 153
152 7.10 CHOLESKY METHOD, BANDED SYSTEMS Matrix Equations
SYMMETRICAL SYSTEM
(a) Three-band system introduced in (7.05a) reduces by (7.08) to (a) DecompOSltlon process developcd by Banachiewicz for thc solution of symmetrieal systcms
assumes that the initial matrices A and B can bc resolved as
"i /,,
1 '.,
x, d2 A = U'U B = U'C
X} = 3 and thc reduced matrix equation becomes
ux = c
X, _d_
whcre A, B arc defined in (7.016) and
whcre '",, 12 UI3 "'" "
,a = r12,/M = r-2,/( 1- rai<12),3< = *W<1 - ^'23) a,., u.,., u
U = C =
and
rf, =m,, rf2 = -rn, + '''2, rf, = -rV!d, +m; rf, = -r^-A-i +.
The reduced equation shown above is again the special case solved in (7.02f).
(b) Coefficients of of U and C are
11.,., = Va22 - a'f2 ;/.,., = Vfl;(, - U'xs 23
"11 = VaM
(b) Five-band system formed by thc governing relation ",. = ll/l! 23 = (23 - "12".3)/22 34 = (34 ~ IJ*H " *23U2*)/U33
Ui3 = A,,/",, 24 = (24 - ,2,4)/"22 "3. = (33 " M ~ 232S)/S3
,_,.v,_, + rkJ.xxk-i + xk + rttt+txt4.x + rk_k42xk4.2
UXm =ajtt a. - (2, - ".2".-)/22 3. = (3- - "I".- - "nfe)/-
reduces by (7.08) to
= Vfl - , - 2, _ ' "' ~ "."'-... ' > '
12 '11 = (e - ",, - 2,2, .-....-.,)/.-. <i. > I
1 23 t,4 2
Ai = 3
f, = A,/h c2 = (A, - r,a,,,)/,, C3 = (3 - Va " *i)/ia
1 '34 '33
Cj = (0i - r,_,,_,., - - V _ ("ii.)/".-
x _"_
(C) Final matrix equation is (fl) in terms of calculated coeflicients in (A) above. The solution of this
where svstcm follows thc special case (7.02r).
/.,, = r2, 22 = I _ r2iria ''3 - rl3 1, = '3,
^42 = r32 ~ '31'12 23 = ('23 - r,,T,,)/!,, 24 = '/'*! (d) Example. Thc reduction ofthc initial symmetrieal system
33 = 1 - r3,r,s - /32/23 T4 3 2 np'l [~20-
3 4 3 2 x, 28
43 ~ r - r'23 t = ('34 - l.vA.d/1-VS '** = r/^' 2 3 4 3 x3 32
/4, = 1 - r4a<j4 - /43<34 Ll 2 3 4. _x4_ _30j
a) in terms of (A) is
r 2.000 00 1.500 00 1.000 00 0.500 00- r 10.000 001
= I ~ rkJt-2lk-3Ji - Y*-I**-U 1.322 88 1.133 89 0.944 91 9.827 04
.
r
T 3 ^H m i ii '1
154 7.12 INVERSIN BYPARTITIONING Matrix Equations X Matrix Equations 7.13 GAUSS-SEIDEL ITERATION METHOD iss
(a) Matrix equation (7.10A) can be expressed in partition form with a = p + qas (a) Iteration system
X, ~m,~
rA I A^-X,! = fB/i '..
m2
La* I aJIx,J Lb,J r2 *2
'3- *3
OTj
where A, = matrix ofa,-,r, [p Xp], i = \,2,... ,p,j = 1. 2,... ,p
Aw= matrix oaik,[p Xa],i= 1,2 p,k = p + \,p + 2,... ,n 1 xm -m-_
A# = matrix ofa,,, [a X], =/ + l,/> + 2,...,n,j = \,2,...,p obtained from (7.0IA) by dividing each equation by a so that
A= matrix ofaw, [a Xa],* = + \,p + 2,... ,n,l = p+ l,p + 2 a ri2 = i2/uri3 = i3/.ir.4 = aX4/axx, ... ,r,m = axJaxx,mx = A,/an
X, = [xx x2 *,lr X, = l*/+1 *, xm]T r2l = 2l/22 >r23 = 2s/22 r24 = 24/22 - . r2r. = 2./22 m1 ~ ./22
B, =[A, A2 A,]r Bf = [,, A,+2 bm]T '31 = 3l/33. f32 = 32/33. r34 = m/33 r3. = 3-/3J m3 = As/a33
(b) SubmatriX form of (a) yields thc final solution, r.i = anXfam,r2 = an2/am,r3 = aHJam,... , r_Xj, = a.,-n/a. m. = A/aBa
is called a strong diagonal system ifthe sum of otT-diagonal elements r (i ^j) in each row is less
fX,-| _ r M I -MAA;'-|fB,-| ihan 1. If this condition is satisfed, thc system can be solved by thc Gauss-Scidel iteration
LxJ ~ L-NA^A-' I JLBfJ method described below.
whcre A~', A~' arc the inverses of A , Aw . rcspectively, and (b) Starting vales of this iteration process may bc taken as
M = [A -A^A^AJ" N = [A -AVA-'AN]- ...o _ _,
x, mx, x2
(0) _ _, (0) _ (O)
ai2, x3 aij , . . . , xm ai,
A A.
21-! U=[:! !1
.1 2. xi x\ , rXjXj ,x.
j-X
x2 j r2jxj ,... ,x.
j-\
x. _, %,*,
7-1
1.0 -0.5"
where s is the number of required iterations. Under certain conditions this process is rapidly
r 0.6 -o.5][- i -i'i r o 0.11 convergent, but is may also converge very slowly.
-MAMAW - -[_Q5 , 0J[ 5 _|J L-0.5 0J
(d) Matrix formulation of this process is
r 1.0 -0.5-ir-l f] _ r o -0.5-1
-NA^A^, - -|__05 06j|__, jj - [ 0 , 0j X = (I + R + R2 + R' + -JM = [I + Rll + R[I + R|I + |]||M
0 -rX2 -ru -"i.
the final solucin is
-r. 0 -r... r...
(a) Tabular form of the iteration method (7.13) in cases of thrcc-bancl and five-band strong
diaconal
diagonal matrix
matrix equations
cqi is called the carry-over method, llustrated by a numerical example
below (Ref. 7.09).
0.30
0
1
0.29
0.20
1
0
0.21
[~x'l
X.,
*J
-30.0
-21.4
8
0 0 0.50 I J -*4- _-50.0_
*l ri *3
X,
EIGENVALUE
-0.20 -0.50 -0.21
EQUATIONS
Carry-over factors -0.30 -0.17 -0.29
+ 2.670 ^* + I.100
Second iteration
+ 0.440
-0.902 /
Third ileration
-0.529
V -0.230*
-0.112 /
Filil iteration
-0.068
V -0.118*
^^-*+0.020 jr ~^>+0.115
Sixth iteration
+ 0.046
i
Kigenvalue Equations 8.02 PROPERTIES OF EIGENVECTORS 159
158 8.01 EIGENVALUES AND EIGENVECTORS Eigcnvalue Equations
a) First orthogonality condition. If X,, X are two cigcnvcccors oforder [a x l| corrcsponding co
(a) Matrix equation ofthc form, two eigenvalues A,, A,, respcclivcly, and L,, L, are their apparent lengths defined in (8.01a"). then
'a, i A aX2 r 0 i ift *j
o
0 x.7X =
a... a-,.. A [L- \i=j
is called thc lirst orthogonality condition.
A_ . x. . . 0 .
A - IA X 0
(b) Second orthogonality condition. IfA is a symmetrieal matrix (normal matrix) of order [a X),
Whcre a =given constants, A,.v, =unknowns, and i,j =1,2,3 , is called thc eigcnvalue X,, X , L,, L. and A,, A, arc che same as in (a), then
matrix equation. f0 if *j
X,7AX= ,... ... .
' 1-A, fi =j
(b) Nontrivial solution of this homogcncOUS matrix equation exists, ifand only ifthe determinant
s called thc second orthogonality condition.
|A - IA| = 0
The expansin of this determinant is a polynomial of nth degrec in Awhich is known as the (c) First normalization condition. Ifthe unit cigenvectors are defined in tcrms of(a) as
characteristic equation (eigcnvalue equation) of A. The roots A,,A,,....A ol this equation are
called the eigenvalues.
" = 7;x' uw.x-
(C) Eiqenvalues of a herrnitian matrix (normal matrix) are real, of an an.ihermitian matrix (lien
(antinormal matrix) are /.ero or pur imaginary, and of auni.ary matnx (orthogonal n.atnx) arc
0 if i #j
all equal to 1.
1 if' =j
(d) Eigenvector. Corrcsponding to each eigcnvalue A4 there is aset of vales of A. which form thc s called the lirst normalization condition.
eigenvector (column matrix)
Xi = {A-u..V,......\;i} (*= 1.2.3.....n) (d) Second normalization condition. If u,,u, are thc same as in (c\ above and A is the matrix
defined in (o), then
Since thc cigenvectors are the soluton of a homogeneous system of equations. only the rclative
vales Of their componente are detern.ined. giving their di.ec.ioi. bul not their magnitud* so that u7/lu. =
0 ii^j
La, ifi=j
U = \ZXft + A';, + + -V;
is called the second normalization condition.
is thc apparent vector length.
(e) Eigenvector spectra. The matrix containing all cigenvectors of Aand thc matrix coniaining all
(e) Numerical methods Tor thc solution of eigcnvalue problems fall into three categories: (I) Direct unit cigenvectors ofA arc called the natural eigenvector spectrum and thc normalized eigenvector
methods, (2) transformation methods. (3) iterative methods. Thc most important methods in each spectrum, rcspcctively; they arc >-
category arc described in this chapter.
A',, A",,. x,.-\ "l l 12
U,., ._,_,
(I) Example. From thc given matrix equation (a), X -
X2l X,, X,
"2 - A -1 0 [A',1 0
_A',, A', A'_ |_ H, U-, "B., _
- 1 2 - A - 1 .V. = 0
0 -1 2 - A. .A'J _0_
(f) Example. For the matrix A in (8.01/).
thC eigcnvalue determinant equation (a) is A:1 - 6Aa + 10A -4 =0. whosc roots arc A, = '4(2 + \fi)
2 + y/2 = 3.411 214. A, = 2, A, = 2 - V2 = 0.585 786. "4
XrX = X'AX =
For A., with Xy, = 1.
For A, with A',| - 1. 4 4(2 - V2)_
-V2 -1 0 0-1 0~| Ti 0
-I 0 -1 A'.. 0 2 + V2
-l -V2 -l
=
V,
At. 0 u'Au
o -i -V^J La'.J 0 -I -1
2-V5.
X, = (1,V2.1} L, =2 X, ={1,0,-1} , =V2
The rcsult ofall congruent transformations defined in (a) to (t") is a diagonal matrix.
ind similarly for A,. with A,, = 1. X, = {1, V2, 1}./-, = 2.
f :,.,,.,......jg|
II
r?a 1
Eigenvalue Equations
fi? Eigenvalue Equations 8.04 SPECIAL EIGENVALUE THEOREMS 161
160 8.03 MATRIX TRANSFORMATIONS
(a) Gerschgorin'S theorem. Thc largcst eigenvalue in modulus of the square matrix A cannot
(a) Correspondence. The square matrix Bis correspondent to another square matrix Aofthe same exceed thc largest sum ofthc moduli ofthecoefficients o any row or any column. Analytically, if
order if there is a nonsingular square matrix C such that
B = CA *m = 2 kl G- = 2 ki
(b) Contracorrespondence. Thc square matrix Bis contracorrespondent to another square matrix then
A ofthesame order ifthere is a nonsingular matrix C such that
|A| s * orR |A,I for Cmj, a R
B = CrA
(C) Congruence. The square matrix Bis congruent to another square matrix Aof the same order if
there is a nonsingular matrix Q such that (b) Brauer'8 theorem. If Pk is the sum ofall the moduli ofthe coefficients of one row or column
exeluding aa, then every eigenvalue ofA lies insidc on the boundary ofat least one ofthecirclcs
B = QrAQ
This transformation is called the congruent transformation, which may take special form under |A - a| = Pk
specialconditions. The eigenvalues ofAare the diagonal coefficients ifAis a diagonal matrix Dofa lower triangular
(i) IfAis asymmetrieal matrix, then Bis also asymmetrieal matrix. matrix L or an upper triangular matrix U.
(2) IfAis a unit matrix I and Bis also a unit matrix I, then
Qr = Q- (c) Cayley-Hamllton's theorem states that a matrix Asatisfics its own characteristic equation, so
and Q is said to be an orthogonal matrix. that if
(3) If B isa diagonal matrix, such that
p(X) = A" + p,X-1 + p2k"~2 + --- + ...A' + p. = 0
DiagB = QrAQ
then
then the columns of Q are said to be orthogonal.
(4) If Bis adiagonal matrix, the square roots ofwhosc coefficients are real numbers, such that (A) = A" + ptA-1 + PiA-2 + - - - + pn.,A + pl = 0
DiagB = Diag L DiagL wherep, ,p2, ,p,are the unknown coefficients. With Y0 = [l 0 0 0] ,
and
Y, = AY0, Y2 = AY,,..., Y. = AY..,
Diag [l/lffiAQ Diag [l/]y = I
the polynomial matrix equation becomes
Er E
thc matrix E issaid to bc normalized with respect to A and .Y_, + />2Y_2 + + A...Y, + A,Y = -Y.
and yields the unknown coefficients p.
E-' = ErA
(d) Contracongruence. The square matrix Bis contracongruent to another square matrix Aif there
is a nonsingular matrix Q such that (d) Exampie. Using the matrix A of (8.01/),
14
B = QAQr 2" o
where Aand Bl tve the same eigenvalues and their respective unit vector* u:. v, are related as 5 2 I -14
-4-1 0 14
u, = Sv, v, = S~'u,
i o oJLpsJ L -6J
(f) Inverse matricef. A-'.B" have the same unit cigenvectors u,.v,. but their respective The solution of this set yields thccoefficients />, = -6, p2 = 10. p} = -4, and
eigenvalues are 1'A,.
(g) Diagonal transformation. If Sis the matrix o unit eigcnvector^Su. th<m-the similarity- /.(A) = A3 - 6A2 + IQA - 4^0
transformation inj (e) reduces A to a diagonal matrix The roots of this cubic equation, calculated in (8.01/), A, = 2+ V2, A., = 2, A3 = 2- V2, must
satisfy thc conditions stated in (a) and (6) above.
S_,AS = DialAJ
_HL_
is an arbitrary initial vector, at least onc component ofwhich is nonzero usually, the initial isagain an arbitrary initial vector, at least one component ofwhich is nonzero (usually, the initial
vector is taken with all components equal to unity). Using thc iteration sequence defined above, vector is taken with all components equal to unity), and A"1 is che inverse ofA in (8.01a). Using
the iteration sequence defined above, thc process stops when thc unit vectors
the process stops when the unit vectors
v<*> s v(*-o
u'," a uj*""
If this condition is satisfied, the smallest eigcnvalue is
If this condition is satisfied, thc largest eigcnvalue is
a. a yifr'yri*,' = r^'vn** = = ntr'7'!'
A, = AT^/At!,-" = A-^/A^" = s -OAr,*n
and the corrcsponding eigenvector is
and che corresponding eigenvector is
Y. a Yi1'
x, ^ xr
This process is also suitablc for matricesof higherorder.
This process is well suitablc for matrices ofhigher order.
(b) Example. Using the matrix A in (8.01/) and seleccing the arbitran inicial vector X0), the (b) Example. Using the inverec ofthe matrix A in (8.01/) and sciccting the arbitrary inicial vector
iteration process becomes Y,0>, the iteration sequence becomes
2 -1 0 1 3 0.514 496
2 r "f ~8~ 0.514 496
-4 -0.*S5 994
AX'01 = XV -1 2 -1 -1 =
A-'Y(' YV 4 2 I 8 0.685 994
0.514 496. 4
0 -1 2_ l_ 3 2 3_ _1_ _6_ .0.514 496.
The last two iteration cyclcs yi :ld The last two iteration cyclcs yield.
y (8)
yOl
-K vi/at:; \-ii
< JTJ/AVJ l-i'J 'S :B: 'i!,Vi'i"J
3.414 211 53 808 0.500000 3.414 213 0.500000 0.585 787 105.113.0 0.500000 0.585 786
15 760 0.500000 61.573 77
3.414 215 -76096 0.707 107 3.414 214 0.707 107 0.585 786 148.652 2 0.707 107 0.585 786
-22 288 0.707 107 87.07844
i
3.414 215 53 808 0.500000 3.414 213 0.585 787 105.1130 0.500000 0.585 786
15 760 0.500000 61.573 i 7 0.500000
(a) Transformatlon of the eigenvalue determinant equation in (8.01a) to the reduced determinant (a) Transformatlon of thc eigenvalue determinant equation in (8.01a) to the reduced determinant
equation proposed by Daniclvsky,
equation proposed by Krylov,
*ii A dx dxi "i.
"ii A kX2 '' xXm
-A 0 0
k2X A k22 |D - IA| = = 0
|K(A)| = = 0 1 -A 0
"o 1 -A
knX A ku2 - *,._
leads to thc direct formulaton of the eigenvalue polynomial equation
leads to the direct formulaton ofthc eigcnvalue polynomial equation in A,
A" - dxxk-x - dX2X'-2 dXm = 0
(kxx - A)C + (* - A2)C21 + + (*.,- A-)C = 0
where a*,, d2,..., dm are thc coefficients of thc first row.
where Cxx, C2X,,..., CmX are the cofactors of|K(A)| and *,,,kX2 * are the coefficients given
below. Ib) Sequence Of matrix operations required for this transformacin is llustrated below.
(b) it COefficientS are calculated by the following recurrent formulas: (I) First reduction
[* kx2 *,] = lu i2 ..I ~bxi bX2 bxl bX4'
II 12 21 22 2.1 24
B = M4AM4-' =
21 22 si b32 bn
[k2l k,2 k-2*] ~ 1*11 "'12 *.. 0 0 I
0 10 0 0 0 0
M4 = m;
[k.x ktt, - *] = [*-,.. *. *-.J 41 42 43
a4X/at:. -a42/a4S l/4:l ~.h/43
0 0 0 1 . 0 0 0 1
-I 2
(2) Second reduction
(C) Example. Using the matrix equation in (8.01/), kcoefficients are fM c)2 c,:, f,4
l* kX2 kx3) = [2 -1 0] c2, c22 c23 c.,4
-1 o" C = MjBM3 =
2 0 10 0
[k2t * k33] = [2 -1 0] -1 2 -1 = [5 -4 1] LO 0 I 0 J
0 -1 2.
where B is given in (1) above and
2 -1 o"
r l 0 0 0 i r i *0 0 0 T
[k3X k32 k33] = [5 -4 1] -1 2 -l = [14 -14 ]
*.ii -b/b-.: \fb32 -bjb3., -b.Jbv,
0 -1 2_ M, = M7' =
0 0 1 o 0
2-A -1 0 o 0 0 I _ 0
and |K(A)| = 5 - A2 -4 1 = 0
(3) Last reduction
14-A3 -14 6
"a*,, a*,,, a*,, dX4~
in which 10 0 0
D = M2CM.;' =
c-t-nt;:;]-.. C,=,-[_-; :]=e c, =-[:;]=-. 0
0
10
0 10
0
Thc eigenvalue equation in Ais then where C is given in (2) above and
(2 - AK-10) + (5 - A)(6) + (14 - A3)(-l) = 0 'c2, c22 c23 c24 lAai -.-.Aai -<\mA"2i -c-iJt-ix~
0 I0 0 0 I 0 0
and reduces to M, = MJ
0 0 10 ooio
AJ - 6A- + 10A - 4 = 0 o o o 1
0 0 0 I J
which is che same resulc as in (8.01/).
m a M M 1 M ~1 1 1 J
'Sfi
166 8.09 PRINCIPAL AXES, GENERAL RELATIONS Eigenvalue Equations W Eigenvalue Equations 8.10 PRINCIPAL AXES, PARTICULAR CASE 167
(a) Dlrection COSineS. Since the discovery of thc cxistence of principal axes by Euler, the directon
i) Statement Of probiem. The principal stresses and directon cosines ofprincipal axes ofthe stress
cosines of these axes have been calculated by the simultaneous solution of three homogeneous licld given by the normal stresses in ksi as
linear equations, = 10 ksi p = -5 ksi s.. = -5 ksi
T -Tp Tv and by thc tangential stresses
/ = 2 ksi / = 2 ksi /., = 0
Tit-TpJlYpJ Loj
shown in the adjacent figure are calculated by thc geometric
and one quadratic equation, method (8.09i).
si - 83*, - 230 = 0
(b) Geometric method. Physically, the matrix equation in (a) reprcsents a bundle of planes in and yields s, = 10.266 59 ksi, s, = -7.120 23 ksi, and s3 = -3.146 36 ksi.
aP,8P, Yf coordinates, whosc line of intersection s the principal axis. The directon normis of
these planes arc in the cartesian vector form, (c) Direction normis and dlrection cosines defined in (8.09>. (8.09c) are:
N = (7* - T)i + T + TXM (1) In terms ofs,, a, = 4
Nj, = 7;. + (T - T)j + 7;ck NM = -0.266 59i + 2j o, = 5.331 80 x |0-' a-, = 9.91080 x 10~'
n,, = 7:,i + t.,j + (r - r,)k N2I = 2i - 15.266 59j + 2k c, = 6.992 02 x 10_ )3, = 1.32107 x 10"'
and the vector of the principal axis Rp must bc normal to N3I = 2j - 15.26659k a*, = 4.035 98 y, = 1.732 42 x I0"J
N^.N^Nj^so that a,, = 4
(2) In terms of s2 ,
oU^
R = N,. X Na = Nto X N = N X N, N,2 = 17.I20 23 + 2j b., = 3.424 05 x 10 a. = 8.467 39 x lO'"'
Thus the vector product of any two rows of [T - AT,] selccted in eyelie order gives thc vector of Ny, = 2i + 2.120 23J + 2k c, = 3.229 88 x 10 ^3 = -7.248 18 x 10"'
thc respective principal axis.
N32 = 2j + 2.120 23k d, = 4.724 01 x 10 Y-, = 6.837 17 X 10"'
(3) ln terms of s3, 3 = 4
(c) Unit vector of this axis is, then,
N = 13.146 36i + 2j b3 = 2.629 27 *x 10 a3= 1.028 66 x 10"'
e, = Rp/Rp = o> + 6pj + y,k B, = -6.761 56 x 10"'
N23 = 2i - 1.853 65j + 2k c3 = 2.836 87 x 10
whcre Rp is the vector length which as a scalar is always positive and ar . /tJ,, y, are the desired a*, = 3.888 56 x 10 y., = -7.295 54 x 10-'
direccin cosines. Their closed algebraic form is chen N: = 2j - 1.853 65k
9
-1 a-A -1 Xm-l 0
kn
FUNCTIONS
<t>t
n + 1
-1 4-A -I x3 = 0
-1 4-A -1 x< 0
are
A, =4^3 AT^l
A2 = 3 A7=^ A4 =5 AyHhV3-
and the eigenvectors in terms of the same <pk are
"| V3 1 V3 I
|Vi$ $V5 o -{y/3 -i\ft
1 0-10 1
2V5 {y/s o Vs -2V5
J Vs 1 -W5 2
/hichshows the ndependence of cigenvectors of a = 4.
"~! '" I i 1 '~~~i 1 '"'1 "I """"1
170 9.01 SERIES OF FUNCTIONS, BASIC CONCEPTS Series of Functions Series ofFunctions 9.02 POWER SERIES, BASIC CONCEPTS 171
(a) Sequence Of functions in x is a set of a functions ft(x),f2{x), ,/.(*> arranged in a prescribed (a) Power Series in the real (orcomplex) variable x is in the standard form
order and formed according to a definite rule. The sequence is finite if a < * or infinite ifn = .
atxk = a0 + a,x + a2x2 + + amx"
(b) Finite series Of functions is the sum of finite sequence defined in a < t as
where a0,ax ,a2,... ,am are real or complex numbers independent of * and a < for a finite
2/Ax) =/,(*) + /,(*) + +/0) = s(x) series or n = for an infinite series. Two nestedforms are shown n (1.07), and thc application of
nesting is presented below.
whcrefk(x) is the Ath function, a < , and *>(*) is thc sum.
(b) Example
(c) Infinite series Of functions is the sum of infinite sequence defined in a s v < as
" xk x x2 x* x"
y-= i +- + - + -+- + 7
0k\ 1! 2! 3! a!
2/*<*> =/.(*>+/*<*> = *-
where = and the sum S(x) may or may not exist (sec below).
-'('((')))-[ +']
(d) Regin Of convergence Of the series in (f) above s che sec of vales of the argument v for which -.+f(.+j(.+f(. +-+-3))-h]
the series converges.. That is, f
exists for a S x b, then che series s said to be convergene al all points of this closed interval. In hm = hm 1*1 = -77 < 1
I a.jr I I a, I r \x\
some cases, the regin of convergence is an open interval a < .v < a. which indicates chac ac thc
endpoints x = a, x = o the series is divergent. and is divergent if
(e) Uniformiy convergent series. If for any preassigned positive number e. however small, there lim
- I . I r |*|
exises an inceger A* such that
where r is a number (including 0).
\S(x) - Sn(x)\ <
for all a > Ar, then che series is uniformiy convergent, and if the sum of absolute vales of thc
same functions for any x in the regin of convergence is convergent, thc series is also absolutely (d) Interval of convergence ofthe power series in (a) is
convergent. -r<x<r
(f) Continuty. If the functions of the uniformiy convergent series are continuous in the regin of where r = lim
convergence of the series, then their sum is also a continuous function in che same regin. * *<*> la.
and r is said to be its raaVa ofconvergence. Thc series is convergent in chis interval and diverges
(g) DIfferentiability and integrability. A uniformiy convergent series of functions can bc outside this interval. It may or may not converge at the endpoints of this interval.
diucrcntated and/or integrated term by term in the reginof its convercence, and the sum of its
derivatives and/or the sum of its integris is equal to che derivative and/or s equal to the integral
of its sum, respeccively, so that (e) Examples. For thc series in (o),
in a s x ^ b. (f) Unlform and absolute convergence. Thc power seres which converges in che interval a < x < b
converges absolutely and uniformiy for every valu o x withn this interval.
(h) Remainder. The dITcrence between S(x) and S.(x) o a uniformiy convergent seres is thc
remainder Rm+,,
(g) Dlfrerentiability^andintegrability^A-power series can-be^difierentated and integrated term by
*,, = S(x) - S.(x) =/b+i(a) + /+2(*) + term in any closed interval ifand only i this interval lies cntircly withn the nccrval ofuniform
as shown in (9.03). convergence of thc power seres.
"~I ""I '"'"" I 1 ' ~~I 1 1 J 1
(a) Standard Taylor's series. If the function y = f(x) is single-valued and continuous, has all (i) Gregory-Newton'S serles is the diflrence form ofthe Taylor's series, written as
derivatives up to d'y/dx" =/"(x) n a S x < a and the derivativef"~:[x) in a < x < b, then it
can be reprcsented in this interval by the power series in x as (x - a)A/(a) _,_ (x - a)2A2f(a) (x - a)3A3/(a)
/(*) =
=Aa)
/(a) +
V.k
+ ZTFi
2!AJ "*" qii.3
3!AJ
(Lagrange's form)
whcre /?.+, S .
^TI)T(x"fl) where for * = I, 2,.... , A*/(a) is the Jtth forward diflerence of/(a) and A= Ax is the diflerence
(x n)"(x - a) (Cauchy's form) interval. The remainder
, (x - ar'A'*'/(i|)
are the estimated remainders and n lies in (a,x).
/?"+l S (a + 1)!A"+'
(b) Nested Taylor's series for/(x) is whcre n is the same as in (c). The diflerence form can also be extended to the MacLaurin's series
with a = 0. Both forms are particularly useful in cases where f{t\x) is avery involved exprcssion.
/(*) =/(a) +(*-) A[/W() +^=-j] +*,
=/(a) +(x - )[/ +^=-* [/"(a) +-^ [/"'(a) ++i^/-'(fl()]]] +*. 9.04 CONVERGENCE AND APPLICATIONS
where /t+, is the same as in (a).
(c) Standard MacLaurn's series is a special form of Taylor's series in (al for a = 0, written as (a) Convergence. The necessary and sulficienl condition for the series in (9.03) to converge at xand
to have a sum S(x) = f(x) is
f(x) =/(0) + -/*(0) +-r(0) + + -/'"'(O) + R..t Hm R+, = 0
n.
Series of Functions
9.05 EXPONENTIAL FUNCTIONS (Continued) 175
174 9.05 EXPONENTIAL FUNCTIONS Seres of Functions
(d) Nested serles ( = Ax,a = Ax ina)
(a) Definltlons ( = Ax, a = bx lna).
ys = a = e = a = e
Ji y*
whcre = 2.718281 828 (2.18), a,A= consunts = 0, Ina = natural logarithm oa, x = real
or complex variable. .-[-]-('-('-('-('--))))+"
10" i '
.-L'-il-O-'-O-iO--;))))^-
1i "
**"' 2"'
whcre /?,+,, rt.*+l are thc respective remainder estimated in (c).
le) Small argument. The numerical evaluation ofthc series in (c), (d) is theoreucally possible for all
vales of Ax; computationally, the respective series becomes inconycnient for u> 2 and v>.
-i ii i The range of practicality of these series is given in (/) by thc number of terms required for 100
aecuracy and the respective remainders.
(b) Derivatives and integris (a = 1,2,...).
(f) Range of practicality
d"y
b"e" 0.5<sl Ks2 2<uS3
Range OSs0.2 0.2<s 0.5
dx" o o
12 17 22
a 7 9
f?-(-r>-
dx o o
t
11 15 IB
n 6 9
4 X 10"l0 8 x I0"9
//-w-?['-;i5]
e
>3 2 x 10"'- 1 x 10"" 7 x IO"10
- = v a x *..
dx" o o
The same ranges apply for a*', a"\ where ais rcplaced by v= Ax ln ain the table above.
ax (g) Large argument. For 2< Ax = 10 or 2< v= 10, the vales eb\ a*' arc calculated as
," = e"*< = ,-V a*' = e* = a+' = V
(c) Standard series
where M,N = integers and c,d = fractions (c < l,d<.\). ln thcse formulas
JL a u a a
(0 < a < *) e
.u
= eee-e
>'
o*! 1! 2! !
Ai times
-f^=1- +!-...(-i)^ +/?:+1 (-00 < i < 0) and /,' are computed by (c), (d). The same applies in cases ofnegative arguments.
" o *! 1! 2! K ' !
(-b)"+i (-y)""*1 , (h) Example
where ^ R, ;$.
(a + 1)! (a + 1)!"" ( + D! ""' < * )! " ^s.6789 _ ^s^.6789 = 292.627 362 7
and 5., S? is the sum ofa terms of thc series preceding rt,+l, R^, rcspcctively, where eh = 148.413 159 I, and by the series c.' 12 terms tMna = 1.971 707 660.
(0Sr<) (i) Very large argument. For Ax > 10, a> 10 t.e logarithmic resolution ofthe argument must bc
^-2f-'*S*5
*-*! 1! 2! +- +h +~> used. *
-l^Tr-<-'>^-= .*<Lt_aS.D)_
where ln 10 = 2.302 585093, M is an int.gcr, and c,d ^fractions {c < \,d< \) so that
where .+,,/??+, have an analogical meaning. M + c = Ax/(ln 10), d = c ln 10. j
3^^^^^^ =i m I 1 "i '1
176 9.06 LOGARITHMIC FUNCTIONS Series of Functions '& Series of Functions 9.06 LOGARITHMIC FUNCTIONS (Continued) 177
1 1
(h) Derivatives ofsin"ax
i i
<i"(sn' ax)/dxm
1
i- i sin'ax neven
n odd
i >
II
AS, -BC,
sino*
..__!
i
L_
filil;'/' \
T.1S/J sin2 a*
2Cl -I*
sin5 ax -^(5,-35.) -f(C,-3C,)
4
(b) Standard and nested series (- < a < =) 4
n 4 5 6 7 8
j(C,+ *C2) <5 +4s)
sin u
5X IO"" 1 X 10"' 4 x 10"" l x 10"" 1 X 10"'
^2+l
4 5 6 7 8
4(CS +5C, + I0C.)
lo
(S, + 5S3 + IOS,)
16
n
cosa
6x 10"" 4 x 10"" 5 x 10",0 1 x IO-"* 7 x 10""
*1
(e) Large argument. For |a| > n/2, the reduction of the argument to the small argument discussed (j) Integris of sinaxand cosax
in (c), (a*) is obtained from thc following relations, in which A= 0. 1. 2 :
sin a = sin (kn + ) = (-!)*sin v cos a = cos {kn + r1= t-1 )* ros v -fxudx)" (n>\)
(f) Example. If a = 35.678, then 35.678/* = 11.356 660 12 and A= 11. anc r = 0.356 660 12*. n odd
The range in {d) requires a = 7, and j /(*)
V, = sinhaa = sin a = \(e' e~") | y2 = coshax = cosh u = h(e" + e ") n = order of derivative or multiplicityof integral
A= 1,2,3,.-- Sk = (ka)" sinh kah C, = (A-a)" cosh kax
whcre a, x arc thc same as in (9.07a).
sinh ax s, c,
sinh' ax k. i.
2 "
- -I.V,,)
sinh av
A - 4C,) i*
(b) Standard and nested series (- < u< ) - 5(7, + 10(7,1
sinh ax
* - 5S, + IOS,.) B*
sinh ax = "_iy =(' +^ (' ++. 5V+" ' +(2n - 2)(2 - 1 + 2,-1
(C) Small argument. Numerical evaluacin of che series in (A) is theorecicaUy possible for all vales cosh"ax
^ i*
ofa; compuutionally, their practcal applica.ions arc rcscnc.ed ,0 -3 * u* 3as shown in ()
below. cosh3 ax i(C, + 3C,) i* + 35,)
4- 4.)
(d) Range Of practicality ofsinh and cosh series cosh1 ax i& +fc tf
Range 0 < |u| 0.2 0.2 < |ii| S 0.5 0.5 < |u| s 1.0 1 ;. < \u\ < 2.0 2.0 < \u\ s 3.0
cosh' ax ((% + 5(7, + I0C,)
1() ' * + 5.V + 10",)
8 10
3 5 6
n
cosh u
2 x 10"'" 4 x 10"" 3 x 10""
6 x 10"" 2 x 10""'
*?
Jf.>...,-
o
where /(., < a.a2""' and the numerical vales ak and Ak are tabulaced below.
= [1 ~ Aku2] =a(l - 2a2(l - A3u2(\ Amu2))) - or/t.+,
"* At
(d) Small argument. The numerical evaluation ofthe series is theorecicaUy possible for all vales ofa 0< M o.i 0.1 < \u\ 0.2 0.2 < \u\ 0.3 0.3 < H < 0.4 0.4 < |u| 0.5
in its interval ofconvergence; computationally, the series becomes inconvenient for |a| > 0.5 as Range
8 9
shown below. R 4 5
lanhu
2 x 10"" 1 x 10",0 1 x 10"' 1 X 10",0 4 x 10",0
0 \u\ 0.1 0.1 < H 0.2 0.2 < |u| 0.3 0.3 < \u\ 0.4 0.4 < |u| x/4 R.+i
Range
5 6 8 11
R 4
tan u
2 x 10"" 1 x 10"' 5 x 10"'" 1 x 10"l0 4 x 10"' (e) Intermedate argument. For |a| S 1.5, the continued fraction
.i
u
tanh u = 75 -
(e) Intermedate argument. For || ^ 1.5, the continued fraccin 1 + u
3 + a2
I -a2 5 +
3 - u2 oflers a very efficient computational model in which A" is the last integer uscd (A' - 1, 3, 5,...),
5 -
1.3 < H 1.5
oflers a very efficient computational model in which A' is the last inceger used (A' - 1,2,5,...). Range 0.5 < M 0.7 0.7 < |u| 0.9 0.9 < |n| 1.1 1.1 < \u\ 1.3
13 15
0.5 < \u\ 0.7 0.7 < M 0.9 0.9 < M l.l 1.1 < |u| 1.3 1.3 < tul 1.5
Range t.inhu
5 x 10" 3 x 10" 2 x 10" 3 x 10"
"Af+a
4 x 10"
13 13
[(4* 4- 1) fu] =un [(4* +3) f.] =j^. tan [(4* +2> 1] -Can [(4* 4- 4) fu] whcre e~2' iscomputed by the procedure introduced in (9.05). If|a| > 10, tanh a as 1.
1
184 9.11 INVERSE CIRCULAR SINE AND COSINE Seres of Functions Seres of Functions 9.11 INVERSE CIRCULAR SINE AND 105
COSINE (Continued)
(a) Notation (a = ax) .
1
(g) Negative argument
a=l ifo>0 a=l+ifi><0
V
(2A)I . ak
1
j"". )
arcsin (ax) = arcsin (a) = arcsin a = {-X + arceos a
At = ! "
k 4k(2k + 1)(A!)2 -i \
arceos (ax) = arceos (a) = x arceos a = jt + arcsin a
"' '/r "
(2k)\
Ck =
Ck 8*(2A + 1)(A!) Ls
The numerical vales ak, Ak,ck, Ck are tabulated in (A) (h) Table of constants
(b) Definitlons k * Ak ck c,
8 8.011 551 8009 0.827 205 882 4 0.000045 124 2 0.413602 9412
9 0.009 761 609 5 0.845 029 2398 0.0000190656 0.422 514 6199
sin a = X a4a21 +l = a + a,a3 + aaa5 + + a.u'2n+l + fl,_ 10 0.008 3903358 0.859 523 809 5 0.000008 193 6 0.429 761904 8
k-0
(I) Example
(d) Range of practicality s restricted to |a| < 0.5. = V2(I + 0.IC,(l + 0.1 C3( 1 + O.IC3(l + 0.1C4(1 + 0.IC,))))) + /?
= 0.4510268118 and R6 < 1 x 10"'"
0 |u| 0.1 0.1 < || 0.2 0.2 < \u\ 0.3 0.3 < |u| 0.4 0.4 < \u\ 0.5
Range
8 10
(j) Continued fraction
3 x 10" 4 x 10" 2 x 10" 2 x 10" 4 x 10"
^+i
aVl - a2
arcsin a = A.
(e) Auxiliary standard and nested series (-1 < v < 1) I - 1 2a2
3 - 1 2a2
w=-2 - V2I\ +
*-i
cku")' =f2 - V5(l +c,a +c2a: +cmu" +Rm+X) 5 - 3 4a2
7 - 3 4a2
=fE _ -^ [i +Ctu] =- - V[| +C,a(l +Ca l +---+Ca))) +i?.+,l
2 -i 2 9 -
0 /2a - 1\
oers a very simple and elficicnl model for small vales of a in which A' is che lase inceger used
where R ^. ^ )" and arcos; = >T - arcsin v. ..V = 1, 3, 5,...) as shown below.
"+l 2a + I \ a /
(f) Range Of practicality ofthis series is rescricted co -1 < v < -0.3 and 0.5 < v < I. Range 0 |u| 0.1 0.1 < |u| 0.2 0.2 < |u| 0.3 0.3 < \u\ 0.4 0.4 < |u| 0.05
N II 11 13 13 15
Range 0 |u| 0.1 0.1 < \u\ 0.2 0.2 < |u| 0- 0.3 < \u\ 0.4 0.4 < \u\ 0.
arcsin u i
arcsin (1 u)
R.+, 1 x 10""'. 7 x 10"" 2 x 10"'" 7 x 10",0 8 X 10",0 ; The same fraction may bc used for the evaluation of arceosa by means of che relation in (A).
1
:M >v:"l ' 1 "~~1 1
(a) Notation (a = bx) (h) Range of practicality of logarithm transforms in (e,) (f) is shown below
ar-M)"*' 6 = (-D* 0.8 < M < 1.0 1.0 < \u\ < 2.0 2.0 < |i>| < 3.0
Range 0.4 < |u| < 0.6 0.6 < |a| < 0.8
(2A - l)2 (A - 1)(2* - 1) 15
Ct = 10
** 2A(2A + 1) 2A2
arcsinh u 2 X 10"
1 X 10" l x 10" l x 10"' 4 X 10""'
Rm+X
(b) Definitions
h'' i
(c) Standard and nested series (-1 < a < I) arcsinh a = ln2a + a"2 A 11 "" Cku~2]
arcsinh u=u^ r, _ ^] =(I - A2u2(l - yt,2(l -4B2))) " "*.. = In2a + >"2(1 - C.a-2(1 - C3a-*(1 C.a"2))) - aR.+,
i) Table of coefficients
arcsinh ti x 10" 2 x 10"
3 X 10" 4 X 10"
.+ , 5 6 7 8 9 10 11
* 2 3 4
(e) Logarithmic transform of arcsinh ufor 0.4 <|a| <3, wich r, =+V"2 +1- 2and wx - Q
1-3 2-5
3-6
37
4-8
4-9
5- 10
511
6 12
6- 13
7-14
7
8
15
16
8- 17
9- 18
9- 19
10-20
10-21
11 -22
2-4
vx/(4 + vx), is
arcsinh u = ln (a + Va2 + 1) = ln (2 + a,) Both series converge very rapidly, and only a few terms are required for 10D aecuracy.
(k) Example. With n = 4 and ln20 calculated by (9.06),
=ln2 +2wx{\ +X(l +X(l +V( ++27~T"'')))) +*+'
arcsinh 10 =In20 +^(1 - i(10)"2(l - ^(10r2(l - i(10)"2))) +R>
4
where Rm+X ^
2a + 1 = 2.998 222 950 3 and rt5 = 0
(f) Logarithmic transform of arccosh o for l< |a| <3, with i, =a+Va2 - l - 2and w2 - For 10000 < |a| < , arcsinh |a| = arccosh |a| = ln|2a|.
v2/(4 + v2), is (I) Continued fraction
arcosh = ln (a + Va2 - 1) = ln (2 + 2) Vi + u2
arcsinh a =
I-2a
=ln2 +2^(1 +\wt(l +W2(l +W(l +"-+_ ?")))) +**' 1 +
3 +
1 2aJ
3 4aJ
5 +
3 4a2
whcre /?,+., ^ 7 +
2a + 1 9 +
(g) Example. With a = 2. v2 = V3. and ta2 = 0.302 169 479 3, oflers again avery simple and efficient model for small vales of a[0 < |a| =s 0.5], wh^re Nis the
last integer used.
arccosh 2 = in 2 + w2(l + K( + f^IO + ^d + "' ' + ft** + *
(m) Example. With A* = 1.3,5,7,9, 11, 13, arcsinh 0.5 =0.481 211 824 3+8XIO"'0 \
= 1.316 957 897 + 0
"r^" a ^i ~i i 1 1
' Vi + 2
^s
arctanO.8 =f- V^l +j^,(l +g^<,( +"" +JT^'.))) +*
6=(-\)t+t
jps
II
s<^\ X
= 0.674 740945 l and Rw < 2.6 X 10-9
(b) Definitions (f) Large argument. For |a| > 1 only arctangent can be considered, and in I < |a| < 3 the
transformation introduced in (d) yields good results. For |a| > 3,
yx = un"' x = tan"1 u= arelan a
j,2 = tanh-1 bx = tanh"' a = arctanh a B
..-?-s
2 *f,(2A- 1)2*-
where a = constant * 0and x = real orcomplex variable. * l J 1_ x - R.+X
(c) Standard and nested series (-1 <a< l). ~2 a + 3s 5a5 (2A - l)a'
1 _i i 2A-3 -I
- /32'- _ _ _3 _s _ - aR.. _2 aAL (2A-l)a2J
arctana= 2,^r-- 3 + 5 2a - 1
(d) Small argument. Thc range of practicality of thcse series is -0.5 aS0.5. K+i 7 X 10" 7 x 10" 8 x 10" x 10"
or
2 X 10" 2 X 10"
arctanh u "i 1 X 10" 6 X 10" 8 x 10' (g) Continued fractions
a
arctanh a =
arctan a = :,
For thc rcmaining vales in 0.5 < || < 1.0. the transformador.* 1 -
1 +
O)2 (2a)2
3 -
arctan u = arcsin
u
= arcsin ,
3 +
(3u)2 (3a)"'
5 -
Vi + " 5 +
7 +
7 +
-fE-VgTi---
"2 (2*-S)a,> 1-*..
V2/,AL1+4(*-l)(2*-l)J
are limited in their practical application to 0< |a| S 0.5 as shown below where A' is the last
integer used.
arctanh a '.= arcsinh = arcsinh j 0.3 < |u| 0.4 0.4 < \u\ 0.5
0.1 < |u| 0.2 0.2 < |t<| 0.3
VP-7 Range 0.0 < |u| 0.1
15
! ; r 2A - 3 .,1 _, II 13
=in2 +2,A[i+<;J +*:+l arctan u
A* 7 9
Seres of Functions
9.15 PARTICULAR PRODUCTS OF POWER SERIES 191
190 9.14 ALGEBRAIC OPERATIONS Series of Functions
WITH POWER SERIES (a) Notation
_ (V2")*cosA0 (-yfo'cos*
(a) Summation theorem. Two power series in xcan be added or subtracted term by term for each a.b consunts > 0 A/M
A!
Nxx A!
Their sum is a ncw power series in x which converges at least in the common interval of a
(b) Multiplication theorem. Two power series in xcan be multiplied term by term for each valu ofx (b) Winkler's functions of the first kind (- < x < )
common to their interval of convergence.
. ^ vfac*)1 sin (0 + kto)
V ()'
tf*> eosAa*
co t" (cos* 4- sinbx) = j Jj
t cos ex = 2/ 7T
(akxk)(bkxk)=ckxk -o
A (exf sin Aa
*
V2(rx)' cos(0 + Aui)
t" sin bx = j r "* (cos 6x - sin bx) = 2
Their product is a ncw power series in xwhich converges ac least in the common interval o i-o *
convergence of the two series. The coefficients ck of the new series arc defined by the following " cosax = 2 A/,j(a*) e"(cosax + sin ax) = 2 M3(ax)
k-o
matrix equation. k-0
(C) Divisin theorem. The quotient of two powcr series in x in their common interval of ~" sindar = 2 A*iji(ax)
i-O
e~"(cos<ur sinax) = 2 A',.i(ax)
convergence,
(d) Factors
t akxk
MXJ Mi.t A'i,
bkxk 4" O
O
is anew powcr series in x, and its interval of convergence -if any) cannot be determined from the 1/1! 1/1! -l/l! 1/1!
-2/2!
interval of convergence of the two series. The coefficients d. of the new series are defined by thc O 2/2! O
2/3! 2/3!
-2/3! 2/3!
following matrix equation. -4/4! O
-4/4! O
4/5! -4/5!
-4/5! -4/5!
O 8/6!
O -8/6!
a, -b, -8/7! -8/7!
8/7! -8/::
16/8! O
a, -b-, -a, 16/8! o
-16/9! 16/9!
16/9! 16/9!
a, b3 a. i O -32/10!
O 32/10!
4 b4 3 2 ~bi 32/11! 32/11!
-32/11! 32/11!
-b, -64/12! O
-a*-b,,-b4 ^j T^ -64/12! O
64/13! ^64/13!
-64/13! "=647131
O 128/14!
O -128/14!
-128/15! -128/15!
128/15! -128/15!
where each dk is evaluated by using che preceding dk_, . </,_: ,dx,da.
i'im
Mi I I
192 9.16 TRANSCENDENTAL OPERATIONS WITH Seres of Functions Series of Functions 9.17 TRANSCENDENTAL FUNCTIONS 0F 193
POWER SERIES POWER SERIES
(a) Exponential theorem. If 5 is a powcr series in x. such as (a) Notation a = constant > 0 /} = auxiliary Bcrnoulli number (A.12)
<44_, cos r;r [i/' tan ax\
S = aQ + a,x + a2x2 + a3x3 + = 2 **' Al ~2*(2* - 2r)!(2r - 1)! Ift-SU
(2A)(2A)
Y /Jj-.sin r;r 41
^ = 77 S4
/,J "*S*(*- r)l(r- I)! (2A)(2A
(4' ~
then e5 = e" 2 <S a.2 ^
rr,A(* -r)t(r- I)! 2A)(2A
where < = 2.718 281 828 (2.18), c0 = 1, c, = a, , and for A > 1. (b) Exponential functions (A = 0,1, 2,.... = a;
/ u1 4b' 31u6 , .,, \ (-,< <=c)
u- 3b' 8hs
or in matrix form, 4- Z),B* 4- (-< <x)
- -
1
1 u2 3' 9a* , (-* < u < |jr)
1!
2 ~ a, <*i
2! '
(c) Coefficients A,, tf,. C4
91 1!
- 2a, c2 ct
3 k '< l>,
3! " S"
0 1.000 000 000 (+00) 1.000000000 (4-00) 1.000 000 000 (+00)
31 1!
4 - 3a, *3,,'
4!
i
4! '
''< 1 -5.000 000 000 (-01) 1.000 000 000 (+00) 1.000 000 000 ( + 00)
4! 2 1.666 666 667 (-01) 5.000 000 000 (-01) 5.000 000 000 (-01)
3 -4.305 555 556 (-02) 0.000 000 000 (+00) 5.000 000 000 (-01)
91 1!
4! 3! 3.750 000 000 (-01)
- 4a., 1 9.399 801 587 (-03) -1.250000000 (-01)
ar 7T4fl4 r:o3
.t!
1'.
5! 5! 5! *
5 -1.806 657 848 (-03) -6.666 666 667 (-02) 3.333 333 333 (-01)
J L 2.458 333 333 (-on
6 3.138 799 536 (-04) -4.166 666 667 (-03)
(-01)
Thc coeflicients ck of thc ncw series, which is convergent in the same interval as S, arc evaluatcd 7 -5.016 685 851 (-05) 1.111 111 111 (-03) 1.902 777 778
7.470 220 791 (-06) 5.381944 444 (-03) 1.373 263 889 (-01)
by using thc preceding ck.x,ck.2 c,. c,. t
1.763 668 430 (-01) 1.132 082 231 (-01)
9 -1.046 251 198 (-06)
(a) Power theorem. The mth powcr ofthe power series 5 in xis
10
or in matrix form,
wb. "'o*
" 'i "
SPECIAL
"
FUNCTIONS
ti
The coefficients ck othe new series, which is convergenc in che same interval as S, are evaluated
by using the preceding f4_i, ck.-2 c-,,c,, c0.
m m ". m *kr
m wkt ',
(c) Example
Vi! 3! 5! / k-o
where *, = 0,i2 = -1/3!. A, = 0,b4 = 1/5!. *5 = 0,...
_ i ... 1 ... = -1 M = 2 IJ = 0 031
= 2 K*4S = B "42 ~
*- =(>)(!)-o '.-()(-5)("*(-)<0,(0,""i5
<, =(i)(0|(i) +(0)(-i)(O| +(-?W-i)-o
196 10.01 INTEGRAL FUNCTIONS, GENERAL RELATIONS Special Functions Special Functions 10.02 INTEGRAL FUNCTIONS, SPECIAL 197
VALES AND GRAPHS
(a) Origin. J. Liouvillc proved in 1833 chac thc following cieht integris cannoc bc expressed in
tcrms of clcmentary functions and can only bc evaluatcd by inCcgraCing their power series (a) Special vales
expansions. They arc:
Si(0) = 0 Ci(0) = 00 Si(0) = 0 Ci(0) = 00
f sin / , ^, /;,
px (b) Graph of Si(x) (c) Graph of Ci(x)
i - x < a: < oo)
s(x)= -r*-2r-n
A_lV2A - 1)(2A - 1)!
f'eos/ . , ^ px
ci(x) =Jf*-r +i-22^5 l(j < x < o=l
;i i L
i \_j-4 js <*-r-
(" sinh t A -v''"' 1' i ow
|-2 < X- < 00) \< <_ -,lH-
S(V) = T* ~i(2* - 1)(2* - 1)! II- 7
o 1 2 3 4 5 6 7 8 9 V
cu,) =__<//
f* cosh i , =y+ln.v
, +i?___
^ x31 (0 < X < 0)
(d) Graphs of Ei(x) and i(x) (e) Graphs of erf (x) and d[erf(x)]/dx
E[{x) =Td! =Y+lnx-%klk lO < X < o=)
> i
i lx)
( - X < x < oo)
IV. i ii ri 1.0
1 1
f I A (lux)* (0 < -r < oo) V
K>l
and in general.
W-a1-=('-= Af9
where Af = aur, N = ( - s)*, m= 7.!!.!).... , and n = (-1)"*'. For lower mthc numerical
E,+M = TTi'H =-U" -xE,(x)]
J\ i r
vales are listed below.
( 1 - x<r*,(x
- 1 "T"
Si (../) 1.851 937 1.418 152 1.674 762 1.492 161 1.633965 1.518034
where r = 1,2.3 and (x) = e '/x. Ci(iV) 0.472000 -0.198410 0.123770 -0.089 564 0.070065 -0.0575)01
.
mam
Special Functions Special Functions 10.03 INTEGRAL FUNCTIONS, NUMERICAL 199
198 10.03 INTEGRAL FUNCTIONS,
EVALUATION (Continued)
NUMERICAL EVALUATION
(a) Small arguments. The range ofpracticality is rcstricced co 0 < x < 1for (d) Rational polynomials uscd in (c) are
m (Ax + x(A. + x(A3 + x(4 + x)))) m x
(2 - 3)x'-' A[> x(/l, + x(/l, + x(A7 + x(AH + *))))
si , ,i -li
2 9 V
-^-i
4 25 \
- 6-49
!X (i (2 - 2)(2n - I)1'
+ R.+i
= (/il+x(/iJ + x(^, + x(/i4 + x))))
(2n - 2)x2 (; x*(/r5 + x(Zjc + x(B1 + *( + *))))
Ci(x) = , - -
4 V
1 --
2x*
3 - 16 V 5-36
4x-
1 -
(2n - l)(2n + 2) i))) +^.
= (. - *( - & - -( - * + |f = 2X io-8i
( " l)x\ a,, - x(a( - x(a, - x(fl - x))))
i /, - (1 --^i -jV (, + x(L. + X(b, + X(4 + X)))) =
(fl - 1)x 6(A' (, + x(, + x(67 + x(6H + x))))
.4-.
r(x) = I - -(c, - (r, - b(c3 - u(c4 - cs))))<--' + |e = 1.5 x 10*;|
in which Ak, /i,, ak, />,, ck arc thc numerical coefficients tabulatcd below and
' ;i+i(l+f(1+-'-+^J ^
(2/i - 3)x*
=il/I_^l1-^l,-^|,
^^v^V'1-^-^ n - \)(2n - 1)
+ *+I 1 + 0.327 591 11 1a.
r> (i 6 i <l 10 (f) Very large arguments. For x > 30, che following asymplocic series yield fase and accurate results.
Eiui
3 X 10""' 6 x 10*" : x io " 6 X 10*'" 2 x 10""
2 x I0~10
K.,i
S(x) = - - (x) cos x - B(x) sin x Ci(x) = (x) sin x - B(x) cos x
t> 8 11
:i 1 5
crfl.l 1 x 10~"
3 x io~" 3 x 10"" 2 x 10"* 8 x io~'u 2 x io""
tf.+i Ei(x)=-(x) Ei(x) = 6(x)- erf (x) = "(x)
X x
(c) Large arguments. For 1< x< 30. che following approximatc formulas developcd by C. i' vi--(l--(i
Hascings (Ref. 9.03) yield good resoles.
Si(*) = - - A(x) eos x - B(x) sin x Ci(.v) = .4(x) sin x - B(x) cos x > +r(-+?(-+?l- i)))) + Q
Special Functions
Speciai Functions 10.05 ELLIPTIC INTEGRALS, CLASSIFICATION 201
200 10.04 FRESNEL INTEGRALS
(a) General forms. Integris of the form
(a) Basle forms. Related to the error functions are Fresnel 1>
52(x) =-,(-x) =^= I -^dt =>/5(2t+i:.4* +3) (3) Integral ofthe third kind
_ r dt
i rcos/J [x (-'V m'a'x) ~i (l +r2)V(l - (\ - k2t2) 10* JO* .** 411" 50* tV Ti"* m*
(a) Notation * = sin w = given modulus (10.056) f) Examples. lk = 0.7 and Ol, cosa, and.4 are the same as in (d), then by the first formula in (e),
E = 1.355 661 136 + (= = 1 X I0-9)
K, E = complete elliptic integris (10.05c)
If it = VO.95 and k,, k2, k3, k4 are the same as in (d), then by thc second formula in (<),
F(k, <p), E(k, <p) = incomplcte elliptic integris (10.05l
E = 1.060473 727 + (e = 5 x IO"10)
2
A = *, =
i + VT-7*-
** =
i + VT-a2 i + vn--*2 (g) Evaluation of incomplete integral ofthe first kind (m = k2)
1 + VT-P For 0 < ft < 20
For r = 0,l,2,
, thc limit must bc evaluated by LHopital's rule and yields with u- 2x,
1 1
1 t j. + - C = + f.-i +
,-'(r-A- 1)! .t
r-l . = .-1 + ~ = -i + n \ n n - 1 r + - 1
YAx)
,v
=-2(y
Xl
+ InaU(x) - "' *-
2 1
' (-1)-|V>(A) + y>(r + A) + 2y] al and ?0 = 2(y + ln u\J0(x) B, = 2(y + ln ^.(x) -
+" *Tc7T7)i J
/ (r-2)A(,
(r- 2)4/ ., (r-
(r - 3)/,
3) , . (L~IT|Y\
Y_t(x) = (-l)'K(x) forr = 0, 1,2,..-, only
where by (2.32rf), (C) Graphs Of Jr(x) and Yr(x) for r = 0, 1. 2are shown in (10.9a) and (10.10a), respectively. Selectcd
special vales of these functions are tabulatcd in (10.96) and (10.10A). Also,
y+y(k) =1+*+5 y+ -ii(r +A) =1+^+i + r + A
yr() = 0 Kr(>) = 0
v,(0) = -y y= 0.577215665,2.19) (d) Numerical evaluation Of Jr(x) and Y,(x) is shown in (10.11). Numerical tabics of0(x)J,(x) and
The Bessel function Y, is also called the Neumann function /V,(x). Y0(x), Yx(x) are given in (A.16), (A.17). Thc vales of higher-order Bessel functions can be
calculated in terms of lower-order vales by the relations (10.07a*) as shown below.
(d) Recurrent relations for all r=2. 3 and with u=\x are (e) Example. Ify0(4) = -0.397 149810 and7,(4) = -0.066043 328, then u= $, and
- 3 l
r - I -I -
u r *<>(*) 1
[*,(*) 1 =
R.gx)\ r
u
r(r - 1)
2 1 -2
- 1
U(*)J 74(4)1 _ -i "' i rjo(4)l
llx(4)\
[0.281 129 0651
lO.132 086 656J
1-
- 1
n tt -Uwi 4 43 _2 21
=T ^=-r
2 4 _ 2 4
fe
where R,(x) isf,(s)or Y,(x). ja
Special Functions Special Functions 10.10 ORDINARY BESSEL FUNCTIONS OF 207
206 10.09 ORDINARY BESSEL FUNCTIONS OF THE
THE SECOND KIND 0F INTEGER ORDER
FIRST KIND OF INTEGER ORDER
(a) Graphs of J0(x)and J,(x) (a) Graphs of Y0(x) and V,(x)
7.w
Numerical vales of Y0(x) and Yx(x) for selected argument xarc cabulaccd in (A.17).
Numerical vales of0(x) and Jt(x) for selccted argumene varc cabulaccd in (A.I6).
c) Derivatives and integris
(c) Derivatives and integris
i I
{ )Ax)dx = 22 Y,+,ltA*) f-F.(x)rfx = -F,(x) +2*-<> Ftt*,(x)
J,(x)dx = 2yf+ai+,(x) -Mx)dx = -/,(.v) +2 2Jn*i(*) J k-0
J X
where che single primes and double primes indicaic thc first and sccind derivatives with respect where the single primes and double primes indicacc che firsc and second derivatives with respece
lo X.
.
W
a,A
Fo(3)=^[Bo +2^(.-f(^-|(fl3-]6(^--- 19
+ , ,, (A.) = -J Yr+in(x) = sphcrical Bessel function of che second kind of order r
= 0.376 850 44 - 4 X IO-7 / .(x), Fr+I/;!(x) = ordinary Bessel functions of order r+ \ (10.07)
whcre and C|, C-, = constants of integration.
f0 = 2(0.57721567 + ln 1.5)./(3) = -0.511 096 89 (c) Graphs of y0(x) and y,(x)
(b) Graphs ol/0(x) and/,(x)
a, = 1 a., = a, + 5 3 = 2 + 1 a, = as + 5
l.lxi v.(.)
, 1
(C) Large argument. In 3< x< <, the following polynomial approximacions incroduced by E. E.
Alien (Rcfs. 10.02, 10.03) yield good results.
where wiih .4 = 3/x and in tcrms ofd, ,bk,ck, <lk tabulated in (d),
a = fl, + .4(<7,, + /!(, + K + + a,A))) + *",
= /;, + A(b2 + A(b3 + A(b4 + + 6,-4))) + -. + e,
(d) Spherical Bessel functions for r = 0,1,2,... can be expressed in terms ofclcmentary functions.
c = r, + -4(r, + .4(r:1 + A(cA + + e7))) + e.
d = dx + A(d, + .4((/j + 4(4, + + -,-4))) + - + et sin x
.>'<>(*) = -
jo(x) =
(d) Factors ak,bklck,dk sin x cosx
cos x sin x
J(6) = 0.150 645 26 + 1 x 10"" I'(61 = -0.288 194 68 - 1 x 10-" (e) Example. For r = 1,
1 /sinx\ mu_x cosx _ cos x
c = 0.801 932 33 d = 3.705 587 20
j,(x) =-j0(x)-j-i(x) or -x(x)=a
y((6) = -0.276 683 86 - 1 x I0~9 )',(6) = -0.17501034 - I X 10"
3 ~ "i
220 10.13 MODIFIED BESSEL FUNCTIONS, Special Functions Special Functions 10.14 MODIFIED BESSEL FUNCTIONS OF 211
INTEGER ORDER
GENERAL RELATIONS w
(a) Origin and definitions. General integris ofthe modified Bessel's diflerential equation (a) Serles representations of lAx). With u = x, A= (x)2 = u2, and r = 0.1,2,...,
x'
dx1
+^-^
dx
+rV-O (r*0) ^-(^('+ri(' +ri(, +"-r-J))) +-
_ rC,/,(x) + C2/_f(x) r* 0,1,2....
arC y ~ lc,/,(x) + C2KAx) r = positive real number
where /r(x) = modified Bessel function ofthe first kind oforder
r(x) = modified Bessel function ofthe second kind oforder
(' +i(TTi)(' +5(7^(' +SfrTs) (' +"' +577^))))+ '
C,,C2 = constants of integration
(b) Series representaton of /C,(x). With u= l2x, A= (x)2, r = 0,1,2 and a = (-1)'*
(b) Modified Bessel functions of the first kind are defined with u=x as
3 ir
r r(r- 1) -- 2 (r- - l)(r -3) A\(4) *o(4) 0.062 229
+ 1
r
+ l
J L
2u l'4*x
u
Ii(x)
J L a
.44^3-
The function K,(x) satisfies the same relations with /r(x) rcplaced by At(x) and by removing the ir *5(4)
L2 4
+ T ^r(4r 0.154 343
/oW = 2E(-i)V,W A'oWi = 2E(-l)ViW 4 -0.010 38637 0.010 797 43 -0.010624 46 0.01504268
5 0.11324884 -0.127 49630 0.005 878 72 -0.007803 53
6 -0.47685867 0.52906130 -0.002 515 40 0.003 256 14
1.14518790 -1.258064 78 0.00053208 -0.000682 45
/,(*) dx = /0(x) KAx) dx = -A'0(x) 7
8 -1.342 361 19 1.456439 23
9 0.599 39583 -0.64168291
/r(x)flx =2(-l)/r+2+,(x) A'p(x)v =2 (-\)kKr+2k+Ax) 1*1 1.9 X 10"' 2.2 x 10"' 1.9 x 10"7 2.2 x 10"7
x'K,_,(x)dx = -x'KAx) x-7Cr+1(x)flx = -x'A,(x) /0(4) = 11.301 92214 - 1.9 x 10"' 7,(4) = 9.759 465 14 + 5 X 10""
c = 1.218 595 34 -fl*_=_L363J51 89
where the single primes and double primes indcate the first and second derivatives with; respect A0(4) =0.011 15968 + 0 A, (4) = 0.012 483 50 - 1 X 10_
tox. \
fra ^^^^^r m n.
r-=?=?
(b) Graphs of /0(x) and /,(*) (c) Graphs ofk0(x) and *.(*)
POLYNOMIALS
(d) Modified spherical Bessel functions for r = 0. 1.2 can be expressed in terms of elementary
functions as
sinhx
k*\x) =
io(x) = 2x
cosh x sinh x
O) =
i-x(x) =
coshx
A-,-)= (i-"-,")
sinh x cosh x *_.x) = (1 5)
i-Ax) = - 2x V x x /
EQUATIONS, BASIC RELATIONS (a) Notation. k,p,q,r = positive integers ak<bk<ck = real numbers
n = order of dcrivativc F(k) = function of A x = parcicular point m = real number
(a) Notation. A: = O, 1,2,3,...
Dky = dky/dxk = th-order derivative of> =y(x) ck = real number (b) Change in limits. If
(b) Basic concept. Thc nth-order linear diflerential equation with variable coeflicients, F(k) =P(p) +F(p + I) + F(p + 2) + + F(q)
k-t
fm(x)D'y +fm-Ax)D"~y + --+f2(x)D2y 4-f,ix Dy+f(x)y =g(x) whcre q p is a positive integer, then for q = , * = r p.
can be solved under condicions defined in (r), (e) by one ofthc following powcr series. 2 akF(k)x"K = a,_,F(r - p)x'
(1) y=Co +C,(X " X) +C2(X ~X0)a +C,(X - V. ;+-+ft(x - X0)' +' ' k-0 ,-p
(2) y=(x - *Tfo - ',(* +*) +c*<* - v'r' " Ck{x " A"):' +" ' +f(V " V,,) +" '' (c) Example. With k = r - 1,
both ofwhich may bc ofasccnding ordescending powcrs ofx.
2(2A +l)**+' = 2'"'(2r - D.v' =x+6xa +20x" +72x< +
(c) Ordinary point. Thc real number x in (I), (2) is said to bc an ordinary point or the given
diflerential equation if and only il Equivalence. If **' = X bkxk then ak = bk
/,(*..) * 0
and
/,_,(x),/n_,(.v).. -,/,(x),/,(.v),/(x) (9) Example. II"
possess aTaylor's series expansin (are analy.ic, in che neighborhood of x0, which may be ,ero_ kckx' =b0x + Ma +*>e*s + b3x4 + +bkx1*' +
' ,s point. che given diflerential equation posse.es linearly independen, senes soluons o
Z type (1) r (2). The radios of convergence of diese series is the d.Stance .rom v to the nearest k-0
f,+2=--T-(^->)
M^^f^d+Ux)y=ig{x) which in many cases reduces to a very simple exprcssion.
where
(e) Final SOlution ofthe given differential equation in (a) becomes
/-/ ^- Yn xk fAx) = m** /->(x) = ^ floJ,x* g(x) = bkxk
i-o
y=C,L,(x) +C2L2(x) |
can be solved under certain conditions (11.01c) by the power series where the twoindependent shapc functions are
y=c0 +c,x +c2x2+ = 2^*' L,(x) = I + c2x2 + c4x4 + s(*) = *+ *s** + 's** +
and C, =c0,C2 = c,. The application of this procedure s shown in (11.04).
with derivatives
fl Frobenius method. Many second-order equations do not lend themselves to the general solution
by the powcr series (1) introduced n (11.01*). In such cases the series (2) n (11.016) may y.cld
che desired solution. Thus for
n which c0 ,cx,c2,... are unknown coeflicients.
-o. -^^r-s2sd,!3
"XTo"toTi toT;1 rfUM. n,ay surit&xr
II brale .uauons ,n wh.ch ,
where
coefficients arc the unknowns. there exists at least one solution of the form
Aoo Aox Avl l~f" "*"
Ax0 Axx AX2 AX3 Cl i y=x- ckxk
k-0
<-., b.
A-m A2X i422 A23 A24 = n the neighborhood of x0, ifthe condition stated n (11.0U) s satisfied.
Ajx A, *)J+* CJ b, (g) Indicial equation of this solution, obtained by equating to zero thc coeflicients of xm. is with
fijo
L - '- L- -J R(x) = 0,
m(m - l) + mxP(x) + x'&x) = 0
where and with x= 0 yields the roots mx,m2. which may be unequal, equal, or difler by an integer.
Consequently, three types o solution may occur as shown in (11.05) to (11.07).
AJJt =k(k - l)2./-*+2 +*ij-k'+i +a-' 1
(h) Modified Frobenius method. At aregular singular point x che solucin muse bc taken n the form
as shown in (c) below. SC
(c) Coefficients \k
an1thc indicial equation in (g) muse be changed to
im(m - 1) + m(x - x0)P(x) + (x - x0)2Q(x) = 0
wl'ich with x= x0 yields the roots ro,.., which may bc of the three types defined in (g) and
"oo agiin three types ofsolution may occur.
+ 2a, 2 3a_v
a + oo 3-4a
<ox
fl|j + <*oi ^Byr*-2tfTr+H,0(r -2^3a-^
3-4a + 4a l0 "4^3ai5
(i) Two Classes Of SOlution are shown in-the^ubsequent sections^according to the type of point at
is + aoa 2aI3 + 2a,2 + a, 2-3aj + ."ii + aoo
2- 3a.., + i:,.. + oi 3 -4a,, + 4an + aoo 4-5a + 5fl| which the series expansin is applied. They are designated as ordinary point solutions and regular
2a,4 + 2a + aj
<l|4 + flOJ siniular point solutions.
^ JP 4 ^ \ m ^1
5 6 7 8
X 2 3 4
0 1
>
0 2 l -2
2-3
where according to (11.03a), a*. = 3, a = 1, a.M = 2, and all the rcmaining a coeflicients are
l 41+2 /.ero. Assuming the series (I) in (11.01*) to be thc correct form ofsolution, the coeflicients .4,< are
2 4-2 + 2 ; 4 calculated by (11.03c) are
4-3 + 2 4-5
3
where am =2, a =4, ax =1and all the rcmaining acoefficients are zero. From this table by ,, = 41+3 4,, = 5-6-2
the relations ( 1.03</) the coefficients ofthe serie* solution are related by
4A + 2
(A = 0.1.2....)
f.+s = ~ from which the recurren! relation [\ 1.03a*) becomes
(A + 1)(A + 2)
By sciccting C, =c0 and C2 =cx as the constants of integration the ccoefficients becomc k + 1
ck = -; ct.3 (A = 2,3,4,...)
2A(A - 1)
c, = C2
c0 = ci
Bv sclecting C, = cQ and C = ?, as the constants ofintegration, the ccoefficients are
*-(-Pi)*
3-3-7 4-6-8^
C7 = T7ZT- C,
*-(-)(-)' <o = Cx C| = _ ,
8(6!)
ci = C,
8(7!)
3-5-7-9 4-6-8-10
C.
f"i =("'^2)("3^4)("5~6)C, (-M-SArifr- (2~ 2(2!) C' 16(8!)
-c,
f:I " 2(3!) C"' 16(9!)
_ 4-6 c
*-(-M-M-Yrifr- 4
-3'5r
O = TTT7T C,
4(4!) 4(5!)
As in (a), the numerical factors in front ofC, and C. are the scaled ccoefficients incroduced in
where the numerical factors in front of C, and C2 arc the scalcd c coefficients introduced in (11.03?)- Because ofthe form of these factors, thc resulting power series may be shown in nested
(11.03,). Because of their form, the nested series solution offers a very compact representaton form.
shown below.
--c,(>-4>-M>-*v--_^m
.^-x-M'-r-A'--i'- ))))) <*('-^('-S-A'-^-A'-^'--m <-='+2> i
l^^^"r^^^^*^^F ' 1 1 1? "1 J
r-JUi
m2 = 6 then the first shape function L,(m,x) is thc same as in (11.05a) and the second shape function
I"""'
are nscrted in the series of (11.03/), the final solution ofthe given diflerential equation becomes must be taken in thc form of
again dL,(m,x)\
L,(m,x) = Lx(m,x) ln (x - x0) +
y = CxLx(a,x) + C2L2(6,x) dm l-o
where the two independent shape functions are where dLx(m,x)/dm = partial dcrivativc ofLx(m,x) with respect to the root m. The final solution
Lx(,x) =(1 +ci.ix +c,,x2 +)** Ufi,x) =(1 +-C2.xx +2.y- +V is
f x dLx(m,x)\ 1
and C, =clflt C2 = c2f, cx, = c,,/C,, c2J = ra,/C9 as shown in (c) below. y= C,Lx(m,x) + C2\ Lx(m,x) ln (x - x0) + 1 J
(b) Recurren! formulas for the calculation ofc and c2 to be used in the final solution are where C,, C2 are thc constants ofintegration and in this case cXfi - l,cM, = 0.
{!}) Example of solution about x =0. In
c,.,=rf ^* - o '*-=f^~(^ 5'<fa*_ *j)
d2r dy n
in which AJ = (k + )(k + a - l)2J-t*2 + * + *),/-*. + oj-* xT + T + xy = 0
dx' dx
fl,, = (jfc + 0)(* + /3 - l)2.,_^_. + * + P)'ij-i+i + <V>-* the selected point x0 = 0 isa regular singular point, since by (11.03/),
and j.k are the row number, column number. rcspcctively, in the matrix equations formally xP(x) = I x'-'t2(x) = x-
identical to (11.036).
are analytic. The indicial equation (l\.03g)
(c) Example ofsolution about x0 =0. In 0
m(m - 1) + 5+/= 0
0 i, .,)->-(>-( -?('-!< ->)))
yields two unequal roots m, =*=0. m,, =0 =4and indicates the solucin given in (a) above. and the second shape function with m = 0 and
The recurrent formulas by (b) are
( -l)\ 1 1 1
Cl.k
dm (2 4- 6--- 2*)-'
2la(2o- - 1) + (4a + 2k + le* + 1)]
fiji+i
c.i
is L,Kx) =l(OT,x)lnx +f(.-^x^(.--x^(,--x^(l -))))
= ~2[/3(2/3- 1) - (46 + 2A- + l>t* + 1)1 (2* + 2)(2* + 3) where thc partial derivative of,<m.x) with respect to mis calculated by means of(11.02/). For
example,
and because ofthc form oftljeae factors, the resulting powcr series may be shown in nested form de, o 3(l)| *,,.,_ 3[l/(" + 2)2J| 1
as
dm dm L-o dm dm
che selected poinc x0 =0 is a regular singular pe,:, bv (11.01/), and the indicial equacion for m *fc n
(Ax) = Legendre polynomial ofthe second kind ofdegree r ,1 -3-5---(2r - I) 0 for r even
M) 2 - 4 6 2r (2,(0)
and C,, C2 = constants of integration. PAO) = (.,r.,--"'*-,> rr,dd
o v ' 2 4 - 6 2r
(b) Regin of ordinary points. Since by (ll.Ok) in (-1,1), /*,(!) = 1 />,(-l) = (-1)' a.(i) = Gr(-i) = (-ir
Mx) = 1- ** = 0 and /,(*) = -2x /(*) = r(r + 1)
are analytic, all points n this regin are ordinary points, and the solution of (11.03a) but also the (b) Graphs PAx) te) Graphs of QAx)
solution of (11.05a) are applicable.
!.()
(c) Indicial equation ofthc second solution is + 1
m(m - 1) =0 nr~T-t,ii -
__L-P,ix>_
whose roots are m, =0 and m., = 1. ' "' /*
(d) Recurrence formulas for thc calculacin of r,., and c2A in (11.05a) arc <+-/ /1+|
"M t\A
!/v /
(r-2A)(r +2*+l) f O- 2k - l)(r + 2* + 2) jrf
V
!'*'- I
(e) Legendre functions oforder rare defined in general with A=x2 as -1
H
- _ (r - 2*)(r + 2 + 1) 1
Px)~aV (2* + D(2* + 2) ' J (d) Algebraic forms
r(r +I) ,/, _(r - 2)(r +3) 4, , _(r - 4)(r +5) ^(1 _...^
5-6 /,(*!
r a.(*)
? r ,r-2*-l)(r + 2 + 2. 1 0 1 L
(2,(x)=xA[l- (2A- + 2)(2A- + 3) 'i 1 X U\(x) - 1
(r - 5)(r + 6) />,(x) - *x
=,(,- (r- l)(r + 2)!,(.-
./. (r~ 3)r +- 4) ,(.- 2 hv*2 -1)
2-3 4-5 6-7
.4(1
')))
)
3 (5x3 - 3*) IA(x) - ?*2 + 5
{(35** - 30xJ + 31 IJ>,(x) - fx3 + x
where r =positive or negative number. For v=0. /> (0) = 1. ,(0) =0. for all r. 4
(f) First derivatives ofthese functions arc 6 i"g(231xb - 315x* + I05x2 - 5) "M*> - ^ + **x' " **
lfe(429x7 - 693xs + 315x3 - 35x)
dPAx) , _, n A l (r - 2k)(r
^Li=-r(r+i)xA[i- (2t+n4- 2k 4- \) 7
] /l + x * x3 . x* . x:
1 :
"T'Al (2* + 2)
220 11.10 LEGENDRE POLYNOMIALS (Continued) Orthogonal Polynomials Orthogonal Polynomials 11.11 TRIGONOMETRIC LEGENDRE 229
POLYNOMIALS (Continued)
(e) Recurrence formulas
(c) Graphs of Pr(costf>) (d) Graphs of Of(cos <p)
(r > iJi
P(CH#I 1<8<
(f) General relations. In -1 < x< 1and with the weichc function w(x) = 1, thc orchogonalicy
condicin is
;e) Algebraic forms in terms Fk = eos 0 with k = 1. 2,3,... are
r" 0
(0 for
lor m ^ n
[/.(*)*<*) * =lAi for m = n r t /'(eos <p Q,(cos 0)
whcre m,/i, rare positive incegcrs and thc orthogonalin coefficient is h, = 2/(2r + 1). 0 1 /.
1 /'i
r*t form
lor m = -.2.4,...
._.+,...
i x'Pr.m(x)dx =\ (2r-> + l)m 2 i(3F3 + 1)
i (:->/-, + 3f,)
l-i \.0o form = 1.3.5,... 3 IA(Px) ~ !#& - Ti
4 ,.'(35y-", + aofj + 9
where ( )!! is che double factorial (2.30). ^,(63/-;, + 35/', - 30F,)
5
A +coc <P^7+r(r+
dd>' dep
llv =0 (-00 < (p < - RAcos(p) = -|cos0/c,(cos0) - K._,(cos d>)l
sin" 0
(r>0)
is y = C,/'r(cos0) + C.,<ZAcos<p)
RAcos<p)d<p = r + l |rt,+I(cosd>) - /t._,(cosd)| (r > 0)
whcre f(CGS (p) = trigonometric Legendre polynomial ;:"the firsi kind of degree r J
<2,(cos <p) = trigonometric Legendre polynomial ::" the second kind of degree r whcre /?,(cos 0) = /^(cos 0) or (2,(cos <p . and Tceos 0) = a*/'(cos <p)/d(p or d(Acos <p)/d<p.
and C,,C, = constants ofintegration.
(g) Asymptotic approximations for large r in A < 0 < jt - A with A jr/10 are
(b) General expressions ofthese polynomials forr = 0..... .. are
Orthogonal Polynomials 11.15 CHEBYSHEV POLYNOMIALS 233
232 11.14 CHEBYSHEV FUNCTIONS Orthogonal Polynomials
(a) Origin and definitions. Thc general integral ofthe dinercncial equacion (a) Chebyshev polynomials. For r = 1,2.3,... , Chebyshcv functions
cions in
in (1I.14)
(1 1.11, reduce co finite
polynomials defined symbolically as
1 U
m(m - 1) =0 x
9 2x" - 1 U(2x)
whosc roocs arc m, =0 and m, 1.
3 lx' - 3x t7(4x- - 1)
4 8x' - 8a- + 1 U(Bx3 - H)
(d) ReCUrrent formulas for che calculacin ofcu and i.. ::-. ' 11.05a) are
5 I6xs - 20*' + 5x <7(I6.v' - 12.V- + 1)
r - (2*)'-' -- - (2k + \y /(32xs - 32a' + 6.x)
6 32x" - 48x4 + 18a- - 1
(2* + l)(2k + 2)'1-4 ""*' : + 2)(2A + 3)
7 64x7 - 112** + 56a' - 7v 17(64*'* - 8tV + 24v- - 1)
8 128xh - 256a" + 160' - 32v' + 1 7(l28*' - I92x* + 80xJ - 8a)
(e) Chebyshev functions oforder r are defined in genera; with A = x2 as
9 256x" - 576x' + 432x5 - 120vl + 9x /(256x8 - 448x'' + 2-10a' - 40x2 + 1)
I i r* - (2kf 10 512x'" - 1280a" + II20x" - 400x* + bOxJ - 1 t7(512x" - 1024*' + 672x* - 160x' + 10x1
=kMl-(2k4.l)(2k +2)A
(e) Recurrence formulas
r i r 41 r - 16
All .11 r A(\ - r..l(x) = 2xf7,(x)- U,_,(x)
1-2 V 3-4 V 5-6 Tr+Ax) = 2xTAx) - Tr_Ax)
r*- (2k + l)a (f) Orthogonality condition defined in 1II.OH6) is applicable wich w(x) = Vi - x'2 and the
C\(x) =x (2A + 2)(2A + 3) orthogonality constants
,1,-zltU-zlMi (r = 1.2.3,...)
2-3 4-5 (i 7 0 23t
234 11.16 LAGUERRE FUNCTIONS AND POLYNOMIALS Orthogonal Polynomials Orthogonal Polynomials 11.17 HERMITE FUNCTIONS AND POLYNOMIALS 235
(a) Origin and definitions. The general integral ofche diflerential (a) Origin and definltions. Thc general integral ofthe diflerential equacion
d'2y , 4y +
,.<+(.
dx'
-x)+rv
dx
=0 -=c < X < x)
dx2
- 2x-f
dx
2rv = 0 (00 < x < oo)
d[L..Ax)\ r+ 1 d(H,4X(x)\
Li(x) = (2r + 1 - x)r(x) - r*Lr-Ax) [I.,+i(x) ~ (r + \)LAx)] rVf+l(x) - 2xHAx) ~ 2r//,_,(x) = 2(r+ \)HAx)
dx
(f) Orthogonality condition defined in (11.08/0 is applicable with w(x) = e ' and, che orthogonality (f) Orthogonality condition delined in (11.086) is applicable wich w(x) = e"' and che orthogonality
conslanc
constant _
h, = (r!
h, = 2'r'V
I
m
,v , ;a-li;;-f(^2'^^('^-|
= (1 +-S^Ur[l +
+ 0(1,
2(y, + ll LEAST-SQUARES
is che ordinarv hvpcrgcomciric funecion (series), u, = function in .v. <p, = factor function in x, and
C,,C2 = conscanes of integration. d,, t, and n. p.. y, cakc on specifie vales and forms
depending on che paramecers a,b,c.
(b) Regular singular points in this differential equation are x = 0, x= 1, and x = ~. Conscquently,
APPROXIMATIONS
cherc arc three kinds of general integris.
(c) General integris of the first kind. If|*| < 1and ,- =0.1,2,..., then
a, = a B, = b
a-, = a - i + I /}, = b - t -
Ifat least one ofthe paramecers a. bis a negative integer, the hypcrgeomccric funecion reduces to
a finite polynomial
(d) General integris of the second kind. If |x - 11 < Iand a + b- c*0.\,2 then
a., = a /}, = b y, = a + b - r + 1 u, = I - x
<P, - i
a, = c b i., = ( - a y, = f - a - b + 1 u. = 1 - *
<p. = (i
r .- Y(c - b - a)
If - a - b > 0 and x = 1. then F(a,b.c. 1) = T ; - a)Y(c - b)
whcre r( ) = gamma funecion (2.28).
(e) General integris of the third kind. If|x| > 1and a - b 0,1,2 then
1
*i = A B, = a - y, = a - b + 1
Pl = *"
i
ti, = b - y., = b - a + 1 u, =
q,, = x <!._ = b A
Ifat lase onc ofche paramecers a.b is a negative integer, the hypcrgcomctric function reduces to
a finite polynomial.
(f) Geometric series. If |x| < 1and a= I, b= c. and r - 0,1,2 the hypergeometric series
reduces to an ordinary geometric series
F(l,c.c,x) = l + x + xa + x3 + -- = j-_
i
^ ^ ^ ^ ^ ^ ^ ^ ^ ^ M r^j I ^1 r~^f ' i
^sltTmean-'square error in the selected interval. The process of finding this polynomial Ifthe functions d>,(x) arc polynomials of degree,/ and if they are mutually orthogonal so that
(substitute function) based on this concept is called the method ofleast squares.
wkd>Axk)d>j(xk) = 0 (i *j\
k-0
y(x) = <J + axx + a2x2 + + ax" = *,* (e) Least-squares polynomial approximation over a given interval. If a given function y(x) is co be
approximated by
fits thcse valesas cise as possible, then the mean-square error
y(x) = a0<p0(x) + a,(px(x) + a.,<?>.(x) + + a.d>,(x) = 2 affifr)
j-0
2 = 2 UAxk) -*]a
k-0 so that thc minimizcd weighted mean-square error in thc interval (a, b) is
is a minimum if
'-' =\ w(x)[y(x)-y(x)]2dx
da0
=o^
' ax
=o^
' da.,
=o,...
'
.^
a
=o
and the functions d>,(x) are polynomials of degrec j and if they arc mutually orthogonal so that
These n conditions yield
w(x)[<t>J(x)]2 dx
= 2* h = S.v-v
This system of linear equations determines uniquely thc desired cocmcicncs, and thc resulting and y(x) is che least-squares polynomial approximation ofy(x). Thcadvantage oy(x) is the same
polynomial f(x) possesses che minimum possible valu of e'2. The applications ofthis process are as in (d). The applications of this process are shown in (12.05) to (12.11).
shown in (12.02), (12.03). .
3
240 12.02 LEAST-SQUARES UNE Least-Squares Approximations Least-Squares Approximations 12.03 LEAST-SQUARES PARBOLA 241
Ti
(a) General relations. Ifa sec of valesy,y .. in (12.01c) is co be approximaccd by a straight (a) General relations. Ifa sec of vales y0,y,, jr3,...,ym in (I2.01c) is co be approximaccd by a
sccond-dcgrcc parbola
linc
thc cocfficiencs 0,, ,s2, t0,/, are determined first as ilie cocfficiencs 0 , , , s-2 , s3 , 4 , / , /, , /_, arc dcccrmincd firsl as
' =7o +.Vi + "' ' + ->'~ 'i = x<0' + x,yx - + xym s., = Xjj + x + + x'i l, = Aav + -v, v, + + .v >,
j = x + x, + + x, /._, x\. + x'i v, + + x%yM
(b) Coefficient matrix equation (12.01c) in cerms ofthese vales becomes S4 = A,', + x4x + + xi
[;: i;][:;H;:] (b) Coefficient matrix equation (12.01r) in ccrms of these vales becomes
from which Si, S, S;
M.L-J * -M S; S, S,
La,J fofj-iL-| J0JL,J
arc che desired coeflicients ofthe least-squares lir. in (a). from which, with D = s0s.,s4 + 2s,s,s3 s,sA s2 s2stl,
(s3sA-s2) -(StS<-s2s3) (sxs3 - s'i)
-(sts< - s,s3) (s0s4 - sfl -(s0s3 - s,s, ':
(c) Example. The following data. i)
L vVs - sl) -(S(s3 - sts.) (ss, - 'f) J
14
2 4
are che desired coeflicients ofthe least-squares parbola in (a)
14
should be fitted by a least-squares line y(x) = a - fl,x. The matrix coeflicients in (a), s (c) Example. Thc following data,
and / . /, . are calculated below.
1.00 1.05 .32 1.51
V,
k *i yi -..'< 0.52 0.73 1.08 1.4-1 1.39
64 10 SO 7.714
1 8
3
60 9.500
/. xk xl x x\ .>' x*Jt xjjt _y4
1 10 100 6
1 12 144 12 .44 11.285
0 1.00 1.00 1.00 1.00 0.52 0.52 0.52 0.909
1 14 196 14 '.96 13.071
6 1 1.05 1.10 1.16 1.22 0.73 0.77 0.80 0.914
\ !)
')
1.15 1.32 1.52 1.75 1.08 1.24 1.43 0.941
yj 7 56 560 54
3 1.32 1.74 2.30 3.01 1.44 1.90 2.51 1.032
4 1.51 2.28 3.14 5.20 1.39 2.10 3.17 1.206
B 1, S-,
5 1.68 2.82 4.74 7.97 1.46 2.45 4.12 1.424
6 1.92 3.68 7.08 13.59 1.58 3.03 5.81 1.835
Thc coefficient matrix in (b) yields
La 9.63 13.91 21.24 33.77 8.20 12.01 18.36
thc polynomial <bk(x) in (12.0 la*) may be taken as where /*,() = Legendre polynomial ofthe first kind of degree r (11.10) and
(b) Five-point least-squares approximation. Ifonly five vales cO are uscd so that M=2. then by ,'c) Square mnimum error of this approximation is
(), / = -2, -1,0, I, 2 and
-2 -1
(f) Powers Of Uin tcrms of PAu) = P. arc
2.6314 4.0457 5.8386
1.386 1.6457
w(x) =
V(*- a)/(b - x)
f' ""TAu) du =fLnLJHD
J-'
L +_ZH!)
lA lm+1)! (m + 2)l (m + 3)1
1J (m + r) even
cransformed as in (12.05a) lo
1
^(a)7;() = [l,TAu) - ,TAu) + /,r IL,
-I < u < 1)
Vi - " whcre m = 1,2,3,..., 7 = d\ T,(u)]/du, T"(u) = d2[TAu)]/du2,.... and
7I(a) = :)'r!Vn?f[(] - -')'-' -'] =2ar,_,(a) - Tr_,(u) and in general, wich p = r,r and q = r,(r I),
(2r)l du'
(e) Derivatives
PA") = -pr
(f) Indefinite integ ais lias che smallcsc lase upper bound for its absolute vale in che
interval 1 u ^ 1. Since che upper bound of che Chebyshcv
f77,(fi) </w =7() +C JTx(u)du - -:r,() + C polynomials is 1,
where 77(a) = shifccd Chebyshev polynomial ofthc first kind of order r, defined in (b), and
77* (a) = 2048a6 - 6144a'' + 6144a4 - 3584a' + 840a2 - 72a + 1 </[77(a)] _ d[T*(u)}
877 () = l2[T*(u) + |]
77(a) = 8192a7 - 28 672a6 + 39 424a'' - 26 880a4 + 9408a3 - 1568a2 + 98a - 1 du du
1 f I /2r
/* = y*(u)du,I* = jy*(u)dudu,I* = ily*(u)dududu,
2*'-'L2\r
77 +
r ^ )77
- 1/
+ 77 + +
O 77., + T,
and che series terminates with the lerm containing D* = 0.
(a) Concept. Cliebyshcv polynomials providc a good approximation ofa funclion with a small error,
but the evaluation of the cocficicnts a, in (12.06) and ofa,* in (12.07) gives risc to rather
complicated integris. If, howcver, a truncated powcr series of the function _)() is rcadily
availablc, the degree of the powcr series can be reduced with littlc loss of aecuracy by expressing
the powcrs of u in terms of T,(u) or T?(u) as shown in (6) below. This series of Cliebyshcv
polynomials can be rcarranged in an ordinary polynomial in u after stopping at the sclcctcd terrn
containing T(u) or T^(u) with a small cocficient. This process devcloped by Lanczos (Rcf. 12.04)
is callcd the tdtscoping (economi^ation) of power series expansin. The rcsulting y(u) is not identieal
with the truc least-squares approximation. but the method is a convenient tool of practical
13
calculations. The following example shows the procedurc of application.
gssf - T +
" ' 2(2!)
(X. + T-,)
4(3!)
(37; + 77,) +
8(4!)
(370 + 47, + Tt)
-4/
APPROXIMATIONS
(107; + 57', 4- T.) + -(I07J 4- 157:, 4- 67, 4- T6)
16(5! 32(b!i
The substitution of the relations (12.06c) for 70, 7,, 7,, 7,, 74, 75 yields the final form of the
economizcd series: f
(b) Basic form. Any singlc-valucd function i d> that is continuous except for a finitc number of
discontinuities in an intcrval (Jt < 0 ^ .t and has a finitc number of mxima and mnima in A / - kJlx - kJCx\ 1 [' knx
this interval may be represented by a convergent Fouricr series as *) = H, + 2 [*i sin 4- B, sin J Bi = l) sm T~
dx
!
i
Change in interval. In the dc%elopnlent of Fourier series the interval can be shifted as shown
y(0) = i4 4- [AK cosk6 + B, sink$ iclow, where L, 7, 0, x, I liave the sa me mcaning as in (<j) and (, C, K are signed numbers. I
i
-2n < 0 < 0 w < 0 < w + lit 0 < t < 2
(C) Principal Characteristic of this series is its :hogonality over any period interval. In particular, in
| (-"-T
(-,t,.t) with j,k = 1,2,3 n, - I _y(0) cos kOdli 1 r(0)cos*t9</0 - i y(0)i.oskOdO
cosj6cos k6d6 = 0 (j * k) I cosjdcos kd d6 = n (J = k) - _>(>) sin kOdt - | r(0) sin k0d6 - i r(t9)sini0/t9
-21. < x < 0 A' < v < A: + 11. 0 < x < 21.
( s\nj6s\nk0d0 =0 (j *k) J smjd sin kOd6 =it (j =k)
1f M ^W - \ y(x) cos dx
zj, j^>-r*
cosj6 sin k6 d6 = 0 (for all j.k 1 " / ^ X-Tv^ 1 1f*** v(x)sin
- i 1 ^wdx 1 f" x ***,,
ZL,/;v(x)s,nT'/v ZJ -'Ws,nX'/* -
(e) Nonperiodic functions that satifv the conc:tions of (b) in a selected interval (jt.-t) can also be
approximated by the truncaied Iourier sene?, but only in this interval. Outside this interval, the ^ = V--1; 4- Bl * = tan-^-=2J Rk = yji + Bl V> = tan--S
series contines to be periodic re .ardless o; :he function to be approximated.
252 13.03 EVALUATION OF FOURIER COEFFICIENTS Fourier Approximations Fourier Approximaons 13.03 EVALUATION OF FOURIER 253
COEFFICIENTS (Continued)
(a) Notatlon. The most common definte integris used in the evaluation of Fourier coefiicients
At, Bkare tabulated below in terms of the following symbols. (d) Range (0 < t/< 1)
k,m,n,r = integers a. b = positive fractions < 1 n = (1)*
| C, A =0
u = x/L = dimensionless variable b = I a w = signed number = 0
(1)
Jo
O)
f**"1^ %U-
A = kjt Ck = cos Aa St = sin Au tu = en Ct = cos <ou S{ sin <ou | S, aa = - ^
(2)
J0bC""' =V (10)
Jo A
(-ir
M, = {-!)* 2 'o(2r- 2/n - 1)!A2" ^Bzr +M). <-t'~-
A2'"' ~o(2r -2m - I)A*- (3)
Jrc** i* (II)
jO*-*^-!
( l)'sin Xa
*,(-!>* 2 'o(2r - 2m)!A"
(-D" AT,=
A2' l ' t'o(2r-2m)A2- (4) (Vd-^.i/, (12) 1f' u*.
Jo
. =
Skdu = (2')!
A
v
Af,
V A - 4" usS, du = 0
(3)
f (u - a)2C rfu = (8)
(3) (H)
(4) (u - af'Ck du = -W Mt (9) ( - af'S, du =-^W,
(4) rrQ du, =-p-
_ 2(2r)! .1/, (12) -'.Sj ato = 0
a'-^'Cjrfu = 0 (13) u St du = ; f,
(5)
f (.-)**>CtA =^-^Ar. (10)
f (u a) St du =
(5)
2jn-sinh-m_ _2Aq_ainh._
W ~Ct du = -5
4
-
a2 + A
Trer "7*~5T2r^~
o-' + A2 fy(u)Sk du =Jof j.()5.''1 ^~fJ ^-(a)^ a
Jo
M
J"
254 13.04 APPROXIMATION OF EVEN FUNCTIONS Fourier Approximations Fourier Approximations 13.04 APPROXIMATION 0F EVEN 255
(a) Notatlon
FUNCTIONS (Continued)
a,b,h,L = positive numbers 17 = (1)* k,m,n,r = positive integers (g) Rectified sine curve (13.03</-7)
u = x/L X = kiX a + b= I C, = cos fon 5^ = sin kmt ^(x) = A sin = A|sinjra| (-l<u<l))
(b) Even function. A functiony(x) is said to be even (symmetrical) if #9
y(-x)=y{x) (-L<x<L) |y40 =A |sin nu\ du = 2h/: 0
Jo /
The Fourier approximation of an even function is Jtodd
1 *
t =2A Jof Ck du =0 y(x) =y(u) = A (-1 < u < -a,a < u < 1) 1(1 1
"
u. a/. "' M
(d)Triangle (!3.03a*-2)
*i0 =h\ du =bh k =2h Ck du = 2A sin Aa
1
A*
^ ) = _ = AU (-!<< 1)
(j) Shifted triangle (13.03-2)
.*(*) =(j) =A2' (-Ku <l! (k) Shifted parbola of 2rth degree (l3.03-4)
^fl
U = A a,ifa = A/(2r 4- I) A*) = ItI("
\b
-flHI i-Ku<-a,a<u<l)
Jo
A f' ., AA 2A f' , 2A(2r)!
k = 2A U-"C4 = 2A(2r)!.A/,A-- A2'A2
^trFm^
Cur(c)-('j>r|dl Omc (c) Cjk- (f) CjwUI<.i-|i>
U<l(l-ClK|j)
Wi
Fourier Approximations 13.05 APPROXIMATION 0F ODD 257
256 13.05 APPROXIMATION OF ODD FUNCTIONS Fouricr Approximations
FUNCTIONS {Continued)
(a) Notation
(g) Quadratic function (13.03</-11,12)
a,b,h,L= positive numbers r = (-1)' k, m,n,r = positive integers
r4Aa(l + ) (-1 <u <0)
u = x/L X- kn a + b= 1 Ct = cos kxu Sk = sin knu y(x) =7(u) |
UAb(I - i/) (0 < u < 1)
16A(1 - n)
(b) Odd function. A function y(x) is said to be odd 4 = 8AJo (I - k)*A = A3
(antisymmetrical) if
y(-x) = -y(x) (-L < x < L)
(h) Cubic function (13.03a'-10.13)
The Fourier approximation of an odd function is
, _, 8Au(I - u-)
y(x)=y(uL)= 4 '- (-!<<!)
y(x) = 2 B*St r-K.
^q-
^fTTl\
i- i
t-S^i-,^*-^ ^QU^
where
are the Fourier coefficients evaluated by the integral -A -1 < u < -a)
tablcs (13.03) and given for selected cases below. y(x) =y(uL) =
Bt =2h\ Sk du = /
-( + Al (-1 < u < -a)
b
y(c) =y(uL) =
^ = 2AJf1 5, du = -
2An
(k) Shifted parbola of (2/- + 1)th degree (13.03e-10)
'u 4- a\:'+*
(-1 < u < -a)
(e) Parbola of (2r + 1)th degree (13.03M3
y(x) = y(uL) = <
(a < u < 1)
.v(x) = A-J - te2'*1 (-1 < "< 1)
,rjp- 2A f1 ., . 2A(2r 4- 1)!
fl, = 2AJf1 u***$kdu
...
= (2r + ')!
\
(I) Superposition
(f) Superposition
t
f
t'
GHr(t).Ciic(d|
rWJ (jK-nl-1'.iwlrl
'(O l4M-(|l
W. /./
Uiir(rl-I .ixiii
w.
Fourier Approximations 13.09 APPROXIMATIONS OF ALTERNATING 261
260 13.08 APPROXIMATIONS OFMONOTONIC Fourier Approximations
TRIANGULAR FUNCTIONS
TRIANGULAR FUNCTIONS
h,L = positive numbers k = 1,2,3,... ,n
a,b = positive fracons h, L= positive numbers * - 1,2,3,....
u=x/L 0 = un j= 1,3,5,. ...
u = x/L <x = an B= bn B= un n = (-1)* j = 1,3,5,... ,,
8A A cosjB
j(x)
a S 1
^r-, 1'J-
,2 J
j-iJ
-1
(2) ,,'
A
... 8A A sinj'0
r i
j s 1
^v ', jM =-i L -75-
oh tA sn*-cos*0 , A A /eos
feos ka _ sin
sinAffN
ka\ .
1 sin kO
// ;a> ,*| : |
01 i .'
1 a < 1
1 "-
0 aA 2A A (l - cos ka) cosAfl
"l* -, , 2A A (-1)*+Isin*0
1 1 r,x> =i+ TPh P *
A") = 2j 7
L
"-O j = i
;.
h 4A A eos/ ' i
/. L
(41 l 1 !
W. aL o/. U. """"""""^ ""
AA 2A A (n - eos Aa) cos fl _, , 2A A (-l)4sinA0
\ 1 1 u) =T +T? 2 *-' y{x) j 7
j = 0
* l * < : !V
A 4A A cosj'fl
1 ' .
2A A sin *0
h 8A A cosj'0
u
(ti) i1
A
.
.. . 2A A sin*0
A 8A A cosjd i .! I'
1
r(', =^2
/. /
Fouricr Approximations Fouricr Approximations 13.11 APPROXIMATIONS OFSINE AND COSINE 263
262 13.10 APPROXIMATIONS OF
264 FUNCTIONS
TRAPEZOIDAL FUNCTIONS
266 13.14 O
a,b = positive fractions h. L= positive numbers k - 1,2,3,... ,n a,b = positive fractions A, /, = positive numbers k = 1,2,3,... ,n
(a) Two Fa
u= x/L a = an B- A.t 6 = un r = (-1)' j = 13, 5,.. . ,s u = x/L 6 = un n = (-1) c noninteger number j* 1, 2, 5,.... s
(11
lll
al. M. a/. o. W. o/.
k *
_r(x) = A |sin 0|
j-AO) '1 1 (2a + b) = 1
1' A
., v 2A 1A A eos 2*0
A 4A A cos ja cosj0 v(.v) = 2j 7^
/ *" \ * K.*f|(24 - 0(2* + I)
with kn
positive . /
* in m in
'-' 1
(b) Sum of
[0): (2a + b) = 1
y(x) = A|cos fl|
4A A
4A A sin/0 2Ai; ^. i (I + n) sin la] sin*0 2A 4A A cos 2*0
m - t*,-.
2 -i- + . .T,
2 + y(x) = -.t - -if,(2*
2 JT - D(2* + I)
where j
c0 -
(c) Produc
I))
al. H al. ti M. al.
-/. < x < -A/2, t/2 < x < /.. v(x) = 0
(2a + A) = 1 -A/2 < v < I./2. y(x) = Acos t
i.(0) '
-p. 4A A sinja sinj0 A A 2All A cos 2*0
i(x) = X ~j
where w
A"
s. * a.-r ,_, .; u f(x) = - + -cos 0
" -t 2 *
'- X
^,(2* - 1)(2* + I)
; i.
Thesc r<
1. /.
1 ' /
JO
(e) Integra
(a + A) = I -/.<t</.. >(*) = Acos r0
(f) Higher
i.. Re.
268 14.01 LAPLACE TRANSFORM, GENERAL RELATIONS Laplace Transforms .aplaceTransforms 14.01 LAPLACE TRANSFORM, GENERAL 269
RELATIONS (Continued)
(a) Definition. IfF(t) is a real piecewisc-continuous function of the real variable / > 0, then the
Laplace Iransform of /*(/) is (g) Exponential function (Ref. 1.20, sec. 19.71)
r r*, , //-""'"v i
Se{F(t)} = f e-F(t)dl =J {/) = e"e-dt = lim /"// = lim I =
J(i
i
where i = real or complex and e = 2.71828 .... (2.18). The variable Ican be x,y, z, or any other and similarly, {e~") =
4- a
variable.
(b) Piecewise COntinuity. A function F(t) is said to be piecewise-coniinuous (seciionally con (h) Trigonometric functions (Ref. 1.20, sec. 9.06)
tinuous) in an interval a < I - i if the narval can be subdivided into a finitc number of
segments, in cach ofwhich the function is continuous and has a finitc limit at both cnds of cach ${s\nwt) = e'"s'mwtdl = lim I e~" sin totdt
segment. The adjacent graph is an examplc ofsuch function.
(c) Exponential order. Afunction F :i is said to be of exponential order S ifwith the real constanl = lim I -5 7. (s sin O)/ 4- w eos fl)/) = -5 r,
'.2 + >- v Jo r 4- w
M > 0 and / -* ,
Fin k
(d) Existence theorem. If F(t) is piecewise-coniinuous V = lim -3 r;(s cos (l - w sin r/) = -5 ;
in every segment of the interva! 0S/SJ2 and of
exponental order A' for I > R. then its Laplace
iransform exists for al] J > A'. These conditions are suflicicnt but nol necessary, since if thcy are
not satisfied the transform may or may not cxisl. (i) Hyperbolic functions (Ref. 1.20, sec. 19.71]
(e) Laplace transforms of elementary functions derived by (a) in (/) to () are tabulated below, ^{sinhai/} = t " smh (ldi = lim dt
Jo **'* Ai
where a, w = signed numbers. i > 0, and n = 1.2.3
r r""-" | r " T = (o
F(t) m) F(l) /(') ~ .I L2(a> - s) 2(( 4- )J0 r - tf2
1
1
(7) sin o
ui
r r,"- + ,-<+*
(1) i
j" + O ifcoshw/} = <- "cosh wt di = lim tfV
Jo f*" Jo -i
j
(2) /" -I
(8) cos Ull
I" + tt>
~ ,-"L2(w - j) 2(w 4- j)J r - o'-
ai
1
(9) sinh tul
(3) e
a j" - a>"
.* L(a - s)~ + w Jo
<{,"} = re'"di = lim I /V"'7
4- a
= lim -<-"/" +
n(n - l)<" h! \T _ ! and similarly, f{,- sin wt) = +^ =^ ^{^"' eos BB) =(j +fl)a +^
I
(b) Linearltyproperty. lf#{*,(/)} =fs),<{F,{t)) =/,(*),..., then X{F(at)) =- f e"w*F(a)rf =*M) MKDl-f'^"* = i/l)
2{aF,(t) bF&) ) = d{FM) bSe{F,(t)) = a/M bUs)
(i) Example (I4.01<-5i
(c) Example (14.01-7,8).
i?"/*siii^-,l =
b i A((j/A - (i)' 4- or]
aw 4- bs
gasino,/ +coso,/} =a-f2 4- 1,-j-, - +-t
) Multiplication by tn. UX{F(t)) =f(s), then
(d) First shifting property. If y{f 0} = /(*), <hen dm
ds
=l\ff-:F(t)dt]=
ds l Jo J
f{-te-"F)\d<
Jo
and in general.
(e) Examples (14.01-2,9,10)
,</('
SE{CF(t)) = 1-1 f"
272 14.03 TRANSFORMS OF DERIVATIVES Laplace Transforms Laplace Transforms 14.03 TRANSFORMS OF DERIVATIVES AND 273
AND INTEGRALS INTEGRALS (Continued)
(a) Notatlon (a) Single integral. If
F(t) = original function f(s) = Laplace transform G(t)
SB{F{i)) =/<*) G(t) =('F(t)dt
Jo
G(0) =0 dt
= F(/)
FM{t) = nth-order derivative of F(t) (n = 1,2,3,...)
F(+0), F("(+0),Fl2)(+0),... ,Fim)(+0) = initial vales then by (14.02) ^{^p} =s{G(t)} - C(+0) =f(s)
u,, 2, u3,..., u, = functions of /
(b) First-order derivativa. If S{F{t)) =/(*) and dF(t)/di = F'(t), then by parts,
from which *{[w*}-*i
S{F'{t)) = Jofr"F'(t) dt =lim [e-"F'(t)dt
<*Jo
(h) Example (14.02m)
'sin wt , ] 2{(sin a>t)/1] 1 , s
=lim \e-'F(t) \'o - f(se-")F(t) dt]
i L Jo J
-> = - cot
s
w
and 2{F'(t)} =sf(s) - Fi+0) | (i) Mltiple integris. For the same function,
i i
(c) Example. ir"(/) = d(sinwt)/(wdt) = eos tuf and <e{(sinwt)/w} = 1/('j 4- ar), then a{//**}-*!. *{///*,)***}-j? ^{///fww}-^
,fsin wt) sin tt(4-0) s
?{cos wt} =^{^-\ + o> r 4- o>" (j) Example. IfF/i = cosh wt eos <o/ and
(f) Oth-order derivative. The transform ofan nth-order derivative is by deduction of (b) and (d),
where the lowcr dcrivatives must be continuous and the highest derivative F""(f) must be
seciionally .jontinuous in every finitc interval of its application.
Jj-
0
J r,) (<//> ~ ( - l)lj('
0 0
T) '^J''1
^ ._^ / . i ^^^^^^^^^^^^^^^^ ^^^^^^^^^
(b) Unit Step function, also called Heaviside's unit function, shown in the adjaccnt graph, is (b) Ramp function Of first degree, shown in the adjacent figure, is
f0 (t
(<<< a) (t<a)
Qlti\ = l v
a) /?,(/ - a) = \ t - a Se{Ry(t -a)}=
(t>a) os
and
(C) Unit Impulse function, also called Dirac Delta function, shown in the adjaccnt graph, is [0
R2(t -a) =\ (t - af
(Ka)
2e- 2
*<*.(' - =757 UUJ
\U.(t - al V(t ~ a ~ )1 1
^-a> =iLm0l J
(d) Ramp function Of /Ith degree, shown in the adjaccnt figure, is
Other designations of this function are 6(t - a) or 9{t - a), and the |(
function is not a function in the truc mathematical sense but a dcfinilion
of distribution. Some of its properties are
(Ka)
2.
Rn(t -a) = i(t- a)' 2{RAt - )) LUJ
v+l
r?(t) dt=\ ^&{t-a)dt=\ b"
= (14-
.,. e-"H(u)du
Jo ' '
= (l - e-T+ e~
'<'' _ )[ e-"H(t) du = "
J() 1 T
_l7- - *(*)
he '" du and A = 2a is the period, then by the geometric superposition (14.05c). (14.05/),
- e~")h
g(s) = - e " (1 - e )s
(I - 2e~" 4- e~h)
h(s) =
(c) Triangular function. II 1 4- <" a(l 4- e~*)ss
[O (a < u < ?i l-l)4 (0 < u < a)
H(t) = <
>1 , . /\
and 6is the period, then b> the geometric superposition (14.05*), ^J
(a < u < b)
(14.05/).
/ I e~" t'"\,
and Ais the period, then by the rclation betwccn g(s) and h(s)
defined in (ai in terms of (14.06c),
W s *:l [\ ~ (as + \)e--\h
Sb) - r^ fl(i - ,-*V 1 - (as 4- l)e-"]h
h(t) =
<i(l 4- _*')2
() Rectified sine half-wave. I:
n (d) Alternating rectified sine half-wave. If with A = 0. 1,2, //i/i
//(o=|'[O"1)4Asin(f") (0<u<a)
(a < a < A)
and b = 2a is the period. then
and A = 2a is the period, then
Ar '" sin \- -\ du an(\ - i-*')/;
anh ajr(l 4- e~")h
A(,) = 7-r
(j) = I - e' (n1 4- a2f2)(l - e" (.7J 4- a'V)(l - e~")(\ 4- f~*') (n- + V)(l 4- "*)
-MW-^ ^^^^^^^^^ ^^^^^^r^^^^^
278 14.08 INVERSE LAPLACE TRANSFORM, GENERAL Laplace Transforms Laplace Transforms 14.09 INVERSE BY HEAVISIDE EXPANSIN 279
RELATIONS AND PROPERTIES (a) Ratlonal fraction. If a rational Laplace transform with m n l,
(8) Definltlon. Ifthe Laplace transform ofa functon F(t) isf(s) defined in (14.01a), then F(t) is the b0+ b,s + b2s7 + + bmsm
inverse of/(). Symbolically,
a0 4- ats + aaj2 + 4- bs"
Srl{f(s)} = F(t)
then by the fundamental theorem of algebra, the denominator v:i
Although in the strict sense of the uniqueness offunctions the Laplace transform is not unique, in Q(s) = (s - qt)(s - a2) (j - ?_,)(* - a.)
engineerng applications this transform isessentially unique.
where a,, a2,..., qare the roots, which according to (6.01a*) can be real and distinct, or real and
(b) Procedure of finding this inverse is usually simplified by using tables ofLaplace transform pairs. repeated, or complex conjgate and distinct, or complex conjgate and repeated. The roots of
Extensive tables are avaable in the literaturc (Refs. 1.01, 1.20, 14.01-14.06). In addion, certain Q(s) are calculated by one of the methods introduced in Chap. 6.
properties of inverse Laplace transforms useful in finding the inverse are defined below.
(C) Inverse properties based on (14.02) and (14.03) are tabulated below. (b) Heaviside's formula, distinct roots. Ifat (A = 1,2,...,n) are distinct, the inverse off(s) in (a)
is
(I) *
"13- m*-{$}- ai ^{}-^ ,P()] = f P(qt)
P(to)
\<(s)i *%<l'(qt >'
(4) <r'(f(s + *)) = *F{t) (5) JT'{/(i - a)) = d-F{l) where P(qt) = b0 4- bxqk + b2q2k 4- 4- bmqk
Q'(li)=-[Q,(s)],mil
(6) Jr-{/w>-;'(;) (7) 2"'{^)} =aFW
= o, 4 2a2at 4- 3o3aJ 4- 4- naMql~
(8) ^-x{e"f(s)) -F-t + a) (9) sr'ir-As)) = F(t-a)
(c) Example. If
and
(b) Double poles. If (b) Double factors. Ifwith a-,/S defined in (a),
o
On +
T *i*
"s _ -\
b- 4- p,j 4- A2J2 4- SJ3 _(a + Bi _(a-Bi
JW-r^ '--* A*) [(, - a)2 B2)2 h-2 "1*4-/3 34 ~\a-B
then by (I4.08c-5,10). then by (a .
j-.f ' 1 ^fsinfr
\.s-a)2p") &lsinh/3/
and by convolution (14.08c-15)
and ^-.J*i*i} =[(A0 4- ,,,)' 4- A,K S-AIs - a)2\ 4- B2; (s - a)2\ + B2)
^1 =iB2)0|V('~r)sn/3(/ - T^sinBTdr
=p Jof '[cosB(t
2
- 2v) - cosBx]dx =^(sin/3/
P
- BtcosBt) =</>,(/)
(c) Triple poles. If
Similarlv.
A04- ,f 4- b:
(l = ?- = J = )
V-' L
* {,f _ af - B2 (s - a)2 - B2)
X- -1 =-^-r
2BiKH
(Bt cosh Bt
H
- sinh 0/) =V
= Vi(0
/(')
F(i) /( /'(/)
(2/)'
ORDINARY
V7, (2r - 1)!!V.T/
I i
Vi + a
1 crf Va/
V7/'
a, crfVa/
' "VT
DIFFERENTIAL
jVj + o (- a)Vi
(2/)V
(21)'''"
(2r - 1)!! V.T/ V(7^ a)2" (2r - 1)!! Vjt/
/,,(')
EQUATIONS
vV O
Joto*t)dl
,V7 {'< nl)dl
{/,(/) //,(a/)
V .-"-r-"..-')' V7 - "-f
//(<")
vV + V VV - - r
I{at) + alisal)
7o(>0 " a'J"t) V(J- - a-f
V(r + -r'P
cxp(-a/i)
exp (a/i) (-Xj^y/a) (k > i)
i ,
k
TF
Ordinary Differential Equations Ordinary Differential Equations 15.02 FOURIER SOLUTION, GENERAL 285
284 15.01 GENERAL CONCEPTS
CASE (Continued)
(a) Definitions. A differential cquation is an algbrale or transcendental cquality involving
differentials or derivatives. The order of a difTercntial cquation is the order of the highest (b) Assumed solution ofthe differential equation (a) is then
derivative, and its degree is the degrec of the highest derivative. The number of independent
variables defines the differential cquation as ordinary differential equation (single variable) or partid I V> - knX V" r k7CX ^ A ,,
}' = -,an + i al eos- + 2/ bt sin 4- G,h, v)
differential equation (two or sevcral variables). Adifferential cquation which is ofthe first degrec in 1-1 /- i -1 '- i -1
the dependent variable and in all its dcrivatives is called the linear differential equation. In general,
where the last sum is (lie complementary solution given for large number of cases in Ref. 1.20, pp.
179-183.
a,(x)d-=g.
.t, dx
ir = 0,l,2,...,a)
(c) Constants ,l,bk are detcrmincd by inserting this solution in the diflerential equation and
where Y= dependent variable, x = independent variable, n = order, ar(x) = variable coefficient equating the coefficients ofthe respective trigonometric functions. The remaining constants C, are
(given function in x . and g(x) = input or forcing function (given function in x). obtained from r conditions (initial vales or boundary vales).
(b) Special forms. The general form (a) is then designated as the nonhomogcncous ordinary linear (d) Example. The motion ofa given mass m (kg) suspended on an elastic spring of stilfness k (N/m)
differential equation with variable coefficients. If^x) = 0, the cquation (a) is called homoge- and acted on by a periodic forc // of square-wave variation (N),
neous, and ifa,(.v) = a, (constant), the cquation in (a) is designated with constant coefficients.
-h -'.T< / < 0
gil) =
(C) General SOlutlon ofihe ordinary diflerential equation (a) is h 0 < t<T
where T period (s) and / = time (s). With the forcing frequeney 2 = 2n/T, this equation
(d) Closed-form Solutions are avaable only in a few simple cases but lead frequcntly to becomes. in terms ofthe Fourier expansin introduced in (13.07-1),
approximate solutions of more complex problems. Selected cases are tabulated in Ref. 1.20, pp.
176-183, and are not repeated here. d:Y .,- 4/i A sinkQl
+ <o-Y = ET (k = 1.3.5 p)
dt nm i i k
(e) Serles solutions appear in the form of powcr series introduced in Chap. 11 and of Fourier series
defined in Chap. 13 and applied in (15.02) below. and its assumed solution is taken in the form of(A) as
(f) Laplace transform solutions are tabulated for selected cases in Ref. 1.20, pp. 248-280. The Y- 2) bk sin kQt + C cos wl 4- C. sin wt
derivations ofthese solutions and the evaluation ofthe respective convolution integris are shown
in the subsequent sections of this chapter. After inserting Yin the diflerential equation, the constants bk are detcrmincd by (c) as
- _ !//
(g) Numerical solutions shown in the last sections of this chapter are based cither on the Taylor's kn(w~ k'2Q')m
series expansin or on the collocation polynomials, whosc indeterminate coefficients rcsult in the
lcast error ofthe assumed approximation. and the approximate solution becomes
4// A sin kQt
Y - C, cos wl + C,sin wt + ^ X 7, ,.,_.,,
nmk.\ k(w* k*Qr)
15.02 FOURIER SOLUTION, GENERAL CASE
If the initial conditions are given as F(0) = Y0,dY(Q)/dt = 0,
(a) Linear differential equation with constant coefcients and an arbitran- input function has a
Fouricr solution if ; k) can be approximated in the selected intcrval (-L.L) by a truncated Y i = C, 0 = wC, +
4//Q A
2 71
'
\d-Y(x)\ y("-l<
= Yo' Y'0' =
'o
Y
'o Y'-' C"-"(x) L\"-"(x) '-l\x) r"W r'H
I dx' I,.o
H;
(b) Laplace transform ofthe differential equation in (a) is then, according to (14.03/), H, T*i.{*)
/(s)M(s) - S(s) = h(s) relates the state vector ofthe right end HK in the interval 10, x] to the state vector ofthe lefi end
from which H, of the same interval by
\(s) + h\s)
/(') = T*Ax) = transport matrix [(n + I) X (n + \)\
Mis)
Y = Y9Lt(x\ + Yo"L.2(x) + Y0*%(x) + Y03%(x) + + FjT"%(*) + <?M which shows thatonly the first nfunctions are required for theconstruction ofthe L(.v) submatrix.
here
"
(g) Load functions in (c) are rclated by the recurrent relations
,(*) a, a_| a,__i fl3 a3 "l .sr il
dx dx
a, a_ i t
(*) 4 3 J
which shows that only the first load function (convolution integral) is required for the
se construction ofthe submatrix G(.v).
M*) 5 4 a.l
M(
1 (h) Transport Chain. Once the transpon matrix is avaable for
(I se-
A/() single segment, the extensin to a multisegmcnt domain is
J L
accomplished by matrix multiplication. Considcring the domain
are the shape functions and given in the adjacent figure by segments 01, 12, 23 of lengths I-T)
/,./..:':. rcspectively, the state vectors at 1, 2, 3 are
rt , _ a-i
H = T,H
(d) Convolution integral in (c), Ha = t,(1f0
H. = Tj.T.oHo
i
G(.t) = ,,.v - r)5(r)a-r = f L(x)g{x - r) </r H, = T:,,T,,T,()H0 H, = TtllH
can be evaluated by Goursat's formula (4.17A). Particular cases ofthis integral are tabulated in where T, = T,T10 and TU) = TjjTgjT,,,; Thesc relations show that the product of two. three,
and anv number of scgmental matrices is ; lways a neiv transport matrix.
(15.04).
288 15.04 LAPLACE TRANSFORM SOLUTION, Ordinary Differential Equations Ordinary Differential Equations 15.04 LAPLACE TRANSFORM SOLUTION, 289
CONVOLUTION INTEGRALS CONVOLUTION INTEGRALS (Continued)
(a) Convolution integris (15.03a1) are evaluated below for selected cases of g(x) in terms ofthe (b) Shifting along the xaxis. If^C*) is shifted along x, then G(x) may be expressed in terms ofthe
followinc results in (a) by rcplacing xbyx-aorx-A and L,(x) by A = L,(x - a) or /,(* - A) = B,.
following equivalents.
cqi The infinite series in (3) to (5) may have a closed-form sum msomc cases. respcctively.
w = signed number i. _ I 2 . . n a,b,c,a,p = positive constants
Input function ,?(x) Convolution integral C(x)
+i(*) = Ux)dx B = n/b 4> =n/2 Ln+k(x) = f /-.*->(x)dx
i
0 (0 =S x < a)
Convolution integral G(.\
Input function g(x)
H
T 11
'
,
.., (a x)
/-.. ,<x)
r(>)
<.-l'
0 (0 =S x < n)
Ib - a)1
1 rf
f- ^-
w.W b
(A-a)'-4"*"1 (k = 1,2,... a s .v)
(c) Geometric superposition of convolution integris tabulated in (a), (A) may be used
J. _
a-[ (-/})'"'/^..(r) sin [/i(v - r) - U- !)?> evaluation of convolution integris of piecewise-coniinuous functions as shown below.
(0 = x < a)
rW r
ll
L . {i (->'-'/..,4(r)coS[(* - r) - (k - 1)01 rrr
^ZU7 vi..
hr1 1
-A. (0 x < a)
(0 s x < A)
1- t -(/--.,. ---I. . -a/i..,) (A x)
PI.Ax - b)
(*x)
(0 r < a)
f-:1
(0 x < b)
CSC "-U..I ---2 (a S x < )
where the shape functions (I5.03e) are and its solution by (15.05c), with the initial condition A = 0, is
V'f } ='_A'
(s + AJ
(A'>0> A= A0-a' + - e~M'-T)Fdx = -(I - i"*')
nJo k
cw-^-ilfel-sf^-^* (a) Standard linear model consisting of two elastic springs of stiflhesscs kx , k., and one dashpot of
viscosity t]x is shown in the adjaccnt figure. If the applied forc is given by the diagram shown
below and the same constitutive equations are valid as in (15.06a), then
Particular cases of these integris are tabulated in Ref. 1.20, sec. 16.36.
F = Fx + F2
'H
(d) Example. The Laplace transform solution of
A = As,i + ADJ = AS2
10F' + 20}' = 13.sin3* Y0 = 4
A = s, + DX + S2 __ i
is by (e) for A" = 20 > 0 and A= ts = 2. ,'>
(b) Governing differential equation of this model in terms
v = 4^"-' - r-2<'-t,sin3Trfr of {H 11
10 Jo
Fx = F - F2 Ft = F - F2 U-^l
where the convolution integral by Ref. 1.20, sec. 9.06, is
F2 = k2A F2 = *,
ni \ - '3 f ,--* sin 3t dx =-^-^ [ / ., (2 sin 3r - 3cos 3r) I becomes
_ :<;.sin/3x - g(cos/3x - ,(s)] The solution by (15.05c) with A0 = ocn and with / > a is then
C(a) ' ~" M(X2 + B2)
A = oy-*' + -(1 - a"*1) - -11 -
*)r *i V i
I0(2a + 3a)
'L
^^^^^^^^^^^^^^^^^^g^^^^^^^^^^^^^^^^^y
292 15.08 SECOND-ORDER DIFFERENTIAL EQUATION Ordinary Differential Equations Ordinary Differential Equations 15.09 F0RCED VIBRATION WITHOUT DAMPING 293
OFTHE FIRST KIND, LTSGENERAL CASE (a) Hookes' model of spring constant k (N/m) and mass tv)
m (kg) is acted on by a forcing function defined in the
(8) Notation. Y= solution M,N =signcd numbers * 0 p=AT/A A= y/\N/M\ adjacent graph. The initial vales of this motion are
l
and for A' < 0. f(s) = Y0-r_A*
, _ 2+" Y'0V - A2 bis)
M(s2 - A2) \-L
am
(Osisj) L4 = (wt sin wt)/w3
where/(i), hs have the same meaning as in (15.05).
C(l) =< (L4-2A4) (a = t < 2a) A4 = [w(t - a) - sin w(t - a)\/wi
am
(C) Inverse Laplace transform of (A) according to (15.03c) is the Laplace transform solution (LTS)
of the diflerential equation given symbolically as (L4-2A4 + B4) (2a < /) B4 = [w(t - 2a) - sin w(t - 2a)]/wf
.ara
N>0 A'<0
15.10 BEAM DEFLECTI0N
(a) Cantilever beam of length / (m), moment of inertia / (m4), elastic modulus E (Pa) is acted on bv
a triangular load of mximum intensity/> (N/m), as shown in the adjacent figure.
-. !_} _ !if
*W-W' A (b) Deflection differential equation of this beam is
where the convolution integral by Ref. 1.20, sec. 9.04-2, is For x = /, the deflection and slope of the elastic curve at B ar;
0
dx) =-^ I r2sin2(x - r)a*r Ytf=Y0+ Y0l- \20EI V) 24EI
r ,cos2(x - r) sin2(x - t) _ cos2(x - t)V from which the end slope and deflection at A are
= T7,\T 9 +T A : 8 Jo
= t- cos2x + 2x2 - I) Io=- JL
0 24/ 30/
or directly bv Ref. 1.20, sec. 16.38, case 3.
294 15.11 SECOND-ORDER DIFFERENTIAL EQUATION Ordinary Differential Equations Ordinary Differential Equations 15.12 SECOND-ORDER DIFFERENTIAL EQUATION 295
0F THE SECOND KIND, LTS SHAPE FUNCTIONS
OF THE SECOND KIND, LTS GENERAL CASE
(a) Notation. 5' = solution M, N, R= signed numbers #0 p= N/m q= R/M (a) Table of shape functions (15.11a")
> = hP
A= p' - 4a eo = r.VJAJ I) Case I: A < 0, V = tan-1 -,E = e *% r = VA2 + w1
(b) Laplace transform ofthe sccond-ordcr-dilferential cquation
Er E .
LA*) sin (wx + xji) LAx) sin wx
2ft) (O
= g(*) l i ' = initial vales
ax- rfx
L3(x) [i?sin (wx + V) - sin V']
is by (I5.0U.. /(*) =*,!+ +, +}V+ps +q+,U(r 4- ps +a) (j) Wr
i + 2A (A = 0)
+
As) = { 1". ,- +TT^+
A)2 Jo"(s + A)' M(j + A)2 (2) Case II: A = 0, ip = 0,E = _A*
+ 2A ... 1 AM
- + Yo ; + (A >0) (1 4- Ax) /,.(*) x
/'(, + A)' - o>2 (s + A)2 - ar M[(s + A)2 - a2] /-,(*)
(c) Inverse Laplace transform of (b) by (15.030 is the Laplace transform solution (LTS) ofthe /-:,(*) -1[(1 +Xx)E- 1)
diflerential ecu.uion given symbolically as
A = 0
Acomplete sel of shape functions in terms ofthe respective phase angles is tabulated in (15.12). f^(*)A =LJ+Ax) forj =2,3,...
^^^^^^=^^^^^^^^^^^^^ '-^1 Pp
^^^^^^^^^^^^^^^^^^^^^^f^^^^^^^^^fl^^
296 15.13 SECOND-ORDER DIFFERENTIAL EQUATION Ordinary Differential Equations Ordinary Differential Equations 15.14 F0URTH-0RDER DIFFERENTIAL EQUATION 297
OFTHE SECOND KIND, CONVOLUTION INTEGRALS OFTHE FIRST KIND, LTS GENERAL CASE
(a) Notation. m= 0,1,2,... b, w,a,B = signed numbers (a) Notation. Y = solution M,N = signed numbers /> = AT/A/
p,q: sec (15.11a) A,B,C,D = constants
A=Vf '4A/ '' VA/
iilliii
(b) Algbraleinput function. If(x) = w{x/b)m, then in (15.1le),
.f I m\w 1 w f
(b) Laplace transform of the fourth-order differential equation
cw x"Wt.!?) =5rI" trWt)'r d*Y
andby (4.17),
M4 + NY= g(x) Y0, Y'0, Yo, y" = initial vales
C(x) = [ix - x)-L,ix) + aix - r)4(T) + m(m - I)(x - T)-2s(r) is, by (15.03*),
s3 s2 1 s *(*)
4- 4- m!+3(r)J0 = -jjLm+i(x) J/(*) = y0-4
s* 4- p + Y~4
s* +p + YM0-.
V + p + Y?-
s* + p . M(s4 + p)
where fis) and h(s) have the same meaning as in (15.05o).
where Lw.;ix) is the respective shape function defined in (15.12a), (15.I2).
"Y
V +j34A4 i yV 4-''4A4 i S4
r +J4A4 ii""1
% + 4A4 i A/(/h{s)
+ 4A4) (/>>0)
(c) Harmonic input function. If(x) = a> sin/3x, then the convolution integral in (15.1 le) becomes As) =
y +r 4- r '
too involved and the solution by pardal fractions is much simpler.
'o/ - 4 + S4 - n* + rs* - n4 +i Yor / -' t,4 +A/(f4
i A(j)
- n4) (p<0)
,, . VI
.1/
/ ) Zr,l_Lt.+iJl}
U2 + ps 4- qs2 + B2) M is2 4- ps 4- a s2 4- B2) (C) Inverse Laplace transform of (b) by (15.03c) is the Laplace transform solution (LTS) of the
where with u = q B2, v = pB, differential equation given symbolically as
In terms o these constants, where ,(x), 2(x), j(x), 4(x) are the shape functions (15.03c) Usted in (15.15), (15.15c). and
in terms of the respective 4(x),
_ w Avs + vp - up _ vs_-JfP\ '! \\m
(X} ~ M(u2 + v2) \s2+ps + q s2 + B2i
uu>[sinBx BL2(x)] vw[cos Bx Lx(x)]
^"^'{TTTJ^r^-'W"^
= M(u2 + v2) is the convolution integral (15.03rf). Particular cases of these integris are tabulated in Ref. 1.20. sec.
16.44. i
(d) Exponential input function. If g(x) = we~"', then the solution by partial fractions is even
simpler. (d) Transport matrixequation introduced in (15.030 is, for all p,
w r A 1 II u',(As + B C ) 1 r i o o o o 1
f/-\ = se~ \ i = se~ i 4- t
.1/ l2 4- ps 4- qs + al M \s2 + ps 4- a s + ai y(x) dx) ,(x) LAx) LAx) LAx)
where with u = ar2 pa 4- q, Y'ix) G'(x) -pL4ix) ,(x) LAx) 3(x) Y'o
Y"ix) G"(x) -pL>(x) -pL4ix) ,(x) ,(x) n
a-p
A = - C = - Lcm(x)
Ln*)J -pLtix) -pL3ix) -pL4ix) ,(x). L>rJ
296 15.15 FOURTH-ORDER DIFFERENTIAL EQUATION Ordinary Differential Equations Ordinary Differential Equations 15.16 FOURTH-ORDER DIFFERENTIAL EQUATION 299
OF THE FIRST KIND, LTS SHAPE FUNCTIONS OFTHE FIRST KIND, CONVOLUTION INTEGRALS
.(Ax)4'
l.,tx &
U4 +4A4j C,C, S (")' (4r)! C(x) = - x)mLAx) + m(x -- t)s(t) + m{m - l)(x - x)-%ix)
m\w
(C,5, +,C,)
L,lx
b4 4- 4A4J 2A A ,t1 '(4r4-l)! + 4-m!.,+s(T)]0 = m+5(x)
I.,n sr
'l4 4- 4A4J 2A2'0'
4,A2 (-4)'-^^
~0* ' (4r4-2)! where M+S(x) is the respective shape function in (15.15o), (15.15c).
rt - r--^3 + ^ + -J\ 1i
G{x)-Mip + p*r i s* +p +s2 + p2
(c) Case II 15.14o), (15.14c) p < 0
41 r = 0,l,2,, it = 5,6, = M(p
, "4- p 4,[sinfr
)
- ^(*) +/33,(x)]
j(C2 + o Y (g*)4'
,<* r
'l4-4J4l o (40!
(d) Exponential input function. If g(x) = we~', then the solution by partial fractions is even i'II
I y (>?x)4^'
/,.(
*-fc^?} 5 <*.+*> / ,-u(4r+l)!
simpler.
!i.
G(x) _ ii <?-{_! !_) = 2-i(A + & + C, + D __)
/..,..
V - 4IJ4J 2t/:
(Q-C,)
i?2 ,ro(4r4-2)! G{X)~M* \s* +ps + al M* l /+/) , +*/
where with u = p + a*,
1 y (nx)1'*3
/.,, ^
l^v) 4f r? ,t,(4r4-3)! A = -l/ B = a/u C= -2/u D = af/u :; = 1/u
,! !
300 15.17 ELASTIC FOUNDATION Ordinary Differential Equations Ordinary Differential Equations 15.18 PONDING PROBLEM 301
following. First, the beam defleets under the applied load and
1 Ly^-L Y deflection (in) the slab forms a dry pond which during the first rain s fillcd
dx* " El El with water forming the ponding load of intensity gYw, where
Yw is the initial deflection due to applied loads. The first pond
(C) General solution of this equation by (15.14c) is load produces the first increment in Yu. called YfX, which in
turn provides space for additional water to form a second
} = YtL,(x) + Y'LLAx) + ylL,(x) 4- Y^LAx) + G(x) ponding load. Thus the initial deflection and all the incremen Uru imrroirM of 1'
ta! deflections due to ponding form the final elastic curve i "" -M ! =
where LAx), j(x), LAx), 4(x) are the shape functions given in (15.15A) in terms of which is defined by the following equation.
A = 0.0236/in, and by (15.166) with m = 0, o = 1, M = El, Setimd porKttK kud
zr
(c) Governing differential equation of this systcm in terms of >
G y = deflection (in) is
-EILM Smmd iiirrrniriu ! I
/T
1- J-Y = -L
(d) End vales ofthe beam are dx* El El
Second poutlin}(U*.u\
}', = left-cnd deflection Y'L = left-end slope = 6L iM I i= =-7^
7
Gil) LAD (/> LAD LAD (e) End vales of this beam are designated by the symbols introduced in (15.17a1) on the opposite
LAD LAD 0l page, and their zeros are, in this case,
G'ii) -PLAD LAD
and \:tii these results, the fourth and fifth row yicld and with these results, the third and fifth row yicld,
The uuiidation reaction is then (0.2)(2OO) - 9.3 = 3TT7X Sincc the appleclload is-40 k7 th~e pbhdi^loaclis V~+ VK 40 = 2k.
302 15.19 FOURTH-ORDER DIFFERENTIAL EQUATION Ordinary Differential Equations Ordinary Differential Equations 15.20 FOURTH-ORDER DIFFERENTIAL EQUATION 303
OF THE SECOND KIND, LTS GENERAL CASE 0F THE SECOND KIND, LTS SHAPE FUNCTIONS
M,N = signed numbers # 0 p = N/M (a) Equivalents. 5 = sin Ax C = cos Ax 5 = sinh nx C = cosh nx
(a) Notation. y = solution
F
A
'"-rS '""Vi*
* = ve (b) Case I (15.196), (15.19c) 0 A=^ r=0,I,2,... * = 5,6,7,.
is. by (15.036),
r (A)"4
A(i) mx)
7h} i' A,-o ' (2r + I)
V + ) 4-' *
C"
s2is2 4- ) AftV + />)
where/(i) and A() are the Laplace transforms ofyand gix), respcctively. LAx) x- pd-C)
Us* + AJ)J A2,^* ' (2r +21!
" 1 ! 1 1 6(j) ip>0)
y0j +y0-2 + ns{s* +A,} + V +A2) +A/f*( +a2)
1 1 1 1 hjs)
L,(x) <
' UV +A1)) ('"I) A3,^1 ' (2r +3>!
Yo- +r0- +nj(fi _ ^ +itjV _ ^ +/jV _ ^ ip<0)
LA*) LAx) dx
Jo p[?-H A4 .t/ ' (2r + 41!
(C) Inverse Laplace transform of (6) by (15.03c) is the Laplace transform solution (LTS) ofthe , (AvW"
differential equation given symbolically as Lk(x) I' Lk.x(x)dx
Jo
[(TTiiJ-^-H _i_y -n--^
A1"1,ti1 ' <2r + <- 1)!
) = y0 4- Y0x + Yimx) 4- C4yv) 4- Gjx) I
where LAx), 4(x) are the shape functions (15.03c) listed in (15.206), (15.20c), and in terms of
L4ix . (C) Case II (15.196), (15.19r) P <0 n
va r = 0, 1,2, A- = 5.6.7,
is the convolution integral (15.03rf). Particular cases ofthese integris are tabulated in Ref. 1.20, sec.
16.46. Three typical derivations are shown in (15.21). L,(x) se-
r^} ~0l(-1)'^'
' <2r)! ni'
I
M
304 15.21 FOURTH-ORDER DIFFERENTIAL EQUATION Ordinary Differential Equations Ordinary Differential Equations 15.22 BEAM-COLUMN PR0BLEM sos
OFTHE SECOND KIND, CONVOLUTION INTEGRALS
(a) System. A straight prismatic bar of moment of
(a) Notation. m = 0,1,2,... 6, w,a,B signed numbers inertia / = 200 in4, elastic modulus E = 3(10)4ksi
is simply supported at L, fixed at R, and acted on
ir
p:scc (15.19) A, B, C, D,E,F = constants
by an axial forc P = 150k and a transversc
uniform load of intensity w = 0.2 k/in as shown in
(b) Algbrale InpUt function. Ugix) = w(x/n)m, then in (15.19c) by (15.03a*), the adjacent figure, where / = 480 in.
where m+5(x) is the respective shape function defined in (15.206), (15.20c) (C) GeneralSOlution of this equation by (15.19c) is
(c) Harmonic input function. If gix) = wsin/Sx, then again the convolution integral in (15.19c) y = YL + Y'Lx 4- Yimx) + YmLL4ix) 4- G(x)
becomes too involvcd and the solution by partial fractions is much simpler.
where LAx), LAx) are the shape functions given in (15.206) in terms of A = 0.005/in, and bv
Cix)=jMs2is2\p)s2p2) (15.216) with m = 0, 6 = I, M = El,
v ^[As'
As* + Bs +
4- Bs2 Cs + D Es + F)
Gix) = -JjLAx)
- Mu* 1 sHsis2 4- p)
where with u = (/?2 p)P~,
(d) End vales of the beam-column are
A = C = = 0 B = - F = YL left-end deflection = 0 Y'. = left-end slope = 8,
ip2 - p)p D=P (P2 ~ P)P
Y"L = left-end bcnding moment// = MJEl = 0
- Y" = left-end shcar/7 4- axial forc times slope/El - (V, 4- P6L)/EI
G{x)~Mu* \s2 + p2 s2s2 +s2(r+p)
s2(r+p).
YR = right-end deflection = 0 Y'K = right-end slope = 6R = 0
-,[sinpx- px + p3L4(x)) Y"R = right-end bending moment// = MJEl
M(p2 - p)p
YR = right-end shear/7 4- axial forc times slope/El = (1 4- P6R)/EI
(d) Exponential input function. If g(x) = we~a", then the solution by parta! fractions is even (e) Transport matrix equation (15.19a1) adjusted to the end conditions given in (a*) above is
simpler.
1 1 0 0 0 0
306 15.23 FOURTH-ORDER DIFFERENTIAL EQUATION Ordinary Differential Equations Ordinary Differential Equations 15.24 FOURTH-ORDER DIFFERENTIAL EQUATION 307
OF THE THIRD KIND, LTS GENERAL CASE 0F THE THIRD KIND, LTS SHAPE FUNCTIONS
M,N,R = signed numbers ^ 0 (a) Basic transform. As apparent from the Laplace transform of the fourth-order differential
(a) Notation. Y = solution equation in (15.236), the basic transform is
N R
A = p2 - 4a
P= M q= M
/<(') =TTyT
(b) Laplace transform of the fourth-order diflerential equation
and its inverse is the highest shape function of LTS designated as LAx). The remaining shape
d*Y *Y functions required in the solution are then the derivatives and the integris of Z.4(x) as indicated
Yo, Y'0, Y"0, C = initial vales
WT7
dx + *T?
dx + RY= *W in (15.23c). Since the denominator of/4(j) may be expressed in six distinct algebraic forms
(depending on the magnitude and sign op,q), six distinct forms of LAx) are avaable (Ref. 1.20,
i*, by (15.036), sec. 16.48).
/(*) = %[/.(') + PfA')\ + rol/A*) + />/(*)] + W3W + K/A*) +/A>)*(s) (b) Case I. \tp > 0 andp2 - 4a = 0, then with A= \fp~J2, by (14.116),
where
(C) Inverse Laplace transform of (6) by (I5.03c) is the Laplace transform solution (LTS) of the
2-l{/As)} =^"'{(J2 _ 2)'^ _ g)} =i(^ cosh nx - sinh nx) =4(x)
ciderential cquation given symbolically as
(d) Case III. If p > 0 and p2 - 4a > 0, then with w = V/>2 - 4a, A = V(/> - w)/2, n
y = y0[,(x) + pLAx)] + Y'0[LAx) + pLAx)) + Y0LAx) + C4(*) + <?(*) \/ip + a>)/2. by (14.116),
where ,(x) = T'^)} 0 = 1,2,3,4)
re the shape functions (15.03a") derived in (15.24) and *"'</. - x-ijFTjfcr} (^ - ^ -
Gix) =Se-,{/As)kis)} =I 4(X - T)(T)rfT (e) Case IV. If p < 0 and p2 - 4a > 0, then with w = V/>2 - 4a, A = V(|/>| - a>)/2, n =
V(|| 4- <u)/2, by (14.116),
is the convolution integral (15.03a*). Particular forms of this integral may be derived by the
Drocedurcs introduced in the preceding sections, and the closed-form evaluations are given in
Ref. 1.20, sec. 16.49.
*""< - *1> - W -,)H(^ - ^ =LM I
'IIll
M
(d) Properties f shape functions. For k = 1,2,3,..., (f) Case V. If p > 0 and p2 - 4q < 0, then with o> = V?, * = 5V2G2 - p, r\ = V2a)2 4- p,
<p = tan"' (A/n),by (14.116),
(a) Classification Of methodS. If the explidt solution of an ordinary differential equation is too (a) Concept The simplcst and yet seldom-uscd mcthod originally suggested by Euler is applicablc
involvcd or cannot be constructed, the effort must be made to compute the vales of Y at certain to the solution of the first-order diflerential equation
discrete points, usually selected at equidistan! spadng h to fadlitate the future interpolaron.
Although the literaturc of this topic is veryvoluminous, with the availability of digital computers,
four classes of methods providc all the necessary tools for the solution of most engineerng =/>> l'(O) Y0 = known condition
probiems. They are:
If the function /(x, Y) is continuous, sufliciently difiercntiablc, and bounded over the interval of
1) Euler's and Taylor's series methods integration, then by using the forward difierence representation of the first derivative, the vales
of y becomes, by successive substitution,
<2) Runge-Kutta methods
3) Predictor-corrector methods
4) Finite-difference methods y, = y0 + a/0 , y2 = y, + a/, ,..., y,+, = y4 4- a/,
where/0 =/(x0, Y0),fx =/(x,, Yx),... ,/k =/(x,, Yk) and A is the equal spacing introduced in
Their theorctical background and selected applications are shown in the remaining sections of (15.256). Geometrically the mcthod rcplaces the desired solution (which is a continuous curve) by
this chapter. a string polygon whose slope of cach side is the approximation of the first derivative of }'.
Although the method is very simple, its results are so inaecurate that it is almost never used in
engineering calculations. If combined with Richardson extrapolation (4.06/), howevcr, it may
yicld good results as shown below.
(b) Numerical solution. The first stepin the numerical solution ofa given diflerential equation is the
divisin ofthe interval of nvestigation [a, b) into r segments along the axis ofthe argument x (or
any other argument) by the points designated as (b) Example. The first-order diflerential equation
where/,. =d/J9x,/ky = d/JdY,/kx, =?/J<fx,.... This approach is called the Taylor's method (3) Fourth-order method
oforder n, and, as apparent, Euler's method is a particular case of this mcthod for n- 1. Ihe local
truncation error .+l can be estimated by the last term ofthe expansin.
K4+i = Yk+-(KA + 2KB 4- 2KC 4- KD)
o
(b) Example. The first-order diflerential equation dY/dx 4- 1' = x + 1 in [0, 1] with the initial KA=/ixk,Yk) Kc=/(xk +hb,Yk + ^hK)
condition y0 = 1 and A= 0.1 is solved for n= 2 and n = 4. The derivatives of/= >' = KD=/ixk+h,Yk + hKc)
- y 4- x 4- I are
KB=/ixk + h,Yk + hKA)
r =-(-y + x - l) = -y + i = y - x A = Yt ~ x* The cstimates of errors are complicated and very often unsatisfactory. The spacing Ain (3) is to
J x be chosen so small that
A = -Yt+xk 5KB
r = -(y - x) = y - i = -y 4- x
K"-kc- 100
r =i.(_y +x) = _r + i = y-x /r =; - *
ax
and
(b) Example. For the diflerential equation (15.276), the fourth-order Rungc-Kutia mcthod based
on relations (a-3) with A= 0.1 yields the following approximations of v,, Y2, and ys.
*k n f(x, Y) = x - Y + 1 n+.
(a) Classification by Steps. The numerical methods avaable for the solution of first-order (a) Predictor-corrector procedure. First, the starting vales
diflerential equations gencrally fall into two categorics:
Y0,Y,,Y2,...,Yk
(1) Single-step methods introduced in (15.26) to (15.28) requiring one initial valu Y at the left
end of the interval of integration which is divided into r subintervals (usually of equal and / =/(x0,YQ),fx =/(x,, Yt),f2 =/(x2, y2) /, =/(x4, y,)
length). The calculation ofYk+X nvolves only the preceding Yk,xk, and A.
ofthe respective equation are calculated by one ofthe preceding methods (15.27), (15.28). Next
(2) Multistep methods introduced in (c) to () below requiring several vales Y0, Yx,... , Yk on the the predictor-corrector set is selected from (15.29c) to (5.29) and the following iteration procedure
left of yi+l in the interval ofintegration, divided again into rsubintervals (usually ofequal is applied.
spacing A). The calculation of Yk+X involves the appropriate preceding vales of Y, x, and A.
Step P0: Yklx is calculated by the predictor formula in termsofthe starting vales.
(b) Multistep methodS are governed by the general formula Step F0: fklx is calculated by/(x, Y) in terms of xi+l and Ylklx.
StepC,: l'iVi is calculated by the corrector formula in terms ofthe appropriate starting
vales and x+1, K^./jB,.
Step Fx: /+| is calculated by/(x,^) in terms of xk+x and Yjk,2x.
where Ais a given point in the interval of integration and mis the order of approximation ir
6 =0 the designation ofexplicit or predictor formula for Y"k+X is used; in turn, when 60 * 0 the StepC2: }'i+, is calculated by the corrector formula in terms ofthe appropriate starting
desienation of implicit or corrector formula for YUi used. The recurrent application of these vales and x4+l, Yk'x,/[*>,.
two formulas is called the predictor-corrector method described in (15.30). The most comrnonly
used formulas are listed below. The most important aspect of these methods is their abihty to
estmate the truncation error.
The iteration process for l't+l stops with C, C,_, = 0. If Yk4.2 is desired, the same procedure is
repeated with Yi +l used as another starting valu.
YUi = K-3 + 4*(2/4 -/,-, + 2/i_2)/6 (b) Example. The application of Hamming's mcthod in the solution of dY/dx = x Y 4- I with
x0 = 0, Y0 1, and A = 0.1 is shown below. The starting vales by the Rungc-Kutta method
(15.286) are
Yk+i = n_,+ *(/+. + 4/*+/*-i)/3
y0 = i.oooooo y, = 1.004 84 y2 = i.oi8 77 y, = 1.04082
*. = YUi " (y* " Y'+M29 = r+l " ***'
/0 = 0 /, = 0.09516 /2 = 0.181 23 /s = 0.259 18
n+, = YUi ~ >W+. - >'*.)/270 = YUx ~ ^i f,: /i" = 0.4 - 1.07032 4- 1 = 0.329 68
(b) Second-order differential equation and the sum of this series of matrix functions is the transport matrix,
J2y Ay
Txx(x) TX2(x) TX3ix) TX4ix)-]
A2-^ + Ax+ A0Y =gix)
T2X(x) 7-2,(x) T2Ax) T24{x)
T(x)
reduces to
T3lix) T32ix) T33ix) T^ix)
lT4Xix) T42ix) T(x) T(x)J
aM
dx
4- AXU 4- A0Y =gix) dx
reduces to
and the sum of this seriesof matrix functions is the same matrix as in ig) but the argument is -x.
dV
A3^ +A2U +A,V +A0Y =gix) dx dx
= U
dx (i)C Convolution in tegral G(x) is then
(d) Fourth-order differential equation -r(x) rl2(x) t,Ax) TX4ixy rTu(-T)g(r)-\ -Gxoix)
Tixix) T(x) T23ix) T24ix)
G(x) =-J- T2Xi-T)gix) GM(x)
A4 T*\(x) T32ix) T3Ax) T34(x) f T3A~T)gix)
dt =
Gyoix)
_r41(x) r42(x) r43(x) r^x). -T4A-x)gix)_ -GUx)
reduces to
dW
A4 + A3U + A2V + Ax W + i40y = gix)
dx =' dx -" (j) Closed-form transport matrixequation written for the subinterval of length A is
i. !
1 1 0 0 0 0
(e) Differential matrix equation constructed from these reductions becomes in case id) with
Bk = Ak/A4. A <?io(A) T,Ak) Tl2ih) TX3ib) TX4ih) Uo
dU/dx -B3 -B2 -Bx -B0 T/" ~gix)/A~ K GM(k) T2Xih) T(h) T23ih) T2Ah)
dV/dx 0 10 0 V 0 Wx GMih) T3lib) T32ib) T33ih) T34ih)
4-
dW/dx 0 0 10 w 0 Yt G(k) T4Xih) T4Ab) T43ib) 7-(A)_ \
dY/dx 0 0 0 1 y 0
H, Tl0
dW/dx D H L(x)
where H, is the state vector at x = A.
(f) Transport matrix equation. The solution of the diflerential matrix equation in () is
(k) Transport Chain. The final solution ofthe diflerential equation (6) is
H=exp (Dx) Ho 4- exp (Dx) Jo exp (-Dr) L(r) a*r =T(x)Ho 4- G(x) H| = TloHo, H2 = T2,H,,..., H. = T_,H,_,
where H, Ho are the state vectors at x > 0, x = 0, respcctively, D is the matrix of constant where the T coefficients qfmatrices T,0, Ta, ,^^,I_, are the same numbers and the G function vtf
-^oeffidentsrT()--thc-transport-matrixHn^ corresponds to the g functions of the respective subintervals. The numerical example of this
integral vector defined in (i) below. method is shown in (15.32). j
316 15.32 SYSTEM OF FIRST-ORDER DIFFERENTIAL Ordinary Differential Equations Ordinary Differential Equations 15.33 SYSTEM OF FIRST-ORDER DIFFERENTIAL 317
(a) RedUCtion. The second-ordcr differential equation (a) Differential matrixequation. If the coefiicients A0,AX ,A2,A3,A4 in (15.31a") are functions ofthe
argument x, then the differential equation has variable coefiicients and the matrix equation
(15.31) has also variable coefiicients in D.
3-42-4r-o
dt2 dt
dU/dx-\ B3(x) -B2(x) -BAx) -BAx)~\ r/-
with t0 = 0, v0 = I, Y'o = 2, and A= 0.1 in [0, 1] is reduccd by (15.316) to dV/dx 0 10 0 V
dW/dx 0 0 10 w
^+4/-4y = 0 dt
dY/dxJ 0 0 0 1 y.
dt
In such a case a substtute model with constant coefficients in each subinterval must be used or
(b) Differential matrix equation (15.31*) is then the solution of this matrix equation must be approximated by the Runge-Kutta method, which is
well suited for this purpose.
M/dl D H
(b) Runge-Kutta method applied in the solution of the
diflerential equation shown in (a) divides the subinterval A
into two segments, shown in the adjacent figure. The
(c) Transport matrix in series form (15.31j?) becomes coordinates of their endpoints are x0(, x0>, x(W and the D
481 /
matrices at these points are
ri o r-4 4i / r 12 -i6i /-' r-
-32
r/3-| __ r 0.654984 0.3274931-0.268 1211 = I" 0.2195371 K,, = |(D, 4- 4D, 4- Dk) Kc = i(D,D2 + D2D,)
Ly,J ^ L-0.081876 0.982483JL 1.206598J L I.2074I4J Kfl = D.D/D,
KB i(D,D, + D2 + DjDk)
and similarly H4 = TH H5 = TH4,..., H, = TH.,. These results can be compared to__the_
exact vales which are L, = -0T9M?77rr, = -0.268 128. U3 = 0.219 525. Yx = 1.146 223, and HHt are the state vectors related by the transpon matrix Ttl. The inclusin of the
y2 = 1.206576, y, = 1.207 385. convolution integral follows the pattern introduced in (15.31i).
I 1
318 15.34 HIGHER-ORDER DIFFERENTIAL EQUATIONS, Ordinary Differential Equations Ordinary Differential Equations 15.34 HIGHER-ORDER DIFFERENTIAL EQUATIONS, 319
FINITE-DIFFERENCE METHOD (Continued)
FINITE-DIFFERENCE METHOD
(a) Concept. The finite-diffrence method replaces the interval ofintegration [a,6] by a selected sct (e) Joint input functions. Since the vales of Yand its derivatives are related to LL, L, C, R, RR,
of points (joints, nodcs, pivotal points) and approximates the diflerential equation and its the input function gix) must also be related to these joints bycquivalents called the joint loads.
initial-value equations or its boundary-value equations by the respective diflerence equations For this purpose cach segment Ais treated as a simple beam whose reactions are thejoint loads.
related to these points. This substitution leads to a sct of algebraic equations written for all The tablc (/) gives the most common cases ofjoint loads expressed in termsof
selected points, and the solution ofthis sct yields the approximate vales of y at these points. The P = conccntratcd load Q = applied couple
spacing of these points is usually equidistant, but uneven spacing may be used in special
situations. Bccausc of higher aecuracy, only central diflerences (3.02d), (3.020 are commonly w = intensityof distributed load G = joint load
used. Since odd-order diflerential equations are hard to handlc by finitc diflerences, the
transformation to even-order equations is performed by diflerentiation or integration. a stringangle A= segment length +=l m,n = fractions
L[xoc ~ b, YL\
0^
^1" ('.
Gc GCi. 4- GCR
Gon = mP
nP
T t"
C[xoc>Yc] t_ 4-
R[xoc 4- A, YR)
(C) Difference operators at Cbased on relations (3.02a1), (3.02) are, with k = 1,2,
<5<>yf; = yc <5>-<+,>yc = -(2i,r, 4- iwyk
6tyc = -y, + YR 6^2>YC = ,2"yt - 2iWYc + 6<2"YR G,.c = 6(2tf/. 4- wc)h
d2Y-\ % YL - 2YC + YR fa^n _ ,4>yc _ Yu. - *yl +6ir - *yr + yri ara Gc = Gcr 4- GCR = h(u>t. 4- 10av + w)h
U2J~ A2 h2 lax4JV A4 A4 ,j ..,jj,.. Grc ~ i*(~wi. 4" 6"<7 4- 7wR)h
and in general,
du^Yl - r5<21+"yf;
fL_Il Y(: = -m)YL + d'2"YR
2A2 r
^
Gt.c = k(2wL + wc)h + k<*Lcb'
r^^yi
a^yi 6l2l+2yYc
(24+2,yc = (2*+"yf - 26{2t+"Yc 4- l2l+"YR ; rfl -
jiy y 1 - y = iox
M(x) -f-2 4- Nix) + Rix)Y = gix) dx2
dx2 dx
in [0,8J with the end conditions y(0) = 0, y(8) = 0, r = 4, and A = 2 is solved by the standard
is at C, KLYL - KCYC + KRYR = Gc finite-diflerence method introduced in (15.35a). The governing diflerence cquation with M = 1.
N = 0, R = -1 and KL = 1, Kc = 2 + 4, KR = 1 becomes
wherewith A(xoc) = Mc, A^(xoc) = Nc, ^(xoc) = Rc,
YL - 6YC +YR = 40xc
Nc R<Jr
Kr = 2 - X,= l + By the recurrent use,
Kl ~ ' 2Mc Mr 2Mr
1 -3 I n 20
1 -3 1 Y3 30
(c) Endpoint equations. Two endpoint equations are required in the solution of the second-order
equation. Thcy must beselected from the following set. I -3 I Y4 = 40
1 -3 Y,_ 70
(d) NumerOV'S method. If W(x) = 0 in (a), then the diflerential equation may be written and y(0) = y0 = 0 y(5) = y5 = -45.543 (-46.017)
symbolically as y(l) = y, = -9.920 (-9.937) y(6) = y6 = -48.328 (-49.173)
(PY y(2) = Y2 = -19.759 (-19.805) y(7) = y7 = -39.443 (-40.570)
y (3) = y, = -29.352 (-29.462) y(8) = y8 = o
and its diflerence equation becomes y(4) = y4 = -38.298 (-38.535)
n-*re+i,--li(25F
A2 / R,
+iOjre
Rc
+JJi.)
Rr .. \
+1r-
Gcb~ The exact solution ofthe diflerential equation is }'(x) = 80 sinh x (sinh8)"' lOx and the exact
vales of y(x) are listed in parentheses above.
(b) NumerOV'S method applied to the same problem with r = 4, A = 2 yields after scaling
RLh2 \0Rch2 R*h2
from which A', = 1 + Kr = 2 - KR=\ + -8 1 0" "y," "120"
\2Mt, 12MC \2MB
1 -8 1 Y2 = 240
This mcthod uses the second-order diflerence as the approximation ofthe second derivative of Y 0 1 -8_ LY3] _360_
and the parabolic joint load formula for the approximation of the variation of }'. The better from which-y(2) = y, = -19.839, y(4) =-yj-=38.710, y(6' = Y3 = -49.839. The com- I
aecuracy of Numerov's mcthod is shown in_(JJL3fiA.l. parison of these results shows that Numerov's method is far superior to the standard method.
s^ J^^^^J ^^^^^^^^^^^^^^^^^^^^F ^ m*
Hgi
W
322 15.37 FOURTH-ORDER DIFFERENTIAL EQUATION, Ordinary Differential Equations Ordinary Differential Equations 15.38 FOURTH-ORDER DIFFERENTIAL EQUATION, 323
is at C,
in [0, 8] with the end conditions y(0) = 0, y'(0) = 0, y(8) = 0, 1"(8) = 0, r = 4. and A = 2 is
Ku.Yu. ~ Rifi + KcYc ~ KrYr + KRRYRR = Gc solved by the method introduced in (15.37a). The governing difference equation with Mc 1,
Nc = 0, Rc = 4 and KLL= 1, K,. = 4, A'c = 70, KR = 4, KRR = I becomes
where with tpj = iNjh2)/Mj and y> = (RJk*)/Mit
Yu. ~ *YL + 70YC - 4YR + YRR = 100
Kt = 4- <Pl *c = 6 " 2</>c + Ve ** = 4 - <*> ^* 1
and the endpoint equations are Y_x = Yx, Y0= Y4- 0, l's = Ij. The use of these endpoint
and Gc = Gch3/M, = scaled joint load (15.34/).This equationapplies at all intermedate points, equations and the recurrent use of the governing diflerence cquation yields
and as in (15.35a) for the cndpoints 0 and r the endpoint equations shown in (6) below must be
used. 71 -4 1 Yt 160
-4 70 -4 Y, = 160
(b) Endpoint equations. Four such equations are required for the solution of the fourth-order L i -4 71_ Iy'J _160
equation. They must be selected from the following sets.
from which Yx = 2.364, 12 = 2.556, Y3 = 2.364. If the number of segments is increased to r = 8,
Y 4- y then with A = I, the governing diflerence equation becomes
YiO) = a0 Y0= a0 Y'(0) =or, Y0 =^ ' = ax
Yu - 4114- ioyc - 4y 4- YR 10
y_, - 2y0 4- y,
y"(0) = a, Y"0 =-^' = a2 and
y(rA) =/30 y, =/S0 y'(rA) =/3. ^ =^^f^2 =/?. y (3) = y3 = 2.578 (2.617)
-(24 - 100) (12 + (p) 0 The first three roots of this determinant equation are the eigenvalues
(12 4- <p) -(24- 100) (12 + 0) = 0 1.092 . 4.087 . 6.648
A, = A,
0 (12 + 0) -(24- 100) L "L' L
As in (a), the expansin of this determinant is the characteristic equation The exact eigenvalue equation is
-(24 - 10A2A2)3 + 2(12 4- A2A2)2(24 - lOA'A-') = 0 cos A = 0
The roots of this equation are the eigenvalues and the exact eigenvalues are
4 /246 - V4I472 3.139 4/24 4 f.264 - V41472 9.254 _ K 1.571 4.712 _ 5,T _ 7.854
98 98 ,~2L~ L 2~ L 3 L L
(c) Matrix method given in (8.11a) yields the simplest and the most accurate procedure for the (b) StSSi'S determinant of the same equation bv the same endpoint equations and by the relation
calculation of eigenvalues. The closed-form solution ofthe matrix equation in a) is (15.37c) with r = 4, A = /4, and 0 = h'2X1f\2 is
I k (7 - 190) -(4 - 80) (I - 0)
Xk = rJ2 - 2 cos
-(4-80) (6-180) -(4-80) (1-0)
Forr = 10, A", = 3.128 87,X, = 6.180 34.A3 = 9.079 81....
(1-0) -(4-80) (6-180) -(4-80) (1-0)
r = 20, , = 3.138 36, X\ = 6.257 38,A3 = 9.337 81,... = 0
(i - 0) -(4 - 80) (6 - 180) -(4 - 80) 1-0
r = 30, , = 3.140 I6,, = 6.271 71,A3 = 9.38607,...
1 (2 - 120) 0 -(2 - 120) I
r = 50, A, = 3.141 08, X2 = 6.279 05,A3 = 9.410 83,...
1 - 2 1 0
r = 100, A", = 3.141 46, X = 6.282 15, ", = 9.421 29,...
The first three roots of this determinant equation are the eigenvalues
r=IOOO, A, = 3.141 59,A2 = 6.283 18, A~3 = 9.424 74,...
1.568 4.669 7.635
A, = A, = _A^=
The aecuracy of these results depends only on the computer capability in the evaluation of tltc
-cosine-functiorrfor veiysmaihargameTTts.
which shows again the superiority of Stssi's mcthod.
1 ~~3 ~1 ' 1
(a) Numerical methodS for well-poscd probiems must sattsfy the conditions of consistency, stability,
and convergence.
(1) Consistency. If in the limit A* 0 the substitute function and the true solution are equivalent,
then the numerical solution is said to be consisten!.
(2) Stability. If the errors at one stage of calculations remain finitc all the time (do not cause
larger errors as the computation contines), then the numerical method is said to be stable.
(3) Convergence. If the results of the numerical method approach the true vales as Aapproaches
zero, then the numerical method is said to be convergent.
(b) One-Step methodS such as Taylor's series and Runge-Kutta methods do not show numerical
16
instability if the step sizes of integration are small. The Taylor's series method is the simplcst
method, and it can be easily adjusted to the diflerential equation and to the desired aecuracy.The
only diflkulty arising in the application is the difierentiation of the given diflerential equation.
The higher derivatives can be convenicntly calculated by the extensive tables given in Ref. 1.20.
Runge-Kutta formulas are the most widcly used procedures for the numerical solution of
PARTIAL
ordinary diflerential equations. Their primary advantages are their easy programming, good
stability, and step-size changes without complications. Their disadvantages are the significantly
longer required computer time and the lack of dcpendablc error cstimation.
(c) Multistep methods such as Adams-Bashforth-Moulton method and Hamming's method are
DIFFERENTIAL
strongly stable for a sufnciently small A, and their great advantage is their requirement of one
derivative calculation. They can be easily adjusted to the diflerential equation and to the desired
aecuracy.
(d) Finite-diflerence methods providc a simple procedure for the solution of boundary-value
EQUATIONS
probiems, and if combined with the extrapolation procedure, theyofler results of good aecuracy.
An extensive evaluation of finite-diflerence solutions of ordinary diflerential equations can be
found in Ref. 15.03.
,!r
lili
Partial Differential Equations 16.02 METHODS OF SOLUTION 329
328 16.01 GENERAL CONCEPTS Partial Differential Equations
(a) Types Ofsolutions. There are many methods by which the partial diflerential equations can be
(a) DefinltionS. Apartial diflerential equation oorder nis a/unctional equation containing atleast one solved. The most commonly used methods are:
partial derivative of the unknown function
(1) Separation-of-variables method (4) Finite-diflerence method
4> = 4>(x, ,x2,...,xj
(2) Laplace transform mcthod (5) Finhc-clement mcthod
oftwo or more variables x,, x2 ,... ,xm. The order ofa partial diflerential equation is theorder of (3) Complcx-variable mcthod
the highest derivative of 4>.
The concepts of the first, second, and fourth method are introduced below.
(b) Solution ofa partial diflerential equation oforder ngiven in general form as (b) Separation-of-variables method. Although this method is not universally applicablc, it ofiers a
convenient and simple solution to many important equations in engineerng and physical
scienecs. The basic idea is to assumc the solution in the form of the product of two or several
3<t> a<t> 9<p a2* ajo a2* _ functions, each of which contains one variable only. The insertion of this product in the given
Fixx,x2,...,xm,*, ^^-^'^''^')
dxl diflerential equation seprales this equation into two or several ordinary diflerential equations.
The product of the solution of these equations is then the solution of the partial diflerential
equation. Particular applications of this method are given in (16.03) to (16.12).
is any function which satisfies the equation identically. The general solution is a solution which
contains a number of arbitran- functions equal to the order of the equation. Aparticular solution (c) Laplace transform method. The basic idea of this method is to obtain the Laplace transform of
(particular integral) is a special case othe general solution with specific functions substituted for the respective partial diflerential equation and ofthe associated end conditions with respect to one
the arbitrary functions. These specific functions are generated by the given conditions (initial variable. By this transformation, the partial diflerential equation is reduced to an ordinary
conditions, boundary conditions). Many partial diflerential equations admit singular solutions diflerential equation whosc solution is found by taking its Laplace transform inverse. The
(singular integris) unrclated to the general solution. following relations are used in this method. If
(C) Linear partial differential equations and systems of such equations appear in the description o 2{Fix,t)) = fe-Fix,l)dt =/ix,s)
engineerng and scicnce probiems. In this chapter. the classical and numerical solutions of the Jo
second-order partial diflerential equations (which domnate the application* >are presented. The
general linear partial differential equation of order two in two independent variables xand ,v is of the is the definition of the Laplace transform of Fix, t), then the transforms of the dcrivatives of
form Fix, t) are
(d) Classification. Because ofthe nature of their solution, the partial diflerential equations in (c) are o^Fjx.t)) d/jx,s) 9F(x. 4-0)
designated as: l Bxdy I 9x
(3) Parabolic equations if B2 - 4.4 C = 0 (d) Finite-difference method (as in the case of ordinary diflerential equations) rcplaces the regin of
Elliptic equations are ofa steady-state type associated with conditions which require that * or its the partial diflerential equation by a specific grid and approximatcs the governing equation and
normal deriv; tives, or both, take on the closed boundary some specific vales and the dosed the end condition equations by the diflerence equations, in terms of the unknown vales <t>,,
boundary prercribes the regin of application of O. Parabolic and hyperbolic equations are ofthe which are related to the nodal points of the selected grid. This substitution leads to a sct of
non-steady-st;'tc type (propagation type), and their solutions are allowed to propgate in the algebraic equations, and the solution of these equations yields the numerical approximations of
%. Finite-diflerencemolecules and their applications are tabulated in (16.17) to (16.31).
open flnmain il Iras' in "n<* variahlc ~~~
330 16.03 LAPLACE'S EQUATION, SOLUTION Partial Differential Equations Partial Differential Equations 16.04 LAPLACE'S DIFFERENTIAL EQUATION, 331
BY SEPARATION OF VARIABLES EXAMPLE BY SEPARATION OF VARIABLES
(a) Notatlon (a) Temperature distribution U(x,y) in a rectangular slab of sides
a,fi = separation constants
a, b (shown in the adjacent figure) is governed by the Laplace's
A,B,C = integration constants equation
x,y = cartesian coordnales r, 6 polar coordinates
d2!/ d2U _
V = !'(r, 0) = polar solution
U = U{x,y) = cartesian solution dx2 + dy2 ~
(b) Cartesian Laplace's equation in two-dimensions, and the edge conditions are given as
admits a solution U = A'(x)y(x) = XY, which when inscrted in this equation yields (b) Separation-of-variables solutions are, by (16.03c),
ifX^_lY Af(x) = Axe"" + A2e~,c" = x cos ax 4- j42sin ax
Xdx2 Ydy2
Yiy) = Bxea' 4- B^0 = Bx cosh cr> + 52sinh ay
Since cach side of this cquation is a function of one variable only, each side must be equal to the
same constant, denoted as a2 (which can be real or complex). or in terms of phase arguments x0 ,y0,
(c) TWO ordinary differential equations resulting from this separation are Xix) = Csn aix + x0) Yiy) = sinh a(_y -.)
d X ., .. Y 2v n (c) Particular edge conditions. Ufx(j) =Aiy) =A(x) = 0and/3(x) =/(x), then
77 - or2}' =0
dy (\)'U(0.v) =0 = Xi0)Y(j) (3) U(x.0) =f(x) = X(x)Y(0)
and their solutions are A' = Axe'" 4- A2e" Y = B,e> 4- B:e~" (2) U(a.y) = 0 = Xia)Y(j) (4) U(x.b) =0 = A'(x)y(6)
(d) General solution is then From (I), Y(y) * 0, X(0) = 0 = x and from (2), Y(y) # 0. Af(a) = 0 = i, sin aa. where with
/12 = 0, sin aa = 0, which implies
riAxeitt' + A2e-ea)(Btear ia*0) ak = Lr/a (k = 1,2,3,...)
U= XY = \
lM, +A,x)iB, + Brf) (* = 0) and yields A',(x) = A2J sin akx.
From (4), A'(x) * 0, y(6) = 0 = Z)4sinhrA(6 -y0), which requires y0 = 6. Thus from (1),
(2), and (4). with ak = st/a, 2ji/a, Zn/a,...,
(e) Polar Laplace's equation in two-dimensions,
dr2 + ~rdr+r2dff2~
= 2 * sm *xsmn<**(.>' *)
admits a solution V = R{r)QA6) = RQ,, which when inserted in this equation yields after
separation two ordinary diflerential equations which must satisfy (3) as
2R \dR B2R n
d7+r7r- = Q 2*- Uix, 0) = 2 * sin t*,x sinh a,(-6) =/(x)
<-i
and their solutions are R = Axrfi 4- A2r~^ Q = Bxe** 4- B*-** where t sinh a(-6) is the sinc Fouricr coefficient of/"(x). eiven in (13.02) as
-(g^'ConstantSTlHntegratlom,, A2 and A,, #, are to be detcrmincd from the given conditions as Uix, y) = 2 k sinakx sinh akiy 6)
shown in (16.04c).
Partial Differential Equations 16.06 WAVE EQUATIONS, EXAMPLE 333
332 16.05 WAVE EQUATION, SOLUTION BY Partial Differential Equations
BY SEPARATION OF VARIABLES
SEPARATIONOFVARIABLES
(a) Vibration of tightly stretched string of length a{m) fixed at both
(a) Notation cnds and subjected to a uniform tensin 5 (shown in the adjaccnt
A,B,C = integration constants / = time c = signed number figure) is governed by the displacemcnt equation
?.*-.
(2) U(a,t) = 0 = X(a)T(t) (4) 'u.0) = /(x) = X(x)T(0)
-T
dt'
+iac)2T =0
From (1), 7(/) =t 0, Af(0) = 0 = .4, and from (2), T(t) * 0, X(a) = 0 = .4,sin era, where with
Y = Bxeu" 4- B2(- (ea = are) A2 * 0, sin oa = 0, which implies
and their solutions are A' = Axe"" 4- A2e~'a
ak = A;r/a (A = 1,2,3,...)
(d) General solution in one dimensin is then
and yields Xkix) = j4.a>i sin akx.
(V* + -'"""KV" + B2e-"' (a =0) From (3), AT(x) * 0, dT(Q)/dt = 0, B2 = 0 and yields
U= XT=\ ia = 0) 7(/) = 5, cosca,/ cu, = akc = cAjr/a
lAx 4- i42x
Then
(e) Cartesian wave equation intwo dimensions
U(x, t) = 2 ^t(*)7*(0 = 2 -^sin ax, , cos wkt
k-\ -l
334 16.07 DIFFUSION EQUATION, SOLUTION Partial Differential Equations Partial Differential Equations 16.08 DIFFUSION EQUATION, EXAMPLE 335
BY SEPARATION OF VARIABLES BY SEPARATION OF VARIABLES
(a) Notation (a) Variation Of VOltage In SUbmarine cable of length a (m) with grounded cnds and with initial
voltage distribution/(x) is governed by
A,B,C = integration constants / = time c = signed number
where A,,A2,... have the same mea'iing as in (16.03). where Pk = (2A - l)jr/a and A, = B\/RC.
336 16.09 BEAM VIBRATION, SOLUTION BY Partial Differential Equations
Partial Differential Equations 16.10 BEAM VIBRATION, EXAMPLE BY 337
SEPARATION OF VARIABLES SEPARATION OF VARIABLES
t = time
(a) Transverse vibratlon without dampingof a simple beam is govemed by the diflerential equation
(a) Notation given in (16.096) and by the conditions
E = modulus of elasticity (Pa) / = moment of inertia (m4)
... _. 5ioa2x(a x)
x,y = cartesian coordnales m = mass per unit length (kg/m) V(Q, /) = 0 v(*>) zhr,-=A*)
96EI
Via,Q) = 0
(c) TWO ordinary differential equations resulting from this separation are Accordingto (16.09c) and by (1) to (4), A, = A3 = 0. .4., = A"(0), A4 = A""(0). and the transport
matrix (15.03c) reduces for x = a to
fT 2rr n f LAa) 4(a)irA"(0)-| TOl
4- w2T = 0
2?-- dr
I.A44(a) 2(a)JLA""(0)J LoJ
wherc A = ymar/EI. The determinant of this matrix cquation yields the frequeney equation
(d) Laplace transform SOlutionS of these ordinary diflerential equations are the eigenfunction sin Aa = 0 A, = kn/a (A = 1,2,3,...)
and the time function T = J9,A',(/) 4- B2KAD A'A (jc) = Ck sin A,x
where From (5), 7"(0) = 0 = B2, and with wk = EIk23l2/ma:. the time function becomes
TAI) = BXJt cos wkt
,(x) = 5(coshAx 4- cosAx) L(x) = 5(sinhAx 4- sinAx)/A
The sum of all solutions is then
LAx) = (cosh Ax - cos Ax)/A2 4(x) = '(sinhAx sinAx)/A
Kxit) = cosca/ K2it) = (sin wt)/w K(x, /) = 2 Gk sinXkxBx cos wkt = 2, Dk sinA4x cos wkt
.. d2Z w dZ
M2-2 +
">7x + (A/o 4-Ar0)Z = gix)
ax2 c2 di2 ~
(6 = constant)
where A is the separation constant and A', Y, Z are the integris of the respective diflerential ICUiO.t) = 0, /(a,/) = 0, /(x,0) = 0, /(x,0) = 0. and with
equations. U = Xix)Tit) + Zix)
the separation yields with w = ac,
(c) Solutionof the second kind with G = giy) assumes X Z
dx2
4- a2X = 0 -?+w2T=0
dt2 dx'
= -6x
<D, = Xix)Yiy) <t>2 = Ziy)
so that 4>, is the solution of the homogeneous equation and The solution ofthe first two equations is given in (16.06c), and the integral ofthe third cquation
is
<j> = 4>, + 4>2
Z = -6xJ/6
s the solution of the nonhomogcncous cquation. After separation.
The complete solution is then with A, = kst/a and wt = ckst/a,
A/2^'+A/,^4-(A/fl-A2)A' =0
dx' dx
U(x, /) = 2 G^u cos ^* + 2Jt sin Xkx)iBXJt cos wkt 4- B2 sin wkt)
bx3
6a3 ba3wk
N2?-i + Xx^- + iNo + X2)Y = 0 where AXJ = 0 An D\ 2JkB2
dy2 +' ldy 6 sin A,a 6 sin Xka
(4)
sinh (sx/e)
j2cosh isa/e)
-+-^Kr.^cos^
cosh (sx/e) <7 2 ^. n <r"uix,s)
(5) s sinh (safe)
- + - > - f.r. ax sin ore/
a a..i
and becomes
x2
=-![2(x,j) - /(x,0) - "(x,0)]
(6)
cosh (sx/c) +--5-:,-m(1
2a ar <_, o"
-cos or) where a(x,) = j e~"U(x,t)dt
j2sinh (sale) Jo
cosh (sx/e) is the transform of U(x, t) and by (3), (4) above, ."(x. 0) = /0sin (stx/a), U'(x.O) = 0.
(7)
1+ -T ^cp'xcospV/
s cosh (sa/c)
(c) Transformed equation is
(8)
cosh (sx/c) 4--.-o*^xsnpV/ (Pujx.s) s2 s2 . XX
i* cosh (sa/c)
TI 2(x,s) = -/0-sin
ax c c a
sinh (xys/e) 2c2 "
T na exp(o-2*2/) sindx
(9) sinh (aVs/c) and its general solution (Ref. 1.20, p. 180) is
(10)
sinh (x\s/t) -f i?3-/5Vl)sin^ / \ a usx __ a usx j.1- Ua-z-r
a(x,x) = A cosh 4- B sinh
ti s sin (xx/a)
rr^i
VTcosh (aVs/e) v ' c c c2s2/c + .T'/r
sinh (xV7/r) -+-25p -/JV/)*in/?x (d) Integration constants are from (1) and (2) in (6). A = 0, B - 0, and the final form of the
(N) a a ,.,0
s sinh (avs/c) integral of the transformed diflerential cquation is
sinh (xvT/r) x/ 2 ^ i) ., , s,i . jsin (ixx/a)
(12)
+ r, 2, ~ exP (~oc e'(\ sin Ox u(x,s) = UQ
asinh(<iV7A) a ar" 4_t a
s + c Jt /a
2r"
cosh (x\s/c) (e) Inverse Laplace transform of this cquation taken from (14.0Ic-8) is the solution of the wave
(13)
cosh (a\s/e) a -1
equation in (6).
cosh (xVs/c) ' 2r ^> 2 j..
(14)
- + N r;xd t or c /) cos ax 7(x,/) = /0sin (Jrx/a) cos (rt/a)
VTsinh (aVi/r) a a "^,
cosh (xV7/r) 1+?V^fxp(-/52c20cos^x (f) Alternativo solution. There is always the possibility to take the Laplace transform of the given
(15)
cosh (aVs/e) a._,P diflerential equation with respect to x rather than / as done above. This choice would, however,
lead to certain difficulties as would become apparent upon the complction of this altemative ti
(16)
cosh (x\s/c) * 2r-'
~ fl +/
' "
2 'y ^ <rxp (-B2c2t) cos/Jx
ar'-" '^', p" r ' ^
approach. In practice, it may be desirablc to take the transform with respect to each variable and
s' cosh (aVj/r) compare which of these variables oflers the best simplfication.
342 16.14 HEAT EQUATIONS SOLUTION BY Partial Differential Equations Partial Differential Equations 16.15 VIBRATION EQUATION SOLUTION BY 343
LAPLACE TRANSFORM METHOD LAPLACE TRANSFORM METHOD
(a) Notation (16.07a) a, c, L'r. = positive constants (a) Notation (16.05a) a, c,E,P = positive constants
/(x, /) = temperatura t/(x, 0> = initial temperatura (/ = 0) 7(x, /) = displacemcnt 7(x,0) = initial displacemcnt (/ = 0)
7(0,/) = left end temperatura U(a, 11 = right end temperatura 7(0, /) = left end displacemcnt U(a, t) = right end displacemcnt
(b) Cartesian heat equation (16.076) in one-dimension, (b) Vibratlon equation (16.056) in one dimensin,
and becomes
a^ux.j) 1 . a*f(x,j) 1 .,
j = -[u(x,j) - /(x,0)J and becomes , = -z[s'uix,s) sUix.0) "(x,0)1
x r ax c
is the transform of U(x,t) and by (3), /(x,0) = 0. is the transform of 7(x,/) and by (3) and (4), '(x.0) = 0, /'(x,0) = 0.
d2u(x,s) s a*a(x,.Q
2u(x,s) = 0 -m(x,j) = 0
x r dx2
and its general solution (Ref. 1.20, p. 180) is and its general solution (Ref. 1.20, p. 180) is
u(x, s) = i4 cosh (xVj/c) + Bsinh (xV7/c) u{x,s) = A cosh (xj/c) 4- B sinh (x/c)
(d) Integration constants ara from (1) and (2) in (6), (d) Integration constants from (1) and (2) in (6) are
1 cP
A = 0 B = V A =0 B =
s sinh iavs/c) Es1 cosh ias/c)
and the final form of the integral of the transformed diflerential equation is and the final form of the integral of the transformed diflerential equation is
, ,. sinh (xV7/f) P c sinh (x/c)
u(x,j) = /0TTTTTTT uix.s) = 7-5-
j sinh (avs/c) E s2cosh ias/c)
(e) Inverse Laplace transform of this equation taken from (16.12M1) is the solution of the heat (e) Inverse Laplace transform of this cquation taken from. (16.126-4) is the solution ofthe vbration
equation in (6). equation in (6). '
/(x.0 = Ji
._ La
+-3p(-V0
a*_(x
Uix, 0=I [* +\ 2.sin Bx cos efit]
where er = kn/a and n = ( I)* where B = (2* - l)/2a and n = (-I)1.
344 16.16 FINITE-DIFFERENCE METHOD IN Partial Differential Equations Partial Differential Equations 16.16 FINITE-DIFFERENCE METHOD IN CARTESIAN 345
CARTESIAN SYSTEM SYSTEM (Continued)
(a) Solution of a cartesian second-order partial difler (C) Forcing function ofthe partial diflerential equation (16.01c),
ential equation in two dimensions (16.01c) by the G = G(x,y)
finite-diflerence method requires first the replacement
of the regin of integration by a rectangular grid of n i i ' -
at the same point C is taken as the joint valu
horizontal lines and m vertical Unes whosc points of
Gc = Gixc,yc)
ntersection are the nodal points (pivotal points, joints) | | i II !
shown n the adjacent figure, where each point , j is or is approximated more accuratcly by the equivalent joint valu in terms of the respective
given by I I ' I 4
formulas given below.
l t l fc
x=xc4- (1 -a)p y=yc+H-b)q
yj=M (j = 0,l,...,)
as shown in the adjacent figure is rcplaccd by
which is the solution ofthe pardal diflerential cquation in (16.01c), are next expressed by the
finite-diflerence approximations (3.02rf) related lo a particular point. If the point Cshown in the Pl
adjaccnt graph is that point, then the partial dcrivatives of* in diflerence form become where a, 6 are the same as in (d).
a2* A q\ - 2<&c 4- <pr (g) Constant function G rapresented by the diagram of the
dx2 p adjacent figure is replaced by
Gr = G
d2 = __ Qlb - E/r ~ Qrb + *rt
dx dy ~ " "" 4/ia
(h) Variablefunction G(x, y) rapresented by the diagram of
the adjaccnt figure is replaced by
* ^ <PW - 24><: 4- <I>r
-rr, = - i / G,.T+ \GT+ GRT\
dy q-
Gc = 36 \++ 4C + l6Gc + 4C/
GtJ> 4- 4GB+ GRBJ
where the subscripts define the location of <t> as L= left point, C= central point, R- nght
point, B= bottom point, T= top point, LB = left-botiom point, RB = right-bottom point, where GLT, GT,..., GB, GRB are the vales of G at the
LT = left-top point, and RT = right-top point. These equations apply to all points of the grid, respective joints and the terms in braces represent an --f:
including the points on its boundary. algebraic sum.
^^^^^ f
346 16.17 ELLIPTIC EQUATION IN CARTESIAN SYSTEM, Partial Differential Equations Partial Differential Equations 16.17 ELLIPTIC EQUATIONS IN CARTESIAN SYSTEM, 347
FINITE-DIFFERENCE RELATIONS FINITE-DIFFERENCE RELATIONS (Contnued)
(a) Notation (g) Stifmess factors in (/) are
U = &ix,y) = solution ofdiflerential cquation in regin Rwith boundary 5
KT = ir*Cc + hrpEc)/p2
k = grid points L, T, C, B, R, depicted in (16.166)
Uk = valu of7 atA Kk = stiflhess coeflicient atk K,. = (Ac - hpDc)/P* Kc = i-2Ac - 2r2Cr 4- p2Fc)/p2 Kr = iAc + >Dc)/p2
A,C,D,E,F = given functions Ac, Cc, Dc. Ec,FC = particular functions at C KB = ir2Cc - \\rpEc)/p2
p, a = spacing ofgrid lines r - p/q = grid ratio
G= given forcing function Gc = valu of(7 at C (h) Modified difference equation is required in cases where the
grid lines and the boundary ofthe regin R do not intersect at
(b) General form ofthe cartesian elliptic equation in two dimensions (16.01</) is the grid joints. In such cases, depicted by the adjacent graph,
the stiflhess factors of the molecule (/) become v y.
| AU
AU, + CU 4- DU, + EU, + FU = G
,' *
where U, U, 7,, U, are the partial dcrivatives of Uand .4 > 0, C> 0, F s 0 so that the i y
elliptic condition is'satisficd. As such, this cquation states always aboundary-value problem, and
its boundary conditions must be one of the following types.
(C) First boundary condition called the Dirichlet condition prescribes the valu of/on the boundarics
as
A 2CC 4- qxEc
U* = <J>(x, ,y.) ?a(i + ?-')
A,. 2AC - p2Dc n _2Af+pxDc
where x, ,y, are the coordinates ofthe boundaries and "* is continuous on S.
L PPx + />) * PPx + Pz)
(d) Second boundary condition called the Neumann condition prescribes the valu of the normal V: = Fe ~ TT Wc + ^i _ A..)^cJ [2Cf + (a, ~ 1*)Ec\
derivatives of U on the boundaries as P\Pl ?l?2
;,. 2Q - a2c
U* = d4>(x, ,y,)/dn
?.(?. + 1i)
where x, ,y,are the same as in (c) and 7* iscontinuous on 5.
wherepx ,q2,qx ,q2 are the new spacings shown in the figure.
(e) Third boundary condition called the mixed condition prescribes the valu
U* 4- aU* = 3<t>(x, ,y,)/dn + ar(x, ,y,)<t>ix, ,y.) (i) Modified forcing function adjusted to this new spacing is
derived by (16.16a1) to (16.16), and for the variable function
where the symbols are the same as in (c) and (</) and a(x,.v) > 0. G(x,y) the joint valu becomes
/ + Px12GLT + 2(px 4- p,)q:Gr + p2q2GRT N
(f) Governing difference equation at C(16.166) na molecule form ( +2(9l +q.,)pxGL +4(px 4- p.,)(qx 4- q:)Gc +2(a, +q2)p2GR
+ P\qxGLB + 2(px + p2)qlGB + p2qxGRB /
KTUT Gr- =
36(, 4- ^2)(a, + a,)
KLUL KcUc KrUr > = Gc
where the numerator of the fraction is an algcbraic sum in which GLB, 7^, GRB, . have the
KBUB same meaning as their counterparts in (16.16A).
applics at all points ofthe grid. The molecule which forms the left side othe equation represents
an algcbraic sum of producs of the stiflhess coefiicients K .KC,KR,... defin d in ig) on the (j) System matrix equation consisting of the diflerence equations written for all joints of the grid
opposite page and ofthe respective vales oUL,UCrUR defined in (a). Th< right side ofthe (including the boundary points) involves as manyunknown Uk as there are joints in the grid and
equation includes the joint valu Gc calculated by the formulas given in (16.16). yields the approximate vales of these unknowns as shown in (16.18) and (16.19).
348 16.18 LAPLACE'S EQUATION IN CARTESIAN Partial Differential Equations Partial Differential Equations 16.19 POISSON'S EQUATION IN CARTESIAN 349
SYSTEM, FINITE-DIFFERENCE SOLUTION SYSTEM, FINITE-DIFFERENCE SOLUTION
(a) Poisson's equation
(a) Laplace's equation in cartesian system
II 12 II i:. l IT S
N. ni 7 + 7 = -I
7 4- / = 0
.1.". * .i; s J-.i m 11
in the rectangular regin shown in the adjaccnt figure is
in the triangular regin shown in the adjaccnt figure is subject i*
the subject of the Dirichlet condition 2K Si l .11 .12 11 l
of the Dirichlet conditions
N IS li- \_i Uix,,y.) = 0 21 22 21 21 2."> 2ti 27
Uix,,y.) = x, -y,
where x, ,y, are coordinates of the boundaries. The grid
where x,,y, are the coordinates of the boundary. The grid of I) 1
of the rectangular regin is formed by p = a = l, and
H i;> li IT l> 1!' 211
(b) Intermedate point 15 is the only regular point. With .4 = C=I, D = E = F = G = 0, and only six vales of 7remain to be calculated.
r = pfq = | and KL = KB = KT = KR = 1, Kc = -4, the dierence equation at 15 by (16.17/)
(c) Intermedate points 8 to 10,16,17,24. With .4 = C = I, D = E = F = 0. G<: = \.r = p/q =
1 and //., /a, /Tc, /Tj-, /^ = I, Kc = 4, the respective diflerence equations (16.17/) become
Uxi /. tf.
-2 -4/IS t/.6 = 0 -4/8 -47 = -1
0 u9 /8 f.o
2 0 0
(C) Points 16,17 are near to the boundary with points 6 to 9 not being the points ofthe regular grid. u /.7
Thus thediflerence equations at these points must be modifico according to (16.17A). At 16, with u9 -4t/,o /9 -I w> -4/, /|7 = -I
Px = 2,p2= \,qx = 2,q,= I, /9
0
Kl = AT, = h K% = K*6 =\ Kc = K*6 = -2 K% = K* = { K'r = K* =\
and at 17 with the same p's and a's, /24 /,7
2 -4 2 /,6 1
(a) Notation. 7 = 4>ix,y) = solution of differential equation in regin R with boundary S. The (a) Wave equation in cartesian system I
symbols used below and on the opposite pageare those defined in (16.17a). 06 16 26 56 46 .56 .66
t/--5/ = 0 05 15 25 ,55 45 55 , 65
(b) General form ofthe cartesian hyperbolic equation in two dimensions (16.01*/) is
04 14 ,24 54 44 54 64
in the rectangular regin shown in the adjacent figure
AU - CUV 4- DU, 4- EU, + F/ = G [0 : x S a, / & 0] is subject to the conditions 05 ,15 25 55 45 55 65
where U,A, U, U,, U, are the partial derivatives of ". The solution of this equation is not 7(0,/) = /(a,/) = 0 02 ,12 ,22 .52 ,42 52 62
restricted to a closed regin and frequcntly is allowed to propgate in one or two directions with
given boundary conditions. 7(x, 0) = sin Jtx 7,(x,0) = 0 ,01 , ,11 21 SI .4! 51 61
KT = i-r2Cc + l2rpEc)fp2
(b) Governing difference equation in molecule form is
A', = (At. - lpDc)/p* Kc = i-2Ac - 2R2CC 4- p2Fcx;p2 Kr = iAc + hpDc)/p*
-XUT
KB = i~r2Cc - l2rpEc)/p2 UR = 0
uL -sUc
where r = p/q. -A/
u = u,
10 II 12 IJ
where U ,U,,U, are the partial derivatives of 7. The solution of this equation is not restricted 7(x, 0) = sin nx
to a closed regin and frequently is allowed to propgate in one or two directions with boundary
conditions given.
where / = time (s) and a = length (m). The grid is formed by four vertical and three horizontal
lines with p = 5 and a = .
(c) Governing difference equation at C in molecule form mav be one of the following types: (b) System matrix equation of the first type (16.22a*-1) with a = 0.25 is
f7nl (-0.5 0.25ir/011 = r0.649 521
(1) Forward difference form
L/I2J Lo.25 0.5 JL/02J L0.63952J
where 70l = sinn/3 and Um = sin 2^/3. Similarly,
-prEcUT
r/21j = (-0.5 0.25-1^,,-J _ T0.487 14]
(Ac - hpDc)Ut. -2AC - prEc - p2Fc)Uc A + \pDc)UR = G(:p L/22J LO.25 0.5 Jl/l2J Lo.487 14J
where 7,, , UX2 are the vales calculated by the preceding matrix equation.
(2) Backward difference form
iAc-hpDc)UL -(2AC 4- prEc - p2Fc)Uc --A - hf>Dc)UR = Gcp2 (c) System matrix equation of thesecond type (l6.22a*-2) with a = 4 is
prEcUB r/01j =r 1.5 -0.25-1^,,-j r/u] _ rO.685 71 0.1l4 29-|r7lll] ro.692821
lUj L-0.25 1.5 iuj lux2\ L0.I1429 0.6875lJ[/02J =lo.69282J
(3) Average difference form where 70l, 7OT are thesame as in (b) above. Similarly,
(d) System matrix equation ofthe third type (16.22a*-3) with a = 0.25, p = 4 is
(d) Reduced equations for the special caseA = 1, D = F = 'j. and E = c~'2 become in terms of
r 5 -o.5-ir/i = 1-3 0.51f/lol
a = c2/pr B = pr/c2 r = p/q
L-0.5 5 JL/I2J Lo.5 3 lUj
(1) UT = aUL + (1 - 2a)Uc 4- aUR - aGc from which by inversin
(2) U = -aUL + (1 + 2a)Uc - ccUR + aG, r/n-| _ fO.616 16 0.161 61 irU0X-] = fO.673 341
L7I2J ~ L0.I6161 0.616 I6JL70,J [o.67334J
(3) -Wu- + (I + P)Ut - kuRr = hVL 4- (p - l)Uc + WR-- pGcq
where Uox, Um are the same as in the preceding cases (b) and (c). By the same procedure
The construction ofthe system matrix equations based on these relations is shown n (16.23) on L'2X = U = 0.523 70. v
the opposite page.
(e) Exact solution of this equation is U(x.t) = exp (-Jt2t) sin jcx and Uxx = /,.. = 0.658 36
(TT3ad~vafu"eS m cases (c) and (d) are calculated~by formula given in (16.16). 72f=/22 = 0.50050, which calis for a fincTgiTd lh (A) t^a*^ " - -
1 .1 "1
354 16.24 BIHARMONIC EQUATION IN CARTESIAN Partial Differential Equations Partial Differential Equations 16.25 BIHARMONIC EQUATION IN CARTESIAN 355
SYSTEM, FINITE-DIFFERENCE RELATIONS
SYSTEM, FINITE-DIFFERENCE SOLUTION
(a) Approximate partial derivativos of 7 = 4>(x,^)
at the point C of the adjacent grid can be found
(a) Notation. 7 = <t>(x,y) = solution ofdiflerential cquation in regin R
by exprcssing each central diflerence in terms of Uk = valuof 7 at * Kk = stiflhess factorat k
the vales of 7corresponding to the nodes of this
grid. The equidistant spacings are the same as A,B,C, D,E,F,G = given functions Ac,Bc,Cc, Dc,Ec,Fc = particular fijnctions at C
before, and only the central diflerences of the
respective orders are considerad in the ap
r = p/q = grid ratio parameter Gc = forcing function at C
proximations listed in (b) below, where the
subscripts define the location of 7. These rela
tions apply to all nodes of the grid including the (b) General form of thecartesian bharmonic equation in two dimensions is
points on the boundary. Their utilization in the
approximation of the bharmonic cquation is AU,,,, 4- BU 4- CU + DU + EL' + FU = G
shown on the opposite page in (16.25).
where 7, U,^,... are the partial derivatives of 7 and A, B Gare functions in x,y or
constants.
uB - u, KltVlt
K-ltUlt KtUt
KTUT KrA'rt
*Rr*- rt
dx \c ~ Ip l 3* le 2V K,j.Vu. Wl KCUC k/)'- R KrrUrr >= Gr
^LO^LX KBUB ^Rfi' RB
rf| -."t- Wc + <-* KbbUb
P1
\32U] = U*r ~ Vl.T " U** +Uu applies at all points ofthe grid defined in 116.24a). The molecule which forms the left sideof this
equation represents an algcbraic sum of producs of the stiflhess coefficients Ku_ ,KL,... defined
in id) below and ofthe respective vales of7. The right side is the forcing function at C.
\?U] _ UKK - 2VK 4- IV,. - Uu. C/] Urr - 2UT + 2L - U,
[dx3\cr 2p> IV Je V
f M] __ U*r - 2> + ULT - U + 2UB - l\ (d) Stiffness factors in (c) are
[dx'dylc 2p*q
rjV] URT - 2VK +uRB-i Kj-r r*Cc/p4
[dxdy2\c 2pq2
Klt ~ r'Bc/p4 KT = (-2r2Z?c - Ar*Cc 4- r2Ec)/p* KRT r Bc/pA
T^/l URR - WR + 6UC - W,. + Vu [?in Utt- Wt + U'c - U' 4- UBB */./. = Ac/P* KL = (-4ilc - 2rB, + Dc)/pA = KR Krr = Ac/p*
Ur4Jc q*
Kc = (6/lc 4- 4r2fic 4- 6r4Cc - 2DC - 2r2c + n/p*
f &U] - V*RT - WrT - U**B + W*B ~ Uu.T + U;.T + UUJt - ULB
A K-LB rBc/p4 KB = (-2r2fic - 4r4Ct- 4- r2Ec)/p4 krb T'Bc/p*
_ UIT - 1LT + URT - 2U,. + AUC - 2LR + l,* - 2UB 4- URB Kbb = r'CJp*
pW
356 16.26 FINITE-DIFFERENCE METHOD IN Partial Differential Equations Partial Differential Equations 16.27 POISSON'S EQUATION IN POLAR SYSTEM 357
POLAR SYSTEM FINITE-DIFFERENCE SOLUTION
(a) Solution of a polar second-order partial (a) Notation. U = <t>(r, 0) = solution ofdiflerential equation in regin R. The symbols used below
diflerential equation by the finite-diflerence method are thosc defined in (16.17a) with modifications introduced on the opposite page. Also, r is the
raquires first the replaccment of the regin of radius ofthe cirele ofthe central point C.
integration by a polar grid of m radial lines and n
concentric circlcs whosc points of intersection are (b) General form ofPoisson's equation 116.19) in polar coordinates is
the nodal points (pivotal points, joints) shown in the
adjaccnt figure, where cach point i,j is given by 1
7 4-
r*
Uno + l-u.
r
= G
6, = ip (i o 0, l,...,m)
'j = 11 0 = 0,1,...,*) where / , Un , U, are the respective partial dcrivatives of /and Gis the forcing function in r, 0.
and p, q are the angular and linear spacngs of the (c) Governing difference equation at Cin molecule form s formally identieal to (16.17/) and may be
grid lines and circlcs, respcctively. applied similarly. The stifness factors used n its molecule are
.. _ 1 + 1/27
r
, 1 - 1/2/
which is the solution ofthe partial diflerential equation. are next expressed by the finite-diflerence (d) Polar grid introduced in (I6.26a) acates aspecial condition at the center 0, where the governing
approximations (3.02a*) related to the particular points. If the point C shown in the adjacent difference equation has the following special form (Refs. 16.08, 16.21):
graph is that point, then the partial derivatives of <t> in difference form become
30 " 'Ip
** = q> s a>7 ~ fl
di iq
(a) Consistency. If in the limit p -* 0, a * 0 the diflerence and diflerential equation become
equivalcnt, thenthediflerence equation is saidto beconsistent with thediflerential cquation.
(2) Stability. If the errors at one stage of calculations remain finitc all the time (do not cause
larger errors as the computation contines), then the finite-diflerence method is said to be
stable.
(3) Convergence. If the results of the finite-diflerence method approach the truc vales as both p
and a approach zero, then the finite-diflerence method is said to be convergent.
Appendix A
(b) Elliptic equation. Since the finite-diflerence solution of this cquation (16.17) is based on the
solution of finite linear system of equations, no stability problem is present. As almost exclusvely
somc iterative method is used, only the convergence of this method is required. The local
discretization error is of order 0(p2 4- a2).
NUMERICAL
(C) HyperboiiC equation. Since the finite-diflerence solution of this cquation (16.20) is based upon
the transport chain, only the stability may become a problem. The procedure is stable if Ain
(16.216) is equal to or less than 1. The local discre2aiion error is oforder Oip'2 + a2). TABLES
(d) Parabolic equation. Solution of (16.22) by the forward diflerence method (explicit method)
requires for convergence and stability that a given in (16.22a*) be equal to or less than |, which
implies small a and necessitates an enormous amount of numerical work. In turn the backward
difference method (implcit mcthod) is unconditionallv stableand convergent. The only limitation
is the size op and a since the discretization error is oforder Oip2 4- a). The desire to climinate
small a's led to the dcvelopmcnt of the average dhfercnce method (Crank-Nicholson method),
which is not only uncondtionally stable and convereent but since Oip2 4- q2), it allows a to beof
the same magntude as p.
Pl
5* pl in r. Pl in rN cn
- n S Pl Pl O
+ !>
in Pl In Cf. - f n a i
h - 5l Pl l> O - _
Pl 2; s * f iO o <A iO tC
Cl Cl Cl .t T <*. O
00 00 00 00 00 00 00 09 00
io io in tn >n lO o S is n r t^ i^. r^. r
00 00 03 00 00 00 SO 00 00 00
oooooo ao co oo ce co
3
C
a in pi in en n
S SsSo
n 55 - in O M ffl 51 in en Pl
n rs. 5; g 2 eo i n n Pl 01 pi r* . Pl IN. |N. O) Pl M 01 fN IN
in in oo co oo
82 = cn m
N N
n eo n
N
*
fl
o e in to 91 en en m xi 10 ie in eo
00 03 00 00 00 03 00 00 00 00 N W N N N ni e>< pi pi pi pi pi pi pj pi
v> 00 00 00 00 00 00|CO 00 00 00 00 00 00 00 00 00 00 03 00 00 00 00 00 00 00
oc
IN IN Pl Ol p in en 01 Pl in pi 51 M Pl IN Cl Pl 01 cn pi
O CM 61 Pl f ^- ^* ul t O) Pl rN pi pi n en o pi rN
in pi
n o Pl |n Pl |N
ul m ui ui 1 ul ul O 0 ein* * tp in 01 pi in en
n |N
|N r< ^
|N .l: l: /i
L.- ui
rN|N|N|N|N
ui
i/ - IK)
in pn S
in S
in fl
p
Si O
(O tO IN
O -
|N |N
w ci ** ui
rNrNrNiNrN
ui io
CNrNooooco
CM Cl -"
eo 00
ul to
oo
rN r
oo
IN |N fN r* |N fN pN |N in pn in in r, iNiNrNrNfN R^"" co 00 en en en
p? ~ pi en in en co r*. o> c co 8 in en en tN in pi in pi cn cn fN
r en pi
8000S
r* fN. r* r* r*
S88S2 es cm CM
r in in r
cm Ul
ui ui in 00 en o ei pi "" ui 00 cn o o CN Pl Pl
01
* m
01 pi
i s ffl -
in
pi in cn
pi in ui ui
IN p.
r- |N |N |N |N ~~" " IN fN fN fN.
RRRRf! S*tg gggg pi pi pi pi pi
ggggff RSSSg ' -r -r
In |N |N S fN |N fN |N
N Cl in O cn en pi
^ 00 en o Pl Ul Ul (O IN Ci Cl in n o ~ pi Pl Pl IN O Pl IN pi o _
Sr- 9 en o o O In - In in 01 Cl IN
1/1 ui ui to to S t o tp tp to tp tp in P nInnnSn O In SO Cl Cl o cn ni ei pi f i to to |n Cl O -1" "t- ui <p tp
IN Pl |N
o S o 10 to uo o <o to (O UO to to c n en cn en cn ci en cn en en ci en ci
to to to (JO to to to to to to to to to
*" 3 3 |N r- in
ul
n
to to
o
rN Cl o
r- pi 01 rN Pl Pl O ^ fN f> Cl Cl Pl Cl IN
TZ 72 CN CM CM CM CM Cl Pl "* <n ui to to
3 3 3 3 to w q ev w CM CM CM Cl Pl pi pi pi pi pi S S 5 "1 ; "* 2 O
o1 N N * ul ul to to
to to to to to 5 Es o o 2 to to to to to
pi pi pi pi pi
tototototo tototototo to to to to nS SnSSSnS nSS SU!
tO
ul
tO
ul
tO
Pl Cl Cl Pl - n n - r. O pi cn fN
-t *t- *t in 00 Cl rs i S P1
In |N In P: p. Q Cl ni Pl f f ul ul t to 'O In r:
flo ;n <n
ai ai .*
r. 5*
o> Cl in pi in pi rN pi ai pi > pi
o in in n ii o un ir> ,r, ,,*, 00 00
ul .11 .1*1 i ,r>
00 00 00 00 SO 03
<r> .f> ^ ,n .7S
00 OS O O O O S Si 5 i j o in in fe en o ci pi pi
.O .o r ir, i#l n ir> .o o ir,
3 333 28 3 3 8 3 3 O 'O -O 'O >o
rN pi en pi pi rN _. . in cn . rN M e^
00 en o o ni Pl ^* ul 03 00 cn rN en pi pi rN en In - IN P) O - in n in
Ni M Pl Pl Pl Pl Pl Pl Pl pl pi pi * 5- in 00 oo-w CM Pl Ul tO tO tN 00 O NJ Pl
ui ui ul ui ui ul m ui ui u-> Ul Ul Ul Ul Ul Ul ul Ul Ul Ul
"***
* *
Ul Ul Ul ui
<?
Ul ui
Ul
ul Ul ul ul ul
Ul ul ul ui ul
Ul Ul Ul Ul I?) Ul Ul Ul to to
* * m m o
tO to to to to
ui Ul ul Ul Ul un Ul Ul Ul Ul ul ul ul ul ul ul
cn pi in Pl -" In IN O Pl en 01 in rN in pi en IN M IN O)
^* Ul Ul tp tp O IN C| |N .
SI 5I cn cn cn en cn
* * * *
Ul Ul Ul Ul ul
N5 N Pl ui ui in ep co
O
Ul
Q O 3 3
ul Ul Ul Ul ul
o
ui
o
ul
ul ul
*
ul ul
ui to in
ui ui ui
rN 00 O O
M
M NI
Ul Ul Ul ul
cn
Cl
Ul
52
| CM
ul ui ul ui ui
Pl _ CN _ MI - (1 Pl in cn in pi
N gt
CM CM oUl itoooooc o22S "t % tp 10 o?
IN pi Cl Pl Cl " IN pi pi en pi
?$ii? l l i ?5?s ui ul ul ui i FF2 H^ff tpiNiNcno cm M M Pl O
CM
IN |N |N |N |N
Pl IN _
pl
^ Pl IN Pl O - IN O - IN Pl Cl - Pl Cl rN 01 Pl Cl Pl rN 01 rN
f$2 CJJMiiQui in CM CM Pl Pl - in C O l - " Pl Pl O IN in rN cn en in Pl Pl |N
ul ui ui in o CM Pl * * Ul O tO
M CM M CM M
in in 00 00 cn CM Pl Pl Ul tO In O O
CM CM CM CM CM CM CM M CM fl pi pi pi Pl
pi Pl
pi Pl Pl Pl Pl
psi
^ co 00 r^ co CO fN O Pl IN r+ o 9* 1^* co 91 ^ r* 9i 9* co f* o*) ^ co co r*N i_* *{q co o. ct> r^
SCO 91 o -" CM Pl Pl - r O M 01 O r- t*N o*
in o 3 p tO tO to tp
ui in in in en Ci O O
O cn rs
-W
I*. fN
co co to to ^o
r in f* r* r
Cn O C> Q M cm co j m jft Sin m g - - - WN f j w 6
co co co co 60 Pl Pl pl p co
S8 Sc o !S S n n r% 5 eo 00 a 00 <7> 9) 9> 9> 9) 9* 9* 9* 9* 9 9>
CO co co co co co co co co co co co co co co co co co co CO CO co co co co CO CO CO CO CO co co co
cn - en cj rs O- n n Cn C7> rN. CO 9 CO 9
si 00 9tp O - co tN. ai CO 9i Cn CO rN o ci rN in 01 Pl Cl
CM CM *0 *0 iO O tN, 9 O O CM CM co *** n t i*, n 00 9 O in in en
co
CM CM CM
co co co co
CM
CO
CM CM
co co
CM CM
co co
CM CM CM CO CO co co co co co co co co co co co co co * ^* 3
^
1 S2
-F T
2
.9" *1*
tO tO Cl Cl
* * * *
Ifl
Ul
CM fl Pl Pl
ul Ul Ul Ul
*
Ul
*
Ul Ul
in in
Ul Ul
h
CO CO CO CO CO co co co co co co co co co co co co co co co Pl Pl Pl Pl pl pl Pl Pl Pl pl Pl Pl Pl Pl Pl Pl Pl Pl pl Pl
O Pl IN IN O rN pi o Ol IN |N Pl in in en pi Cl Pl IN 01 r~ 01 Pl 01 O c
"f ul to in eo O O o o Sin 00 cn o CM Pl Ul rN pi in en
ininininin in r. n eo 09 > CM Pl
CM CM CM CM CM CM CM CM CM M
00 00 03
CM CM CM
00 00
CM CM
00
C-l CM
00 00
CM CM
01
CM
o en o ci
CM CM CM CM
Si 8pl_ o
Pl
o
Pl
o
Pl
o
Pl
sssss 00 00 O
o o
Cl
Pl tO tO tO 03
Pl Pl Pl pl pl pl Pl Pl pl Pl Pl Pl Pl Pl Pl
in _ pi en pi en > in pi in . in 01 rN Pl rN pi -. _
f> in 00 00 00 Pl In in Ol Pl IN Pl |N
en 01 cm Pl -f Ul tp rN in q M Pl ul ul in 01 O Q Cl Pl t Ul Ul
pl pl
M Cl
Pl Pl
Cl Cl
Pl
Cl
pi pi f "t- f * -f -f ^ C -f 1 ul ui Ul ll Ul Ul Ul 9 O to tO to
co en Ci o
tO tO tO In In
|N
IN
|N
S |N
pi |N
-f
Cl Cl Cl Cl ci ci ci ci ci ci ci ci ci ci ci Cl Cl CM CM Cl Cl Cl Cl CM CM M Cl Cl Cl Cl Cl Cl Cl Cl Cl Cl CM Cl CM
p
Pl IN Ol Pl rN rN o O Pl Pl Cl - Cl IN p pi Cl Pl rN pi IN O Pl in en pi in Cl n |n
Cl CM Pl MI
Cl O Ol O o
Cl Cl Cl Cl ci
o p p SSspSo ss::2 22222 to
in
o
Cl
Cl
Cl
Cl Pl Pl -f 1 to 10 in
ci ci ci
00
ci
co
ci
en O
Cl Cl Ti
Pl
Pl
*
Pl
f
Pl
Ul
Pl
Ul
Pl
O IN |N |N
r^> Pl IN O. Pl IN Pl IN O. "- Pl In in
ci cm pi un tp tp tp O Ol C. Cl Cl Pl f Ul ul in O Pl C. In
S S S 3 to to to to to tO (O tO to to 'O IN In |n |n |n ,n |n |n |n
00 00 00 fN. rN
CO 00
i-* r>. 00
00 00 co
0
9>
0
0%
-<
91
eo co
9* 9
f i in in eo
c. Cl O Cl Cl
In 01 Ci IN Ol Pl Ol In _ pi in o
Pl Pl -f Ul IN fN 03 O O CM Cl tO tO In O Q Cl Cl ui
pi Pl O Pl O <n . O Pl
CM CM CM CM CM pi pi -r 00 03 en 01 Cl Pl f -f Ul Ul tp IN Pn 00
Cl CM M CM CM CM Pl Pl Pl Pl Pl Pl Pl Pl Pl -f f -f r -r f f .o Ul Ul Ul Ul Ul
f^i ul in ui ui ul
r* pi rN pn en 01 in IN Pl |N .. pN Pl IN
ci pi * * ul to rN p.
iNcnpi 01 pi en en pi 01 in pi rN p) ciiNpien~ m m -. n Pl ^ Pl IN pl
; 00
S S
00 9
OO SB
00 O Cl Cl O O Cl Cl O O O Cl S8S 3SSSS SS88 gS2 2=222 22^22 en
CM
CM
CM
cm
M
LU
IN IN pi O in rN pi gi tN pi rN 01 -" Pl IN Pl Ol Pl IN pl ci o rN Ol Pl IN O Pl IN IN
MC Ul tO IP IN r> 00 O Ol o S pi S* + in m
Pl in O O Pl In O Pl
Ul Ul Ul Ul Ul S I N In 03 OI O CM Pl Tf Ul Ul IO tO In 03 O CM CM CM M Pl Ul Ul Ul to
ul 10 Ul ul to tO to to to to to to to to to to to 03 03 03 00 00 00 00 00 00 co
UJ Pl Ol IN pi en in Pl Pl |N n p^ fN .. In
Pl Pl f Ul Ul Pl Cl In Pl O Pl O in Ol
IO tO IN In 00 00 en pi
pi pi f -f lO 1 tO In |n 00 00 O o o CM Pl Pl Pl
12 5 i $
tOlNOOOCl QCICM1-
ce
CM CM CM CM CM CM CM Cl CM CM CM CM Pl Pl Pl pi pi pi pi pi Pl Pl Pl Pl Pl n tn n *r *r -r -f -f uiuiuiuiui
u
CL
SO
CM Pl ul IN
"* Pl rN 01 Pl c i IN Pl |N Pl o rN Pl C Pl O In --PIINOPI r in O O ^INplrNp^ Cl Pl in O Pl
CM M en n * * * ulultOtOIN N in X GO Cl O CMPIPIPI^- UlultOtOIN in 00 Ci S Cl Ol ci
N CN S
CM CM CM CM
p*i
r^i ^ -^j f==j r^
96 4656 299 536 21678 336 1673 507632 134 571883 776
46 1081 33 511 1 168 561 43 463 767 1683896401
97 4753 308945 22591009 1762036913 143159224033
47 1128 35 720 ! 272 384 48 343448 1913241408
98 4851 1854273 729 452198432001
48 1176 38024 J8976 53651864 2168045376
99 4950 328 350 24 502 500 1950333 330 161708332500
49 1225 40 425 500625 59 416 665 2 450520625
100 5050 338 350 25 502 500 2050333 330 171708 332500
50 1275 42925 i625625 65 666665 2 763020625
Numerical Tables A.05 DIGAMMA AND TRIGAMMA FUNCTIONS (Continued) 367
366 A.05 DIGAMMA AND TRIGAMMA FUNCTIONS Numerical Tables
dxfi(x)/dx = triiramma function (2.34a) V'(-v) = digamma function (2.32a) d\i(x)/dx = trgamma function (2.34a)
y(x) = ligamma function (2.32a)
X V'(<) dq>(x)/dx X
VW dV'(x)/dx
X VW dq>(x)/dx X H>(x) dq<(x)/dx
0.500 0.036 489 974 0.934 802 201 0.750 0.247 472 454 0.764 101 870
0.000 -0.577 215 665 1.644934 067 0.250 -0.227 453 533 1.197 329 155
0.505 0.041 153 654 0.930 675 759 0.755 0.251 285 956 0.761 302 377
0.005 -0.569 020 911 1.632994 157 0.255 -0.221 183 427 1.190 725 058
0.510 0.045 796 790 0.926 584 114 0.760 0.255 085 510 0.758 522 687
0.010 -0.560 885 458 1.621 213 528 0.260 -0.215 546 169 1.184 189 480
0.515 0.050 119 553 0.922 526 843 0.765 0.258 871 215 0.755 762 595
0.015 -0.552 808 516 1.609 589 182 0.265 -0.209 64! 419 1.177 721 403
0.520 0.055 022 115 0.918 503 527 0.770 0.262 643 169 0.753 021 900
0.020 -0.514 789 311 1.598 118 192 0.270 -0.203 76R 844 1.171 319830
0.525 0.059 604 644 0.914 513 755 0.775 0.266 401 466 0.750 300 404
0.025 -0.536 827 083 1.586 797 699 0.275 -0.197 928 112 1.164 983 782
0.530 0.064 167 307 0.910 557 125 0.780 0.270 146 204 0.747 597 911
0.030 -0.528 921087 1.575 624 915 0.280 -0.192 118 898 1.158712299
0.535 0.068 710 270 0.906 633 236 0.785 0.273 877 477 0.744 914 227
0.035 -0.521 070 592 1.564 597 116 0.285 -0.186 340 883 1.152 504 439
0.540 0.073 233694 0.902 741 698 0.790 0.277 595 378 0.742 249 162
0.040 -0.513274879 1.553 711643 0.290 -0.180 593 749 1.146 359 276
0.545 0.077 737 740 0.898 882 125 0.795 0.281 299 999 0.739 602 527
0.045 -0.505 533 243 1.542965 897 0.295 -0.174 877 187 1.140 275 903
0.550 0.082 222 568 0.895 054 137 0.800 0.284 991 133 0.736 974 138
0.050 -0.497 844 991 1.532 357 342 0.300 -0.169 190 889 1.134 253 435
0.555 0.086 688 333 0.891 257 360 0.805 0.288 669 771 0.734 363 809
0.055 -0.190 209 445 1.521883 500 0.305 -0.163 534 553 1.128 290 992
0.560 0.091 135 193 0.887 491 425 0.810 0.292 335 101 0.731 771 362
0.060 -0.482 625 936 1.511541 950 0.310 -0.157 907 880 1.122 387718
0.565 0.095 563 298 0.883 755 970 0.815 0.295 987 514 0.729 196 617
0.065 -0.175 093 809 1.501 330 326 0.315 -0.152 310 578 1.116542 771
0.570 0.099 972 802 0.880 050 638 0.820 0.299 627 097 0.726 639 397
0.070 -0.467612420 1.491 246 316 0.320 -0.146 742 357 1.110 755 325
0.575 0.104 363 854 0.876 375 077 0.825 0.303 253 937 0.724 099 530
0.075 -0.460 181 137 1.481 287 662 0.325 -0.141 202 931 1.105 024 568
0.580 0.108 736 602 0.872 728 940 0.830 0.306 868 121 0.721 576 843
0.080 -0.452 799 338 1.471 452 156 0.330 -0.135 692 018 1.099 349 704
0.585 0.113091 193 0.869 111 887 0.835 0.310 469 734 0.719 071 166
0.085 -0.445 466 414 1.461 737 638 0.335 -0.130 209 342 1.093 729 950
0.590 0.117 427 769 0.865 523 582 0.840 0.314 058 860 0.716 582 333
0.090 -0.438 181 761 1.452 141 999 0.340 -0.124 751628 1.088 164 538
0.595 0.121 746 475 0.861 963 692 0.845 0.317 635 585 0.714 110 179
0.095 -0.430 944 799 1.442 663 176 0.345 -0.119 327 607 1.082 652 714
0.600 0.126 047 453 0.858 431 893 0.850 0.321 199 999 0.711 654 540
0.100 -0.423 754 940 1.433299 151 0.350 -0.113928013 1.077 193 736
0.605 0.130 330 841 0.854 927 863 0.855 0.324 752 157 0.709 215 255
0.105 -0.416611 619 1.424 047 951 0.355 -0.108 555 583 1.071 786 877
0.610 0.134 596 777 0.851 451 286 0.860 0.328 292 169 0.706 792 165
0.110 -0.409 514 276 1.414 907 648 0.360 -0.103 210058 1.066 431 423
0.615 0.138 845 399 0.848 001 849 0.865 0.331 820 106 0.704 385 114
0.115 -0.402 462 361 1.405 876 354 0.365 -0.097 891 184 1.061 126 667
0.620 0.143 076 840 0.844 579 246 0.870 0.335 336 047 0.701 993 946
0.120 -0.395455334 1.396952 221 0.370 -0.092 598 708 1.055 871 929
0.625 0.147 291 235 0.841 183 173 0.875 0.338 840 071 0.699 618 509
0.125 -0.388 192 663 1.388 133 445 0.375 -0.087 332 383 1.050 666 522
0.630 0.151 488 716 0.837 813 333 0.880 0.342 332 258 0.697 258 651
0.130 -0.381 573 827 1.379 418 257 0.380 -0.082 091 962 1.045 509 783
0.635 0.155 669 412 0.834 469 432 0.885 0.345 812 684 0.694 914 224
0.135 -0.374 698 311 1.370 801929 0.385 -0.076 877 205 1.040 401 058
0.640 0.159 833 453 0.831 151 179 0.890 0.349 281 426 0.692 585 079
0.140 -0.367 865 611 1.362291 767 0.390 -0.071 687 872 1.035 339 701
0.645 0.163 980 966 0.827 858 290 0.895 0.352 738 560 0.690 271 072
0.145 -0.361075 229 1.353 877 115 0.395 -0.066 523 730 1.030 325 088
0.650 0.168 112 078 0.824 590 483 0.900 0.356 184 161 0.687 972 058
0.150 -0.354 326 678 1.345 559 352 0.400 -0.061 384 545 1.025 356 591
0.655 0.172 226 912 0.821 347 480 0.905 0.359 618 305 0.685 687 897
0.155 -0.347619477 1.337 336 890 0.405 -0.056 270 088 1.020 433 600
0.660 0.176 325 593 0.818 129 009 0.910 0.363 041065 0.683 418 447
0.160 -0.340 953 153 1.329 208 175 0.410 -0.051 180 134 1.015555517
0.665 0.180 408 243 0.814 934 800 0.915 0.366 452 514 0.681 163 570
0.165 -0.334 327 241 1.321 171 686 0.415 -0.046 114 459 1.010 721 752
0.670 0.184 474 981 0.811 764 588 0.920 0.369 852 724 0.678 923 129
0.170 -0.327 741 285 1.313225 932 0.420 -0.041072 843 1.005 931 724
0.675 0.188 525 928 0.808 618 109 0.925 0.373 241 769 0.676 696 990
0.175 -0.321 194 833 1.305369 455 0.425 -0.036 055 070 1.001 184 864
0.680 0.192 561 202 0.805 495 108 0.930 0.376619 718 0.674 485 019
0.180 -0.314687444 1.297600825 0.430 -0.031060 924 0.996 480 611
0.685 0.196 580 918 0.802 395 328 0.935 0.379 986 642 0.672 287 085
0.185 -0.308 218 681 1.289 918 642 0.435 -0.026 090 194 0.991818415
0.690 0.200 585 193 0.799 318 520 0.940 0.383 342 612 0.670 103 056
0.190 -0.301 788 116 1.282 321 536 0.440 -0.021 112 670 0.987 197 733
0.695 0.204 574 141 0.796 264 435 0.9-15 0.386 687 696 0.667 932 804
0.195 -0.295 395 326 1.274 808 162 0.445 -0.016218 148 0.982 618 032
0.700 0.208 547 875 0.793 232 830 0.950 0.390 021963 0.665 776 203
0.200 -0.289 039 897 1.267377 205 0.450 -0.011 316 423 0.978 078 789
0.705 0.212 506 506 0.790 223 464 0.955 0.393 345 481 0.663 633 127
0.205 -0.282 721 419 1.260027 376 0.455 -0.006 437 293 0.973 579 487
0.710 0.216 450 146 0.787 236 101 0.960 0.396 658 316 0.661 503 451
0.210 -0.276 439 490 1.252 757 400 0.460 -0.001 580 562 0.969 119 622
0.715 0.220 378 904 0.784 270 506 0.965 0.399 960 537 0.659 387 054
0.215 -0.270 193711 1.245 566 061 0.465 +0.003 253 968 0.964 698 692
0.720 0.224 292 887 0.781 326 449 0.970 0.403 252 209 0.657 283 813
0.220 -0.263 983 700 1.238 452 130 0.470 0.008 066 489 0.960 316 209
0.725 0.228 192 204 0.778 403 701 0.975 0.406 533 397 0.655 193 610
0.225 -0.257 809 065 1.231 414 426 0.475 0.012 857 193 0.955971 690
0.730 0.232 076 959 0.775 502 040 0.980 0.409 804 166 0.653 116 327
0.230 -0.251 669 131 1.224 451 770 0.480 0.017 626 268 0.951 664 659
0.735 0.235 947 259 0.772 621242 0.985 0.413 064 582 0.651 051 845
0.235 -0.245 564 424 1.217 563 025 0.485 0.022 373 901 0.947 394 651
0.74O 0.239 803 206 0.769 761 092 0.990 0.416 314 706 li IvlQ H0O l\\ X
0.240 -0.239 493 679 1.210747071 0.490 0.027 100 276 0.943 161 205
0.245 -0.233 456 834 1.204 002 806 0.495 0.031 805 574 0.938 963 870 | 0.745 0.243 644 904 0.766 921 371 0.995 0.419 554 603 | 0.646 960 829
F~~T "T 1 ~1 i i "~~T
364 A.03 FACTORIALS OF INTEGERS* Numerical Tables Numerical Tables A.04 GAMMA AND RELATED FUNCTIONS* 365
ni = n factorial (2.09a) n\\ = n double factorial (2.30a*) T(tt 4- 1) = a! = gamma function (2.28a*) ri() = a! = pi function (2.31a)
n a! !! n n! !! t>!
II u <! u o! i> ir!
1 1.000000000(00) 1.000000000(00) 51 1.551 118 753(066) 2.980227914(33) 0.005 0.997 138535 0.255 0.905385 766 0.505 0.886 398 974 0.755 0.920209222
2 2.000000000(00) 2.000000000(00) 52 8.065817 517(067) 2.706443 182(34) 0.010 0.994 325851 0.260 0.904 397 118 0.510 0.886591685 0.760 0.921374885
3 6.000000000(00) 3.000000000(00) 53 4.274883 284(069) 1.579520 794(35) 0.015 0.991 561 289 0.265 0.903 426 295 0.515 0.886804 980 0.765 0.922 559518
4 2.400000000(01) 8.000000000(00) 54 2.308 436973(071) 1.461479318(36) 0.020 0.988 844 203 0.270 0.902 503065 0.520 0.887038 783 0.770 0.923 763 128
5 1.200000000(02) 1.500000000(01) 35 1.269640335(073) 8.687 364 368 (36) 0.025 0.986173963 0.275 0.901 597 199 0.525 0.887 293 023 0.775 0.924985 721
6 7.200000000(02) 4.800000000(01) 56 7.109985878(074) 8.184 284 181(37) 0.030 0.983549951 0.280 0.900 718 477 0.530 0.887 567838 0.780 0.926227 306
7 5.040000000(03) 1.050000000(021 3/ 4.052691950(076) 4.951797 690(38) 0.035 0.980971561 0.285 0.899866677 0.535 0.887 862529 0.785 0.927 487 893
8 4.032000000(04) 3.840000000(02) 38 2.350561331(078) 4.746884 825(39) 0.040 0.978438201 0.290 0.899041586 0.540 0.888 177 659 0.790 0.928 767 490
9 3.628800000(05) 9.450000000(02) 39 1.386831 185(080) 2.921560637(40) 0.045 0.975949292 0.295 0.898 242995 0.545 0.888 512953 0.795 0.930066 112
10 3.628800000(06) 3.840000000(03) 60 8.320987 112(081) 2.848 130895(41) 0.050 0.973504 266 0.300 0.897 470696 0.550 0.888 868 348 0.800 0.931383 771
II 3.991680000(07) 1.039500000(04) 61 5.075802139(083) 1.782151989(42) 0.055 0.971 102566 0.305 0.896 724490 0.555 0.889 243 783 0.805 0.932 720 481
12 4.790016000(08) 4.608000000(04) 32 3.146997 326(085) 1.765 841 155(43) 0.060 0.968 743650 0.310 0.896004 177 0.560 0.889638 199 0.810 0.934 076259
13 6.227 020800(09) 1.351350000(05) 63 1.982608 315(087) 1.122 755 753(44) 0.065 0.966426982 0.315 0.895 309 564 0.565 0.890054 539 0.815 0.935 451 120
14 8.717 829 120(10) 6.451200 000(05) 64 1.268869 322(089) 1.130 138 339(45) 0.070 0.964 152043 0.320 0.894 640463 0.570 0.890489 746 0.820 0.936845 083
15 1.307674368(12) 2.027 025 000(061 63 8.247 650 562 (090) 7.297912 393(46) 0.075 0.961918319 0.325 0.893996687 0.575 0.890 944 769 0.825 0.938258 168
16 2.092278989(13) 1.032192000(07) 66 5.443 449 391 (092) 7.458913 039(46) 0.080 0.959 725 311 0.330 0.893 378054 0.580 0.891419554 0.830 0.939690395
17 3.556874281(14) 3.445942 500(07) 67 3.647 111 092 (094) 4.889601304(47) 0.085 0.957 572527 0.335 0.892 784 385 0.585 0.891914 052 0.835 0.941 141 786
18 6.402 373 706(15) 1.857 945600(08) 68 2.480035 542(096) 5.072 060 866 (48) 0.090 0.955459 488 0.340 0.892 215 507 0.590 0.892 428 214 0.840 0.942612 363
19 1.216451004(17) 6.547 290 750(08) 69 1.711224 524(098) 3.373824 899(49) 0.095 0.953 385 723 0.345 0.891 671 249 0.595 0.892961995 0.845 0.944 102 152
20 2.432902008(18) 3.715891200(09) "0 1.197 857 167(100) 3.550442606(50) 0.100 0.951350 770 0.350 0.891 151 442 0.600 0.893 515 349 0.850 0.945611 176
21 5.109094217(19) 1.374 931058(10) 1 8.504 785 855(101) 2.395 415 679(51) 0.105 0.949541 178 0.355 0.890655924 0.605 0.894 098234 0.855 0.947 129 464
22 1.124000 728(21) 8.174 960640(10' *2 6.123 445837(103) 2.556 318 677(52) 0.110 0.947 955 504 0.360 0.890 184 532 0.610 0.894 680609 0.860 0.948687042
23 2.585 201674(22) 3.162341432(11) "3 4.470 115461 (105) 1.748653 445(53) 0.115 0.945 474 315 0.365 0.889 737 112 0.615 0.895 292433 0.865 0.950 253 939
24 6.204 484 017(23) 1.961 990534 (12) *4 3.307 885 441 (107) 1.891675821 (54) 0.120 0.943 590 186 0.370 0.889313807 0.620 0.895 923 669 0.870 0.951 840 186
25 1.551 121004(25) 7.905853 581 (12) "5 2.480914 081 (109) 1.311490084(55) 0.125 0.941 742 700 0.375 0.888913 569 0.625 0.896 574 280 0.875 0.953 445 813
26 4.032 914611(26) 5.101 175439(13) "6 1.885494 702(111) 1.437 673 624(56) 0.130 0.939931450 0.380 0.888537 149 0.630 0.897 244 233 0.880 0.955070853
27 1.088886945(28) 2.134 580467(14) "7 1.451830920(113) 1.009847 365(57) 0.135 0.939138036 0.385 0.888 184 104 0.635 0.897 933 493 0.885 0.956 715 340
28 3.048883446(29) 1.428329123(15) 78 1.132 428 118(115) 1.121385427(58) 0.140 0.936416066 0.390 0.887 854 292 0.640 0.898 642 030 0.890 0.958 379 308
29 8.841 761 994 (30) 6.190283 354(15) "9 8.946182 130(116) 7.977 794 181 (58) 0.145 0.934 711 134 0.395 0.887 547 575 0.645 0.899 369814 0.895 0.96OO62 793
30 2.652528598(32) 4.284987 369(16) SO 7.156945 704(118) 8.971083 412(59) 0.150 0.933 CO 931 0.400 0.887 263818 0.650 0.900 116816 0.900 0.961 765832
31 8.222838654(33) 1.918987 840(17) SI 5.797 126020(120) 6.462013 287(60) 0.155 0.931405022 0.405 0.887002888 0.655 0.900983010 0.905
32 2.631308369(35) 1.371 195858(181 S2 4.753 643337(122) 7.356 228 389(61) 0.963488 463
0.160 0.929803 967 0.410 0.886 764658 0.660 0.901668 371 0.910
8.683 317 619(36) 6.332659 871(18) 3.945 523 970(124) 0.965 230 726
33 83 5.363 471028(62) 0.165 0.928 234 712 0.415 0.886548999 0.665 0.902472875 0.915 0.966992661
34 2.952 327990(38) 4.662066 258(19) S4 3.314 240134(126) 6.179282254(63)
1.033 314 797(40) 2.216430955(20) 2.817 104 114(128)
0.170 0.926699611 0.420 0.886355 790 0.670 0.903 2% 500 0.920 0.968 774 309 I
35 85 4.558950374(64) 0.175 0.925 197 423 0.425 0.886 184908 0.675 0.904 139 224 0.925 0.970 575 713
36 3.719933 268(41) 1.678343 853(21) 36 2.422 709638(130) 5.314 182 739(65) 0.180 0.923 727 814 0.430 0.886036236 0.680 0.905001030 0.930 0.972 396918
37 1.376 375309(43) 8.200 794 533 (21) 37 2.107 757 298(132) 3.966 286825(66) 0.185 0.922 290459 0.435 0.885909659 0.685 0.905 881900 0.935 0.974 237 967
38 5.230226175(44) 6.377 706640(22) S8 1.854 826422(134) 4.676480810(67) 0.190 0.920885037 0.440 0.885 805064 0.690 0.906 781816 0.940 0.976098908
39 2.039 788208(46) 3.198 309868(23) S9 1.650 795 516(136) 3.529995274(68) 0.195 0.919 511 234 0.445 0.885 722 340 0.695 0.907 700 765 0.945 0.977 979 786
40 8.159152832(47) 2.551082656(24) 90 1.485 715964(138) 4.208832 729(69) 0.200 0.918 168 742 0.450 0.885661380 0.700 0.908638 733 0.950 0.979 880651
41 3.345 252661 (49) 1.311307046(25) 91 1.352001528(140) 3.212 295 700(70) 0.205 0.916857 261 0.455 0.885622080 0.705 0.909 595 708 0.955 0.981 801 552
42 1.405006118(51) 1.071454 716(26) 92 1.243 841405(142) 3.872 126 111(71) 0.210 0.915 576493 0.460 385 604 336 0.710 0.910 571680 0.960
5.638620297(26) 0.983 742540
43 5.041526306(52) 93 1.156 772 507(144) 3.987 435001(72) 0.215 0.914 326150 0.465 '. 885608050 0.715 0.911566639 0.965 0.985 703 666
44 2.658271575(54) 4.714400 749(27) 94 1.087 366 157(146) 3.639 798544(73) 0.220 0.913 105948 0.470 0 385633 122 0.720 0.912580 578 0.970 0.987 684 984
45 1.196 222209(56) 2.537 379 134(28) 95 1.032997 849(148) 2.838063 251 (74) 0.225 0.911 915607 0.475 0 385679458 0.725 0.913 613 490 0.975 0.989686546
46 5.502622160(57) 2.168624 344(29) 96 9.916 779348(149) 3.494 206602(75) 0.230 0.910 754856 0.480 0 385 746 965 0.730 0.914 665 373 0.980 0.991 708408
47 2.586232415(59) 1.192 568 193(30) 97 9.619 275 968(151) 2.752921353(76) 0.235 0.909623 427 0.485 0 385835 552 0.735 0.915 736217 0.985
1.241391553(61) 1.040939685(31) 98 9.426890448(153) 3.424 322 470 (77) 0.993 750627
48
0.240 0.908521058 0.490 -8945132 .740 0.916826025 0990
5.843_585-145431) -9.333621544(155) 2r?2fr393-t40f78) 0 995813 260
49 6.082818640(62) -99
0.245 0.907 447 492 0.495 0 186075 617 0.745 0.917 934 795 0.995
50 3.041409320(64) 5.204 698 426(32) 100 9.332621544(157) 3.424 322 470 (79) 0.997 896364
0.250 0.906402477 0.500 0 86 226926 0.750 0.919 062 527 1.000 1.000000000 1
Ref. 1.20, p. 456. Ref. 1.20. p. 457.
368 A.06 BINOMIAL COEFFICIENTS Numerical Tables Numerical Tablcs A.06 BINOMIAL COEFFICIENTS {Continued) 369
'
(;) (;) (;) (5) 0 (;) 0 (;) C) (;) ti) ti) ti) ti) ti) ti) -
0
0
1 i
o- ocfficicnu (2.10*)
2 2 1
< - l)(fl - 2)-- (/ - X + 1) 2
("\
3 3 3 1
\i) *!(. "'- i>. 3
4 i 6 1 1
4
5 10 10 5 i
(;) (;)- (.:,)- 0- 5
6 6 15 20 15 > 1
7 7 21 35 35 21 i -- (;) (.:.)
For cocflicicmj noi s 7
8
H 8 28 56 70 56 28 a 1
9 9 36 81 126 126 84 36 9 1
i-ic.(-:;i (*)-(?) 12 650 9
15 15 105 155 1 365 3 003 5005 6435 6 435 5 005 3003 1 365 455 105 15 1 15
16 16 120 560 1 820 4 368 8 008 11 441) 12 870 II 14(1 8 008 4 368 1 820 560 120 16 16
17 17 136 680 2 380 6 188 :.' 376 19 448 24 310 24310 19 448 12 376 6 188 2 380 6111) 136 17
18 18 153 816 3 060 8568 S564 31 824 4 1 758 48 620 43 758 31 824 18 564 8 568 3 060 Hit, III
19 19 171 969 3 876 11 628 . - 132 50388 75 582 92 378 92 378 75 582 50 388 27 132 11 628 3 876 19
20 20 190 1 140 4 845 15 504 ".4760 77 520 125 970 167 960 184 756
167 960 123 970 77 520 38 760 15 501 20
-'I 21 210 1 330 5 985 20 349 -f 264 116 280 203 490 293 930 352 716
352 716 293 930 203 490 116 280 54 263 21
22 231 1 540 7 315 26 3.14 4613 1 70 544 319 770 497 420 646 646 705 432 646646 497 420 319 770 1 70 544
23 23 253 1 771 8 855 33 649 '947 245 157 490 314 817 190 1 114 066 1 352 078 1352 078 1 144 066 817 190 490 314 23
24 24 276 2 024 10626 42 504 M596 346 104 735 471 1 307 504 1961 256 2 496 144 2 704 156 2 196 144 1 961 256 1 307 504 24
25 25 300 2 300 12 650 53 130 :100 480 700 1 081 575 2 042975 3 268 760 4 457 400 5 200 300 5200300 4457400 3 268 7<<)
26 26 325 2 600 14 950 65 780 :->230 657 800 1 562 275 3 124 550 5 311 735 7 726 160 9 657 700 10400600 9657 700 7 726 160 26
27 351 2 925 17 550 80 730 .--.lllll 888 030 2 220 075 4 686 825 8 436 285 13 037 895 17 383 860 20 058 300 20058 300 17 383 860
28 28 378 3 276 20 475 98 280 .:-6 740 1 184 040 3 108 105 6 906 900 13 123 110 21 474 180 30 421 755 37 442 160 40116600 37 442 160 28
29 29 406 3 654 23 751 118 755 f"5 020 1 560 780 4 292 145 10 015 005 20 030010 34 597 290 51 895 935 67 863 915 77 558760 77 558 760 29
30 i 30 435 4 060 27 405 142 506 ,:-3 775 2 035 800 5 852 925 14 307 150 30 045 015 54 627 300 86 493 225 119 759 850 145 422 675 155 117 520 30
f ^^^^^T *i % f^ m
r~
* r^m
<f\ = Stirling number ofthe first kind (2.1 la) X\ = factorial polynomial (2.1 la) 9\** Stirling number ofthe second kind (2.1 la*) x> = pth power ofx (2.1 la*)
*> = &\X\l> + &fWt2) + v**?* + + P/W
AT =x(x - \)ix - 2)ix - 3) ix - p+ 1) =x*T +x2W +x3Sf? + +*>ST
For numbers not shown below use the recurrent relation 9*km) = k9^km~l> + ^"7"
For numbers not shown below use the recurrent rclaton St*"* = 5^"7" "~ (m - O^"
k yi' yj2' yiJ> y<4> yj" yf> y<7> y" y<9>
k y," y" S? y4) y? yi6> y" yi8' y
I i 1 i i i 1 1 1 1
2 -6 24 -120 720 -5040 40320
1 i -i 2 1 3 7 15 31 63 127 255
i -3 11 -50 274 -1764 13068 -109584
2 3 1 6 25 90 301 966
1624 -15 132 118121 3025
1 -6 35 -225
3
4 1 10 65 350 1701 7 770
1 -10 85 -735 6 769 -67284
4 5 1 15 140 1050
175 -1960 22449 6950
l -15
5 6 1 21 266
-21 322 -4 536 2646
1
6
7 1 28 462
1 -28 546
7 8 1
1 -36 36
8 9 1
1
9
BoW 1
(*) = Bcrnoulli polynomial (2.23a)
l,(x) _ i
1
, = Hirnoiilli number (2.22a, A.10)
h,(x) l -1 1 *w - (;.y 4. - i v
l-ll x'*/ /l-ll
1 _J
For example,
,(x) 0 2
1
4(x) _ 1
0 1 -2 1
l'l,(x) =x*Ba +Q*v/J, +Q*/J, +Q/i,
= !,v - V 4 v'
0 0 1
i
_ 1 1 1
/'l'l IV
0 0
B,(x) 0 :, 0
_7
0
7
2
-l 1
11
0 0 -1 1
(*) -,' 0 1
.::, 0 -3 0 5 0 -7 0 J' 5 1
(*)
_ U _ u 1
0 0 II 0 -11 0
Puto 0 o
_6M ni na 1001 n
2 0 0 0 0 0 -7 l
uW 0 10 2 SO
II, = auxiliary Bcrnoulli number (2.226) II, = Bcrnoulli number ('2.22a) ' = 0,1, 2, a = (-1)"
o 1.000 000 000(4-00) '116 -sa 1.511031 577(4-10) 54 -..., 2.583419885 (+42)
-II, 5.000000000 (-01) "l 7 m 4.21)0 140 431 (4-11) 35 >i 3.100 693 503 ( + 44)
- in 1.371 105521 (4-13) 36 -ll 4.092 733821 (+46)
II, II, 1.666666 667 (-01)
II, -I, 3.333 333 333 (-02) 'I,., M 4.883323 190(+14) 17 74 5.600 261 977 (+48)
Z?, II,. 2.380 952 381 (-02) w - 1.929657 934 (+16) M ~?6 8.085 808 705 (+50)
II,, M 8.416930476 (+17) 39 7,1 1.230 124459 (+53)
< - 3.333 333 333 (-02)
:, 5,o 7.575 757 576 (-02)
22 "44 4.033 807 185(4-19)
2.531 135531 (-01) 40 -80 1.969275 126 (+55)
21 46 2.115074864 (+21) II41 82 3.313 183806 (+57)
11,, -4H 1.208662 652(4-23) -, 8.851 176 136 (+59)
1*7 h 1.666 606 607(4-00)
a -16 7.092 156863(4-00)
Il 25 5o 7.500 866 746(4-24) 41 86 1.083429 888 (+62)
,a 5.497 117 794(4-01)
26 -w 5.038 778 101 (4-26) 44 -M 2.101 081 990 (+64)
B 54 3.652 877 638(4-28)
J -Ilm 5.291 242 424(4-02)
4S ,0 4.263 004 386 (+66)
i, ll,i 0.192 123 188(4-03)
-B24 8.658025 311 (4-04) 28 2.849 (176 930 (+30)
., 2.386 542 750 (+32) Series representation for r 2:
M -, 2.139994 926 (+34)
IS w 1.425517 107(4-06)
14 -Bm 2.729 823 107(4-07) Pr4TiV*/
15 II i., 6.015808 739(4-08) 11 I... 2.017 031 483 ( + 30) Rccurrcncc relations Cor r s 32:
-64 2.060 029 615 (+38)
M 66 2.238 571 753 (+40) i. - a,r - l
fi, = B = /}, = = () (2,)' "<-'
A.11 EULER POLYNOMIALS
,
,(*N. f,i .
t. 'i c. e., ,, fM <, 'II. 'll 'ia 'n '.. 15
E0(x) i
E,(x) = Euler polynomial (2.25n)
_i E, = Euler number (2.24n), (A.12)
E,(x) 1
(*) 0 -1 1
" =rrl"-2'C IV"*- %y
,(*) i
0
_ i
1
For example,
<
0 -3 0 0 3 1
,.(*) 5
l _21 31 _z
,(*) 0 2
0 0 1
Mx) 0 17 0 -28 0 14 0 -4 1
_ai 0
153
0 -63 0 21 0 -i 1
,(*) 2
k,w 621
0
1705
2 0 2801
4
0 -231 0 isa 0 _Ji
2 1
(*) 929569
II, ' 9
-173405
0 213211
0 _ lilil
0
109395
0 _2PJJ3
2 0 na
0 _ii
2 1
E, = auxili iry Euler num jer (2.244) '> = Euler number (2.24<i) r = 0, 1,2,... or = (-lp' 0 = (-l)'+1
n 1.000000 000 (+(MI) ,.. /,, 1.775 193 910 (+29) E,,
. 0.000 (HMI(H)O ( + 00) na 1.456 184439 (+83)
/'i; -14 8.0722329.92 (+31) l--v. - 71, 2.850517834 (+86)
'i ii 11. 4.122206033 (+34) /'II. 72 5.905 747 212 (+89)
':, -E, 1.000 000000 (+00)
E, 4 5.000 000 000 (+00)
3
,., "f. 2.348 958 053 (+37) ,7 -74 1.292973 665 (+93)
6.100 000 000 (+01)
40 1.485 115 077 (+40) M 76 2.968 928 185 (+96)
"42 1.030 402 273 (+43) Ex -, 7.270601 719(+99)
4 . 1.385 000 000 (+ 03)
5 -Exo 5.052 100 000 (+ 04) Erl 44 7.947 579 423 (+45)
., ,2 2.702 765 000 (+06) 40 ^80 1.862291 577 (+104)
E -1., 0.007 537 517 (+ 48) ., 5.013 104944 (+107)
E 48 6.096 278646 (+51)
7
42 84 1.416525577 (+111)
-,4 1.993 609 810 (+08)
Bt 16 1.939 151 215 (+10)
E,, Er, "50 6.053 285248 (+54)
-18 2.404879 675 (+12) 41 -86 4.196 643 167 (+114)
En, 52 6.506 162 487(+57) 44 ^88 1.302 159592 (+118)
C27
-.5, 7.546659939 (+60) 45 ';90 4.227240691 (+121)
,o 20 3.703 711882 (+14)
P.i -E,, 6.934 887 439 (+16)
E,, 24 1.551 453 416 (+19) v. 9.420 321 896 (+63)
En -sa 1.262 201 925 (+67) Series rc prcsctilation for r & 2:
E ,,0 1.810891 150(+70)
fs
fu
-26
2
4.087 072 509(421)
1.252259041 (+ 24)
E, = -, =2(-) (2r)!i?(I)
15 -3(1 E,, "62 2.775 710 171 (+ 73)
4.415438 932 (+20) Recurre nce relations Cor r a 32:
En Ka
E,, 7.886 284 211 ( + 79)
1 = S = , 0
.
E, =
---(1) 2r(2r- l)|j,_,|
1 1
376 A.13 CIRCULAR SINE AND COSINE INTEGRALS Numerical Tables Numerical Tables A.13 CIRCULAR SINE AND COSINE INTEGRALS (Continued) 377
/"sin*
S(x) = l'sJHj,
Jo X
(10.0l4) C(*) =1 dx (10.014) s(*) = 1
Jo x
dx (10.014) Ciix) = cos X
dx (10.014)
-0.824 6*5 062 0.75 0.726954 247 0.152 163 601 2.5 1.778 520 173 0.285871 196 7.5
0.25 0.249133 570 1.510681531 0.115633 203
0.259025 534 -0.786 71.'386 0.76 0.736030907 0.161809 783 2.6 1.800394 451 0.253 336616 7.6
0.26 1.523 313 791 0.119 597 529
0.268908891 -0.750 28" 386 0.77 0.745083 666 0.171239811 2.7 1.818212077 0.220084 879 7.7
0.27 1.536109238 0.122 245832
0.278 783 309 -0.715 28* 095 0.78 0.754 112 249 0.180458334 2.8 1.832096589 0.186488 390 7.8
0.28 1.548937 458 0.123 585954
0.288648469 -0.681 610 153 0.79 0.763 116 380 0.189469829 2.9 1.842 190 195 0.152895 324 7.9
0.29 1.561 671 070 0.123 638007
-0.649172933 0.80 0.772095 786 0.198278616 3.0 1.848652 528 0.119629 786 8.0
0.30 0.298504 044 1.574 186 822 0.122433 883
0.308402 750 -0.617 8 321 0.81 0.781 050 192 0.206888861 3.1 1.851659 308 0.086991831 8.1
0.31
0.789979 329 0.215304586 3.2 1.851400897
1.586 366622 0.120016673 1
0.318185138 -0.587 7(* 639 0.82 0.055257 412 8.2
0.32
0.798882928 0.223 529675 3.3 1.848080 783
1.598098511 0.116440006 1
-0.55854*724 0.83 +0.024678285
0.33 0.328010011
0.807 760 719 0.231567 882 3.4 1.841913 983
8.3 1.609277542 0.111767 293 1
0.34 0.337 827 399 -0.530 355 151 0.84 -0.004518078 8.4 1.619806597 0.106070920
0.347626 791 -0.503 075 569 0.85 0.816612437 0.239422837 3.5 1.833 125 399 -0.032128549 8.5
0.35 1.629597 100 0.099431 359
0.357 418056 -0.47666! 125 0.86 0.825 437817 0.247 098049 3.6 1.821948116 -0.057 974 352 8.6
0.36 1.638569645 0.091 936 240
0.367 197 475 -0.451 Oto 976 0.87 0.834236595 0.254596915 3.7 1.808621681 -0.081 901 001 8.7
0.37 1.646654 531 0.083 679 370
0.376951556 -0.426 251 S55 0.88 0.843008 510 0.261 922 726 3.8 1.793 390355 -0.103 778 150
0.38 8.8 1.653 792186 0.074 759 720
0.386219499 -0.4021" 704 0.89 0.851 753 309 0.269078669 3.9 1.776501360 -0.123 499349 8.9
0.39 1.659933 505 0.065 280385
0.396461465 -0.378 b\* 346 0.90 0.860470 711 0.276067 831 4.0 1.758203 139 -0.140981698 9.0
0.40 1.665040076 0.055 347 531
-0.356114 201 0.91 0.869160481 0.282893 207 4.1 1.738 743 627 -0.156165 392
0.41 0.406190 310
0.289557 702 4.2 1.718368 564
9.1 1.669084 306 0.045 069 333 1
0.415905 717 -0.334 0*: 035 0.92 0.877822 356 -0.169013 157
0.42
0.296064136 4.3 1.697 319851
9.2 1.672049448 ).034 554 9I3 1
0.425582945 -0.312624 "40 0.93 0.886456084 -0.179 509573
0.43 9.3 1.673929 528 1.023 913 305 1
0.44 0.435294 951 -0.291 77* 135 0.94 0.895061411 0.302415246 4.4 1.675833 959 -0.187660287 9.4 1.674 729173 1.013 252 419 |
0.45 0.444968 145 -0.271 4*: "99 0.95 0.903638088 0.308613691 4.5 1.654 140414 -0.193491 122 9.5 1.674463 342 + 1.002678059
0.46 0.454626684 -0.251 74-5913 0.96 0.912 185866 0.314662055 4.6 1.632460 353 -0.197047080 9.6 1.673156980
-
Fot x < 0.01 use the formula in (10.03o). For jr> 9.9 use the formula in (10.03c).
i i I I 1 "i i
37a A.14 EXPONENTIAL INTEGRALS Numrica! Tables Numerical Tables A.14 EXPONENTIAL INTEGRALS (Continued) 379
mx) = f dx
J. X
(10.014) i(,) =f^ -dx (10.014)
-2.468051020 -2.3*7 884 599 0.55 -0.503364081 0.615290657 1.05 -0.201872813 2.031 087 184 1.55 -0.092882108
0.05 3.451954503
-2.295306918 -2.125282916 0.56 -0.493019959 0.646677490 1.06 -0.198 572 347 2.058309800 1.56 -0.091523960
0.06 3.482 405 289
-2.150838180 -2.010800064 0.57 -0.482960034 0.677818642 1.07 -0.195 335403 2.085 547825 1.57 -0.090187 917
0.07 3.512965580
-2.026941003 -1.8*6884103 0.58 -0.473173 433 0.708725 720 1.08 -0.192 160479 2.112803672 1.58 -0.088873566
0.08 3.543637024
-1.918 744 770 -1.738663 750 0.59 -0.463649849 0.739409 764 1.09 -0.189046 118 2.140079 712 1.59 -0.087580504
0.09 3.574 421266
0.10 -1.822923958 -1.622812814 0.60 -0.454 379 503 0.769881290 1.10 -0.185990905 2.167 378280 1.60 -0.086308334 3.605 319949
-1.737 106694 -1.51*958 751 0.61 -0.445 353 112 0.800150320 1.11 -0.182993 465 2.194 701672 1.61 -0.085056670
0.11 3.636334 719
-1.659541752 -1.419 349669 0.62 -0.436561854 0.830226417 1.12 -0.180052 467 2.222052 152 1.62
0.12 -0.083825133 3.667 467221
-1.588899 305 -1.328665020 0.63 -0.427997 338 0.860118 716 1.13 -0.177 166615 2.249 431949 1.63 -0.082613354
0.13 3.698 719099
-1.524 145 722 -1.243 560654 0.64 -0.419651581 0.889835949 1.14 -0.174 334 651 2.276843 260 1.64 -0.081420970
0.14 3.730091999
0.15 -1.464 461671 -1.1*4086417 0.65 -0.411516976 0.919386468 1.15 -0.171555 354 2.304 288252 1.65 -0.080 247 627 3.761587 569
-1.409186699 -1.SS731238 0.66 -0.403 586275 0.948 778277 1.16 -0.168827 535 2.331 769062 1.66 -0.079092978
0.16 3.793 207456
-1.357 780653 -1.017 234 290 0.67 -0.395 852 563 0.978019042 1.17 -0.166150040 2.359287 800 1.67
0.17 -0.077956684 3.824 953 310
-1.309 796 135- -0.949 147 505 0.68 -0.388 309 243 1.007 116 121 1.18 -0.163 521 748 2.386846 549 1.68
0.18 -0.076838412 3.856826 783
-1.264858424 -0.854 095 487 0.69 -0.380950010 1.036076576 1.19 -0.160941567 2.414447 367 1.69 -0.075 737839
0.19 3.888829528
-1.222650544 -0.621 760587 0.70 -0.373 768 843 1.064907 195 1.20 -0.158 408 437 2.442092 285 1.70
0.20 -0.074 654 644 3.920963 201
-1.182 901986 -0.7*1 871 623 0.71 -0.366 759981 1.093614501 1.21 -0.155 921324 2.469 783 315 1.71
0.21 -0.073 588 518 3.953 229462
-1.145 600 055 -0.704 195 224 0.72 -0.359917 914 1.122 204 777 1.22 -0.153 479 226 2.497 522442 1.72
0.22 -0.072539154 3.985 629972
-1.109881 139 -0*4S 529 102 0.73 -0.353 237 364 1.150684069 1.23 -0.151081 164 2.525 311634 1.73 -0.071506255
0.23 4.018 166395
-1.076235415 -0.594 696 758 0.74 -0.346 713 279 1.179058208 1.24 -0.148 726 188 2.553152836 1.74 -0.070489527
0.24 4.050840400
-1.044 282634 -0.542 543 264 0.75 -0.340340813 1.207 332816 1.25 -0.146413 373 2.581047974 1.75
0.25 -0.069488685 4.083653659
-1.013 888 737 -0.491931883 0.76 -0.334115 321 1.235 513 319 1.26 -0.144 141 815 2.608998956 1.76 -0.068503447
0.26 4.116607847
-0.984 933 104 -0.442 941312 0.77 -0.328032 346 1.263604960 1.27 -0.141 910639 2.637007 673 1.77 -0.067533539
0.27 4.149 704645
-0.957 308300 -0.394 863 444 0.78 -0.322087 610 1.291612805 1.28 -0.139 718 989 2.665 075997 1.78 -0.066578691
0.28 4.182 945 736
-0.930918246 -0.348 201510 0.79 -0.316277 004 1.319541753 1.29 -0.137 566032 2.693 205 785 1.79
0.29 -0.065638641 4.216 332809
-0.905676652 -0.302 668 539 0.80 -0.310 596 579 1.347 396548 1.30 -0.135 450958 2.721398880 1.80
0.30 -0.064713 129 " 4.249867 557
-0.881 505 746 -0.258 186075 0.81 -0.305042 539 1.375 181783 1.31 -0.133 372975 2.749657 110* 1.81
0.31 -0.063801903 4.283 551681
-0.858335189 -0.214683096 0.82 -0.299611236 1.402901910 1.32 -0.131331314 2.777 982287 1.82
0.32 -0.062904 715 4.317 386883
-0.836 101 162 -0.1 "2 095 092 0.83 -0.294299155 1.430561245 1.33 -0.129 325 224 2.806376214 1.83
0.33 -0.062021 320 4.351374872
-0.814 745580 -0.150363 293 0.84 -0.289102918 1.458163978 1.34 -0.127353 972 2.834840677 1.84 -0.061 151 482
0.34 4.385517 364
-0.794217 357 -0.089434001 0.85 -0.284019269 1.485 714176 1.35 -0.125416844 2.863 377 453 1.85
0.35 -0.060294967 4.419816080
-0.774462 218 -0.049 258 017 0.86 -0.279045070 1.513 215 791 1.36 -0.123513146 2.891 988 308 1.86
0.36 -0.059451545 4.454 272 746
-0.755 441 428 -0.CO9 790I48 0.87 -0.274 177 301 1.540672664 1.37 -0.121642198 2.920674 997 1.87
0.37 -0.058620994 4.488889097
-0.737 112144 +0.029011221 0.88 -0.269413046 1.568088534 1.38 -0.119803 337 2.949439263 1.88
0.38 -0.057803091 4.523666872
-0.719436652 0.0*" 184 501 0.89 -0.264 749 496 1.595467036 1.39 -0.117 995 919 2.978 282844 1.89
0.39 -0.056997623 4.558607817
-0.702 380119 0.104 265 219 0.90 -0.260183939 1.622811714 1.40 -0.116219313 3.007 207 464 1.90
0.40 -0.056204 378 4.593 713687
-0.685910311 0.141 786 307 0.91 -0.255 713 758 1.650126019 1.41 -0.114 472903 3.036214843 1.91
0.41 -0.055423149 4.628986242
-0.669997 342 0.195 347657 0.92 -0.251336425 1.677 413 317 1.42 -0.112 756090 3.065306691 1.92
0.42 -0.054653 731 4.664 427 249
-0.654613 448 0.214 269 821 0.93 -0.247049 501 1.704676891 1.43 -0.111068 287 3.094484 712 1.93
0.43 -0.053895927 4.700038485
-0.639 732 798 0.249 787 245 0.94 -0.242850627 1.731919946 1.44 -0.109408923 3.123 750601 1.94
0.44 -0.053 149540 4.735821734
-0.625331316 0.2M 855 406 0.95 -0.238 737 524 1.759 145 612 1.45 -0.107 777 440 3.153 106049 1.95
0.45 -0.052414380 . 4L771 778 785
-0.611386530 0.319 497 483 0.% -0.234 707 988 1.786 356947 1.46 -0.106173 291 3.182552 741 1.%
0.46 -0.051690257 4.807911438
-0.597877 426 0.355 735 1% 0.97 -0.230 759 890 1.813556941 1.47 -0.104595946 3.212092355 1.97
0.47 -0.050976988 4.844 221501
-0.584 784 345 0.37 588 924 0.98 -0.226891 167 1.840 748519 1.48 -0.103044 882 3.241726566 1.98
0.48 -0.050274392 4.880 710 791
-0.572088836 0.421077 818 0.99 -0.223099826 1.867 934543 1.49 -0.101519593 3.271457042 1.99*
0.49 -0.049582291 4.917381 131
-
380 A.15 ERROR FUNCTION AND ITS DERIVATIVE Numerical Tables Numerical Tables A.15 ERROR FUNCTION AND ITS DERIVATIVE (Continued) 381
d[criix)]=2_., d[crr{x)] 2 ., |
crf(x) e-''dx (10.014) dx y/x
erf(x)
=""** (,00,4) dx -vr~" <,oo,)|
crf(x) d[er(x)]/dx X erf(jr) d[crfix)]/dx x erfW dlcrt-(x)]/dx |
erf(x) <[crf(*)]/*
1.00 0.842 700 793 0.415 107 497 1.50 0.966 105 147
1.128 379167 0.50 0.520499878 0.878 782579 0.118930289
0.00 0.000000000 1.01
0.869951547 0.846810496 0.406847 132 1.51 0.%7 276 748
0.011283416 1.128266335 0.51 0.529243620 0.115403827
0.01" 1.02 0.850838018 0.398671399 1.52
1.127 927 906 0.52 0.537 898631 0.861 037 034 0.968 413 497 0.111959536
0.02 0.022564575
1.03 0.854 784 212 0.390581837 1.53 0.%9 516209
0.033841222 1.127 364083 0.53 0.546464097 0.852043 444 0.108 5% 320
0.03 1.04 0.858649947 0.382 579899
0.554939251 0.842975 181 1.54 0.970585690 0.105313068
0.045111 106 1.126575204 0.54
0.04
0.833836647 1.05 0.862 436 106 0.374 666957 1.55 0.971622 733
0.056371978 I 125561742 0.55 0.563 323 366 0.102 108658
0.05 1.06 0.866143 587 0.366844 303
0.571 615 764 0.824632 240 1.56 0.972628 122 0.098981951
0.067 621594 1.124 324 305 0.56
0.06 1.07 0.869 773 297 0.359 113149
0.579815806 0.815 366346 1.57 0.973602 628 0.095931800
0.078857 720 1.122 863 633 0.57
0.07 1.08 0.873 326158 0.351474625
0.587922900 0.806043 343 1.58 0.974 547009 0.092957 046
0.090078126 1.121 180600 0.58
0.08 1.09 0.876803 102 0.343929 783
0.595936497 0.796667 591 1.59 0.975 462016 0.090056 524
0.101 280 594 1.119276213 0.59
0.09
0.787 243 432 1.10 0.880 205 070 0.336479598 1.60 0.976 348383
0.112462916 1.117 151607 0.60 0.603856091 0.087 229059
0.10 1.11 0.883 533012 0.329124 %7 1.61
1.114 808050 0.61 0.611681219 0.777 775 185 0.977 206837 0.084 473 470
0.11 0.123 622 8%
0.768267 144 1.12 0.886 787 890 0.321866 710 1.62 0.978038088
0.134 758 352 1.112246938 0.62 0.619 411462 0.081 788 571
0.12 1.13 0.889970 670 0.314 705 574
0.63 0.627 046443 0.758 723 576 1.63 0.978842 840 0.079 173 173
0.13 0.145867 115 1.109 469 793
0.749 148 716 1.14 0.893 082 328 0.307642230 1.64 0.979621780
0.156947 033 1.106478265 0.64 0.634585829 0.076626082
0.14
0.739 546 763 1.15 0.896 123 843 0.300677 276 1.65 0.980 375 585
0.167995971 1.103 274 127 0.65 0.642029 327 0.074 146 103
0.15 1.16 0.899096 203 0.293 811239
0.649 376680 0.729921881 1.66 0.981 104 921 0.071 732040
0.179011813 1.099859 273 0.66
0.16 1.17 0.902 000399 0.287 044 575 1.67
1.096 235 719 0.67 0.656627 702 0.720278193 0.981810442 0.069 382 697
0.17 0.189992 461
1.18 0.904 837 427 0.280377 670
0.18 0.200935 839 I.092 4O5 601 0.68 0.663 782203 0.710619 778 1.68 0.982 492 787 0.067 096878 1
0.700950672 1.19 0.907 608 286 0.273 810844 1.69 0.983 152 587
0.211839892 1.088371 168 0.69 0.670840062 0.064 873 390
0.19
0.691274860 1.20 0.910 313 978 0.267 344 347 1.70 0.983 790459
0.222 702589 1.084 134 787 0.70 0.677 801 194 0.062 711041
0.20 1.21 0.912955 508 0.260978 366
0.71 0.684 665 550 0.681596279 1.71 0.984 407 008 0.060 608644
0.21 0.233 521923 1.079698934
0.671918811 1.22 0.915 533 881 0.254 713 024 1.72 0.985002827
0.244 295912 1.075 066 196 0.72 0.691 433 123 0.058 565016
0.22 1.23 0.918050104 0.248 548 381
0.73 0.698 103 943 0.662246284 1.73 0.985 578 500 0.056 578979
0.23 0.255022600 1.070 239 267
0.652 582 467 1.24 0.920505 184 0.242 484 434 1.74 0.986134 595
0.265 700059 l.(-65 220 945 0.74 0.704678078 0.054649 361
0.24
0.642931069 1.25 0.922900128 0.236521 122 1.75 0.986671671
0.276 326390 1.060014 129 0.75 0.711 155634 0.052 774 9%
0.25 1.26 0.925 235942 0.230658328
0.76 0.717 536 753 0.633295 740 1.76 0.987 190 275 0.050954 726
0.26 0.286899 723 1.054 621819
0.623680063 1.27 0.927 513 629 0.224895875 1.77 0.987690942
0.297 418212 1.049047 110 0.77 0.723821614 0.049 187 401
0.27 1.28 0.929 734 193
0.730010431 0.614087 551 0.219233 532 1.78 0.988 174 1% 0.047471879
0.307880067 1.(43 293 189 0.78
0.28 1.29 0.931898633
0.736 103 454 0.604521670 0.213 671015 1.79 0.988640549 0.045 807 027
0.318283496 1.037363 333 0.79
0.29
0.742100965 0.594985 786 1.30 0.934007 945 0.208207987 1.80 0.989090 502
0.328626 760 1.031260910 0.80 0.044 191 723
0.30 1.31 0.936063 123
0.81 0.748003 281 0.585483 216 0.202 844062 1.81 0.989524545 0.042624 854
0.31 0.338908150 1.024 989 366
0.576017197 1.32 0.938065155 0.197 578805 1.82 0.989943157
0.349125995 1.018 552 231 0.82 0.753810 751 0.041 105 319
0.32 1.33 0.940015026
0.759523 757 0.566590894 0.192411733 1.83 0.990 346805 0.039 632 026
0.359 278655 1.011953 112 0.83
0.33 1.34 0.941 913 715
0.765 142 712 0.557 207 397 0.187 342 317 1.84 0.990 735 948 0.038203897
0.369 364 529 1.(05 195689 0.84
0.34
0.770668058 0.547 869 717 1.35 0.943 762 196 0.182369987 1.85 0.991 111030
0.379382054 0.998283 712 0.85 0.036819865
0.35 1.36 0.945 561437
0.776100268 0.538580 792 0.177 494 126 1.86 0.991472488 0.035 478877
0.389 329 701 0.991 221 000 0.86
0.36 1.37 0.947 312 398
0.87 0.781439846 0.529343477 0.172 714081 1.87 0.991820 748 0.034 179 892
0.37 0.399 205984 0.984011434
0.520160557 1.38 0.949016035 0.168029 157 1.88 0.992 156 223
0.409009 453 0.976658954 0.88 0.786687 319 0.032921881
0.38 1.39 0.950673 2%
0.89 0.791843 247 0.511034 712 0.163 438622 1.89 0.992479318 0.031 703 831
0.39 0.418 738 700 0.9*9 167 559
1.40 0.952 285 120 0.158941708 1.90 0.992 790 429
0.428392 355 0.9*1 541 299 0.90 0.796908212 0.030 524 740
0.40 '.41 0.953 852439
0.801882826 0.154 537 613 1.91 0.993 089940 0.029 383624
0.437 969090 0.953 784 273 0.91
0.41 .42 0.955 376179
0.92 0.806 767 722 0.150225 503 1.92 0.993 378 225 0.028279 510
0.42 0.447467 618 0.945900626
0.811563 559 1 43 0.956857 253 0.146 004511 1.93 0.993655650
0.456886695 0.937 894 544 0.93 0.027 211441
0.43 1 44 0.958296 570
0.94 0.816271019 0.141873 741 1.94 0.993 922571 0.026 178 475
0.44 0.466 225 115 0.929 770 254
384 A.18 MODIFIED BESSEL FUNCTIONS Numerical Tables Numerical Tables A.19 MODIFIED BESSEL FUNCTIONS 385
1
X
X <-'/(*) <-/,(*) X <"/,(*) '"*/,(*) *
'**.,<*) ''*,(*)
0.0 00 eo
5.0
1.000000000 0.000000000 5.0 0.183540813 0.163972 267 0.547807564 0.600273859
0.0
0.1 2.682326102 10.8% 182683 5.1
0.%7 100926 0.045298447 5.1 0.181615 102 0.162663855 0.542635 352 0.593625046
0.1
0.179 749488 0.161383 285 0.2 2.140 757323 5.833386037 5.2 0.537607 354
0.2 0.826938552 0.082283124 5.2 0.587 188606
0.3 1.852627 301 4.125 157 762 5.3 0.532 716974
0.757580625 0.112 377 561 5.3 0.177940865 0.160129 791 0.580953609
0.3
0.158902615 0.4 1.662682089 3.258673880 5.4 0.527958033
0.697 402 171 0.136 763 224" 5.4 0.176186348 0.574909887
0.4
0.5 1.524 109 386 2.731009 708 5.5 0.523 324 732
0.645035271 0.156420803 5.5 0.174483 256 0.157 701009 0.569047 974
0.5
0.156524241 0.6 1.416 737 621 2.373 920038 5.6 0.518811625
0.6 0.599327 203 0.172 164420 5.6 0.172829095 0.563 359039
0.7 1.330123656 2.115011313 5.7 0.514413 594
0.7 0.559305 527 0.184 669983 5.7 0.171221536 0.155 371592 0.557 834 835
0.154242 364 0.8 1.258 203 122 1.917930299 5.8 0.510125818
0.8 0.524148942 0.194498693 5.8 0.169658406 0.552467 650
0.9 1.197 163 380 1.762388220 5.9 0.505943 758
0.9 0.493162966 0.202 116531 5.9 0.168137 673 0.153135874 0.547 250264
0.143 185 175 1.9 0.861450617 1.067 470930 6.9 0.469 101 637
1.9 0.318243 163 0.216611 911 6.9 0.154895609 0.502009947
rnri
E a>; K E
a>; K
Appendix B
0
1 1.570915958 1.570676 709 46 1.869147546 1.341806058
4 1.572 712 435 1.568883 380 49 1.917997 546 1.314 729 560
6 1.573 113 608 1.566 4% 788 51 1.953 864809 1.296 278 008
8
9
7 1.576 677 %2
1.578 486 578
1.580540934
1.564947 563
1.563162 230
1.561 141 745
52
53
54
1.972882266
1.992669 756
2.013 266565
1.286953 739
1.277 573948
1.268146531 GLOSSARY
10 1.582 842 804 1.558887 197 55 2.034 715312 1.258679625
1.585 394 164 1.556 399 798 56 2.057 062323 1.249181621
OF SYMBOLS
11
12 1.588197 213 1.553680892 57 2.080358067 1.239661 175
% co 1.000000000
-
1 r "1 ' ~1 ' 1 ' "1 ' 1 ~ 1 "1 "1 """1 " I r' "~1 " ~i
Glossary of Symbols
Glossary of Symbols
388 B.1 RELATIONSHIPS B.5 ANALYSIS 389
(a, 4) Bounded open intcrval
or : Equals lfl b] Bounded closed interval
=j= or # Does not equal fix), Fix) Functon ofx
> Greater than f~ (x),F~l(x) Inverse function ofx
< Less than
a Greater than or equal
Less than or equal J>, Sum of terms f] tt
<-i
Product oterms
B Idntica!
= Approximately equal
A [1 ,!) =(I ,(> ,,(l ,,(1 .,)))) Ntslcd sum oa + IIEnn,
B.2 NUMERICAL CONSTANTS
A Incrcmcntofi, du DMrential ofu
3t Archimedes constant (2.17a) G Cataln constant (2.274)
e Euler's constant (2.18a l y Euler's constant (2.19a)
B, Bernoulli number (2.22a) B, Auxilliary Bernoulli number (2.224) >First-order derivative oy = fix) with respect to .v
E, Euler number (2.24a) E, Auxilliary Euler number (2.244)
9\f) Stirling number ofthe first kind (2.1 la) a, 8 Golden mean numbers (2.214)
9\p) Stirling number ofthe second kind (2.1 la*)
F. Fibonacci number (2.21 b i $S>.r,
nth orderderivative ofy =/(x) with respect to x
d-fjx)
,f"(x),D<r>/{x)
B.3 ALGEBRA dx"
4,
a2
b2
Determinant
i a *
Matrix d2/ \Second-order partial derivative ofa; =f(x,j,...) with respect to *
f F
dx2,J"'t"
u = bx Variable v = bx ln a Variable
JAx) Ordinary Bessel function ofthe first kind oforder r (10.074)
e',e~m,a',a~",e',t Exponential functions (9.05a) YAx) Ordinary Bessel function ofthe second kind oforder r (10.07c)
Inx, logx Logarithmic functions (9.06a) jAx) Sphercal Bessel functon ofthe first kind oforder r (10.12a)
sin x Circular sine (9.07a) yAx) Sphercal Bessel function ofthe second kind of order r (10.12a)
eos x Circular cosine (9.07a) L(x) Modified Bessel function ofthe first kind oforder r (10.134)
sinh x Hyperbolic sine (9.08a) KAx) Modified Bessel function ofthe second kind oforder r (10.13c)
cosh x Hyperbolic cosine (9.08a) tr(x) Modified sphercal Bessel functon ofthe first kind oforder r (10.17a)
tan* Circular tangent (9.09a) kAx) Modified sphercal Bessel functions ofthe second kind of order r (10.17a)
B(x) Beta function (2.314) crf(x) Error function (10.014) LAx) Laguerre polynomial ofthe first kind of order r (11.16a)
Wix) Digamma function (2.32a) 4"""(x) Polygamma functon (2.34a) Ltix) Laguerre polynomial ofthe second kind oforder r (11.16a)
Z(x) Complementary zeta function (2.27a) HAx) Hermite function ofthe first kind oforder r (11.16a)
Z(x) Zeta function (2.26a)
Htix) Hermite function ofthe second kind of order r (11.16a)
Si(x),S(x) Circular and hyperbolic sine integris (10.014)
HAx) Hermite polynomial ofthe first kind oforder r (11.16a)
Ci(x),Ci(x) Circular and hyperboliccosine integris (10.014)
Hfix) Hermite polynomial ofthe second kind oforder r (11.16a)
Ei(x),i(x) Alternating and monotonicexponental integris (10.014)
Li(x) Logarithmic integral (10.014)
Six),Cix) Fresnel integris (10.04a)
Fik,x) Incomplete elliptic integral ofthe first kind (10.054)
Eik,x) Incomplete elliptic integral ofthe second kind (10.054) B.10 GREEK ALPHABET
Uik,x) Incomplete elliptic integral ofthe third kind (10.054) A a Alpha H " Eta N V Nu T T Tau
n,n(jt,i) Completeelliptic integral ofthe third kind (10.054). A 6 Delta K K K.appa n jr Pi X X Chi
K Zeta M " Mu 2
P
a Sigma Q
V
(O Omega
1 ' 1 ~~~% r~ 1 I ' I 1 1
392 B.11 INTERNATIONAL SYSTEM OFUNITS (SI UNITS) Glossary of Symbols Glossary of Symbols
B.13 U.S. CUSTOMARY SYSTEM OF 393
UNITS (FPS SYSTEM)
Symbol ame Quantity Relations
Symbol* ame Quantity Relations
cd candela luminous intensity basic unit
kg kilogram mass basic unit deg degree plae angle supplementary unit
Im lumen luminous flux cd sr ft foot length basic unit
Ix lux illumination Im m~2 i standard gravity acceleration 32.l74ftsec-2
m meter length basic unit hp horsepower powcr 550 lbf-ft-sec-'
mol mole amount of substance basic unit Ib pound-mass mass lbf--
rad radian plae angle supplementary unit Ibf pound-force forc basic unit
s second time basic unit pd poundal forc Ib ft sec-2
sr steradian solid angle supplementary unit sec second time basic unit
A ampere elcctric current basic unit si slug mass
lbf-ft-'-sec2
C coulomb electric charge As Btu British thermal unit work, encrgy 778.128ftlbf
F farad elcctric capacitance A-s-V :F degrec Fahrenheit temperature basic unit
H henry electric inductance V-s-A"'
Hz hertz frequeney s_l For A.C,F,...referto(B.ll).
J joule work, encrgy N-m
K kelvin temperature degree basic unit B.14 METRIC SYSTEM OF UNITS (MKS SYSTEM)
X newton forc kg m s~"
Pa pasca1 pressure, stress Nm-2
Svmbol* ame Quantity Relations
S siemens electric conductance Q-1
T tesla magnetic flux density Wb m-2 cjI calore work, encrgy 0.42665 m-kgf
Y volt voltagc, electromotive forc W-A-' deg degrec plae angle supplementary unit
\V watt power Js- dvn dyne fbrec lO-'kgf
Wb weber magnetic flux Vs ere " work, encrgy 10"8 m-kgf
Q ohm electric resistance V-A-' S standard gravity acceleration 9.806 65 m-sec"2
hp horscpower power 75kgf- m sec"1
1.06 Collatz, L.: "Functional Analysis and Numerical Mathcmatcs," Academic, New York, 1966. C.4 NUMERICAL INTEGRALS
1.07 Demidovich, B. P., and I. A. Marn; "Computational Mathematics," Mir, Moscow, 1981. 4.01 Davis, S., and P. Rabinowitz: "Methods ofNumerical Integration," Academic, New York, 1975.
1.08 Froburg,C. E.: "Numerical Mathematics," Benjamin-Cummings, Reading, Mass., 1985. 4.02 Gradshteyn, I. S., and I. M. Ryzhik: "Tables ofIntegris, Series and Products," 4th ed., Academic, New
York, 1965.
1.09 Hamming, R.: "Numerical Methods for Sciensts and Engineers," 2d ed., McGraw-Hill, New York, 1973.
1.10 Hildebrand, F. B.: "Introduction to Numerical Analysis," McGraw-Hill, New York, 1956.
4.03 Grobner, W., and N. Hofreiter "Intcgraltafcln," 4th cd., vols. I, II,Springer, Vienna, 1965.
1.11 Jain, M. K., S. R K. Inycngar, and R K. Jain: "Numerical Methods for Sdentifc and Engineerng
4.04 Krylov, V. I.: "Approximate Calculation of Integris," Macmillan, New York, 1962.
Computations," Halstead, New York, 1982. 4.05 Nikosky, S. M.: "Quadrature Formulas," Fizmatgiz, Moscow, 1958.
1.12 Johnson, L. W., and R D. Riess: "Numerical Analysis," 2d ed., Addison-Wesley, New York, 1982. 4.06 Spiegel, M. R.: "Mathematical Handbook," McGraw-Hill, New York, 1968.
1.13 Korn, G. A., and T. M. Korn: "Mathematical Handbook for Scientists and Engineers," 2d ed. 4.07 Spiegel, M. R: "Schaum's Outline of Finite Diflerences and Diflerence Equations," McGraw-Hill New
McGraw-Hill, New York, 1968. York, 1971. '
1.14 Marchuk, G. I.: "Methods of Numerical Mathematics," Springer, New York, 1975. 4.08 Stroud, A.: "Approximate Calculation ofMltiple Integris," Prcnticc-Hall, Englewood Clifls, NJ., 1971.
1.15 Pachner,J.: "Handbook of Numerical Analysis Applications," McGraw-Hill, New York, 1984. 4.09 Stroud, A, and D. Secrest: "Gaussian Quadrature Formulas," Prentice-Hall, Englewood Clifls, NJ., 1966.
1.16 Ralston, A., and P. Rabinowitz: "A First Course in Numerical Analysis," McGraw-Hill, New York, 1978.
C.5 SERIES AND PRODUCTS OF CONSTANTS
1.17 Scarborough, J. B.: "Numerical Mathematical Analysis," 3d ed., Johns Hopkins, Baltimore, 1955.
1.18 Scheid, F.: "Schaum's Outline of Numerical Analysis," McGraw-Hill, New York, 1968. 5.01 Adams, E. P.: "Smithsonian Mathematical Formulas," The Smithsonian Institute, Washington, 1922.
1.19 Tod, J.: "Survey of Numerical Analysis," McGraw-Hill, New York, 1962.
5.02 Bromwich, T.J.: "Introduction ofthe Theory ofInfinite Series," Macmillan, London, 1926.
1.20 Turna, J. J.: "Engineerng Mathematics Handbook," 3d ed., McGraw-Hill, New York, 1987.
5.03 Erdclyi, A: "Asymptotic Expansions," Dover, NewYork, 1956.
1.21 Turna, J. J.: "Handbook of Physical Calculations," 2d ed., McGraw-Hill, New York, 1983.
5.04 Fort, T.: "Infinite Series,"Oxford, New York, 1930.
1.22 Turna, J. J.: "Technology Mathematics Handbook," McGraw-Hill, New York, 1975.
5.05 Jolley, L. B.: "Summation of Series," 2ded.,Dover, New York, 1948.
1.23 Young, D., and R. T. Grcgory: "A Survey of Numerical Mathematics," vols. I, II, Addison-Wesley,
5.06 Knopp, K.: "Theory and Application of Infinite Series," Blackie, London, 1963.
Reading, Mass., 1972. 6.07 Smail, L. L.: "Elements ofthe Theory ofInfinite Processcs," McGraw-Hill, New York, 1923.
C.2 NUMERICAL CONSTANTS C.6 ALGEBRAIC AND TRANSCENDENTAL EQUATIONS
2.01 Artin, E.: "The Gamma Function," Holt, New York, 1964. 6.01 Acton, F. C: "Numerical Methods ThatWork,", Harper & Row, New York, 1970.
2.02 Beckmann, P.: "A History on," 3d ed., St. Martin's New York, 1974. 6.02 Bairstow, L.: "Investigations Rclating to the Stability ofthe Aeroplane," Repon and Memoranda No. 154
AEA, Washington, D.C., 1914.
2.03 Dwigrft, H. B.: "Tables of Integris andOther Mathematical Data," 4thed., Macmillan, New York, 1961.
2.04 Jahnke, E., and F. Emden: "Tablcs of Functions," Dover, New York, 1945.
6.03 Dickson, L. E.: "Algcbraic Theories," Dover, New York, 1959.
2.05 Losen, F.,and F. Schoblik: "Die Fakutat und Vervandte Funktionen," Teubner, Leipzig, 1962.
6.04 Housholdcr, A.S.: "Principies ofNumerical Analysis," McGraw-Hill, New York, 1953.
2.06 Niclscn, N.: "HandbuchderTheorieder Gammafunktion," Teubner, Leipzig, 1906.
6.05 Ropa], Z.: "Numerical Analysis," Wilcy, NewYork, 1955.
2.07 Perron, O.: "Die Lehrc von Kettenbrchen," vols. I, II, Teubner, Stuttgart, 1957.
6.06 Kunz, K. S.: "Numerical Analysis," McGraw-Hill, New York, 1957.
2.08 Sibagaki, W.:"Theoiy andApplications of the Gamma Function," Iwanami Syoten, Tokyo, 1952.
6.07
^vlI'JL*" ThC Eva,uaton of Zcros of Hh DcSne Polynomials, Pkilos. Trtms. R. Soe. Lmdon, Ser. A.,
2.09 Titchmarsh, E. C: "The Zeta Function ofRiemann," Cambridge, London, 1930. 6.08 Upensky, J. V.: "Theory of Equations," McGraw-Hill, New York, 1948.
2.10 Wahl, H. S.: "Analytical Theory of Continued Fractions",Van Nostrand, New York, 1948. 6.09 Zurmhl, R.: "Praktische Mathematik," 5thed., Springer, Berln, 1965.
C.3 NUMERICAL DIFFERENCES
C.7 MATRIX EQUATIONS
3.01 Boole, G.: "Finite Differences," 3d ed., Stechert, New York, 1931.
3.02 Fort, T: "Finite Diflerences and Difference Equations," Oxford, New York, 1948.
7.01
E?TT' KP^'I *nd'''J' Tuma: Influence Uncs for Continuous Beams, OUa., State Urna. Agri. Appt. Sci., En?.
3.03 Jordn, C: "Calculus of Finite Diflerences," Chelsea, New York, 1947. 7.02
l^^^Jh-^, ?' .S?" "ComPutd<"> Solution of Linear Algebraic Systems," Prentice-Hall,'
3.04 Kunzmann, J.: "Mthodes Numriques," Dunod, Paris, 1959. Englewood Clifls, NJ., 1967.
3.05 Milne, W. E.: "Numrica!Calculus," Princeton, Princeton, NJ., 1949.
7.03 Gantmacher, F. R.: "Matrix Theory," vols. I, II, Chelsea, New York, 1950.
3.06 Milne-Thompson, L. N.: "Calculus of Finite Diflerences,"Macmillan, London, 1951.
7.04 Or^a,J., and.W. Reinboldt: "Iterativc Solution ofEquations ofSeveral Variables," Academic, New York,
-.&
mmmit >RW
i i i
8.04 National Bureau of Standards: "Simultancous Linear Equations and the Dctcrmination of Eigenvalues,"
Applied Mathematics Series, no. 29, Washington, 1953. C.12 LEAST-SQUARES APPROXIMATIONS
8.05 National Bureau of Standards: "Simultancous Linear Equations and the Dctcrmination of Eigenvalues," 12.01 Clenshaw, C.W.: "ANote onthe Summation ofChebyshev Series," Malh. Tables Aids Comp., vol. 9, 1955.
Applied Mathematics Series, no. 38, Washington, 1954.
12.02 Davis, P.J.: "Intcrpolation and Approximation," Dover, New York, 1975.
8.06 Willcinson, J.: "The Algcbraic Eigenvalue Probiems,", Oxford, London, 1965.
12.03 Fox, L.,and I. B. Parker: "Chebyshev Polynomials in Numerical Analysis," Oxford, London, 1968.
C.9 SERIES OF FUNCTIONS 12.04 Lanczos, C: "Applied Analysis," Prentice-Hall, Englewood Clifls, N.J., 1956.
9.01 Calson, B., and M. Goldstcin: "Rational Approximation of Functions," I-os Alamos Scicntific Laboratory 12.05 Lawson, C. L., and R. J. Hanson: "Solving Least-Squares Probiems," Prentice-Hall, Englewood Clifls,
N.J., 1974.
LA-1943, Los Alamos, N. Mex., 1955.
9.02 Clenshaw, C. \V.: Polynomial Approximations to Elementary Functions, Math. TabUs Aids Comp vol 8 12.06 Powell, M.J.: "Approximation Theoryand Methods," Cambridge, London, 1981.
1954. 12.07 Rivlin, T. K.: "An Introduction to the Approximation of Functions," Dover, New York, 1981.
9.03 Hastings, C: "Approximations for Digital Computors," Princeton Univ. Press, Princeton, N.J., 1955.
9.04 Mangulis, V.: "Handbook or Series," Academic, New York, 1965. C13 FOURIER APPROXIMATIONS
9.05 Meinardus, G.: "Approximations of Functions," Springer, New York, 1967. 13.01 Carslow, H. S.: "Theory of Fouricr Series and Integris," Dover, New York, 1930.
9.06 Ralston, A., and H. F. Wilf (cds.): "Mathematical Methods for Digital Computers," vol. I, Wilcy New 13.02 Churchill, R. V.: "Fouricr Series and Boundary Valu Probiems," 2ded., McGraw-Hill, New York, 1963.
York, 1960.
9.07 Rice,J. R.: "Approximations of Functions," Addison-Wesley, Reading, Mass., 1964.
13.03 Franklin, P.: "Introduction toFourier Methods and the Laplace Transforms," Dover, New York, 1958.
9.08 Schwartz,I. J.: "Introduction to the Operations with Series," Chelsea, New York, 1924.
13.04 Jackson, D.: Fouricr Series and Orthogonal Polynomials, Cana Math. Monogr. no. 6, Mathematical
Assodation of America, 1941.
13.05 Rogosinski, W.: "Fourier Series," Chdsca, New York, 1950.
CIO SPECIAL FUNCTIONS
13.06 Tolstov, G. P.: "Fourier Series," Dover, New York, 1976.
10.01 Andrews, L. C: "Special Functions," Macmillan, New York, 1985.
13.07 Zygmund, A.: "Trigonomctric Series," Dover, New York, 1955.
10.02 Alien, E. E.: Analytical Approximations, Math. Tables Aids Comp., vol. 8, 1954.
10.03 Alien, E. E.: "Polyrjomial Approximations to Some Modified Bessel Functions," Math. Tables Aids Comb,
vol. 10, 1956. C.14 LAPLACE TRANSFORMS
10.04 Bell, W. W.: "Special Functions," Van Nostrand, London, 1968. 14.01 Churchill, R. V.: "Operational Mathematics," 2d cd., McGraw-Hill, New York, 1958.
10.05 Carlson, B. C: "Special Functions of Applied Mathematics," Academic, NewYork, 1977. 14.02 DiStcfano III, J. J., A. R. Stubbcrud, and I. J. Williams: "Schaum's Oudinc of Fccdback and Control
Systems," McGraw-Hill, New York, 1967.
10.06 Erdclyi ct al.: "Higher Transcendental Functions, ' vols. I to III, McGraw-Hill, New York, 1955.
10.07 Klein, F.: "Vorlcsungcn ber die Hypergcomctrisc 4c Funktion," Teubner, Berln, 1933. 14.03 Doetsdi, G.: "Thcoric und Anwendung der Laplace-Transformation," Springer, Bcrlin, 1937.
10.08 Lebedev, N. N.: "Special Functions and Their Ap;>lications," Dover, New York, 1972. 14.04 Doctsch, G., H. Knicss, and D. Voelkcr "Taballen zur Laplace-Transformation and Anldtung zum
Gebrauch," Springer, Bcrlin, 1947.
10.09 Magnus, W., F. Oberhettingcr, and R. P. Soni: "Formulas and Theorems of the Spcdal Functions of .
Mathematical Physics," Springer, New York, I96C 14.05 Erdclcyi, A, W. Magnus, F. Oberhcttingcn, and F. Tricomi: "Tablcs of Integral Transforms." vols I II
McGraw-Hill, New York, 1954.
10.10 McLachlan, N. Wj; "Bessel Functions for Enginecs," Oxford, London, 1955.
14.06 Spiegel, M. R.:"Schaum's Oudinc of Laplace Transforms," McGraw-Hill, New York, 1965.
10.11 Oberhettingcr, F., and W. Magnus: "Anwendum der Elliptischea Funktionen in Physk und Tcchnik," >
Springer, Berln, 1949. 14.07 Widdcr, D. V.: "TheLaplace Transform," Princeton Univ. Press, Princeton,' NJ., 1941.
^ "1 1 D ' r_r^i
Beam-column, 15.22 Candela (cd, SI unit, luminous Ddu (6, A, Greek letter), B.10 Drvergencc: Euler's constant y. Finite series (Cot.):
Beam deflection, 15.10 intensity), B.11 Derivative(s): produets, 2.05 definition, 2.19 powersof fractions, 5.15, S.I6
Beam vbration: Cany-over mcthod, 7.14 dreular (trigonometric) series 2.01 table, frontcndpaper powers ofintegers, 5.11,5.12
by Laplace transforms, 16.15 Cataln constant G= 2(2), 2J27 functions, 9.07 Divideddiflerences, 3.01, S.09 Euler's method, 15.27 produetsof fractions, 5.25,5.26
by separation ofvariables, 16.09 Cauchy's test ofconvergence, 2.02 definition, 3.05 Double factorial (in), (2n-1)!!: Everett'sinterpolation: producs ofnumbers, 5.27,5.28
Bcrnoulli number Bm: Cayley-Hamilton's theorcm, 8.04 by diflerences, 3.20,3.21 definition, 2.30 application, 3.14 sinc terms, 550
definition, 2.22 Cdsius degrec(*C,MKS unit, exponential functions, 9.05 table, A.03 formula, 3.11 transcendental terms, 532
taUe, A.10 temperature), B. 14 Fourier series, 13.14 Dync(MKS unit, forc), B.15 Exa-(SI prefix, E= 1018), B.10 transformations, 5.07
Bcrnoulli porynomial Bm(x): Central diflerence 6: higher (su Highcr dcrivatives) Exponential function *', a", 9.05 First-orderordinary diflerential
definition, 2.23 definition, 3.01 hyperbolic functions, 9.08 Exponential integris, Ei(x), cquation:
table, A.09 operations, 3.02 Laplace transforms, 14.02 Eigenvalu: (x): Adams-Bashforth-Moulton
Bessd functions: rdations, 3.03 logarithmic functions, 9.06 definition, 8.01 definition, 10.01 method, 15.29
ordinary,,/,!*), YAx): table, 3.04 power series, 9.02 diflerence equations, 8.11 graphs, 10.02 Euler's method, 15.27
dcfinitions, 10.08 Central mcans pt: Determinant solution (equations), diflerential equations, 15.39, numerical evaluation, 10.03 Hamming's method, 15.29
graphs, 10.09.10.10 definition, 3.01 7.02 15.40 table, A14 Laplace transform method,
numerical evaluation, 10.11 operations, 3.02 Diagonal matrix, 7.02 methods, 8.05 to 8.11 Extrapolation,4.06 15.02, 15.03
tables, A16.A.I7 relations, 3.03 Diflerentialequation: properties, 8.02
predictor-corrector method,
modified, /,(x), K,(x): Chebyshev approximation, 12.06 ordinary: theorems, 8.04
15.30
definitions, 10.13 Chebyshev functions T,(x), U,(x), dassification, 15.01 Eigenvector Factortheorcm, 6.13 Runge-Kutta method, 15.28
graphs, 10.15 11.14 dosed-form solution, 15.01 definition, 8.02 Factorial ( )!: transport method, 15.31
numerical evaluation, 10.16 Chebyshev integration formula, definition, 15.01 properties, 8.03 definition, 2.09 Foot (ft, FPS unit, length), B.15
tables, A.18, A19 4.13 dgenvalucs, 15.39 to 15.41 Elastic foundation, 15.17 gamma function, 2.28 Forccd vibratlon, 15.07
spherical,j',(x),_r,(x): Chebyshev polynomials: finite-diflerence methods and Electron charge e, back cndpaper series, 5.17, 5.18 Forward diflerences A (su Finite
dcfinitions, 10.12 regular T,(x), /,(x), 11.15 solutions, 15.34 to 15.40 Electron mass mt, back cndpaper table, A.03 diflerences)
graphs, 10.12 shifted T?(x), 12.07 first-order (su First-order Ellipticequations, 16.01,16.07 Factorial polynomial X^"*, Al-"': Fourierapproximations:
trigonometric formulas, 10.12 Chi (X, X, Greek letter), B.10 ordinary diflerential Elliptic integris E(k, <f>), F(k, <f>), definition, 2.11 coeffiricnts, 13.03
sphercal modified,i,(x), *,(x): Cholcsky mcthod, 7.08 equation) n(k,a,4>): operations, 2.12,2.30 definition, 13.01
definitions, 10.17 Circular mili (FPS unit, arca), Fouricr series solution, 15.02 ' definition, 10.05 series, 5.17, 5.18 even functions, 13.04
graphs, 10.17 back cndpaper fourth-order equations, 15.14, numerical evaluation, 10.06 Factorization theorcm, 6.13 exponential form, 13.12
trigonometric formulas, 10.17 Comparison test ofconvergence, 15.19,15.23,15.37 table, A20 Fahrenhdt degree (*F, FPS unit, odd functions, 13.05
Bessd's integration formula, 4.09 2.02 Laplace transform solutions, English system (U.S.customary): temperature), B.13 rectangular functions, 13.06,
Bcssel's intcrpolation formula, Conditionally convergent: 15.03 to 15.24 conversin factors, B.14 Farad (F, SI unit, capadtance), 13.07
3.12 produc, 2.05 methods ofsolution, 15.01 units, B.12 B.ll
trapezoidal functions, 13.10
Beta (B, B, Greek lener), B.10 series, 2.01 multistep methods, 15.29 to Epsilon (e, E, Greekletter), B.10 Faradayconstant 9, back trigonometric functions, 13.11
Beta function B(m,), 2.31 Congrucnce, 8.03 15.31 Equations: cndpaper Fourier series:
Biharmonic equations, 16.24 Continued fractions, 2.07 numerical solutions, 15.25 algcbraic,6.01, 6.07,6.16 Fibonacd number, 2J2.X concept, 13.01
Convergence: second-order equation (ser dgenvalue, 8.01 Fikm'sintegration,4.18
Binomial cocfficknt I j: ofproduct ofconstant terms, Second-order ordinary matrix, 7.01 Finitediflerences A, V, 6:
operations, 13.14
definition,2.09 ' Fourth-order ordinarydiflerential
2.05 differential equation) ordinarydiflerential(su applications:
product, 2.15 equations:
ofseries ofconstant terms, 2.02 series solutions, 11.01 to 11.07 Differential equation, equations, 15.34to 15.40, finite-diflerence method, 15.37,
series, 2.15,2.16
ofseries of functions, 9.01,9.02 single-stcp methods, 15.26 to ordinary) 16.16 to 16.27
Binomial equations, 6.02 15.38,15.40
Convergent, 2.07 . 15.28,15.33 partialdifferential(su interpolation, 3.07 to 3.19
Binomial function: Laplace transform method,
Convolution integral, 15.04 systems ofequations, 15.31 to , Differential cquation, series, 9.03
definition, 2.11 15.14 to 15.24
Cosme circular (eos), 9.07 15.33 ^ partial) definition, 3.01
operations, 2.12 Fresnd integris,10.04
Cosine hyperbolic (cosh), 9.08 transport method, 15.31,15.32 transcendental, 6.07 error, 3.05
relation to gamma function, 2.30
Cosine integris, 4.17,9.07 partial: Error function erf (*): operations, 3.05
Binomial series:
Cosine series: dassification, 16.01,16.02 '. definition, 10.01 operators, 3.02
particular cases, 2.14 Galln (FPS unit, volunte), B.15
finite, 5.32 definition, 16.01 ; \ graph, 10.02 rdations, 3.03,3.20 to 3.22
relations, 2.13 Gamma (y, I\ Greek letter), B.10
infinite, 5.33 fimte-duTerencc mcthod and A table, A.15 tables, 3.04
Biscction mcthod, 6.08 Gamma function T(x):
Cramer's rule, 7.02 solutions, 16.16 to 16.28 . A Eta (n, H, Greekletter), B.10 Finitc series:
Bohr magneton ftB, back cndpaper definition, 2.28
Cubic equation, 6.03 hyperbolic, 16.01,16.20 Euler-MacLaurin formula: approximation sum, 5.04
Boundary-value problem: numerical evaluation, 2.29
Laplace transform method :-~; integration, 4.10 arithmetic, 5.08
ordinary difierential equations, table, A.03, A.04
Danidvsky's method, 8.08 and solutions, 16.12 to seriessuremation, 5.05 arithmogeometric, 5.09,5.10
15.25 Gauss-Chebyshevquadrature, 4.14
Decadic (common, Brigg's) 16.16 Euler number ,: binomial coefficients, 2.15, 2.16
partial diflerential equations, GausseUmination method, 7.06
logarithm (log), 9.06 methods ofsolution, 16.02 definition, 2.24 classification, 5.01
16.01 Gauss-Legcndrc quadrature,4.12
Dea-(SI prefix, d = 10_l),B.12 separation-of-variables '. table, A.12 cosine terms, 5.31 Gaussquadratures,4.11
Brauer' theorcm, 8.04
Definte integral: method and solutions, Euler poh/nonal ,(x)t dedmal numbers, 5.19 to 5.22 Gauss-Seidd method, 7.13
Brtish thermal unit (BTU, FPS
definition, 4.01 16.03 to 16.11 definition, 2.25 definition, 5.01 Geometric series:
unit, encrgy), B.15 table, A.11
numerical methods, 4.03 Digarama function S>(x): exact sum, 5.03 baste case, 5.08
properties, 4.02 definition, 2.32 Euler's constant f. ( factorial polynomials, 5.17,5.18 finitealternating, 5.30
Calorie(cal, MKS unit, encrgy), Defiation theorcm, 6.13 numerical evaluation, 233 definition, 2.18 geonactriof&OB' . : - finitcrnonotonicT 5.29 ,.
B*l& "Degree (angte)7lrotndpaper ~taoTerA.05~ table,fiont endpaper harmonic,5.I3.-5.14',:., ... -^ ... infinh^SJSS^^^'^di:!
^Sllt)^
^ '~^1 ' "1 ~H V":H f^=^
^ ^ m
15.05 Frank,R., und R. V. Mises: "Die Diflerential und Integral Glechungcn der Mechanik und Physik," vols I
II, Dover, New York, 1961.
15.06 Goldbcrg, S.: "Introduction to Diflerence Equations," Dover, New York, 1986.
INDEX
15.07 Grintcr, L. E. (ed.):"Numerical Methods of Analysis in Engincering," Macmillan, New York, 1949.
15.08 Henrid, P.: "Discrctc Variable Methods in Ordinary Diflerential Equations," Wilcy, New York, 1962.
15.09 Hildebrand, F. B.: "Finite-diflerence Equations and Simulations," Prentice-Hall, Englewood Clifls N I
1968.
15.10 Incc, E. L.: "Ordinary Diflerential Equations," Dover, New York, 1956.
15.11 Kamkc, E.: "Differential Gleichungen," 3d cd., Chelsea, New York, 1948. References are made to section numbers for the text material. Numcrs preceded by the letters Aand B
15.12 Lcvy, H., and E. A. Baggolt: "NumericalSolutions of Differential Equations," Dover, New York, 1950. refer to the Appcndixcs. In the dcsignation ofunits and systcms ofunits the following abbrcviations are
used:
15.13 Milnc, W. E.: "Numerical Solution of Differential Equations," Dover, New York, 1970.
avd = avoirdupois thm = thermochemical
liq = liquid try = troy(apothccary)
C.16 PARTIAL DIFFERENTIAL EQUATIONS
nat = nautical FPS = English system
16.01 Ames, W. F.: "Numerical Methods for Partial Differential Equations," Academic, New York, 1977. phs = physical IST = intcrnational steam tablcs
16.02 Batcman, H.: "Partial Differential Equationsof Mathematical Physics,"Cambridge, London, 1959. stu = statute MKS = metric system
tec = tcchnical SI = intcrnational system
16.03 Bcndcr, L., and S. Orzag: "Advanced Mathematics for Engineers," McGraw-Hill, New York, 1978.
16.04 Carricr, G. F., and C. E. Pearson: "Partial Difierential Equations," Academic, New York, 1976.
16.05 Duffy, D. G.: "Solution of Partial Differential Equations," TAB, Bluc RidgeSummit, Pa., 1986.
16.06 Forsythc, G. E., and W. R. Watson: "Finitc Difference Methods for Partial Differential Equations," Wilcy Abel's transformation, 5.07 Approximation of constants Asymptotic series in evaluation:
New York, 1960. Absolutcly convergent: (Cont.): of constants, 5.05
powcr series, 9.02 by produets, 2.17 integris, 4.10
16.07 Gladwell, I., and R. Wait: "Survey of Numerical Metods for Partial Differential Equations," Oxford New
York, 1979. product of constants, 2.04 Approximation of functions: Atmosphere (atm, MKS unit,
series of constants, 2.01 by continued fractions, 9.10 to pressure), B.15
16.08 Havner, K. S.: "Algcbriac Carry-Over in Two-Dimcnsional Systems," dissertaiton, Oklahoma State series of functions, 9.01
University Library, Stillwater, 1959. 9.13 Atomic mass unit u, back
Acre (FPS unit, arca), B.15 by finitc diflerences, 3.07, 9.03 cndpaper
16.09 Jain, M. K.: "Numerical Solution of Differential Equations," Halstcad, New York, 1984. Adams-Bashforth-Moulton by Fouricr series, 13.01 Atto- (SI prefix, a = 10-IB), B.12
16.10 Kardcstuncer, H., and D. H. Norric (cds.): "Finitc Elcment Handbook," McGraw-Hill, New York, 1987. mcthod, 15.29 by least squares, 12.01 Auxiliary Bcrnoulli numberBm:
Algcbraic equations: by powers series, 9.03 definition, 2.22
16.11 Kcttcr, R L., and S. P. Prawel, Jr.: "Modera Methods of Engincering Computation," McGraw-Hill New
York, 1969. definition, 6.01 by rational polynomial, 10.03 table, A.10
roots, 6.16 Arccosinc drcular (arceos,eos-1), Auxiliary Euler number Em:
16.12 Miller, K. S.: "Partial Differential Equations in Engincering Probiems," Prentice-Hall, Englewood Clifls, Algcbraic polynomials: 9.11 definition, 2.24
N.J., 1953.
operations, 6.15 Arccosinc hyperbolic (arecosh, table, A12
16.13 Odcn,J. T.: "Finite Elcmcnts in Nonlincar Continua," McGraw-Hill, New York, 1972. properties, 6.13 cosh-1), 9.12 Avogadro constant A", back
16.14 Panov, D. J.: "Formulas for Numerical Solution of Partial Differential Equations by the Method of synthctic divisin, 6.14 Archimedcs constant t: cndpaper
Diflerences," Ungar, New York, 1963. Alpha (o-, A, Greek letter), B,10 definition, 2.17
16.15 Salvadori, M. G., and M. L. Barn: "Numerical Methods in Engincering," Prentice-Hall, Englewood Clifls, Alternating product, 2.04 table, front cndpaper Backward difference V:
NJ., 1961. Alternating series, 2.01 Arcsinc drcular (aresin,sin-1), definition, 3.01
16.16 Shaw, F. S.: "An Introduction to Rclaxation Methods," Dover, New York, 1953.
Angstrom (, MKS unit, length), 9.11 operations, 3.02
B.15 Arcsinc hyperbolic (aresinh, relations, 3.01
16.17 Soare, M.:"Application of Finitc Difference Equations to Shell Analysis," Pcrgamon, London, 1967. Antisymmetrical equations, 6.06 sinh-1), 9.12 table, 3.04
16.18 Sommerfdd,A: "Partial Differential Equationsof Physics," Academic, New York, 1949. Antisymmctrical functions: Arctangent ireular (arctan, Bairstow's mcthod, 6.18
Fourier series, 13.05 tan- 1,9.13 Banded matrix, 7.05, 7.10
16.19 Southwdl, R V.: "Rclaxation Methodsn Engineerng Srience," Oxford, New York, 1940.
Laplace transforms, 14.17 Arctangent I yperbolic (arctanh, Bar (b, MKS unit, pressure), B.15
16.20 Tikhovov, A N., and A. A.Samarski: "Equations of Mathematical Physics," Macmillan, New York, 1963. Approximate sum, 5.05 tanh '), 9.18 Baseof natural logarithm e:
16.21 Turna,J. J.: "Handbookof Structural and Mcchanical Matrices," McGraw-Hill, New York, 1988. Approximation of constants: Arithmcric s -res, 5.08 definition, 2.18
16.22 Vladimirov, V. S.: "Equations of Mathematical Physics," Dekker, NewYork, 1971. by asymptotic series, 2.29, 2.33, Arithmogeo etric series: table, front cndpaper
2.35 alternating, 5.10 Basic units:
16.23 Zicnkicwicz, O. C: "The FiniteElement Method," 3d cd., McGraw-Hill, New York, 1977. by fractions, 2.18 to 2.21 monotonic 5.09 FPS system, B.13
by powcr series, 2.17 to 2.22, Astronomical unit (MKS unit, MKS system, B.14
2.24, 2.26, 2.27 length ,B.I5 SI system, B.ll