Applications of Potential Theory in Mechanics: Selection of New Results

APPLICATIONS
OF POTENTIAL THEORY
IN MECHANICS
Selection of new results
V.I. Fabrikant
To my parents

1
PREFACE
It is not easy to find something new in mathematics. It is difficult to

find something new in something very old, like the Potential Theory, which was
studied by the greatest scientists in the past centuries. And it is extremely
difficult to make this find on an elementary level, with no mathematical
apparatus involved which would be considered new even in the times of, say,
Poisson. This is exactly what the author claims to have done in this book: a
new and elementary method is described for solving mixed boundary value
problems, and their applications in engineering. The method can solve
non-axisymmetric problems as easily as axisymmetric ones, exactly and in closed
form . It enables us to treat analytically non-classical domains. The major
achievements of the method comprise the derivation of explicit and elementary
expressions for the Green’s functions, related to a penny-shaped crack and a
circular punch, development of the Saint-Venant type theory of contact and crack
problems for general domains , and investigation of various interactions between
cracks, punches, and external loadings. The method also provides, as a bonus, a
tool for exact evaluation of various two-dimensional integrals involving distances
between two or more points. It is believed that majority of these results are
beyond the reach of existing methods.
For over twenty years, since being a graduate student in Moscow, the
author had been perplexed by an inconsistency between various solutions to the
problems in Potential Theory and the way those solutions have been obtained,
namely, that the solution was quite elementary, while the apparatus used was
very complicated, involving various integral transforms or special functions
expansions, which are beyond the comprehension of an ordinary engineer. The
author’s search for a new method was based on the conviction that an
elementary result should be obtainable by elementary means. The method has
been found and is presented here in detail. One may just wonder why the
method was not discovered at least a century ago.
2
3
The book is addressed to a wide audience ranging from engineers, involved

in elastic stress analysis, to mathematical physicists and pure mathematicians.
While an engineer can find in the book some elementary, ready to use formulae
for solving various practical problems, a mathematical physicist might become
interested in new applications of the mathematical apparatus presented, and a pure
mathematician might interpret some of the results in terms of fractional calculus,
investigate the group properties of the operators used, or, having noticed the fact
that no attention is paid in the book to a rigorous foundation of the method,
might wish to remedy the situation. Due to the mathematical analogy between
mixed boundary value problems in elasticity and in other branches of engineering
science, the book should be of interest to specialists in electromagnetics,
acoustics, diffusion, fluid mechanics, etc. Though several such applications have
been published by the author, the space considerations did not allow us to
include them, but references are given at appropriate places in the book.
The book is accessible to anyone with a background in university
undergraduate calculus, but should be of interest to established scientists as well.
Though the method is elementary, the transformations involved are sometimes very
non-trivial and cumbersome, while the final result is usually very simple. The
reader who is interested only in application of the general results to his/hers
particular problems may skip the long derivations and use the final formulae
which require little effort. The reader who wants to master the method in order
to solve new problems has to repeat the derivations which are given in sufficient
detail. The exercises are important in this regard. They vary from very simple
to quite difficult. Some can be used as a subject for a graduate degree thesis.
The book is based entirely on the author’s results, and this is why the
work of other scientists is mentioned only when such a quotation is inevitable
for some reason, like numerical data needed to verify the accuracy of
approximate results, comparison with existing results, or pointing out some errors
in publications. There are several books and review articles presenting an
adequate account of the state-of-the-art in the field. Appropriate references are
given for the reader’s convenience. The purpose of this book was neither to
repeat nor to compete with them.
The development of the method can by no means be considered completed,
this is just a beginning. The results presented in the book may be compared to
the tip of an iceberg, taking into consideration numerous applications which are
still to come. The solution of fundamental problems in a simple form enables
us to consider various more complicated problems which were not even attempted
before. The method can be expanded to spherical, toroidal and other systems of
coordinates, so that more complex geometries may benefit from it. The method
proved useful in the generalized potential theory as well. Some of these results,
though already published by the author, could not be included in the book due
to severe restrictions on the book volume.
4
For the reader’s convenience, it was attempted to make each chapter (and
section, wherever possible) self-contained. The reader can skip several sections
and continue reading, without loosing the ability to understand material. On the
other hand, this resulted in repetition of some definitions and descriptions. The
author thinks that the additional convenience is worth several extra pages in the
book.
The author is grateful to Professor J.R. Barber from Michigan, Professor
B. Noble from England, and Professor J.R. Rice from Harvard who agreed to
read the manuscript and expressed their opinion.
The book contains so much new material that some misprints and errors are
inevitable, though every effort was made to eliminate them. The author would
be grateful for every communication in this regard. All the readers’ comments
are welcome. The address is
V.I. Fabrikant
prisoner # 167 932D
Archambault jail
Ste-Anne-des-Plaines
Quebec, Canada J0N 1H0
CONTENTS
INTRODUCTION 7
CHAPTER 1. DESCRIPTION OF THE NEW METHOD

1.1 Integral representation for the reciprocal
of the distance between two points 15
1.2 Properties of the L-operators 22
1.3 Some further integral representations 25
1.4 Internal mixed boundary value problem for a half-space 31
1.5 External mixed boundary value problem for a half-space 43
1.6 Some fundamental integrals 63
CHAPTER 2. MIXED BOUNDARY VALUE PROBLEMS

IN ELASTICITY
2.1 General solution 73
2.2 Point force solutions 77
2.3 Internal mixed problem of type I 81
2.4 External mixed problem of type I 92
2.5 Integral representation for q 2/ R 3 101
2.6 Internal mixed problem of type II 105
2.7 External mixed problem of type II 111
2.8 Inverse crack problem in elasticity 129
CHAPTER 3. MIXED-MIXED BOUNDARY VALUE PROBLEMS

3.1 General formulation of the problem 142
3.2 Internal axisymmetric mixed-mixed problem 145
3.3 External axisymmetric mixed-mixed problem 171
3.4 Generalization for non-homogeneous half-space 177
3.5 Effect of a shearing force and a tilting moment
on a bonded circular punch 182
3.6 Non-axisymmetric internal mixed-mixed problem 188
3.7 Non-axisymmetric external mixed-mixed problem 200
Appendix A3.1 207
Appendix A3.2 211
Appendix A3.3 215
5
6
CHAPTER 4. APPLICATIONS IN FRACTURE MECHANICS

4.1 Flat crack under arbitrary normal loading 220
4.2 Point force loading of a penny-shaped crack 227
4.3 Concentrated load outside a circular crack 229
4.4 Plane crack under arbitrary shear loading 231
4.5 Penny-shaped crack under uniform pressure 243
4.6 Penny-shaped crack under uniform shear loading 248
4.7 Asymptotic behavior of stresses and displacements
near the crack rim 252
4.8 Flat crack of general shape 259
4.9 General crack under uniform shear 273
4.10 Close interaction of pressurized coplanar circular cracks 275
4.11 Close interaction of coplanar circular cracks
under shear loading 292
Appendix A4.1 307
Appendix A4.2 311
Appendix A4.3 312
Appendix A4.4 317
CHAPTER 5. APPLICATION TO CONTACT PROBLEMS

5.1 Contact problem for a smooth punch. 324
5.2 Flat centrally loaded circular punch 328
5.3 Inclined circular punch on an elastic half-space 332
5.4 Flat punch of arbitrary planform under the action
of a normal centrally applied force 336
5.5 Inclined flat punch of general shape 347
5.6 Curved punch of general planform 367
5.7 Flat flexible punch of general planform
under shifting load 387
5.8 Reissner-Sagoci problem for general domains 393
5.9 Interaction between punches subjected to normal pressure 402
5.10 Interaction between flexible punches under shifting loading 412
5.11 Contact problem for a rough punch 422 Appendix A5.1
425
REFERENCES 439
SUBJECT INDEX 447

INTRODUCTION
A short survey of methods for solving mixed boundary value problems of

potential theory is given, and some of their limitations are pointed out. The
necessity for developing a new method is justified. A concise description of
each chapter is given for the reader’s convenience.
Various applications of potential theory in electrostatics, heat transfer,
elasticity, diffusion, and other branches of engineering science are well known,
and have attracted significant attention of scientists like Laplace, Poisson, Green,
Beltrami, Kirchhoff, Lord Kelvin, Hobson, and others who made a significant
contribution to the field during past centuries. The boundary value problems
with mixed conditions are the most difficult to solve, and, at the same time,
they are among the most important in various engineering applications. Recall
the celebrated problem of an electrified circular disc. It is next to impossible
even to name every scientist who solved the problem by one method or another.
We can specify two categories of methods. The first one requires construction
of the Green’s function, after which each particular solution can be presented in
quadratures. The second category encompasses various integral transform methods.
Let us briefly discuss both categories.
Hobson (1899) has constructed the Green’s function for a circular disc and
a spherical bowl using a method due to Sommerfeld. He has used toroidal
coordinates τ, σ, φ (in our notation), which are related to the cartesian x , y , and
z as follows:
a sinhτ cosφ a sinhτ sinφ a sinσ

x = , y = , z = .
coshτ − cosσ coshτ − cosσ coshτ − cosσ
Here a is the disc radius. The potential function V at the point (τ0,σ0,φ0)
7
8 INTRODUCTION
external to the disc, which on the disc takes the values v (τ,φ), can be expressed
(1 + coshτ) cos(σ /2)

1 ⌠⌠ 0
V (τ0,σ0,φ0) = 2 
π ⌡ ⌡ 2 aR [cosh2(α/2) − sin2(σ0/2)]1/2
S 
z (1 − cosσ0)1/2
+ tan-1  v (τ,φ) d S , (0.1)
R3 (coshα + cosσ0)1/2

where R is the distance between the points (τ0,σ0,φ0) and (τ,σ,φ), and α is
defined by coshα = coshτ0coshτ − sinhτ0sinhτ cos(φ−φ0). The practical value of
expression (0.1) is quite limited, since there seems to be no way to evaluate the
integral involved, even in the simplest case when v (τ,φ) is constant. Hobson had
to use a very ingenious method in order to find the potential function for
v =const and v =µ x , µ=const. On the other hand, it turns out that the integral in
(0.1) is computable in elementary functions for any polynomial v (in Cartesian
coordinates). This was one of the reasons that prompted the author to look for
an alternative approach, which would be as general as (0.1), and, on the other
hand, would allow elementary and straightforward computation of the integrals
involved.
Consideration of the mathematically equivalent problem of a circular punch
pressed against an elastic half-space leads to the integral equation (Galin, 1953)
2π a
σ(ρ0,φ0)
ω(ρ,φ) = H ⌠ ⌠ ρ0dρ0dφ0. (0.2)
⌡⌡ R
0 0
Here a is the punch radius, H is an elastic constant, ω is the normal

displacement under the punch (a known function), σ stands for the pressure
exerted by the punch (an unknown function), and R is the distance between the
points (ρ,φ) and (ρ0,φ0). Leonov (1953) has obtained a closed form exact
solution of integral equation (0.2) by a very ingenious method. His result reads
in our notation
2π a
ω(ρ0,φ0)
1  ⌠⌠
σ(ρ,φ) = − ∆ ⌡ ρ0dρ0dφ0
4π2 H ⌡ R
 0 0
INTRODUCTION 9
2π a
ω(ρ0,φ0) 
2 ⌠ ⌠ R
− tan-1( )
R
+ ρ0dρ0dφ0. (0.3)
π ⌡ ⌡ η η R3
0 0 
Here η=[( a 2−ρ2)( a 2−ρ20)]1/2/ a and ∆ is the two-dimensional Laplace operator. One
can observe the same handicap in (0.3): difficulty to evaluate the integrals
directly, even in the simplest case when ω is constant. Again, it is clear that
the integrals are computable in elementary functions for any prescribed polynomial
displacement. This indicates a gap in our knowledge which needs to be filled.
The integral transform method, involving dual integral equations, was
originated, probably by Weber (1873) and Beltrami (1881), and continued by
Busbridge (1938) and others. Significant achievements in the systematic
application of the method to various problems belong to Sneddon. The reader is
referred to the books by Sneddon (1951, 1966) and Ufliand (1967) for additional
references. Some quite remarkable results were obtained by Ufliand (1977).
Despite this success, it has always been the author’s conviction that our use of
integral transforms generally indicates our inability to solve problems directly. To
this end, two illustrative examples are presented.
Here is how the problem of a circular disc, charged to a potential
v 0=const, is solved by the dual integral equation method. It is necessary to find
a harmonic function V , vanishing at infinity, and subjected to the mixed
boundary conditions on the plane z =0:
∂V
V = v 0, for ρ≤ a ; = 0, for ρ> a . (0.4)
∂z
The solution is presented in the form
V (ρ, z ) = ⌠ A ( t )e-tz J 0( t ρ)
dt
. (0.5)
⌡ t
0
Here J 0 is the Bessel function of zero order, and A ( t ) is the as yet unknown
function which should be chosen to satisfy (0.4). Substitution of the boundary
conditions (0.4) in (0.5) leads to the dual integral equations
10 INTRODUCTION
⌠ A ( t ) J ( t ρ) d t = v , for 0≤ρ≤a ;
⌡ 0 t 0
0
⌠ A ( t ) J ( t ρ)d t = 0, for ρ> a . (0.6)

⌡ 0
0
By using the discontinuous Weber-Schafheitlin integrals, one can deduce that
2
A(t) = v sin( at ). (0.7)
π 0
The solution (0.5) can now be rewritten as
∞
2 ⌠ -tz dt
V (ρ, z ) = v e sin( at ) J ( t ρ) , for z ≥0, (0.8)
π 0⌡ 0 t
0
and the charge density σ over the disc is given by
v ∞
⌠ J ( t ρ) sin( at ) d t .
0
σ(ρ) = (0.9)
π2 ⌡ 0
0
Both integrals (0.8) and (0.9) are computable in elementary functions. For
example, the last integral yields
v
0
σ(ρ) = . (0.10)
π2( a 2 − ρ2)1/2
There seems to be a discrepancy between the simplicity of the final result and
the apparatus used to obtain it. The general idea, that an elementary result
should be obtained by elementary means, calls for a search for a new and
elementary approach.
The second example comes from consideration of the simplest case of a
penny-shaped crack of radius a , subjected to axisymmetric pressure p (ρ). The
corresponding potential function f can be found in (Kassir and Sih, 1975) as
follows:
INTRODUCTION 11
f (ρ, z ) = ⌠ A ( s ) J 0(ρ s ) exp(− sz)

ds
, (0.11)
⌡ s
0
where
a t
1 ⌠
sin st d t ⌠ 2
rp(r) d r
A ( s) = − 2 1/2 . (0.12)
πµ ⌡ ⌡ (t − r )
0 0
The user has to substitute the explicit expression for p in (0.12), and to evaluate
two consecutive integrals. The result is to be substituted in (0.11), and an
infinite integral with Bessel function is to be evaluated. It seems natural to try
to spare one integration by substituting (0.12) in (0.11), changing the order of
integration and evaluating the integral (Gradshtein and Ryzhik, 1963, formula
6.752.1):
⌠ sinst J (ρs) exp(−sz) d s = sin-1 t , (0.13)

⌡ 0 s l (t)
2
0
where the notation
1
l (t) = {[(ρ + t )2 + z 2]1/2 + [(ρ − t )2 + z 2]1/2}. (0.14)
2 2
was introduced. Formulae (0.11) and (0.12) can be combined to give
a t p ( ρ ) ρ dρ
1 ⌠ -1 t
dt ⌠ 2
0 0 0
f (ρ, z ) = − sin . (0.15)
πµ ⌡ l (t) ⌡ ( t − ρ0)
2 1/2
2
0 0
Note that expression (0.15) contains no trace of the integral transform, and thus
gives us a hint that a direct and elementary solution is indeed possible. For
example, in the case of a uniform loading, p =const., and the potential function
(0.15) will take the form:
a
p ⌠ t
f (ρ, z ) = − t sin-1 dt. (0.16)
πµ ⌡ l (t)
2
0
12 INTRODUCTION
The integral in (0.16), though looking formidable due to (0.14), can be evaluated
in elementary functions, namely,
p  2
(2 a + 2 z 2 − ρ2)sin-1( ) + l (ρ2 − l 21)1/2 − 2 z ( a 2 − l 21)1/2.
a
f (ρ, z ) = −
4πµ l 1 
2
(0.17)
Here the abbreviations l and l stand for l ( a ) and l ( a ) respectively, with l
1 2 1 2 2
defined by (0.14), and
1
l (t) = {[(ρ + t )2 + z 2]1/2 − [(ρ − t )2 + z 2]1/2}. (0.18)
1 2
One can show that arbitrary polynomial loading in (0.15) will lead to an
elementary expression for the potential function, and thus to an elementary
complete solution.
The situation can now be summarized. The Green’s function approach is
the most general, the main impediment being the inability of direct derivation of
results which were usually constructed due to some ingenious considerations. By
contrast, the integral transform method allows a straightforward derivation of the
results, but it is the least general, since each particular problem has to be solved
from beginning to the end. The method is best suited to axisymmetric problems.
In the general case, separate solutions have to be obtained for each harmonic.
Non-axisymmetric problems involving various interactions (several arbitrarily
located charged discs, interaction of punches and cracks, etc.) are extremely
difficult to solve by the integral transform method. A new method has to be
found which would be as general as the Green’s function method, and, at the
same time, it has to be elementary and straightforward, with no integral
transforms or special function expansions involved.
When one problem has been solved by a complicated method, it is often
possible to find another method to solve the same problem more simply. The
new method would have been of little value if all it could do were to solve
more easily the already solved problems. The main advantage of the new
method presented here is its ability to solve non-axisymmetric problems as easily
as axisymmetric ones, which in turn opens up new horizons, and allows us to
solve some problems which were not even considered before, namely, the
analytical treatment of nonclassical domains, and the solution to various interaction
problems.
The general description of the method is given in Chapter 1. It starts
with a derivation of the basic integral representation for the reciprocal of the
distance between two points, followed by several generalizations. A closed form
exact solution is given to the non-axisymmetric mixed problem of potential theory
INTRODUCTION 13
for a half-space, with Dirichlet conditions prescribed inside a circle, and Neumann
conditions given on the outside, and vice-versa. Some integrals, which are of
fundamental value to the method, are evaluated in elementary functions.
Chapter 2 is devoted to the mixed boundary value problems of an elastic
half-space. The general solution is expressed in terms of three harmonic
functions. A classification of internal and external mixed problems of type I and
type II is introduced. The problems of the first type are characterized by mixed
conditions with respect to normal parameters (the pressure and normal
displacement), with the shear stress prescribed all over the boundary. These
problems are solved for a non-homogeneous half-space, with the elasticity modulus
assumed to be a power function of the depth. The case when the boundary
conditions are mixed with respect to tangential displacements and stresses, with
the normal stress being prescribed all over the boundary, is classified as the type
II problem. Each type is considered separately. Exact closed form solution and
various examples of punch and crack problems are presented.
The problem is called mixed-mixed when the boundary conditions are mixed
with respect to both normal and tangential parameters. This kind of problem is
the most difficult to solve due to the coupling of the governing integral
equations, which can no longer be solved separately. The axisymmetric and
non-axisymmetric internal and external problems are considered in Chapter 3.
The exact solution is given in terms of Fourier series expansions. A flat punch,
bonded to a transversely isotropic elastic half-space and subjected to general
loading is considered in detail. The interaction of exterior loading with the
punch is also investigated.
While the problems solved in Chapter 2 deal primarily with the stresses and
displacements in the plane z =0, the complete solution to various crack problems
is the subject of Chapter 4. The solution is called complete, when explicit
expressions for the field of stresses and displacements is defined in the whole
space. The cases of a penny-shaped crack under arbitrary normal and tangential
loadings are considered separately. Explicit expressions are derived for all the
Green’s functions involved. All the results are given in terms of elementary
functions. These solutions enable us to solve more complicated problems of
interaction of a penny-shaped crack and an exterior load. A set of non-singular
governing integral equations is derived for the interaction between coplanar
circular cracks, subjected to normal pressure and shear loading. An approximate
analytical solution is presented for the case of a flat crack of general shape,
subjected to a uniform pressure or a uniform shear. The solution is exact for
an ellipse, and is expected to be satisfactory for a wide variety of shapes. A
comparison is made with various numerical results, available in the literature, and
a very good accuracy is established in many cases. The method is less accurate
for domains with sharp angles and the aspect ratio far away from unity.
14 INTRODUCTION
A general solution in terms of one harmonic function is given in Chapter 5

to the problem of a smooth punch, penetrating a transversely isotropic elastic
half-space. The main potential function and all the Green’s functions are
expressed in terms of elementary functions for a circular punch of arbitrary
profile. It is shown that the complete solution is also presentable in elementary
functions for a general polynomial profile. The knowledge of a complete
solution, combined with the reciprocal theorem, enables us to solve more
complicated problems of interaction of exterior loadings with punches, and the
interaction between punches. The general method is applied to the analytical
treatment of nonclassical contact problems. Again, the solution is exact for an
elliptical punch, and is expected to be satisfactory for a punch of arbitrary
planform. The cases of a flat centrally and noncentrally loaded punch, and the
case of a curved punch are considered in detail. An extensive comparison is
made with the numerical results, available in the literature. As it was in the
case of crack problems, the agreement is quite satisfactory, except for the
domains with sharp angles and the aspect ratio far away from unity. An
analytical solution is given to the problem of a non-smooth punch, subjected to
normal and shear loading, with the Coulomb friction law assumed between the
punch base and the elastic half-space.
The best way to master a new method is through the exercises. Each
chapter contains a certain number of them, the majority of exercises are supplied
with an answer, a hint or a complete solution. The reader is encouraged to do
them all. The transformations involved are elementary, though sometimes very
non-trivial, and require some ingenuity.
CHAPTER 1
DESCRIPTION OF THE NEW METHOD
Some integral representations, which are of fundamental value to the

method, are derived. An exact closed form solution is given to the mixed
boundary value problem of potential theory for a half-space, with a circular line
of division of boundary conditions.
1.1 Integral representation for the reciprocal

of the distance between two points
The author has decided to start with a derivation of a new integral

representation for the reciprocal of the distance between two points located in the
plane z =0 since this quantity is very important in potential theory. Here we
repeat the derivation leading to such a representation, as it was given in
(Fabrikant 1971e). Consider the expression
1 1
1+u = , (1.1.1)
R ( ρ2 + ρ20 − 2ρρ0cos(φ − φ0))(1+u)/2
where u is a constant and -1< u <1. The standard expansion of (1.1.1) in Fourier
series will take the form
∞ 2π
in(φ-φ0) ψ
1
R 1+u
= Σ
n=-∞
e
2π
⌠ e-in dψ
⌡ ( ρ2 + ρ2 − 2ρρ cosψ)(1+u)/2
0 0 0
∞
in(φ-φ )
Σ 2πΓ[ n + (1 + u )/2]  ρ n 1 + u ρ2
0
e 1 + u
= F ( , n + , n +1; ).
2πρ1+u
0
Γ[(1 + u )/2] Γ( n + 1)ρ0 2 2 ρ20
n=-∞
(1.1.2)
15
16 CHAPTER 1, Description of the new method
Here F stands for the Gauss hypergeometric function. By using another integral
representation
1 + u 1 + u
F( , n + , n + 1; z )
2 2
1
2Γ( n + 1) ⌠ t (1 − t )
2n+u 2 -(1+u)/2
= dt,
Γ[ n + (1 + u )/2] Γ[1 − (1 + u )/2] ⌡
0 (1 − zt )
2 (1+u)/2
expression (1.1.2) can be transformed into
∞ min (ρ0,ρ)
in(φ-φ0)
R
1 2 πu
1+u = π cos 2 Σ e ⌠
(ρρ0)n ⌡
x 2n+ud x
. (1.1.3)
(ρ2 − x 2)(ρ2 − x 2)(1+u)/2
n=-∞ 0
 0

Summation in (1.1.3) finally gives
1 1
1+u =
R ( ρ2 + ρ20 − 2ρρ cos(φ − φ ))(1+u)/2
0 0
x2
λ( , φ−φ ) x ud x
min (ρ0,ρ) ρρ0 0
2 πu ⌠
= cos . (1.1.4)
π 2 ⌡
0
(ρ2 − x 2)(ρ2 − x 2)(1+u)/2
 0

Here the notation was introduced
1 − k2
λ( k ,ψ) = . (1.1.5)
1 + k 2 − 2 k cosψ
After one cumbersome derivation is finished, we can always find a way to do it

much simpler. Indeed, if we introduce a new variable
η( x ) = [(ρ2 − x 2)(ρ20 − x 2)]1/2/ x , (1.1.6)
expression (1.1.4) may be rewritten as

Integral representation for the reciprocal distance 17
∞
1 2 π u ⌠ η-udη
1+u = π cos 2 2. (1.1.7)
⌡R + η
2
R
0
The integral in (1.1.7) can be evaluated by using formula (3.241.4) from

(Gradshtein and Ryzhik, 1963), thus proving the identity. Note that parameter η
will be used throughout the book also for the case when x >max(ρ,ρ0), and
expression (1.1.6) in this case is interpreted as
η( x ) = ( x 2 − ρ2)1/2( x 2 − ρ20)1/2/ x .
One can deduce from (1.1.7) that in the particular case when u =0, the
integral in (1.1.4) can be evaluated as indefinite, and we have a very important
representation
x2
λ( , φ−φ0) d x
ρρ0
(ρ2 − x 2)1/2(ρ20 − x 2)1/2
⌠ tan-1 .
1
= − (1.1.8)
⌡ (ρ2 − x 2)1/2(ρ20 − x 2)1/2 R  xR 
All the results above are related to the distance between two points in the
plane z =0. We need to generalize them to represent
1 1
1+u = . (1.1.9)
R0 [ρ + ρ0 − 2ρρ0cos(φ−φ0) + z 2](1+u)/2
2 2
One can observe that representation (1.1.4) remains valid if we formally substitute
ρ and ρ0 by arbitrary quantities l 1 and l 2. We need to choose them so that
ρ2 + ρ20 − 2ρρ0cos(φ−φ0) + z 2 = l 21 + l 22 − 2 l 1 l 2cos(φ−φ0). (1.1.10)
This leads to two equations
l 21 + l 22 = ρ2 + ρ20 + z 2, l 1 l 2 = ρρ0. (1.1.11)
The solution will take the form
1
l 1(ρ0,ρ, z ) = {[(ρ + ρ0)2 + z 2]1/2 − [(ρ − ρ0)2 + z 2]1/2}, (1.1.12)
2
1
l 2(ρ0,ρ, z ) = {[(ρ + ρ0)2 + z 2]1/2 + [(ρ − ρ0)2 + z 2]1/2}. (1.1.13)
2
Hereafter the following abbreviations will be used:
l 1( x ) ≡ l 1( x ,ρ, z ), l 2( x ) ≡ l 2( x ,ρ, z ), (1.1.14)
l 1 ≡ l 1( a ,ρ, z ), l 2 ≡ l 2( a ,ρ, z ). (1.1.15)
Note the limiting properties
lim l 1( x ) = min( x ,ρ), lim l 2( x ) = max( x ,ρ). (1.1.16)

z→ 0 z→ 0
In view of the properties above, the representation (1.1.4) can be generalized
1 1
1+u =
R0 [ρ + ρ0 − 2ρρ0cos(φ−φ0) + z 2](1+u)/2
2 2
x2
λ( , φ−φ0) x ud x
l1(ρ 0) ρρ0
2 πu ⌠
= cos . (1.1.17)
π 2 ⌡ {[ l 21(ρ ) − x 2][ l 22(ρ ) − x 2]}(1+u)/2
0 0 0
Formula (1.1.17) simplifies when u =0
1 1
= 2
R0 [ρ + ρ0 − 2ρρ cos(φ−φ ) + z 2]1/2
2
0 0
x2
λ( , φ−φ ) d x
l1(ρ 0) ρρ 0
0
2 ⌠
= . (1.1.18)
π ⌡ {[ l 21(ρ ) − x 2][ l 22(ρ ) − x 2]}1/2
0 0 0
Again, one can notice that the integral in (1.1.18) may be evaluated as indefinite
x2
λ( , φ−φ ) d x
ρρ0 0
{[ l 21(ρ ) − x 2][ l 22(ρ ) − x 2]}1/2
⌠ = −
1
tan-1
0 0
.
⌡ {[ l 21(ρ0) − x 2][ l 22(ρ0) − x 2]}1/2 R0 xR 0
(1.1.19)
The last representation is very important and will be widely used throughout the
book.
Another series of useful formulae can be obtained from those above by a
simple change of variables, namely,
ρρ0
λ( , φ−φ0) d x
x2 {[ x 2 − l 21(ρ0)][ x 2 − l 22(ρ0)]}1/2
⌠ 1
tan-1
1/2 = R ,
⌡ {[ x − l 1(ρ0)][ x − l 2(ρ0)]}
2 2 2 2
0
xR 0
(1.1.20)
1 1
1+u =
R0 [ρ + ρ0 − 2ρρ0cos(φ−φ0) + z 2](1+u)/2
2 2
ρρ0
∞ λ( 2 , φ−φ0) x ud x
x
=
2
cos
πu ⌠ , (1.1.21)
π 2 ⌡ {[ x 2 − l 21(ρ )][ x 2 − l 22(ρ )]}(1+u)/2
l (ρ )
0
0 0
2
1 1
= 2
R0 [ρ + ρ0 − 2ρρ0cos(φ−φ0) + z 2]1/2
2
ρρ0
∞ λ( , φ−φ0) d x
x2
=
2 ⌠ , (1.1.22)
π ⌡ {[ x 2 − l 21(ρ )][ x 2 − l 22(ρ )]}1/2
l (ρ )
0
0 0
2
ρρ0
λ( , φ−φ0) d x
x2 ( x 2 − ρ2)1/2( x 2 − ρ20)1/2
⌠ tan-1 .
1
= (1.1.23)
⌡ ( x 2 − ρ2)1/2( x 2 − ρ20)1/2 R  xR 
The representations above are useful for solving external mixed boundary value
problems.
Several modifications of (1.1.19) are available. For example, we can write
x2
λ( , φ−φ0) d x
ρρ0
(ρ2 − x 2)1/2[ρ20 − g 2( x )]1/2
⌠ = −
1
tan-1 .
⌡ (ρ2 − x 2)1/2[ρ20 − g 2( x )]1/2 R0 xR 0
(1.1.24)
Here
g ( x ) = x [1 + z 2/(ρ2 − x 2)]1/2. (1.1.25)
It is important to notice that the function g ( x ) is inverse to l 1 for x 2<ρ2, and is

inverse to l 2 for x 2>ρ2+ z 2. Introduction of a new variable x = l 1( y ), y = g ( x )
transforms (1.1.24) into
[ l 22( y ) − y 2]1/2 l 2( y )
⌠ λ 1 , φ−φ  d y
⌡ (ρ20 − y 2)1/2[ l 22( y ) − l 21( y )]  ρρ0 0
(ρ20 − y 2)1/2[ l 22( y ) − y 2]1/2

1
= − tan-1 (1.1.26)
R0 yR 0
A particular case of (1.1.18), when z =0, reads
1 1
= 2
R [ρ + ρ0 − 2ρρ0cos(φ−φ0)]1/2
2
x2
λ( , φ−φ0) d x
min (ρ0,ρ) ρρ0
= ⌠
2
. (1.1.27)
π⌡ (ρ2 − x 2)1/2(ρ20 − x 2)1/2
0
The same result takes another form due to (1.1.22)
ρρ0
∞ λ( , φ−φ0) d x
x2
1
=
2 ⌠ . (1.1.28)
R π ⌡ ( x 2 − ρ2)1/2( x 2 − ρ20)1/2
max (ρ0,ρ)
The integral representations of this section can be generalized for a sphere (see
Fabrikant, 1987d), and yet another modification in toroidal coordinates is also
possible. The book volume limitation does not allow us to go into detail.
Exercise 1.1
1. Prove the identity (η is defined by 1.1.6)
ρ2ρ20 − x 4
dη
= − .
dx x 3η
2. Prove the identity ( R is defined by the first line of 1.1.27)
x2 xη dη
λ( , φ−φ ) = − 2 .
ρρ 0 R + η dx
2
0
Hint : use the identity: x 2 + ρ2ρ20/ x 2 − 2ρρ cos(φ−φ ) = R 2 + η2, and the result
0 0
above.
3. Prove the identity
ρρ
0 xη dη
λ( , φ−φ ) = 2 dx .
x 2 0 R + η
2
4. Prove the identities
( l 22 − ρ2)1/2( l 22 − a 2)1/2 = z l , ( a 2 − l 21)1/2(ρ2 − l 21)1/2 = z l ,

2 1
( a 2 − l 21)1/2( l 22 − a 2)1/2 = za , ( l 22 − ρ2)1/2(ρ2 − l 21)1/2 = z ρ.

Reminder : l and l are understood as l ( a ,ρ, z ) and l ( a ,ρ, z ) respectively. Hint :
1 2 1 2
use (1.1.11)
5. Prove that g ( x ) is inverse to both l and l , namely, prove that g ( l )= a , and

1 2 1
g ( l )= a .
2
6. Prove the identities

∂l zl ∂l zl
1 1 2 2
= − 2 , = ,
∂z l 2 − l 21 ∂z l 22 − l 21
∂l1 a l 2 − ρl 1 ρ( a 2 − l 21) ∂l2 ρl 2 − a l 1 ρ( l 22 − a 2)

= = , = = .
∂ρ l 22 − l 21 l 1( l 22 − l 21) ∂ρ l 22 − l 21 l 2( l 22 − l 21)
Hint : use the properties above.
7. Evaluate the integral
[ l 22( x ) − x 2]1/2 l ( x) x
⌠ dx
λ 1 , φ − ψ .
⌡ (ρ20 − x 2)1/2 l 22( x ) − l 21( x )  l 2( x )ρ0 
(ρ20 − x 2)1/2[ l 22( x ) − x 2]1/2

1
Answer: − tan-1 .
R0 xR 0
Hint : use (1.1.26)
( x 2 − l 21( x ))1/2 ρρ
⌠ dx 
λ 2
0
, φ−φ .
⌡ ( x − ρ0) [ l 2( x ) − l 1( x )] l 2( x )
2 2 1/2 2 2 0
( x 2 − l 21( x ))1/2( x 2 − ρ20)1/2

1
Answer: tan-1 .
R xR
0 0
Hint : use (1.1.20)
9. Establish the integral representation
min (ρ0,ρ)
cos[κ(ρ2 − x 2)1/2(ρ20 − x 2)1/2/ x + (πν/2)]
λ , φ−φ0 x νd x = 1+ν.
2⌠ x2 e-κR
π⌡ [(ρ2 − x 2)(ρ20 − x 2)](1+ν)/2 ρρ0  R
0
Note : try to use this integral representation for solving the Klein-Gordon equation.
1.2 Properties of the L-operators
Introduce the integral L-operator as follows:

1.2 Properties of the L-operators 23
∞ 2π ∞
-in φ
L( k ) f (φ) =
1
2π Σ n=-∞
k |n|e ⌠ e
in φ
⌡
0
f (φ0)dφ0 = Σk
n=-∞
|n| in φ
f ne . (1.2.1)
0
Here f n is the n -th Fourier coefficient of the function f . In the case when k <1,
formula (1.2.1) can be rewritten as
2π
1⌠
L( k ) f (φ) = λ( k , φ−φ0) f (φ0) dφ0, (1.2.2)
2π⌡
0
where λ(⋅,⋅) is defined by (1.1.5). The L-operator may also be called Poisson
operator, since it was introduced by Poisson for solving the two-dimensional
Dirichlet problem for a circle.
The following properties of the L-operators are valid
L( k 1)L( k 2) = L( k1k2 ), limL( k ) f = f . (1.2.3)

k→1
The proof is elementary, and left to the reader. These properties are widely
used in various transformations throughout this book, and are essential to the new
method.
Exercise 1.2
1. Prove the identity L( k )λ( m ,φ)=λ( km ,φ), for k <1 and m <1.
Hint : use (1.2.3)

2π
⌠ dφ
, for ρ>ρ0 and r > r 0.
⌡ [ρ + ρ0 − 2ρρ0cos(φ−φ0)][ r2 + r20 − 2 rr0cos(φ−ψ)]
2 2
0
2π(ρ2 r 2 − ρ20 r 20)

Answer: .
(ρ2 − ρ20)( r 2 − r 20)[ρ2 r 2 + ρ20 r 20 − 2ρρ0 rr0cos(φ0−ψ)]

2π
⌠ dφ
, for k <1 and k 1<1.
⌡ [1 + k − 2 k cos(φ−φ0)]2[1 + k 21 − 2 k 1cos(φ−ψ)]
2
0
 2k2
2π
Answer: 
1 + k 2 k 21 − 2 kk 1cos(φ0−ψ) (1 − k 2)3

k 21 1 − k 4 k 21 
+
1  + .
1 + k 2 k 21 − 2 kk 1cos(φ0−ψ) 1 − k 21 (1 − k 2)2

2π
iφ
⌠ e dφ
, for k <1 and k 1<1.
⌡ [1 + k − 2 k cos(φ−φ0)]2[1 + k 21 − 2 k 1cos(φ−ψ)]
2
0
iψ
 k 31e
2π
Answer: 
[1 + k 2 k 21 − 2 kk 1cos(φ0−ψ)]2 1 − k 21

iφ
] + k ei ψ(1 − 3 k 2)
- i (ψ-φ0)
ke 0
[2(1 + k 4 k 21) − kk 1(1 + k 2)e
1
+ .
(1 − k 2)3

2π
2 iφ
⌠ e dφ
, for k <1 and k <1.
⌡ [1 + k − 2 k cos(φ−φ0)] [1 + k 21 − 2 k 1cos(φ−ψ)]
2 2 1
0
2 iψ
 k 21e
2π
Answer: 
[1 + k 2 k 21 − 2 kk cos(φ −ψ)]2 1 − k 21
1 0 
2 iφ 2 iφ
[e2 i ψ( k 4−3 k 2+1) − k 2e
- i (ψ-φ0)
2 kk 1e 0
] − k 2e 0
( k 2 + k 21−3 − 3 k 2 k 21)
+ .
(1 − k 2)3

1.3 Some further integral representations
An integral representation for z / R 30 is presented here, for R 0=[ρ2 + ρ20 −

2ρρ0cos(φ−φ0) + z 2]1/2. Consider the integral:
ρ
0 [ x 2 − l 21( x )]1/2 l ( x) x
λ , φ−φ .
1 1 d xdx 1
I = L( ) ⌠ (1.3.1)
1 ρ
0
ρ dρ ⌡
0 0
(ρ20 − x 2)1/2 l 2( x ) − l 1( x )  2
2 2 l ( x ) 0
0
The integral (1.3.1) looks cumbersome, and the first impression is that one should
be lucky just to express it in elliptic integrals. It will be shown below that the
integral is quasi-elliptic, and is computable in elementary functions. By using the
rule of differentiation
ρ ρ
0 0
d
⌠ f( x) dx  f( x)  + ρ ⌠ dx d f ( x )
2 1/2 = lim 2 1/2 d x x ,
dρ (ρ0 − x ) x→0  x  (ρ20 − x)  
2
⌡ ⌡
0
0
0 0
(1.3.2)
the integral (1.3.1) can be rewritten as
ρ
d  [ x − l 1( x )]
0 2 2 1/2 l ( x) x
I = ⌠
dx
λ 1
, φ−φ0. (1.3.3)
(ρ20 − x 2)1/2 d x  l 22( x ) − l 21( x )  l ( x )ρ 
⌡
1
2 0
0
Introduce a new variable:
(ρ20 − x 2)1/2 [ l 22( x ) − x 2]1/2

j = ,
x
dj [ l 22( x ) − x 2]1/2 [ρ20 l 22( x ) − x 2 l 21( x )]

j´ = = − . (1.3.4)
dx x 2[ l 22( x ) − l 21( x )] (ρ20 − x 2)1/2
Let us transform the expression for λ:
l ( x) x l 2( x ) l 2( x ) − ρ2ρ2/ l 2( x )
λ , φ−φ  = λ , φ−φ  = − 2
1 1 1 0 1
l 2( x )ρ0 0  ρρ0 0 l 1( x ) + ρ ρ0/ l 1( x ) − 2ρρ0cos(φ−φ0)

2 2 2
− l 21( x ) + ρ2ρ20/ l 21( x )

=
l 21( x ) + ρ2ρ20/ l 21( x ) − 2ρρ cos(φ−φ ) + ρ2+ρ20+ z 2 − l 21( x )− l 22( x ) + x 2−ρ20
0 0
x 2 l 21( x ) − ρ20 l 22( x ) [ l 22( x ) − l 21( x )] (ρ20 − x 2)1/2 j´

= − = − (1.3.5)
x 2
( R 20 + j) 2
[ l 22( x ) − x]
2 1/2
R 20 + j2
The substitution of (1.3.4) and (1.3.5) in (1.3.3) allows us to continue the

transformations:
ρ
d  [ x − l 1( x )] (ρ0 − x )
0 2 2 1/2 2 2 1/2
I = − ⌠
dx j´ 
(ρ20 − x 2)1/2 dx  [ l 22( x ) − x 2]1/2 R 20 + j 2 
⌡
1
0
ρ
d  xz(ρ0 − x )
0 2 2 1/2
= − ⌠
dx j´ 
(ρ20 − x 2)1/2 d x  l 22( x ) − x 2 R 20 + j 2 
⌡
0
ρ
d  (ρ0 − x ) j ´ 
0 2 2 3/2
dx
= −z ⌠
⌡ (ρ20 − x 2)1/2 d x  xj 2( R 20 + j 2) 
0
ρ
 0 (ρ20 − x 2) j ´ r
dj( x) 
= −z   2 2 − ⌠ 
  xj ( R 0 + j 2) ⌡ j 2( R 20 + j 2)

0 0
ρ
 (ρ20 − x 2) j ´
0
1 1 j  π z
= −z   2 2 2 + 2 + 3 tan-1
 = (1.3.6)
  xj ( R 0 + j ) R0j R0 R0 2 R 30

0
Finally, (1.3.6) allows us to write the required representation:
ρ
0 [ x 2 − l 21( x )]1/2 l ( x) x
λ , φ−φ .
z 2 1 d xdx 1
3 = πρ L(ρ ) dρ ⌠ (1.3.7)
R0 0 0 0 ⌡
(ρ20 − x 2)1/2 l 2( x ) − l 1( x )  l 2( x )
2 2 0
0
The second integral to consider is
a
[ l 22( x ) − x 2]1/2 ρ
d ⌠
λ 2 , φ−φ0.
1 xdx
I2 = − L(ρ0) (1.3.8)
ρ0 dρ0 ⌡ ( x − ρ0)
2 2 1/2
l 2( x ) − l 1( x ) l 2( x )
2 2

ρ
0
Make use of the rule of differentiation
a a
d ⌠ F (ρ) dρ dρ d F (ρ)
+ x⌠ 2
F(a)x
= −
d x ⌡ (ρ − x ) a(a − x ) ⌡ (ρ − x ) dρ ρ 
2 2 1/2 2 2 1/2 2 1/2
x x
a
F(a)a 1⌠ ρdρ d
= − 2 1/2 + x 2 1/2 dρ F (ρ). (1.3.9)
x( a − x ) ⌡ (ρ − x )
2 2
x
Expression (1.3.8) will take the form
( l 22 − a 2)1/2 ρρ
 0 , φ−φ 
I2 = 2 λ
( a − ρ0) ( l 2 − l 1)  l 22
2 2 1/2 2 0
a
[ l 2( x ) − x 2]1/2 ρρ
−
⌠ dx d 2  0

⌡ ( x − ρ0) d x  l 2( x ) − l 1( x ) l 2( x ) , φ−φ0.
2 2 1/2 2 2 λ 2 (1.3.10)
ρ
0
Introduce a new variable
( x 2 − ρ20)1/2[ x 2 − l 21( x )]1/2

h = ,
x
[ x 2 − l 21( x )]1/2[ x 2 l 22( x ) − ρ20 l 21( x )]

dh
h´ = = . (1.3.11)
dx x 2( x 2 − ρ20)1/2[ l 22( x ) − l 21( x )]
The expression for λ can be presented in the manner, similar to (1.3.5), namely,
ρρ0 [ l 22( x ) − l 21( x )]( x 2 − ρ20)1/2

λ 2 , φ−φ0 =
h´
. (1.3.12)
l 2( x )  [ x 2 − l 21( x )]1/2 R 20 + h 2
Substitution of (1.3.11) and (1.3.12) in (1.3.10) yields
( l 22 − a 2)1/2 ρρ
I2 = λ 0 , φ−φ 
( a 2 − ρ20)1/2( l 22 − l 21)  l 22 0
a
xz( x 2 − ρ20)1/2 h ´
−
⌠ dx d 
⌡ ( x 2 − ρ20)1/2 d x [ x 2 − l 21( x )]( R 20 + h 2)
ρ
0
( l 22 − a 2)1/2 ρρ
 0 , φ−φ 
= 2 λ
( a − ρ0) ( l 2 − l 1)  l 22
2 2 1/2 2 0
a
( x 2 − ρ20)3/2 h ´
− z
⌠ dx d . (1.3.13)
⌡ ( x − ρ0) d x  xh ( R 0 + h ) 
2 2 1/2 2 2 2
ρ
0
Integration by parts in (1.3.13) yields
a
[ l 22( x ) − x 2]1/2
−
1
L(ρ0)
d ⌠ xdx
λ ρ , φ−φ 
ρ0 dρ0 ⌡ ( x 2 − ρ20)1/2 l 22( x ) − l 21( x ) l 22( x ) 0
ρ
0
z R 0
+ tan-1 .
h
= 3 (1.3.14)
R0  h R 0
Here h stands for h ( a ), as defined by the first expression of (1.3.11). In the

limiting case, when a →∞, expression (1.3.14) gives yet another representation for
z / R 30, namely,
∞
[ l 22( x ) − x 2]1/2
z
= −
2
L(ρ0)
d ⌠ xdx
λ ρ , φ−φ .
R 30 πρ0 dρ0 ⌡ ( x 2 − ρ20)1/2 l 22( x ) − l 21( x ) l 22( x ) 0
ρ
0
(1.3.15)
The last expression is useful in external problems, while its equivalent (1.3.6) is
needed in solving internal ones. Since the integral in (1.3.1) was evaluated
earlier as indefinite, we may consider its generalization
ρ
0 [ x 2 − l 21( x )]1/2 l ( x) x
I3 =
1 1
L( )
d
⌠
xdx
λ1 , φ−φ0. (1.3.16)
ρ0 ρ0 dρ0
⌡ (ρ20 − x 2)1/2 l 2( x ) − l 1( x )  2
2 2 l ( x ) 
a
Make use of the rule of differentiation
x x
d ⌠ F (ρ) dρ
=
F(a)x
2 1/2 + x
⌠ 2 dρ 2 1/2 d F (ρ)
d x ⌡ ( x 2 − ρ2)1/2 a(x − a ) ⌡ ( x − ρ ) dρ ρ 
2
a a
x
F(a)a 1⌠ ρdρ d
= 2 1/2 + x 2 1/2 dρ F (ρ). (1.3.17)
x( x − a )
2
⌡ (x − ρ )
2
a
Expression (1.3.16) will take the form
( a 2 − l 21)1/2 l 21
I3 = λ , φ−φ0
(ρ20 − a 2)1/2[ l 22 − l 21] ρρ0 
ρ
d  [ x − l 1( x )]
0 2 2 1/2 l ( x) x
+ ⌠
dx
λ 1 , φ−φ0. (1.3.18)
(ρ20 − x 2)1/2 d x  l 22( x ) − l 21( x )  l 2( x )ρ0 
⌡
a
By introducing the notation
z  R0
+ tan-1 ,
j(y)
F ( y) = 3 (1.3.19)
R 0 j(y)  R 0 
where j ( y ) is defined according to (1.3.4), expression (1.3.18) can be rewritten as
ρ
ρ20 − a 2 0 (ρ20 − y 2)3/2
dF(a) dy d d F ( y )
I = + ⌠ . (1.3.20)
3 a da
⌡ (ρ0 − y )
2 2 1/2 d y
 y dy 
a
Integration in (1.3.20) can be performed by parts, with a simple result F ( a ),

which means establishment of another integral representation
z R 0 -1 
j
+ tan
R 30  j R 0
ρ
0 [ x 2 − l 21( x )]1/2 l ( x) x
λ , φ−φ0.
1 1 d xdx 1
= L( ) ⌠ (1.3.21)
ρ0 ρ0 dρ0
⌡ (ρ20 − x 2)1/2 l 2( x ) − l 1( x )  2
2 2 l ( x ) 
a
Exercise 1.3
1. Prove that h in (1.3.14) can be defined by any of the expressions
( a 2 − ρ20)1/2( a 2 − l 21)1/2 z ( a 2 − ρ20)1/2
h ≡ h(a) = =
a ( l 22 − a 2)1/2
[ l 22 − l 21(ρ0)]1/2[ l 22 − l 22(ρ0)]1/2 ( a 2 − ρ20)1/2( l 22 − ρ2)1/2

= = .
l2 l2
2. Prove that j in (1.3.4) can be defined in several equivalent ways:

(ρ20 − a 2)1/2( l 22 − a 2)1/2 (ρ20 − a 2)1/2(ρ2 − l 21)1/2
j ≡ j(a) = =
a l1
z (ρ20 − a 2)1/2 [ l 21(ρ0) − l 21]1/2[ l 22(ρ0) − l 21]1/2

= = .
( a 2 − l 21)1/2 l1
3. Establish (1.3.15)
1.4 Internal mixed boundary value problem for a half-space
The material in this section follows the paper (Fabrikant, 1986h). Introduce
a set of cylindrical coordinates (ρ,φ, z ). Consider the problem of finding a
potential function V , harmonic in the half-space z ≥0, vanishing at infinity, and
subject to the boundary conditions on the plane z =0
V = v (ρ,φ), for ρ≤ a , 0≤φ<2π;
∂V
= 0, for ρ> a , 0≤φ<2π. (1.4.1)
∂z
The problem (1.4.1) can be interpreted as an electrostatic one of a charged disc,

with a certain potential prescribed on its surface, or it can be interpreted as an
elastic contact problem of a circular punch pressed against an elastic half-space;
other interpretations are also possible. We call the problem internal because the
non-zero conditions are prescribed inside the disc. The potential function V can
be represented through the simple layer as follows:
2π a
σ(ρ ,φ ) 2π ∞
σ(ρ ,φ )
V (ρ,φ, z ) = ⌠ ⌠ ρ dρ dφ + ⌠ ⌠
0 0 0 0
ρ dρ dφ . (1.4.2)
⌡⌡ R
0
0 0 0 ⌡⌡ R
0
0 0 0
0 0 0 a
Here
1 ∂V
R = [ρ2 + ρ20 − 2ρρ cos(φ−φ ) + z 2]1/2, and σ = − .
0 0 0 2π ∂ z z=0
Substitution of (1.1.18) and (1.1.22) in (1.4.2) yields, after changing the order of
integration
l1 a ρ dρ
V (ρ,φ, z ) = 4⌠ 2
dx
⌠ 2
0 0
 x  σ(ρ ,φ) 2
1/2 L ρρ
⌡ (ρ − x )
2 1/2
⌡ [ρ0 − g ( x )]
2
 0 0
0 g(x)
∞ g(x)
ρ dρ ρρ
+ 4⌠ 2  0 σ(ρ ,φ).
⌠ dx 0 0
2 1/2 L (1.4.3)
⌡ ( x − ρ ) ⌡ [ g ( x ) − ρ0]
2 1/2 2
 x2  0
l a
2
Here the L-operator is defined by (1.2.1), g is given by (1.1.25), the

abbreviations l and l are understood as l ( a ,ρ, z ) and l ( a ,ρ, z ) respectively; and
1 2 1 2
the following rule is used for changing the order of integration:
a l1(ρ 0) l1 a ∞ ∞ ∞ g(x)
⌠ dρ ⌠ d x = ⌠ d x ⌠ dρ , ⌠ dρ ⌠ d x = ⌠ d x ⌠ dρ .
⌡ 0 ⌡ ⌡ ⌡ 0 ⌡ 0 ⌡ ⌡ ⌡ 0
0 0 g(x) a l (ρ0) l a
0 2 2
(1.4.4)
The rule is illustrated by Fig. 1.4.1, where the domains of integration are shaded.
Fig. 1.4.1 Domains of integration.
Substitution of the boundary condition (1.4.1) in (1.4.3) leads to the

governing integral equation
ρ a ρ dρ
4⌠ 2 ⌠ 2
dx
0 0
 x  σ(ρ ,φ) = v (ρ,φ). 2
2 1/2 L ρρ (1.4.5)
⌡ (ρ − x ) ⌡ (ρ0 − x )
2 1/2
 0 0
0 x
Expression (1.4.5) is now presented as a sequence of two Abel-type operators and

one L-operator. We recall that the general Abel integral equation
⌠ 2 F ( y ) 2d(1+u)/2
y
= f( x) (1.4.6)
⌡ (y − x )
x
has the solution

a
2cos(π u /2) d ⌠ f( x) xdx
F ( r) = − . (1.4.7)
π d r ⌡ ( x 2 − r 2)(1-u)/2
r
Since the variables in the Abel operators of (1.4.5) are interwoven with those of
the L-operator, we need to apply their combination, in order to invert (1.4.5).
The first operator to be applied to both sides of (1.4.5) is
L  L .
ζ d ⌠ ρdρ ρ
(1.4.8)
 t  dt ⌡ (t − ρ )
2 2 1/2
ζ 
0
Here we introduced a dummy parameter ζ in order to make the parameter of the

L-operator dimensionless, and also in order to claim it being less that unity
almost everywhere in the interval of integration which is a precondition for usage
of the properties (1.2.3). We call the parameter ζ ’dummy’ because it was
introduced for formal reasons only; it will disappear in the final result, and has
no real bearing on the transformations to follow. Of course, the introduction of
dummy parameter does not rigorously validate our use of the properties (1.2.3).
Such a validation is beyond scope of this book: the author is satisfied by the
fact that the final result everywhere is proven to be correct. In order to make
the intermediate transformations rigorous, one has to prove the theorem stating
that one can use the properties (1.2.3) in the mathematical manipulations with an
expression of the type L( k )M L( k )M L( k )..., where M are certain linear
1 1 2 2 3
operators, if the product k1k2k3 ... is less than unity, thus allowing for any
particular k to be greater than unity. We hope that some readers might be
willing and able to prove the theorem.
The result of application of (1.4.8) to both sides of (1.4.5) is
a ρ0dρ0 t
L  σ(ρ0,φ) = L  ρ v (ρ,φ).

ζ d ⌠ ρdρ
2π⌠
t
2 1/2 L ζ (1.4.9)
⌡ (ρ0 − t )
2 2 1/2
 0
ρ  
t d t ⌡ (t − ρ )
2
 
t 0
The second operator to be applied to both sides of (1.4.9) is
L  L ,
y d⌠ tdt ζ
ζ d y⌡ ( t − y )
2 2 1/2
t 
y
with the result
a t
σ( y ,φ) = − 2 L  L 2  ρ
1 y d⌠ tdt ζ2 d ⌠ ρdρ
2 1/2 L ζ v (ρ,φ).
πy  
ζ d y ⌡ (t − y )
2 2 1/2
t  d t ⌡ (t − ρ )
2
 
y 0
(1.4.10)
The rules of differentiation of integrands and the properties of the L-operators
allow us to rewrite (1.4.10)
σ( y ,φ) = 2  2 ⌠ 2 d t 2 1/2 d Φ( t , y ,φ).

1 Φ( a , y ,φ)
2 1/2 − (1.4.11)

π (a − y ) ⌡ (t − y ) dt 
y
Here
t
d  ρ y 
v (ρ,φ).
1 ⌠ ρdρ
Φ( t , y ,φ) = ρL (1.4.12)
t ⌡ ( t − ρ ) dρ  t 2 
2 2 1/2

0
Note that the dummy parameter ζ has disappeared from the final solution, and
the combined parameter of the L-operator is less than unity, as it should be. In
the future we shall no longer use the dummy parameter explicitly, assuming that
the use of the properties (1.2.3) is justified. Using integration by parts and the
fact that λ( k ,ψ) satisfies the two-dimensional Laplace equation in polar
coordinates, the following identity can be established
t
d
Φ( t , y ,φ) = ⌠ 2
ρdρ ρy  ∆ v (ρ,φ),
2 1/2 L 2 (1.4.13)
dt ⌡ (t − ρ ) t 
0
where ∆ is the two-dimensional Laplace operator in polar coordinates.

Substitution of (1.4.13) in (1.4.12) leads to another form of solution, namely,
a t
σ( y ,φ) = 2  2 ⌠ 2 d t 2 1/2 ⌠ 2 ρdρ 2 1/2 Lρ2y  ∆ v (ρ,φ),

1 Φ( a , y ,φ)
2 1/2 −

π (a − y ) ⌡ (t − y ) ⌡ (t − ρ ) t  
y 0
(1.4.14)
Interchange of the order of integration in (1.4.14) and integration with respect to
t (see (1.1.23)) yields
1  Φ( a , y ,φ)
σ( y ,φ) = 
π2 ( a 2 − y 2)1/2

2π a
1⌠ ⌠ -1
( a 2 − ρ2)1/2( a 2 − y 2)1/2  ∆ v (ρ,ψ) ρdρdψ 
a [ρ2 + y 2 − 2ρy cos(φ−ψ)]1/2 [ρ2 + y 2 − 2ρy cos(φ−ψ)]1/2
− tan .
2π⌡ ⌡
0 0 
(1.4.15)
The solution obtained here consists of two parts: the first part is singular at the
boundary while the second one vanishes at the boundary. In various applications
it is required that the solution be nonsingular at the boundary. The necessary
and sufficient condition then is Φ( a , a ,φ)=0. In elastic contact problems this
condition defines the radius of the contact domain. Notice also that in the case
when v is a two-dimensional harmonic function, the non-trivial solution is
singular.
Now it is of interest to express the potential V in the half-space directly
through its value v prescribed inside the disc ρ= a . Substitution of (1.4.10) in
(1.4.3) yields, after subsequent integration
l1 g(x)
V (ρ,φ, z ) = ⌠ 2
2 dx
L  x2  d ⌠ rdr
L( r ) v ( r ,φ).
π ⌡ (ρ − x )
2 1/2
ρg ( x ) d g ( x )⌡ [ g ( x ) − r2]1/2
2 2
0 0
(1.4.16)
Here the following property of the Abel operators was used
a a
⌠ 2 d r 2 1/2 d ⌠ 2tf ( t ) d2 t1/2 = − π f ( y ). (1.4.17)

⌡ (r − y ) dr⌡ (t − r ) 2
y r
Introduction of a new variable t = g ( x ), x = l 1( t ), transforms (1.4.16) into
a d l 1( t ) l 21( t ) t ρ0dρ0
V (ρ,φ, z ) =
2⌠
L   d ⌠ L(ρ0) v (ρ0,φ). (1.4.18)
π⌡ [ρ2 − l 21( t )]1/2  ρ t 2  d t ⌡ ( t 2 − ρ20)1/2
0 0
By changing the order of integration in (1.4.18), according to the rule

a r a a
⌠ F ( r)d r d ⌠ ρ2 f (ρ) d2 ρ1/2 = −⌠ f (ρ)dρ d ⌠ 2F ( r) rd2r 1/2, (1.4.19)

⌡ dr ⌡ (r − ρ ) ⌡ dρ ⌡ ( r − ρ )
0 0 0 ρ
the following expression can be obtained
2⌠  ρ 
a t d l 1( t )
d ⌠ 
π⌡  0 dρ ⌡ ( t 2 − ρ20)1/2[ρ2 − l 21( t )]1/2 l 22( t ) 0
V (ρ,φ, z ) = − L ( ρ ) L v (ρ ,φ)dρ0.
0  
0ρ
0
(1.4.20)
The integral in curly brackets can be evaluated using (1.3.14), with the result
2π a
R0
1
V (ρ,φ, z ) = 2 ⌠ ⌠  + tan-1  3 v (ρ0,φ0) ρ0dρ0dφ0.
h z
(1.4.21)
π ⌡ ⌡ h R 0 R 0
0 0
Here
R = [ρ2 + ρ20 − 2ρρ cos(φ−φ ) + z 2]1/2, h = ( a 2 − l 21)1/2( a 2 − ρ20)1/2/ a .
0 0 0
(1.4.22)
Formulae (1.4.18) and (1.4.21) define the potential function V in the half-space
z ≥0, expressed directly through its value v prescribed inside the disc ρ= a , z =0.
Expression (1.4.18) is useful when an explicit evaluation of the integrals is
possible, while expression (1.4.21) is more convenient for numerical integration.
Note that in the limiting case, when z =0, equation (1.4.21) transforms into
a known result, namely,
V (ρ,φ,0) = v (ρ,φ), for ρ≤ a ; and
(ρ2 − a 2)1/2
2π a
v (ρ ,φ )ρ dρ dφ
V (ρ,φ,0) = ⌠⌠ 0 0 0 0 0
, for ρ> a .
π2 ⌡ ⌡ ( a −ρ0) [ρ + ρ0 − 2ρρ0cos(φ−φ0)]
2 2 1/2 2 2
0 0
The solution of the first mixed boundary value problem is completed. The main
results are given by formulae (1.4.10) and (1.4.18).
Consider now another internal problem, characterized by the following mixed
conditions on the boundary z =0:
∂V
= − 2πσ(ρ,φ), for ρ≤ a , and 0≤φ<2π;
∂z
V = 0, for ρ> a , and 0≤φ<2π. (1.4.23)
The problem (1.4.23) can be interpreted as an electrostatic one of a charged disc

ρ≤ a inside an infinite grounded diaphragm ρ> a . Mathematically similar problem
arises in the consideration of a penny-shaped crack subjected to an arbitrary
pressure σ.
Substitution of (1.4.23) in (1.4.3) leads to the integral equation, for ρ> a ,
a a ρ dρ
⌠ 2 d x ⌠
0 0
L  x 2  σ(ρ ,φ)
⌡ (ρ − x ) ⌡ (ρ0 − x )
2 1/2 2 2 1/2
ρρ0 0
0 x
∞ x ρ dρ ρρ
+ ⌠ 2
dx
⌠ 2
0 0
 0 σ(ρ ,φ) = 0.
2 1/2 L (1.4.24)
⌡ ( x − ρ ) ⌡ ( x − ρ0)
2 1/2
 x2  0
ρ a
Notice that σ in the first term of (1.4.24) is known from (1.4.23), while σ in
the second term is yet to be determined. By using the integral representations
(1.1.27) and (1.1.28), equation (1.4.24) can be rewritten as
∞ x ρ dρ ρρ
⌠ ⌠ 2
0dx
0
 0 σ(ρ ,φ)
2 1/2 L
⌡ ( x − ρ ) ⌡ ( x − ρ0)
2 2 1/2
 x2  0
ρ a
∞ a ρ dρ ρρ
= −⌠ 2 ⌠
0 dx
0
L  0 σ(ρ ,φ). (1.4.25)
⌡ ( x − ρ2)1/2 ⌡ ( x 2 − ρ20)1/2  x 2  0
ρ 0
Operation on both sides of (1.4.25) by
L 
d ⌠ ρdρ 1
L( t )
d t ⌡ (ρ2 − t 2)1/2 ρ
t
leads to
t ρ0dρ0 ρ a ρ dρ ρ
⌠ 2 L  σ(ρ0,φ) = −⌠ 2  0 σ(ρ ,φ).
0 0 0
L (1.4.26)
⌡ ( t − ρ0)
2 1/2
t ⌡ ( t − ρ0)
2 1/2
t 0
a 0
The next operator to apply is
ρ
L 
1 d ⌠ tdt
L( t ),
ρ dρ ⌡ (ρ − t 2)1/2
2
a
and the final result takes the form
a(a2 − ρ20)1/2ρ0dρ0 ρ
⌠ L  σ(ρ ,φ)
2 0
σ(ρ,φ) = −
π(ρ − a 2)1/2 ⌡
2
ρ2 − ρ20 ρ 0
0
1
2π a 2
(a − ρ20)1/2σ(ρ0,φ0) ρ0dρ0dφ0
= − ⌠⌠ . (1.4.27)
π (ρ − a )
2 2 2 1/2
⌡ ⌡ ρ2 + ρ20 − 2ρρ0cos(φ−φ0)
0 0
Formula (1.4.27) defines the value of σ outside the circle ρ= a directly through
its value inside. Now σ is known all over the plane z =0, and substitution of
(1.4.27) in the second term of (1.4.3) allows us to express the potential function
V directly through the prescribed value of σ. The first integration yields
l1 a ρ0dρ0
V (ρ,φ, z ) = 4⌠
dx
⌠ 2  x  σ(ρ ,φ)
2
1/2 L ρρ
⌡ (ρ − x )
2 2 1/2
⌡ [ρ0 − g ( x )]
2
 0 0
0 g(x)
∞
ρ0dρ0 ρρ a
− 4⌠ 2
dx
⌠ L  0 σ(ρ ,φ). (1.4.28)
⌡ ( x − ρ2)1/2 ⌡ [ g 2( x ) − ρ20]1/2  x 2  0
l 0
2
Here the following integral was employed
ρ
⌠ 2 ydy π
2 = , for r < a .
⌡ (ρ − y ) ( y − a ) ( y − r ) 2(ρ − r ) ( a − r 2)1/2
2 1/2 2 2 1/2 2 2 2 1/2 2
a
(1.4.29)
The first term in (1.4.29) can be transformed by using (1.1.22), in the following
manner:
l1 a ρ dρ
⌠ 2
dx
⌠ 2
0 0
 x  σ(ρ ,φ) 2
1/2 L ρρ
⌡ (ρ − x )
2 1/2
⌡ [ρ0 − g ( x )]
2
 0 0
0 g(x)
∞
a ρρ
= ⌠ ρ0dρ0 ⌠ dx  0 σ(ρ ,φ)
2 1/2 L
⌡ ⌡ ( x − ρ ) [ g ( x ) − ρ0]
2 2 1/2 2
 x2  0
0 l (ρ )0
2
l2 g(x)
ρ dρ ρρ
⌠ dx
⌠ 0 0
 0 σ(ρ φ)
= 2 1/2 L
⌡ ( x − ρ ) ⌡ [ g ( x ) − ρ0]
2 2 1/2 2
 x2  0
l (0) 0
2
∞ a ρ dρ ρρ
+ ⌠ 2  0 σ(ρ φ).
dx
⌠ 2
0 0
2 1/2 L (1.4.30)
⌡ ( x − ρ ) ⌡ [ g ( x ) − ρ0]
2 1/2
 x2  0
l 0
2
Substitution of (1.4.30) in (1.4.29) yields
l2 g(x)
ρ dρ ρρ
V (ρ,φ, z ) = 4 ⌠ 2 ⌠ dx
0 0
 0 σ(ρ φ).
2 1/2 L (1.4.31)
⌡ ( x − ρ ) ⌡ [ g ( x ) − ρ0]
2 1/2 2
 x2  0
l (0) 0
2
Introduction of a new variable t = g ( x ), x = l 2( t ), transforms (1.4.31) into
a d l 2( t ) t ρ dρ ρρ
V (ρ,φ, z ) = 4⌠ 2 ⌠ 2
0 0
 0 σ(ρ ,φ).
2 1/2L 2 (1.4.32)
⌡ [ l 2( t ) − ρ ] ⌡ ( t − ρ0) l 2( t ) 0
2 1/2
0 0
An interchange of the order of integration in (1.4.32), and integration with

respect to t (see 1.1.23), yields
2π a
tan-1  σ(ρ0,φ0) ρ0dρ0dφ0.

2⌠ ⌠ 1 h
V (ρ,φ, z ) = (1.4.33)
π⌡ ⌡ R 0  0
R
0 0
Formulae (1.4.31−33) give three equivalent representations of the potential function

V , the first two being more convenient for explicit evaluation of the integrals
involved, while the third one has some advantages for numerical integration.
Two examples are considered below.
Example 1. Let the potential prescribed inside the disc be v (ρ,φ)= v nρncosn φ,
v n=const. The solution due to (1.4.16) is
2vn l
Γ( n + 1) x 2nd x
1
V (ρ,φ, z ) = cosn φ ⌠ 2
√π ρn Γ( n + 1) ⌡ (ρ − x )
2 1/2
2 0
2Γ( n + 1) ( l 2 − a ) l2 − a2
2 2 1/2

= v nρ cosn φ 1 −
n 1 1 3 2
F −n, ; ; . (1.4.34)
 √πΓ( n + )
1 l 2
2 2 2 l 2 
2
The hypergeometric function in (1.4.34) can be expressed in elementary functions

(Bateman and Erdelyi, 1955)
(1 − ζ)n+1/2 dn  ζn-1/2
sin-1√ζ.
1 1 3
F( − n , ; ; ζ) = (1.4.35)
2 2 2 Γ( n + 1) dζn √1 − ζ 
Example 2. Let the charge distribution be prescribed in the form
σ(ρ,φ)=σnρncosn φ, σn=const. The solution is given by (1.4.32)
b
Γ( n + 1) x 2n+2d x
V (ρ,φ, z ) = 2√π σnρncosn φ ⌠ 2
⌡ (x + z )
3 2 n+1
Γ( n + )
2 0
Γ( n + 1) b 2n+3 3 5 b2
= √π σnρncosn φ 2n+2 F ( n + 1, n + ; n + ; − 2 ), (1.4.36)
5 z 2 2 z
Γ( n + )
2
where b =( a 2 − l 21)1/2, and the hypergeometric function can be expressed in

elementary (Bateman and Erdelyi, 1955)
3 5 2 n + 3 dn 1 1 1 + √ζ .
F ( n +1, n + ; n + ; ζ) =
2 2  ln
Γ( n + 1) dζn ζ 2√ζ 1 − √ζ
− 1

 
(1.4.37)
Bibliographical note. The degree of effectiveness of the new approach can
be established by comparison with the results reported in the literature.
Formulae, similar to (1.4.21), can be found in the works of Lord Kelvin.
Heine, Hobson and others. As was mentioned before, their practical use was
quite limited. The equivalent expressions, like (1.4.18), allow us to evaluate the
integrals involved in a straightforward and elementary manner. This was
demonstrated by two examples above. The general solution has now become a
working tool, available to anyone with even an undergraduate background in
mathematics.
Copson (1947) was very close to the discovery of the new method. He
established the identity
2π
⌠ cosn (φ − ψ) dφ
⌡ [ρ2 + ρ20 − 2ρρ cos(φ−φ )]1/2
0 0 0
4cosn (φ − ψ) min (ρ0,ρ)

0
⌠ x 2nd x
= . (1.4.38)
(ρρ0)n ⌡ (ρ2 − x 2)1/2(ρ20 − x 2)1/2
0
Copson has obtained a solution, similar to (1.4.10), in the form of a Fourier

expansion. It remains unclear why he did not perform the summation in (1.4.38)
leading to (1.1.27) which, in a way, is the foundation of the new method.
Galin (1953) obtained the nonsingular part (corresponding to the second term
in (1.4.15) of the general solution of equation (1.4.5). Leonov (1953) obtained
the closed form complete solution (0.3). Both solutions had the same limitation
as the previous classical solutions: difficulty in practical use.
Another type of solution of equation (1.4.5) can be expressed through the
z -derivative of (1.4.21) for z =0, due to the relationship
1 ∂V
σ = − . (1.4.39)
2π ∂ z z=0
The result is
2π a
1  ⌠⌠ 1 η
σ(ρ,φ) = 3 −∆ tan-1( ) ω(ρ0,φ0) ρ0dρ0dφ0
2π
 0⌡ ⌡
R R
0
2π a
1 − tt ω(ρ ,φ ) 
−
a ⌠⌠ 0 0
ρ d ρ d φ 
( a 2 − ρ2)3/2 ⌡ ⌡ (1 − t )(1 − t ) ( a 2 − ρ20)1/2 0 0 0
0 0 
ρ0dρ0 ρρ0
1  ⌠
a x
= 2 −∆
dx
⌠ 2   ω(ρ ,φ)
2 1/2 L
π
 ⌡ ( x − ρ ) ⌡ ( x − ρ0)
2 2 1/2
x 
2 0
ρ 0
ρ0dρ0 ρρ0
 ω(ρ ,φ ),
a
a
− ⌠ 2 L ( (1.4.40)
( a 2 − ρ2)3/2 ⌡ ( a − ρ0)
2 1/2
a2  0 0
0 
where
ρρ0
i(φ-φ0)
t = 2 e , (1.4.41)
a
and the overbar everywhere indicates the complex conjugate value. Formula
(1.4.40) corresponds to the solution by Mossakovskii et al. (1985). The
developed apparatus can be used for solving the problem of several charged
coaxial (Fabrikant, 1987e) and arbitrarily located (Fabrikant, 1988a) circular discs.
Exercise 1.4
1. A circular conducting disc is kept at the potential v 0. Find the potential
function V .
2 2 a
Answer: V (ρ, z ) = v sin-1( l 1/ρ) = v sin-1( ).
π 0 π 0 l2
Hint : use formula (1.4.18)
2. Subject to the conditions of the previous problem, find the charge distribution
σ by using formulae (1.4.10) and (1.4.39). Prove that in both cases the result
is the same.
v0
Answer: σ = 2 2 .
π ( a − ρ2)1/2
3. Solve problems 1 and 2 for v = v 1ρcosφ, v 1= const.
v 1ρcosφsin-1( ) − √1 − ( a / l 2)2,
2 a a
Answer: V (ρ,φ, z ) =
π  l2 l2 
2 v 1ρcosφ
σ (ρ,φ) = .
π2( a 2 − ρ2)1/2
4. A uniform charge density σ=σ0=const is prescribed over a circular disc of

radius a . Find the potential function.
Answer: V (ρ, z ) = 4σ0( a 2 − l 21)1/2 − z sin-1( ).
a
 l 
2
5. Solve the previous problem for the case where σ=σ1ρcosφ, σ1=const.
σ ρcosφ ( a 2 − l 21)1/2 − 2 − z sin-1( ).

8 3 a2 3 a
3 1  2 2l2  2 l
2

6. The potential function is given by the expression
V (ρ,φ, z ) =
8  2 1/23
σ ρcosφ ( a − l 1)
2 a2 
− 2 −
3
z sin-1( ).
a
3 1  2 2 l 2 2 l
2

Find the charge distribution on the plane z =0.

Answer: σ = σ ρcosφ, for ρ≤ a ; σ=0, for ρ> a .
1
Hint : use (1.4.39).
7. Solve the mixed boundary value problem of potential theory for a sphere.
Hint : see (Fabrikant, 1987i).
1.5 External mixed boundary value problem for a half-space
The material in this section follows the paper (Fabrikant, 1986f). The
problem is called external when non-zero boundary conditions are prescribed
outside the disc. As in the previous section, we consider two types of problem.
Problem 1. It is necessary to find a function, harmonic in the half-space
z ≥0, vanishing at infinity, and subject to the mixed boundary conditions on the
plane z =0, namely,
∂V
= 0, for ρ< a , 0≤φ<2π;
∂ z z=0
V = v (ρ,φ), for ρ≥ a , 0≤φ<2π. (1.5.1)
The problem (1.5.1) can be interpreted as an electrostatic one of a charged

diaphragm, or as an external elastic contact problem. The potential V is
presented through a simple layer distribution (1.4.3). Substitution of the boundary
conditions (1.5.1) in (1.4.3) leads to the governing integral equation
∞ x ρ dρ ρρ
4⌠ 2
dx
⌠
0 0
L  0 σ(ρ ,φ) = v (ρ,φ). (1.5.2)
⌡ ( x − ρ2)1/2 ⌡ ( x 2 − ρ20)1/2  x 2  0
ρ a
Its solution is obtained in exactly the same manner as that of (1.4.5), and is
ρ ∞ ρ dρ
1
σ(ρ,φ) = − 2 L 1 d ⌠ xdx d ⌠ 0 0
1
2 1/2 L ρ v (ρ0,φ).
2
L( x )
πρ ρ dρ ⌡ (ρ − x )
2 2 1/2 d x ⌡ (ρ0 − x )
2
 0
a x
(1.5.3)
The rules of differentiation (1.3.2) and (1.3.9) allow us to rewrite (1.5.3) as
follows
ρ
1  χ( a ,ρ,φ) ⌠
dx ∂ 
σ(ρ,φ) = − 2 + χ( x , ρ, φ) , (1.5.4)
π (ρ2 − a 2)1/2 ⌡ (ρ2 − x 2)1/2 ∂ x
 a 
where
∞ dρ
∂   x2 
χ( x ,ρ,φ) = x⌠ v (ρ0φ).
0
L (1.5.5)
⌡ (ρ0 − x )
2 2 1/2 ∂ρ  ρρ  
0 0
x
The following transformation can now be performed:
∞ dρ
∂ ⌠
(L v )´
∂ 0
χ( x ,ρ,φ) = x
∂x ∂ x  ⌡ (ρ0 − x 2)1/2
2

x
∞ dρ0
= ⌠ 2 2 1/2 [(L v )´ + ρ0(L v )´´ − 2(L´ρ0 v )´]
⌡ (ρ0 − x )
x
∞ ρ dρ
= ⌠ 2
0 0
Lv ´´ + 1 v ´ − L´´ + 1 L´v . (1.5.6)
⌡ (ρ0 − x )   ρ0   ρ0  
2 1/2
x
Here, for the sake of brevity, the primes (´) indicate the partial derivatives with
respect to ρ0, L stands for L( x 2/ρρ0), v ≡ v (ρ0,φ), and the following identity was
used
ρ0
∂  x2 
= − 2 
 L .
∂ x2
L
∂ x ρρ0  x  ∂ρ0 ρρ0
Since
1 ∂2 v 1 ∂2L
L = v,
ρ20 ∂φ2 ρ20 ∂φ2
its addition to and subtraction from (1.5.6) yields
∞ ρ0dρ0
∂
χ( x ,ρ,φ) = ⌠ L∆ v − (∆L) v , (1.5.7)
∂x ⌡ 0( ρ2
− x 2 1/2
)  
x
where ∆ is the two-dimensional Laplace operator in polar coordinates. Since λ

is a harmonic function, ∆L=0, and (1.5.7) simplifies to
∞ ρ0dρ0
∂ ⌠
χ( x ,ρ,φ) = 2 1/2 L∆ v . (1.5.8)
∂x ⌡ (ρ0 − x )
2
x

ρ ∞ ρ0dρ0
1  χ( a ,ρ,φ) ⌠ d x ⌠  x 2 ∆ v (ρ ,φ).
σ(ρ,φ) = − 2 2 + L
π (ρ − a 2)1/2 ⌡ (ρ − x ) ⌡ (ρ0 − x ) ρρ0
2 2 1/2 2 2 1/2 0
 a x 
(1.5.9)
It should be noticed that the first term in (1.5.9) becomes singular when ρ→ a ,
while the second term vanishes at the edge of the disc. In the case of v being
a harmonic function, the second term in (1.5.9) vanishes, and the solution is
represented by the first term only. Further integration with respect to x becomes
possible in (1.5.9), after changing the order of integration and using (1.1.8). The
result is
1  χ( a ,ρ,φ)
σ(ρ,φ) = − 
π2 (ρ2 − a 2)1/2

2π ∞
∆ v (ρ0,φ0) ρ0dρ0dφ0 (ρ2 − a 2)1/2(ρ20 − a 2)1/2 
1⌠⌠
.
-1
+ tan
2π⌡ ⌡ [ρ2 + ρ20 − 2ρρ cos(φ−φ )]1/2 a [ρ2 + ρ20 − 2ρρ cos(φ−φ )]1/2
0 a 0 0 0 0 
(1.5.10)
Solutions, like (1.5.3) and (1.5.9), are appropriate for use when an exact
evaluation of the integrals is possible, while the solution in the form (1.5.10) has
some advantages when numerical integration is to be employed.
Now we can express the potential function V directly through its boundary
value v . Since σ=0 inside the circle ρ= a , the potential function (1.4.3) takes
the form
∞ g(x)
ρ dρ ρρ
V (ρ,φ, z ) = 4⌠ 2  0 σ(ρ ,φ).
dx
⌠ 0 0
2 1/2 L
⌡ ( x − ρ ) ⌡ [ g ( x ) − ρ0]
2 1/2 2
 x2  0
l a
2
(1.5.11)
Substitution of (1.5.3) in (1.5.11) yields, after the first integration,
∞ ∞ ρ dρ
V (ρ,φ, z ) = −
2⌠ dx
L ρ g 2
( x )  ∂ ⌠
0 0
L  v (ρ0,φ).
1
π⌡ ( x − ρ )  x ∂ g ( x ) ⌡ [ρ0 − g ( x )] ρ 
2 2 1/2 2 2 2 1/2
l 0
2 g(x)
(1.5.12)
Here the properties of the L-operators (1.2.3) were used, along with the following
identity, valid for the Abel-type operators
ρ x
⌠ 2 d x 2 1/2 d ⌠ f2( t ) t d2 t1/2 = π f (ρ). (1.5.13)

⌡ (ρ − x ) d x ⌡ ( x − t ) 2
a a
Introduction of a new variable y = g ( x ), x = l ( y ), in (1.5.12), allows us to rewrite

2
(1.5.12)
∞ dl ( y) l 21( y ) ∞ ρ dρ
V (ρ,φ, z ) = −
2⌠ 2
L   d ⌠ 0 0
L  v (ρ ,φ).
1
π⌡ [ l 22( y ) − ρ2]1/2  ρ  d y ⌡ (ρ20 − y 2)1/2 ρ  0
0
a y
(1.5.14)
Interchange of the order of integration in (1.5.14) yields
∞ ρ
2⌠  1 d 0 ydl ( y)
2

l 21( y ) 
 v (ρ ,φ)dρ .
V (ρ,φ, z ) = −
π⌡ Lρ  dρ ⌠ 2 2 1/2 L
 ρ  0
0 ⌡
(ρ0 − y ) [ l 2( y ) − ρ ]
2 1/2 2 0
a  0 a
(1.5.15)
Here the general formula was used
∞ ∞ ∞ x
⌠ F (ρ)dρ d ⌠ 2xf ( x )d2x 1/2 = −⌠ f ( x ) d x d ⌠ ρ2F (ρ) 2dρ1/2 . (1.5.16)

⌡ dρ⌡ ( x − ρ ) ⌡ dx ⌡ (x − ρ )
a ρ a a
The integral in curly brackets of (1.5.15) can be evaluated, according to (1.3.21),

with the result
2π ∞
R
1 ⌠⌠ z 0 -1 
j
V (ρ,φ, z ) = 2 + tan v (ρ ,φ )ρ dρ dφ . (1.5.17)
π ⌡ ⌡ R 30 j R 0 0 0 0 0 0
0 a
Here R is defined by (1.4.22), and

0
(ρ20 − x 2)1/2[ l 22( x ) − x 2]1/2

j( x) = . (1.5.18)
x
The abbreviation j in (1.5.17) stands for j ( a ). In the particular case, when z =0,
expression (1.5.17) simplifies to
2π ∞
v (ρ0,φ0) ρ0dρ0dφ0
1 2 1/2⌠ ⌠
V (ρ,φ,0) = 2 ( a − ρ )
2
,
π ⌡ ⌡ (ρ20 − a 2)1/2[ρ2 + ρ20 − 2ρρ0cos(φ−φ0)]
0 a
for ρ< a ;
V (ρ,φ,0) = v (ρ,φ), for ρ≥ a . (1.5.19)
The general solution is completed. The charge density σ is given by the two
equivalent expressions (1.5.3) and (1.5.10), while the potential is in the two
forms (1.5.14) and (1.5.17), the first one being more convenient for exact
evaluation of the integrals involved, while the second is better suited for
numerical integration.
Problem 2. Consider the problem of finding a harmonic function, vanishing
at infinity, and subject to the mixed conditions on the plane z =0
V = 0, for ρ≤ a , 0≤φ<2π;
∂V
= − 2πσ(ρ,φ), for ρ> a , 0≤φ<2π. (1.5.20)
∂z
The problem may be interpreted as an electrostatic one of a charged infinite

diaphragm, with a grounded disc inside, or as an external crack problem in
elasticity. Substitution of the boundary conditions (1.5.20) in (1.4.3) leads to the
governing integral equation
ρ a ρ dρ
⌠ 2
dx
⌠ 2
0 0
 x  σ(ρ ,φ) 2
2 1/2 L ρρ
⌡ (ρ − x ) ⌡ (ρ0 − x )
2 1/2
 0 0
0 x
∞ x ρ dρ ρρ
= − ⌠
dx
⌠
0 0
 0 σ(ρ ,φ).
2 1/2 L (1.5.21)
⌡ (x − ρ )
2 2 1/2
⌡ ( x − ρ0)
2
 x2  0
a a
One should notice that σ in the second term of (1.5.21) is known from the
boundary condition (1.5.20), while the value of σ in the first term is yet to be
determined. The right hand side of (1.5.21) can be transformed, by using
(1.1.26),
ρ a ρ dρ
⌠ 2
dx
⌠ 2
0 0
 x  σ(ρ ,φ) 2
2 1/2 L ρρ
⌡ ( ρ − x 2 1/2
) ⌡ 0
( ρ − x )  0 0
0 x
ρ ∞ ρ dρ
= − ⌠ 2 ⌠
0 0 dx  x  σ(ρ ,φ), 2
2 1/2 L ρρ
⌡ (ρ − x ) ⌡ (ρ0 − x )
2 1/2 2
 0 0
0 a
with an immediate result
a ρ0dρ0 ∞ ρ0dρ0
⌠ 2 L  x  σ(ρ ,φ) = − ⌠ L  σ(ρ0,φ).
x
(1.5.22)
⌡ (ρ0 − x )
2 1/2
ρ0 0
⌡ (ρ0 − x )
2 2 1/2
ρ0
x a
Application of the operator
L 
d ⌠ xdx 1
L(ρ)
dρ ⌡ ( x 2 − ρ2)1/2 x 
ρ
to both sides of (1.5.22) gives, after necessary transformations
∞
(ρ20 − a 2)1/2
2 ⌠ ρ
σ(ρ,φ) = − L( ) σ(ρ ,φ) ρ dρ , for ρ< a ,
π( a 2
− ρ2)1/2 ⌡ ρ20 − ρ 2 ρ 0 0 0
0
a
(1.5.23)
or, interpreting the L-operator, we obtain
2π ∞
(ρ20 − a 2)1/2σ(ρ ,φ ) ρ dρ dφ
σ(ρ,φ) = − 2 2
1 ⌠⌠ 0 0 0 0 0
. (1.5.24)
π ( a − ρ2)1/2 ⌡ ⌡ ρ +
2
ρ20 − 2ρρ0cos(φ−φ0)
0 a
Now the value of σ is known all over the plane z =0, and (1.4.3) can be used
in order to express the potential V directly through the prescribed σ.
Substitution of (1.5.23) in (1.4.3) yields, after the first integration
l1 ∞ ρ dρ
V (ρ,φ, z ) = − 4⌠ 2 ⌠ 2
0 0 dx  x  σ(ρ ,φ) 2
1/2 L ρρ
⌡ (ρ − x ) ⌡ [ρ0 − g ( x )]
2 1/2 2
 0 0
0 a
∞ g(x)
ρ dρ ρρ
+ 4⌠ 2 ⌠ 0 0 dx 0 σ(ρ ,φ).
2 1/2 L (1.5.25)
⌡ ( x − ρ ) ⌡ [ g ( x ) − ρ0]
2 1/2 2
 x2  0
l a
2
The second term in (1.5.25) is equivalent to the second term in (1.4.2), which,
in turn, can be represented by using (1.1.18), as
∞ l (ρ )
 1 0  x 2  σ(ρ ,φ) ρ dρ .
4⌠  ⌠
dx
L
⌡  ⌡ (ρ − x ) [ρ0 − g ( x )]
2 2 1/2 2 2 1/2
ρρ0 0 0 0
a 0
The following scheme of changing the order of integration is enacted
l1(ρ 0) l (∞)
∞ l1 ∞ 1 ∞
⌠ dρ ⌠ d x = ⌠ d x ⌠ dρ + ⌠ d x ⌠ dρ ,
⌡ 0 ⌡ ⌡ ⌡ 0 ⌡ ⌡ 0
0 l
a 0 a 1 g(x)
and the second term in (1.5.25) can be rewritten as
l1 ∞ ρ dρ
4⌠ 2 ⌠ 2
0 0 dx  x  σ(ρ ,φ) 2
1/2 L ρρ
⌡ (ρ − x ) ⌡ [ρ0 − g ( x )]
2 1/2 2
 0 0
0 a
l (∞)
1 ∞ ρ dρ
⌠ dx
⌠ 2
0 0
 x  σ(ρ ,φ). 2
+ 4
⌡ (ρ − x 2)1/2 1/2 L ρρ (1.5.26)
l
2
⌡ [ρ0 − g ( x )]
2
 0 0
1 g(x)
Substitution of (1.5.26) in (1.5.25) gives, by virtue of l 1(∞)=ρ,
ρ ∞ ρ dρ
V (ρ,φ, z ) = 4⌠ 2  x  σ(ρ ,φ).
dx 2
⌠ 2
0 0
1/2 L ρρ (1.5.27)
⌡ (ρ − x 2)1/2 ⌡ [ρ0 − g ( x )]
2
 0 0
l g(x)
1
Interchange of the order of integration in (1.5.27), and integration with respect to

x , according to (1.1.24), results in
2π ∞
2⌠ ⌠ 1 j
V (ρ,φ, z ) = tan-1( ) σ(ρ0,φ0) ρ0dρ0dφ0, (1.5.28)
π⌡ ⌡ R0 R0
0 a
where R 0 is defined by (1.4.22), and j stands for j ( a ), as defined by (1.3.4).
The second problem is now solved. Expression (1.5.24) defines the charge
density σ inside a circle directly in terms of its values outside. The potential V
is given by two equivalent expressions (1.5.27) and (1.5.28), the first one to be
used for exact evaluation of the integrals, while the second has some advantages
in the case of numerical integration. Some specific examples are considered
below.
Example 1. Consider an external mixed problem with the following
boundary conditions at z =0
V = v 0/ρn, for ρ≥ a , 0≤φ<2π;
∂V
= 0, for ρ< a , 0≤φ<2π. (1.5.29)
∂z
The conditions (1.5.29) correspond to those of Problem 1. The solution is given

by (1.5.14) and (1.5.3). Substitution of (1.5.29) in (1.5.14) yields, after the
integration
2 v 0 Γ[( n + 1)/2] ∞
V (ρ,φ, z ) = ⌠ dx
, (1.5.30)
√π Γ( n /2) ⌡ ( x − ρ2)1/2 g n( x )
2
l
2
where g ( x ) is defined by (1.1.25), and the following integral was employed

(Gradshtein and Ryzhik, 1963)
⌠ n 2 dρ 2 1/2 = √π Γ( n /2)
. (1.5.31)
⌡ ρ (ρ − x ) 2Γ[( n + 1)/2] x n
x
The integral in (1.5.30) can be evaluated in terms of elementary functions for

any integer n , but the procedure is slightly different for even and odd values of
n . For example, for even n =2 k , the problem reduces to the evaluation of the
integral
⌠ (x − ρ ) dx ,
2 2 k-1/2
⌡ x 2k( x 2 − ρ2 − z2)k
l
2
which can be evaluated by introduction of a new variable t = x /( x 2 − ρ2)1/2. The

final result is
2 v 0 Γ[( n + 1)/2]  k Am
V (ρ,φ, z ) =
√π Γ( n /2) z n Σ
 2 m − 1 [1 − Q 0 ]
 m=1
2m-1
( Q m-1 − Q m-1) − ( Q m-1 − Q m-1),

k B
Σ
m
+ 2 B 1ln Q + (1.5.32)
1 − m  1 2 3 4

m=2
where
1 dm-1 (η − 1)k-1
A = , for η=0, and r 2=1+ρ2/ z 2;
k-m+1 ( m − 1)! dηm-1  ( r 2 − η)k 
1 dm-1  ( t 2 − 1)k-1 
B k-m+1 = , for t =√1+ρ2/ z 2;
( m − 1)! d t m-1 t 2k( r 2 + t )k
( l 22 − ρ2)1/2 l 2[(ρ2 + z 2)1/2 + ( l 22 − a 2)1/2]

Q0 = , Q = ,
l2 a [(ρ2 + z 2)1/2 + z ]
z [(ρ2 + z 2)1/2 + ( l 22 − a 2)1/2] z [(ρ2 + z 2)1/2 − ( l 22 − a 2)1/2]

Q1 = , Q2 = ,
l 21 l 21
z [(ρ2 + z 2)1/2 + z ] z [(ρ2 + z 2)1/2 − z ]

Q3 = , Q4 = . (1.5.33)
ρ2 ρ2
For the case of an odd n =2 k +1, the integration can be performed by using the
substitution t =( x 2−ρ2− z 2)1/2, and the final result is
v 0 Γ[( n + 1)/2]  k Cm k+1


V (ρ,φ, z ) =
√π Γ( n /2)  Σ
 m=1 (2 m − 1)( a − l 1)
2 2 m-1/2 + Σm=1
D m E m ,

(1.5.34)
where
1 dk-m  ( t + z 2)k ,
Cm = for t =0;
( k − m )! d t  ( t + ρ + z ) 
k-m 2 2 k+1
1 dk+1-m ( t + z 2)k
Dm = , for t =−(ρ2+ z 2);
( k + 1 − m )! d t k+1-m  tk 
(−1)m dm-1  1 √t ,
Em = tan-1 2 for t =ρ2+ z 2.
( m − 1)! d t √ t
m-1
( a − l 21)1/2
(1.5.35)
Substitution of (1.5.29) in (1.5.3) yields, after integration
v 0Γ[( n +1)/2] 1 n (ρ2 − a 2)1/2 1 1 3 a 2

σ(ρ,φ) =  n 2 − F
2
n +1, ; ; 1 − ,
π3/2Γ( n /2) ρ2
a ( ρ − a ) ρn+2
2 1/2 2 2

(1.5.36)
and the Gauss hypergeometric function can be expressed in elementary functions
(Bateman and Erdelyi, 1955), namely, for even n =2 k , k =1,2,3 ...,
F  k + 1,
1 3  1 dk  k-1/2 1 + √ t 
; ; t = t ln ,
 2 2  2k! dtk  1 − √t 
and for odd n =2 k +1, k =0,1,2, ...,
F k +
3 1 3  √π dk  t k+1/2 
, ; ; t = .
 2 2 2 
2√ t Γ( k + ) d t √1 − t 
3 k
2
(1.5.37)
The charge density distribution, evaluated due to (1.5.36), is non-negative for
n =1, and changes sign when n ≥2, its negative maximum increases with n , while
the total charge stays at zero.
Example 2. Consider the boundary conditions at z =0
in φ
V = ( v /ρn) e , for ρ≥ a , 0≤φ<2π;
n
∂V
= 0, for ρ< a , 0≤φ<2π, (1.5.38)
∂z
where v is constant. The solution is given by (1.5.3) and (1.5.14).

n
Substitution of (1.5.38) in (1.5.14) yields, after integration
∞
2Γ( n + 1/2) n in φ ⌠ dx
V (ρ,φ, z ) = ρe . (1.5.39)
√πΓ( n ) ⌡ x ( x − ρ2)1/2
2n 2
l
2
The final integration gives
2vn n
V (ρ,φ, z ) =
√πρn
in φ
e Σ
k=1
(−1)k-1Γ( n + 1/2)
(2 k − 1)Γ( k )Γ( n + 1 − k )
(1 − Q 2k-1
0 ), (1.5.40)
where Q 0 is defined by (1.5.33). Substituting (1.5.38) in (1.5.3), we get, after

integration,
Γ( n + 1/2) v nein φ
σ(ρ,φ) = . (1.5.41)
π3/2Γ( n ) ρn(ρ2 − a 2)1/2
Evidently, expression (1.5.41) can also be obtained by differentiation of (1.5.40)

with respect to z for z =0.
Example 3. Consider a case related to Problem 2, with the boundary
conditions
V = 0, for ρ≤ a , 0≤φ<2π;
σ0
∂V
= −2π n for ρ> a , 0≤φ<2π; (1.5.42)
∂z ρ
The solution is given by (1.5.23) and (1.5.27). Substitution of (1.5.42) in

(1.5.27) yields, after integration using (1.5.31),
ρ
Γ[( n − 1)/2] ⌠ dx
V (ρ,φ, z ) = 2√πσ0 , (1.5.43)
Γ( n /2) ⌡ (ρ − x 2)1/2 g n-1( x )
2
l
1
where g ( x ) is defined by (1.1.25). The technique used in the previous example

can be employed here for further integration. The final result depends on the
value of n being even or odd. For even n =2 k , k =1,2,3, ..., the potential is
Γ[( n − 1)/2] 
V (ρ,φ, z ) = 2√πσ0
Γ( n /2) 2 B 1ln Q

k-1 Am ( a 2 − l 21)1/2
+ Σ(2m
m=1
− 1) z 2m-1
1 − 
  a
2m-1
 
k-1 B m+1 
+ Σ mz
m=1
m [ Q 1 − Q 2 − ( Q 3 − Q 4 )].
m m m m

(1.5.44)
Here Q , Q , Q , Q , and Q are defined by (1.5.33), and

1 2 3 4
1 dm-1  ( t − z 2)k-1 
A = for t =0;
k-m ( m − 1)! d t m-1 (ρ2 + z 2 − t )k
1 dm-1  ( t 2 − z 2)k-1 ,
B = for t =−(ρ2 + z 2)1/2.
k-m+1 ( m − 1)! d t m-1 t 2k-2[(ρ2 + z 2)1/2 − t ]k
(1.5.45)
For odd n =2 k +1, the result is
σ Γ[( n − 1)/2] k
 Gm ( l 22 − a 2)1/2 2m-1 
Σ  
0
V (ρ,φ, z ) = 2√π  2m − 1  
− H L .
Γ( n /2) z n-1 a m m
m=1  
(1.5.46)
Here
1 dm-1 (1 + t )k-1
G k-m+1 = , for t =0, ξ=(ρ2 + z 2)/ z 2;
Γ( m ) d t m-1  (ξ + t )k 
1 dm-1 (1 + t )k-1

H k-m+1 = , for t =−(ρ2+ z 2)/ z 2;
Γ( m ) d t m-1
 t k

(−1)m dm-1  1
tan-1 ( l 22 − a 2)1/2,
√t
Lm = for t =(ρ2+ z 2)/ z 2.
Γ( m ) d t m-1 √ t a 
(1.5.47)
Example 4. Consider the boundary conditions on the plane z =0:
V = 0, for ρ≤ a , 0≤φ<2π;
∂V
= − 2π(σ /ρn)ein φ, for ρ> a , 0≤φ<2π; (1.5.48)
∂z n
Substitution of (1.5.48) in (1.5.27) yields, after integration,
σ Γ( n − 1/2) 
ein φ( l 22 − a 2)1/2 − z
n
V (ρ,φ, z ) = 2√π
Γ( n )ρn

n

− z Σ
k=2
(−1)kΓ( n )
Γ( k ) Γ( n − k +1)(2 k − 3)
[1 − (1 − l 21/ a 2)k-3/2],

(1.5.49)
and on the plane z =0
σ Γ( n − 1/2)
ein φℜ(ρ2 − a 2)1/2.
n
V (ρ,φ,0) = 2√π
Γ( n )ρn
The symbol ℜ indicates the real part sign. The charge density is defined,
according to (1.5.23),
σ Γ( n − 1/2) 
ein φℜ1 −
n a
σ(ρ,φ) =
√π Γ( n )ρ n
( a − ρ2)1/2
2

n

+ Σk=2
(−1)kΓ( n )
Γ( k ) Γ( n − k +1)(2 k − 3)
[1 − (1 − ρ2/ a 2)k-3/2].

(1.5.50)
A more general case of boundary conditions, namely,
V = 0, for ρ≤ a , 0≤φ<2π;
∂V
= − 2π(σ /ρj)ein φ, for ρ> a , 0≤φ<2π; (1.5.51)
∂z jn
can also be considered, by using the same technique as in the previous examples,
and the final result can always be expressed in elementary functions. The form
of the result will be different for ( j + n ) even, and for ( j + n ) odd. As an
example, the following expression can be obtained by substituting (1.5.51) in
(1.5.23), for the case when j + n =2 k
σjn Γ( m − 3/2) ρ2m-2

k
σ(ρ,φ) = j e ℜ1 −
ρ

in φ a 1 −
( a − ρ2)1/2
2 Σ
m=2
2√πΓ( m ) a  
,

(1.5.52)
and for odd j + n =2 k +1
2σjn  -1 ρ √πΓ( m − 1) ρ2m-2

k
σ(ρ,φ) =
πρj e ℜ

in φ
sin (a ) − 2
a 
2 1/2 1 −
(a − ρ )  Σm=2
4Γ( m + 1/2) a  

,
(1.5.53)
Expressions (1.5.52) and (1.5.53) represent general formulae which cover all the
particular cases considered in Examples 3 and 4.
The examples above have demonstrated the simplicity of the method. The
generation of the solution is reduced to a straightforward and elementary
procedure.
Exercise 1.5
1. The following boundary conditions are prescribed at z =0
V = v 0/ρ, for ρ≥ a , 0≤φ<2π;
∂V
= 0, for ρ< a , 0≤φ<2π.
∂z
Find the potential function and the charge distribution.
2v0
-1(ρ + z ) 
2 2 1/2
Answer: V (ρ,φ, z ) = sin ,
π(ρ2 + z 2)1/2  l2 
v0a
σ(ρ,φ) = .
π2ρ2(ρ2 − a 2)1/2
The total charge is equal v .
0
V = v 0/ρ2, for ρ≥ a , 0≤φ<2π;
∂V
= 0, for ρ< a , 0≤φ<2π.
∂z
v0 ( l 22 − ρ2)1/2
Answer: V (ρ,φ, z ) = 1 −
ρ2 + z 2  l2
l 2[(ρ2 + z 2)1/2 + ( l 22 − a 2)1/2]

+
z
ln ,
(ρ + z 2)1/2
2
a [(ρ + z )
2 2 1/2
+ z] 
v0
σ(ρ,φ) =  1
−
1 ρ + (ρ2 − a 2)1/2
ln .
2πρ2(ρ2 − a 2)1/2 ρ a 
V = v 0/ρ3, for ρ≥ a , 0≤φ<2π;
∂V
= 0, for ρ< a , 0≤φ<2π.
∂z
4v0 2 l 21( a 2 − l 21)1/2
Answer: V (ρ,φ, z ) = 2  z
−
(ρ + z 2)2 ( a 2 − l 21)1/2 2a2
ρ2 − 2 z 2 -1(ρ2 + z 2)1/2
+ sin .
2(ρ2 + z 2)1/2  l2 
Note that the potential at the coordinate origin is finite, namely,
V (0,0,0)=4 v 0/(3π a 3).
2 v 0(2 a 2 − ρ2)
σ(ρ,φ) = .
π2 a ρ4(ρ2 − a 2)1/2
V = v 0/ρ4, for ρ≥ a , 0≤φ<2π;
∂V
= 0, for ρ< a , 0≤φ<2π.
∂z
ρ2 − z2 
3v0 ( l 22 − ρ2)1/2

Answer: V (ρ,φ, z ) =  1 −
2(ρ2 + z 2)2 ρ2 + z 2  l2 

( l 22 − ρ2)1/2 3 l 22
1     ( l 22 − a 2)1/2 − z 
z
− 1 − +
3  l2  2(ρ + z )a
2 2 2

z (2 z 2 − 3ρ2) l [(ρ2 + z 2)1/2 + ( l 22 − a 2)1/2]

2
− ln ,
2(ρ + z )
2 2 3/2
a [(ρ2 + z 2)1/2 + z ]
The last expression simplifies at z =0:
v ( a 2 − ρ2)1/2 ρ2( a 2 − ρ2)1/2


V (ρ,φ,0) = 4 1 −
0
− , for ρ≤ a ;
ρ  a 2a3 
and V (ρ,φ,0) = v /ρ4, for ρ> a . Note that the potential at the coordinate
0
origin is finite, namely, V (0,0,0)=3 v /(8 a 4).
0
 3 a 2 − ρ2
3v
0 3 ρ + (ρ2 − a 2)1/2
σ(ρ,φ) =  − ln
ρ  .
8πρ4 a 2(ρ2 − a 2)1/2 a
 
iφ
V = ( v 1/ρ) e , for ρ≥ a , 0≤φ<2π;
∂V
= 0, for ρ< a , 0≤φ<2π.
∂z
Find the potential function.
v ( a 2 − l 21)1/2
1 iφ  
Answer: V (ρ,φ, z ) = e 1 −
ρ  a 
iφ
V = ( v 2/ρ2) e2 , for ρ≥ a , 0≤φ<2π;
∂V
= 0, for ρ< a , 0≤φ<2π.
∂z
3v
2 2 iφ
 ( a 2 − l 21)1/2
 1  
( a 2 − l 21)1/2 3 
 .
Answer: V (ρ,φ, z ) = 2 e 1 − 
−

1 −
 
2ρ a 3 a
 
V = ( v /ρ3) e3 i φ, for ρ≥ a , 0≤φ<2π;

3
∂V
= 0, for ρ< a , 0≤φ<2π.
∂z
15 v  ( a 2 − l 21)1/2
e3 i φ  1 − 
3 1
4ρ3 2 a 

( a 2 − l 21)1/2 3 ( a 2 − l 21)1/2 5 
1
1 −    + 1 −   .
1
−
3  a  10  a  
8. Let the following boundary conditions be prescribed at z =0:
V = 0, for ρ≤ a , 0≤φ<2π;
∂V
= − 2πσ /ρ2, σ =const, for ρ> a , 0≤φ<2π.
∂z 0 0
2πσ l [(ρ2 + z 2)1/2 + ( l 22 − a 2)1/2]
0 2
Answer: V (ρ,φ, z ) = 2 2 1/2 ln ,
(ρ + z ) a [(ρ2 + z 2)1/2 + z ]
σ σ
σ(ρ,φ) = 2 ℜ1 − ,
0 a 0
σ(0,0) = − .
ρ  ( a − ρ2)1/2
2
2a2
V = 0, for ρ≤ a , 0≤φ<2π;
∂V
= − 2πσ /ρ3, σ =const, for ρ> a , 0≤φ<2π.
∂z 0 0

4σ0 ( l 22 − a 2)1/2 (ρ2 + z 2)1/2
Answer: V (ρ,φ, z ) = 2  − 2
z
sin-1
,
ρ + z2  a (ρ + z 2)1/2  l2 
2σ0
σ(ρ,φ) = 1 sin-1(ρ) − 1 
2 1/2 , for ρ< a ; σ(0,0) = −2σ0/(3π a ).
3
πρ 
2 ρ a (a − ρ ) 
2
V = 0, for ρ≤ a , 0≤φ<2π;
∂V
= − 2πσ0/ρ4, σ0=const, for ρ> a , 0≤φ<2π.
∂z
πσ0 2 z 2 l 22
Answer: V (ρ,φ, z ) =  ( a − l 1) − 3 z + 2 ( l 2 − a )
2 1/2 2 2 1/2
2(ρ2 + z 2)2 a a

ρ2 − 2 z 2 l 2[(ρ2 + z 2)1/2 + ( l 22 − a 2)1/2]

ln ,
(ρ2 + z 2)1/2 a [(ρ2 + z 2)1/2 + z ]

σ0 σ0
σ(ρ,φ) = 4 ℜ1 −
2 a 2 − ρ2 
, σ(0,0) = − .
ρ  2 a ( a 2 − ρ2)1/2 8a4
11. Consider the boundary conditions on the plane z =0:
V = 0, for ρ≤ a , 0≤φ<2π;
∂V iφ
= − 2π(σ1/ρ)e , for ρ> a , 0≤φ<2π.
∂z
iφ
Answer: V (ρ,φ, z ) = 2π(σ1/ρ) e [( l 22 − a 2)1/2 − z ],
iφ
σ(ρ,φ) = (σ1/ρ)e ℜ[1 − a /( a 2 − ρ2)1/2].
V = 0, for ρ≤ a , 0≤φ<2π;
∂V
= − 2π(σ /ρ2)e2 i φ, for ρ> a , 0≤φ<2π.
∂z 2
Answer: V (ρ,φ, z ) = π(σ /ρ2) e2 i φ[( l 22 − a 2)1/2 − 2 z + z ( a 2 − l 21)1/2/ a ],
2
σ(ρ,φ) = (σ /ρ2)ei φℜ1 −

2 a 2 − ρ2 
.
2  2 a ( a 2 − ρ2)1/2
V = 0, for ρ≤ a , 0≤φ<2π;
∂V
= − 2π(σ /ρ3)e3 i φ, for ρ> a , 0≤φ<2π.
∂z 3
3πσ
3 3 i φ 2 8
Answer: V (ρ,φ, z ) = 3 e ( l 2 − a 2)1/2 − z
4ρ  3
( a 2 − l 21)1/2 3
z 1   ,
+ 2 ( a − l 1) − z
2 2 1/2
a 3  a 
σ
e3 i φℜ1 −
3 3a 3( a 2 − ρ2)1/2 ( a 2 − ρ2)3/2
σ(ρ,φ) = − + .
ρ3  8( a 2 − ρ2)1/2 4a 8a3 
14. Prove that the total charge Q in Problem 2 (1.5.20) can be expressed
T
directly in terms of the given charge density σ as
2π ∞
2⌠ ⌠ a
Q = σ(ρ,φ) cos-1( ) ρdρdφ.
T π⌡ ⌡ ρ
0 a
Hint : integrate (1.5.23).
15. Solve the problem above in the case when σ=σ /ρn.
0
2σ √π Γ[( n − 1)/2]
0
Answer: Q = .
T ( n − 2) Γ( n /2) a n-2
1.6 Some fundamental integrals
The integrals are called fundamental because of their primary importance to

the new method, and also because almost all the integral representations, derived
earlier, are just particular cases of the fundamental ones to be evaluated here.
Consider three points in the system of cylindrical coordinates, namely, M (ρ,φ, z ),
M 0(ρ0,φ0, z 0), and N ( r ,ψ,0). The following notation is introduced:
1
l 1(t) = {[(ρ + t )2 + z 2]1/2 − [(ρ − t )2 + z 2]1/2} , (1.6.1)
2
1
l 2(t) = {[(ρ + t )2 + z 2]1/2 + [(ρ − t )2 + z 2]1/2} , (1.6.2)
2
1
l 10(t) = {[(ρ0 + t )2 + z 20]1/2 − [(ρ0 − t )2 + z 20]1/2} , (1.6.3)
2
1
l 20(t) = {[(ρ0 + t )2 + z 20]1/2 + [(ρ0 − t )2 + z 20]1/2}. (1.6.4)
2
According to the earlier convention, l 10 stands as an abbreviation for l 10( a ), etc.;

R (⋅,⋅) denotes the distance between two points.
Consider the following integral:
2π a
h
I1 = ⌠ ⌠ 3 tan-1   rd rdψ ,
z 1 0
(1.6.5)
⌡ ⌡ R ( M,N) R ( N,M0)  R ( N,M0) 
0 0
where
h 0 = [ a 2 − l 10]1/2 [ a 2 − r 2]1/2/ a (1.6.6)
Make use of the integral representation (1.1.23)
h0 a d l 20( x ) ρ0 r
-1   , ψ−φ0,
1
tan = ⌠ 2 1/2 λ 2
R ( N,M )
0
R ( N,M )
0 ⌡ 20
[ l 2
( x ) − ρ2 1/2
0 ] ( x 2
− r ) l 20 ( x ) 
r
(1.6.7)
where λ(⋅,⋅) is defined by (1.1.5).
The substitution of (1.6.7) in (1.6.5) yields, after changing the order of

integration:
2π a
d l 20( x ) x ρ0 r
⌠ dψ ⌠ ⌠
rdr  , ψ−φ0 3
z
I1 = 2 1/2 λ 2 .
⌡ ⌡ [ l 20( x ) − ρ0]
2 2 1/2
⌡ (x − r )
2
l 20( x )  R ( M,N)
0 0 0
(1.6.8)
By substituting the integral representation (1.3.7),
z z
≡
R 3( M,N ) [ρ2 + r 2 − 2 r ρ cos(φ − ψ) + z 2]3/2
r
[ t 2 − l 21( t )]1/2 l 1( t ) t
=
2 1 d ⌠
L( )
tdt
λ , φ−ψ, (1.6.9)
πr r dr ⌡ (r − t )
2 2 1/2
l 2( t ) − l 1( t )  l 2( t )
2 2

0
in (1.6.8), the following result can be obtained, after integration with respect to
ψ:
a d l 20( x ) x r
dr d tdt
I 1 = 4⌠ ⌠ 2 ⌠ 2
⌡ [ l 20( x ) − ρ0] ⌡ ( x − r ) d r ⌡ ( r − t 2)1/2
2 1/2 2 1/2
0 0 0
[ t 2 − l 21( t )]1/2 l 1( t ) t ρ0
× 
λ 2 , φ−φ0. (1.6.10)
l 22 − l 21 l 20( x ) l 2( t ) 
Here the following property of the L-operators was used:
L( k ) λ( k 1,⋅) = λ( k k 1,⋅) , for k , k 1 <1 (1.6.11)
The well known property of the Abel-type operators, namely,
x r
⌠ dr d ⌠ f(t) tdt π
2 1/2 = 2 f ( x ) , (1.6.12)
⌡ (x − r ) ⌡ (r − t )
2 2 1/2 d r 2
0 0
allows us to simplify (1.6.10) significantly:

a
[ x 2 − l 21( x )]1/2 [ x 2 − l 210( x )]1/2 l ( x) l ( x)
I 1 = 2π⌠ 2 λ 1 10 , φ−φ d x .
⌡ l 2( x ) − l 1( x ) l 20( x ) − l 10( x )  2 20 0
2 2 2 l ( x ) l ( x )
0
(1.6.13)
It is noteworthy that the integrand in (1.6.13) is symmetric with respect to the
points M and M while it did not look so in the original expression (1.6.5).
0
The integrand in (1.6.13) is a perfect differential, so that the integral can be
evaluated as indefinite:
[ x 2 − l 21( x )]1/2 [ x 2 − l 210( x )]1/2 l ( x) l ( x)

⌠ λ 1 10 , φ−φ  d x
⌡ l 22( x ) − l 21( x ) l 220( x ) − l 210( x ) l 2( x ) l 20( x ) 0
Θ ( x) Θ ( x)
1 1 1 2
-1 -1
= tan + tan , (1.6.14)
2R R 2R R
1 1 2 2
where
R = [ρ2 + ρ20 − 2ρρ cos(φ − φ ) + ( z − z )2]1/2 ,

1 0 0 0
R = [ρ2 + ρ20 − 2ρρ cos(φ − φ ) + ( z + z )2]1/2 ,

2 0 0 0
Θ ( x ) = θ( x ) + zz /θ( x ) , Θ ( x ) = θ( x ) − zz /θ( x ) ,
1 0 2 0
θ( x ) = [ x 2 − l 21( x )]1/2 [ x 2 − l 210( x )]1/2/ x . (1.6.15)
Notice that when z =0, θ( x ) transforms into h ( x ) as it is defined by (1.3.11),

0
and the integral (1.6.14) reduces to the one considered in Exercise 1.1.8.
Correctness of the integral in (1.6.14) can be verified by differentiation. The
algebra involved is not trivial. Here we present some intermediate
transformations:
θ( x ) [ l 22( x ) l 220( x ) − l 21( x ) l 210( x )]

∂
θ( x ) = , (1.6.16)
∂x x [ l 22( x ) − l 21( x )] [ l 220( x ) − l 210( x )]
l ( x) l ( x) l 22( x ) l 220( x ) − l 21( x ) l 210( x )

λ , φ−φ  =  
1 10 1 1
+ 2
l ( x ) l ( x )
2 20
0 2x 2
R 1 + Θ1( x )
2 2
R 2 + Θ22( x )
(1.6.17)
Formula (1.6.14) allows us to evaluate the integral (1.6.5):
2π a
h
tan-1   rd rdψ
⌠⌠ z 1 0
⌡ ⌡ R ( M,N) R ( N,M0)
3
 R ( N,M0) 
0 0
Θ Θ
| z |  1  -1 1
| zz | | zz | 
= π  tan −
π 0
 +
1  -1 2
tan +
π 0
,
z R 1 R 1  2 zz0  R 2 R 2  2 zz 
 0 
(1.6.18)
where the contractions Θ and Θ stand for Θ ( a ) and Θ ( a ) respectively. Note
1 2 1 2
an important particular case when z 0=0. Formula (1.6.18) in this case transforms
into:
2π a
( a 2 − r 2)1/2 ( a 2 − ρ20)1/2
⌠⌠ z 1
tan -1   rd rdψ
⌡ ⌡ R 3( M,N) R ( N,N0)  a R ( N,N0) 
0 0
tan-1   , for ρ < a ,

2π h
= (1.6.19)
R ( M,N 0)  R ( M,N 0
)  0
and the integral vanishes when ρ0≥ a . Here the point N 0 has the cylindrical
coordinates (ρ0,φ0,0), and h is defined by (1.4.22).
The second fundamental integral to be considered is:
2π a
z R ( N,M 0) h0
I2 = ⌠ ⌠ 3
0
 + tan -1  rd rdψ ,
⌡ ⌡ R ( N,M0)  h0 R ( N,M 0)  R ( M,N )
0 0
(1.6.20)
where h 0 is defined by (1.6.6). Make use of the integral representation for the
reciprocal distance
r
[ l 22( x ) − x 2]1/2 l 1( x ) x
1
=
2 ⌠ dx
λ , φ−ψ,
R ( M,N ) π ⌡ (r − x )
2 2 1/2
l 2( x ) − l 1( x )  l 2( x ) r
2 2

0
(1.6.21)
which is a variation of (1.1.26). By substituting (1.6.21) in (1.6.20), the
following expression results, after changing the order of integration:
2π
[ l 2( x ) − x 2]1/2
a 2 a
rdr
2 ⌠
I2 = dψ ⌠ 2 dx ⌠
π ⌡ ⌡ l 2( x ) − l 1( x )
2
⌡ (r − x )
2 2 1/2
0 0 x
l ( x) x z R ( N,M ) h
× λ , φ − ψ  3   ,
1 0 0 0
-1
+ tan (1.6.22)
 l 2( x ) r  R ( N,M0)  h
0
R ( N,M ) 
0
We use the integral representation (1.3.14):
z R ( N,M ) h
 0 0
-1
+ tan
0

R ( N,M ) 
3 h
0
R ( N,M ) 
0
0
a
[ l 220( t ) − t 2]1/2 l (t)
= −
L( r ) d ⌠ tdt
λ 10
, φ −ψ.
r d r ⌡ ( t 2 − r 2)1/2 l 220( t ) − l 210( t )  t l ( t ) 0 
20
r
(1.6.23)
The substitution of (1.6.23) in (1.6.22) yields, after integration with respect to ψ:
a
[ l 22( x ) − x 2]1/2 a
dr d tdt
a
I 2 = −4 ⌠ dx ⌠ ⌠ 2
⌡ l 2( x ) − l 1( x ) ⌡ (r − x ) d r ⌡ ( t − r 2)1/2
2 2 2 2 1/2
0 x r
[ l 220( t ) − t 2]1/2 l ( x) l ( t) x
× 2 λ 1 10 , φ−φ . (1.6.24)
l 20( t ) − l 10( t )  l 2( x ) l 20( t ) t
2 0
We recall another well known property of the Abel operators:
a a
⌠ dr d ⌠ f(t)tdt π
2 1/2 = − 2 f ( x ). (1.6.25)
⌡ (r − x ) ⌡ (t − r )
2 2 1/2 d r 2
x r
Application of (1.6.25) to (1.6.24) yields:
a
[ l 22( x ) − x 2]1/2 [ l 220( x ) − x 2]1/2 l ( x) l ( x)
I 2 = 2π⌠ 2 λ 1 10 , φ−φ d x .
⌡ l 2( x ) − l 1( x ) l 20( x ) − l 10( x )  l 2( x ) l 20( x ) 0
2 2 2
0
(1.6.26)
Note certain similarity between (1.6.26) and (1.6.13). The integrand in (1.6.26)
is a perfect differential, and can be evaluated in elementary functions:
[ l 22( x ) − x 2]1/2 [ l 220( x ) − x 2]1/2 l ( x) l ( x)

⌠ λ 1 10 , φ − φ0  d x
⌡ l 2( x ) − l 1( x ) l 20( x ) − l 10( x )  l 2( x ) l 20( x )
2 2 2 2

Ξ1( x ) Ξ2( x )
1 1
= − tan-1 − tan-1 , (1.6.27)
2R1 R1 2R2 R2
where R 1 and R 2 are defined by (1.6.15), and
Ξ1( x ) = ξ( x ) + zz0/ξ( x ) , Ξ2( x ) = ξ( x ) − zz0/ξ( x ) ,
ξ( x ) = [ l 22( x ) − x 2]1/2 [ l 220( x ) − x 2]1/2/ x (1.6.28)
Again, one can notice that in the case when z 0=0, ξ( x ) transforms into j ( x ) as it
is defined in (1.3.4), and the integral (1.6.27) coincides with (1.1.26). As
before, correctness of the integration can be verified by differentiation, using
(1.6.17) and the property:
ξ( x ) [ l 22( x ) l 220( x ) − l 21( x ) l 210( x )]

∂
ξ( x ) = − , (1.6.29)
∂x x [ l 22( x ) − l 21( x )] [ l 220( x ) − l 210( x )]
Finally, the integral (1.6.20) can be evaluated as follows:
2π a
z R ( N,M 0) h0
⌠⌠ 0
 + tan -1  r d r dψ
⌡ ⌡ R ( N,M0) 
3 h0 R ( N,M 0)  R ( M,N)
0 0
| z 0|  1 π Ξ1 1  π Ξ2 
   .
z0  R 1  2 R 2 
= π − tan-1
+ − tan-1
(1.6.30)
R1  R2  2
 
According to our convention, Ξ1 and Ξ2 denote Ξ1( a ) and Ξ2( a ) respectively.
Consider a particular case, when z 0=0, and ρ0> a . Due to the relationship
az 0
→ (ρ20 − a 2)1/2 , for z 0→0 ,
(a 2
− l 210)1/2
the integral (1.6.30) will take the form:
2π a
(ρ2 − a 2)1/2
r d r dψ
⌠⌠ 0
⌡ ⌡ ( a − r ) R ( M,N) R 2( N0, N)
2 2 1/2
0 0
( l 22 − a 2)1/2(ρ20 − a 2)1/2
2π  π .
= − tan-1 (1.6.31)
R0  2 aR 0 
Here R 0= R ( M,N 0)=[ρ2 + ρ20 − 2ρρ0cos(φ−φ0) + z 2]1/2. The integration in (1.6.30)

for z 0=0 and ρ0< a yields π / R 0. 2
The case of z =0 can be considered in a
similar manner.
The integrals evaluated above may be called internal because the domain of
integration was the interior of a disc. We can also evaluate relevant external
integrals. For example, consider the integral
2π ∞
z 1 j
I3 = ⌠ ⌠
0
tan-1 r d r dψ, (1.6.32)
⌡ ⌡ R ( M,N) R ( N,M0)
3 R ( N,M 0)
0 a
where j 0 = ( r 2 − a 2)1/2( l 220 − a 2)1/2/ a . (1.6.33)
The integral representations due to (1.1.26) and (1.3.15), namely,
j0 r
[ l 220( x ) − x 2]1/2 d x l 210( x )
λ , φ0−ψ,
1
tan-1 = ⌠
R ( N,M 0) R ( N,M 0) ⌡ ( r − x ) [ l 20( x ) − l 10( x )]
2 2 1/2 2 2
 ρ 0
r 
a
(1.6.34)
∞
[ l 22( t ) − t 2]1/2 ρ
λ 2 , φ−ψ,
z 2 d ⌠ tdt
= − L( r )
3
R ( M,N ) π r d r ⌡ (t − r )
2 2 1/2
l 2( t ) − l 1( t ) l 2( t )
2 2

r
(1.6.35)
can be substituted in (1.6.32) yielding
∞ ∞
[ l 220( x ) − x 2]1/2 [ l 22( t ) − t 2]1/2 l 2 ( x )ρ
I = −4⌠ dx ⌠
dr
λ 10 , φ−φ0
3
⌡ l 220( x ) − l 210( x ) ⌡ (t − r )
2 2 1/2
l 2( t ) − l 1( t )  l 2( t )ρ
2 2 2

a x 0
∞
[ l 22( x ) − x 2]1/2 [ l 220( x ) − x 2]1/2 l ( x) l ( x)
= 2π⌠ λ 1 10 , φ−φ0  d x . (1.6.36)
⌡ l 22( x ) − l 21( x ) l 220( x ) − l 210( x )  l 2( x ) l 20( x ) 
a
This is the integral which was already evaluated in (1.6.27), so we may write
the final result
2π ∞
z 1 j
⌠⌠ tan-1
0
r d r dψ
⌡ ⌡ R ( M,N) R ( N,M0)
3 R ( N,M )
0
0 a
Ξ Ξ
| z |  1  -1 1
| zz0| | zz0| 
π  1  -1 2 π  ,
z R  zz 
= π tan − + tan +
1  R
1
2 zz 
0
R
2
R
2
2 0 
(1.6.37)
Comparison with the relevant internal integral (1.6.18) indicates similarity, except
for substitution of Θ by Ξ.
The second external integral is
2π ∞
z R ( N,M 0) j0
I4 = ⌠ ⌠
0
 + tan -1  rd rdψ ,
⌡ ⌡ R ( N,M0) 
3 j0 R ( N,M 0)  R ( M,N )
0 a
(1.6.38)
where j 0 is defined by (1.6.33). Make use of the integral representations (see
Exercise 1.1.8 and (1.3.21))
∞
[ x 2 − l 21( x )]1/2 l 1( x ) r
1
=
2 ⌠ dx
λ , φ − ψ,
R ( M,N) π ⌡ ( x 2 − r 2)1/2 l 22( x ) − l 21( x )  l 2( x ) x 
r
(1.6.39)
z0 R ( N,M 0) j0
 + tan -1 
R 3( N,M 0)  j0 R ( N,M 0) 
r
[ x 2 − l 210( x )]1/2 l ( x) x
=
1 1 d ⌠
L( )
xdx
λ 10
, ψ−φ0.
r r d r ⌡ ( r 2 − x 2)1/2 l 220( x ) − l 210( x )  l ( x ) 
20
a
(1.6.40)
Substitution of (1.6.39) and (1.6.40) in (1.6.38) leads to
∞
[ x 2 − l 21( x )]1/2 [ x 2 − l 210( x )]1/2 l ( x) l ( x)
I = 2π⌠ 2 λ 1 10 , φ−φ d x .
4
⌡ 2
l ( x ) − l 2
1 ( x ) l 2
20 ( x ) − l 2
10 ( x ) l
2
( x ) l
20
( x ) 0
a
(1.6.41)
This integral was evaluated in (1.6.14), and the final result is
2π ∞
z R ( N,M ) j
⌠⌠ 0
 0
+ tan -1  rd rdψ
0
⌡ ⌡ R ( N,M0) 
3 j
0
R ( N,M )  R ( M,N )
0
0 a
|z |  1 π Θ 1  π Θ 
= π
0
  − tan-1
1
 + − tan-1
2
. (1.6.42)
z R  2 R  R  2 R 
0  1 1 2 2 
One can notice the same similarity between the internal (1.6.30) and the external
(1.6.42) integrals. The similarity goes further. By using the property
( l 22( x ) − x 2)( x 2 − l 21( x )) = x 2 z 2,
we deduce, that for zz >0,

0
Ξ ( x ) = ξ( x ) + θ( x ), Ξ ( x ) = ξ( x ) − θ( x ),
1 2
Θ ( x ) = θ( x ) + ξ( x ), Θ ( x ) = θ( x ) − ξ( x ).
1 2
(1.6.43)
This means that Ξ =Θ and Ξ =−Θ for zz >0. In the case when zz <0, the
1 1 2 2 0 0
relationships change, namely, Ξ =−Θ and Ξ =Θ .
1 1 2 2
Exercise 1.6
Introduce the following points: M (ρ,φ, z ), M (ρ ,φ , z ), N ( r ,ψ,0), N (ρ ,φ ,0),

0 0 0 0 0 0 0
P (ρ,φ,0); as before, R (⋅,⋅) stands for the distance between two points.
2π a
z R ( N,M 0) h0
⌠⌠ 0
 + tan -1  rd rdψ , for ρ> a .
⌡ ⌡ R ( N,M0)
3 h0 R ( N,M 0)R ( P , N )
0 0
( l 220 − a 2)1/2(ρ2 − a 2)1/2

Answer:
2π  π − tan-1 .
R ( M 0, P )  2 aR ( M 0, P ) 
Hint : use (1.6.30) for z =0.
2. Evaluate the integral above for ρ< a .
Answer: π2/ R ( M 0, P ).
2π ∞
( r 2 − a 2)1/2(ρ20 − a 2)1/2
⌠⌠ tan-1   rd rdψ.
z 1
⌡ ⌡ R ( M,N)
3 R ( N,N 0
)  a R ( N,N0) 
0 a
(ρ20 − a 2)1/2( l 22 − a 2)1/2

2π -1 .
Answer: tan
R ( M,N0)  a R ( M,N0) 
2π ∞
( a 2 − ρ2)1/2
r d r dψ
⌠⌠ 0
⌡ ⌡ ( r − a ) R ( M,N) R 2( N0, N)
2 2 1/2
0 a
( a 2 − l 21)1/2( a 2 − ρ20)1/2
Answer:
2π  π − tan-1 .
R ( M,N 0)  2 a R ( M,N 0) 
CHAPTER 2
MIXED BOUNDARY VALUE PROBLEMS IN ELASTICITY
Elastic half-space has been proven to be a useful mathematical model for

consideration of various contact and crack problems in finite bodies, provided that
the domain of contact or the crack size is much smaller than the characteristic
dimension of the body. A general solution in terms of three harmonic functions
is presented for the case of transverse isotropy. Exact closed form solutions are
given to the mixed problems of the first and second type, with various
applications considered. The material in this Chapter is based on the papers
(Fabrikant, 1970, 1971b, 1971c, 1985b, 1986g).
2.1 General solution
Consider a transversely isotropic elastic body which is characterized by five

elastic constants A ik defining the following stress-strain relationships:
∂ux ∂uy ∂w
σx = A 11 + ( A 11 − 2 A 66) + A 13 ,
∂x ∂y ∂z
∂ux ∂uy ∂w
σy = ( A 11 − 2 A 66) + A 11 + A 13 ,
∂x ∂y ∂z
∂ux ∂uy ∂w
σz = A 13 + A 13 + A 33 ,
∂x ∂y ∂z
∂ux ∂uy ∂uy ∂w

τxy = A 66( + ) , τyz = A 44( + ) ,
∂y ∂x ∂z ∂y
73
74 CHAPTER 2. MIXED BOUNDARY VALUE PROBLEMS IN ELASTICITY
∂w ∂ux
τzx = A 44( + ). (2.1.1)
∂x ∂z
The equilibrium equations are:
∂σx ∂τxy ∂τzx ∂τxy ∂σy ∂τyz

+ + = 0 , + + = 0 ,
∂x ∂y ∂z ∂x ∂y ∂z
∂τzx ∂τyz ∂σz

+ + = 0 . (2.1.2)
∂x ∂y ∂z
Substitution of (2.1.1) in (2.1.2) yields:
∂2 u x ∂2 u x ∂2 u x ∂2 u y ∂2 w
A 11 + A 66 + A 44 + ( A 11 − A 66) + ( A 13 + A 44) = 0 ,
∂x2 ∂y2 ∂ z2 ∂x∂y ∂x∂z
∂2 u y ∂2 u y ∂2 u y ∂2 u x ∂2 w
A 66 + A 11 + A 44 + ( A 11 − A 66) + ( A 13 + A 44) = 0 ,
∂x2 ∂y2 ∂ z2 ∂x∂y ∂y∂z
∂2 w ∂2 w  ∂2 w ∂2 u x ∂2 u y
A 44    = 0 .
2 + + A 33 2 + ( A 44 + A 13) +
 ∂x ∂y 
2
∂z  ∂ x ∂ z ∂ y ∂ z 
(2.1.3)
Introduce complex tangential displacements u = u x+i u y, and u = u x−i u y. This will
allow us to reduce the number of equations in (2.1.3) by one, and to rewrite
these equations in a more compact manner, namely,
1 ∂2 u 1 ∂w
( A 11 + A 66)∆ u + A 44 + ( A 11 − A 66)Λ2 u + ( A 13 + A 44)Λ = 0,
2 ∂z 2 2 ∂z
∂2 w 1 ∂
A 44∆ w + A 33 2 + 2( A 13+ A 44) ∂ z (Λu + Λ u ) = 0. (2.1.4)
∂z
Here the following differential operators were used:
∂2 ∂2 ∂ ∂
∆ = + , Λ = + i , (2.1.5)
∂x2 ∂y2 ∂x ∂y
and the overbar everywhere indicates the complex conjugate value. Note also
2.1 General solution 75
that ∆=ΛΛ. One can verify that equations (2.1.4) can be satisfied by
∂F1 ∂F2
u = Λ( F 1+ F 2 + i F 3), w = m1 + m2 (2.1.6)
∂z ∂z
where all three functions F k satisfy the equation (Elliott, 1948):
∂2 F k
∆ F k + γ2k = 0, for k = 1,2,3, (2.1.7)
∂ z2
and the values of m k and γk are related by the following expressions (Elliott,
1948):
A 44 + m k( A 13+ A 44) m k A 33
= = γ2k, for k =1,2;
A 11 m k A 44 + A 13 + A 44
γ3 =  A 44/ A 661/2. (2.1.8)

 
Introducing the notation z k= z /γk, for k =1,2,3, we may call function F k= F ( x,y,zk )
harmonic. Note the property m 1 m 2=1, which seems to have escaped the attention
of previous researchers, and which will help us to simplify various expressions to
follow. The other elastic constants which will be used throughout the book are:
G 1 = β + γ 1γ 2 H , G 2 = β − γ 1γ 2 H ,
(γ1 + γ2) A 11 ( A 11 A 33)1/2 − A 13 γ3

H = , α = , β = .
2
2π( A 11 A 33 − A 13) A 11(γ1 + γ2) 2π A 44
(2.1.9)
Introduce the following inplane stress components:
σ1 = σx + σy , σ2 = σx − σy + 2iτxy , τz = τzx + iτyz.

(2.1.10)
This will simplify expressions (2.1.1), namely
∂w
σ1 = ( A 11 − A 66)(Λu + Λ u ) + 2 A 13 , σ2 = 2 A 66Λ u ,
∂z
τz = A 44 + Λ w .
1 ∂w ∂u
σz = A (Λu + Λ u ) + A 33 ,
2 13 ∂z  ∂z 
(2.1.11)
We have now only four components of stress, instead of six, as it was in
(2.1.1). The substitution of (2.1.6) in (2.1.11) yields:
∂2
σ1 = 2 A 66 2 {[γ1 − (1 + m 1)γ3] F 1 + [γ2 − (1 + m 2)γ3] F 2} ,
2 2 2 2
∂z
σ2 = 2 A 66Λ2( F 1 + F 2 + i F 3) ,
∂2
σz = A 44 2 [(1 + m 1)γ1 F 1 + (1 + m 2)γ2 F 2]
2 2
∂z
= − A 44 ∆[(1 + m 1) F 1 + (1 + m 2) F 2] ,
∂
τz = A 44Λ [(1 + m 1) F 1 + (1 + m 2) F 2 + i F 3] . (2.1.12)
∂z
Here we used the fact that each F k satisfies equation (2.1.7), and the relation:
A 11γ2k− A 13 m k= A 44(1+ m k), (for k =1,2) which is an immediate consequence of (2.1.8).
Expressions (2.1.6) and (2.1.12) give a general solution, expressed in terms of
three harmonic functions F k. It is very attractive to express each function F k
through just one harmonic function as follows:
F k( x,y,z) = c k F ( x,y,zk ) , (2.1.13)
where z k= z /γk, and c k is an as yet unknown complex constant. As we shall see

further, this is possible indeed. All the results obtained in the book are valid
for isotropic solids, provided that we take
1 − ν2 1 − 2ν
γ1 = γ2 = γ3 = 1, H = , α = ,
πE 2(1 − ν)
1 + ν (2 − ν)(1 + ν) ν(1 + ν)
β = , G1 = , G2 = ,
πE πE πE
(2.1.14)
where E is the elastic modulus, and ν is Poisson coefficient.
Exercise 2.1
1. Establish the equivalence of (2.1.3) and (2.1.4)
2. Prove that the solution (2.1.6) satisfies (2.1.3), provided that the condition
(2.1.7) is met.
3. Prove the identity m 1 m 2=1.
4. Prove the identity A 11γ2k− A 13 m k= A 44(1+ m k).
2.2 Point force solutions
The field of stresses and displacements due to a concentrated load is

important for the integral equation formulation of various mixed boundary value
problems. Two cases are considered here: an arbitrary point load in a
transversely isotropic elastic space, and the action of an arbitrary concentrated
force on the boundary of a similar half-space. Though these problems have been
solved by many authors, we follow here the results given in (Fabrikant, 1970).
The main reason for this is the simplification of the elastic coefficients, which
seems to have escaped the attention of other authors. Here is an example: one
of the coefficients in (Chen, 1966) reads
A 13 + A 44 A m1
 33 − A 13.
A 11 A 44(γ1 − γ2)  γ21
2 2

This expression, after simplification, reduces to 1/(1− m 2).

Let a point force, with components T x, T y, and P in Cartesian coordinates
be applied at the point N 0 inside a transversely isotropic elastic space. We may
assume, without loss of generality, that the polar cylindrical coordinates of N 0 are
(ρ0,φ0,0). We need to find the field of stresses and displacements at the point
M (ρ,φ, z ). Introduce the complex tangential force T = T x+i T y. The general solution
can be expressed through the three potential functions:
F1 =
1 1 γ m (Λχ + Λχ ) + P ln( R + z ),
4π A 44( m 1 − m 2) 2 1 2 1 1 1 1

F2 = −
1 1 γ m (Λχ + Λχ ) + P ln( R + z ),
4π A 44( m 1 − m 2) 2 2 1 2 2 2 2

γ3
F3 = i (Λχ3 − Λχ3).
8π A 44
(2.2.1)
χk( z ) = χ( z k), R k = [ρ2 + ρ20 − 2ρρ0cos(φ−φ0) + z 2k]1/2, for k =1,2,3;
χ( z ) = T [ z ln( R 0 + z ) − R 0]. (2.2.2)
The displacements are defined by (2.1.6) as follows:
1
γ1 m 2− 
1 T q2T
u =  
+
4π A 44( m 1 − m 2) 2 R R 1( R 1 + z 1) 
2
 1
 − P  1 − 2 
z z
γ2 m 1−
1 T q2T
− +
2  R2 R 2( R 2 + z 2)2 q R 1 R 2

γ3
+  T + q2T , (2.2.3)
8π A 44 R 3 R 3( R 3 + z 3) 
2
1  1T T 
z1 z2
 
m1 m2 
.
w = 
4π A 44( m 1 − m 2) 2q
+ − + + P −
q   R1 R 2 γ1 R 1 γ 2 R 2
 
(2.2.4)
Here
φ iφ
q = ρei − ρ0e 0. (2.2.5)
The stress field is defined by (2.1.12). We shall need only the expressions for
σz and τz. Here they are:
1 1
γ1 γ2
 
4π2
σz = − (T q + Tq) +
 ( m 1 − 1) R 31 ( m 2 − 1) R 32
,
m1 z1 m2 z2
+ P  +
( m 1 − 1) R 31 ( m 2 − 1) R 32

(2.2.6)
z1 z2 z3
T 
τz = + − 3
8π ( m 1 − 1) R 1
3
( m 2 − 1) R 2
3
R 3
2R 1 + z1 2R 2 + z2 2R 3 + z3
Tq 2  
− + +
8π ( m 1 − 1) R 31( R 1 + z 1)2 ( m 2 − 1) R 32( R 2 + z 2)2 R 33( R 3 + z 3)2
m1 m2
Pq  
− + 3 . (2.2.7)
4π γ1( m 1 − 1) R 31 γ2( m 2 − 1) R 2
Consider a transversely isotropic elastic half-space z ≥0. Let a concentrated
force, with components T x, T y, and P , be applied at the point N 0(ρ0,φ0,0). We
need to find the field of stresses and displacements in the half-space. The
potential functions are defined by
H γ1
F1 = 1 γ (Λχ + Λχ ) + P ln( R + z ),
m 1 − 1 2 2 1 1 1 1

H γ2
F2 = 1 γ (Λχ + Λχ ) + P ln( R + z ),
m 2 − 1 2 1 2 2 2 2

γ3
F3 = i (Λχ3 − Λχ3). (2.2.8)
4π A 44
γ3
u = T + q2T 
4π A 44 R 3 R 3( R 3 + z 3)2
H γ21  T q2T  + Pq 
+  γ1 −
m2 − 1 2  R2
+
R 2( R 2 + z 2)
R 2( R 2 + z 2)2
 
H γ11  T q2T  + Pq 
+  γ2 −
m1 − 1 2  R1
+
R 1( R 1 + z 1)
,
R 1( R 1 + z 1)2
 
(2.2.9)
1 γ2 m 1 γ1 m 2
w = H (T q + Tq)  + 
2
 ( m 1 − 1) R 1( R 1 + z1) ( m 2 − 1) R 2( R 2 + z 2)
.
m1 m2
+ P  + (2.2.10)
( m 1 − 1) R 1 ( m 2 − 1) R 2

We shall need expressions for the following stress components:
1 1 
γ1γ2( T q + T q ) + Pz − 3 + 3,
1 1
σz = 
2π(γ1 − γ2) 2  R 1 R 2
 
γ2 Tz T q 2(2 R + z ) γ Tz T q 2(2 R + z )
τz =  1 − 1 1
 − 1
 2 − 2 2

4π(γ1 − γ2)  R 1
3
R 1( R 1 + z 1) 
3 2 4π( γ 1 − γ 2  R2
) 3
R 2( R 2 + z 2) 
3 2
1 Tz 3 T q 2(2 R 3 + z 3)
− + 3  + Pq − 1 + 1 . (2.2.11)
4π  R 33 R 3( R 3 + z 3) 2
 2π( γ 1 − γ 2 
) R 31 R 32
Expressions (2.2.9) and (2.2.10) simplify for the case when z =0
1 T 1 Tq2 P
u =
2
G1
R
+ G2
2 3 − Hα , (2.2.12)
R q
w = H α ℜ  + H .
T P
(2.2.13)
q  R
Here H , α, G 1, and G 2 are defined by (2.1.9), and
R = [ρ2 + ρ20 − 2ρρ0cos(φ−φ0)]1/2. (2.2.14)
Expressions (2.2.12) and (2.2.13) will be used for the integral equation
formulation of various mixed boundary value problems in an elastic half-space.
The following classification of mixed boundary value problems may be
suggested. The problem is called internal mixed when the normal/tangential
displacements are prescribed inside a finite domain, while the relevant tractions
are prescribed on the rest of the half-space boundary. In the case when the
displacements are given outside a finite domain, the problem is called external.
We can specify two types of internal problems. The internal problem of type I :
the normal displacements are prescribed inside a finite domain S , the normal
traction is given outside the domain S , while the tangential tractions are known
all over the plane z =0. The internal problem of type II : the tangential
displacements are prescribed inside S , and the shear tractions are given outside,
while the normal traction is known all over the plane z =0. The external
problems of types I and II are defined in the same way as the internal ones
above, with an interchange of the terms traction and displacement. These are
the four types of problems which will be considered in this chapter. We shall
call a problem mixed-mixed when the boundary conditions are mixed with respect
to both normal and tangential components. These problems will be considered in
the next chapter.
Exercise 2.2
1. Establish (2.2.3) and (2.2.4).
2. Verify (2.2.6) and (2.2.7).
3. Derive expressions for σ1 and σ2 in both cases of concentrated load,

considered in section 2.2.
4. Derive the equivalent solutions for an isotropic body.
5. Consider the case of arbitrary point loading applied inside an elastic

half-space.
2.3 Internal mixed problem of type I.
Consider a transversely isotropic elastic half-space z ≥0. Introduce a set of

polar cylindrical coordinates (ρ,φ, z ). Let the following boundary conditions be
prescribed on the plane z =0:
w = w (ρ,φ), ρ≤ a , 0≤φ<2π,
σ = σ(ρ,φ), ρ> a , 0≤φ<2π,
τ = τ(ρ,φ), 0≤ρ<∞, 0≤φ<2π,

(2.3.1)
Here σ stands for the normal loading, and τ is the complex shear loading,
namely, τ=τzx+iτyz. The governing integral equation can be written by using
(2.2.13), namely,
2π a
σ(ρ ,φ ) ρ dρ dφ
H⌠ ⌠
0 0 0 0 0
= f (ρ,φ). (2.3.2)
⌡ ⌡ [ρ2 + ρ20 − 2ρρ0cos(φ−φ0)]1/2
0 0
We use the same notation σ for the unknown normal loading inside the circle
ρ≤ a , as well as for the prescribed function σ outside the circle. This should
not create any confusion since the argument (ρ0,φ0) provides a clear distinction.
Function f is known from the conditions (2.3.1), and is
2π ∞
σ(ρ0,φ0) ρ0dρ0dφ0
f (ρ,φ) = w (ρ,φ) − ⌠ ⌠
⌡ ⌡ [ρ2 + ρ20 − 2ρρ0cos(φ−φ0)]1/2
0 a
2π ∞
τ(ρ0,φ0) ρ0dρ0dφ0
− H αℜ⌠ ⌠ . (2.3.3)
⌡⌡ ρe iφ
− ρ0e
i φ0
0 0
As soon as equation (2.3.2) is solved, and the value of σ inside the circle
becomes known, the tangential displacements in the plane z =0 can be defined by
(2.2.12):
2π ∞
1 ⌠⌠
u = G1
2 ⌡ ⌡ [ρ2 + ρ20 − 2ρρ0cos(φ−φ0)]1/2
0 0
i φ0
[ρei φ − ρ0e
2π ∞ 2π ∞
]2τ(ρ0,φ0) ρ0dρ0dφ0 σ(ρ0,φ0) ρ0dρ0dφ0
1 ⌠⌠
+ G − H α⌠ ⌠ .
2 2⌡ ⌡ [ρ2 + ρ20 − 2ρρ0cos(φ−φ0)]3/2 ⌡⌡ -i φ -i φ0
0 0 0 0
ρe − ρ0e
(2.3.4)
Integral equation (2.3.2) was solved in section 1.4. It seems useful to consider
here a more general case:
2π a
H⌠ ⌠ = f (ρ,φ), (2.3.5)
⌡ ⌡ [ρ2 + ρ20 − 2ρρ0cos(φ−φ0)](1+κ)/2
0 0
where −1<κ<1. This type of equation arises in the problems of nonhomogeneous

elastic half-space, with the modulus of elasticity E being a power function of z ,
namely, E = E 0 z κ. Of course, in the nonhomogeneous case H is no longer defined
by (2.1.9). The reader is referred to the paper by Rostovtsev (1964) for details.
Rostovtsev (1964) obtained an exact solution of (2.3.5) in Fourier series. Here
we present a closed form solution.
By using the integral representation (1.1.4), integral equation (2.3.5) can be
rewritten as
ρ κ a
ρ dρ
πκ ⌠ x dx
⌠
0 0
x  2
4 H cos 2 (1+κ)/2 L ρρ σ(ρ0,φ) = f (ρ,φ). (2.3.6)
2 ⌡ (ρ2 − x 2)(1+κ)/2 ⌡ (ρ0 − x )
2
 0
0 x
Integral equation (2.3.6) represents a sequence of two Abel operators and one
L-operator. The solution procedure is similar to that of (1.4.5). The first
operator to be applied to both sides of (2.3.6) is
L 
1 d ⌠ ρdρ
2 (1-κ)/2 L(ρ). (2.3.7)
 
t d t ⌡ (t − ρ )
2
0
a
ρ0dρ0 t
L  σ(ρ0,φ) = L 
κ⌠ t 1 d ⌠ ρdρ
2π Ht L(ρ) f (ρ,φ). (2.3.8)
⌡ (ρ0 − t )
2 2 (1+κ)/2
ρ0  t  d t ⌡ ( t − ρ2)(1-κ)/2
2
t 0
L ,
d⌠ t 1-κd t 1
L( y )
d y⌡ ( t − y )
2 2 (1- κ)/2
t 
y
with the result
a
cos(πκ/2) d⌠ t 1-κd t
σ( y ,φ) = − L( y )
π Hy
2 d y⌡ (t2 − y 2)(1-κ)/2
y
×L 2
1 d ⌠ ρdρ
2 (1-κ)/2 L(ρ) f (ρ,φ).
t  d t ⌡ (t − ρ )
2
0
(2.3.9)
allow us to rewrite (2.3.9) in the form
a
cos(πκ/2)  Φ( a , y ,φ) ⌠ 2 d t2 (1-κ)/2 d Φ( t , y ,φ).
σ( y ,φ) = 2 (1-κ)/2 − (2.3.10)
πH2
(a − y )
2
⌡ (t − y ) dt 
y
Here
t
⌠ 2 ρ d2ρ(1-κ)/2 d ρ1-κLρ2y f (ρ,φ).

1+κ
1
Φ( t , y ,φ) = (2.3.11)
t 1+κ ⌡ (t − ρ ) dρ t  
0
Yet another form of solution can be found in (Fabrikant, 1971e). The problem
solved above has two major applications: contact problems of a smooth punch
pressed against an elastic half-space, and that of an external circular crack in an
infinite elastic body. Let us consider both cases in more detail.
Example 1. The smooth punch problem. In elastic contact problems, we
have σ=0, for ρ> a , and τ=0 all over the plane z =0, so that function f = w (see
2.3.9). It becomes possible to compute the resultant force P and the tilting
moments M x and M y directly in terms of the prescribed displacement w . Since
2π a
P = ⌠ ⌠ σ(ρ,φ) ρdρdφ, (2.3.12)

⌡⌡
0 0
substitution of (2.3.9) in (2.3.12) yields directly the resultant force
2π a
cos(πκ/2) ⌠ w (ρ,φ) ρdρdφ
P = ⌠ . (2.3.13)
π2 H ⌡ ⌡ ( a 2 − ρ2)(1-κ)/2
0 0
For computation of the tilting moments M x and M y, it is convenient to introduce

the complex parameter
2π a
M = M x + i M y = −i ⌠ ⌠ σ(ρ,φ) ei φ ρ2dρdφ. (2.3.14)

⌡⌡
0 0
By using (2.3.9), the final expression for the tilting moment is found to be
φ 2π a
2cos(πκ/2)
⌠ w (ρ,φ) ei ρ2dρdφ
M = −i 2 ⌠ . (2.3.15)
π H (1 + κ) ⌡ ⌡ ( a 2 − ρ2)(1-κ)/2
0 0
Expressions (2.3.14) and (2.3.15) are in agreement with similar results of

Rostovtsev (1964).
By reviewing the derivation of expression (2.3.6), one may find that it is
valid for evaluating the normal displacements outside the contact region, if the
upper limit of integration ρ is replaced by a . Substitution of (2.3.9) into the
modified form of (2.3.6) results in
a x
ρ0dρ0 ρ
w (ρ,φ) =
2cos(πκ/2) ⌠ dx d ⌠
L 0w(ρ ,φ),
π ⌡ (ρ − x )
2 2 (1+κ)/2 d x
⌡ ( x − ρ0)
2 2 (1-κ)/2
ρ 0
0 0
for ρ> a . (2.3.16)
Performing differentiation of the integrand, and then integrating by parts, we
obtain
2π
a
w (ρ0,φ0) ρ0dρ0dφ0
1 πκ 2 2 (1-κ)/2⌠ ⌠
w (ρ,φ) = 2cos( )(ρ − a ) ,
π 2 ⌡ ⌡ ( a 2−ρ20)(1-κ)/2[ρ2 + ρ20 − 2ρρ0cos(φ−φ0)]
0 0
for ρ> a . (2.3.17)
Here the following identities were employed (Bateman and Erdélyi, 1955)
d  (1+κ)/2  1+κ 1+κ 3+κ  1+κ -(1-κ)/2

ζ F , ; ; ζ = ζ (1 − ζ)-(1+κ)/2,
dζ  2 2 2  2
x x
d ⌠
= f (0) x κ + x ⌠
f(t)t dt df(t)
d x ⌡ ( x 2 − t 2)(1-κ)/2 2 (1-κ)/2. (2.3.18)
⌡ (x − t )
2
0 0
All the quantities of interest, namely, the pressure exerted by the punch σ, the
resultant force P , the tilting moment M , and the normal displacement outside the
punch, can be expressed directly through the prescribed normal displacement w by

formulae (2.3.9), (2.3.13), (2.3.15), and (2.3.17) respectively.
Example 2. External crack in non-homogeneous elasticity. Consider a
κ
non-homogeneous elastic space with modulus of elasticity E = E 0| z | , E 0=const,
|κ|<1. This space is weakened by a circular external crack ρ≥ a . An arbitrary
pressure σ(ρ,φ) is applied to both faces of the crack in opposite directions. The
problem is to find the normal stress in the crack neck, the normal displacements
of the crack faces, the stress intensity factor, and the work required to open up
the crack.
Due to the symmetry of the problem, it may be reduced to the mixed
boundary value problem of a half-space, subject to the boundary conditions at
z =0:
w = 0, τ = 0, for 0≤ρ≤ a , 0≤φ<2π;
σ = σ(ρ,φ), τ = 0, for a <ρ<∞, 0≤φ<2π. (2.3.19)
The governing integral equation takes the form (2.3.5), with the known function
2π ∞
f (ρ,φ) = − H ⌠ ⌠ . (2.3.20)
⌡ ⌡ [ρ2 + ρ20 − 2ρρ0cos(φ−φ0)](1+κ)/2
0 a
Its solution can be found in exactly the same manner as that of (1.5.21), and is
2π ∞
cos(πκ/2) (ρ20 − a 2)(1-κ)/2σ(ρ0,φ0) ρ0dρ0dφ0
σ(ρ,φ) = − ⌠⌠ . (2.3.21)
π (a
2 2
− ρ)
2 (1-κ)/2
⌡⌡ ρ2 + ρ20 − 2ρρ0cos(φ−φ0)
0 a
Expression (2.3.21) gives the normal stress in the crack neck through the pressure
applied to the crack faces. Note that (1.5.24) may be considered as a particular
case of (2.3.21), when κ=0.
The normal displacement of the crack faces can be evaluated as a
superposition of the displacement caused by the applied pressure, and the
displacement due to the normal stress in the crack neck. By using a procedure
analogous to the one described in section 1.5, we can obtain the expression
∞
 x κd x ρ0dρ0 ρρ
x
πκ ⌠
w (ρ,φ) = 4 H cos  ⌠ L  0σ(ρ ,φ)
2
 ρ⌡ ( x − ρ )
2 2 (1+κ)/2
⌡ ( x − ρ0)
2 2 (1+κ)/2
 x2  0
a
a κ ρ0dρ0 a

⌠
x dx
⌠  x2 
+ 2 (1+κ)/2 L ρρ σ(ρ0,φ), for ρ> a . (2.3.22)
⌡ (ρ − x )
2 2 (1+κ)/2
⌡ (ρ0 − x )
2
 0 
0 x
Substitution of (2.3.21) in (2.3.22) leads, after simplification, to
ρ ∞
x κd x ρ0dρ0
L σ(ρ0,φ),
πκ ⌠ ⌠ x2
w (ρ,φ) = 4 H cos
2 ⌡ (ρ2 − x 2)(1+κ)/2 ⌡ (ρ20 − x 2)(1+κ)/2 ρρ0
a x
for ρ> a . (2.3.23)
The normal displacements of the crack faces are now defined in terms of the
applied pressure.
Introduce the stress intensity factor as
k 1(φ) = lim[( a − ρ)(1-κ)/2σ(ρ,φ)]. (2.3.24)

ρ→ a
Substitution of (2.3.21) in (2.3.24) gives
∞
2cos(πκ/2) ρ dρ
k 1(φ) = ⌠ 2
0 0
a 
2 (1+κ)/2 L ρ σ(ρ0,φ).
π(2 a )(1-κ)/2 ⌡ (ρ0 − a )  0
a
Introduce the stress intensity function :
∞
2cos(πκ/2) ρ dρ
K 1(ρ,φ) = ⌠ 0 0
ρ
2 (1+κ)/2 L ρ σ(ρ0,φ). (2.3.25)
π(2ρ)(1-κ)/2 ⌡ (ρ0 − ρ )
2
 0
ρ
It is obvious that the limiting case of the stress intensity function, when ρ→ a , is
the stress intensity factor. By using the property of the L-operators (1.2.3) we
may rewrite (2.3.23) as
ρ κ ∞
ρ0dρ0
L  ⌠ L σ(ρ0,φ)
πκ ⌠ x dx x x
w (ρ,φ) = 4 H cos
2 ⌡ (ρ − x )
2 2 (1+ κ)/2
ρ ⌡ (ρ20 − x)
2 (1+κ)/2
ρ0
a x
ρ
πH ⌠
x (1+κ)/2d x x 
2 (1+κ)/2 L ρ K 1( x ,φ).
(3-κ)/2
= 2 (2.3.26)
⌡ (ρ − x ) 2
 
a
The energy W may be defined by the integral
2π ∞
W = ⌠⌠ σ(ρ,φ) w (ρ,φ)ρ dρdφ. (2.3.27)

⌡⌡
0 a
2π ∞ ρ
W = 2(3-κ)/2π H ⌠ dφ ⌠ σ(ρ,φ)ρ dρ ⌠
x (1+κ)/2 d x x 
2 (1+κ)/2 L ρ K 1( x ,φ)
⌡ ⌡ ⌡ (ρ − x )
2
 
0 a a
2π ∞ ∞
π H ⌠ dφ ⌠ x (1+κ)/2d x ⌠ 2
σ(ρ,φ)ρ dρ x 
2 (1+κ)/2 L ρ K 1( x ,φ)
(3-κ)/2
= 2
⌡ ⌡ ⌡ (ρ − x )  
0 a x
2π ∞ ∞
π H ⌠ dφ ⌠ K 1( x ,φ) x (1+κ)/2d x ⌠ 2
ρ dρ x 
2 (1+κ)/2 L ρ σ(ρ,φ).
(3-κ)/2
= 2
⌡ ⌡ ⌡ (ρ − x )  
0 a x
Here the interchange of the order of integration was used twice. Now
comparison of the last expression with (2.3.25) yields the final result
2π ∞
2 πH ⌠⌠
1-κ 2
W = [ K 1(ρ,φ)]2ρ dρ dφ. (2.3.28)
cos(πκ/2) ⌡ ⌡
0 a
Expression (2.3.28) interprets the stress intensity function squared as being

proportional to the energy per unit area required to open up the crack. In the
case of axial symmetry, formula (2.3.28) simplifies to
∞
22-κπ3 H ⌠
W = [ K (ρ)]2ρ dρ,
cos(πκ/2) ⌡ 1
a
with
∞
2cos(πκ/2) σ(ρ ,φ )ρ dρ
K 1(ρ) = ⌠ 2 0 0 20 (1+κ)/20 .
π(2ρ)(1-κ)/2 ⌡ (ρ0 − ρ )
ρ
All the results obtained become valid for a transversely isotropic space provided
that κ=0, and H is defined by (2.1.9). When H is defined by (2.1.14), we
have the results for isotropic body, namely,
2π ∞
1 − ν2 ⌠ ⌠
W = 2π [ K 1(ρ,φ)]2ρ dρ dφ, (2.3.29)
E ⌡⌡
0 a
with
∞
K 1(ρ,φ) =
√2 ⌠ x dx ρ σ( x ,φ).
2 1/2 L x (2.3.30)
π√ρ ⌡ (x − ρ )
2
 
ρ
The case of axial symmetry simplifies (2.3.29) and (2.3.30) as follows:
∞
1 − ν2 ⌠
W = 4π 2
[ K 1(ρ)]2ρ dρ, (2.3.31)
E ⌡
a
with
∞
√2 ⌠ σ( x ) x d x
K 1(ρ) = . (2.3.32)
π√ρ ⌡ ( x 2 − ρ2)1/2
ρ
Exercise 2.3.
1. The normal displacements under a flat circular punch are given by
w (ρ,φ)= w 0+θρcosφ, with w 0=const, and θ=const. Find the traction distribution σ
exerted by the punch.
cos(πκ/2) w 0 + (2θρcosφ)/(1 + κ)
Answer: σ(ρ,φ) = .
π2 H ( a 2 − ρ2)(1-κ)/2
2. In the problem above find the relationships between the applied force P , the
tilting moment M , and the punch settlement w 0 and the inclination angle θ.
2 w 0 a 1+κcos(πκ/2)
4θ a 3+κcos(πκ/2)
Answer: P = , M = .
π H (1 + κ) π H (1 + κ)(3 + κ)
3. The normal displacements under a paraboloidal punch are w (ρ,φ)= w 0− c ρ2,

with w 0=const, and c =const. Find the traction distribution σ and the radius of
contact a .
Solution : utilization of (2.3.9) yields
cos(πκ/2)  2 c [ a 2(1 − κ) − 2ρ2]
σ(ρ) =  0
w + .
π2 H ( a 2 − ρ2)(1-κ)/2 (1 + κ)2
 
The radius of contact a is found from the condition σ( a )=0. The result is
a =[(1 + κ) w 0/(2 c )]1/2, and the final expression for the traction is
2 w 0cos(πκ/2)
σ(ρ) = ( a 2 − ρ2)(1+κ)/2.
π a H (1 + κ)
2 2
4. In the problem above find the relationship between the punch settlement w 0
and the applied force P .
4 a 1+κcos(πκ/2)
Answer: P = .
π H (1 + κ)(3 + κ)
5. Consider an external circular crack ρ> a . Find the traction distribution σ in

the crack neck due to the action of a pair of equal concentrated forces P ,
applied normally to the crack faces in opposite directions at the point ( b ,ψ).
Answer: σ(ρ,φ) = − 2 cos   2
P πκ b 2 − a 2(1-κ)/2 1
.
π   a − ρ 
2 2
ρ + b − 2 b ρcos(φ−ψ)
2 2
6. Consider an external circular crack ρ> a . Let a uniform pressure σ0 be

applied in opposite directions to the annulus b ≤ρ≤ c , ( b > a ), the rest of the crack
faces being traction free. Find the stress distribution σ in the crack neck and
the stress intensity factor k 1.
2σ0cos(πκ/2) c
(ρ20 − a 2)(1-κ)/2ρ0 dρ0
Answer: σ(ρ) = − ⌠ , for ρ< a .
π( a 2 − ρ2)(1-κ)/2 ⌡ ρ20 − ρ2
b
In general, the last integral can be computed in terms of hypergeometric
functions. In the particular case of an isotropic body (κ=0), the integral is
computable in elementary functions:
σ0
2 ( c 2 − a 2)1/2 − ( b 2 − a 2)1/2
σ(ρ) = −
π  ( a 2 − ρ2)1/2
− tan-1 2
c 2 − a 2 1/2 -1b
2
− a 21/2
+ tan .
a − ρ2 a 2 − ρ2 
The stress intensity factor is

2σ0cos(πκ/2)[( c 2 − a 2)(1-κ)/2 − ( b 2 − a 2)(1-κ)/2]
k1 = − .
π(2 a )(1-κ)/2(1 − κ)
7. In the preceding problem find the crack opening displacement w .

min(ρ,c)
4cos(πκ/2)  ⌠ ( c 2 − x 2)(1-κ)/2 κ
Answer: w (ρ) = σ0  x dx
1 − κ ⌡ (ρ2 − x 2)(1+κ)/2
 a
min(ρ,b)
⌠ ( b 2 − x 2)(1-κ)/2 κ 
−
⌡
x d x , for ρ> a .
(ρ2 − x 2)(1+κ)/2
a 
8. By using formula (2.3.9), prove the identity

2π a
⌠ ⌠ σ(ρ,φ)ρ1+|n| ein φ dρ dφ
⌡⌡
0 0
2π a
(1 + κ)Γ(1 + | n |) πκ ⌠ ⌠ f (ρ,φ)ρ1+|n| ein φ dρ dφ
= cos .
2π2Γ[| n | + (1 + κ)/2] 2 ⌡⌡ ( a 2 − ρ2)(1-κ)/2
0 0
Note : in the particular cases n =0 and n =−1, the last identity transforms into
(2.3.13) and (2.3.15) respectively.
9. Define the stress intensity function K 1(ρ,φ) in terms of the displacement w .

ρ
Answer: K 1(ρ,φ) = (1-κ)/2 2 (1+κ)/2 L 
cos(πκ/2) 1 d ⌠ x dx
2 (1-κ)/2 L( x ) w ( x ,φ).
2 π Hρ ρ  dρ ⌡ (ρ − x )
2
a
Hint : perform the inversion of (2.3.26).

 d ρ( x 2 − a 2)(1-κ)/2  π(1 − κ) f ( a )
lim  ⌠ 2 f ( x ) d x  = .
d ρ ⌡ (ρ − x )
2 (1-κ)/2 2cos(πκ/2)
 
ρ→ a
a
Hint : use the substitution t =(ρ2− x 2)/( x 2− a 2).
11. Use the identity above to express the stress intensity factor in terms of the
displacement w .
lim
1 − κ w (ρ,φ) 
Answer: k 1(φ) = 2-κ .
2 πH ρ → a (ρ − a )(1-κ)/2
Hint : compute the limit ρ→ a of the result in Exercise 9.
2.4 External mixed problem of type I.

The problem is characterized by the following mixed boundary conditions on the
plane z =0:
w = w (ρ,φ), ρ> a , 0≤φ<2π,
σ = σ(ρ,φ), ρ≤ a , 0≤φ<2π,
τ = τ(ρ,φ), 0≤ρ<∞, 0≤φ<2π,

(2.4.1)
Here the same notation is used as in the previous section. The governing
integral equation can be written by using (2.2.13), namely,
2π ∞
H⌠ ⌠ = f (ρ,φ). (2.4.2)
⌡ ⌡ [ρ2 + ρ20 − 2ρρ0cos(φ−φ0)]1/2
0 a
We use again the same notation σ for the unknown normal loading inside the
circle ρ≤ a , as well as for the prescribed function σ outside the circle. Function
f is known from the second condition (2.4.1), and is
2π a
f (ρ,φ) = w (ρ,φ) − ⌠ ⌠
⌡ ⌡ [ρ2 + ρ20 − 2ρρ0cos(φ−φ0)]1/2
0 0
2π ∞
− H αℜ⌠ ⌠ . (2.4.3)
⌡⌡ ρe
iφ
− ρ0e
i φ0
0 0
As soon as equation (2.4.2) is solved, and the value of σ inside the circle
becomes known, the tangential displacements in the plane z =0 can be defined by
(2.3.4). Integral equation (2.4.2) was solved in section 1.5. We consider again
a more general case:
2π ∞
H⌠ ⌠ = f (ρ,φ), (2.4.4)
⌡ ⌡ [ρ2 + ρ20 − 2ρρ0cos(φ−φ0)](1+κ)/2
0 a
where −1<κ<1. The new method allows us to present a closed form solution.
By using the integral representation (1.1.21), for z =0, the integral equation
(2.4.4) can be rewritten as
∞
x κd x ρ dρ ρρ
x
πκ ⌠ ⌠
0 0
 0σ(ρ ,φ) = f (ρ,φ).
4 H cos 2 (1+κ)/2 L (2.4.5)
2 ⌡ ( x 2 − ρ2)(1+κ)/2 ⌡ ( x − ρ0)
2
 x2  0
ρ a
Integral equation (2.4.5) represents a sequence of two Abel operators and one
L-operator. The solution procedure is similar to that of (1.5.2). The first
operator to be applied to both sides of (2.4.5) is
L .
d ⌠ ρdρ 1
L( t ) (2.4.6)
d t ⌡ (ρ − t )
2 2 (1- κ)/2
ρ
t
t
ρ0dρ0 ρ0
−2π Ht κ ⌠ L  σ(ρ0,φ)
⌡ ( t − ρ0)
2 2 (1+κ)/2
t
a
L  f (ρ,φ).
d ⌠ ρdρ 1
= L( t ) (2.4.7)
d t ⌡ (ρ2 − t 2)(1-κ)/2 ρ
t

L 
1 d⌠ t 1-κd t
L( t ) ,
y  d y⌡ ( y 2 − t 2)(1-κ)/2
a
with the result
y
cos(πκ/2) 1 d ⌠ t 1-κd t
σ( y ,φ) = − L
π2 Hy y  d y ⌡ ( y 2 − t 2)(1-κ)/2
a
L  f (ρ,φ).
d ⌠ ρdρ 1
×L( t 2) (2.4.8)
d t ⌡ (ρ − t )
2 2 (1- κ)/2
ρ
t

allow us to rewrite (2.4.8) in the form
y
cos(πκ/2)  Φ( a , y ,φ) ⌠ 2 d t2 (1-κ)/2 d Φ( t , y ,φ).
σ( y ,φ) = − 2 (1-κ)/2 − (2.4.9)
πH2
( y − a )
2
⌡ (y − t ) dt 
a
Here
∞
⌠ 2 dρ2 (1-κ)/2 d L t  f (ρ,φ).

2
Φ( t , y ,φ) = t 1-κ
(2.4.10)
⌡ (ρ − t ) dρ ρ y  
t
We can consider again two major applications: contact problems of a smooth

punch pressed against an elastic half-space, and that of a penny-shaped crack in
an infinite elastic body. Let us consider both cases in more detail.
Example 1. The smooth punch problem. In elastic contact problems we
have σ=0, for ρ< a , and τ=0 all over the plane z =0, so that function f = w . It
becomes possible to compute the resulting force P and the tilting moments M x
and M y directly in terms of the prescribed displacement w . Since
2π ∞
P = ⌠⌠ σ(ρ,φ) ρdρdφ, (2.4.11)

⌡⌡
0 a
substitution of (2.4.8) in (2.4.11) yields directly the resultant force

∞ 2π
 cos(πκ/2) y 
⌠ 2 dρ2 (1-κ)/2 d ⌠ w(ρ,φ) dφ.
2-κ
P = lim− ⌠ t d t
πH2
⌡ (y − t )
2 2 (1-κ)/2
⌡ (ρ − t ) dρ ⌡
y→∞
 a t 0 
(2.4.12)
The tilting moment can be found in a similar manner. We can also express the
normal displacement inside the circle ρ≤ a directly in terms of the prescribed
displacement w outside the circle. We substitute (2.4.8) in (2.4.5) keeping in
mind that, for ρ≤ a , the lower limit of integration of the first integral will be a
instead of ρ. By using the properties of Abel operators and the L-operators, the
following expression can be obtained
∞ ∞
ρ0dρ0
L  w (ρ0,φ).
2 πκ ⌠ dx d ⌠ ρ
w (ρ,φ) = − cos
π 2 ⌡ ( x 2 − ρ2)(1+κ)/2 d x ⌡ (ρ20 − x)
2 (1-κ)/2
ρ0
a x
(2.4.13)
Carrying out the differentiation of the integrand, interchanging the order of
integration, and then integrating with respect to x yields
∞
ρ20 − a 2
w (ρ,φ) = −
2cos(πκ/2) ⌠  (1+κ)/2
π(1 + κ) ⌡ ρ20 − ρ2
a
ρ2 − a 2
×F 1 + κ 1 + κ 3 + κ 0
, ; ; 2  d L ρ  w(ρ ,φ) dρ .
 2 2 2 ρ0 − ρ  dρ0 ρ0
2 0
 0
Integration by parts and the differential properties of the Gauss hypergeometric

functions (2.3.18) allow us to simplify the last expression, namely,
∞
ρ0dρ0
2 πκ
w (ρ,φ) = − cos 2 (1-κ)/2⌠
(a − ρ ) ρ
2 L ρ w (ρ0,φ)
2
π 2 ⌡ (ρ0 − a )
2
(ρ0 − ρ )  0
2 (1-κ)/2 2
a
∞ 2π
w (ρ0,φ0)ρ0 dρ0 dφ0
1
= − 2 cosπκ 2 2 (1-κ)/2⌠ ⌠
( a −ρ ) .
π 2 ⌡ ⌡ (ρ20−a 2)(1-κ)/2[ρ2 + ρ20 − 2ρρ0cos(φ−φ0)]
0 a
(2.4.14)
Expression (2.4.14) gives the normal displacements inside a circle ρ≤ a directly in
terms of the prescribed displacement outside the circle. We note a certain
similarity between (2.3.17) and (2.4.14).
Example 2. Penny-shaped crack in non-homogeneous elasticity. Consider

κ
a non-homogeneous elastic space with modulus of elasticity E = E 0| z | , E 0=const,
|κ|<1. This space is weakened by a penny-shaped crack ρ≤ a . The crack is
opened by arbitrary pressure σ(ρ,φ). The problem is to find the normal stress
on the plane z =0 outside the crack, the crack opening displacement, the stress
intensity factor, and the work required to open up the crack.
Due to the symmetry of the problem, it may be reduced to the mixed
boundary value problem of a half-space, subject to the boundary conditions at
z =0:
w = 0, τ = 0, for a <ρ<∞, 0≤φ<2π;
σ = σ(ρ,φ), τ = 0, for 0≤ρ≤ a , 0≤φ<2π. (2.4.15)
The governing integral equation takes the form (2.4.2), with the known function
2π a
σ(ρ ,φ ) ρ dρ dφ
f (ρ,φ) = − H ⌠ ⌠
0 0 0 0 0
. (2.4.16)
⌡ ⌡ [ρ2 + ρ20 − 2ρρ0cos(φ−φ0)](1+κ)/2
0 0
Its solution can be found in exactly the same manner as that of (1.4.25), and is
2π a
cos(πκ/2) ( a 2 − ρ20)(1-κ)/2σ(ρ0,φ0) ρ0dρ0dφ0
σ(ρ,φ) = − ⌠⌠ . (2.4.17)
π2(ρ2 − a 2)(1-κ)/2 ⌡ ⌡ ρ2 + ρ20 − 2ρρ0cos(φ−φ0)
0 0
Again, one should notice the similarity between (2.3.21) and (2.4.17). Expression
(2.4.17) gives the normal stress in the plane z =0 outside the crack in terms of
the pressure applied to the crack faces. Note that (1.4.27) may be considered as
a particular case of (2.4.17), when κ=0.
The crack opening displacement can be evaluated as a superposition of the
displacement caused by the applied pressure, and the displacement due to the
normal stress (2.4.17) outside the crack. By using a procedure analogous to the
one described in section 1.4, we obtain the expression
∞
πκ ⌠ x κd x ρ0dρ0 ρρ
x
w (ρ,φ) = 4 H cos  ⌠ L  0σ(ρ ,φ)

2
 ⌡ (x − ρ )
2 2 (1+κ)/2
⌡ ( x − ρ0)
2 2 (1+κ)/2
 x2  0
a a
ρ
x κd x ρ0dρ0
 x 2 σ(ρ ,φ), for ρ< a .
a
+ ⌠ ⌠ L (2.4.18)
⌡ (ρ − x )
2 2 (1+κ)/2
⌡ (ρ0 − x )
2 2 (1+κ)/2
ρρ0 0 
0 x
Substitution of (2.4.17) in (2.4.18) leads, after simplification, to
x κd x ρ0dρ0 ρρ
a x
w (ρ,φ) = 4 H cos
πκ ⌠ ⌠ L  0σ(ρ ,φ),
2 ⌡ ( x 2 − ρ2)(1+κ)/2 ⌡ ( x 2 − ρ20)(1+κ)/2  x 2  0
ρ 0
for ρ< a . (2.4.19)
The crack opening displacements are now defined in terms of the applied
pressure.
k 1(φ) = lim[(ρ − a )(1-κ)/2σ(ρ,φ)]. (2.4.20)

ρ→ a
2cos(πκ/2)
a
ρ dρ ρ
k 1(φ) = ⌠ 2
0 0
 0
2 (1+κ)/2 L a σ(ρ0,φ). (2.4.21)
π(2 a )(1-κ)/2 ⌡ ( a − ρ0)  
0
Introduce the stress intensity function :
ρ
2cos(πκ/2) ρ dρ ρ
K 1(ρ,φ) = ⌠ 2
0 0
L  0σ(ρ ,φ). (2.4.22)
π(2ρ)(1-κ)/2 ⌡ (ρ − ρ0)
2 (1+κ)/2
ρ 0
0
One can see that the limiting case of the stress intensity function, when ρ→ a , is
the stress intensity factor. By using the property of the L-operators (1.2.3) we
may rewrite (2.4.19) as
x κd x ρ0dρ0 ρ
a x
w (ρ,φ) = 4 H cos
πκ ⌠
L ρ ⌠
L  0σ(ρ ,φ)
2 ⌡ ( x 2 − ρ2)(1+κ)/2 x  ⌡ ( x 2 − ρ20)(1+κ)/2  x  0
ρ 0
πH ⌠
x (1+κ)/2d x ρ
2 (1+κ)/2 L x K 1( x ,φ).
(3-κ)/2
= 2 (2.4.23)
⌡ (x − ρ )
2
 
ρ
The energy W may be defined by the integral
2π a
W = ⌠ ⌠ σ(ρ,φ) w (ρ,φ)ρ dρdφ. (2.4.24)

⌡⌡
0 0
2π a a
(3-κ)/2
π H ⌠ dφ ⌠ σ(ρ,φ)ρ dρ ⌠
x (1+κ)/2 d x ρ K ( x ,φ)
W = 2 2 (1+κ)/2 L x
⌡ ⌡ ⌡ (x − ρ )
2
  1
0 0 ρ
2π a x
π H ⌠ dφ ⌠ x (1+κ)/2 d x ⌠
(3-κ)/2 σ(ρ,φ)ρ dρ ρ K ( x ,φ)
= 2 2 (1+κ)/2 L x
⌡ ⌡ ⌡ (x − ρ )
2
  1
0 0 0
2π a x
π H ⌠ dφ ⌠ K 1( x ,φ) x
(3-κ)/2 (1+κ)/2
dx ⌠
ρ dρ ρ σ(ρ,φ).
= 2 2 (1+κ)/2 L x
⌡ ⌡ ⌡ (x − ρ )
2
 
0 0 0
Here the interchange of the order of integration was used twice. Now
comparison of the last expression with (2.4.22) yields the final result
2π a
2 πH ⌠⌠
1-κ 2
W = [ K (ρ,φ)]2ρ dρ dφ. (2.4.25)
cos(πκ/2) ⌡ ⌡ 1
0 0
Expression (2.4.25) interprets the stress intensity function squared as being

proportional to the energy per unit area required to open up the crack. In the
case of axial symmetry, formula (2.4.25) simplifies to
a
22-κπ3 H ⌠
W = [ K (ρ)]2ρ dρ,
cos(πκ/2) ⌡ 1
0
with
ρ
2cos(πκ/2) σ(ρ0,φ0)ρ0 dρ0
K 1(ρ) = ⌠ 2 (1+κ)/2 .
π(2ρ)(1-κ)/2 ⌡ (ρ − ρ0)
2
0
All the results obtained are valid for a transversely isotropic space provided that
κ=0, and H is defined by (2.1.9).
Exercise 2.4
1. Find the equivalent of (2.4.25) for an isotropic body.
2π a
1 − ν2 ⌠ ⌠
Answer: W = 2π [ K 1(ρ,φ)]2ρ dρ dφ,
E ⌡⌡
0 0
with
ρ
K 1(ρ,φ) =
√2 ⌠ x dx x  σ( x ,φ).
2 1/2 L ρ
π√ρ ⌡ (ρ − x )2
 
0
2. Solve the problem above for the case of axial symmetry.

a
1 − ν2 ⌠
Answer: W = 4π 2
[ K 1(ρ)]2ρ dρ,
E ⌡
0
with
ρ
√2 ⌠ σ( x ) x d x
K 1(ρ) = .
π√ρ ⌡ (ρ2 − x 2)1/2
0
Note : these results were obtained by Sneddon (1965).
3. A penny-shaped crack in a non-homogeneous elastic space is opened by the

pressure σ(ρ,φ)=σ0+σ1ρcosφ, with σ0=const and σ1=const. Find the normal stress
in the plane z =0 outside the crack.
2cos(πκ/2) a 3-κ 
σ F ,
3-κ 3-κ 5-κ a 2
π(3 − κ) ρ  0  2
Answer: σ(ρ,φ) = − ; ; ( )
2 2 ρ 

+
2 a 2 F 3-κ, 5-κ; 7-κ; (a )2 σ ρcosφ, for ρ> a .
5 − κ ρ 2 2 2 ρ  1

The result can be expressed in elementary functions for a homogeneous body:
2 
− sin-1( )σ0
a a
π(ρ2 − a 2)1/2
σ(ρ,φ) = −
ρ


+  -1 a 
3ρ2 − a 2 ρ
3ρ(ρ2 − a 2)1/2
− 3
a
sin ( ) σ
ρ 1
ρcosφ.

4. In the problem above find the crack opening displacements w .
− ρ2)(1-κ)/2 σ0 + σ1ρcosφ.
4cos(πκ/2) 2 2
Answer: w = 2 (a
(1 − κ)  3 − κ 
5. Consider a penny-shaped crack ρ≤ a . Find the stress distribution σ in the

plane z =0 outside the crack due to the action of a pair of equal concentrated
forces P , applied normally to the crack faces in opposite directions at the point
( b ,ψ), b < a .
Answer: σ(ρ,φ) = − 2 cos   2
P πκ a 2 − b 2(1-κ)/2 1
.
π  2  ρ − a 2
 ρ + b − 2 b ρcos(φ−ψ)
2 2
6. Prove the identity for a penny-shaped crack:

2π a 2π ∞
⌠ ⌠ σ(ρ,φ)ρ |n|+1 in φ
e dρ dφ = − ⌠ ⌠ σ(ρ,φ)ρ|n|+1 ein dρ dφ.
φ
⌡⌡ ⌡⌡
0 0 0 a
Note : the identity states that the normal stress in the plane z =0 is in equilibrium,
which is not the case for an external crack.
7. Express the stress intensity function K 1(ρ,φ) in terms of the displacement w .

a
cos(πκ/2)
Answer: K 1(ρ,φ) = − (1-κ)/2 2 (1+κ)/2L(ρ)
d ⌠ x dx 1
2 (1-κ)/2L x w ( x ,φ).
2 π Hρ dρ ⌡ (x − ρ )
2
 
ρ
Hint : perform the inversion of (2.4.23).

 d 
a
( a 2 − x 2)(1-κ)/2 π(1 − κ) f ( a )
lim  ⌠ f ( x ) d x  = − .
dρ ⌡ ( x − ρ )
2 2 (1- κ)/2
2cos(πκ/2)
 
ρ→ a
ρ
Hint : use the substitution t =( x 2−ρ2)/( a 2− x 2).
9. Use the identity above to express the stress intensity factor in terms of the
displacement w .
lim
1 − κ w (ρ,φ) 
Answer: k 1(φ) = 2-κ .
2 π H ρ→a ( a − ρ)(1-κ)/2
Hint : compute the limit ρ→ a of the result in Exercise 7.
10. Consider a transversely isotropic elastic space weakened by a penny-shaped

crack ρ≤ a in the plane z =0, subjected to arbitrary pressure σ(ρ,φ). Find the
complex tangential displacements u in the plane z =0.
2π a
( a 2 − ρ20)1/2 σ(ρ0,φ0) ρ0dρ0dφ0
Answer: u = − H α ⌠ ⌠ , for ρ≤ a ;
⌡ ⌡ ( a 2 − ρρ e-i(φ-φ0) )1/2 ρe -i φ -i φ0
− ρ0e
0 0 0
2π a
( a 2 − ρ20)1/2
Hα ⌠ ⌠
2
u = −
π ⌡ ⌡ ( a 2 − ρρ e-i(φ-φ0) )1/2
0 0 0
-i(φ-φ0) 1/2
( a 2 − ρρ0e ) σ(ρ0,φ0) ρ0dρ0dφ0
×tan-1  , for ρ> a .
 (ρ2 − a 2)1/2 ρe
 -i φ
− ρ0e
-i φ0
Hint : see (Fabrikant, 1987a) for details.
2 3
2.5 Integral representation for q / R
While it was sufficient to know the integral representation for 1/ R in order

to solve the mixed boundary value problems of type I, this is no longer the
case for the problems of type II. The need to know the integral representation
for q 2/ R 3 is quite obvious from (2.2.12). We recall that q is defined by (2.2.5),
and R is given by (2.2.14). The original derivation was made by the author
many years ago. It was very long and cumbersome. Here we present only the
idea used, and the final result.
Since
q 2/ R 3= q /( q R ), (2.5.1)
we may use the following expansion:
φ i φ0
ρei − ρ0e  ∞
ρ i(φ-φ0) 
Σ
2 i φ0
= e  ρ 2  ρ k -ik(φ-φ0)
1 − ρ   ρ  e − e , for ρ<ρ0;
φ -i φ0 ρ0
ρe-i − ρ0e  0
k=0
0

φ i φ0
ρei − ρ0e      e 0 − 0e-i(φ-φ0) , for ρ>ρ ;
ρ0 2 ∞ ρ0 k ik(φ-φ ) ρ
ρe-i
φ
− ρ0e
-i φ0
= e  1 −


2 iφ
ρ  ρ
k=0
ρ Σ 
0
(2.5.2)
Now we have to substitute (2.5.2) and (1.1.27) in (2.5.1). This procedure yields,
for ρ<ρ0,
2 i φ0
 ρ k e-ik(φ-φ0) − ρ ei(φ-φ0) 
∞
q2
R3
= e  1 −  ρ 2
 ρ0 

Σk=0
ρ0 ρ0

ρ
∞
( x 2/ρρ0)|n|d x
× Σ
n=-∞
2 in(φ-φ0) ⌠
π
e
⌡ (ρ − x ) (ρ0 − x )
2 2 1/2 2 2 1/2 .
0
(2.5.3)
A very tedious procedure follows: grouping together the terms belonging to each
harmonic. The derivation does not end there: we shall also need to use the
identities:
n
min (ρ0,ρ)
(ρ20 − x) 2
Σρ
k=0
2(n-k) 2k
x − ρ2n x 2
⌠ dx
⌡ (ρ2 − x 2)1/2(ρ20 − x 2)1/2
0
min (ρ0,ρ)
x 2n[(2 n + 1)ρ20 − (2 n + 2) x 2]
=⌠ dx,
⌡ (ρ2 − x 2)1/2(ρ20 − x 2)1/2
0
(2.5.4)
and
min (ρ0,ρ)
(2 n − 1)ρ2 − 2 n x 2
ρ20 ⌠ 2 1/2 x
2n-2
dx
⌡ (ρ − x ) (ρ0 − x )
2 2 1/2 2
0
min (ρ0,ρ)
2 n ρ2 − (2 n + 1) x 2
= ⌠ x 2nd x , for n =1,2,3, ... (2.5.5)
⌡ (ρ2 − x 2)1/2(ρ20 − x 2)1/2
0
The first identity may be proven by the mathematical induction method, the
second one may be established by using the property
min (ρ0,ρ)
⌠ d[ x 2n-1(ρ2 − x 2)1/2(ρ20 − x 2)1/2] = 0.

⌡
0
A similar procedure is required for the case ρ>ρ0. The final result is
min (ρ0,ρ)
2 e e
∞
2 i φ in(φ-φ0)
Σ (2 n + 1)ρ2 − (2 n + 2) x 2 2n 
2
q
=  ⌠ x dx
R3 π ρ2 (ρρ0)n ⌡ (ρ2 − x 2)1/2(ρ20 − x 2)1/2 
 n=0 0
min (ρ0,ρ)
∞ 2 i φ0 -in(φ-φ0) (2 n + 1)ρ20 − (2 n + 2) x 2
+ Σ
n=0
e
2
e ⌠
ρ0  (ρρ0)n ⌡ (ρ − x )
2 2 1/2
(ρ20 − x)
2 1/2 x 2n d x 

0
min (ρ0,ρ)
i (φ+φ )
x2 dx 
⌠
0
e
−
ρρ0 . (2.5.6)
⌡ (ρ2 − x 2)1/2(ρ20 − x 2)1/2

0
Expression (2.5.6), though looking cumbersome, will prove very useful for solving
internal mixed boundary value problem of the type II. We need yet another
integral representation which is useful in external problems. The procedure is as
tedious as the one described above, with the final result
∞
 ∞
(2 n + 1) x 2 − (2 n + 2)ρ20
Σ
e i φ (e 0 ρρ0)n ⌠ 
2
q 2 i( φ - φ )
3 = π
2
2 1/2 d x
R
 n=0  ⌡ x (x 2n+2 2
− ρ2 1/2 2
) (x − ρ0)

max (ρ0,ρ)
∞ ∞
Σ
2 i φ0
+ e (e -i(φ-φ0)
ρρ0)n ⌠
(2 n + 1) x 2 − (2 n + 2)ρ2
dx
 ⌡ x 2n+2( x 2 − ρ2)1/2( x 2 − ρ20)1/2 
n=0 max (ρ0,ρ)
∞

ρρ0 ⌠
i (φ+φ ) dx
− e 0
. (2.5.7)
⌡ x 2( x 2 − ρ2)1/2( x 2 − ρ20)1/2

max (ρ0,ρ)
Here the following identities were used:

n
∞
( x 2 − ρ2) Σx ρ0
2k 2(n-k)
− ρ2 x 2n
⌠
k=0
dx
⌡ x 2n+2( x 2 − ρ2)1/2( x 2 − ρ20)1/2
max (ρ0,ρ)
∞
(2 n + 1) x 2 − (2 n + 2)ρ2
= ρ2n ⌠ dx,
0
⌡ x 2n+2( x 2 − ρ2)1/2( x 2 − ρ20)1/2
max (ρ0,ρ)
and
∞
2 n ρ20 − (2 n − 1) x 2
⌠ dx
⌡ x 2n( x 2 − ρ2)1/2( x 2 − ρ20)1/2
max (ρ0,ρ)
∞
(2 n + 1)ρ20 − 2 n x 2
= ρ ⌠ 2
dx.
⌡ x 2n+2( x 2 − ρ2)1/2( x 2 − ρ20)1/2
max (ρ0,ρ)
These identities can be derived from (2.5.4) and (2.5.5) by a formal substitution
of x by ρρ0/ x . As usual, when the final result is achieved, one can find an
easier way to do it. We shall discuss further some generalizations of the
integral representations derived here (see section 2.7).
Exercise 2.5
1. Prove the identities (2.5.4) and (2.5.5).
2. Establish the representation (2.5.6).
3. Establish the representation (2.5.7).
2.6 Internal mixed problem of type II
The material in this section follows essentially the paper (Fabrikant, 1971c).
Consider a transversely isotropic elastic half-space z ≥0. Let the normal traction
σ be prescribed all over the plane z =0. An arbitrary tangential displacement
u = u x+i u y is specified inside a circle ρ= a , while the complex shear loading τ is
known outside the circle. The problem is to find the shear traction inside the
circle. The mathematical formulation of the boundary conditions is
σ = σ(ρ,φ), for 0≤ρ<∞, 0≤φ<2π;
τ = τ(ρ,φ), for a <ρ<∞, 0≤φ<2π;
u = u (ρ,φ), for 0≤ρ≤ a , 0≤φ<2π; (2.6.1)
The governing integral equation can be written due to (2.2.12):
2π a 2π a
τ(ρ0,φ0) ρ0dρ0dφ0 q τ(ρ0,φ0) ρ0dρ0dφ0
1 ⌠⌠
+ G2⌠ ⌠
1
G = χ(ρ,φ). (2.6.2)
2 1⌡ ⌡ R 2 ⌡⌡ qR
0 0 0 0
Function χ is known from the boundary conditions (2.6.1):
2π ∞
χ(ρ,φ) = u + H α⌠ ⌠
⌡⌡ ρe-i φ − ρ0e
-i φ0
0 0
2π ∞ 2π ∞
− G1⌠ ⌠ G ⌠⌠
1 1
− .
2 ⌡⌡ R 2 2⌡ ⌡ qR
0 a 0 a
(2.6.3)
Though a closed form exact solution of (2.6.2) is possible, we present first its
exact solution in Fourier series. Assume validity of the expansions:
∞ ∞
τ(ρ,φ) = Στ (ρ)
n=-∞
n
in φ
e , χ(ρ,φ) = Σχ (ρ)
n=-∞
n
ein φ. (2.6.4)
Substitution of (2.5.6) and (2.6.4) in (2.6.2) leads to an infinite set of integral

equations
2G1 ρ x 2n+2d x
a τn+1(ρ0) dρ0
⌠ ⌠
ρn+1 ⌡ (ρ2 − x 2)1/2 ⌡ ρn0(ρ20 − x 2)1/2
0 x
2G2 ρ x 2n-2d x
a
(2 n − 1)ρ2 − 2 n x 2
+ ⌠ ⌠ τ-n+1(ρ0) dρ0 = χn+1(ρ),
ρn+1 ⌡ (ρ2 − x 2)1/2 ⌡ ρn-2
0 (ρ0 − x )
2 2 1/2
0 x
(2.6.5)
2G1 ρ x 2n-2d x
a τ-n+1(ρ0) dρ0
⌠ ⌠
ρn-1 ⌡ (ρ2 − x 2)1/2 ⌡ ρn-2
0 (ρ0 − x )
2 2 1/2
0 x
2G2 ρ x 2n-2d x
a
(2 n − 1)ρ20 − 2 n x 2
+ ⌠ ⌠ τn+1(ρ0) dρ0 = χ-n+1(ρ),
ρn-1 ⌡ (ρ2 − x 2)1/2 ⌡ ρn0(ρ20 − x 2)1/2
0 x
(2.6.6)
Equations (2.6.5) and (2.6.6) are valid for n =1,2,3, ... . In the case of axial
symmetry, n =0, and the integral equation takes the form
ρ
x 2d x
a
G 1τ1(ρ0) − G 2τ1(ρ0)
2 ⌠ ⌠ dρ0 = χ1(ρ). (2.6.7)
ρ ⌡ (ρ2 − x 2)1/2 ⌡ (ρ20 − x 2)1/2
0 x
Its solution is elementary, namely,
G 1χ1(ρ0) + G 2χ1(ρ0)
a x
d ⌠
dρ0.
2 d ⌠ dx
τ1(ρ) = − 2 2 2 1/2 d x x
2 dρ
π ( G 1 − G 2) ⌡ (x − ρ )
2
 ⌡ ( x 2 − ρ20)1/2 
ρ 0
(2.6.8)
The general solution of the system (2.6.5) and (2.6.6) can be presented in the
form
a
f (t) dt G
τ-n+1(ρ) = ρ ⌠
-n+1
n-1
+  2 C + D ρn-1
,
⌡ ( t 2 − ρ2)1/2 G 1 n n ( a 2 − ρ2)1/2
ρ
a a a
f n+1( t ) d t f n+1( t ) d t
τ (ρ) = ρ ⌠ + n+1⌠ y 2n-1d y ⌠
n-1 2n
n+1
⌡ ( t 2 − ρ2)1/2 ρ ⌡ ⌡ ( t 2 − y 2)1/2
ρ ρ y
+ C n 2
ρn-1 2n ⌠ y 2n-1 d y 
+ .
( a − ρ2)1/2 ρn+1 ⌡ ( a 2 − y 2)1/2
ρ
(2.6.9)
Here f k are the as yet unknown complex functions, and C n and D n are the as
yet unknown constants. By substitution of (2.6.9) in (2.6.5) and (2.6.6), we
obtain, after interchanging the order of integration and integration with respect to
ρ0,
ρ
πG1 a
ρ ⌠ t 2n-1 f ( t ) d t + C ρa 2n-1 − ⌠ (ρ2 − t 2)1/2 t 2n-1 f ( t ) d t 
ρn+1  ⌡ n+1 n
⌡ n+1 
0 0
ρ
πG2
+ ⌠ (ρ2 − t 2)1/2 t 2n-1 f ( t ) d t = χ (ρ), (2.6.10)
ρn+1 ⌡ -n+1 n+1
0
1 1
πG1 ρ a Γ( ) Γ( n − )
 ⌠ x 2n-2d x 2 2 2n-2
⌠ f (t) dt + Dn ρ
ρn-1
 ⌡ (ρ2 − x 2)1/2 ⌡ -n+1 2Γ( n ) 
0 x
ρ
πG2 x 2n-2d x
a
− ⌠ ⌠ f ( t ) d t = χ (ρ). (2.6.11)
ρn-1 ⌡ (ρ2 − x 2)1/2 ⌡ n+1 -n+1
0 x
Expression (2.6.10) can be simplified if we define C n as
Cn = −a -2n+1 ⌠ t 2n-1 f ( t ) d t . (2.6.12)

⌡ n+1
0
Application of the operator

r
-2n+2 d ⌠ ρndρ
r
d r ⌡ ( r − ρ2)1/2
2
0
to both sides of (2.6.11) yields
a a rχ (ρ)ρndρ
π ⌠
G f ( t ) d t − G 2⌠ f n+1( t ) d t + G 1 D n = r -2n+2
2
d ⌠ -n+1
.
2  1 ⌡ -n+1 ⌡  d r ⌡ ( r 2 − ρ2)1/2
r r 0
(2.6.13)
Since expression (2.6.13) should stay valid in the limiting case of r → a , this
defines D n as follows:
aχ (ρ)ρndρ
2 d ⌠ -n+1
Dn = 2 1/2. (2.6.14)
π G1a ⌡ (a − ρ )
2 2n-2 d a 2
0
Both constants are now defined. Inversion of (2.6.10) and differentiation of

(2.6.13) lead to the system
− G 1 f n+1( r ) + G 2 f -n+1( r ) = ψn+1( r ),
− G 1 f -n+1( r ) + G 2 f n+1( r ) = ψ-n+1( r ).

(2.6.15)
Here the notations were introduced
r d[ρn+1χn+1(ρ)]
2 d ⌠
ψn+1( r ) = 2 2n-1 ,
πr dr ⌡ ( r 2 − ρ2)1/2
0
dρ r ρnχ (ρ)
2 d  -2n+2 d ⌠  -n+1
ψ-n+1( r ) = 2 r 2 1/2 .
π d r  d r ⌡ (r − ρ ) 
2
0
(2.6.16)
Solution of the system (2.6.15) yields
G 1ψ-n+1( r ) + G 2ψn+1( r )
f -n+1( r ) = − ,
G 21 − G 22
G 1ψn+1( r ) + G 2ψ-n+1( r )
f n+1( r ) = − . (2.6.17)
G 21 − G 22
The general solution is now completed. It is given by formulae (2.6.9), with

the constants C and D defined by (2.6.12) and (2.6.14), and functions f defined
by (2.6.17) and (2.6.16).
Example. Consider a transversely isotropic elastic space weakened by an
external circular crack ρ≥ a in the plane z =0. Two identically oriented equal
concentrated forces P are applied normally to the crack faces at the points
(ρ0,φ0,0±), ρ0> a . Let us find the tangential stress in the crack neck. Due to
the antisymmetry of the applied load, the problem can be reduced to the one of
a half-space, with the tangential displacements and the normal stress vanishing in
the crack neck. The mathematical formulation of the boundary conditions is
σ = P δ(ρ−ρ0)δ(φ−φ0)/ρ for 0≤ρ<∞, 0≤φ<2π;
τ = 0, for a <ρ<∞, 0≤φ<2π;
u = 0, for 0≤ρ≤ a , 0≤φ<2π; (2.6.18)
Here δ(⋅) is the Dirac delta-function. The governing integral equation corresponds
to (2.6.2), with the right hand side χ, defined by (2.6.3), i.e.
∞
PH α PH α
Σ(e ρ n
-i(φ-φ0)
χ(ρ,φ) = = − ).
-i φ0 -i φ0 ρ0
ρe-i φ − ρ0e ρ0e n=0
The general solution, presented above, yields the following results
i φ0
PH αe i φ0 ρ n-1
χ-n+1(ρ) = − (e ) , χn+1(ρ) = 0,
ρ0 ρ0
2 PH αΓ( n )
f -n+1 = f n+1 = C n =0, Dn = − -i φ0 n
π3/2 G 1Γ( n − 12)(ρ0e )
(2.6.19)
Substitution of (2.6.19) in (2.6.17) and (2.6.9) leads to the solution
ρne-in φ
2 PH α
τ(ρ,φ) = − 3/2
π G1 Σ n=0 (ρ0e
Γ( n + 1)
-i φ0 n+1 1 (a − ρ )
) Γ( n + 2)
2 2 1/2.
The summation can be performed, according to the scheme

∞
ΣΓ(n
n=0
Γ( n + 1)
+ 2)
1
ζ = π
n -1/2 1
F (1, 1; ; ζ) =
2
π-1/2 
1 − ζ
1 + 
1
ζ 1/2
− ζ 
sin-1√ζ.

Here we have used the well known property of the hypergeometric functions
(Bateman and Erdélyi, 1955). Now the final result will take the form
1 
1 + 
b 1/2
sin-1√ b ,
2 PH α
τ(ρ,φ) = − -i φ0 1 − b 1 − b  
π2 G 1ρ0e ( a 2 − ρ2)1/2
-i(φ-φ )
where b =(ρ/ρ0)e 0 . In the case of isotropy, the last formula coincides with
the result of Ufliand (1967).
Exercise 2.6
1. Consider a transversely isotropic elastic half-space z ≥0. The tangential
displacement u = u 0=const is prescribed inside a circle ρ= a . The shear traction
outside the circle is equal to zero, and the normal pressure vanishes all over the
plane z =0. Find the shear traction inside the circle.
2u0
Answer: τ(ρ) = 2 .
π G 1( a 2 − ρ2)1/2
2. In the problem above find the normal displacement w in the plane z =0.
-i φ
4ℜ ( u 0 e ) H α a
Answer: w = , for ρ> a ;
π G 1ρ
-i φ
4ℜ( u 0e ) H α[ a − ( a 2 − ρ2)1/2]
w = , for ρ≤ a ;
π G 1ρ
3. Subject to the conditions of the first problem, find the tangential

displacement outside the circle ρ= a .
Answer: u =  u 0sin-1( ) + u 0 2
2 a G a (ρ2 − a 2)1/2 2 i φ
e .
π ρ G1 ρ2 
2.7 External mixed problem of type II
Consider a transversely isotropic elastic half-space z ≥0. Let the normal

traction σ be prescribed all over the plane z =0. An arbitrary tangential
displacement u = u x+i u y is specified outside a circle ρ= a , while the complex shear
loading τ is known inside the circle. The problem is to find the shear traction
outside the circle, the tangential displacement inside, and the normal displacement
at the plane z =0. The mathematical statement of the boundary conditions is
σ = σ(ρ,φ), for 0≤ρ<∞, 0≤φ<2π;
τ = τ(ρ,φ), for 0≤ρ< a , 0≤φ<2π;
u = u (ρ,φ), for a ≤ρ<∞, 0≤φ<2π. (2.7.1)
The governing integral equation can be written due to (2.2.12):
2π ∞ 2π ∞
G1⌠ ⌠ G 2⌠ ⌠
1 1
+ = χ(ρ,φ). (2.7.2)
2 ⌡⌡ R 2 ⌡⌡ qR
0 a 0 a
Function χ is known due to the boundary conditions (2.7.1):
2π ∞
χ(ρ,φ) = u + H α⌠ ⌠
⌡⌡ ρe-i φ − ρ0e
-i φ0
0 0
2π a 2π a
G ⌠⌠ G⌠⌠
1 1
− − .
2 1⌡ ⌡ R 2 2⌡ ⌡ qR
0 0 0 0
(2.7.3)
Once equation (2.7.2) has been solved, the normal displacement is found from
2π ∞ 2π ∞
 τ(ρ0,φ0) ρ0dρ0dφ0 σ(ρ0,φ0) ρ0dρ0dφ0
w (ρ,φ) = H α ℜ ⌠ ⌠  + H ⌠ ⌠ .
⌡
 0 0⌡ iφ
ρe − ρ0e
i φ0
 ⌡ ⌡ R
0 0
(2.7.4)
We present the exact solution of (2.7.2) in Fourier series. Assume validity of
the expansions:
∞ ∞
τ(ρ,φ) = Στ (ρ)
n=-∞
n
e ,
in φ
χ(ρ,φ) = Σχ (ρ)
n=-∞
n
e .
in φ
(2.7.5)
Substitution of the integral representation (2.5.7) and (2.7.5) in (2.7.2) leads to an

infinite set of coupled integral equations
∞ ∞
dx
x τn+1(ρ0)ρn+2
0 dρ0 dx
2 G 1ρn+1⌠ ⌠ + 2 G 2ρn+1 ⌠
⌡x 2n+2
(x − ρ )
2 2 1/2
⌡ (x − 2
ρ20)1/2 ⌡x 2n+2
( x − ρ2)1/2
2
ρ a ρ
x
2 n x 2 − (2 n + 1)ρ20
×⌠ τ-n+1(ρ0)ρn0 dρ0 = χn+1(ρ), for n =0,1,2, ... (2.7.6)
⌡ (x − ρ20)1/2
2
a
∞ xτ ∞
dx (ρ0)ρn0 dρ0 (2 n −1) x 2 − 2 n ρ2
2 G 1ρ ⌠ ⌠ + 2 G 2ρ ⌠
-n+1
n-1 n-1
dx
⌡ x 2n-2( x 2 − ρ2)1/2 ⌡ ( x 2 − ρ20)1/2 ⌡ x 2n( x 2 − ρ2)1/2
ρ a ρ
x τn+1(ρ0)ρn0 dρ0
×⌠ = χ-n+1(ρ), for n =1,2,3, ... (2.7.7)
⌡ ( x 2 − ρ20)1/2
a
Here
ρ a
⌠ σn(ρ0) x 2nd x
− n+1 ⌠
2
χn+1(ρ) = u n+1(ρ) + 2π H αρ -n-1
ρn+1
0 dρ0
⌡ ρ ⌡ (ρ − x )
2 2 1/2
0 0
a G 1 x 2τn+1(ρ0) + G 2[2 n ρ2 − (2 n + 1) x 2]τ-n+1(ρ0)

×⌠ dρ0, (2.7.8)
⌡ ρn0(ρ20 − x 2)1/2
x
and
∞ a
⌠ σ (ρ ) x 2n-2d x
− n-1 ⌠
2
χ-n+1(ρ) = u -n+1(ρ) + 2π H αρ n-1
ρ-n+1
0 dρ0
⌡ -n 0 ρ ⌡ (ρ − x )
2 2 1/2
ρ 0
a G 1ρ20τ-n+1(ρ0) + G 2[(2 n − 1)ρ20 − 2 n x 2]τn+1(ρ0)

×⌠ dρ0, (2.7.9)
⌡ ρn0(ρ20 − x 2)1/2
x
The case of axial symmetry corresponds to n =0, and we have only one equation
(2.7.6) to solve. The general solution of the system (2.7.6) and (2.7.7) may be
presented in the form
ρ f (t) dt
Dn
1  ⌠ n+1

τn+1(ρ) = n+1 2 1/2 + 2 1/2 , for n =0,1,2, ...
ρ  ⌡ (ρ − t )
2
(ρ − a ) 
2
a
ρ y f -n+1( t )
d  2n ⌠
τ-n+1(ρ) = n+1 ⌠ dt
1 ydy
2 1/2 d y y
ρ ⌡ (ρ − y )
2
 ⌡ t 2n+1 
a a
ρ
+ C n n+1 2
1 2n ⌠ y 2nd y 
+ , for n =1,2,3, ...
ρ (ρ − a 2)1/2 ρn+1 a 2n+1 ⌡ (ρ2 − y 2)1/2
a
(2.7.10)
Here functions f are to be determined, and C and D are the as yet unknown
complex constants. We used the same notation as in the previous section hoping
that this would not produce any confusion, and that the reader would understand
clearly that, for example, C n of this section is not equal to C n from the
previous one. Substitution of (2.7.10) in equations (2.7.6) yields, after
simplification,
∞ x
πρn+1 ⌠
dx G  ⌠ f ( t ) dt + D 
⌡ x ( x − ρ2)1/2  1 ⌡ n+1
2n+2 2 n

ρ a
− G 2 ⌠ f -n+1( t ) d t + C n = χn+1(ρ), (2.7.11)

 ⌡ 
a
∞ x f -n+1( t ) d t Cn
πρ G 1 ⌠
n-1  ⌠ xdx
+ 
⌡ ( x 2
− ρ2 1/2
)  ⌡ t 2n+1 a 2n+1
ρ a
∞
( x 2 − ρ2)1/2
+ πρ G 2 ⌠n-1
f n+1( x ) d x = χ-n+1(ρ). (2.7.12)
⌡ x 2n+1
ρ
The following identity was used during the procedure of integration by parts
d  ( x 2 − ρ2)1/2 (2 n − 1) x 2 − 2 n ρ2
− = .
dx x 2n  x 2n+1( x 2 − ρ2)1/2
Looking at the first integral in (2.7.12), one may conclude that it converges only
if the term in square brackets tends to zero when x →∞. This condition defines
C n as follows:
Cn = −a 2n+1 ⌠ f ( t ) t -2n-1d t . (2.7.13)

⌡ -n+1
a
By dividing both sides of (2.7.11) by ρn(ρ2− r 2)1/2, integrating with respect to ρ

from r to ∞, differentiating with respect to r , and multiplying the result by
r 2n+2/π, we obtain
r r
G 1 ⌠ f n+1( t ) d t + D n − G 2 ⌠ f -n+1( t ) d t + C n
 ⌡   ⌡ 
a a
∞ χn+1(ρ) dρ
2 d ⌠
= − 2 r 2n+2 . (2.7.14)
π dr ⌡ ρn(ρ2 − r 2)1/2
r
Proceeding to the limit as r → a in (2.7.14) gives us the formula for evaluating

D n, namely,
∞ χn+1(ρ) dρ
1  2 d ⌠ .
Dn = G C − 2 a 2n+2 (2.7.15)
G1  2 n π d a ⌡ ρ (ρ − a ) 
n 2 2 1/2
a
Inversion of (2.7.12) and differentiation of (2.7.14) lead to the system of

equations
G 1 f -n+1( r ) − G 2 f n+1( r ) = ψ-n+1( r ),
G 1 f n+1( r ) − G 2 f -n+1( r ) = ψn+1( r ). (2.7.16)
Here
∞ χ (ρ)
2 d ⌠ dρ d  -n+1 
ψ-n+1( r ) = − 2 r 2n+1 ,
π d r ⌡ (ρ2 − r 2)1/2 dρ ρn-1 
r
∞ χn+1(ρ) dρ
2 d  2n+2 d ⌠ .
ψn+1( r ) = − 2 r (2.7.17)
π dr dr ⌡ ρ (ρ − r ) 
n 2 2 1/2
r
The following rule of differentiation was used during the above transformations:
∞ ∞
d ⌠
= ρ⌠
f( x) dx d[ f ( x )/ x ]
. (2.7.18)
dρ ⌡ ( x − ρ2)1/2
2
⌡ ( x 2 − ρ2)1/2
ρ ρ
Solution of the system (2.7.16) is
G 1ψn+1( r ) + G 2ψ-n+1( r )
f n+1( r ) = ,
G 21 − G 22
G 1ψ-n+1( r ) + G 2ψn+1( r )
f -n+1( r ) = . (2.7.19)
G 21 − G 22
The set of expressions (2.7.10), (2.7.13), (2.7.15), (2.7.17), and (2.7.19) gives a
complete solution to the problem. Since the shear traction is now known
throughout the plane z =0, the tangential displacements inside the circle can be
defined by (2.7.2), and the normal displacement is defined by (2.7.4). We

consider further the case of a penny-shaped crack in more detail, including the
derivation of closed form solutions.
Example: Penny-shaped crack. Let the crack of radius a be located in the
plane z =0. Both crack faces are loaded by arbitrary shear tractions acting in
opposite directions. The boundary conditions are antisymmetric, so the problem
can be reduced to a mixed one for a half-space, with the boundary conditions:
σ = 0, for 0≤ρ<∞, 0≤φ<2π;
τ = τ(ρ,φ), for 0≤ρ< a , 0≤φ<2π;
u = 0, for a ≤ρ<∞, 0≤φ<2π. (2.7.20)
The shear tractions outside the crack, the tangential displacement inside, and the
normal displacement in the plane z =0 are to be determined. The crack problem
can be considered as a particular case of a more general one solved above.
However, we can show that the crack problem has a simpler solution. Indeed,
the integral equations (2.7.6) and (2.7.7) remain unchanged, but instead of (2.7.8)
and (2.7.9) we have
a
2n
aG
1x
2
τn+1(ρ0)+ G 2[2 n ρ2−(2 n +1) x 2]τ-n+1(ρ0)
2 ⌠ x dx ⌠
χ (ρ) = − dρ0,
n+1 ρ ⌡ (ρ − x 2)1/2⌡
n+1 2
ρn0(ρ20 − x 2)1/2
0 x
(2.7.21)
and
1ρ0τ-n+1(ρ0)+ G 2[(2 n −1)ρ0−2 n x ]τn+1(ρ0)
a aG 2 2 2
2n-2
2 ⌠ x dx ⌠
χ (ρ) = − dρ0.
-n+1 ρ ⌡ (ρ − x 2)1/2⌡
n-1 2
ρn0(ρ20 − x 2)1/2
0 x
(2.7.22)
The structure of integral equations indicates the possibility of expression the yet
unknown τ outside the crack through the prescribed τ inside. Such a
representation takes the form
2
a ( a 2 − t 2)1/2τ ( t ) t n+2d t An
⌠
n+1
τn+1(ρ) = − + ,
πρn+1(ρ2 − a 2)1/2 ⌡ ρ2 − t 2 ρn+1(ρ2 − a 2)1/2
0
for n =0,1,2, ...
2
a ( a 2 − t 2)1/2τ ( t ) t nd t
⌠
-n+1
τ (ρ) = − , for n =1,2,3, ...
-n+1 πρn-1(ρ2 − a 2)1/2 ⌡ ρ2 − t 2
0
(2.7.23)
Here A n is the as yet unknown complex constant. Substitution of (2.7.23) in

(2.7.6) yields, after intermediate integration,
∞ a τn+1( t ) t n+2d t
2 G 1ρn+1 ⌠
dx  −⌠ π 
2 1/2 + 2 A n
⌡ x 2n+2
(x − ρ )  ⌡ (x − t )
2 2 1/2
2

ρ 0
∞ a
−⌠ 2 n x − (2 n + 1) t τ ( t ) t n d t
2 2
dx
+ 2 G 2ρn+1 ⌠ 2n+2 2
⌡ x ( x − ρ2)1/2 ⌡ ( x 2 − t 2)1/2 -n+1
ρ 0
+ (2 n + 1) ⌠ ( a 2 − t 2)1/2 τ ( t ) t n d t  = χ (ρ). (2.7.24)

⌡ -n+1 n+1 
0
The following identity can be verified by a formal substitution of x by ρ t / x :
∞ t
⌠ dx 1 ⌠ x 2n+2d x
= . (2.7.25)
⌡ x 2n+2( x 2 − ρ2)1/2( x 2 − t 2)1/2 (ρ t )2n+2 ⌡ (ρ2 − x 2)1/2( t 2 − x 2)1/2
ρ 0
Interchanging the order of integration in (2.7.24), substituting (2.7.25), and yet

another interchange lead to the expression
2G1 a
x 2n+2d x
a τn+1( t ) d t
− ⌠ ⌠ n 2
ρn+1 ⌡ ρ − ⌡ t (t − x )
2 2 1/2 2 1/2
( x )
0 x
2G2 a
2 n ρ2 − (2 n + 1) x 2
a τ (t) dt
⌠ x dx ⌠
-n+1
− 2n
ρn+1 ⌡ (ρ2 − x 2)1/2 ⌡ t (t − x )
n 2 2 1/2
0 x
+ ρ n+1 ⌠ dx πG A
⌡ x ( x − ρ2)1/2  1 n
2n+2 2
ρ
+ 2 G 2(2 n + 1) ⌠ ( a 2 − t 2)1/2 τ ( t ) t nd t  = χ (ρ). (2.7.26)

⌡
-n+1 n+1 
0
Comparison of (2.7.26) and (2.7.21) shows that the equation is satisfied if A n is

defined by
= − (2 n + 1)( G 2/ G 1) ⌠ τ (ρ)ρn( a 2 − ρ2)1/2dρ.

2
An (2.7.27)
π ⌡ -n+1
0
Since τ is defined for n ≥1 only, we may conclude that A 0=0. A similar

-n+1
procedure of substitution of (2.7.23) in (2.7.7) leads to
∞ a τ-n+1( t ) t n d t
2 G 1ρn-1 ⌠
dx −⌠ 
⌡ x 2n-2
(x − ρ )
2 2 1/2
 ⌡ (x − t ) 
2 2 1/2
ρ 0
∞
(2 n − 1) x 2 − 2 n ρ2
a τn+1( t ) t n
+ 2 G 2ρn-1 ⌠ d x −⌠ dt
⌡ x (x − ρ )
2n 2 2 1/2
 ⌡ ( x 2 − t 2)1/2
ρ 0
a
1 ⌠
A n = χ (ρ).
π
+ ( a 2 − t 2)1/2 τ ( t ) t n+2d t + (2.7.28)
ax  ⌡ n+1 2  -n+1
0
It may be noted that the integral
∞ ∞
⌠ (2 n −1 ) x − 2 n ρ d x = ⌠ d− ( x − ρ )  = 0, for n ≥1.

2 2 2 1/2 2
⌡ x 2n+1( x 2 − ρ2)1/2 ⌡  x 2n 
ρ ρ
(2.7.29)
By using (2.7.25) and (2.7.29), equation (2.7.28) can be transformed into
2G1 a
x 2n-2d x
a τ-n+1( t ) d t
− ⌠ ⌠ n-2 2
ρn-1 ⌡ ρ − ⌡ t (t − x )
2 2 1/2 2 1/2
( x )
0 x
a a
2G2 x 2n-2d x (2 n − 1) t 2 − 2 n x 2
− ⌠ ⌠ τn+1( t ) d t = χ-n+1(ρ).
ρn-1 ⌡ ( ρ2
− x 2 1/2
) ⌡ t n( t 2 − x 2)1/2
0 x
Comparison of the last expression with (2.7.22) proves that equation (2.7.7) is
satisfied, and that (2.7.23) is indeed the solution to the crack problem. A closed
form solution can be obtained by summation of (2.7.23) and (2.7.27), with the
result
2π a
1 ( a 2 − ρ20)1/2τ(ρ0,φ0) ρ0dρ0dφ0
τ(ρ,φ) = − ⌠⌠
π (ρ − a )
2 2 2 1/2
⌡⌡ ρ2 + ρ20 − 2ρρ0cos(φ−φ0)
0 0
G 2e2 i φ
2π i(φ-φ0)
a
3 − (ρ0/ρ)e
− ⌠ ⌠ 2
− ρ20)1/2τ(ρ0,φ0)ρ0dρ0dφ0.
i(φ-φ0) 2( a
π G 1ρ (ρ − a ) ⌡ ⌡ [1 − (ρ /ρ)e
2 2 2 2 1/2
]
0 0 0
(2.7.30)
One can notice that the first integral in (2.7.30) corresponds to the solution for
the case of normal loading of a penny-shaped crack. Define the complex stress
intensity factor as
k (φ) = lim[(ρ − a )1/2τ(ρ,φ)]. (2.7.31)

ρ→ a
2π a
( a 2 − ρ20)1/2τ(ρ0,φ0) ρ0dρ0dφ0
k (φ) = −
1 ⌠⌠
π2√2 a ⌡ ⌡ a 2 + ρ20 − 2 a ρ0cos(φ−φ0)
0 0
(2.7.32)
2 iφ 2π i(φ-φ0)
G 2e a
3 − (ρ / a )e
− ⌠ ⌠ 0
2
− ρ20)1/2τ(ρ0,φ0) ρ0dρ0dφ0.
i(φ-φ0) 2 ( a
π G 1 a √2 a
2 2
⌡ ⌡ [1 − (ρ0/ a )e ]
0 0
Since our definition of τ contains both x- and y- components, so will the

expression for the stress intensity factor k = k x+i k y. If we require the expression
for the radial and tangential components, we have to use the relationship
φ
τzx + iτyz = (τzρ + iτφz)ei .
This allows us to rewrite (2.7.32) in terms of the second and third mode stress
intensity factors as follows:
e-i φ
2π a
( a 2 − ρ20)1/2τ(ρ0,φ0) ρ0dρ0dφ0
k2 + ik3 = − 2 ⌠⌠
π √2 a ⌡ ⌡ a 2 + ρ20 − 2 a ρ0cos(φ−φ0)
0 0
iφ 2π i(φ-φ0)
G 2e a
3 − (ρ0/ a )e
− ⌠ ⌠ 2
− ρ20)1/2 τ(ρ0,φ0) ρ0dρ0dφ0.
i(φ-φ0) 2 ( a
π G 1 a √2 a ⌡ ⌡ [1 − (ρ / a )e
2 2
]
0 0 0
(2.7.33)
While using (2.7.33), one should remember that τ is still defined in terms of the
cartesian coordinate components.
In order to define the tangential displacements inside the crack directly in
terms of the prescribed shear loading, equations (2.7.6) and (2.7.7) have to be
rewritten for ρ≤ a . They will have a similar form, with a difference only in the
limits of integration, namely,
∞
dx
x τn+1(ρ0)ρn+2
0 dρ0
u n+1(ρ) = 2 G 1ρn+1 ⌠ ⌠
⌡ x 2n+2( x 2 − ρ2)1/2 ⌡ ( x 2 − ρ20)1/2
a a
∞
dx
x
2 n x 2 − (2 n + 1)ρ20
+ 2 G 2ρn+1⌠ ⌠ τ-n+1(ρ0)ρn0 dρ0
⌡x (x − ρ ) ⌡ (x − ρ20)1/2
2n+2 2 2 1/2 2
a a
ρ a τn+1(ρ0) dρ0
2G1 x 2n+2d x
+ ⌠ ⌠ n 2
ρn+1 ⌡ (ρ − x ) ⌡ ρ0(ρ0 − x )
2 2 1/2 2 1/2
0 x
ρ a
2G2 x 2n d x 2 n ρ2 − (2 n + 1) x 2
+ ⌠ ⌠ τ-n+1(ρ0) dρ0,
ρn+1 ⌡ ρ − ⌡ ρn0(ρ20 − x 2)1/2
2 2 1/2
( x )
0 x
for n =0,1,2, ... (2.7.34)
∞
dx
x τ-n+1(ρ0)ρn0dρ0
u -n+1(ρ) = 2 G 1ρn-1 ⌠ ⌠
⌡ x (x − ρ ) ⌡ ( x 2 − ρ20)1/2
2n-2 2 2 1/2
a a
∞ x
dx (2 n − 1) x 2 − 2 n ρ2
+ 2 G 2ρ ⌠ 2n 2
n-1 ⌠ τn+1(ρ0)ρn0 dρ0
⌡ x ( x − ρ2 1/2
) ⌡ (x −
2
ρ20)1/2
a a
ρ a τ-n+1(ρ0) dρ0
2G1 x 2n-2d x
+ ⌠ ⌠ n-2 2
ρn-1 ⌡ ( ρ2
− x 2 1/2
) ⌡ ρ0 (ρ0 − x )
2 1/2
0 x
ρ
2G2 x 2n-2 d x
a
(2 n − 1)ρ20 − 2 n x 2
+ ⌠ ⌠ τn+1(ρ0) dρ0,
ρn-1 ⌡ (ρ2 − x 2)1/2 ⌡ ρn0(ρ20 − x 2)1/2
0 x
for n =1,2,3, ... (2.7.35)
The first two terms in (2.7.34) and (2.7.35) represent the displacement inside the
crack due to the shear traction outside, while the remaining terms give the
displacement caused by the shear tractions inside. Substitution of (2.7.23) in
(2.7.34) yields, after integration with respect to ρ0,
∞
dx
a τn+1( t ) t n+2d t
u n+1(ρ) = −2 G 1ρn+1⌠ ⌠
⌡ x (x − ρ ) ⌡ (x − t )
2n+2 2 2 1/2 2 2 1/2
a 0
∞ a
dx 2 n x 2 − (2 n + 1) t 2
− 2 G 2ρ n+1 ⌠ 2n+2 2 ⌠ τ-n+1( t ) t nd t
⌡ x (x − ρ ) ⌡ (x − t )
2 1/2 2 2 1/2
a 0
2G1 x 2n+2d x
+ ⌠ ⌠ n 2
ρn+1 ⌡ (ρ − x ) ⌡ ρ0(ρ0 − x )
2 2 1/2 2 1/2
0 x
ρ a
2G2 x 2nd x 2 n ρ2 − (2 n + 1) x 2
+ ⌠ ⌠ τ-n+1(ρ0) dρ0.
ρn+1 ⌡ ρ − ⌡ ρn0(ρ20 − x 2)1/2
2 2 1/2
( x )
0 x
(2.7.36)
Transform the third term in (2.7.36) by using (2.7.25). The procedure is as
follows:
2G1 x 2n+2d x
⌠ ⌠ n 2
ρn+1 ⌡ ( ρ2
− x 2 1/2
) ⌡ ρ0(ρ0 − x )
2 1/2
0 x
ρ τn+1(ρ0) dρ0 ρ0
2G1
 ⌠ x 2n+2d x
= ⌠
ρn+1  ⌡ ρn0 ⌡ (ρ2 − x 2)1/2(ρ20 − x 2)1/2
0 0
a
τn+1(ρ0) dρ0 ρ
 x 2n+2d x
+ ⌠ ⌠
⌡ ρn0 ⌡ ( ρ2
− x 2 1/2 2
) ( ρ0 − x 2 1/2
) 
ρ 0
ρ τn+1(ρ0) dρ0 ∞
2G1 (ρρ0)2n+2d x
=  ⌠ ⌠
ρn+1  ⌡ ρn0 ⌡ x 2n+2( x 2 − ρ2)1/2( x 2 − ρ20)1/2
0 ρ
∞
a
τn+1(ρ0) dρ0 (ρρ0)2n+2d x
+ ⌠ ⌠ 
⌡ ρn0 ⌡ x 2n+2( x 2 − ρ2)1/2( x 2 − ρ20)1/2
ρ ρ0
∞ a τn+1(ρ0)ρn+2
0 dρ0
= 2 G 1ρn+1 ⌠
dx
⌠
 ⌡ x 2n+2( x 2 − ρ2)1/2 ⌡ ( x 2 − ρ20)1/2
a 0
a x τn+1(ρ0)ρn+2
0 dρ0
+ ⌠
dx
⌠ . (2.7.37)
⌡ x 2n+2
(x − ρ )
2 2 1/2
⌡ (x − 2
ρ20)1/2 
ρ 0
An analogous transformation of the last term in (2.7.36) leads to the identity
ρ a
2G2 x 2nd x 2 n ρ2 − (2 n + 1) x 2
⌠ ⌠ τ-n+1(ρ0) dρ0
ρn+1 ⌡ (ρ2 − x 2)1/2 ⌡ ρn0(ρ20 − x 2)1/2
0 x
∞ a
2 n x 2 − (2 n + 1)ρ20
= 2 G 2ρ n+1  ⌠ dx
⌠ τ-n+1(ρ0)ρn0 dρ0
 ⌡ x ( x − ρ2)1/2 ⌡
2n+2 2
(x −
2
ρ20)1/2
a 0
2 n x 2 − (2 n + 1)ρ20
a x
τ-n+1(ρ0)ρn0 dρ0.
dx
+ ⌠ ⌠ (2.7.38)
⌡ x 2n+2
(x − ρ )
2 2 1/2
⌡ (x −2
ρ20)1/2 
ρ 0
Back substitution of (2.7.37) and (2.7.38) in (2.7.36) results in
a
dx
x τn+1(ρ0)ρn+2
0 dρ0
u n+1(ρ) = 2 G 1ρn+1 ⌠ ⌠
⌡ x 2n+2( x 2 − ρ2)1/2 ⌡ ( x 2 − ρ20)1/2
ρ 0
2 n x 2 − (2 n + 1)ρ20
a x
dx
+ 2 G 2ρ n+1 ⌠ ⌠ τ (ρ0)ρn0 dρ0,
⌡ x 2n+2( x 2 − ρ2)1/2 ⌡ (x −2
ρ20)1/2 -n+1
ρ 0
for n =0,1,2, ..., and ρ≤ a . (2.7.39)
A similar procedure can be applied to (2.7.35). Substitution of (2.7.23) in
(2.7.35) yields, after an integration with respect to ρ0,
∞
dx
a τ-n+1( t ) t n d t
u -n+1(ρ) = −2 G 1ρn-1 ⌠ ⌠
⌡ x (x − ρ ) ⌡ (x − t )
2n-2 2 2 1/2 2 2 1/2
a 0
∞
(2 n − 1) x 2 − 2 n ρ2
a τ (t) dt
+ 2 G 2ρ ⌠ d x −⌠
n+1
n-1
⌡ x 2n( x 2 − ρ2)1/2  ⌡ ( x 2 − t 2)1/2

a 0
a
1 ⌠
A n
π
+ ( a 2 − t 2)1/2τn+1( t ) t n+2 d t +
ax  ⌡ 2 
0
2G1 x 2n-2d x
+ ⌠ ⌠
ρn-1 ⌡ (ρ2 − x 2)1/2 ⌡ ρn-2
0 (ρ0 − x )
2 2 1/2
0 x
ρ
2G2 x 2n-2d x
a
(2 n − 1)ρ20 − 2 n x 2
+ ⌠ ⌠ τn+1(ρ0) dρ0.
ρn-1 ⌡ (ρ2 − x 2)1/2 ⌡ ρn0(ρ20 − x 2)1/2
0 x
(2.7.40)
The following identities may be established by using the procedures identical to
those used for deriving (2.7.37) and (2.7.38)
2G1 x 2n-2d x
⌠ ⌠
ρn-1 ⌡ (ρ2 − x 2)1/2 ⌡ ρn-2
0 (ρ0 − x )
2 2 1/2
0 x
∞ a τ (ρ0)ρn0
= 2 G 1ρn-1 ⌠
dx
⌠
-n+1
dρ
 ⌡ x 2n-2( x 2 − ρ2)1/2 ⌡ ( x 2 − ρ20)1/2 0
a 0
a x τ-n+1(ρ0)ρn0
dρ0,
dx
+ ⌠ ⌠ (2.7.41)
⌡ x 2n-2( x 2 − ρ2)1/2 ⌡ ( x 2 − ρ20)1/2 
ρ 0
ρ
2G2 x 2n-2d x
a
(2 n − 1)ρ20 − 2 n x 2
⌠ ⌠ τ (ρ0) dρ0
ρn-1 ⌡ (ρ2 − x 2)1/2 ⌡ ρn0(ρ20 − x 2)1/2 n+1
0 x
∞
(2 n − 1) x 2 − 2 n ρ2
a τn+1(ρ0)ρn0 dρ0
= 2 G 2ρn-1 ⌠ dx ⌠
 ⌡ x 2n( x 2 − ρ2)1/2 ⌡ ( x 2 − ρ20)1/2
a 0
a
(2 n − 1) x 2 − 2 n ρ2
+ ⌠ dx ⌠ . (2.7.42)
⌡ x (x − ρ )
2n 2 2 1/2
⌡ (x −2
ρ20)1/2 
ρ 0
The back substitution of (2.7.41) and (2.7.42) in (2.7.40) yields
a
dx
x τ-n+1(ρ0)ρn0
u (ρ) = 2 G 1ρn-1 ⌠ ⌠ dρ
-n+1 ⌡ x 2n-2( x 2 − ρ2)1/2 ⌡ ( x 2 − ρ20)1/2 0
ρ 0
a
(2 n − 1) x 2 − 2 n ρ2
+ 2 G 2ρn-1 ⌠ dx ⌠ + B n( a 2 − ρ2)1/2.
 ⌡ x (x − ρ )
2n 2 2 1/2
⌡ (x −2
ρ20)1/2 
ρ 0
(2.7.43)
Here
a
Bn = a -2n-1 ⌠ t n( a 2 − t 2)1/2[τ ( t ) − (2 n + 1)( G 2/ G 1)τ ( t )] d t . (2.7.44)

⌡ n+1 -n+1
0
Formulae (2.7.39), (2.7.43), and (2.7.44) give the tangential displacement of the
crack faces in terms of the prescribed shear tractions. Note that the displacement
vanishes outside the crack. The complex tangential displacement u can be
represented in terms of its harmonics as
∞ ∞
u (ρ,φ) = Σu
n=0
n+1
(ρ)e
i ( n +1)φ
+ Σu
n=1
-n+1
(ρ)e
-i ( n -1)φ
, (2.7.45)
where u n+1 and u -n+1 are defined by (2.7.39) and (2.7.44). The summation will
be performed for the terms containing G 1 and G 2 separately. Substitution of
(2.7.39) in (2.7.45) and summation of the terms with G 1 gives
2π
λ(ρρ0/ x 2, φ−φ0) d x τ(ρ0,φ0) ρ0dρ0
a x
G1
⌠ dφ ⌠ ⌠ . (2.7.46)
π ⌡ 0⌡ ( x 2 − ρ2)1/2 ⌡ ( x 2 − ρ20)1/2
0 ρ 0
Here we should change the order of integration according to the scheme
a x a a ρ a
⌠ d x ⌠ dρ = ⌠ dρ ⌠ d x + ⌠ dρ ⌠ d x . (2.7.47)
⌡ ⌡ 0 ⌡ 0⌡ ⌡
0
⌡
ρ 0 ρ ρ0 0 ρ
Integration with respect to x in (2.7.46) can be performed as in (1.1.23), and

yields
2π a
G1
⌠ ⌠  1 tan-1η   τ(ρ ,φ ) ρ dρ dφ . (2.7.48)
π ⌡⌡  R R   0 0 0 0 0
0 0
where R and η( x ) are defined by (2.2.14) and (1.1.6) respectively. We recall

that, according to our convention, the abbreviation η stands for η( a ). Summation
of the terms with G 2 in (2.7.39) and (2.7.43) is slightly more complicated, but
generally it reduces to the series
∞
Σ
n=0
nz n =
z
(1 − z )2
.
The result of summation is
2π
λ(ρρ0/ x 2, φ−φ0) d x
a
G2 2 i φ0
⌠e dφ0⌠
π ⌡ ⌡ ( x 2 − ρ2)1/2
0 ρ
-i φ
 x
4 iq ρe 0 sin(φ − φ0)  τ(ρ0,φ0) ρ0dρ0
×⌠  1 +  .
⌡  x 2(1 − ζ)(1 − ζ)
 ( x 2 − ρ20)1/2
0
(2.7.49)
i(φ-φ0)
Here ζ=(ρρ0/ x )e , and ζ is a complex conjugate of ζ. The integral in
2
(2.7.49), though looking formidable, is a perfect differential and can be computed

as indefinite
-i φ
λ(ρρ0/ x 2, φ−φ0) d x  4 iq ρe 0 sin(φ − φ0) 
⌠  1 + 
⌡ ( x 2 − ρ2)1/2 ( x 2 − ρ20)1/2  x 2(1 − ζ)(1 − ζ)

-i φ0
-2 i φ0
2 i ρe sin(φ − φ0) η( x )
tan-1
q η( x )
= e + .
qR  R  x 2 q (1 − ζ)(1 − ζ)
(2.7.50)
It is noteworthy, that (2.7.50) can be considered as a generalization of the
integral representation for q 2/ R 3, given by (2.5.7). Indeed, such a representation
can be obtained from (2.7.50) by taking a definite integral, namely,
∞ 2 i φ0 -i φ
e λ(ρρ0/ x 2, φ−φ0) d x  4 iq ρe 0 sin(φ − φ0) 
2 ⌠  1 +  =
q
.
π ⌡ ( x 2 − ρ2)1/2 ( x 2 − ρ20)1/2 x (1 − ζ)(1 − ζ)
2
qR
max (ρ0,ρ)  
By changing the order of integration in (2.7.49) and integrating with respect to

x , according to (2.7.50), we obtain
2π a i φ0
G2 2 i ρe sin(φ − φ )η
⌠ ⌠  q tan-1η  + 0
 τ(ρ ,φ )ρ dρ dφ , (2.7.51)
π ⌡ ⌡  qR R  a q (1 − t )(1 − t ) 
2 0 0 0 0 0
0 0
where t is defined by (1.4.41). The remaining step is the summation of B n

(2.7.44), with the result
2π a 2 i φ0
G2 τ(ρ0,φ0) e G 2 (3 − t ) τ(ρ0,φ0)
⌠⌠  −  η ρ dρ dφ .
πa ⌡ ⌡ 
2
(1 − t ) G1 (1 − t ) 2
 0 0 0
0 0
(2.7.52)
Finally, the summation of (2.7.48), (2.7.51) and (2.7.52) leads to
2π a 2
⌠ ⌠  1 tan-1η  − 2 (3 − t ) η  τ(ρ ,φ ) ρ dρ dφ
G1 G
u (ρ,φ) =
π ⌡⌡  R R  G 21 a 2(1 − t )2  0 0 0 0 0
0 0
2π a
G2 2 iφ
⌠ ⌠  q tan-1η  + η[( q / q ) − t e ]  τ(ρ ,φ ) ρ dρ dφ .
0
+
π ⌡ ⌡  qR R  a 2(1 − t )(1 − t )  0 0 0 0 0
0 0
(2.7.53)
Since the normal tractions vanish in the plane z =0, the normal displacement
w will be defined by (2.7.4) as
2π ∞
w (ρ,φ) = H α ℜ⌠ ⌠ ,
⌡⌡ ρei
φ
− ρ0e
i φ0
0 0
which is equivalent to the following series representation
 e-i( n +1)φ ρ ∞
w (ρ,φ) = 2π H α ℜ 
n=0 
ρ n+1 Σ ⌠ τ (ρ0) ρn+1
⌡ -n 0 dρ0
0
a
τn+1(ρ0) ∞ τn+1(ρ0) 
− e ρ  ⌠
in φ n
dρ0 + ⌠ dρ0  , for ρ≤ a ;
 ⌡ ρn0 ⌡ ρn0 
ρ a
(2.7.54)
 e-i( n +1)φ  a
∞
w (ρ,φ) = 2π H α ℜ 
n=0 
ρ n+1 Σ 
⌠ τ (ρ0) ρn+1
⌡ -n 0 dρ0
0
ρ ∞
τn+1(ρ0) 
+ ⌠ τ-n(ρ0) ρn+1 dρ0  − e ρn ⌠
in φ
dρ0 , for ρ≥ a .
⌡
0
 ⌡ ρn0

a ρ
(2.7.55)
By substituting formulae (2.7.23) and (2.7.27) in (2.7.54) and (2.7.55), and
performing the summation and integration, we obtain
2π a
 ( a 2 − ρ20)1/2 τ(ρ0,φ0)
w = H α ℜ ⌠ ⌠ ρ0dρ0dφ0
⌡
 0 0⌡ a (1 − t )1/2 q
2π a
 ( a 2 − ρ2)1/2 τ(ρ ,φ )eiφ0 dρ dφ , for ρ≤a ;
G2
+ ⌠⌠  1
− 1
G 1 a ⌡ ⌡ (1 − t )3/2  0 0 0 0 0
0 0 
and
2π a
 ( a 2 − ρ20)1/2 a (1 − t )  1/2 τ(ρ0,φ0)
H α ℜ ⌠ ⌠ tan  2
2
w = -1
ρ0dρ0dφ0
π
0⌡ ⌡ a (1 − t )1/2 (ρ − a 2)1/2 q
0
2π a
G2
⌠  1  a (1 − t )1/2 
+ ⌠ tan-1
−
1
sin-1
a
G 1 ⌡ ⌡  a (1 − t )3/2 (ρ2 − a 2)1/2 a ρ
0 0
-i φ0
 ( a 2 − ρ2)1/2 τ(ρ ,φ ) eiφ0 dρ dφ , for ρ> a .

ρ0e (ρ2 − a 2)1/2
−
q a 2(1 − t )  0 0 0 0 0

Exercise 2.7
1. Uniform shear tractions τ=τ0, where τ0 is a complex constant, are applied
antisymmetrically inside a penny-shaped crack of radius a . Find the tangential
displacements of the crack faces.
Answer: u = 2τ0[( G 21 − G 22)/ G 1]( a 2 − ρ2)1/2.
2. In the example above find the shear tractions in the plane z =0 outside the
crack.
G 2 a 3e2 φ
2 -1 a 
i
Answer: τ(ρ,φ) =  sin ( ) − 2

a τ − τ .
π  ρ (ρ − a 2)1/2 0 G 1ρ2(ρ2 − a 2)1/2 0
 
3. Subject to the conditions of the first example, find the normal displacement
w in the plane z =0.
Answer: w (ρ,φ) = π H αρℜτ0e φ + ( G 2/ G 1)τ0e φ, for ρ≤ a ;
-i i
 
w (ρ,φ) = 2 H αℜτ0e φ + ( G 2/ G 1)τ0e φρsin-1( ) − (ρ2 − a 2)1/2, for ρ≥ a .

-i i a a
  ρ ρ 
2.8 Inverse crack problem in elasticity
The usually considered elastic crack problems assume that the stress
distribution on the crack faces is known, and the crack opening displacements are
to be determined. Investigation of materials with rigid inclusions leads to another
formulation of the crack problem, namely, the displacements are prescribed on the
crack faces, and the stress distribution is to be determined. The problem so
formulated is called the inverse crack problem. We shall consider two types of
problem: the case of a smooth rigid inclusion (normal displacements are
prescribed, tangential stresses vanish) is called the inverse crack problem of type
I. The second type corresponds to the case when the normal stress is equal to
zero over the plane z =0, and the antisymmetric tangential displacements are
prescribed inside the crack. The stress distribution is to be determined in each
case. Strictly speaking, the inverse crack problem does not belong to the class
of the mixed problems, and its exact solution is known for a general crack.
We show below how some specific results can be obtained by the new method.
Smooth rigid inclusion problem. Consider a penny-shaped crack of radius
a in a transversely isotropic space. Let the crack be opened by a rigid smooth
inclusion. The crack opening displacements are prescribed as
w = w (ρ,φ), for z = 0+, 0≤ρ≤ a ;
-
w = − w (ρ,φ), for z = 0 , 0≤ρ≤ a ;
Due to symmetry, the problem can be reduced to the one of a half-space, with
the boundary conditions at the plane z =0
w = w (ρ,φ), for ρ≤ a , 0≤φ<2π;
w = 0, for ρ≥ a , 0≤φ<2π;
τ = 0, for 0≤ρ<∞, 0≤φ<2π. (2.8.1)
The general relationship between the crack opening displacement w and the
applied pressure σ was established in (2.4.19), and in this case may be written
as
ρ dρ ρρ
a x
w (ρ,φ) = 4 H ⌠
dx
⌠ 2
0 0
 0σ(ρ ,φ).
2 1/2 L
⌡ ( x 2 − ρ2)1/2 ⌡ ( x − ρ0)  x2  0
ρ 0
(2.8.2)
Since in our case the displacement w is known, and σ is unknown, we can
interpret (2.8.2) as an integral equation. It can be solved exactly. The first
operator to be applied is
L ,
d ⌠ ρ dρ 1
L( t )
d t ⌡ (ρ − t )
2 2 1/2
ρ
t
with the result
t
ρ0dρ0 ρ a
L σ(ρ0,φ) = −L( t ) ⌠ 2 1 
ρ dρ
2π H ⌠
0 d
2 1/2 L ρ w (ρ,φ).
⌡ ( t − ρ0)
2 2 1/2
 
t d t ⌡ (ρ − t )  
0 t
(2.8.3)
The next operator to be applied is
L 
1 d ⌠ t dt
2 1/2 L( t ).
 
y d y ⌡ (y − t )
2
0
The final result reads

y a
L  1 
1 1 d ⌠ t dt 2 d ⌠ ρ dρ
σ( y ,φ) = − 2 2 1/2 L( t )d t 2 1/2 L ρ w (ρ,φ).
π Hy  
y d y ⌡ (y − t )
2
⌡ (ρ − t )
2
 
0 t
(2.8.4)
Formula (2.8.4) is valid inside the crack only. The normal traction outside the
crack can be expressed directly in terms of the prescribed crack opening
displacement due to the relationship between the normal stresses inside and
outside the crack (2.4.17), which in this case takes the form
a
2 ⌠ ( a 2 − y 2)1/2 y 
σ(ρ,φ) = − L σ( y ,φ) y d y , for ρ> a .
π(ρ2 − a 2)1/2 ⌡ ρ2 − y 2 ρ
0
(2.8.5)
Substitution of (2.8.4) in (2.8.5) yields, after integration with respect to y ,
σ(ρ,φ) = 2 2 2
L(1/ρ) ⌠ a − t ρ2 − a 21/2d t
π H ρ (ρ − a 2)1/2 ⌡   ρ2 − t 2  
0
a
ρ0dρ0
d 2 d ⌠
L  w (ρ0,φ).
1
× t L( t )
dt d t ⌡ (ρ0 − t )
2 2 1/2
ρ0 
t
Integration by parts in the last expression leads to
a a
ρ0dρ0
L  w (ρ0,φ).
L(1/ρ) ⌠ xdx d ⌠ 1
σ(ρ,φ) = L ( x 2
)
π H ⌡ (ρ − x )
2 2 2 3/2 d x ⌡ (ρ0 − x )
2 2 1/2
ρ0
0 x
(2.8.6)
Expression (2.8.6) can also be represented in the form
a
ρ0dρ0
a
σ(ρ,φ) = −
L(1/ρ) d ⌠ xdx 2 d ⌠ 1
L ( x ) 2 1/2 L ρ w (ρ0,φ).
π Hρ
2 dρ ⌡ (ρ − x )
2 2 1/2 d x ⌡ (ρ0 − x )
2
 0
0 x
(2.8.7)
Comparison of (2.8.7) and (2.8.4) indicates that the stress outside the crack can
be represented in almost the same form as the stress inside, with the only
difference in the upper limit of the first integral. By using the rule (1.3.9) and
the condition w ( a ,φ)=0, expression (2.8.6) can be rewritten as
a a
ρ0dρ0
1 ⌠ dx ⌠ d   x 2 w(ρ ,φ).
σ(ρ,φ) = 2 ρ L
π H ⌡ (ρ2 − x 2)3/2 ⌡ (ρ20 − x 2)1/2 dρ0 0 ρρ0 0

0 x
(2.8.8)
Interchange of the order of integration and integration by parts yields
2π
a
1 ⌠⌠
σ(ρ,φ) = − .
4π2 H ⌡ ⌡ [ρ2 + ρ20 − 2ρρ0cos(φ−φ0)]3/2
0 0
(2.8.9)
The last expression can also be rewritten as
2π
a
σ(ρ,φ) = −
1
∆ ⌠⌠ .
4π2 H ⌡ ⌡ [ρ2 + ρ20 − 2ρρ0cos(φ−φ0)]1/2
0 0
(2.8.10)
Here ∆ is the two-dimensional Laplace operator. It will be shown later that
formula (2.8.10) is of general nature: it is valid for an arbitrarily shaped flat
crack all over the plane z =0.
We can also express some integral characteristics in terms of the crack
opening displacement. The resultant force P is defined by
2π a
P = ⌠ ⌠ σ(ρ,φ) ρdρdφ. (2.8.11)

⌡⌡
0 0
Substitution of (2.8.4) in (2.8.11) leads to
2π a a
1 ⌠ d ⌠ w (ρ,φ)ρ dρ
dφ ⌠
xdx
P = − 2 1/2. (2.8.12)
π2 H ⌡ ⌡ ( a 2
− x 2 1/2 d x
) ⌡ (ρ − x )
2
0 0 x
Integration by parts in (2.8.12) yields yet another representation
2π a a
P = 2 ⌠ dφ ⌠
a 2
dx ⌠ w (ρ,φ)ρ dρ
. (2.8.13)
πH ⌡ ⌡ (a 2
− x 2)3/2 ⌡ (ρ2 − x 2)1/2
0 0 x
Several additional forms can be obtained by integration in (2.8.13) with respect

to x , and a consequent use of various integral representations for the complete

elliptic integrals (see Exercise 2.8.2).
One can also evaluate the resultant moment of the traction exerted by the
rigid inclusion. Introducing the complex moment M = M +i M , we can deduce
x y
2π a
M = −i⌠ ⌠ σ(ρ,φ)e ρ2 dρdφ.

iφ
(2.8.14)
⌡⌡
0 0
2π a a
w (ρ,φ) dρ
M = 2 ⌠ e dφ ⌠
i iφ x 3d x d ⌠
. (2.8.15)
πH ⌡ ⌡ (a − x )
2 2 1/2 d x
⌡ (ρ2 − x 2)1/2
0 0 x
Yet another representation can be obtained from (2.8.15) (see Exercise 2.8.3).
Example. It is of interest to consider a general case where the
displacements can be presented as an expansion
∞
w (ρ,φ) = ( a − ρ ) 2 2 1/2
Σw ρ
n=-∞
n
|n| in φ
e . (2.8.16)
Substitution of (2.8.16) in (2.8.4) and (2.8.7) yields

∞
σ(ρ,φ) =
1
2√πH Σ
n=-∞
Γ(| n | + 3/2)
Γ(| n | + 1)
φ
w nρ|n|ein , for ρ≤ a .
∞
a 2|n|+3 w nein φ
σ(ρ,φ) = −
1
2π H Σ
n=-∞
(2| n | + 3)ρ|n|+3
F  , +| n |; +| n |; 2,
3 3
2 2
5
2
a2
ρ
for ρ> a .
(2.8.17)
The Gauss hypergeometric function can be expressed in terms of elementary
functions (Bateman and Erdélyi, 1955)
(−1)n(3 + 2 n ) dn (1−ζ)n  1−ζ1/2sin-1√ζ.

F  , + n ; + n ; ζ =
3 3 5
2 2   ζ 
1 −
 ζ  
2 n ! (1 − ζ)1/2 dζn
 
(2.8.18)
The traction in (2.8.17) becomes singular when ρ→ a . +
Using integration by
parts in (2.8.6), the singular and nonsingular parts can be separated as follows:
∞
σ(ρ,φ) = −
1
2π H Σn=-∞
a 2|n|+1  ρ
ρ|n|+1 (ρ2 − a 2)1/2
− F 1, 1+| n |; 3+| n |; a 2w ein φ,
2 2 2 ρ2 n
(2.8.19)
which yields the stress intensity factor as
∞
k1 =
√a
2√2π H Σw a
n=-∞
n
|n| in φ
e . (2.8.20)
The hypergeometric function in (2.8.19) can be expressed in elementary functions:
F  , +| n |; +| n |; ζ = (−1)n
dn  (1 − ζ)n-1/2
sin-1√ζ.
1 1 3 2n + 1
√1 − ζ
2 2 2  n! dζ 
n
√ζ 
(2.8.21)
The total force P and the tilting moment M can be determined from (2.8.12)
and (2.8.15) respectively
πa2
P = w,
4H 0
3π a 4
M = −i w
16 H -1
Inverse crack problem of type II. Let the tangential displacements be
prescribed on the crack faces
u = u (ρ,φ), for ρ≤ a z = 0+ ;
-
u = − u (ρ,φ), for ρ≤ a z = 0. (2.8.22)
It is required to find the shear stress distribution in the plane z =0. Due to
antisymmetry, the problem can be reduced to that of a half-space, subjected to
the boundary conditions:
u = u (ρ,φ), for ρ≤ a , 0≤φ<2π;
u = 0, for ρ≥ a , 0≤φ<2π;
σ = 0, for 0≤ρ<∞, 0≤φ<2π. (2.8.23)
The relationship between the tangential displacement u and the shear traction τ
was established in (2.7.39) and (2.7.43). Since in the inverse problems u is

known, and τ is to be determined, we should treat these expressions as a set of
integral equations to be solved for the two unknowns τn+1 and τ-n+1. Assume
the solution in the form
ρ
τn+1(ρ) = G 1 f n+1(ρ) + G 2  f -n+1(ρ) − n+1 ⌠ f -n+1( x ) x nd x , for n = 0, 1, 2 ...
2n
 ρ ⌡ 
0
ρ
τ-n+1(ρ) = G 1 f -n+1(ρ) + G 2  f n+1(ρ) + 2 n ρn-1 ⌠ n+1n d x  + C nρn-1, for n = 1, 2, 3 ...
f ( x)
 ⌡ x 
0
(2.8.24)
Here f is an as yet unknown complex function and C n is an as yet unknown
complex constant. Substitution of (2.8.24) in equation (2.7.39) and subsequent
integration results in
a x
⌠ 2n+2 d2x 2 1/2 ⌠ f n+1(ρ2 0)ρ20 1/2dρ0
n+2
u n+1(ρ) = 2( G 21 − G 22)ρn+1 (2.8.25)
⌡ x ( x − ρ ) ⌡ ( x − ρ0)
ρ 0
from which the value of f n+1 can be found as
ρ a
2 d ⌠ x 2n+3d x d ⌠ u n+1(ρ0) dρ0
f n+1(ρ) = − 2 2 (2.8.26)
π ( G 1 − G 22)ρn+2 dρ⌡ (ρ2 − x 2)1/2 d x⌡ ρ0n (ρ20 − x 2)1/2
0 x
Substitution of (2.8.24) in equation (2.7.43) gives, after simplification
a x
⌠ f -n+1(2ρ0)ρ20 1/2dρ0 + 2 D nρn-1( a 2 − ρ2)1/2

n
u -n+1(ρ) = 2( G 21 − G 22)ρn-1 ⌠ dx
⌡ x ( x − ρ ) ⌡ ( x − ρ0)
2n-2 2 2 1/2
ρ 0
(2.8.27)
Here D n is a complex constant defined by
a x
d x f (ρ) ρ dρ
n
D n = G 2 2 nG 1 ⌠ 2n ⌠ n+12 2 1/2
 ⌡ x ⌡ (x − ρ )
0 0
+ 2 nG 2 ⌠ f -n+1(ρ)
ρn -1 a  √π Γ( n )
2n cosh (ρ) dρ + B n + G 1 C n 2Γ( n + 1/2) (2.8.28)
⌡ a 
0
Now the purpose of introducing the constant C n in (2.8.24) becomes clear: we

can choose C n so that D n vanish. In this case expression (2.8.27) can be
inverted, with the result
ρ a
2 d ⌠ x 2n-1d x d ⌠ u -n+1(ρ0) dρ0
f -n+1(ρ) = − 2 2 (2.8.29)
π ( G 1 − G 22)ρn dρ⌡ (ρ2 − x 2)1/2 d x⌡ ρn-2
0 (ρ0 − x )
2 2 1/2
0 x
We can substitute (2.8.26) and (2.8.29) in (2.8.28) in order to find the value of
the as yet unknown constant C n from the condition D n = 0. The mathematical
transformations involved are cumbersome though elementary. Some of the
integrals are presented below.
a x a a
f (ρ) ρn dρ 2 n + 1⌠ u n+1(ρ) dρ − √π Γ( n ) ⌠ u (ρ) ρn-1dρ,
⌠ d2nx ⌠ n+12 2 1/2 = 1
⌡ x ⌡ (x − ρ ) π( G 1 − G 2)  2 n ⌡ ρn+1 2 a 2n Γ( n + 1/2) ⌡ 
2 2 n+1
0 0 0 0
(2.8.30)
a a
⌠ f -n+1(ρ) ρncosh-1(a ) dρ = √π Γ( n ) ⌠ u -n+1 (ρ) ρn-1dρ. (2.8.31)

⌡ ρ π( G 1 − G 2) Γ( n − 1/2)⌡
2 2
0 0
Substitution of (2.8.24) in (2.7.44) gives the following expression for B n
1 ⌠
a
 G 22
B n = 2n+1 − 2 nG 2 f -n+1(ρ) + G 1 − (2 n + 1) f (ρ)
a ⌡
  G 1 n+1
0
ρ ρ
2 nG 2 G 22 n-1 fn+1( x )  2 2 1/2 n
− n+1 x f -n+1( x ) d x − 2 n (2 n + 1) ρ ⌠
⌠ n
d x  ( a − ρ ) ρ dρ
ρ ⌡ G1 ⌡ x
n
0 0 
(2.8.32)
The following integrals need to be computed

a a
⌠ f -n+1(ρ)ρ ( a − ρ ) dρ = 1 √π Γ( n + 1/2)⌠
n 2 2 1/2
u -n+1(ρ)ρndρ (2.8.33)
⌡ π( G 1 − G 2)
2 2 Γ ( n ) ⌡
0 0
a a
⌠ f n+1(ρ)ρ ( a − ρ ) dρ = a √π Γ( n + 1)⌠
n 2 2 1/2
2 Γ( n + 1/2) u n+1(ρ)ρn-1dρ (2.8.34)
⌡ π( G 1 − G 2)
2
⌡
0 0
a ρ a
⌠ ( a 2 − ρ2)1/2dρ⌠ f -n+1( x ) x nd x = 1  a √π Γ( n ) ⌠ u (ρ)ρn-1dρ

⌡ ρ⌡ π( G 1 − G 2) Γ( n − 1/2) ⌡
2 2 -n+1
0 0 0
u -n+1(ρ)ρndρ
√π Γ( n + 1/2) ⌠
− (2.8.35)
Γ( n ) ⌡ 
0
a ρ a
⌠ρ 2n-1
(a − ρ )
2 2 1/2
dρ⌠
f n+1( x )
dx =
1 a 2n+1⌠ u n+1(ρ) dρ −
⌡ ⌡ xn π( G 21 − G 22)  2 n ⌡ ρn+1
0 0 0
u n+1(ρ)ρn-1dρ
a √π Γ( n ) ⌠
− (2.8.36)
2Γ( n + 1/2)⌡ 
0
Substitution of (2.8.33-2.8.36) in (2.8.32) yields
a a
B n = 2n 2
1 − 2 G √π Γ( n + 1)⌠ u (ρ)n-1dρ + G √π Γ( n + 1) ⌠ u (ρ)ρn-1dρ
π a ( G 1 − G 2)  2 Γ( n − 1/2)
⌡
1 Γ( n + 1/2)
-n+1
⌡
2 n+1
0 0
2a
2n G 2 (ρ)  G √π Γ( n )
− (2 n + 1) a ⌠ u n+1n+1 dρ − C n 2 . (2.8.37)
G 1⌡ ρ  G 1 2Γ( n + 1/2)
0
Substitution of (2.8.30), (2.8.31) and (2.8.37) in (2.8.28) gives a fairly simple

formula for C n, namely,
a
4 G 2Γ( n + 3/2) u n+1(ρ) dρ
Cn = − ⌠ (2.8.38)
π 3/2
( G 21 − G 22) Γ( n ) ⌡ ρn+1
0
The problem, in principle, is now solved but a certain simplification is possible

due to the following integral:
ρ ρ
f n+1( x ) 1
⌠ n dx = − 2 2
2
2  2n
⌠ 2 d x 2 1/2
⌡ x π ( G 1 − G 2) ρ ⌡ (ρ − x )

0 0
d  2n+1⌠ u n+1(ρ0) dρ0  √π Γ( n + 3/2)⌠ u n+1(ρ) dρ

a a
Γ( n + 1) ⌡ ρn+1 
× x 2 1/2 − (2.8.39)
dx  ⌡ ρ0 (ρ0 − x ) 
n 2
x 0 
Note that the last term in (2.8.39) will cancel C n when substituted in (2.8.24).
Finally, formulae (2.8.26), (2.8.29) and (2.8.38) lead to
ρ
 G 1 d x 2n+3 d x d u n+1(ρ0) dρ0 a
τn+1(ρ) = − 2 2
2
 ⌠ ⌠
π ( G 1 − G 22) ρn+2 dρ⌡ (ρ2 − x 2)1/2 d x ⌡ ρ0n (ρ20 − x 2)1/2
 0 x
ρ
d  1 ⌠ x 2n-1d x d ⌠ u -n+1(ρ0)dρ0 
a
+ G 2ρ 2 1/2 
n
dρ ρ2n⌡ (ρ2 − x 2)1/2 d x ⌡ ρn-2
0 (ρ0 − x ) 
2
0 x 
for ρ ≤ a , n = 0, 1, 2, ... (2.8.40)
ρ a
τ-n+1(ρ) = − 2 2
2 G 1 d ⌠ x 2n-1 d x d ⌠ u -n+1(ρ0) dρ0
π ( G 1 − G 22)  ρn dρ⌡ (ρ2 − x 2)1/2 d x ⌡ ρn-2
0 (ρ0 − x )
2 2 1/2
0 x
ρ a
u n+1(ρ )dρ0 
+ n2 ⌠ 2 2 1/2  x 2n+1⌠ n 2 0 2 1/2
G d dx d
ρ dρ ⌡ ( ρ − x ) d x  ⌡ ρ0 (ρ0 − x ) 
0 x
for ρ ≤ a , n = 1, 2, 3, ... (2.8.41)
Expressions (2.8.40) and (2.8.41) are valid inside the crack only. In order to
express the shear traction outside the crack, we recall formulae (2.7.23) and
(2.7.27) which relate one to the other. Substitution of (2.8.41) in (2.7.27) allows
us to compute
a
4 G 2Γ( n + 3/2)
An = − ⌠ u -n+1(ρ) ρndρ (2.8.42)
π3/2( G 21 − G 22) Γ( n )⌡
0
Finally, substitution of (2.8.40-2.8.42) in (2.7.23) yields
a a
τn+1(ρ) = − 2 2
2  G 1 d ⌠ x 2n+3 d x d ⌠ u n+1(ρ0) dρ0
π ( G 1 − G 22) ρn+2 dρ⌡ (ρ2 − x 2)1/2 d x ⌡ ρ0n (ρ20 − x 2)1/2
0 x
a a
G2 d
⌠ x 2n-1d x d ⌠ u -n+1(ρ0)dρ0 
+ n
0 (ρ0 − x ) 
ρ dρ⌡ (ρ2 − x 2)1/2 d x⌡ ρn-2 2 2 1/2
0 x
for ρ > a , n = 0, 1, 2, ... (2.8.43)
a a
τ-n+1(ρ) = − 2 2
2 G 1 d ⌠ x 2n-1 d x d ⌠ u -n+1(ρ0) dρ0
π ( G 1 − G 22)  ρn dρ⌡ (ρ2 − x 2)1/2 d x⌡ ρn-2
0 (ρ0 − x )
2 2 1/2
0 x
a a
u n+1(ρ )dρ0 
+ n2 ⌠ 2 2 1/2  x 2n+1⌠ n 2 0 2 1/2
G d dx d
ρ dρ⌡ (ρ − x ) d x  ⌡ ρ0 (ρ0 − x ) 
0 x
for ρ > a , n = 1, 2, 3, ... (2.8.44)
Formulae (2.8.43) and (2.8.44) can be used to define the stress intensity factor
directly in terms of the tangential displacements prescribed inside the crack. We
recall that the stress intensity factor was given by (2.7.31). By using the
properties
ρ
lim (ρ − a )1/2 ⌠ 2 2 1/2 = −
d f( x) dx f(a)
,
ρ→ a  dρ⌡ (ρ − x )  √2 a
0
a
d ⌠ f( x) dx π f (ρ)
lim
ρ 2 1/2 = − 2 lim 2 1/2 , (2.8.45)
ρ→ a d ⌡ (x − ρ )
2
ρ→ a ( a − ρ )
2
ρ
the appropriate harmonics of the stress intensity factor can be written as
k n+1 = −
a
lim G 1 u n+1(ρ) + G 2 u -n+1(ρ),
π( G 21 − G 22)√2 a ρ→a  ( a 2 − ρ2)1/2 
G u -n+1(ρ) + G 2 u n+1(ρ)
lim  1
a
k -n+1 = − .
π( G 21 − G 22)√2 a ρ→ a  ( a 2 − ρ2)1/2 
Their summation can be performed in an elementary manner, to give finally
a G 1 u (ρ, φ) + G 2e2 i φ u (ρ, φ)

k (φ) = − lim . (2.8.46)
π( G 21 − G 22)√2 a ρ→a  ( a 2 − ρ2)1/2 
The simple expression (2.8.46) for the stress intensity factor will be very useful
in the investigation of crack interaction since it is much easier to solve the
integral equation involved in terms of the displacements than in terms of the
stresses which are singular at the crack boundary. Expression (2.8.46) can be
made more symmetric by introduction of the polar displacements
u (ρ) = u ρ + iu φ = e-i φ u , and similar shear traction τ(ρ) = τρz + i τφz = e-i φτ. The
corresponding stress intensity factor k (ρ) will take the form
(ρ) a G 1 u (ρ)(ρ, φ) + G 2 u (ρ)(ρ, φ)

k =− lim . (2.8.47)
π( G 21 − G 22)√2 a ρ→a  ( a 2 − ρ2)1/2 
Note that k (ρ) is proportional to the combination k 2 + ik 3 of the second and the
third mode stress intensity factors.
Summation of (2.8.40-2.8.41) and (2.8.43-2.8.44) leads to another simple
result
τ=−
1  ⌠ ⌠ u d S + G Λ2⌠ ⌠ u d S .
2 G 1∆ (2.8.48)
2π ( G 1 − G 2) 
2 2
⌡⌡R 2
⌡⌡R 
S S
It will be shown further, that the last expression is valid not just for a
penny-shaped crack, but for a general flat crack all over the plane z = 0.
Exercise 2.8
1. The crack opening displacement w is prescribed by expression

w = w 0( a 2 − ρ2)1/2, with w 0 = const. Find the normal stress distribution σ in the

plane z = 0.
Answer: σ = w 0/(4 H ), for ρ ≤ a ;
w0 
− sin-1( ) , for ρ > a
a a
σ=−
2π H ( a 2 − ρ2)1/2 ρ
2. Prove that the total force P exerted by an inclusion can be defined by
2π a a
w (ρ, φ)ρ dρ
P = 2 ⌠ dφ⌠ ( a 2 − x 2)1/2d x ⌠ 2 2 2 2 1/2.
1
π H⌡ ⌡ ⌡ ( a − ρ )(ρ − x )
0 0 x
Hint : use (2.8.13)
3. Prove that the tilting moment M can be defined by
2π a a
x 2(3 a 2 − 2 x 2) ⌠ w (ρ, φ)dρ
M = − 2 ⌠ ei φdφ ⌠
i
2 3/2 d x 2 1/2.
π H⌡ ⌡ (a − x )
2
⌡ (ρ − x )
2
0 0 x
Hint : integrate (2.8.15) by parts.

CHAPTER 3
MIXED-MIXED BOUNDARY VALUE PROBLEMS
The mixed-mixed problems in elasticity theory are among the most

complicated due to the coupling between the normal and tangential parameters.
We should mention the works of Mossakovskii (1954) and Ufliand (1956) among
the first published exact solutions for the isotropic half-space, obtained by using
various integral transforms. A more compact solution has been reported by
Kapshivyi and Masliuk (1967), who used a special apparatus of p -analytical
functions. The first elementary exact solution for a transversely isotropic elastic
half-space was published in (Fabrikant, 1971a). We present here a general
formulation of the internal and external mixed-mixed problems in terms of
two-dimensional integral equations. Four kinds of exact solution to the internal
axisymmetric problem are given, and yet another kind of solution is presented for
the external axisymmetric problem. The action of a general loading on a flat
circular bonded punch is considered in detail. A general solution to the
non-axisymmetric internal and external problems is presented as a Fourier series
expansion. The material in this chapter follows the results published in the
papers (Fabrikant, 1971d, 1972, 1974b, 1975, 1976, 1986j).
3.1 General formulation of the problem
Consider a transversely isotropic elastic half-space z ≥0. Let the following

boundary conditions be prescribed on the plane z =0:
u = u (ρ,φ), for 0≤ρ≤ a , 0≤φ<2π;
w = w (ρ,φ), for 0≤ρ≤ a , 0≤φ<2π;
σ = σ(ρ,φ), for a ≤ρ≤∞, 0≤φ<2π;
142
3.1 General formulation of the problem 143
τ = τ(ρ,φ), for a ≤ρ≤∞, 0≤φ<2π. (3.1.1)
The problem, so defined, will be called internal mixed-mixed . The system of

governing integral equations is formulated due to (2.2.12) and (2.2.13), and is
2π a τ(ρ ,φ ) ρ dρ dφ 2π a σ(ρ0,φ0) ρ0dρ0dφ0

H αℜ⌠ ⌠ + H⌠ ⌠
0 0 0 0 0
iφ
= ω1(ρ,φ), (3.1.2)
⌡ ⌡ iφ ⌡ ⌡ R
0 0 ρe − ρe 0
0 0
0
2π a τ(ρ0,φ0) ρ0dρ0dφ0 2π a q τ(ρ0,φ0) ρ0dρ0dφ0

1 ⌠ ⌠
+ G 2⌠ ⌠
1
G
2 1⌡ ⌡ R 2 ⌡ ⌡ qR
0 0 0 0
2π a σ(ρ0,φ0) ρ0dρ0dφ0
− H α⌠ ⌠ -i φ
= ω2(ρ,φ). (3.1.3)
⌡ ⌡ -i φ
0 0 ρe − ρ0e 0
It is reminded that the notations q and R are defined by (2.2.5) and (2.2.14)
respectively. Functions ω1 and ω2 are known from the boundary conditions
(3.1.1):
2π ∞ τ(ρ0,φ0) ρ0dρ0dφ0
ω1(ρ,φ) = w (ρ,φ) − H αℜ⌠ ⌠ iφ
⌡ ⌡
0 a ρei φ − ρ0e 0
2π ∞ σ(ρ0,φ0) ρ0dρ0dφ0
− H⌠ ⌠ , (3.1.4)
⌡ ⌡ R
0 a
ω2(ρ,φ) = u (ρ,φ) + H α⌠ ⌠ -i φ
⌡ ⌡ φ
0 a ρe-i − ρ0e 0
2 π ∞ τ( ρ , φ ) ρ0dρ0dφ0 2 π ∞ q τ( ρ , φ ) ρ0dρ0dφ0
− G1⌠ ⌠ G⌠ ⌠
1 0 0 1 0 0
− .
2 ⌡ ⌡ R 2 2⌡ ⌡ qR
0 a 0 a
(3.1.5)
The external mixed-mixed boundary value problem for a transversely isotropic
144 CHAPTER 3 MIXED-MIXED BOUNDARY VALUE PROBLEMS
elastic half-space can be formulated in a similar manner. The boundary

conditions are:
u = u (ρ,φ), for a ≤ρ<∞, 0≤φ<2π;
w = w (ρ,φ), for a ≤ρ<∞, 0≤φ<2π;
σ = σ(ρ,φ), for 0≤ρ≤ a , 0≤φ<2π;
τ = τ(ρ,φ), for 0≤ρ≤ a , 0≤φ<2π. (3.1.6)
The system of governing integral equations in this case takes the form
2π ∞ τ(ρ0,φ0) ρ0dρ0dφ0 2π ∞ σ(ρ0,φ0) ρ0dρ0dφ0

H αℜ⌠ ⌠ iφ
+ H⌠ ⌠ = ω1(ρ,φ), (3.1.7)
⌡ ⌡ iφ ⌡ ⌡ R
0 a ρe − ρ0e 0
0 a
2π ∞ τ(ρ0,φ0) ρ0dρ0dφ0 2π ∞ q τ(ρ0,φ0) ρ0dρ0dφ0

1 ⌠ ⌠
+ G 2⌠ ⌠
1
G
2 1⌡ ⌡ R 2 ⌡ ⌡ qR
0 a 0 a
− H α⌠ ⌠ -i φ
= ω2(ρ,φ). (3.1.8)
⌡ ⌡ -i φ
0 a ρe − ρ0e 0
Functions ω1 and ω2 are known from the boundary conditions, and are
2π a τ(ρ0,φ0) ρ0dρ0dφ0
ω1(ρ,φ) = w (ρ,φ) − H αℜ⌠ ⌠ iφ
⌡ ⌡ iφ
0 0 ρe − ρ0e 0
− H⌠ ⌠ , (3.1.9)
⌡ ⌡ R
0 0
ω2(ρ,φ) = u (ρ,φ) + H α⌠ ⌠ -i φ
⌡ ⌡ -i φ
0 0 ρe − ρ0e 0
2 π a τ( ρ , φ ) ρ0dρ0dφ0 2 π a q τ( ρ , φ ) ρ0dρ0dφ0
− G1⌠ ⌠ G⌠ ⌠
1 0 0 1 0 0
− .
2 ⌡ ⌡ R 2 2⌡ ⌡ qR
0 0 0 0
(3.1.10)
The exact closed form solution of these equations is not known at the moment,
though we strongly believe that the new method is capable of furnishing such a
solution. In the sections to follow we consider separately the axisymmetric case
and the general one. The exact solution to both internal and external problems
is obtained for the n -th harmonic, with the assumption that all the functions
involved can be represented as Fourier expansions.
Exercise 3.1
1. Verify the derivation of (3.1.2−3.1.5)
2. Verify the derivation of (3.1.7−3.1.10)
3. Derive the governing integral equations for the internal mixed-mixed problem
in the case of an isotropic half-space.
3.2 Internal axisymmetric mixed-mixed problem
In order to demonstrate versatility of our method, we present here four

kinds of exact solution. The first kind is more convenient for the stress
evaluation, while the second one has certain advantages for calculating the
displacements outside the circle ρ= a . Since it is important to have one kind of
solution easily transformed into another, the necessary relationships are established.
Application of this technique to a bonded flat-ended circular punch under the
action of a normal force and expanding in the radial direction is considered.
The influence of an arbitrary axisymmetric normal and tangential tractions field,
applied outside the punch, is investigated.
The boundary conditions in the case of axial symmetry are
u = u (ρ), for 0≤ρ≤ a , 0≤φ<2π;
w = w (ρ), for 0≤ρ≤ a , 0≤φ<2π;
σ = σ(ρ), for a ≤ρ≤∞, 0≤φ<2π;
τ = τ(ρ), for a ≤ρ≤∞, 0≤φ<2π. (3.2.1)

The set of governing integral equations will take the form
ρ a σ(ρ )ρ dρ
 0 
a
2 H −πα ⌠ τ(ρ0)dρ0 + 2⌠ ⌠
d x 0 0
 = ω1(ρ),
 ⌡ ⌡ (ρ2 − x 2)1/2 ⌡ (ρ20 − x 2)1/2

ρ 0 x
(3.2.2)
ρ τ(ρ0) dρ0 ρ
2H 
a
x 2d x
⌠ ⌠ ⌠
ρ  1 2⌡ (ρ2 − x 2)1/2 ⌡ (ρ20 − x 2)1/2  = ω2(ρ).
2γ γ − πα σ( ρ ) ρ dρ
 ⌡ 0 0 0

0 x 0
(3.2.3)
The functions ω1 and ω2 are known from the boundary conditions, and are
defined by
∞ ∞ x σ(ρ )ρ dρ
 0 
ω1(ρ) = w (ρ) + 2 H πα ⌠ τ(ρ0)dρ0 − 2⌠ ⌠
d x 0 0
,
⌡ ⌡ ( x 2 − ρ2)1/2 ⌡ ( x 2 − ρ20)1/2
 a a a 
(3.2.4)
∞
dx
x τ(ρ0) ρ20dρ0
ω2(ρ) = u (ρ) − 4 H γ1γ2ρ⌠ ⌠ 2 1/2. (3.2.5)
⌡ x 2 2
( x − ρ2 1/2
) ⌡ ( x − ρ0)
2
a a
Multiply both sides of equation (3.2.2) by ρ( r 2−ρ2)-1/2dρ, integrate with respect to

ρ from zero to r and differentiate with respect to r . The result is
a r τ(ρ )dρ a σ(ρ )ρ dρ

2 ⌠  = χ ( r).
−α τ(ρ0)dρ0 + α r ⌠ 2 ⌠
0 0 0 0 0
+ (3.2.6)
π ⌡ ⌡ ( r − ρ2 1/2
0 ) ⌡ 0
( ρ2
− r 2 1/2
)  1
0 0 r
Here
r ω (ρ)ρ dρ
1 d ⌠ 1
χ1( r ) = 2 . (3.2.7)
π H d r ⌡ ( r 2 − ρ2)1/2
0
A similar transformation can be applied to equation (3.2.3), with the result
a τ(ρ0)dρ0 α
r σ(ρ )ρ dρ
2  = χ ( r),
√γ1γ2r ⌠ ⌠
0 0 0
− (3.2.8)
π ⌡ ( ρ2
0 − r 2 1/2
) √γ γ ⌡ ( r 2
− ρ 2 1/2
0 )  2
r 2 2 0
where
r ω2(ρ)ρ2dρ
1 1 d ⌠
χ2( r ) = 2 . (3.2.9)
π H √γ1γ2 r d r ⌡ ( r 2 − ρ2)1/2
0
The set of modified integral equations (3.2.6) and (3.2.8) is solved below by
four different methods. It is shown that all the solutions are consistent with
each other, and give effectively the same solution.
Solution of the first kind. Assume the solution to the set of equations
(3.2.2) and (3.2.3) to be defined by
ρ f 1( t ) t d t ρ f 2( t ) d t
1 d ⌠ d ⌠
σ(ρ) = , τ(ρ) = (γ1γ2) -1/2
.
ρ dρ ⌡ (ρ2 − t 2)1/2 dρ ⌡ (ρ2 − t 2)1/2
0 0
(3.2.10)
Here f 1 and f 2 are the as yet unknown stress functions. Substitution of (3.2.10)
in (3.2.6) yields, after integration,
a f 2( t ) t d t
α 2 2 2 1/2 ⌠
f ( r) + ( a − r ) 2 1/2 = χ1( r ) + b .
√γ1γ2 2 π ⌡ ( t − r )( a − t )
2 2 2
0
(3.2.11)
a f 2( t ) d t
2α ⌠
b = . (3.2.12)
π√γ1γ2 ⌡ ( a 2 − t 2)1/2
0
It will be shown later that we may assume b =0, without loss of generality.
Hereafter the following identities are used
r a a
⌠ 2 d x 2 1/2 d ⌠ 2f ( t ) d2t 1/2 = r ⌠ f 2( t ) d t 2 ,

⌡ (r − x ) dx ⌡ (t − x ) ⌡ t(t − r )
0 x 0
a x a
⌠ 2 d x 2 1/2 d ⌠ f(t) dt
= ( a 2 − r 2)1/2⌠
f(t) dt
2 1/2,
⌡ (x − r ) dx ⌡ (x − t ) ⌡ ( t − ρ )( a − t )
2 2 1/2 2 2 2
r 0 0
a a
⌠ dx d ⌠ f(t) dt π f ( r)
2 1/2 = − 2 ,
⌡ (x − r ) ⌡ (t − x )
2 2 1/2 d x 2 r
r x
r x
= 
⌠ 2 d x 2 1/2 d ⌠ f(t) dt π f ( r ) − f (0)
. (3.2.13)
⌡ (r − x ) dx ⌡ (x − t )
2 2 1/2 2 r 
0 0
The identities (3.2.13) can be verified by using (1.3.2) and (1.3.9). Now
substitution of (3.2.10) in (3.2.8) gives, after simplification,
a f 2( t ) d t
α
r ( a 2 − r 2)1/2⌠
2
− f 1( r ) + = χ2( r ).
√γ γ π ⌡ ( t 2 − r 2)( a 2 − t 2)1/2
1 2 0
(3.2.14)
Let
f 1( t ) = − f 1(− t ), f 2( t ) = f 2(− t ). (3.2.15)
These assumptions allow us to rewrite equations (3.2.11) and (3.2.14) in the form
a
f 1( t ) d t
α
f ( r) + ( a − r ) ⌠
1 2
2 1/2 = χ1( r ) + b ,
2 1/2
√γ1γ2 2 π ⌡ ( t − r)( a − t )
2
-a
a
f 2( t ) d t
α 1 2 ⌠
− f ( r) + ( a − r )
2 1/2
2 1/2 = χ2( r ).
√γ1γ2 1 π ⌡ ( t − r)( a − t )
2
-a
(3.2.16)
By introducing the complex functions f = f 1+ if 2 and χ=χ1+ i χ2, the system depicted
in (3.2.16) can be reduced to one singular integral equation, namely,
a
α f(t) dt
f ( r ) + ( a 2 − r 2)1/2 ⌠
1
−i 2 1/2 = χ( r ).
√γ1γ2 π ⌡ ( t − r)( a − t )
2
-a
(3.2.17)
-i θ iθ
Multiply both sides of equation (3.2.17) by ( a + r ) ( a − r ) ( r − y ) d r , where θ is an -1
as yet unknown constant, and integrate with respect to r from − a to a . The

result is
a
-i θ -i θ
−i
α ⌠ a + r f ( r) d r − πf ( y )a + y 
√γ1γ2 ⌡ a − r  r − y a − y 
-a
a
a + y -i θ
− i(a − y )  tanh(πθ) ⌠
2 2 1/2 f(t) dt
a − y  ⌡ ( t − y )( a − t )
2 2 1/2
-a
a a
iθ -i θ
+ i tanh(πθ) ⌠ 
a + t f(t) dt ⌠ a + r χ( r) d r .
=
⌡ a − t  t − y ⌡ a − r  r − y
-a -a
(3.2.18)
Defining
tanh(πθ) = α/√γ γ , (3.2.19)
1 2
equation (3.2.18) may be simplified as follows:
a
-i θ
−π
a + y f ( y ) + i tanh(πθ)( a 2 − y 2)1/2 ⌠ f(t) dt 
a − y   π ⌡ ( t − y )( a − t ) 
2 2 1/2
-a
a
-i θ
= ⌠ a + r χ( r) d r .
⌡ a − r  r − y
-a
(3.2.20)
The singular integral may be eliminated from (3.2.20) by using (3.2.17), and the
exact solution becomes available in the form
a
1 a + y i θ ⌠ a + r -i θ χ( r ) d r 
f ( y ) = −cosh (πθ) i tanh(πθ)χ( y ) + 
2
.
 πa − y  ⌡ a − r  r − y
-a
(3.2.21)
The following rule of interchanging the order of integration in singular integrals
was employed (Muskhelishvili, 1946)
a a a a
⌠ dr ⌠ f ( r, t )d t = −π2 f ( y , y ) + ⌠ dt ⌠ f ( r , t )d r
. (3.2.22)
⌡ r − y ⌡ t − r ⌡ ⌡ ( r − y )( t − r)
-a -a -a -a
The other integrals used here can be found in Appendix A3.1. Strictly speaking,
the complete solution of (3.2.17) is given by (3.2.21) plus the term
c ( a + y )i θ( a − y )-i θ, which represents the homogeneous solution of (3.2.17), with c
being an arbitrary constant. The value of c is to be chosen to satisfy the
condition b =0, where b is defined by (3.2.12). The appropriate integration of
expression (3.2.21) shows that the condition b =0 is satisfied when c =0, and this
is the reason why the final solution is written in the form (3.2.21).
The general solution is now completed, and we can consider in more detail
the case of a bonded axisymmetric punch, with no tractions applied outside the
punch. The total force P may be obtained by integration
a a f 1( t ) t d t
P = 2π⌠ σ(ρ)ρdρ = 2π⌠ 2 1/2 . (3.2.23)
⌡ ⌡ (a − t )
2
0 0
We recall that f 1 is an odd function, and that f 1=ℜ f . Taking this into
consideration, we obtain, after substitution of (3.2.21) in (3.2.23)
P = πcosh(πθ)ℜ ⌠ a + r-iθ χ( r)d r. (3.2.24)

⌡ a − r 
-a
The field of displacements outside the punch can be obtained by repeating the
derivation of (3.2.2) and (3.2.3) for ρ> a , which results in
a a
σ( x )d x
w (ρ) = 4 H ⌠ ⌠
dy
, for ρ> a ;
⌡ ρ − ⌡ ( x − y 2)1/2
2 2 1/2 2
( y )
0 y
a a
2H y 2d y τ( x )d x α 
u (ρ) = 2γ 1 γ 2 ⌠ ⌠ 2 1/2 − 2 P , for ρ> a .
ρ ⌡ ( ρ2
− y 2 1/2
) ⌡ (x − y )
2

0 y
(3.2.25)
Substitution of (3.2.10) in (3.2.25) and integration with respect to x leads to
a a f (t)tdt
⌠ ( a 2
− y 2 1/2
) ⌠
1
w (ρ) = 4 H d y 2 1/2 ,
⌡ (ρ − y ) ⌡ ( t − y )( a − t )
2 2 1/2 2 2 2
0 0
a a f ( t )d t
2H ( a 2 − y 2)1/2 y 2d y ⌠ α 
2√γ γ ⌠
2
u (ρ) = 2 1/2 − 2 P .
ρ  1 2⌡ (ρ − y )
2 2 1/2
⌡ ( t − y )( a − t )
2 2 2

0 0
(3.2.26)
The singular integrals in (3.2.26) can be evaluated from (3.2.17) and (3.2.21),
and the final result may be written as
a
ℜ{ Z ( y )}
w (ρ) = 2 H ⌠ 2 1/2 d y ,
⌡ (ρ − y )
2
0
u (ρ) = √γ γ 2⌠ d y − P tanh(πθ), for ρ> a .

H y ℑ{ Z ( y )}
ρ 1 2  ⌡ (ρ − y )
2 2 1/2

0
(3.2.27)
Here
a
a + y i θ -i θ
Z ( y ) = πcosh (πθ)χ( y ) − tanh(πθ) ⌠ a + r χ( r)d r .
2 i
 π a − y  ⌡ a − r  r − y 
-a
(3.2.28)
Formulae (3.2.10) and (3.2.21) are the main results of this section.
Solution of the second kind. Assume solution to the problem in the form
a a
F (t)tdt F ( t )d t
1 d ⌠ 1 1 d ⌠ 2
σ(ρ) = , τ(ρ) = . (3.2.29)
ρ dρ ⌡ ( t 2 − ρ2)1/2 √γ γ dρ ⌡ ( t 2 − ρ2)1/2
1 2
ρ ρ
Again, F and F are the as yet unknown stress functions. Substitution of

1 2
(3.2.29) in (3.2.6) and (3.2.8) results in
a F ( t )d t a F ( t )d t
2 α ⌠
F ( r ) = χ ( r ),
2 α r ⌠
2 2 π
+ −
π√γ γ ⌡ t √γ γ ⌡ t(t 2
− r) 2 2 1  1
1 2 0 1 2 0
a F ( t )d t
2 π α r ⌠ 1
 = χ ( r).
− F 2( r ) − (3.2.30)
π 2 √γ γ ⌡ t − r 
2 2 2
1 2 0
Assuming that
F ( t ) = F (− t ), F ( t ) = − F (− t ), (3.2.31)
1 1 2 2
equations (3.2.30) can be rewritten as
a
F ( t )d t
α ⌠
2
−F ( r) + = χ ( r ),
1 π√γ γ ⌡ t − r 1
1 2
-a
a
F ( t )d t
α ⌠
1
−F ( r) − = χ ( r ). (3.2.32)
2 π√γ γ
1 2 ⌡ t − r 2
-a
Introducing the complex functions
F ( r ) = F ( r ) + iF ( r ), χ( r ) = χ ( r ) + i χ ( r ), (3.2.33)
1 2 1 2
the system (3.2.32) can be reduced to a single equation, namely,
a
iα F ( t )d t
F ( r) + ⌠ = − χ( r ). (3.2.34)
π√γ γ
1 2 ⌡ t − r
-a
Multiplication of (3.2.34) by ( a + r )-i θ( a − r )i θ( r − y )-1 and integration with respect to

r from − a to a leads to
a
-i θ -i θ -i θ
⌠ a + r F ( r)d r + i α −π2a + y  F ( y ) + a + y  πcoth(πθ)
⌡ a − r  r − y π√γ γ  a − y 
1 2
a − y  i
-a
a a a
a + t -i θ F ( t )d t  a + r -i θ χ( r )d r
− coth(πθ) ⌠  = − ⌠
π
× ⌠
F ( t )d t
.
⌡ t − y i ⌡ a − t  t − y  ⌡ a − r  r − y
-a -a -a
(3.2.35)
Here formula (3.2.22) and the integrals from Appendix 3.1 were used. Taking,
as before, tanh(πθ)=α/√γ γ , equation (3.2.35) can be simplified significantly,
1 2
namely,
a a
a + y i θ ⌠ a + r -i θ χ( r )d r
F ( y ) + coth(πθ) ⌠ = − coth(πθ)
i F ( t )d t i
.
π ⌡t − y π a − y  ⌡ a − r  r − y
-a -a
(3.2.36)
The singular integral can be eliminated from (3.2.36) by using (3.2.34), and the
final result is
a
iθ -i θ
F ( y ) = cosh2(πθ)−χ( y ) + tanh(πθ) 
i a + y ⌠ a + r χ( r)d r .
 π a − y  ⌡ a − r  r − y 
-a
(3.2.37)
The general solution may be considered completed. Some additional results are
presented for the case of a bonded punch. The total force, exerted by the
punch is obtained by integration of σ:
a a
P = −2π⌠ F 1( t )d t = −π ⌠ F 1( t )d t . (3.2.38)
⌡ ⌡
0 -a
P = πcosh(πθ)ℜ ⌠ a + r-iθ χ( r)d r.

⌡ a − r 
-a
As expected, this result is identical to (3.2.24). The displacements outside the

punch can be found by substitution of (3.2.37) and (3.2.29) in (3.2.25), with the
result
a
ℜ{ F ( y )}
w (ρ) = −2π H⌠ 2 1/2 d y ,
⌡ (ρ − y )
2
0
u (ρ) = − √γ γ 2π⌠ 2 d y + P tanh(πθ), for ρ> a .

H y ℑ{ F ( y )}
ρ 1 2  ⌡ (ρ − y 2)1/2 
0
(3.2.39)
Comparison of (3.2.28) and (3.2.37) shows that Z ( y )=−π F ( y ), and this means that
formulae (3.2.39) actually repeat (3.2.27).
Solution of the third kind. Let the stresses in question be expressed
through two new and as yet unknown functions q and q as follows
1 2
ρ q (t)tdt a
q ( t )d t
1 d ⌠ 1 1 d ⌠ 2
σ(ρ) = , τ(ρ) = .
ρ dρ ⌡ (ρ2 − t 2)1/2 √γ γ dρ ⌡ ( t 2 − ρ2)1/2
1 2
0 ρ
(3.2.40)
Substitution of (3.2.40) in (3.2.6) and (3.2.8) leads to the system
a q (t)tdt a q (t)tdt
α ⌠ 2
2 1/2⌠
1 π
2 + (a
2
− r) 2 1/2 = 2 χ1( r ),
√γ γ ⌡ t − r
2
⌡ ( t − r )( a − t )
2 2 2
1 2 0 0
α
−q ( r) − q ( r ) = χ ( r ). (3.2.41)
2 √γ γ 1 2
1 2
Function q can be expressed from the second equation (3.2.41) as

2
α
q ( r) = − q ( r ) − χ ( r ). (3.2.42)
2 √γ1γ2 1 2
Substitution of (3.2.42) into the first equation (3.2.41) gives
a q (t)tdt a q (t)tdt
α2 ⌠
− r) ⌠
1 1
(a 2 2 1/2
− = ψ( r ). (3.2.43)
⌡ ( t − r )( a − t )
2 2 2 2 1/2 γ γ ⌡ t − r2
2
1 2
0 0
Here
a χ2( t ) t d t
π α ⌠
ψ( r ) = χ ( r) + . (3.2.44)
2 1 √γ γ ⌡ t 2 − r2
1 2 0
An exact solution of the integral equation (3.2.43) can be obtained in the

following manner. Introduce the notation
Y ( r ) = sinθln|
a + r
Y ( r ) = cosθln|
a + r
| , | . (3.2.45)
s  a − r c  a − r
Multiply both sides of (3.2.43) by Y ( r )/( r 2− x 2) and integrate with respect to r

c
from 0 to a . The result is
a Y (t) Y ( x ) q ( t )d t
π2  α2  π ⌠  s
a 2 − x 21/2 s  1
− 1− q ( x ) Y ( x ) + tanh(πθ) −
4 x 2 γ 1γ 2 1 c 2 ⌡ t  a 2 − t 2  x ( x 2 − t 2)
0
a q (t) Y (t) Y ( x)
−
πα2
coth(πθ)⌠ 2
1
 s
−
s
d t = X ψ( x ).
2γ γ
1 2 ⌡ x − t
2
 t x  c
0
(3.2.46)
Hereafter the following integral operators are introduced
a Y ( r )ψ( r )d r a Y ( r )ψ( r ) r d r
X ψ( x ) = ⌠ X ψ( x ) = ⌠
c s
, . (3.2.47)
c
⌡ r − x
2 2 s
⌡ r2 − x2
0 0
Again, assuming tanh(πθ)=α/√γ γ , equation (3.2.46) can be simplified as

1 2
π2 q ( x ) Y ( x ) a
2 1/2 q 1( t )d t
−
1 π 
+ tanh(πθ) Y ( x )⌠ 1 −
c
a 2
− x   = x X ψ( x ).
4 x cosh (πθ) ⌡  a − t  x 2 − t 2
2 2 s 2 2 c
0
(3.2.48)
A similar procedure of multiplication of (3.2.43) by rY ( r )/( r 2
− x ) leads, after
2
s
simplification, to
π2 q 1( x ) Y s( x ) a q ( t )d t
−
π ⌠ 
− tanh(πθ) Y c( x ) 1 − a 2 − x 21/2 1
= Xsψ( x ).
4 x cosh2(πθ) 2 ⌡  a 2 − t 2  x 2 − t 2
0
(3.2.49)
Equations (3.2.48) and (3.2.49) finally give the solution
Y c( x )
4  
q 1( x ) = − 2 cosh (πθ) xY c( x )Xcψ( x ) + Y s( x )Xsψ( x ) + b 0
2
.
π   x
(3.2.50)
The last term in (3.2.50) represents the homogeneous solution, and b 0 is an
arbitrary constant. If the stresses defined by (3.2.40) are to be nonsingular at
ρ=0, then b 0 should be equal to zero.
Substitution of (3.2.44) in (3.2.50) allows us to express the solution in

terms of χ1 and χ2:
q 1( x ) = −cosh2(πθ)
2 2
xY c( x )Xcχ1( x ) + Y ( x )Xsχ1( x ) − tanh(πθ)χ2( x )
 π π s
Y s( x )Xc{ x χ2( x )}.

2 2
+ xY c( x )Xs{χ2( x )/ x } − (3.2.51)
π π 
Here a modified version of formula (3.2.22) was employed:
a a a a
⌠ 2 d t 2 ⌠ f (2t , r)d r2 = − π f ( x2, x ) + ⌠ d r ⌠ 2 f ( t2, r)d2t

2
2 .
⌡ t − x ⌡ r − t 4x ⌡ ⌡ ( t − x )( r − t )
0 0 0 0
(3.2.52)
The second stress function q 2 may be obtained from (3.2.42) and (3.2.51) in the
form
 2
q 2( x ) = cosh2(πθ) tanh(πθ)[ xY c( x )Xcχ1( x ) + Y s( x )Xsχ1( x )
π


+ xY c( x )Xs{χ2( x )/ x } − Y s( x )Xc{ x χ2( x )}] − χ2( x ).

(3.2.53)
The general solution is completed, and we can derive some additional results for
the case of a bonded punch. The total force P may be obtained by integration
of the first expression in (3.2.40):
a q 1( t ) t d t
P = 2π ⌠ 2 1/2 . (3.2.54)
⌡ (a − t )
2
0
Substitution of (3.2.51) in (3.2.54) gives, after integration,
P = 2πcosh(πθ)⌠ [χ ( r ) Y ( r ) + χ ( r ) Y ( r )]d r . (3.2.55)

1 c
⌡ 2 s
0
The result (3.2.55) coincides with (3.2.24) when χ is an even function and χ
1 2
is odd. The displacements outside the punch may be found by substitution of
(3.2.40) in (3.2.25), and are
a a q (t)tdt
w (ρ) = 4 H ⌠  2
a 2 − y 21/2
dy ⌠ 2
1
2 1/2 ,
⌡ ρ − y  ⌡ ( t − y )( a − t )
2 2 2
0 0
a q ( y) ydy
1  ⌠ 2

u (ρ) = − H √γ γ 2π 2 1/2 + P tanh(πθ) , for ρ> a .
1 2 ρ  
⌡ (ρ − y )
2
0
(3.2.56)
One can prove that expressions (3.2.56) are in agreement with (3.2.28).
Solution of the fourth kind. Let the solution be
a
Q (t)tdt ρ Q ( t )d t
1 d ⌠ 1 1 d ⌠ 2
σ(ρ) = , τ(ρ) = .
ρ dρ ⌡ ( t − ρ2)1/2
2
√γ1γ2 ρ ⌡
d (ρ − t 2)1/2
2
ρ 0
(3.2.57)
By using the same methods as before, the following set of equations can be
obtained
α
Q ( r) − Q ( r) = χ ( r) + b,
√γ1γ2 2 1 1
a Q 2( t )d t a Q ( t )d t
α ⌠ π
(a − r ) ⌠
1
2 2 1/2
− 2 = 2 r χ2( r ),
⌡ ( t − r )( a − t ) √γ γ ⌡ t − r
2 2 2 2 1/2 2
0 1 2 0
(3.2.58)
where
a Q 2( t )d t
2α ⌠
b = . (3.2.59)
π√γ γ ⌡ ( a 2 − t 2)1/2
1 2 0
It will be shown later, that Q 2 can always be chosen in such a way that b =0.
Now Q 1 can be expressed from the first equation (3.2.58) and substituted in the
second one, with the result
a Q 2( t )d t a
Q 2( t )d t
α ⌠ 2
(a 2
− r) ⌠
2 1/2
− = Ψ( r ).
⌡ ( t − r )( a − t )
2 2 2 2 1/2 γ 1γ 2 ⌡ t 2 − r 2
0 0
(3.2.60)
Here
a χ1( r ) + b
π α ⌠
Ψ( r ) = χ2( r ) − dt. (3.2.61)
2r √γ1γ2 ⌡ t 2 − r2
0
Equation (3.2.60) is similar to (3.2.43), so its solution can be written down as
Q 2( x ) = −
4  x2Y ( x )X Ψ( x ) + xY ( x )X Ψ( x ) + b Y ( x ).
2 cosh (πθ)
2
π  c c s s  2 c
(3.2.62)
The last term in (3.2.62) represents the homogeneous solution, and b 2 is an
arbitrary constant to be chosen from the condition b =0. Appropriate integration
of (3.2.62), using the integrals from Appendix 3.1, allows us to define b 2 as
follows
b 2 = − 2cosh2(πθ) ⌠ Ψ( x ) Y c( x )d x .
4
(3.2.63)
π ⌡
0

Q 2( x ) = −
4  
2cosh (πθ) Y c( x )Xc{ x Ψ( x )} + xY s( x )XsΨ( x ) .
2 2
(3.2.64)
π  
Function Q 2 can be expressed in terms of χ1 and χ2 by using (3.2.61) and

(3.2.59) as follows:
2
Q 2( x ) = − cosh2(πθ){tanh(πθ)χ1( x ) + [ xY ( x )Xcχ1( x )
π s
− Y c( x )Xsχ1( x ) + Y c( x )Xc( x χ2( x )) + xY s( x )Xs(χ2( x )/ x )]}.

(3.2.65)
Consequently, from the first expression of (3.2.58), we have
2
Q 1( x ) = −cosh2(πθ){χ1( x ) + tanh(πθ)[ xY s( x )Xcχ1( x )
π
− Y c( x )Xsχ1( x ) + Y c( x )Xc( x χ2( x )) + xY s( x )Xs(χ2( x )/ x )]}.

(3.2.66)
The total force exerted by a bonded punch is defined by
P = −2π⌠ Q 1( t )d t .
⌡
0
One can show that this result coincides with (3.2.55). The displacements outside
the punch are
a Q 1( y )d y
w (ρ) = −2π H ⌠ 2 1/2,
⌡ (ρ − y )
2
0
a a Q 2( t )d t
2H  2
− 
2 1/2
α 
2√γ γ ⌠ y dy ⌠ 2
a y
u (ρ) = 2
2 1/2 − 2 P .
ρ  1 2 ⌡ ρ2 − y 2 ⌡ ( t − y )( a − t )
2 2

0 0
(3.2.67)
Expressions (3.2.67) are in agreement with (3.2.27).
Example 1. Consider a circular flat-ended punch of radius a , bonded to a
transversely isotropic half-space, and acted upon by an axial force P . The
boundary conditions (3.2.1) in this particular case will take the form
u (ρ) = 0, w (ρ) = w 0 = const., for ρ< a ;
σ(ρ) = 0, τ(ρ) = 0, for ρ> a .
The stress functions are
w0
a + y i θ
f( y) = i coth(πθ)1 − cosh(πθ)  ,
πH
2
 a − y  
w0 w0
a + y i θ
F ( y ) = − 2 cosh(πθ) , Q 1( y ) = − cosh(πθ) Y c( y ),
πH a − y  π2 H
w0
Q 2( y ) = coth(πθ)[1− cosh(πθ) Y c( y )].
π2 H
w 0cosh2(πθ) w0
q 1( y ) = Y s( y ), q 2( y ) = − cosh(πθ) Y s( y ),
π2 H sinh(πθ) π2 H
One can notice that q 1=ℜ f , q 2=ℑ F , Q 1=ℜ F , and Q 2=ℑ f , and this is why we
have only two different representations for the surface tractions, namely,
w 0cosh2(πθ) ρ Y s( t ) t d t
1 d ⌠
σ(ρ) = 2
π H sinh(πθ) ρ dρ ⌡ (ρ2 − t 2)1/2
0
a
w0 Y c( t ) t d t
1 d ⌠
= − 2 cosh(πθ) ,
πH ρ dρ ⌡ ( t 2 − ρ2)1/2
ρ
w 0cosh(πθ) ρ Y c( t )d t w 0cosh(πθ) a
Y s( t )d t
d ⌠ d ⌠
τ(ρ) = − = − .
π2 H α dρ ⌡ (ρ2 − t 2)1/2 π2 H √γ1γ2 dρ ⌡ ( t 2 − ρ2)1/2
0 ρ
The equivalence of these representations can be verified by using the identities

ρ Y c( t )d t a
Y ( t )d t
⌠ ⌠ s π
= tanh( πθ) 2 1/2 + 2cosh(πθ), (3.2.68)
⌡ (ρ − t )
2 2 1/2
⌡ (t − ρ )
2
0 ρ
ρ Y s( t ) t d t a
Y c( t ) t d t
⌠ πθ a ⌠
= − tanh(πθ) . (3.2.69)
⌡ (ρ − t )
2 2 1/2 cosh(πθ) ⌡ ( t 2 − ρ2)1/2
0 ρ
These and other similar identities can be established by using the general
relationship:
a ρ a
⌠ f ( x )d x 2 ⌠
2 1/2 = π
dx ⌠ 2f ( t ) t d t 2 + π lim[ tf ( t )].
⌡ (x − ρ )
2
⌡ (ρ − x ) ⌡ t − x
2 2 1/2 2ρ t→0
ρ 0 0
(3.2.70)
The relationship between the total force P and the punch settlement w 0 is
2w0 aθ
P = . (3.2.71)
H tanh(πθ)
The surface displacements outside the punch are given by
a Y c( x )d x
w (ρ) =
2
w cosh(πθ) ⌠ 2 1/2,
π 0 ⌡ (ρ − x )
2
0
a Y s( x ) x d x
w √γ γ cosh(πθ)⌠
2 πaθ  1
u (ρ) = − .
π 0 1 2 ⌡ (ρ − x )
2 2 1/2 cosh(πθ) ρ
0
(3.2.72)
The integrals in (3.2.72) can be evaluated (Gradshtein and Ryzhik, 1963)
a Y c( x )d x ∞ k
πθ
Σ(−1)(2(kρ /+a 1)(− k1)!) Π( θ
k 2 2 -k-1/2
I c(ρ) = ⌠ = 2
+ m 2),
⌡ (ρ2 − x 2)1/2 sinh(πθ) 2
k =0 m=1
0
(3.2.73)
a Y s( x ) x d x ∞ k
I s(ρ) = ⌠
⌡ (ρ − x )
2 2 1/2 =
π a θ2
sinh(πθ) Σ
k =0
(−1)k(ρ2/ a 2 − 1)-k-1/2
(2 k + 1) k !( k + 1)! Π( θ
m=1
2
+ m 2).
0
(3.2.74)
For real materials, the physical constant θ<1. For example, for an isotropic
material
1
θ = ln(3 − 4ν). (3.2.75)
2π
Since the Poisson coefficient ν≤0.5, this means that 0≤θ<0.2 for isotropic
materials. By using the approximation
k
( k !) -2
Π( θ
m=1
2
+ m 2) = 1 + O(θ2),
the summation in (3.2.73) and (3.2.74) can be performed, and the results are
πθ a
I (ρ) ≈ sin-1( ), (3.2.76)
c sinh(πθ) ρ
πθ2  2
(ρ − a 2)1/2ln1 − 2 + 2 a sin-1( ).
a2 a
I (ρ) ≈
s sinh(πθ)  ρ ρ
(3.2.77)
Direct numerical computations show that the relative error of (3.2.76) is less than
6% when (ρ/ a )>1.01, and θ≤0.2; the relative error of (3.2.77) is less than 4%.
The same accuracy can be obtained for (ρ/ a )>1.1, and θ≤0.3. The relative error
of both (3.2.76) and (3.2.77) rapidly decreases when ρ increases, for example, the
relative error is less than 4% and 2% respectively, for θ<0.9 and (ρ/ a )>3.
Example 2. Consider the case where no forces act upon the punch bonded
to a half-space, but the punch itself expands, so that the radial displacement is
proportional to the radius, with k as the coefficient, namely, inside the circle
ρ= a the boundary conditions are
w (ρ) = w , u (ρ) = k ρ.
0
Here w is the as yet unknown punch settlement. The stress functions are
0
coth(πθ) 2 ky a + y iθ 2 k (2 a θ + iy ) + w ,

f( y) =  iw − − i cosh(πθ)
a − y  √γ1γ2 0
π2 H 0 √γ1γ2
 
cosh(πθ) a + y i θ
(2 a θ + iy ) 
2k
F ( y) = − w + ,
πH  0
2
√γ1γ2  a − y 
coth(πθ) 2k 
q ( y) =  w cosh(πθ) Y ( y ) + [cosh( πθ) ( 2 a θ Y ( y )+ yY ( y ) ) − y ] ,
1 π2 H 0 s √γ1γ2 s c
 
cosh(πθ) 2k 
q ( y) = −  w Y ( y ) + [2 a θ Y ( y ) + yY ( y )] ,
2 π2 H 0 s √γ γ s c
 1 2 
cosh(πθ) 2k 
Q ( y) = −  0 c
w Y ( y ) + [2 a θ Y ( y ) − yY ( y )] ,
1 π2 H √γ γ c s
 1 2 
coth(πθ) 2k 
Q ( y) =  0
w [1 −cosh(πθ) Y ( y )] − cosh(πθ) [2 a θ Y ( y ) − yY ( y )] .
2 π2 H c √γ γ c s
 1 2 
The stresses can be defined two ways
ρ q ( x) xdx a
Q ( x) xdx
1 d ⌠ 1 1 d ⌠ 1
σ(ρ) = = ,
ρ dρ ⌡ (ρ2 − x 2)1/2 ρ dρ ⌡ ( x − ρ2)1/2
2
0 ρ
ρ Q ( x )d x a
q ( x )d x
1 d ⌠ 2 1 d ⌠ 2
τ(ρ) = = .
√γ γ dρ ⌡ (ρ2 − x 2)1/2 √γ γ dρ ⌡ ( x − ρ2)1/2
2
1 2 1 2
0 ρ
The total force P is defined by
coth(πθ) w +
2aθ 2 ka θ
P = . (3.2.78)
H  0 √γ γ 
1 2
When no force acts upon the punch, its normal displacement is equal to
w = −2 ka θ/√γ γ .
0 1 2
The force, needed to provide a zero normal displacement is
4 ka 2θ2
P = .
Hα
The displacements outside the punch are
a Q 1( x )d x
w (ρ) = −2π H ⌠ ,
⌡ (ρ − x )
2 2 1/2
0
a q ( x) xdx
u (ρ) = − H √γ1γ22π ⌠  1.
2
2 1/2 + P tanh(πθ)
 ⌡ (ρ − x )
2
 ρ
0
One can calculate the stress at the point ρ=0 in elementary functions
coth(πθ)θ k 
σ(0) =
πH  w cosh(πθ) + [(1+4θ2)cosh(πθ) − 1], τ(0) = 0.
a 0 √γ1γ2
 
Discussion. One could notice that the four kinds of solution considered
represent all combinations of the Abel type integrals with limits from 0 to ρ and
from ρ to a . Solutions of the first and the second kind are more compact than
the others, but they are only convenient to use when w (ρ) is an even function
and u (ρ) is an odd one. The solution of the second kind is preferable when
one is interested mainly in the displacements outside the punch, while the
solution of the first kind has definite advantages when one is interested in the
stress distributions. Formulae (3.2.10) are more convenient for numerical
integration than (3.2.29), especially in the domain close to either ρ=0 or ρ= a : (i)
the differentiation of the integrand can be performed in (3.2.10), thus avoiding
numerical differentiation, which is less accurate; the differentiation in (3.2.29) is
rather difficult because F ( a ) usually is not defined; (ii) formulae (3.2.10) allow
us to determine easily the stress at ρ=0, directly through the stress function,
namely,
π
σ(0) = f ´(0), τ(0) = f 2´(0)/√γ1γ2,
2 1
while the result of (3.2.29) is rather difficult to use, for example,

a F ( x ) − F (0)
1 d F 1( x ) x d x 
a
F 1(0)
σ(0) = lim ⌠  = ⌠
1 1
dx − .
ρ→0 ρ dρ ⌡ ( x 2 − ρ2)1/2 ⌡ x2 a
 ρ  0
(3.2.79)
The solutions of the third and the fourth kind are more general because they do
not require either w (ρ) to be an even function or u (ρ) to be an odd one. The
same logic of preference is applicable here: the integral representations with limits
from 0 to ρ are more convenient for evaluation of the stresses while the
representations with limits from ρ to a are preferable for the displacement
evaluation.
It is of interest to establish relationships between the various kinds of
solution. Some are obvious, due to the uniqueness of the solution, namely,
f 1 = q 1, f 2 = Q 2, F 1 = Q 1, F 2 = q 2.
(3.2.80)
The other relationships may be found from (3.2.70) and the following identity
ρ a a
⌠ f ( x )d x 2 ⌠ xdx ⌠ ( a 2 − y 2)1/2 f ( y )d y
= − .
⌡ (ρ − x )
2 2 1/2 π ⌡ ( x 2 − ρ2)1/2 ⌡ ( a 2 − x 2)1/2( y 2 − x 2)
0 ρ 0
(3.2.81)
Comparison of (3.2.57) and (3.2.40) with (3.2.70) and (3.2.81) yields
a ( a 2 − y 2)1/2 yq 1( y )
2 ⌠
Q 1( t ) = − dy,
π ⌡ ( a 2 − t 2)1/2( y 2 − t 2)
0
a ( a 2 − y 2)1/2 Q 2( y )
2 ⌠
q 2( t ) = − t dy,
π ⌡ ( a 2 − t 2)1/2( y 2 − t 2)
0
(3.2.82)
a 2 a
y Q ( y) y q ( y)
2 ⌠ 1 2 ⌠ 2
q (t) = dy, Q (t) = dy.
1 π t ⌡ ( y − t 2)
2 2 π ⌡ ( y − t 2)
2
0 0
(3.2.83)
Comparison of (3.2.56) and (3.2.67) leads to slightly different expressions:
a ( a 2 − t 2)1/2 yq 1( y )d y
2 ⌠
Q 1( t ) = − ,
π ⌡ ( a 2 − y 2)1/2( y 2 − t 2)
0
a ( a 2 − t 2)1/2 Q 2( y )d y
2 ⌠
q 2( t ) = − t .
π ⌡ ( a 2 − y 2)1/2( y 2 − t 2)
0
(3.2.84)
Expressions (3.2.84) differ from (3.2.82) by the term const/( a − t ) and 2 2 1/2
t ⋅const/( a − t )
2 2 1/2
respectively. One may notice from (3.2.57) and (3.2.40) that
addition of these terms to Q 1 and q 2 respectively does not affect the stresses and
therefore, expressions (3.2.82) and (3.2.84) are equivalent. The same argument is
applicable to q 1. Since the addition to q 1 of a term const/ t does not affect the
solution, an alternative to the first expression of (3.2.83) may be suggested
a Q 1( y )d y
2 ⌠
q 1( t ) = t . (3.2.85)
π ⌡ y2 − t2
0
Comparison of (3.2.39) and (3.2.26) allows us to construct the relationship

between the complex stress functions
⌠ ( a2 − t2 )1/2 f ( y )d y .
2 2 1/2
1
F(t) = − (3.2.86)
π ⌡ (a − y ) (y − t)
-a
The inverse relationship takes the form
f(t) =
1 ⌠ F ( y )d y . (3.2.87)
π ⌡ y − t
-a
One can also deduce from (3.2.82) the following expression which is equivalent
to (3.2.86):
⌠ ( a2 − y2 1/2
2 2 1/2
1 ) f ( y )d y
F(t) = − . (3.2.88)
π ⌡ (a − t ) (y − t)
-a
Considerable simplifications occur when θ=0. In the case of an isotropic body

this condition corresponds to the Poisson coefficient ν=1/2. The stress functions
will be defined by
f( y) = −
1 ⌠ χ( r)d r , F ( y ) = −χ( y ),
π ⌡ r − y
-a
a χ ( r )d r
2 ⌠ 1
q ( y) = − y q ( y ) = −χ ( y ),
1 π ⌡ r − y2
2 2 2
0
a χ2( r ) r d r
2 ⌠
Q 2( y ) = − Q 1( y ) = −χ1( y ).
π ⌡ r2 − y2
0
The stress distributions are
a
χ1( x ) x d x a
χ2( x )d x
1 d ⌠ 1 d ⌠
σ(ρ) = − , τ(ρ) = − .
ρ dρ ⌡ ( x 2 − ρ2)1/2 √γ1γ2 dρ ⌡ ( x 2 − ρ2)1/2
ρ ρ
The displacements outside the punch simplify as follows:
a χ2( x ) x d x a χ1( x )d x
1 ⌠
u (ρ) = 2π H √γ1γ2 , w (ρ) = 2π H ⌠ 2 1/2.
ρ ⌡ (ρ2 − x 2)1/2 ⌡ (ρ − x )
2
0 0
(3.2.89)
Notice that here the normal parameters are decoupled from the tangential ones,
namely, the normal displacements affect the normal pressure, and the tangential
displacements produce the shear tractions only. Substitution of (3.2.7) and (3.2.9)
in (3.2.89) furnishes a direct relationship between the displacements inside and
outside the circle ρ= a
a
2 (ρ2 − a 2)1/2 ⌠ u ( x ) x 2d x
u (ρ) = 2 , for ρ> a ;
π ρ ⌡ ( a − x ) (ρ − x )
2 2 1/2 2
0
(ρ2 − a 2)1/2 ⌠
2 w( x) xdx
w (ρ) = 2 , for ρ> a .
π ⌡ ( a − x ) (ρ − x )
2 2 1/2 2
0
The last expressions demonstrate a certain mathematical similarity between the

normal and tangential displacements.
Exercise 3.2
1. Prove that the settlement of a smooth flat circular punch is greater then or
equal to that of a bonded punch.
Hint : prove that (πθ)≥tanh(πθ).
Note : a more general statement can be proven from the energy consideration.
2. An axisymmetric pressure σ=σ(ρ) is applied to the annulus b ≤ρ≤ c outside a

flat bonded punch of radius a . Investigate its influence on the punch settlement
w 0 and the traction distributions under the punch.
Solution: The right hand side in the integral equations (3.2.2) and (3.2.3) will
take the form
ρ c σ(ρ )ρ dρ
ω1(ρ) = w 0 − 4 H ⌠ ⌠
dx 0 0 0
2 1/2, ω2 = 0.
⌡ (ρ − x ) ⌡ (ρ0 − x )
2 2 1/2 2
0 b
Now, from (3.2.7) and (3.2.9)
w0 c σ(ρ )ρ dρ
2 ⌠ 0 0 0
χ1( r ) = 2 − , χ2 = 0.
πH π ⌡ (ρ0 − r )
2 2 1/2
b
Substitution of the last expressions in the chosen kind of solution gives the
stresses and the displacements. The total force can be defined as
w0 aθ c
P = 2cosh(πθ) − 2⌠ I c(ρ) σ(ρ)ρdρ,

H sinh(πθ) ⌡ 
b
where I c is defined by (3.2.73). When the punch is not subjected to a direct

loading, its settlement will be
sinh(πθ)⌠ I (ρ)σ(ρ)ρdρ.
2H
w =
0 aθ ⌡ c
b
If no displacements are allowed for the punch, then the total force should be
P = −4cosh(πθ)⌠ I (ρ) σ(ρ)ρdρ.

⌡ c
b
By using the approximation (3.2.76), the results can in many cases be expressed
in elementary functions. For example, in the case when σ=σ =const,
0
c πθσ 
⌠ I (ρ)σ(ρ)ρdρ ≈
0 a a
2sinh(πθ)  c 2sin-1( ) − b 2sin-1( )
⌡ c

c b
b

+ a [( c 2 − a 2)1/2 − ( b 2 − a 2)1/2].

3. Investigate the influence of radial tangential tractions τ=τ(ρ), applied at the
surface of an annulus b ≤ρ≤ c , outside a bonded punch of radius a .
Solution : We have from (3.2.2) and (3.2.3)
ω (ρ) = w + 2π H α⌠ τ(ρ)dρ,
1 0
⌡
b
ρ c τ(ρ )dρ
2
1⌠ x xd ⌠
0 0
ω (ρ) = −4 H γ γ 1/2.
1 2 ρ
⌡ (ρ − x ) ⌡ (ρ0 − x )
2 2 2 1/2 2 2
0 b
Substitution in (3.2.7) and (3.2.9) yields
w c c
2 ⌠ τ( x )d x
χ ( r ) = − √γ γ r⌠
0 2
χ ( r) = 2 + α τ(ρ)dρ, .
1 πH π ⌡ 2 π 1 2 ⌡ ( x 2 − r 2)1/2
b b
Since χ is an even function, and χ is an odd one, we can use any of the
1 2
four kinds of solution considered above. The total force is given by (3.2.55):
 aθ 
c

c

P = 2cosh(πθ) w + 2π H α⌠ τ(ρ)dρ − 2√γ γ ⌠ I (ρ)τ(ρ)dρ.
H sinh(πθ)  0 ⌡  1 2
⌡ s
 b b 
Here I is defined by (3.2.74). When the punch is not loaded, its settlement
s
will be
c c
w = 2 H −πα ⌠ τ(ρ)dρ + √γ γ ⌠ I (ρ)τ(ρ)dρ.

sinh(πθ)
0  ⌡ a θ 1 2
⌡ s 
b b
The value of the axial force to provide a zero punch displacement is
 πa θα c c

P = 4cosh(πθ) ⌠ τ(ρ)dρ − √γ γ ⌠ I (ρ)τ(ρ)dρ.
sinh(πθ) ⌡ 1 2
⌡ s
 b b 
4. A normal concentrated load P is applied at the point (0,0, z ) underneath a

circular punch of radius a , bonded to a transversely isotropic elastic half-space.
Find the traction distribution at the punch base and its settlement w .
Answer: as an illustration, the solution is given according to (Fabrikant, 1971a).
The stresses are defined by
ρ
1 d ⌠
f (t)tdt a
f (t)tdt 
σ(ρ) = 
1
+ ⌠ 2
,
ρ dρ ⌡ (ρ − t )
2 2 1/2
⌡ ( t − ρ2)1/2
2
 0 ρ 
ρ
d
a
f ( t )d t f ( t )d t 
α ⌠ 1 1⌠ 2
τ(ρ) =
dρ −
γ γ ⌡ ( t 2 − ρ2)1/2
+
α⌡ (ρ2 − t 2)1/2
.
1 2
ρ 0 
Here
2 m tY ( t )sinhξ + z Y ( t )coshξ
cosh2πθ
Σ  ,
k c k k s k
f (t) =
1 π2 ( m − 1)sinhπθ z 2k + t 2

k=1 k
2 γk zkY c( t )sinhξk + tY s( t )coshξk

coshπθ
f 2( t ) = 2
π √γ1γ2 Σ
k =1

m k − 1 z 2k + t 2
, ξ = 2θtan-1( a ).
 k zk
The resultant of the stresses is
2 m sinhξ γ (coshξ − 1)
N = − P cothπθ  Σ .
k k k k
+
 k
m − 1 √γ1γ2( m k − 1) 
k =1
When no force is applied to the punch directly, its settlement due to the load P
is
2 m sinhξ γ (coshξ − 1)
Σ  k .
HP k k k
w = +
2aθ  k
m − 1 √γ1γ2( m k − 1) 
k =1
Note : verify that in the case z →0 N =− P .
3.3 External axisymmetric mixed-mixed problem
We choose this problem to demonstrate yet another kind of solution, which

features two stress functions, introduced in such a way that they decouple the
integral equations, so that each equation can be solved independently.
The boundary conditions in the case of axial symmetry are
u = u (ρ), for a ≤ρ≤∞, 0≤φ<2π;
w = w (ρ), for a ≤ρ≤∞, 0≤φ<2π;
σ = σ(ρ), for 0≤ρ≤ a , 0≤φ<2π;
τ = τ(ρ), for 0≤ρ≤ a , 0≤φ<2π. (3.3.1)
The set of governing integral equations will take the form
∞ ∞x σ(ρ )ρ dρ
 0 
2 H −πα ⌠ τ(ρ )dρ + 2⌠ d x ⌠
0 0
 = ω1(ρ),
 ⌡ 0 0 ⌡ ( x 2 − ρ2)1/2 ⌡ ( x 2 − ρ20)1/2
ρ ρ a
(3.3.2)
∞
τ(ρ )ρ20dρ ρ
2H 
x
dx
2γ γ ρ2⌠ ⌠ ⌠
0 0
ρ 1 2 ⌡  = ω2(ρ).
− πα σ( ρ ) ρ dρ
x 2( x 2 − ρ2)1/2 ⌡ ( x − ρ0)
2 2 1/2
⌡ 0 0 0
 ρ a  a
(3.3.3)
The functions ω and ω are known from the boundary conditions, and are
1 2
defined by
a a σ(ρ )ρ dρ
ω (ρ) = w (ρ) − 4 H⌠ ⌠
dx 0 0 0
,
1
⌡ (ρ2 − x 2)1/2 ⌡ (ρ20 − x 2)1/2
0 x
(3.3.4)
x 2d x τ(ρ )dρ
a a a
H ⌠ ⌠
0 0 Hα ⌠
ω (ρ) = u (ρ) − 4γ γ 2 1/2 + 2π ρ σ(ρ )ρ dρ .
1 2ρ
⌡ (ρ − x ) ⌡ (ρ0 − x ) ⌡
2 2 2 1/2 2 0 0 0
0 x 0
(3.3.5)
We shall seek the solution of the system (3.3.2) and (3.3.3) in the form
∞
f (t)tdt ρ f (t)tdt
1 d ⌠ ,
+ ⌠
1 2
σ(ρ) =
ρ dρ ⌡ (t 2
− ρ) 2 1/2
⌡ (ρ − t ) 
2 2 1/2
ρ a
ρ f ( t )d t ∞
f ( t )d t C Da
d ⌠  +
+ C⌠
1 2 1
τ(ρ) = C .
dρ 1⌡ (ρ2 − t 2)1/2 2⌡ (t − ρ ) 
2 2 1/2
ρ(ρ − a 2)1/2
2
a ρ
(3.3.6)
Here f and f are the as yet unknown functions, and C , C , and D are the
1 2 1 2
constants to be determined. Substitution of (3.3.6) in (3.3.2) and (3.3.3) leads to
two uncoupled equations which can be solved independently, provided that the
constants are defined by
C = α /γ γ , C = −1/α. (3.3.7)
1 1 2 2
The equations in question are
ρ f ( t )d t
 πα2 
D sin ( ) − ⌠ 2
1
-1 a
2 H −
γγ ρ ⌡ ( ρ − t 2 1/2
) 
 12 a
∞ ∞
2
− 2 1/2 f (t)tdt 
+ 2⌠ ( x a )
d x ⌠
1
2  = ω1(ρ).
⌡ ( x 2 − ρ2)1/2 ⌡ (t − a ) (t − x )
2 2 1/2 2

ρ a
2γ γ ∞ ∞
2π H α 1 2⌠ dx ⌠ ρ2 a 2( t 2 − x 2) − x 4( t 2 − a 2)
ρ  f ( t )d t
πα2 ⌡ ( x 2 − ρ2)1/2( x 2 − a 2)1/2 ⌡ x 2( t 2 − a 2)1/2( t 2 − x 2) 2
 ρ a
∞ ρ f (t)tdt
  
+ ⌠ ⌠
t 2
− 1 f ( t )d t + aD − 2 1/2 = ω2(ρ).
⌡ ( t − a )  1 ⌡ (ρ − t ) 
2 2 1/2 2
a a
(3.3.8)
Let us solve the first equation (3.3.8). Divide both sides by ρ(ρ − r ) , integrate 2 2 1/2
with respect to ρ from r to ∞, multiply the result by r , and differentiate with

respect to r . The result is
∞ f ( t )d t ∞ ( r 2 − a 2)1/2 f ( t ) t d t
α2 ⌠
− ⌠
1 1
= ψ ( r ).
γ 1γ 2 ⌡ t 2
− r 2
⌡ r( t − a ) ( t − r )
2 2 2 2 1/2 1
a a
(3.3.9)
Here
∞ ω (ρ)dρ
1 d ⌠ 1
 − α2 D lnr + a .
ψ1( r ) = r (3.3.10)
2π H d r  ⌡ ρ(ρ2 − r 2)1/2 γ γ 2 r r − a 
1 2
r
Equation (3.3.9) can be solved in a manner similar to that of (3.2.43). Multiply

both sides of (3.3.9) by Y ( r )/( r 2− x 2) and integrate with respect to r from a to
c
∞. We use in this section the notation Y and Y , as it was defined in
c s
(3.2.45). The result of integration is
∞ f 1( t )d t
π2  α2  πa
1 − Y x f x −
( ) ( ) ⌠
4 x 2 γγ c
1 2
1 2 x 2cosh(πθ)⌡ t ( t 2 − a 2)1/2
a
∞ t ( x 2 − a 2)1/2 Y s( x ) Y s( t ) f ( t )d t
− tanh(πθ) ⌠  2 2 −  1
⌡  x (t − a ) t  t2 − x2
2 1/2
a
∞ Y s( x ) Y s( t ) f 1( t )d t ∞ψ ( r ) Y ( r )d r
coth(πθ)⌠  
π α2
= ⌠
1 c
+ − .
2 γ 1γ 2 ⌡ x t  t2 − x2 ⌡ r2 − x2
a a
(3.3.11)
Again, expression (3.3.11) can be simplified significantly, assuming
tanh(πθ)=α/√γ1γ2,
π2 Y c( x ) Y s( x ) ∞ f ( t )d t
f ( x) +
π
tanh(πθ) 
⌠ 1 − t ( x 2
− a 2 1/2
)  1
4 x cosh (πθ) 1
2 2 2 x ⌡  x( t − a )  t − x2
2 2 1/2 2
a
∞ f ( t )d t ∞ψ ( r ) Y ( r )d r
πa
⌠ = ⌠
1 1 c
− . (3.3.12)
2 x 2cosh(πθ) ⌡ t ( t 2 − a 2)1/2 ⌡ r2 − x2
a a
Multiplication of (3.3.9) by rY ( r )/( r 2− x 2) and transformations similar to those

s
above lead to
π2 Y s( x ) f 1( t )d t∞
f ( x) −
π 
tanh(πθ) Y ( x ) ⌠ 1 −
t ( x 2
− a 2 1/2
) 
4 x cosh2(πθ) 1 2 c
⌡ x ( t 2 − a 2)1/2 t 2 − x 2
a
∞ψ ( r ) Y ( r ) r d r
= ⌠
1 s
. (3.3.13)
⌡ r2 − x2
a
Equations (3.3.12) and (3.3.13) give the final solution
∞ψ ( r ) Y ( r )d r ∞ψ ( r ) Y ( r ) r d r
f 1( x ) = 2cosh2(πθ) x  xY c( x )⌠ + Y ( x)⌠ .
4 1 c 1 s
π  ⌡ r − x
2 2 s
⌡ r − x
2 2

a a
(3.3.14)
Strictly speaking, we should have added a term BY c( x ), representing the

homogeneous solution, where B is an arbitrary constant. It will be shown
further, that we may assume B =0, since the constant D , introduced earlier,
actually plays the same role.
The second equation in (3.3.8) can be solved in a similar manner. At the
first stage, it is reduced to
γ 1γ 2 ∞ r ( r 2 − a 2)1/2 f 2( t )d t ∞ f 2( t )d t ∞f (t)tdt
 ⌠ ⌠ 2  − ⌠ 2
2 + 2 = ψ2( r ),
α  ⌡ (t − a ) (t − r )
2 2 2 1/2 2
⌡ (t − a ) 
2 1/2
⌡t − r
2
a a a
(3.3.15)
with
∞ ω (ρ)dρ
1 d ⌠ 2
.
ψ2( r ) = r (3.3.16)
2π H α d r  ⌡ ( ρ − r ) 
2 2 1/2
r
Its solution is
∞ ψ2( r ) Y c( r )d r ∞ ψ2( r ) Y s( r ) r d r
f ( x ) = − 2sinh2(πθ) xY ( x⌠ + Y ( x)⌠ .
4
2 π  c ⌡ r2 − x2 s
⌡ r2 − x2 
a a
(3.3.17)
One could notice that terms of the form const./ρ were lost during transformation
of the second equation (3.3.8), due to differentiation. This means that the
solution (3.3.17) satisfies the second equation (3.3.8), except for the
abovementioned terms. And here the role of the constant D becomes clear: it
has to be chosen so that the equation be satisfied. We show below how this is
done.
Example. Let the exterior of a circle ρ= a be clamped, so that w = u =0, for
ρ> a . A uniform pressure σ is applied inside the circle. The stress distribution
0
outside the circle and the displacements inside are to be determined. The
general solution described above yields for this particular case
a
( a 2 − x 2)1/2d x a2
ω (ρ) = −4 H σ ⌠ 2 1/2 , ω (ρ) = π H ασ ;
0ρ
⌡ (ρ − x )
1 0 2 2
0
ψ1( r ) = 2π H σ01 −
( r 2 − a 2)1/2
ln
α2 D r + a
− , ψ2( r ) = 0;
 r  γ γ
1 2
2 r r − a 
2
f 1( t ) = coth(πθ)σ0[ tY s( t ) − 2 a θ Y c( t )] − D [1 − Y c( t )], f 2( t ) = 0.
π
(3.3.18)
Substitution of (3.3.18) in the second equation (3.3.8) defines the constant D :
2
D = a θσ0coth(πθ). (3.3.19)
π
Formulae (3.3.6), (3.3.18) and (3.3.19) give the complete solution for the traction
distribution. The displacements inside the circle are given by
2 coth(πθ) ∞ tY s( t ) − a θ[1 + Y c( t )]
u (ρ) = 2π H ασ0 ⌠ tdt
π ρ  ⌡ ( t 2 − ρ2)1/2
 a
ρ
− a θ( a 2 − ρ2)1/2 −
2
,
 
 ∞ ( t − a ) − tY c( t ) − a θ Y s( t )
2 2 1/2
w (ρ) = 4 H σ0 ⌠ dt
⌡
 a ( t 2 − ρ2)1/2
ρ
( a 2 − t 2)1/2 π 
+ ⌠ d t − a θtanh( πθ) .
⌡ (ρ2 − t 2)1/2 2

0
The integrals above may be computed in elementary functions at the centre ρ=0,
namely,
1 + cosh(πθ)
u (0) = 0, w (0) = 4π Ha θσ0 .
sinh(2πθ)
The reader who is interested in numerical results is referred to (Fabrikant, 1972),

where the field of tractions and displacements was computed for steel, concrete
and sandstone.
The method of solution presented in this section is neither the only method
available, nor the simplest one. The reader is encouraged to try various
modifications of the approach presented in section 3.2.
3.4 Generalization for a non-homogeneous half-space 177
Exercise 3.3
1. Consider a transversely isotropic elastic half-space z ≥0. Let the exterior of a
circle ρ= a be clamped, so that u = w =0, for ρ> a . An axisymmetric pressure σ(ρ)
is applied in the annulus b ≤ρ≤ c , with c ≤ a . Find the traction distribution outside
the circle and the displacements inside.
Hint : use solutions presented in Exercise 3.2 as an example.
2. Subject to the conditions of the exercise above, prove that the stresses in the
plane z =0 are in equilibrium.
Note : this is generally not true in internal problems.

circle ρ= a be clamped, so that u = w =0, for ρ> a . An axisymmetric tangential
traction τ(ρ) is applied in the annulus b ≤ρ≤ c , with c ≤ a . Find the traction
distribution outside the circle and the displacements inside.

circle ρ= a be clamped, so that u = w =0, for ρ> a . Find the traction distribution
outside the circle and the displacements inside due to a concentrated force P
applied in the positive Oz direction at the point with cartesian coordinates x =0,
y =0, z = b .
3.4 Generalization for a non-homogeneous half-space
Popov (1973) has considered an internal mixed-mixed problem for the case
of a non-homogeneous isotropic half-space, with elastic modulus E = E z κ,
κ 0
E 0=const, and 0≤κ<1. He reduced the problem to a generalized Abel integral
equation, which he solved in terms of Jacobi polynomial expansion. The
interested reader is referred to the original paper for details. We present here
only the closed form solution of the generalized Abel integral equation, as it was
given in (Fabrikant, 1976).
Consider the integral equation
x a
⌠ F ( t )d t + A ⌠ F ( t )d t = f ( x ), for 0<κ<1, b ≤ x ≤ a .
⌡ ( x − t )κ ⌡ ( t − x )κ
b x
(3.4.1)
Here A is a known constant, f is a known function, and function F is to be
determined. Make use of the following integral representations (Gradshtein and
Ryzhik, 1963)
t
δ
1 ( x − b ) ( t − b )1-δ-κ ⌠ ( y − b )κ-1( t − y )δ-1
= dy,
(x − t)
κ B(κ,δ) ⌡ ( x − y )δ+κ
b
(3.4.2)
x
δ
1 ( x − b ) ( t − b )1-δ-κ ⌠ ( y − b )κ-1( t − y )δ-1
= dy.
( t − x)
κ B(κ,1−δ−κ) ⌡ ( x − y )δ+κ
b
(3.4.3)
Let us define the value of δ from the condition
B(κ,1−δ−κ)
= A. (3.4.4)
B(κ,δ)
By using the properties of Beta-functions, expression (3.4.4) can be simplified
sin(πδ)
= A.
sin[π(δ+κ)]
The value of δ can be found from the last expression as
tan-1
1 A sin(πκ) 
δ = . (3.4.5)
π 1 − A cos(πκ)
Substitute (3.4.2) and (3.4.3) in (3.4.1) and interchange the order of integration.
The result is
x a
δ
(x − b) ⌠ ( y − b ) κ+dδy ⌠
κ-1
F ( t )d t
B(κ,δ) 1-δ = f ( x ).
⌡ ( x − y) ⌡ ( t − b) ( t − y)
δ+κ-1
b y
(3.4.6)
The generalized Abel integral equation is now represented as a sequence of two
Abel type operators, and each can be inverted. The solution is
a
sin(πδ) sin[π(κ + δ)] d ⌠ ( r − b )1-κd r
F ( t ) = −B(κ,δ) δ
π2( t − b )1-δ-κ dt ⌡ (r − t)
t
r
d ⌠ f ( x )d x
× .
d r ⌡ ( x − b )δ( r − x )1-δ-κ
b
(3.4.7)
The form of solution given by (3.4.7) is not the only one which can be derived.
Indeed, one can use the integral representations:
a
1 ( a − t )δ( a − x )1-δ-κ ⌠ ( a − y )κ-1( y − x )δ-1
= dy,
( x − t )κ B(δ,κ) ⌡ ( y − t )δ+κ
x
(3.4.8)
a
1 ( a − t )δ( a − x )1-δ-κ ⌠ ( a − y )κ-1( y − x )δ-1
= dy.
( t − x )κ B(1−δ−κ,κ) ⌡ ( y − t )δ+κ
t
(3.4.9)
Substitution of (3.4.8) and (3.4.9) in (3.4.1) leads to
a y
( a − x )1-δ-κ ⌠ ( a − y )κ-1d y ⌠ ( a − t )δ F ( t )d t
= f ( x ).
B(δ,κ) ⌡ ( y − x)
1-δ
⌡ ( y − t )δ+κ
x b
(3.4.10)
The solution will now take the form
t
sin(πδ) sin[π(δ+κ)] d ⌠ ( a − r )1-κd r
F ( t ) = −B(κ,δ)
π2( a − t )δ d t ⌡ ( t − r )1-κ-δ
b
a
d ⌠ f ( x )d x
× .
d r ⌡ ( a − x )1-δ-κ( x − r )δ
r
(3.4.11)
We leave it to the reader to establish the equivalence of the solutions (3.4.7)
and (3.4.11).
If one compares integral equations (3.2.30) and (3.4.1), the first impression
is that they are so different that there is no way to relate them. This is not
so. Consider a set of equations
a a
−⌠ 
1 sign( r − t ) πκ α ⌠ 1
+ F ( t )d t + cot( )
⌡ ( r + t ) κ
|r − t| 
κ 1 2 √γ γ ⌡ | r − t |κ
1 2
0 0
r ω (ρ)ρdρ
−
1  F ( t )d t = 2cos(πκ/2)Γ(1 − κ) ⌠ 1
2 1/2,
(r + t)  2
κ π2 H ⌡ (r − ρ )
2
0
(3.4.12)
a
α ⌠ 1
− κ F 1( t )d t
πκ 1 2
− cot( ) +
2 √γ γ ⌡ | r − t |κ
(r + t)
κ
t 
1 2 0
+ ⌠
1 sign( r − t ) 1 sign( t )
− − κ − F ( t )d t

⌡ (r + t) κ
|r − t|
κ
t t  2
κ
0
r ω (ρ)dρ
2cos(πκ/2)Γ(1 − κ)
r ⌠ 2
2
= 2 1/2 .
π H √γ1γ2
2
⌡ (r − ρ )
0
(3.4.13)
In the limiting case of κ→0, equations (3.4.12) and (3.4.13) transform into
r a r ω (ρ)ρdρ
α
−⌠ F 1( t )d t + ⌠ F ( t ) ln| ⌠ 2
r + t 1 1
| d t = ,
⌡ π√γ1γ2 ⌡ 2 r − t π H ⌡ ( r − ρ2)1/2
2
0 0 0
(3.4.14)
a r ω2(ρ)dρ r
α ⌠ t2 ⌠ 1 ⌠
− F ( t ) ln 2 dt − F ( t )d t = 2 r 2 1/2.
π√γ1γ2 ⌡ 1 | r − t 2| ⌡ 2 π H √γ1γ2 ⌡ (r − ρ )
2
0 0 0
(3.4.15)
One can easily verify that differentiation of (3.4.14) and (3.4.15) with respect to
r leads to (3.2.30). Thus, the connection is established. Introducing the
complex function F = F 1+ iF 2, equations (3.4.12) and (3.4.13) may be unified as
follows:
r a
−1 + cot( )⌠ + 1 − cot( )⌠

iα πκ F ( t )d t iα πκ F ( t )d t
 √γ1γ2 2 ⌡ ( r − t ) κ  √γ1γ2 2 ⌡ ( t − r)
κ
-a r
r √γ1γ2ρω1(ρ) + ir ω2(ρ)
2cos(πκ/2) Γ(1 − κ) ⌠
= dρ
π2 H √γ1γ2 ⌡ ( r 2 − ρ2)1/2
0
a F 2( t )d t a F 1( t )d t
+ 2i ⌠ .
α πκ
− cot( ) ⌠ (3.4.16)
 ⌡ t κ √γ1γ2 2 ⌡ t κ 
0 0
We obtained the generalized Abel integral equation of the type (3.4.1), with
b =− a , the solution of which is available in the forms (3.4.7) and (3.4.11), with
the parameter δ, defined as
κ
δ = − + i θ, (3.4.17)
2
and θ given by (3.2.19). The stress function for a flat circular bonded punch is
a + t i θ
w 0( a 2 − t 2)κ/2
sin(πκ) cosh(πθ) Γ(1 − κ)
F(t) = − .
π κH
3
a − t 
The last result is in agreement with those of paragraph 3.2. The reader can
derive several new modifications of the governing integral equations by using the
integral representations
min (ρ , ρ)
0
x κ+1d x
⌠
⌡ [(ρ − x )(ρ20 − x 2)](κ+1)/2
2 2
Γ(κ/2) Γ[(1 − κ)/2] 1 1 ,

= −
4√π |ρ − ρ |κ (ρ + ρ0) κ
0
min (ρ , ρ)
0
x κ-1d x
⌠
⌡ [(ρ2 − x 2)(ρ20 − x 2)](κ+1)/2
0
Γ(κ/2) Γ[(1 − κ)/2] 1 1 .

= +
4√πρρ0 |ρ − ρ0| κ
(ρ + ρ0) 
κ
Some additional representations can be obtained by simple addition, subtraction,

integration or differentiation of those above.
Exercise 3.4
1. Find a solution of the generalized Abel equation (3.4.1), with f ( x )= C =const.
C sin[π(δ+κ)]
Answer: F ( t ) =
π( t − b )1-δ-κ( a − t )δ
2. Find a solution of the generalized Abel equation (3.4.1), with f ( x )= Cx, with
C =const.
C sin[π(δ+κ)][ t − b − δ( a − b ) + κ b ]
Answer: F ( t ) =
πκ( t − b )1-δ-κ( a − t )δ
3. Find a solution of the generalized Abel equation (3.4.1), with f ( x )= C x 2, with

C =const.
C sin[π(δ+κ)][2( t − b )( t + D ) + D 2 − δ( a 2 − b 2) + κ b 2]
Answer: F ( t ) = .
πκ(1 + κ)( t − b )1-δ-κ( a − t )δ
Here D =κ b −δ( a − b ).
3.5 Effect of a shearing force and a tilting moment

on a bonded circular punch.
Consider a circular flat punch of radius a , bonded to a transversely

isotropic elastic half-space z ≥0. The punch is subjected to a shearing force T ,
acting in the Ox direction, and a tilting moment M . We may assume, without
loss of generality, that the moment is oriented along the Oy axis. We need to
find the traction distribution under the punch, and to relate the translational ( u 0)
and angular (δ) displacements of the punch to the applied loading parameters.
The problem is considered in a separate section due to its practical importance.
The problem is characterized by the following boundary conditions on the
plane z =0:
u = u 0, for 0≤ρ≤ a , 0≤φ<2π;
w = −δρcosφ, for 0≤ρ≤ a , 0≤φ<2π;
σ = τ = 0, for a ≤ρ<∞ 0≤φ<2π. (3.5.1)
The governing integral equations, due to (2.5.6), (3.1.2) and (3.1.3), take the
3.5 Effect of a shearing force and a tilting moment 183
form
2G1 ρ x 4d x
a τ2(ρ0)dρ0 ρ
⌠ ⌠ 2π H α ⌠
− σ1(ρ0)ρ20dρ0
ρ2 ⌡ (ρ2 − x 2)1/2 ⌡ ρ0(ρ20 − x 2)1/2 ρ2
⌡
0 x 0
2G ρ ρ2 − 2 x 2
a τ0(ρ0)ρ0dρ0
⌠ dx ⌠
2
+ 2 1/2 = 0,
ρ2 ⌡ (ρ2 − x 2)1/2 ⌡ (ρ0 − x )
2
0 x
ρ a (ρ20 − 2 x 2) τ (ρ )dρ a
2G ⌠ ⌠ + 2π H α⌠ σ (ρ )dρ
dx 2 0 0
2
⌡ (ρ − x ) ⌡
2
2 1/2
ρ0(ρ20 − x 2)1/2 ⌡-1 0 0
0 x ρ
ρ a τ0(ρ0)ρ0dρ0
dx
+ 2G ⌠ 2 ⌠ 2 1/2 = u 0,
1
⌡ ( ρ − x 2 1/2
) ⌡ (ρ0 − x )
2
0 x
e-iφ ρ dρ a
⌠ i φ⌠ 0
2π H α ℜ τ0(ρ0)ρ0dρ0 − ρe τ2(ρ0)
ρ ⌡ ⌡ ρ0
 0 ρ
ρ
x 2d x
a σ1(ρ0)ei φ + σ-1(ρ0)e-i φ
4H⌠ ⌠
+ dρ0 = −δρcosφ.
ρ ⌡ (ρ2 − x 2)1/2 ⌡ (ρ20 − x 2)1/2
0 x
(3.5.2)
The structure of equations (3.5.2) is such that we may assume that σ = σ .
1 -1
The solution may be represented in the form
ρ
d⌠ f ( t )d t
σ1(ρ) = σ-1(ρ) = ,
dρ⌡ (ρ − t 2)1/2
2
0
a
C d ⌠ f(t)tdt D
τ0(ρ) = − + ,
ρ dρ ⌡ ( t 2 − ρ2)1/2 ( a − ρ2)1/2
2
ρ
d 1⌠ f(t)tdt 
a
2 a 2 − ρ2
τ (ρ) = − C ρ 
dρ ρ2 ⌡ ( t 2 − ρ2)1/2  − D .
2 ρ2( a 2 − ρ2)1/2
 ρ 
(3.5.3)
Here f is the as yet unknown stress function, and C and D are the constants to
be determined. Substitution of (3.5.3) in the first two equations of (3.5.2) make
them identities, provided that the following conditions are satisfied
a
α
D = ⌠ f ( t )d t ,
C
C = , (3.5.4)
γγ a⌡
1 2
0
a
π2 D
( G 1 + G 2) + 2π H α⌠
f ( t )d t
2 1/2 = u 0. (3.5.5)
⌡ (a − t )
2 2
0
Expressions (3.5.4) and (3.5.5) look contradictory: we have only two constants to
satisfy three equations. This will be clarified further. An additional constant,
representing the homogeneous solution, will appear in the expression for f .
Substitution of (3.5.3) in the third equation (3.5.2) leads to
2π H α  
a a
 −2 C ⌠ f(t)tdt
+ C ⌠ f ( t )d t + aD 
ρ ⌡ ( t 2 − ρ2)1/2 ⌡
 ρ 0 
ρ a
8H⌠ x 2( a 2 − x 2)1/2d x ⌠ f ( t )d t
+ = −δρ.
ρ⌡ (ρ2 − x 2)1/2 ⌡ (t 2
− x )( a 2 − t 2)1/2
2
0 0
Multiply the last expression by ρ2( r 2−ρ2)-1/2, and integrate with respect to ρ from
0 to r . The result is
a a
α2 ⌠ f ( t )d t δ
( a 2 − r 2)1/2⌠ 2
f ( t )d t
− = − .
⌡ (a − t ) (t − r )
2 1/2 2 2 γ 1γ 2 ⌡ t − r
2 2 πH
0 0
(3.5.6)
Equation (3.5.6) is similar to (3.2.60), with the general solution given by
3.5 Effect of a shearing force and a tilting moment 185
(3.2.62). The solution in this particular case is
δcosh2(πθ) 
f(t) = − tY s( t ) − θ aY c( t ) + AY c( t ). (3.5.7)
πH
2
sinh(πθ)  
The last term in (3.5.7) represents the homogeneous solution, with A being an
arbitrary constant. Substitution of (3.5.7) in (3.5.4) and (3.5.5) allows us to
define all the constants, namely,
πθα
D = A,
γ1γ2sinh(πθ)
πθ( G + G )
A = u0 +
δ a θα  π2 H α  1 2
-1.
1 +
 tanh(πθ)cosh(πθ) tanh(πθ)( G − G )
1 2
(3.5.8)
Formulae (3.5.7), (3.5.8) and (3.5.3) determine completely the traction distribution
under the punch. Now we need to relate the applied loading to the punch
displacements. Make use of the equilibrium conditions:
a 2π a
T = 2π⌠ τ0(ρ)ρdρ, M = −⌠ ⌠ [σ1(ρ)ei φ + σ-1(ρ)e-i φ]ρ2cosφ dρdφ.

⌡ ⌡ ⌡
0 0 0
After carrying out the calculations, we obtain
aθ α 4δ a 3θ(1 + θ2) 4π2 a 2θ2

T = 4π2 A , M = + A.
sinh(πθ) γ γ 3 H tanh(πθ) cosh(πθ)
1 2
(3.5.9)
Expressions (3.5.8) and (3.5.9) enable us to determine the displacements of the
punch
1
( G − G ) T −
(1 + 4θ2)tanh(πθ) 3Hα
u0 = π( G 1 + G 2) + M,
8a θ(1 + θ )2 1 2  4 a (1 + θ2)
2
δ =
3Hα  −T + M .
4 a (1 + θ )
2 2
a θ√γ γ 
1 2
(3.5.10)
In the case of isotropy, formulae (3.5.10) are in agreement (except for some
signs) with the results of Ufliand (1967), who seems to have used a different
sign convention. It is noteworthy that the tilting moment produces translational
displacement of the punch, even in the absence of the shearing force. The
shearing force, in turn, tilts the punch, even when no tilting moment is applied.
We shall see further (section 5.11) that a similar situation takes place in the case
of a finite friction between the punch and the elastic half-space.
Exercise 3.5
1. A flat circular punch of radius a is bonded to a transversely isotropic elastic
half-space z ≥0. A shearing force T is applied in the Oy direction. Find the
tilting angle δ.
3Hα
Answer: δ = T , with tilting about Ox axis. Note that the angle is
4 a 2(1 + θ2)
positive.
2. Subject to the conditions above, find the tilting moment needed to prevent
tilting.
Answer: M = − a θ√γ γ T .
1 2
3. A tilting moment M about the axis Ox is applied to a flat circular punch,

bonded to a transversely isotropic elastic half-space. Find the translational
displacement u and its direction.
3Hα
Answer: u = M , in the Oy direction.
4 a (1 + θ2)
2
4. A flat circular bonded punch is subjected to a shearing force T , acting in

the Ox direction, and a tilting moment M about the Oy axis. Find the normal
displacements outside the punch.
 4δcosh(πθ) a x Y c( x ) + θ axY s( x )
2
Answer: w (ρ,φ) = − ⌠ dx
πρ ⌡ (ρ2 − x 2)1/2
 0
a xY ( x )d x 
1 πθ a 
− ⌠
s
+ 4π HA tanh(πθ)
ρcosh(πθ) 2 1/2 cosφ.
⌡ (ρ − x ) 
2
0
5. Express the answer above in terms of the shearing force T and tilting
moment M .
a x 2 Y ( x ) + θ axY ( x )
 3 H αcosh(πθ)  M
−T  ⌠
c s
Answer: w (ρ,φ) = − dx
πρ a (1 + θ )θ a √γ γ
2 2
 ⌡ (ρ − x )
2 2 1/2
 1 2 0
a xY s( x )d x 
Hα  cosh(πθ) ⌠ 
+ T 1 −
ρ  2 1/2 cosφ.
πθ a ⌡ (ρ − x ) 
2
0 
6. Express the normal displacement w outside the punch in terms of the stress
function f .
8 H a 2
a
( a 2 − x 2)1/2 f ( t )d t
Answer: w (ρ,φ) =  ⌠ x d x ⌠
ρ ⌡ (ρ2 − x 2)1/2 ⌡ ( a 2 − t 2)1/2( t 2 − x 2)
 0 0
a
+ 4π H α D cosφ.
ρ

7. Express the tangential displacement u outside the punch in terms of the
stress function f .
Answer: u (ρ,φ) = u (ρ) + u (ρ)e2 i φ,
0 2
where
a
α ⌠ f ( x )d x -1 a
u (ρ) = π( G − G ) 2 1/2 + π( G 1 + G 2) D sin (ρ),
2 γ γ
1 2 ⌡
0 1 (ρ − x )
2
0
1
a
u 2(ρ) = 22π H α⌠ f ( x )d{ x [( a 2 − x 2)1/2 − (ρ2 − x 2)1/2]}

ρ
 ⌡
0

+ π( G + G ) Da (ρ2 − a 2)1/2.
1 2

8. Investigate the interaction of an arbitrary concentrated force, applied at some
point inside the transversely isotropic half-space, and a flat bonded circular punch.
3.6 Non-axisymmetric internal mixed-mixed problem
The general formulation of this problem is given in section 3.1, with the
boundary conditions (3.1.1), and the governing integral equations (3.1.2−3.1.5).
The closed form exact solution is not known at the moment. We assume that
all the parameters involved can be expanded in a Fourier series. An exact
solution for the n -th harmonic is presented below. The governing integral
equations for the n -th harmonic are
ρ 2n a G 1 x 2τn+1(ρ0) + G 2[2 n ρ2 − (2 n +1) x 2]τ-n+1(ρ0)

2 ⌠ x dx
⌠ dρ
ρn+1 ⌡ (ρ2 − x 2)1/2 ⌡ ρn0(ρ20 − x 2)1/2 0
0 x
ρ
2π H α ⌠
− σn(ρ0)ρn+1
0 dρ0 = F n+1(ρ), for n ≥0.
ρn+1 ⌡
0
ρ a G ρ20τ (ρ ) + G [(2 n −1)ρ20 − 2 n x 2]τ (ρ )

2 ⌠ x 2n-2d x
⌠
1 -n+1 0 2 n+1 0
dρ0
ρn-1 ⌡ (ρ2 − x 2)1/2 ⌡ ρn0(ρ20 − x 2)1/2
0 x
a
σ (ρ )dρ
+ 2π H αρ ⌠
-n 0 0
n-1
= F -n+1(ρ), for n ≥1;
⌡ ρn-1
0
ρ
e-in φ ρ τ (ρ )dρ 
a
⌠ n in φ⌠
n+1 0 0
2π H α ℜ n τ (ρ )ρ0dρ − ρ e
n

 ρ ⌡ -n+1 0 0 ⌡ ρ0
n

0 ρ
ρ σ-n(ρ0)e φ + σn(ρ0)e φ
a -in in
4H ⌠ x 2nd x
+ ⌠ dρ = ℜ{ein φΦ (ρ)},
ρn ⌡ (ρ − x )
2 2 1/2
⌡ ρn-1
0 (ρ0
2
− x)
2 1/2 0 n
0 x
for n ≥0. (3.6.1)
Here the right hand sides are known from the boundary conditions, and are
∞
F (ρ) = u (ρ) − 2ρ n+1⌠ 2n+2 2d x 2 1/2

n+1 n+1
⌡ x (x − ρ )
a
x G 1ρ20τn+1(ρ0) + G 2[2 n x 2 − (2 n +1)ρ20]τ-n+1(ρ0)

⌠ ρn0dρ0,
⌡ (x −
2
ρ20)1/2
a
∞ σ-n(ρ0)dρ0
F -n+1(ρ) = u -n+1(ρ) − 2π H αρ n-1 ⌠
⌡ ρn-1
0
a
∞ x G 1 x 2τ-n+1(ρ0) + G 2[(2 n −1) x 2−2 n ρ2]τn+1(ρ0)

− 2ρn-1⌠ 2n 2 ⌠
dx
ρn0dρ0,
⌡ x (x − ρ ) ⌡ ( x 2 − ρ20)1/2
2 1/2
a a
in φ in φ -in φ
 in φ n ∞ τn+1(ρ0)dρ0
ℜ{Φn(ρ)e } = w n(ρ)e + w -n(ρ)e + 2π H α ℜ  e ρ ⌠ 
⌡ ρn0
 a 
in φ -in φ
∞ x σn(ρ0)e + σ-n(ρ0)e
− 4 H ρn⌠ 2n 2 ⌠
dx
ρn+1
0 dρ0.
⌡ − ρ ⌡ ( x 2 − ρ20)1/2
2 1/2
x ( x )
a a
(3.6.2)
The case of axial symmetry ( n =0) was considered in detail in paragraph 3.2, and
is not discussed here. The solution is sought for n ≥1. We may assume,
without loss of generality, that the first two equations (3.6.1) are homogeneous.
This can be achieved by addition of some special solutions to the parameters
τ-n+1 and τn+1. These special solutions, satisfying the right hand sides of the first
two equations (3.6.1), can be obtained from the results of section 2.6. Of
course, this procedure will make the right hand side of the third equation (3.6.1)
more complicated.
Assume the solution of the set (3.6.1), with the first two equations
transformed into homogeneous, in the form
ρ t 2n-1d f n( t )
1 ⌠
σn(ρ) = σ-n(ρ) =
ρn ⌡ (ρ2 − t 2)1/2
0
ρ
1 d ⌠ t 2n-2[(2 n − 1)ρ2 − 2 n t 2]
= − n+1 f n( t )d t ;
ρ dρ ⌡ (ρ2 − t 2)1/2
0
a
f n( t ) t d t D nρn-1
d ⌠
τ-n+1(ρ) = − C ρn-2 + ;
dρ ⌡ ( t 2 − ρ2)1/2 ( a 2 − ρ2)1/2
ρ
 f n( t ) t d t 
a a
d 1 ⌠ d ⌠
dρ ρ2n⌡ d y ⌡ ( t 2 − y 2)1/2
τn+1(ρ) = C ρn 2n-2
y dy
 ρ y 
d 1 ⌠ t 2n-1d t 
a
dρ ρ2n ⌡ 
+ D nρ n
( a 2 − t 2)1/2
 ρ 
d ⌠ 
a a a
f n( t )d t
+ (2 n − 1)⌠ 2n 2 ⌠
dy
dρ ⌡ 
= −Cρ n
f ( t ) t 2n-2
d t
 t ( t 2 − ρ2)1/2 ⌡ y ( y − ρ2)1/2 ⌡ n 
ρ ρ y
− D n 2
ρn-1 2n ⌠ t 2n-1d t 
+ .
( a − ρ2)1/2 ρn+1 ⌡ ( a 2 − t 2)1/2
ρ
(3.6.3)
Here f n is the as yet unknown complex stress function, and C , D n are the
constants to be determined. In the following derivations we give in some cases
two equivalent expressions of the same parameter, for the sake of convenience in
the procedure of substitution into the governing equations (3.6.1). We present
first some intermediate results related to the substitution of (3.6.3) in the first
equation (3.6.1):
a τ (ρ )dρ f ( x) a f ( t )d t
⌠ n 2 πC  + π D,
− ⌠
-n+1 0 0 n n
=
⌡ ρ0(ρ0 − x )
2 1/2 2x  x ⌡ t 2
 2 ax n
x x
(3.6.4)
a τn+1(ρ0)dρ0 f n( x ) a
⌠ n 2 πC 2n − 1 ⌠ 2n-2  π a 2n-1

= + f ( t ) t d t − D.
⌡ ρ0(ρ0 − x )
2 1/2 2x  x x 2n ⌡ n  2 x 2n+1 n
x x
(3.6.5)
Substitution of (3.6.4) and (3.6.5) in the first equation (3.6.1) yields
ρ ρ
πC x 2n-2[2 n x 2 − (2 n −1)ρ2] 2π H α ⌠
( G 1− G 2) n+1 ⌠ f n( x )d x − σ (ρ )ρn+1
0 dρ0 = 0,
ρ ⌡ (ρ − x ) ρ ⌡ n 0
2 2 1/2 n+1
0 0
(3.6.6)
provided that the following condition holds
a
(2 n − 1) C ⌠
D = f ( t ) t 2n-2d t . (3.6.7)
n a 2n-1 ⌡ n
0
It is now easy to verify that substitution of the first expression (3.6.3) in (3.6.6)
makes it an identity if
α
C = . (3.6.8)
γγ
1 2
Here are some intermediary results related to the procedure of substitution of

(3.6.3) in the second equation (3.6.1):
a τ (ρ )dρ
⌠ n-2 2
-n+1 0 0 π 
2 1/2 = 2 C f n( x ) + D n , (3.6.9)
ρ
⌡ 0 0( ρ − x )  
x
a
(2 n − 1)ρ20 − 2 n x 2
π
⌠ τ (ρ )dρ = − C f ( x ) + D . (3.6.10)
⌡ ρn0(ρ20 − x) 2 1/2 n+1 0 0 2 n n
x
We used in transformations some general formulae from Appendix A3.3.

Substitution of (3.6.9) and (3.6.10) in the second equation (3.6.1) reduces it to
ρ [ C ( G − G ) f ( x ) + ( G + G ) D ] x 2n-2d x a
σ (ρ )dρ
π ⌠
+ 2π H αρ ⌠
1 2 n 1 2 n -n 0 0
n-1
= 0.
ρn-1 ⌡ (ρ − x )
2 2 1/2
⌡ ρn-1
0
0 ρ
(3.6.11)
Substitution of the first expression (3.6.3) in (3.6.11) makes it an identity, subject
to (3.6.8), and an additional condition
a
√πΓ( n − 1/2) 2Hα ⌠ x 2n-2d x
( G 1 + G 2) D n + 2n-2 f ( x) = 0. (3.6.12)
2Γ( n ) a ⌡ n ( a 2 − x 2)1/2
0
The conditions (3.6.7) and (3.6.12) might look contradictory. It will be shown
further (see 3.6.14) that this is not so, because an additional constant will appear
in the expression for f , corresponding to a homogeneous solution of the singular
n
integral equation (3.6.14).
By now we have satisfied the first two equations (3.6.1). Substitution of
(3.6.3) in the third equation (3.6.1) requires the following transformation:
ρ
a dρ 0 t 2n-1d f ( t ) a
⌠ 2n-1 2 = ⌠ ( a 2 − x 2)1/2( a 2 − t 2)1/2 Q ( x , t )

0 n
2 1/2 ⌠
⌡ ρ0 (ρ0 − x ) ⌡ (ρ20 − t) ⌡ 
2 1/2 n
x 0 0
+ ψ ( x , t ) ln
a | t 2 − x 2|1/2  t 2n-1d f ( t ).
n |t(a2 − x ) − x( a − t ) | 
2 1/2 2 2 1/2 n
(3.6.13)
Here Q ( x , t ) is a polynomial in even negative powers of x and t . Although we
n
cannot write the explicit expression for Q ( x , t ), it can be computed in elementary
n
manner for any particular n . The explicit expression for ψ is
n
n-1
ψn( x , t ) =
1
π tx 2n-1 Σk=0
Γ( n − k −1/2) Γ( k +1/2) x 2k
Γ( n − k ) Γ( k +1) t 
.
Multiply both sides of the third equation (3.6.1) by ρn, differentiate with respect
to ρ, divide the result by ρ2n-2( r 2−ρ2)1/2 and integrate with respect to ρ from 0
to r . This procedure allows us to split the kernel of the integral equation into
two parts: a singular one and a degenerate one. The result takes the form
a f ( t )d t a f ( t )d t
α2 ⌠
(a − r ) ⌠ 2
n n
2 2 1/2
− = χ ( r ).
⌡ ( t − r )( a − t )
2 2 2 1/2 γ 1γ 2 ⌡ t 2
− ρ2 n
0 0
(3.6.14)
Here
r
1 ⌠ dρ d n
χn( r ) = [ρ Φ (ρ)]
4π Hr ⌡ ρ ( r − ρ ) dρ
2n-2 2 2 1/2 n
0
r ρ a
2⌠ dρ d ⌠ x 2n( a 2 − x 2)1/2d x ⌠ 2 2 1/2
− ( a − t ) Q ( x , t ) t 2n-1d f ( t )
π r ⌡ ρ2n-2( r 2 − ρ2)1/2 dρ⌡ (ρ − x )
2 2 1/2
⌡ n n
0 0 0
r ρ a
−
2 ⌠ 2n-2 2dρ 2 1/2 ⌠ 2 dx ⌠ q (ρ, x , t )( a 2 −
π r 3/2
⌡ ρ (r − ρ ) ⌡ (ρ − x ) ( a − x )
2 1/2 2 2 1/2
⌡ n
0 0 0
2 1/2 2n-2
t) t d f ( t ),
n
(3.6.15)
with
n-2
q n(ρ, x , t ) = Σ
k=0
Γ( n − k −1/2)ρ2k
Γ( n − k )  t 
1 3 x2
F (2− n + k , ; − n + k ; 2 ).
2 2 t
(3.6.16)
Note that the hypergeometric function in (3.6.16) is, in fact, a polynomial, and
that all the integrals with respect to x and ρ in the degenerate part of the
kernel (3.6.15) are computable in elementary function for any n . The integral
equation (3.6.14) was solved in paragraph 3.2, and the solution is
a χn( r ) Y c( r )d r
f n( t ) = − 2cosh2(πθ) t  tY c( t )⌠
4
π  ⌡ r2 − t 2
0
a χn( r ) Y s( r ) r d r
+ Y ( t )⌠  + A Y ( t ). (3.6.17)
s
⌡ r − t2 2
 n c
0
The last term in (3.6.17) represents the homogeneous solution, with A being an
n
arbitrary constant. Its value, along with the constant D and others which appear
n
due to the degenerate part of the kernel, can be found from the appropriate
system of linear algebraic equations and the conditions (3.6.7) and (3.6.12). The
general solution may be considered completed. The main handicap of the
solution is the necessity for solving a set of linear algebraic equations whose
order increases with n , thus making the exact solution for higher harmonics very
cumbersome. We are not aware of any other solution for a transversely
isotropic solid. The corresponding problem for an isotropic body was solved by
Ufliand (1967) who used the method of Mehler-Fok integral transform. It has
the same hindrance: one needs to solve a set of linear algebraic equations, whose
order increases with n .
Example. Consider the action of a normal concentrated load P applied
outside a flat circular punch of radius a , bonded to a transversely isotropic
elastic half-space z ≥0. We may assume, without loss of generality, that the force
is applied at the point ρ= b , φ=0 ( b > a ). Thus, the boundary conditions are
u = w =0, for ρ≤ a ;
σ = P δ(ρ− b )δ(φ−0)/ρ, τ = 0, for ρ> a .
The boundary conditions yield
ρn-1
F (ρ) = 0, for n ≥0; F = − PH α , for n ≥1;
n+1 -n+1 bn
ρ
4 PH ⌠ x 2nd x
Φn(ρ) = − .
π(ρ b )n ⌡ (ρ2 − x 2)1/2( b 2 − x 2)1/2
0
(3.6.18)
We present now the explicit solutions for some specific values of n . In the
axisymmetric case n =0, the following results may be obtained:
ρ f 0( t ) t d t a
f 0( t )d t
1 d ⌠ α d ⌠
σ (ρ) = , τ (ρ) = − ,
0 ρ dρ ⌡ (ρ2 − t 2)1/2 1 γ γ dρ ⌡ ( t 2 − ρ2)1/2
1 2
0 ρ
a Y ( r )d r
2 
f ( t ) = 3 P cosh (πθ) tY ( t )
2 ⌠
c
0 π  c ⌡ ( b − r )1/2( r2 − t 2)
2 2
0
a Y ( r) rdr
+ Y ( t )⌠
s

2 . (3.6.19)
⌡ ( b − r ) ( r − t )
s 2 2 1/2 2
0
The results for n =1 are:

ρ f 1( t )d t
d ⌠
σ1(ρ) = ,
dρ ⌡ (ρ2 − t 2)1/2
0
a
f (t)tdt D
α 1 d ⌠ 1 1
τ0(ρ) = − + ,
γ1γ2 ρ dρ ⌡ ( t − ρ )
2 2 1/2
( a − ρ2)1/2
2
a
f (t)tdt
α d 1 ⌠ 1
 − D 2 a 2 − ρ2 ,
τ2(ρ) = − ρ
γ 1γ 2 dρ ρ2 ⌡ ( t 2 − ρ2)1/2 1ρ2( a 2 − ρ2)1/2
ρ
t
f (t) = f ( t ) + A Y ( t ),
1 b 0 1 c
πθ( G 1 + G 2)
Pθ  cosh(πθ)  -1,
D = − 1 − I (b) 1 +
1 π b √γ γ  πaθ s  tanh(πθ)( G 1 − G 2)
1 2
P cosh2(πθ) cosh(πθ)
A = − [ I s( b ) − 2 a θ I c( b )] + D √γ γ .
1 π θ ab
3 πθ 1 1 2
(3.6.20)
We recall that the notations I c,s are defined by (3.2.73) and (3.2.74) respectively.
The case of n =2 is more cumbersome:
ρ
1 d ⌠ 4 t 2 − 3ρ2 2
σ2(ρ) = 3 t f ( t )d t ),
ρ dρ ⌡ (ρ2 − t 2)1/2 2
0
a
f (t)tdt
d
− D ( a 2 − ρ2)1/2,
α ⌠ 2
τ (ρ) = −
-1 dρ γ1γ2⌡ ( t − ρ2)1/2
2 2 
ρ
d 1 
a
τ3(ρ) = ρ  4 2 1/2
α ⌠ ρ2 − 2 t 2 2 a 2 + ρ2 2

2
f ( t ) t d t + D ( a − ρ ) ,
dρ ρ  γ1γ2 ⌡ ( t 2 − ρ2)1/2 2 2 3
 ρ 
a χ ( r ) Y ( r )d r a χ ( r) Y ( r) rdr
f ( t ) = − 2 cosh2(πθ) t  tY ( t )⌠ + Y ( t )⌠ 
4 2 c 2 s
2 π  c ⌡ r − t
2 2 s
⌡ r − t
2 2

0 0
2coth(πθ)2θ
Y ( t ) B .
1
+ A Y (t) + Y c( t ) −
2 c πa  a t s  2
The constant B corresponds to the degenerate part of the kernel:

2
⌠ 2t2 − a2
B2 = 2 1/2 f 2( t )d t .
⌡ (a − t )
2
0
All the constants are determined as follows:
3α  π a 3θ 1 − 2θ2 2 a θ2(1 + 4θ2)cosh(πθ)

D2 = L + A − B ,
γ 1γ 2 a 3 1 2 sinh(πθ) 3 2 3sinh2(πθ) 
cosh(πθ) 4θ(1 + 2θ2)

A2 = L − B ,
2π a 2θ2  3 2 sinh(πθ) 
 2 Pa 2 3π( G 1 + G 2)
B 2 =  2θ  + L 1 − θ2
2 + L2 + L1 4a(G − G )
 2πb  34
1 2
( G 1 + G 2)   πθ(1 + θ2)( G 1 + G 2) -1

(1 − 2θ2)
θ   ,
tanh(πθ)( G 1 − G 2) 
+ 1 +
4( G 1 − G 2) sinh(πθ) 
a
2cosh2(πθ) ⌠ 
2 a θ r 2 + a 2
1 − 2θ2
L1 = Y ( r)
πsinh(πθ) ⌡   3  c
0
+ r (2 a 2θ2 − r 2) Y s( r ) χ2( r )d r ,

a
 21 
a 4 − θ  + 2 Y c( r) + θ arYs( r)χ2( r)d r,
r2
cosh(πθ)⌠
4 2
L2 =
π ⌡
0  
cosh(πθ) ⌠ [( r 2 − 2 a 2θ2) Y c( r ) + 2θ arY s( r )]χ2( r )d r ,

4
L3 =
π ⌡
0
P  1 1 
χ2( r ) = − 2 1/2 + 2 1/2 .
2π b ( b − r )
2 2
b + (b − r ) 
2
In order to provide zero displacements, the punch should be loaded by a normal

force N , a shearing force T , directed in the negative Ox direction, and a tilting
moment M about the Oy axis. Their relationship with the force P can be
established by the statics equations, namely,
a a
T = 2π⌠ τ0(ρ)ρdρ = 2π f 1( t )d t + D 1 a ,

α ⌠
⌡ γ γ
1 2 ⌡

0 0
a a f 0( t ) t d t
N = 2π⌠ σ0(ρ)ρdρ = 2π⌠ ,
⌡ ⌡ (a − t )
2 2 1/2
0 0
2π a a
a2 − 2t2
M = −⌠ ⌠ [σ1(ρ)e iφ -i φ
+ σ-1(ρ)e ]ρ cosφdρdφ = 2π⌠ 2
f ( t )d t .
⌡ ⌡ ⌡ ( a 2 − t 2)1/2 1
0 0 0
Performing the integrations, we get
a Y c( r )d r
cosh(πθ)⌠
2P
T = 4π aD 1, N = − ,
π ⌡ ( b 2 − r 2)1/2
0
a r 2 Y c( r ) + θ arY s( r )
cosh(πθ)⌠
4P
M = d r + 4π a 2θ√γ1γ2D 1.
πb ⌡ ( b 2 − r 2)1/2
0
(3.6.21)
If the punch is not loaded then it will undergo the translational, normal, and
angular displacements. Their values can be determined from (3.2.71), (3.5.10)
and (3.6.21).
Exercise 3.6
1. Subject to the conditions in the example above (page 200), find the normal
traction at the punch centre.
a [ aY ( x ) + 2θ xY ( x )]d x
P cosh2(πθ)⌠ c s πθ 
Answer: σ(0) = 2 1/2 − sinh(πθ) .
π ab ⌡ [ b + ( b − x ) ]( b − x )
2 2 2 1/2 2

0
2. Subject to the conditions in the example above, find the shear traction at the
punch centre.
a Y ( x )d x D πθ A
Pα  2 ⌠
s
 1 1
Answer: τ(0) = 1 − b coth( πθ) + + .
2π b 2γ γ  π ⌡ x( b − x ) 
2 2 1/2 a √γ γ
1 2 0 1 2
3. Subject to the conditions in the example above, find the traction distribution
under the punch in the limiting case α=0.
Answer: σ (ρ) = −
P ( b 2 − a 2)1/2 ρn, τ (ρ) = 0.
n π ( a − ρ ) ( b − ρ ) b 
2 2 2 1/2 2 2 n
4. No loading is applied to a flat circular punch of radius a bonded to a

transversely isotropic half-space. A normal concentrated force P is applied at the
point ρ= b , φ=π/2. Find the normal component w of the punch displacement.
a Y ( r )d r
PH sinh(πθ) ⌠ c
Answer: w = 2 1/2.
πθ a ⌡ (b − r )
2
0
Hint : use (3.2.71) and (3.6.21)
5. Subject to the conditions of Exercise 4, find the tangential component u of

the punch displacement and its direction.
a x 2 Y ( x ) + θ axY ( x )
d x + PH α 1
3 PH αcosh(πθ) ⌠ c s 1
Answer: u =
π b a (1 + θ ) ⌡
2 2
(b − x )
2 2 1/2 b 
0
a xY ( x )d x
−
cosh(πθ) ⌠ s

2 1/2 , in the Oy direction.
πθ a ⌡ (b − x ) 
2
0
6. Subject to the conditions of Exercise 4, find the tilting angle δ of the

punch and its direction.
a Y c( x ) x 2 + θ axY s( x )
3 PH sinh(πθ) ⌠
Answer: δ = dx, in the positive
π a 3 b θ(1 + θ2) ⌡ ( b 2 − x 2)1/2
0
direction about the Ox axis.
7. No loading is applied to a flat circular punch of radius a bonded to a

transversely isotropic half-space. A tangential concentrated force T is applied at
the point ρ= b , φ=0 in the positive Ox direction. Find the normal component w
of the punch displacement.
√γ1γ2 I s( b ) −
TH sinh(πθ) πθ a 
Answer: w = , with I s defined by (3.2.74).
πθ ab  cosh(πθ)
Hint : use (3.2.72) and the reciprocal theorem.
8. Subject to the general boundary conditions (3.1.1), find the tangential

displacements for ρ> a , expressed in terms of the stress function f n.
a (α/γ1γ2)( G 1 − G 2) f n( x ) + ( G 1 + G 2) D n
π
Answer: u -n+1(ρ) = n-1 ⌠ x 2n-2d x ,
ρ ⌡ (ρ − x )
2 2 1/2
0
a
2π H α ⌠
u n+1(ρ) = f ( x )d{ x 2n-1[( a 2 − x 2)1/2 − (ρ2 − x 2)1/2]}
ρn+1 ⌡ n
0
(ρ2 − a 2)1/2
+ π( G 1 + G 2) D n a 2n-1 .
ρn+1
9. Subject to the general boundary conditions (3.1.1), find the normal

displacements for ρ> a , expressed in terms of the stress function f n.
in φ -in φ Γ( n ) a 2n-1 in φ
Answer: ℜ{ w n(ρ)e + w -n(ρ)e } = 2π3/2 H α n ℜ{ D n e }
Γ( n + 1/2)ρ
ρ
 a
x 2n
d x
a dρ0 0 t 2n-1d f n( t ) 
in φ
+ n ℜ e ⌠ ⌠
8H
⌠ .
ρ ⌡ (ρ2 − x 2)1/2 ⌡ ρ2n-1 (ρ20 − x 2)1/2 ⌡ (ρ20 − t 2)1/2
 0 x
0
0

10. Investigate the interaction of an arbitrary tangential force with a bonded

axisymmetric punch.
3.7 Non-axisymmetric external mixed-mixed problem
The general formulation of the problem is given in section 3.1, with the
boundary conditions (3.1.6), and the governing integral equations (3.1.7−3.1.10).
We assume that all the parameters involved can be expanded in a Fourier series.
An exact solution for the n -th harmonic is presented below. The governing
integral equations for the n -th harmonic are
∞
dx
xG ρ2τ (ρ ) + G [2 n x 2 − (2 n +1)ρ20]τ (ρ )
2ρn+1⌠
1 0 n+1 0
⌠
2 -n+1 0
ρn0dρ0
⌡x 2n+2
(x − ρ )
2 2 1/2
⌡ (x −
2
ρ20)1/2
ρ a
ρ
2π H α ⌠
− σn(ρ0)ρn+1
0 dρ0 = F n+1(ρ), for n ≥0.
ρn+1 ⌡
a
∞
dx
xG x 2τ-n+1(ρ0) + G 2[(2 n −1) x 2 − 2 n ρ2]τn+1(ρ0)
2ρn-1⌠ ⌠
1
ρn0dρ0
⌡ x (x − ρ )
2n 2 2 1/2
⌡ (x −
2
ρ20)1/2
ρ a
∞
σ-n(ρ0)dρ0
+ 2π H αρ ⌠ n-1
= F (ρ), for n ≥1;
⌡ ρn-1
0
-n+1
ρ
e-in φ ρ
∞
τ (ρ )dρ 
2π H α ℜ n ⌠ τ (ρ )ρ0dρ − ρ e
n n in φ⌠ n+1 0 0

ρ ⌡
-n+1 0 0
⌡ ρ0
n

a ρ
∞ x σ (ρ )e-in φ + σ (ρ )ein φ
dx
+ 4 H ρn⌠ ⌠ in φ
-n 0 n 0
ρn+1
0 dρ = ℜ{e Φ (ρ)},
⌡ x 2n( x 2 − ρ2)1/2 ⌡ ( x 2 − ρ20)1/2 0 n
ρ a
for n ≥0. (3.7.1)
The right hand sides in (3.7.1) are known from the boundary conditions, and are
a
2π H α⌠
F n+1(ρ) = u n+1(ρ) + σ (ρ )ρn+1dρ0
ρn+1 ⌡ n 0 0
0
a
2n
a G 1 x 2τn+1(ρ0) + G 2[2 n ρ2 − (2 n +1) x 2]τ-n+1(ρ0)
2 ⌠ x dx ⌠
− dρ0,
ρ ⌡ (ρ − x 2)1/2 ⌡
n+12
ρn0(ρ20 − x 2)1/2
0 x
a
x 2n-2d x
F -n+1(ρ) = u -n+1(ρ) − n-1 ⌠ 2
2
ρ ⌡ (ρ − x 2)1/2
0
a G 1ρ20τ-n+1(ρ0) + G 2[(2 n −1)ρ20−2 n x 2]τn+1(ρ0)

×⌠ dρ0,
⌡ ρn0(ρ20 − x 2)1/2
x
in φ in φ -in φ
e-in φ a 
ℜ{Φn(ρ)e } = w n(ρ)e + w -n(ρ)e + 2π H α ℜ n ⌠ τ-n+1(ρ0)ρn0dρ0
 ρ 0⌡ 
in φ -in φ
a a σn(ρ0)e + σ-n(ρ0)e
x 2nd x
− n⌠ 2 ⌠
4H
dρ0.
ρ ⌡ (ρ − x 2)1/2 ⌡ ρn-1
0 (ρ0 − x )
2 2 1/2
0 x
(3.7.2)
The case of axial symmetry ( n =0) was considered in detail in section 3.3, and is
not discussed here. The solution is sought for n ≥1. We may assume, without
loss of generality, that the first two equations (3.7.1) are homogeneous. This
can be achieved by the addition of some special solutions to the parameters τ-n+1
and τn+1. These special solutions, satisfying the right hand sides of the first two
equations (3.7.1), can be obtained from the results of section 2.7. Of course,
this procedure will make the right hand side of the third equation (3.7.1) more
complicated.
Assume the solution of (3.7.1), with the first two equations transformed into
homogeneous ones, in the form
∞
df (t)
σn(ρ) = σ-n(ρ) = ρ ⌠ 2n 2
n
n
,
⌡ t ( t − ρ2)1/2
ρ
ρ f ( t )d t Dn
C d ⌠ n
τn+1(ρ) = n − n+1 2 ,
ρ dρ ⌡ (ρ2 − t 2)1/2 ρ (ρ − a 2)1/2
a
ρ
C d  2n⌠ d y d ⌠
y f n( t )d t 
τ-n+1(ρ) = n ρ 2 1/2
ρ dρ ⌡ y dy ⌡ (y − t )
2n 2
 a a 
ρ
Dn 
d 2n⌠ dt 
+ n  ρ .
ρ dρ ⌡ t 2n+1( t 2 − a 2)1/2
 a 
(3.7.3)
Here f is the as yet unknown complex stress function, and C , D are the
n n
constants to be determined. Substitution of (3.7.3) in the first two equations
(3.7.1) satisfies them identically, provided that the following conditions hold
∞ f (t)
α ⌠ 2n+1 d t .
n
C = , D = −2 n C a 2n+1
(3.7.4)
γ 1γ 2 n
⌡ t
a
D + 2π H α a 2n+2 ⌠  x f ( x )
π3/2Γ( n + 1)
( G 1 + G 2)
2Γ( n + 3/2) n ⌡ n
a
− ⌠ f n( t )d t  2n+2 2
dx
= 0. (3.7.5)
⌡ x ( x − a 2)1/2
a
Some general formulae from Appendix A3.3 were used in transformations. The
conditions (3.7.4) and (3.7.5) might appear contradictory. It will be shown below
(see 3.7.11) that this is not so, because an additional constant will appear in the
expression for f , due to a homogeneous solution of the integral equation (3.7.8).
n
So far, we have satisfied the first two equations (3.7.1). Substitution of

(3.7.3) in the third of equations (3.7.1) requires the following transformation:
∞
x ρ2n+1 dρ df (t) ∞
⌠ 2
0 0 ⌠ n
= ⌠ ( x 2 − a 2)1/2( t 2 − a 2)1/2 Q ( x , t )
⌡ ( x − ρ0)
2 1/2 ⌡ t ( t − ρ0)
2n 2 2 1/2
⌡  n
a ρ a
0
df (t)
|( x 2 − a 2)1/2 + ( t 2 − a 2)1/2| n
+ ψ ( x , t ) ln .
n | t 2 − x 2|1/2  t 2n
(3.7.6)
Here Q ( x , t ) is a polynomial in even powers of x and t . Although we cannot
n
write the explicit expression for Q ( x , t ), it can be computed in elementary
n
manner for any particular n . The explicit expression for ψ is
n
ψn( x , t ) =
x 2n
π Σ k=0
Γ( n − k +1/2) Γ( k +1/2)  t 2k
Γ( n − k +1) Γ( k +1) x 
. (3.7.7)
Divide both sides of the third of equations (3.7.1) by ρn, differentiate with
respect to ρ, multiply the result by ρ2n/(ρ2− r 2)1/2 and integrate with respect to ρ
from r to ∞. This procedure allows us to split the kernel of the integral
equation into two parts: a singular one and a degenerate one. The result takes
the form
∞ f (t)tdt ∞ f ( t )d t
( r 2 − a 2)1/2 ⌠ n α2 ⌠ n
− + = χ ( r ).
r ⌡ ( t − r )( t − a )
2 2 2 2 1/2 γ1γ2 ⌡ t 2 − ρ2 n
a a
(3.7.8)
Here
∞ Φ (ρ)
1 ⌠ ρ2ndρ d n 
χn( r ) =
4π H ⌡ (ρ2 − r 2)1/2 dρ ρn 
r
∞ ∞ ∞ df (t)
2 ⌠ ρ2ndρ d ⌠ ( x 2 − a 2)1/2d x ⌠ 2 n
− ( t − a 2)1/2 Q ( x , t )
π ⌡ (ρ − r ) dρ ⌡
2 2 1/2
x 2n( x 2 − ρ2)1/2 ⌡ n t 2n
r ρ a
∞ ∞ ∞ df (t)
2 ⌠ ρ2n-1dρ ⌠ d x ⌠
n
+ 3/2 q (ρ, x , t )( t − a )
2 2 1/2
2n-2 ,
π ⌡ (ρ − r )
2 2 1/2
⌡ (x − ρ ) (x − a ) ⌡
2 2 1/2 2 2 1/2 n t
r ρ a
(3.7.9)
with
n-1
q (ρ, x , t ) =
n Σ
k=0
Γ( n − k +1/2) t 2k
Γ( n − k +1) ρ
1 1 t2
F ( 1− n + k , ; − n + k ; 2 ) .
2 2 x
(3.7.10)
Note that the hypergeometric function in (3.7.10) is, in fact, a polynomial, and
that all the integrals with respect to x and ρ in the degenerate part of the
kernel (3.7.9) are computable in elementary function for any n . The integral
equation (3.7.8) was solved in paragraph 3.3, and the solution is
∞ χ ( r ) Y ( r )d r ∞ χ ( r) Y ( r) rdr
f ( t ) = 2cosh2(πθ) t  tY ( t )⌠ + Y ( t )⌠  + A Y ( t ).
4 n c n s
n π  c ⌡ r2 − t 2 s
⌡ r2 − t 2  n c
a a
(3.7.11)
The last term in (3.7.11) represents the homogeneous solution, with A n being an
arbitrary constant. Its value, along with the constant D and others which appear
n
due to the degenerate part of the kernel, can be found from the appropriate
system of linear algebraic equations and the conditions (3.7.4) and (3.7.5). The
general solution may be considered completed.
Example. Consider the action of a normal concentrated load P applied at
some point inside the circle ρ= a , with the rest of the plane z =0 being clamped.
We may assume, without loss of generality, that the force is applied at the point
ρ= b , φ=0 ( b < a ). Thus, the boundary conditions are
u = w =0, for ρ> a ;
σ = P δ(ρ− b )δ(φ−0)/ρ, τ = 0, for ρ< a .
We determine in this case
bn
F -n+1(ρ) = 0, for n ≥1; F n+1 = PH α , for n ≥0;
ρn+1
Φ (ρ) = − PH (ρ b )n ⌠ 2n 2
4 dx
.
n π ⌡ x ( x − ρ )1/2( x 2 − b 2)1/2
2
ρ
(3.7.12)
We present now the explicit solutions for some specific values of n . In the
axisymmetric case n =0, the following results may be obtained:
∞ ∞
d f 0( t ) f 0( t ) t d t
σ0(ρ) = ⌠
1 d ⌠
= ,
⌡ ( t 2 − ρ2)1/2 ρ dρ ⌡ ( t 2 − ρ2)1/2
ρ ρ
ρ f 0( t )d t D0
α d ⌠
τ1(ρ) = − ,
γ1γ2 dρ ⌡ (ρ2 − t 2)1/2 ρ(ρ2 − a 2)1/2
a
∞ Y c( t ) rY c( r ) + tY s( t ) Y s( r )
f 0( t ) = 3 cosh2(πθ)⌠
2P
dr
π ⌡ ( r 2 − t 2)( r 2 − b 2)1/2
a
γ 1γ 2
− D0 [1 − Y c( t )],
aα
∞ Y s( r )d r
P αsinh(πθ) .
coth(πθ)⌠
2
D0 = − 1 −
4π θγ1γ2 
2 π ⌡ (r − b ) 
2 2 1/2
a
The results for the first harmonic ( n =1) are
∞
df (t)
σ1(ρ) = σ-1(ρ) = ρ⌠ 2 2
1
,
⌡ t ( t − ρ2)1/2
ρ
ρ f 1( t )d t D1
α 1 d ⌠
τ2(ρ) = − ,
γ1γ2 ρ dρ ⌡ (ρ2 − t 2)1/2 ρ2(ρ2 − a 2)1/2
a
ρ ρ
1 d  2 α ⌠ ,
y f 1( t )d t
dy d ⌠ ⌠ dy
ρ dρ γ1γ2 ⌡
τ0(ρ) = ρ + D
y 2 d y ⌡ ( y 2 − t 2)1/2 ⌡ y ( y − a ) 
1 3 2 2 1/2
 a a a
∞χ ( r ) Y ( r )d r ∞χ ( r ) Y ( r ) r d r
f 1( t ) = 2cosh2(πθ) t  tY c( t )⌠ + Y s( t )⌠  + A Y ( t ),
4 1 c 1 s
π  ⌡ r − t
2 2
⌡ r − t
2 2
 1 c
a a
Pb  1 1  + 21 − ( r2 − a 2)1/2B .
χ ( r) = +
1 2π r ( r 2 − b 2)1/2 r + ( r 2 − b 2)1/2 π r  1
The constants D , A , and B are to be determined from the set of linear

1 1 1
algebraic equations
∞ f ( t )d t ∞
2α a 3 ⌠ ( t 2 − a 2)1/2
= ⌠
1
D1 = − , B d f ( t ),
γ 1γ 2 ⌡ t3 1
⌡ t2 1
a a
∞ x
( G + G ) D + 2π⌠  xf ( x ) − ⌠ f ( t )d t  4 2
2π dx
− − Pb = 0.
3Hα 1 2 1
⌡  1
⌡ 1 x ( x − a 2)1/2
a a
It should be noted that the system of tractions in the clamped part is such that
its resultant vector is exactly equal to P . This can be shown by a direct
integration.
Exercise 3.7
1. Find the solution of the example above (page 211) in the limiting case b =0.
Answer: the only non-zero stress function is
2cosh (πθ) Y ( t ) 2
sinh(2πθ)[1 − Y ( t )]
P s c
f (t) =
0 2π πsinh(πθ) t
−
2πθ[1 + cosh(πθ)] a 
.
 
2. Find the solution of the example above in the limiting case α=0.
Answer: σ (ρ) = − 2 2
P ( a 2 − b 2)1/2 b n, τn = 0.
n π (ρ − a ) (ρ − b ) ρ
2 1/2 2 2
3. Prove that the tractions in the plane z =0 are in equilibrium when the
exterior is clamped.
Note : this is not the case in internal problems.
4. Try to find an exact closed form solution to the mixed-mixed boundary

value problem.
Appendix A3.1 207
Appendix A3.1
Some integrals, related to solving internal mixed-mixed problem, are

presented here. The notations Y c,s are defined by (3.2.45). It is assumed that
0< r < a .
Integrals, containing Y c:
a Y c( x )d x
⌠ 2 π
2 = − 2 r coth(πθ) Y s( r ),
⌡ x − r
0
a Y ( x )d x
⌠  a 2 − r 2 π
2 2 1/2 c
2 = − 2 r tanh(πθ) Y s( r ),
⌡ a − x  x − r
2
0
a Y c( x ) x 2 d x
⌠ πθ a π
= − r coth(πθ) Y s( r ),
⌡ x − r
2 2 sinh(πθ) 2
0
a Y c( x ) x 4 d x
⌠ π 3 πθ a  2 2 1 − 2θ 
2
= − r coth(πθ) Y s( r ) + r + a ,
⌡ x2 − r2 2 sinh(πθ) 3 
0
a ( a 2 − x 2)1/2 Y c( x )d x
⌠ π 2 π
= − ( a − r 2)1/2tanh(πθ) Y s( r ) − ,
⌡ x − r
2 2 2r 2cosh(πθ)
0
a x 2( a 2 − x 2)1/2 Y c( x )d x
⌠ πa2 1
= ( + θ2)
⌡ x2 − r2 cosh(πθ) 4
0
π 2 r2 
− r ( a − r 2)1/2tanh(πθ) Y s( r ) + ,
2 cosh(πθ)
a x 2 Y c( x )d x
⌠ π π r
2 = 2cosh(πθ) − 2 tanh(πθ) 2 Y ( r) ,
⌡ (a − x ) (x − r ) ( a − r 2)1/2 s
2 2 1/2 2
0
a a
⌠ Y ( x )d x = πθ a , ⌠ x 2 Y ( x )d x = πθ(1 − 2θ ) a ,
2 3
⌡ c sinh(πθ) ⌡ c 3sinh(πθ)
0 0
a a Y c( x )d x
⌠ ( a − x ) Y ( x )d x = πa (1 + θ2), π
2
2 2 1/2 ⌠ 2 1/2 = 2cosh(πθ),
⌡ c cosh(πθ) 4 ⌡ (a − x )
2
0 0
a x 2 Y c( x )d x
⌠ πa 1 2
2 1/2 = cosh(πθ)(4 − θ ),
2
⌡ ( a 2
− x )
0
a 1 − Y c( x )
⌠ πθcoth(πθ) − 1
dx = .
⌡
2 a
x
0
Integrals, containing Y s:
a xY s( x )d x
⌠ 2 π π
2 = 2 coth(πθ) Y c( r ) − 2sinh(πθ),
⌡ x − r
0
a Y s( x )d x Y c( r ) − 1
⌠ π
2 = 2 coth(πθ) ,
⌡ x( x − r )
2
r2
0
a x 3 Y s( x )d x
⌠ πθ2 a 2 πr2
= + [cosh(πθ) Y c( r ) − 1],
⌡ x2 − r2 sinh(πθ) 2sinh(πθ)
0
a Y ( x )d x Y ( r)
⌠
s π
πθ  c 1
= tanh( ) 2 1/2 − a ,
⌡ x( a − x ) ( x − r )
2 2 1/2 2 2
2r 2
( a − r )
2

0
Appendix A3.1 209
a x 3 Y s( x )d x
⌠ πθ a πr 2
2 = cosh(πθ) + tanh(πθ) Y c( r ),
⌡ (a − x ) (x − r ) 2( a − r 2)1/2
2 2 1/2 2 2
0
a x Y ( x )d x
⌠  a 2 − r 2 π
2 2 1/2 s
2 = 2 tanh(πθ) Y c( r ),
⌡ a − x  x − r
2
0
a x ( a 2 − x 2)1/2 Y s( x )d x
⌠ π πθ a
= tanh(πθ) ( a 2 − r 2)1/2 Y c( r ) − ,
⌡ x2 − r2 2 cosh(πθ)
0
a a
⌠ x Y ( x )d x = πθ a , ⌠ x 3 Y ( x )d x = πθ (2 − θ ) a ,
2 2 2 2 4
⌡ s sinh(πθ) ⌡ s 3sinh(πθ)
0 0
a x Y s( x )d x a x 3 Y s( x )d x
⌠ πθ a ⌠ πθ(5 − 4θ2) a 3
2 1/2 = cosh(πθ), =
6cosh(πθ)
,
⌡ (a − x ) ⌡ (a − x )
2 2 2 1/2
0 0
a ( a 2 − x 2)1/2 Y s( x )d x
⌠ π πθ a
= a tanh(πθ) − ,
⌡ x 2 cosh(πθ)
0
a Y s( x )d x a
⌠ π ⌠ x ( a 2 − x 2)1/2 Y ( x )d x = πθ(1 + 4θ ) a ,
2 3
2 1/2 = 2 a tanh(πθ), 6cosh(πθ)
⌡ x( a − x ) ⌡
2 s
0 0
a Y c( x )d x
⌠ 2 πsinh[2θtan-1( a / r )]
= ,
⌡ x + r
2 2 r sinh(πθ)
0
a Y ( x) xdx
⌠ 2
s π{cosh[2θtan-1( a / r )] − 1}
= ,
⌡ x + r
2 2sinh(πθ)
0
Integrals, containing combination Y c+ iY s:
a
iθ
⌠ a + x  dx
=
π 1 − cosh(πθ) a + riθ,
⌡ a − x  x − r i sinh(πθ) a − r  
-a
a
iθ π i coth(πθ) a + r i θ
⌠ a + x  dx
= ,
⌡ a − x  ( a + x )( x − r) a + r a − r 
-a
a
iθ π i tanh(πθ) a + r i θ
⌠ a + x  dx
= ,
⌡ a − x  ( x − r)( a − x ) ( a 2 − r 2)1/2 a − r 
2 2 1/2
-a
a
iθ iθ
⌠ a + x  ( a − x ) d x = πi tanh(πθ)( a 2 − r2)1/2a + r − π(2 ia θ + r),
2 2 1/2
⌡ a − x  x − r a − r  cosh(πθ)
-a
a a
iθ iθ
⌠ a + x  2πθ a ⌠ a + x  x d x = 2 i πθ a ,
2 2
dx = ,
⌡ a − x  sinh(πθ) ⌡ a − x  sinh(πθ)
-a -a
a
iθ
⌠ a + x  x 2d x = 2πθ(1 − 2θ ) a ,
2 3
⌡ a − x  3sinh(πθ)
-a
a
iθ
⌠ a + x  dx
= −
iπ
,
⌡ a − x  x + a sinh(πθ)
-a
a
iθ
⌠ a + x  xdx
=
π a ( i + 2θ)
,
-a
Appendix A3.2 211
a
iθ
⌠ a + x  x 2d x
= −
π a 2[2θ + i (1 − 2θ2)]
.
-a
Appendix A3.2
Some integrals, used in solving the external mixed-mixed problem, are

presented here. It is assumed that a < r <∞.
Integrals containing Y c:
∞ Y c( x )d x
⌠ 2 π
2 = 2 r coth(πθ) Y s( r ),
⌡ x − r
a
∞ Y c( x )d x
⌠ 2 2 π πθ
2 = 3 coth(πθ) Y s( r ) − ,
⌡ x (x − r ) ar sinh(πθ)
2
2r
a
∞ xY c( x )d x Y s( r )
⌠ π
2 = 2 tanh(πθ) 2
⌡ (x − a ) (x − r ) ( r − a 2)1/2
2 2 1/2 2
a
∞ Y ( x )d x Y ( r)
⌠
c π 1 s

2 = 2 − a cosh(πθ) + tanh(πθ) 2 2 1/2 ,
⌡ x ( x 2
− a 2 1/2 2
) ( x − r ) 2 r  ( r − a ) 
a
∞ Y c( x )d x
⌠ 3 2 π 1
2 = 4 − a cosh(πθ)
⌡ x (x − a ) (x − r )
2 1/2 2
2r 
a
Y s( r )
+ tanh(πθ)  − π[(1/4) − θ2],
( r 2 − a 2)1/2 a 3 r 2cosh(πθ)
∞ ( x 2 − a 2)1/2 Y ( x )d x
⌠
c π a 
2 cosh(πθ) + tanh(πθ)( r − a ) Y s( r ) ,
2 2 1/2
=
⌡ x( x − r )
2 2
2r  
a
∞Y ( x )d x ∞ Y c( x )d x
⌠
c πθ ⌠ 2 π
= , 2 1/2 = 2 a cosh(πθ),
⌡ x2 a sinh(πθ) ⌡ x( x − a )
a a
∞ Y c( x )d x
⌠ 3 2 π[(1/4) − θ2]
= ,
⌡ x (x − a ) a 3cosh(πθ)
2 1/2
a
⌠ [1 − Y ( x )]d x = a [πθcoth(πθ) − 1],

⌡ c
a
⌠ [1 − Y ( x )] 2 x d x 2 1/2 = πθ a tanh(πθ),
⌡ c (x − a )
a
∞ xY ( x )
⌠ 1 − 2  d x = a [πθtanh(πθ) − 1],
c
⌡  ( x − a 2 1/2
) 
a
⌠ ( x 2 − a 2)1/2d[ Y ( x )] = πθ a tanh(πθ),
⌡ c
a
⌠  2 x 2 1/2 − 1 Y ( x )d x = 2πθ a ,
⌡ ( x − a )  c sinh(2πθ)
a
⌠ 1 − ( x − a )  Y ( x )d x = πa 2πθ a
2 2 1/2
− ,
⌡  x  c 2cosh(πθ) sinh(2 πθ)
a
Appendix A3.2 213
Integrals containing Y :
s
∞ x Y s( x )d x
⌠ π
= coth(πθ)[1 − Y ( r )],
⌡ x − r
2 2 2 c
a
∞ Y s( x )d x
⌠ π 1 
2 = 2 sinh(πθ) − coth(πθ) Y c( r ) ,
⌡ x ( x 2
− r ) 2 r  
a
∞ Y s( x )d x
⌠ πtanh(πθ)
2 = − Y ( r ),
⌡ (x − a ) (x − r ) 2 r ( r 2 − a 2)1/2 c
2 2 1/2 2
a
∞ x 2 Y s( x )d x
⌠ π  r 
2 = 2 tanh(πθ) 1 − 2 1/2 Y c( r ) ,
⌡ (x − a ) (x − r )
2 2 1/2 2
 (r − a )
2

a
∞ Y ( x )d x
⌠ 2 2
s π  tanh(πθ) θ ,
2 = − 2 1/2 Y c( r ) +
⌡ x ( x − a 2 1/2 2
) ( x − r ) r 2
2 r ( r 2
− a ) a 2
cosh(πθ) 
a
∞ ( x 2 − a 2)1/2 Y ( x )d x
tanh(πθ)1 − Y c( r ),
⌠
s π ( r 2 − a 2)1/2
=
⌡ x2 − r2 2  r 
a
∞ ∞
⌠ Y ( x ) d x = π tanh(πθ), ⌠ Y ( x ) d x3 = 2 πθ
2
,
⌡ s x 2 2 ⌡ s x a sinh(πθ)
a a
∞ ∞ Y s( x )d x
⌠ Y ( x ) d x5 = πθ 4(2 − θ ), π
2 2
⌠ 2 1/2 = 2 tanh(πθ),
⌡ s x 3 a sinh(πθ) ⌡ (x − a )
2
a a
∞ Y s( x )d x
⌠ 2 2 πθ
2 1/2 = ,
⌡ x (x − a ) a cosh(πθ)
2
a
⌠ ( x 2 − a 2)1/2d[ Y ( x )] = πa 2tanh(πθ)(θ2 − 1/4),

⌡ s
a
⌠ ( x 2 − a 2)1/2 Y ( x ) d x2 = πtanh(πθ) − 2θ 
⌡ s x 2 cosh(πθ)
a
⌠ ( x 2 − a 2)1/2 Y ( x ) d x4 = 2πθ(2θ + 1/4)

2
⌡ s x 3 a cosh(πθ)
a
⌠  2 x 2 1/2 − 1 Y ( x ) x d x = πa  tanh(πθ) + 8θ2 

2
,
⌡ ( x − a )  s 4  sinh(2πθ)
a
Integrals containing both Y c and Y s:
⌠ [ x Y ( x ) − 2 a θ Y ( x )] d x = πa 2θ2coth(πθ),
⌡ s c
a
⌠ [ x Y ( x ) − 2 a θ Y ( x )] 2 x d x 2 1/2 = πa tanh(πθ)(1 + 4θ2),

2
⌡ s c (x − a ) 4
a
⌠ 1 − ( x − a )  [ x Y ( x ) − 2 a θ Y ( x )] d x
2 2 1/2
⌡ x  s c
a
2πθ2 a 2 πa2 πθ a 2
= + tanh(πθ) − .
sinh(2πθ) 4 cosh(πθ)
Appendix A3.3 215
Appendix A3.3
Some formulae, related to transformation and computation of integrals, are

presented here.
Transformations of the limits of integration:
ρ a a
⌠ f ( x )d x 2 ⌠
2 1/2 = π ρ
dy ⌠ f ( x )d x
⌡ (ρ − x )
2
⌡ ρ2
− y 2
⌡ ( x − y 2)1/2
2
0 0 y
a a y
⌠ f ( x )d x 2 2 2 1/2⌠
2 1/2 = π( a − ρ )
ydy ⌠ f ( x )d x
⌡ (x − ρ )
2
⌡ ( y − ρ )( a − y ) ⌡ ( y − x )
2 2 2 2 1/2 2 2 1/2
ρ 0 0
ρ a
=
2 ⌠ dy ⌠ xf2 ( x )d x2 + π lim[ xf ( x )],
π ⌡ (ρ − y )
2 2 1/2
⌡ x − y 2ρx→0
0 0
a a y
⌠ f ( x )d x 2 2 2 1/2⌠
2 1/2 = π(ρ − a )
ydy ⌠ f ( x )d x
⌡ (ρ − x ) ⌡ (ρ − y )( a − y ) ⌡ ( y − x )
2 2 2 2 2 1/2 2 2 1/2
0 0 0
Simplification of two consecutive integrals:
a a a
⌠ x 2d x d ⌠ f ( y )d y π ⌠ f ( y )d y
2 1/2 = − 2 ρ f (ρ) +
⌡ (x − ρ )
2 2 1/2 d x
⌡ (y − x )
2
 ⌡
ρ x ρ
a 2 1/2 
− 2 1/2 lim[ f ( r )( a − r ) ] ;
2
(a − ρ )
2
r→a 
a a a
⌠ 2 2 d x 2 1/2 d ⌠ π  f (ρ) f ( y )d y 
− 2 + ⌠
f ( y )d y
= .
⌡ x ( x − ρ ) d x ⌡ ( y 2
− x 2
) 1/2 2 ρ ρ ⌡ y3 
ρ x ρ
x a a
⌠ 2 t d t 2 1/2 ⌠ f ( y )d y ⌠ ( x 2 − b 2)1/2 + ( y 2 − b 2)1/2

= f ( y )ln dy,
⌡ (x − t ) ⌡ (y − t ) ⌡ | x 2 − y 2|1/2
2 2 1/2
b t b
a t a
⌠ 2 tdt ⌠ f ( y )d y ⌠ ( a 2 − x 2)1/2 + ( a 2 − y 2)1/2

= f ( y )ln dy,
⌡ (t − x 2)1/2 ⌡ ( t 2 − y 2)1/2 ⌡ | x 2 − y 2|1/2
x b b
a a
⌠ 2 d x2 (1+κ)/2 d ⌠ f ( r) rdr π
2 (1-κ)/2 = − 2cos(πκ/2) f (ρ),
⌡ (x − ρ ) d x ⌡ (r − x )
2
ρ x
ρ x
⌠ 2 d x2 (1+κ)/2 d ⌠ f ( r) rdr
=
π f (ρ) − 1 lim[ rf( r)]
⌡ (ρ − x ) dx ⌡ (x − r )
2 2 (1- κ)/2
2cos(πκ/2) ρ r→0 
0 0
x a a
⌠ 2 d t 2 1/2 d ⌠ f ( y )d y 2 1/2 ⌠
2 1/2 = ( x − b )
2 f ( y )d y
2 1/2,
⌡ (x − t ) ⌡ (y − t ) ⌡ ( y − x )( y − b )
d t 2 2 2 2
b t b
a t a
⌠ 2 dt d ⌠ f ( y )d y
2 1/2 = ( a
2
− x 2)1/2⌠
f ( y )d y
2 1/2,
⌡ (t − x) ⌡ (t − y ) ⌡ ( y − x )( a − y )
2 1/2 d t 2 2 2 2
x b b
x a a
d ⌠ tdt ⌠ f ( y )d y x ⌠ ( y 2 − b 2)1/2 f ( y )d y
= ,
d x ⌡ ( x 2 − t 2)1/2 ⌡ ( y 2 − t 2)1/2 ( x 2 − b 2)1/2 ⌡ y2 − x2
b t b
a t a
d ⌠ tdt ⌠ f ( y )d y x ⌠ ( a 2 − y 2)1/2 f ( y )d y
= ,
d x ⌡ ( t 2 − x 2)1/2 ⌡ ( t 2 − y 2)1/2 ( a 2 − x 2)1/2 ⌡ y2 − x2
x b b
a x a
⌠ dx d ⌠ f ( y )d y 2 1/2⌠
2 1/2 = (ρ − a )
2 f ( y )d y
2 1/2,
⌡ (ρ − x ) ⌡ (x − y ) ⌡ (ρ − y )( a − y )
2 2 1/2 d x 2 2 2 2
0 0 0
Appendix A3.3 217
∞ ∞ ∞ ∞
⌠ xdx ⌠ df(t)
= ⌠
dx d ⌠ f(t)tdt
⌡ ( x − ρ2)1/2 ⌡
2
( t − x 2)1/2
2
⌡ (x − ρ )
2 2 1/2 d x
⌡ ( t − x 2)1/2
2
a x a x
− ρ) ⌠
2 2 1/2 f(t)tdt
= (a .
⌡ (t
2
− a 2)1/2( t 2 − ρ2)
a
Computation and/or transformation of some integrals:
a a
⌠ 2n+1 2dρ 2 1/2 ⌠ t 2n-1d t

=
π( a 2n − x 2n)
,
⌡ ρ (ρ − x ) ⌡ ( a − t )
2 2 1/2
4 nax 2n+1
x ρ
a a
⌠ 2n-1 2dρ 2 1/2 ⌠ t 2n-1d t

=
π( a 2n-1 − x 2n-1)
,
⌡ ρ (ρ − x ) ⌡ ( a − t ) 2(2 n − 1) x 2n-1
2 2 1/2
x ρ
a a
⌠ 2n-1 2 dρ 1 ⌠ ρ2n-1dρ
=
⌡ ρ (a − ρ2)1/2(ρ2 − x 2)1/2 ( ax )2n-1 ⌡ ( a 2 − ρ2)1/2(ρ2 − x 2)1/2
x x
π 1 x2 √πΓ( n − 1/2) 1 3 x2
= F ( 1− n , ; 1; 1− ) = F ( 1− n , ; − n ; ),
2 ax 2n-1 2 a2 2 ax 2n-1Γ( n ) 2 2 a2
a a
a 2 − ρ2)1/2dρ
⌠ (2n+1 1 ⌠ (ρ2 − x 2)1/2ρ2n-1dρ π( a 2 − x 2)
= =
⌡ ρ (ρ − x ) a 2n-1 x 2n+1 ⌡ ( a 2 − ρ2)1/2
2 2 1/2
4 ax 2n+1
x x
1 x2 √πΓ( n + 1/2)( a 2 − x 2) 1 1 x2
× F (1− n , ; 2; 1− 2) = F ( 1− n , ; − n ; )
2 a 2 n ! ax 2n+1 2 2 a2
a a
ρ2 − x 2)1/2dρ
⌠ (2n+1 1 ⌠ ( a 2 − ρ2)1/2ρ2n-1dρ
=
⌡ ρ (a − ρ ) a 2n+1 x 2n-1 ⌡ (ρ2 − x 2)1/2
2 2 1/2
x x
π( a 2 − x 2) 3 x2
= F ( 1 − n , ; 2; 1 − ).
4 a 3 x 2n-1 2 a2
ρ ρ
⌠ x 2nd x
= a 2n+1ρ2n⌠
dx
,
⌡ (ρ2 − x 2)1/2 ⌡ ( x − a 2)1/2
2n+1 2
x
a a
a a
⌠ x 2nd x 2n+1 2n⌠

ρ
dx
2 1/2 = a .
⌡ (x − ρ )
2
⌡ x ( a − x 2)1/2
2n+1 2
ρ ρ
a n-1
dρ (a − x ) (a − t )
Σ a 2 − x 2m
2 2 1/2 2 2 1/2
⌠ 2n-1 2 2 1/2 =
⌡ ρ ( ρ − x 2 1/2 2
) ( ρ − t ) a 2 x 2n-2( x 2 − t 2)  a2 
m=0
x
Γ( n ) dm  m-1/2 sin √ζ

-1
t 2( a 2 − x 2)
× (1 − ζ ) , with ζ = − .
Γ( n − m )( m !)2 dζm √ζ  a 2( x 2 − t 2)
∞ ρ
⌠ 2n+1 2 dx 1 ⌠ x 2n+2d x
2 =
⌡ x (x − a ) (x − ρ ) a ρ ⌡ (ρ2 − x 2)1/2( a 2 − x 2)
2 1/2 2 2n+1 2n+2
a 0
= 2n+1 2n+2
a ρ
1

π a 2n+1
2( a − ρ )
2
√π
2 1/2 − 2 Σ k=1
Γ( k +1/2) 2(n-k) 2k
Γ( k +1)
a ρ ,

Rules of interchanging the order of integration:

a r a a
⌠ F ( r)d r d ⌠ f (ρ)ρdρ ⌠ f (ρ)dρ d ⌠ F ( r) rdr

2 1/2 = − ρ
+
⌡ d r ⌡ (r − ρ )
2
⌡ d ⌡ ( r − ρ2)1/2
2
0 0 0 ρ
 d ⌠ F ( r) rdr   F ( r) rdr 
a a
limρ f (ρ) ⌠
dρ ⌡ ( r 2− ρ2)1/2  ,
+ lim f ( ρ)
ρ→0
  ρ→a
 ⌡ ( r 2− ρ2)1/2

ρ ρ
Appendix A3.3 219
∞ ∞ ∞ x
⌠ F (ρ)dρ d ⌠ f( x) xdx
2 1/2 = −
⌠ f ( x )d x d ⌠ F2 (ρ)ρd2ρ1/2,
⌡ dρ ⌡ (x − ρ )
2
⌡ dx ⌡ (x − ρ )
a ρ a a
a a a a
⌠ 2 d t 2 ⌠ f (2t , r)d r2 = − π f ( x2, x ) + ⌠ d r ⌠ 2 f ( t2, r)d2t

2
2 ,
⌡t − x ⌡ r − t 4x ⌡ ⌡ ( t − x )( r − t )
0 0 0 0
a a a a
⌠ d t ⌠ f (2t , r)d r2 = − π f (0,0) + ⌠ d r ⌠ f (2 t , r)d t2 .

2
⌡ ⌡ r − t 4 ⌡ ⌡ (r − t )
0 0 0 0
CHAPTER 4
APPLICATIONS IN FRACTURE MECHANICS
The great majority of the punch and crack problems solved deals with the
stresses and displacements in the plane z =0 only. Some solutions of this kind
have been presented in previous chapters. There are just a few complete
solutions published (Sneddon, 1951; Elliott, 1949; Westmann, 1965), where explicit
expressions are given for the field of displacements and stresses for the simplest
axisymmetric problems (a circular punch and a penny-shaped crack). The explicit
expressions for the field of displacements due to an elliptic crack can be found
in (Kassir and Sih, 1975). Knowledge of complete solutions is indispensable for
consideration of more complicated problems of crack interactions, influence of
external loads on punches and cracks, etc.
We present in this chapter a complete solution to the problem of a
penny-shaped crack in a transversely isotropic elastic space, subjected to an
arbitrary normal and tangential loading. All the relevant Green’s functions are
given explicitly in terms of elementary functions. An approximate analytical
solution is given for a flat crack of arbitrary shape. The solution’s accuracy is
high, which is mainly due to the fact that it becomes exact in the case of an
elliptical crack. The derivation of non-singular governing integral equations
enables us to consider a very close interaction of coplanar cracks. Some of the
material presented in this Chapter is still unpublished. The rest follows the
papers (Fabrikant, 1987a, 1987b, 1987f, 1987g, 1988b, 1989).
4.1 Flat crack under arbitrary normal loading
A general solution to some mixed problems in terms of three harmonic

functions was given in Chapter 2. We show here that in the case of a flat
crack under normal loading, the general solution can be expressed through just
220
one such function. Consider a transversely isotropic elastic space weakened by a

flat crack S in the plane z =0, with arbitrary pressure p applied to the crack
faces. Due to symmetry, the problem can be formulated as follows: find the
solution to the set of differential equations (2.1.3) for a half-space z ≥0, subject
to the mixed boundary conditions on the plane z =0:
σz = − p ( x,y), for ( x,y)∈ S ; w = 0, for ( x,y)∉ S ;
τz = 0 , for −∞< ( x,y) <∞. (4.1.1)
These conditions can be satisfied by a representation in terms of one harmonic

function. Let us put, according to (2.1.13),
F ( z ) = c F ( z ), F ( z ) = c F ( z ), F ( z) = 0 . (4.1.2)
1 1 1 2 2 2 3
Expressions of the type F ( z ) and F ( z ), etc., everywhere in the book should be

1 1
understood as F ( x,y,z) and F ( x,y,z ) respectively. The substitution of (4.1.2) and
1 1
the last of expressions (2.1.12) in the third condition (4.1.1) yields:
c = −c γ / m γ (4.1.3)
1 2 1 1 2
We can represent the function F as the potential of a simple layer, i.e.
ω( N ) d S
F (ρ,φ, z ) ≡ F ( z ) = ⌠ ⌠ , (4.1.4)
⌡ ⌡ R(M,N)
S
where ω stands for the crack face displacement w ( x,y,0), R(M,N) is the distance
between the points M (ρ,φ, z ) and N ( r ,ψ,0), and the integration is taken over the
crack domain S . Expression (4.1.4) satisfies the second condition (4.1.1)
identically, due to the well known property of the potential of a simple layer.
Inside the crack the same property gives:
∂F
= −2πω = −2π w ( x,y,0) (4.1.5)
∂ z z=0
Now expressions (4.1.2), (4.1.4), (4.1.5), and (2.1.6) give the second equation for
c and c :
1 2
− m c /γ − m c /γ = 1/2π (4.1.6)
1 1 1 2 2 2
222 CHAPTER 4 APPLICATIONS IN FRACTURE MECHANICS
The constants c 1 and c 2 are determined from (4.1.3) and (4.1.6) as
γ1 γ2
c1 = − , c2 = − . (4.1.7)
2π( m 1 − 1) 2π( m 2 − 1)
The potential functions will be given by
γ1 γ2
F 1(z) = − F ( z 1), F 2(z) = − F ( z 2).
2π( m 1 − 1) 2π( m 2 − 1)
(4.1.8)
The substitution of (4.1.8) and (2.1.12) in the first condition (4.1.1) leads to the
governing integral equation:
∆⌠ ⌠
1 ω( N ) d S
p ( N 0) = − , (4.1.9)
4π H ⌡ ⌡ R ( N0,N)
2
where, as before, R ( N0,N) stands for the distance between two points N 0 and N ,
and both N 0, N ∈ S . The following identities were used:
m 1 m 2 = 1, ( m 1 − 1)/( m 1 + 1) = 2π A 44 H (γ1 − γ2). (4.1.10)
We next consider the penny-shaped crack in more detail. We shall return to the
case of a general crack in section 4.8 below.
Green’s functions for a penny-shaped crack. An exact solution in
elementary functions is possible when the crack is circular. Let a be the radius
of the crack. The governing integral equation (4.1.9) can be rewritten in polar
coordinates as follows (see section 2.8)
ρ
1 1 d ⌠ x dx
p (ρ,φ) = − 2 L( ) L( x 2)
π Hρ ρ dρ ⌡ (ρ − x )
2 2 1/2
0
ρ0dρ0
a
d ⌠ 1
× L( ) ω(ρ0,φ).
d x ⌡ (ρ 2 − x 2)1/2 ρ0
x0
(4.1.11)
The integral operator inverse to (4.1.11) is defined by (2.8.2)
a x ρ dρ ρρ
ω(ρ, φ) = 4 H ⌠ d x ⌠
0 0
L  0
 p (ρ , φ). (4.1.12)
⌡ ( x 2
− ρ )
2 1/2
⌡ ( x 2
− ρ )
2 1/2
0
 x 2
 0
ρ 0
Another form of solution can be obtained from (1.4.33), namely,
2π a p (ρ , φ )
tan-1  ρ0dρ0dφ0,
2 ⌠ ⌠ 0 0 η
ω = H (4.1.13)
π ⌡ ⌡ R  
R
0 0
where
R =[ρ2 + ρ20 − 2ρρ cos(φ−φ )]1/2, η = ( a 2 − ρ2)1/2 ( a 2 − ρ20)1/2/ a .

0 0
(4.1.14)
In this chapter, we do not restrict attention to the plane z =0: our purpose is
obtaining a complete solution. We shall call F (ρ,φ, z ), as defined by (4.1.4), the
main potential function since both functions F 1 and F 2 become easily available,
when F is found. The substitution of (4.1.13) in (4.1.4) allows us to express
the main potential function as follows:
2π a
F (ρ,φ, z ) = H⌠ ⌠ K (ρ,φ, z ; ρ ,φ ) p (ρ ,φ )ρ dρ dφ ,
2
(4.1.15)
π ⌡ ⌡ 0 0 0 0 0 0 0
0 0
where the Green’s function K reads:
K ( M;N ) = K (ρ,φ, z ; ρ ,φ )
0 0 0
2π a
( a 2 − r 2)1/2( a 2 − ρ20)1/2
= ⌠ ⌠
1
tan-1  rd rdψ .
⌡ ⌡ R ( N,N0)  a R ( N,N )
0
 R ( M,N)
0 0
(4.1.16)
Here R (⋅,⋅) denotes the distance between respective points: M (ρ,φ, z ), N ( r ,ψ,0), and
N (ρ ,φ ,0). Although we can not compute the integral in (4.1.16) in elementary
0 0 0
functions, all its derivatives can be expressed in elementary functions, due to the
fundamental integral established in section 1.6. Making use of (1.6.19), one can
write:
tan-1 ,
∂K 2π h
= − (4.1.17)
∂z R ( M,N )
0
R ( M,N )
0
where
h = ( a 2 − l 21)1/2( a 2 − ρ20)1/2/ a , (4.1.18)
and the contraction l everywhere in the book stands for l ( a ), as defined by
1 1
(0.18). Note that h tends to η, as defined by (4.1.14), for z →0 and ρ< a .
Expressions (4.1.15) and (4.1.17) allow us to write:
2π a
tan-1  p (ρ ,φ )ρ dρ dφ .
∂F
= −4 H⌠ ⌠
1 h
(4.1.19)
∂z ⌡ ⌡ R ( M,N 0
) R ( M,N 0
)  0 0 0 0 0
0 0
The integral in (4.1.19), although looking difficult to compute even for p =const,
can be expressed in elementary functions for any polynomial loading. This
becomes evident, if we use the equivalent representations through the L-operator
(see 1.4.31):
l2 g(x)
ρ dρ ρρ
∂F
= −8π H ⌠
dx ⌠ 0 0
 0p (ρ ,φ).
2 1/2 L
∂z ⌡ ( x 2 − ρ2)1/2 ⌡ [ g ( x ) − ρ0]
2
 x2  0
l (0) 0
2
(4.1.20)
Here
 z2 1/2,
g( x) = x 1 + 2 (4.1.21)
 ρ − x 
2
and the contraction l everywhere in the paper stands for l ( a ), as defined by

2 2
(0.14). Using the change of variables x = l 2( t ), t = g ( x ), expression (4.1.20) can be
rewritten as follows:
a dl (t) t ρ dρ ρρ
∂F
= −8π H ⌠
2
⌠
0 0
L  0
p (ρ ,φ).
∂z ⌡ [ l 2( t ) − ρ ] ⌡ ( t − ρ0)
2 2 1/2 2 2 1/2
l 2( t ) 0
2
0 0
(4.1.22)
Since the function F vanishes at infinity, it can be determined from (4.1.22) in
the form
z a d l 2( t ) t ρ dρ ρρ
F (ρ,φ, z ) = −8π H ⌠ d z ⌠ 2 ⌠ 2
0 0
L  0 p (ρ ,φ).
⌡ ⌡ [ l 2( t ) − ρ ] ⌡ ( t − ρ0)
2 1/2 2 1/2
l 22( t ) 0
∞ 0 0
(4.1.23)
By using the property
∂ l 2( t ) [ l 22( t ) − ρ2]1/2 ∂ l 1( t )
= − ,
∂t [ρ2 − l 21( t )]1/2 ∂z
which is a consequence of formulae (A4.1.28) and (A4.1.29) from Appendix

A4.1, expression (4.1.23) can be modified as follows:
a t ρ0dρ0 l1(t) y 2ρ0

F (ρ,φ, z ) = 8π H⌠ d t ⌠ ⌠   p (ρ ,φ).
dy
2 1/2 L
⌡ ⌡ (t −
2
ρ20)1/2 ⌡ (ρ − y )
2
 t 2ρ  0
0 0 0
(4.1.24)
Expression (4.1.24) is convenient for exact evaluation of the potential function F
and proves that it can be expressed in elementary functions for arbitrary
polynomial loading. A simple change of variables gives another formula,
equivalent to (4.1.24):
∞
a t ρ0dρ0
ρρ0
⌠ ⌠ ⌠ dx  
F (ρ,φ, z ) = 8π H t d t 2 1/2 ⌡ 2 1/2 L 2 p (ρ0,φ).
⌡ ⌡ ( t − ρ0) l ( t ) x ( x − t )
2 2
x 
0 0
2
(4.1.25)
We can proceed now with the remaining derivatives of the Green’s function
K , defined by (4.1.16). Differentiation of (4.1.16) yields:
2π a φ ψ
ρei − r ei ( a 2 − r 2)1/2( a 2 − ρ20)1/2
Λ K (ρ,φ, z ;ρ0,φ0) = −⌠ ⌠ tan-1  rd rdψ .
⌡ ⌡ 3
R ( M,N )  a R ( N,N0)  R ( N,N0)
0 0
(4.1.26)
This integral is computed in Appendix A4.3. By using (A4.3.11), one can write:
2π  z ( a 2 − ρ20)1/2
Λ K (ρ,φ, z ; ρ0,φ0) = tan-1
h
− tan-1 2
s ,
q R 0 R0 s (l2 − a ) 
2 1/2
(4.1.27)
where Λ is given by (2.1.5), h is defined by (4.1.18), and

-i φ
-i φ -i(φ-φ0) 1/2
q = ρe − ρ0e 0
, s = ( a 2 − ρρ0e ) ,
R 0 = R ( M,N0) = [ρ2 + ρ20 − 2ρρ0cos(φ−φ0) + z 2]1/2 . (4.1.28)
The other derivatives, which will be needed for the complete solution, are:
∂2 z -1 
h h 
ρ2 − l 21 z 2 
K (ρ,φ, z ; ρ0,φ0) = 2π 3 tan − − 2 ,
∂ z2 R0 R 0
 z [ R 20 + h 2]  l 22 − l 21 R0 

(4.1.29)
iφ
iφ
ρe − ρe 0
tan-1 
∂ 0 h
Λ K (ρ,φ, z ; ρ ,φ ) = 2π
∂z 0 0
 R 30 R  0
iφ
iφ
iφ ρe − ρe 0
+
h  ρe
+
0
, (4.1.30)
R 0 + h  l 22 − l 21
2 2
R 20 
iφ
( a − 2
ρ20)1/2 ρe 0
Λ2 K (ρ,φ, z ; ρ ,φ ) = 2π 2 −
0
 tan-1 s 
0 0
 q s q s2  ( l 22 − a 2)1/2
iφ
z (3 R 20 − z) 2 ( a 2 − ρ20)1/2( l 22 − a 2)1/2ρ0e 0
tan-1
h 
− +
q 2 R 30  R  0 q s 2[ l 22 − ρρ0e
-i(φ-φ0)
]
iφ
− 2
zh q
− 2
ρ2e2  . (4.1.31)
R 0 + h 2  q R 20 ( l 2 − l 21)( l 22 − ρ2) 
This concludes the general solution to the problem of a penny-shaped crack

subjected to an arbitrary pressure. Formulae (4.1.17) and (4.1.24−4.1.31) are the
main results of this section.
Exercise 4.1
1. Prove the identity (4.1.10).
4.2 Point force loading of a penny-shaped crack 227
2. Establish (4.1.19).
3. Verify the derivation of (4.1.27) −(4.1.31).
4. Find the Green’s functions for a semi-infinite plane crack in a transversely

isotropic space, subjected to a normal loading.
Hint : consider the limiting case of (4.1.24−4.1.31), when the radius a →∞, and
the coordinate origin moves from the circle centre to its boundary.
4.2 Point force loading of a penny-shaped crack
Consider a penny-shaped crack opened by two equal concentrated forces P

applied in opposite directions at the point (ρ ,φ ,0±), ρ < a . Formulae (2.1.6),
0 0 0
(2.1.12), (4.1.8), (4.1.17), and (4.1.24−4.1.31) give a complete solution for the
field of displacements and stresses in elementary functions, namely,
γ γ
u =
2
HP  1
f (z ) +
2
f (z )  , (4.2.1)
π  m1 − 1 1 1 m − 1 1 2 
2
m m
w =
2
HP  1
f (z ) +
2
f (z )  , (4.2.2)
π  m1 − 1 2 1 m − 1 2 2 
2
2P  γ1 1 
σ1 = 2  2 − γ

f (z )
π (γ − γ ) ( m + 1)γ3 3 1
1 2  1 1
γ2 
 1 
γ 3 2 
− − f ( z ) , (4.2.3)
 ( m 2 + 1)γ23 2 
γ γ
σ =
4
HA P  1
f (z ) +
2
f (z )  , (4.2.4)
2 π 66  m − 1 4 1
1
m − 1 4 2 
2
σ =
P  γ f (z ) − γ f (z )  , (4.2.5)
z π (γ − γ )  1 3 1
2 2 3 2 
1 2
τz =
P  f (z ) − f (z )  , (4.2.6)
π (γ 1 − γ 2)  5 1
2 5 2 
where
1 ( a − ρ0)
2 2 1/2
s z h 
f 1( z ) = tan-1 2 − tan-1
, (4.2.7)
q  s ( l 2 − a 2)1/2 R0 R0 
tan-1
1 h 
f 2( z ) = , (4.2.8)
R0  0 
R
 z -1 h 
h 
ρ2 − l 21 z2  ,
f 3( z ) = − 3 tan + − 2
 R0  R0  z ( R 20 + h 2)  l 22 − l 21 R0 
(4.2.9)
iφ
( a 2 − ρ20)1/2 ρ0e 0
f 4( z ) =  −
2 
tan-1 2
s 
q s  s2 q   ( l 2 − a 2)1/2 
iφ
z (3 R 20 − z 2) ( a 2 − ρ20)1/2( l 22 − a 2)1/2ρ0e 0
tan-1  −
h
+
q 2 R 30 R 0 q s 2[ l 22 − ρρ0e
-i(φ-φ0)
]
iφ

zhq ρ2e2 ,
+ 2 2 − (4.2.10)
R0 + h  qR0
2
( l 2 − l 1)( l 2 − ρ ) 
2 2 2 2
iφ iφ
iφ iφ
ρe − ρ0e ρe − ρ0e
0 0
iφ
f 5( z ) = − tan-1  + 2
h h  ρe +  .
 R 30 R 0 R 0 + h 2 l 22 − l 21 R 20 
(4.2.11)
It is reminded that R 0=[ρ + 2
ρ20 − 2ρρ0cos(φ−φ0) + z ] . 2 1/2
The expression for σz
(4.2.5) simplifies when z =0 and ρ> a , namely,
P ( a 2 − ρ20)1/2
σz = . (4.2.12)
π2 (ρ2 − a 2)1/2[ρ2 + ρ20 − 2ρρ0cos(φ−φ0)]
4.3 Concentrated load outside a circular crack 229
Defining the stress intensity factor
k 1 = lim{(ρ − a )1/2σz} ,
ρ→ a
the following result may be obtained from (4.2.12):
P ( a 2 − ρ20)1/2
k = . (4.2.13)
1 π2(2 a )1/2 a 2 + ρ20 − 2 a ρ0cos(φ−φ0)
One can write for an arbitrarily distributed pressure:
2π a ( a 2 − ρ20)1/2 p (ρ ,φ ) ρ dρ dφ
1
k = ⌠ ⌠ 0 0 0 0 0
,
1 π (2 a )
2 1/2
⌡ ⌡ a 2 + ρ20 − 2 a ρ0cos(φ−φ0)
00
which corresponds to the well known result (Cherepanov, 1974).
Exercise 4.2
1. Derive the solution (4.2.1) −(4.2.6) for the case of an isotropic body.
2. Verify the derivation of (4.2.7) −(4.2.11).
4.3 Concentrated load outside a circular crack
Consider a transversely isotropic space weakened by a penny-shaped crack of

radius a in the plane z =0. Let a concentrated force P be applied at an
arbitrary point (ρ,φ, z ) in the Oz direction. The crack faces are stress-free. Let
us find the crack opening displacement and the opening mode stress intensity
factor k 1.
Consider the second system in equilibrium: two unit concentrated forces Q

applied normally to the crack faces in opposite directions at the point (ρ0,φ0,0±).
Denote the normal displacement in the space due to the forces Q by w ; while
Q
w P is the crack opening displacement due to force P . Application of the
reciprocal theorem to the two systems yields
Qw P = Pw Q ,
which gives the crack opening displacement
m1 m2
w P(ρ0,φ0) =
2
HP  f (z ) + f (z )  , (4.3.1)
π  m1 − 1 2 1 m2 − 1 2 2 
with f 2 defined by (4.2.8). The stress intensity factor can be determined by
w P(ρ0,φ0)
1
k 1(φ0) = lim
8π H ρ0→a ( a − ρ )1/2
0
m m
=
P  1
f ( z ) +
2
f 6( z 2) ,
2(2 a ) π  m 1 − 1
1/2 2 6 1 m2 − 1 
where
f 6( z ) = ( a 2 − l 21)1/2/ r 2a , r 2a = ρ2 + a 2 − 2ρ a cos(φ−φ0) + z 2.
(4.3.2)
The stress intensity factor vanishes as z tends to zero for ρ≥ a .
In the case of an isotropic body, expression (4.3.1) transforms into
P  1 − ν
tan-1 
h
w P(ρ0,φ0) =  R  0
πµ
2 R
0
1  
h ρ2 − l 21
z2  z2 -1 h  

R 0 
− − − tan . (4.3.3)
2  R 20 + h 2  l 22 − l 21 R 20  R 30

Here µ is the shear modulus, and ν is Poisson’s ratio. The corresponding
expression for the stress intensity factor will take the form:
P ( a 2 − l 21)1/2 2 ρ2 − l 21
k 1(φ0) =  1 + 1  z − .
(8 a )1/2π2 r a2  1 − ν  ra
2
2( l 2 − l 1) 
2 2
(4.3.4)
In the case of axial symmetry ρ=0, and formulae (4.3.3−4.3.4) simplify as
follows:
P 
a 2 − ρ20 1/2
w P(ρ0,φ0) = 2 
1 − ν
+ 2
z2  tan 
-1 
π µ  (ρ20 + z 2)1/2 (ρ0 + z 2)3/2   ρ20 + z2 

z 2( a 2 − ρ20)1/2 
+ ,
2(ρ20 + z 2)( a 2 + z 2)

k1 =
Pa 1/2  1 + 1 z2 
,
2π2√2( a 2 + z 2)  1 − ν a 2 + z 2
which is in agreement with the results reported by Collins (1962), who

considered the axisymmetric case only.
Exercise 4.3
1. Verify (4.3.2)
2. Derive a complete solution for the case of an arbitrary point force applied
outside a penny shaped crack in a transversely isotropic space.
4.4 Plane crack under arbitrary shear loading
Consider a transversely isotropic elastic space weakened by a flat crack S

in the plane z =0, with arbitrary shear loading applied to the crack faces
antisymmetrically. The problem can be formulated as follows: find the solution
to the set of differential equations (2.1.3) for a half-space z ≥0, subject to the
mixed boundary conditions on the plane z =0:
τz = −τ( x,y), for ( x,y)∈ S ; u = 0, for ( x,y)∉ S ;
σz =0, for −∞<( x,y)<∞. (4.4.1)
It is no longer possible to present the solution in the form (4.1.2). A more

complicated representation is necessary, namely,
F 1 = c 1(Λχ1 + Λχ1) , F 2 = c 2(Λχ2 + Λχ2) , F 3 = c 3(Λχ3 − Λχ3).(4.4.2)
Here c 1, c 2 and c 3 are the as yet unknown constants; χ1, χ2, and χ3 are the as
yet unknown complex harmonic functions. A bar indicates the complex conjugate
value throughout this book. Introducing the notation z = z /γ , for k =1,2,3, we

k k
assume also that
χ ( z ) = χ( z ), χ ( z ) = χ( z ), χ ( z ) = χ( z ). (4.4.3)
1 1 2 2 3 3
This assumption will allow us to reduce the problem to finding just one
harmonic function which is much easier than searching for three. By substituting
(4.4.3) into the third equation (2.1.12), we obtain the first equation for the
constants, namely,
c + m c = 0. (4.4.4)
1 2 2
The third condition in (4.4.1) is thus satisfied. Substitution of (4.4.2) in (2.1.6)

yields
u = c (Λ2χ + ∆χ ) + c (Λ2χ + ∆χ ) + i c (Λ2χ − ∆χ ),

1 1 1 2 2 2 3 3 3
(4.4.5)
where the differential operators Λ and ∆ are defined by (2.1.5). When z =0,
equation (4.4.5) transforms into
u = ( c + c + i c )Λ2χ + ( c + c − i c )∆χ. (4.4.6)

1 2 3 1 2 3
It is convenient to assume
c + c + i c = 0. (4.4.7)
1 2 3
This assumption simplifies (4.4.6) as follows
u = ( c + c − i c )∆χ, (4.4.8)
1 2 3
and makes it possible to represent
χ( M ) = ⌠ ⌠ ln[ R ( M,N ) + z ] u ( N ) d S . (4.4.9)

⌡⌡ N
S
The representation (4.4.9) satisfies the second condition (4.4.1) identically, and
inside the crack the following equation becomes valid
c + c − i c = 1/2π. (4.4.10)
1 2 3
The solution of the set of equations (4.4.4), (4.4.7), and (4.4.10) gives
1 1 i
c1 = − , c2 = − , c3 = . (4.4.11)
4π( m 1 − 1) 4π( m 2 − 1) 4π
Substitution of (4.4.2) and (4.4.11) in the last of expressions (2.1.12) gives the
following expression for the tangential stress:
A 44 ∂ m 1 + 1
τz = −  (Λ2χ + ∆χ )
4π ∂ z m − 1 1 1
1
m2 + 1
+ (Λ2χ + ∆χ ) + (Λ2χ − ∆χ ) .
m2 − 1 2 2 3 3 
(4.4.12)
Expression (4.4.12) simplifies for z =0
A 44 m + 1
∂χ ∂χ m + 1
τz = −  1
+  Λ2
1
+ 
2
−  ∆  .
1
4π  ( m − 1)γ γ 3 ∂z ( m 2 − 1)γ2 γ 3 ∂ z 
1 1
(4.4.13)
Finally, satisfaction of the first condition (4.4.1) yields the governing
integro-differential equation:
τ( N 0 ) = −
1 G ∆⌠ ⌠ u ( N) d S + G Λ2⌠ ⌠ u ( N) d S ,
2π
2
( G 21 − G 22)  1 ⌡ ⌡ R ( N,N0) N 2 ⌡ ⌡ R ( N,N0) N
S S
(4.4.14)
where the elastic constants G 1 and G 2 are defined by (2.1.9).
Green’s functions in the case of shear loading. The integro-differential

equation (4.4.14) was solved exactly for a penny-shaped crack in section 2.7.
The closed form solution is (see 2.7.53)
G1 2π a 2
⌠ ⌠  1 tan-1 η − 2 (3 − t ) η  τ(ρ ,φ ) ρ dρ dφ
G
u (ρ,φ) =
π ⌡ ⌡  R R G 21 a 2(1 − t )2  0 0 0 0 0
0 0
G2 2π a 2 iφ
⌠ ⌠  q tan-1 η + η [( q/q) − t e ]  τ(ρ ,φ ) ρ dρ dφ ,
0
+
π ⌡ ⌡  Rq R a (1 − t )(1 − t ) 
2 0 0 0 0 0
0 0
(4.4.15)
where R and η are defined by (4.1.14), q is defined by (4.1.28), a bar indicates
the complex conjugate value, and
ρρ
0 i(φ-φ0)
t = 2 e . (4.4.16)
a
The potential functions can be found by substitution of (4.4.15) and (4.4.9)

in (4.4.2) and evaluation of the resulting integrals. This looks at first somewhat
difficult, nevertheless, it will be shown here that all the Green’s functions can be
expressed in elementary functions. Note the following property:
(3 − t ) η 
Λ
1 η
tan-1( ) − 2
 R R a (1 − t )2 
2 iφ
= −Λ
q η η [( q/q) − t e 0] 
tan-1 + 2 . (4.4.17)
 Rq R a (1 − t )(1 − t ) 
Introduce the following notation:
( a 2 − r 2)1/2( a 2 − ρ20)1/2
1 -1  ,
E ( N, N ) = tan
1 0 R ( N,N )
0
 a R ( N,N
0
) 
i (ψ-φ0)
(3 a 2 − r ρ e ) ( a 2 − r 2)1/2( a 2 − ρ20)1/2
0
E ( N, N ) = i (ψ-φ0) 2 ,
2 0
a( a2 − rρ e )
0
iφ
r ei ψ − ρ e 0
( a 2 − r 2)1/2( a 2 − ρ20)1/2
tan  
0
-1
E ( N, N ) =
3 0 -i φ  a R ( N,N ) 
R ( N,N )( r e-i ψ − ρ e 0
) 0
0 0
iφ
a ( a 2 − r 2)1/2( a 2 − ρ20)1/2 r ei ψ − ρ e 0
rρ
+  −
0
e
i (ψ+φ0)
.
0
) r e-i ψ − ρ e 0 
i (ψ-φ ) - i (ψ-φ ) -i φ 2
( a 2 − r ρ e 0 )( a 2 − r ρ e 0 a
0 0
0
(4.4.18)
Here the points N and N are characterized by the cylindrical coordinates ( r ,ψ,0)
0
and (ρ0,φ0,0) respectively. Note the following relationships of symmetry:
E 1( N , N 0) = E 1( N 0, N ), E 2( N , N 0) = E 2( N 0, N ),
E 3( N , N 0) = E 3( N 0, N ). (4.4.19)
Let R ( M,N) denote the distance between the points M (ρ,φ, z ) and N ( r ,ψ,0). By
using (4.4.17) one may write
dS dS
⌠ ⌠ Λ[ E ( N, N ) − E ( N, N )] N
= −⌠ ⌠ ΛE ( N, N ) N
.
⌡⌡ 1 0 2 0 R ( M,N) ⌡⌡ 3 0 R ( M,N )
S S
(4.4.20)
Integration by parts in (4.4.20) leads to an important property:
⌠ ⌠ [ E ( N, N ) − E ( N, N )]Λ 1 d S = −⌠ ⌠ E ( N, N )Λ 1 d S .
⌡⌡ 1 0 2 0 R ( M,N) N ⌡⌡ 3 0 R ( M,N) N
S S
(4.4.21)
Two more properties can be obtained by applying Λ and Λ to both sides of
(4.4.21), namely,
⌠ ⌠ [ E ( N, N ) − E ( N, N )]Λ2 1 d S = −⌠ ⌠ E ( N, N )∆ 1 d S ,
⌡⌡ 1 0 2 0 R ( M,N) N ⌡⌡ 3 0 R ( M,N ) N
S S
(4.4.22)
⌠ ⌠ [ E ( N, N ) − E ( N, N )]∆ 1 d S = −⌠ ⌠ E ( N, N )Λ2 1 d S .
⌡⌡ 1 0 2 0 R ( M,N) N ⌡⌡ 3 0 R ( M,N) N
S S
(4.4.23)
The properties (4.4.21−4.4.23) will allow us to substitute the evaluation of various
integrals involving E 3, which look very formidable, by evaluation of integrals
involving expressions E 1 and E 2, some of which have already been computed
(4.1.27−4.1.31), and the remaining can be evaluated relatively easy (see Appendix
A4.4).
Introduce the notation
X = Λχ + Λχ , Y =Λχ − Λχ . (4.4.24)
In order to obtain the complete solution, we shall need the following expressions
for various derivatives of X and Y : the tangential displacements are defined by

Λ X and Λ Y ; the normal displacements by ∂ X /∂ z ; the field of stresses may be
computed through ∂2 X /∂ z 2, Λ2 X , Λ2 Y , Λ(∂ X /∂ z ), Λ(∂ Y /∂ z ). All the Green’s
functions involved can be expressed as various derivatives of two fundamental
functions, namely,
K 1( M , N 0) = ⌠ ⌠ E 1( N , N 0) ln[ R ( M,N ) + z ] d S N ,
⌡⌡
S
K 2( M , N 0) = ⌠ ⌠ E 2( N , N 0) ln[ R ( M,N ) + z ] d S N . (4.4.25)

⌡⌡
S
Rewrite formula (4.4.15) as
G1 G 22
u( N) = ⌠ ⌠ [ E ( N, N ) − E 2( N , N 0)] τ( N 0) d S N
π⌡ ⌡ 1 0 G 21 0
S
G2
+ ⌠ ⌠ E ( N , N ) τ( N ) d S .
π⌡ ⌡ 3 0 0 N 0
S
(4.4.26)
By substituting (4.4.9) and (4.4.26) in (4.4.24) and using the properties
(4.4.21−4.4.23), we obtain the following results:
G1 − G2 G G
X = Λ⌠ ⌠ K + 2 K  τ d S + Λ⌠ ⌠ K + 2 K  τ d S ,
π  ⌡⌡ 1 G 1 2 ⌡⌡ 1 G 1 2 
S S
G1 + G2 G G
Y = −Λ⌠ ⌠ K − 2 K  τ d S + Λ⌠ ⌠ K − 2 K  τ d S .
π  ⌡⌡ 1 G 1 2 ⌡⌡ 1 G 1 2 
S S
We shall only need the following derivatives of X and Y for the complete
solution:
G1 − G2 G G
ΛX = − ∂2 ⌠ ⌠ K + 2 K  τ d S + Λ2⌠ ⌠ K + 2 K  τ d S ,
π  ∂ z2 ⌡ ⌡  1 G 1 2 ⌡⌡ 1 G 1 2 
S S
(4.4.27)
G1 + G2 G2 G2
ΛY =  ∂2 ⌠ ⌠  K − K 2 τ d S + Λ2⌠ ⌠  K 1 − K 2 τ d S ,
π ∂ z ⌡ ⌡ 
2 1 G 1
 ⌡⌡ G 1
 
S S
(4.4.28)
∂X G1 − G2 G2 G2
∂  ⌠⌠
= Λ K + K  τ d S + Λ⌠ ⌠  K + K  τ d S ,
∂z π ∂z  ⌡ ⌡  1 G 1 2 ⌡ ⌡ 1 G 2 
1
S S
(4.4.29)
∂Y G + G G G
=
1 2 ∂
−Λ ⌠ ⌠  K −
2

K τ dS + Λ⌠ ⌠  K −
2
K  τ d S ,
∂z π ∂z  ⌡ ⌡  1 G 2 ⌡⌡ 1 G 2 
1 1
S S
(4.4.30)
G − G G G
∂  ∂2 ⌠ ⌠   2⌠ ⌠ 
K τd S ,
∂ 1 2 2 2
ΛX = − 2 K + K τd S + Λ K +
∂z π ∂ z ∂ z ⌡ ⌡  1 G 2 ⌡⌡ 1 G 2 
1 1
S S
(4.4.31)
G + G G G
∂  ∂2 ⌠ ⌠   2⌠ ⌠ 
K  τ d S ,
∂ 1 2 2 2
ΛY = K − K τ d S + Λ K −
∂z π ∂ z ∂ z ⌡ ⌡ 
2 1 G  2 ⌡ ⌡ 1 G 2 
1 1
S S
(4.4.32)
∂2 X G − G G G
∂2  ⌠ ⌠   
⌠ ⌠ K  τ d S ,
1 2 2 2
= Λ K + K τ d S + Λ K +
∂ z2 π ∂z  ⌡ ⌡ 
2 1 G 
1
2 ⌡⌡ 1 G 2
1

S S
(4.4.33)
G − G G G
Λ− 2⌠ ⌠  K + K  τ d S + Λ2⌠ ⌠  K + K  τ d S ,
1 2 ∂2 2 2
ΛX =
2
π  ∂z ⌡ ⌡  1 G 2
1
⌡⌡  1 G
1
2 
S S
(4.4.34)
G + G G G
Λ 2⌠ ⌠  K − K  τ d S + Λ2 ⌠ ⌠  K − K  τ d S ,
1 2 ∂2 2 2
Λ2 Y =
π ∂ z ⌡ ⌡  1 G
1
2  ⌡⌡ 1 G
1
2 
S S
(4.4.35)
All the integrals in (4.4.27−4.4.35) are computed in elementary functions in
Appendix A4.4 and in (4.1.27−4.1.31).
The results above may be applied to solving the problem of a tangential
point force loading of a penny-shaped crack. The solution will give us all the
Green’s functions, related to the case. Consider an infinite transversely isotropic

solid weakened in the plane z =0 by a penny-shaped crack of radius a . Let two
equal concentrated forces T = T + iT be applied to the crack faces antisymmetrically
x y
±
at the point N (ρ ,φ ,0 ). The previously obtained results give the complete
0 0 0
solution in elementary functions:
γγH
1   
2 G G
Σ   f ( z ) T 
1 2 2 2
u =
π  − f (z ) +
m − 1 2 k
f (z ) T + f (z ) +
G 7 k  16 k G 8 k
k=1 k 1  1 
β 
G G
+  f (z ) −
2
 
f (z ) T + f (z ) −
2
f ( z )  T , (4.4.36)
π 2 3 G 7 3  16 3 G 8 3
 1 1 
2 m G
Σ  f ( z ) + 2 f ( z ) T ,
2 k
w = Hγ γ ℜ (4.4.37)
π 1 2 ( m − 1)γ  1 k G 9 k
k=1 k k 1
 2γ1γ2 2 G
T ,
Σ  1 1   2
σ = ℜ 2 (−1) k+1
− f ( z ) + f ( z )
1
π ( γ 1 − γ 2 ) γ23( m k + 1) γ2k   5 k G 10 k 
1 
k=1
(4.4.38)
1 
2 G
Σ f ( z ) T
2 2
σ = −
2
A Hγ γ
π 66 1 2  f (z ) +
m − 1 5 k G 13 k 
k=1 k  1

G
  1 
G
f ( z ) T
2 2
+ f (z ) + f ( z ) T  − 2  −f ( z ) +
 11 k G 12 k 
1  π γ 3
 5 3 G 13 3 
1

G 
f ( z )  T ,
2
+ f (z ) − (4.4.39)
 11 3 G 12 3 
1 
 γ 1γ 2 2 G 
Σ  f ( z ) T  ,
2
σ = ℜ 2 (−1) k+1
f (z ) + (4.4.40)
z
π ( γ 1 − γ 2 ) 5 k G 10 k 
1 
k=1
γγ
(−1)k 
2 G G
Σ   f ( z ) T 
1 2 2 2
τ =
γ  f (z ) +
3 k
f ( z ) T + −f ( z ) +
G 14 k   4 k G 15 k 
z 2π2(γ − γ )
1 2 k=1 k  1 1 
1  T  .
G G
+  f ( z ) −
2
f ( z )  T +  f ( z ) +
2
f ( z ) (4.4.41)
2π2  3 3 G 14 3  4 3 G 15 3 
 1 1 
Here ℜ indicates the real part, the elastic coefficients are defined by (2.1.9), and
the functions f with subindex less than 6 are given by formulae (4.2.7−4.2.11),
and the remaining functions are computed in Appendix A4.4, namely,
ha 2 3 3( l 22 − a 2)1/2
tan-1 2 
t s
f ( z) = − − 2 1/2 ,
7 s  s
2 2
l 22 − a t2
s 3
( l 2 − a ) 
(4.4.42)
1 2 (ζ − 1)1/2 -1 1  -1
(a −

2
l 21)1/2
f ( z ) = ( a − ρ0) 
2 1/2
tan − tan
8 q q  (ζ − 1) 

1/2
a (ζ − 1) 
1/2
ei φ ( a − l 1)   − 1 ,
2 2 1/2
ρ2
− 1 + (4.4.43)
ρ a  l 22 − ρρ e 0 
i(φ-φ )

0
( a 2 − ρ20)1/2  a ( l 22 − a 2)1/2
iφ 1 -1 a
f ( z ) = −ρe  t sin (l ) + i(φ-φ )
9 a3 (1 − t )( l 22 − ρρ e 0 )
 2
0
1 -1 a (1 − t ) 
1/2 
( l 22 − a 2)1/2 
− tan , (4.4.44)
t (1 − t )3/2

h ρei φ(3 l 22 − a 2 t )
f ( z) = − , (4.4.45)
10 ( l 22 − l 21)( l 22 − a 2 t )2
iφ
4ρ e 0
1
3 R 40 + 6 R 20 z 2 − z 4
tan-1( ) − ( a 2 − ρ20)1/2  2 −
h z 8 0
f ( z) = 
11 q
 R 30 q 2 R
0
 sq s2 q
2 iφ
3ρ20e 0 iφ iφ
+  tan-1 s  − e 2e + 3
s4  ( l 22 − a 2)1/2 ρ ρ q
3(ζ − 1)1/2  -1 ( a 2 − l 21)1/2

−
1 
1/2 − tan
-1
tan
q 2
 (ζ − 1) a (ζ − 1)  1/2
iφ iφ
2 iφ 2ρ0e 0
iφ ρ0e 0
2 ( l 2 − a ) t 
2 2
+
ha e  − − +  2
2e 2
−
ρs2  s2 ρ q  s q  l 22 − a 2 t 
ρ 3 iφ ei φ( l 22 − ρ2) z2 q 
h  q e 2 i φ

− + − + 2e , (4.4.46)
R 20 + h 2  l 22 − l 21 ρq R 20 q

1 3(ζ − 1)1/2 -1 1  − tan 

( a 2 − l 21)1/2

f ( z) = ( a − ρ0) 
2 2 1/2
tan -1
12 q q2  (ζ − 1)1/2
 a (ζ − 1)1/2
e2 i φ( a 2 − l 21)1/2 l 22 + ρ2 2ρ2( l 22 − a 2)
−  + 
i(φ-φ0) 2 + 1
a ( l 22 − l 21)  l 22 − ρρ ei(φ-φ0) ( l 2 − ρρ e
2
) 
0 0
ei φ3 2ei φ ( a 2 − l 21)1/2


l 22 + 2ρ2
1 ei φ 
ρ 
+ + − + 2 ( + ) ,
ρ q ρ a q ( l 22 − ρρ ei(φ-φ0) ) q
0 
(4.4.47)
a 2 iφ 15( l 22 − a) 2 1/2
f ( z ) = − h  2 ρ e 0 tan-1 2  − 15
s
13
s
0  s5 ( l 2 − a 2)1/2 s4
φ
5 2t  +
ρei (3 l 22 − a 2 t ) 
+ + , (4.4.48)
s 2( l 22 − a 2 t ) ( l 22 − a 2 t )2 ( l 22 − l 21)( l 22 − a 2 t )2

( a 2 − ρ20)1/2  a ( l 22 − a 2)1/2 3( l 22 − l 21 t )
f ( z) =  2 
 ( l 2 − l 1)( l 2 − ρρ0e ) 
14 a 3(1 − t ) 2 2 i(φ-φ0) 1 − t
i(φ-φ0)
ρρ e (2 l 22 + l 21 t − 3ρ2) 
0
 − 3 -1 a (1 − t ) 
1/2
( l 22 − a 2)1/2 
+ tan ,
l 22 − ρρ e
i(φ-φ0)
 (1 − t )3/2
0
(4.4.49)
2 2 iφ i(φ-φ0)
ρe (a 2
− ρ20)1/2( l 22 − a) 2 1/2
(3 l 22 − ρρ e )
0
f ( z) = i(φ-φ0) 2 , (4.4.50)
15
l 22( l 22 − l 21)( l 22 − ρρ e )
0
iφ
ρ0e 0
1
R 20
+ z 2
tan-1( ) + ( a 2 − ρ20)1/2  2
h z 2  -1 s 
f ( z) =  − tan
16 q
 R0q
R
0
s  s q  (l2 − a ) 
2 2 1/2
 + e  − e ha 2.
(ζ − 1)1/2 -1 1 ( a 2 − l 21)1/2 iφ iφ
+ tan − tan-1
q  (ζ − 1)1/2 a (ζ − 1)1/2 ρ ρs2

(4.4.51)
The solution (4.4.36−4.4.51) presents, in fact, the explicit expressions for the
Green’s functions, and allows us to write a complete solution for the case of
arbitrary tangential loading in quadratures. The general results simplify
significantly for z =0, namely,
G 2
 1 tan-1 η − 2 (3 − t ) η T
1 G
u =
π  R R G 21 a 2(1 − t )2 
G 2 iφ
+  q 2
tan-1 η η [( q/q) − t e 0] 
+ 2 T, for ρ< a , (4.4.52)
π  Rq R a (1 − t )(1 − t ) 
-i φ
 1 G e 0 a2
2
 1 
w = H α( a − ρ0) ℜ
a  
2 2 1/2
+ − T for ρ≤ a ,
 qs G ρ  s3
1 0 
-i φ
2  1 -1 s
G a 2e 0
2
 s
w = H α( a − ρ0) ℜ tan 2
2 2 1/2
+ tan-1 2

π qs (ρ − a )
2 1/2 G ρs 3
(ρ − a 2)1/2
1 0
-i φ
(ρ2 − a 2)1/2 e 0 -1 a 

ρ  
− − sin ( ) T , for ρ> a (4.4.53)
s2 q aρ
0 
 γ + γ
σ = 2 ℜ2π H A γ γ −  1 tan-1(η ) + 1
2 1 2 η
1 π  66 1 2

γγ
1 2
qR R a (1 − t )(1 − t )q
2
ρe-i φ 
G
2 ηρe-i φ(3 − t ) T ,
+ 2 + for ρ< a ,
a − ρ2 G ( a 2 − ρ2) a 2(1 − t )2
1 
σ = 0, for ρ> a , (4.4.54)
1
1 
G
1 
f (0) T 
2
σ = 2  2π A H γ γ + f (0) T +
2 π  66

1 2 γ 
3
5 G 12
1

 1 
G 
f (0) T ,
2
+ 2π A H γ γ − f (0) T +
 66 1 2 γ  11
3
G 13
1

σ = 0, (4.4.55)
z
G e2 i φ(3ρ − ρ e
i(φ-φ0)
( a 2 − ρ20)1/2 )T
τ = T +
2 0
, for ρ> a .
z π2(ρ2 − a 2)1/2 R 2 G
1 ρ(ρ − ρ e
i(φ-φ0) 2
) 
0
(4.4.56)
Here R and η are defined by (4.1.14). The second and the third mode stress
intensity factors can be expressed through the decomposition τ(n)=τ + i τ , which is
zn tz
(n) i φ
related to τ by a relationship τ =τ e . Introducing the complex stress intensity
z z
factor
k + ik = lim{(ρ − a )1/2τ e-i φ}, (4.4.57)

2 3 ρ→ a z
one gets from (4.4.56)
( a 2 − ρ20)1/2
 T e-i φ
k + ik =
2 3 π2(2 a )1/2 ρ20 + a 2 − 2 a ρ0cos(φ−φ0)
G 2 ei φ(3 a − ρ0e
i(φ-φ0)
)T
+ .
) 
G1 i(φ-φ0) 2
a ( a − ρ0e
(4.4.58)
In the general case of arbitrarily distributed loading, the stress intensity factor
takes the form
2π a
( a 2 − ρ20)1/2 e-i φτ(ρ ,φ )
k 2 + ik 3 = ⌠ ⌠  0 0
⌡ ⌡ π2(2 a )1/2 ρ20 + a 2 − 2 a ρ0cos(φ−φ0)

0 0
i(φ-φ0) φ
G 2 (3 a − ρ0e ) ei τ(ρ0,φ0)
+ ρ dρ dφ , (4.4.59)
G1 a ( a − ρ0e
i(φ-φ0) 2
)  0 0 0
which is in agreement with (2.7.30).
Exercise 4.4
1. Derive the general solution in the case of isotropy.
2. Find the isotropic equivalent of (4.4.14).
3. Verify the property (4.4.17).
4. Derive the solution (4.4.36) −(4.4.51).
5. Rederive the solution (4.4.36) −(4.4.51) for the case of isotropy.
4.5 Penny-shaped crack under uniform pressure
Let a penny-shaped crack of radius a be opened by the pressure p =const.

In this case one gets from (4.1.12)
ω(ρ,φ) = 4 Hp ( a 2 − ρ2)1/2. (4.5.1)
The potential function F can be obtained by substitution of (4.5.1) in (4.1.4).

The integral can be computed in elementary functions (A4.1.4), giving
2 a 2 − 3 l 21
F = 2π Hp (2 a 2 + 2 z 2 − ρ2)sin-1( ) − ( l 22 − a 2)1/2.
a
 l2 a 
(4.5.2)
The complete solution can be expressed through various derivatives of the
potential function, as prescribed by formulae (2.1.6) and (2.1.12). All the
derivatives are given in Appendix A4.1. The solution is:
 γk  -1 a
2
a ( l 22k − a 2)1/2
,
u = 2 Hp ρe
k=1  k
iφ
Σ
m − 1sin (l ) −
2k
l 22k 
(4.5.3)
 mk 2
( a 2 − l 2 )1/2 − zsin-1( a ),

w = 4 Hp 
k=1 
Σ
mk − 1  1k l 2k 

(4.5.4)
γ2k − ( m k + 1)γ23
a 
2
a ( l 22k − a 2)1/2
σ1 = 8 Hp A 66 Σ
k=1
γk( m k − 1)

 ( l 22k − l 21k)
− sin-1(
l 2k 
) ,

(4.5.5)
2
 γk − a)
l 21k( l 22k 2 1/2
,
σ = −8 HpA66a e
2
2 iφ
Σ m − 1  2 2
k=1  k
l 2k( l 2k − l 21k) 
(4.5.6)
 k+1  a ( l 22k − a 2)1/2

2

Σ -1 a 
2p
σz =
π(γ − γ )  (−1) γ
k
− sin (
l 2k 
) , (4.5.7)
2 k=1 
l 22k − l 21k
1 
2 pa 2ρei φ  k ( a 2 − l 21k)1/2 
2
τz =
π(γ − γ )
1 2
Σ (−1) 2 2
k=1 
l 2k( l 2k − l 21k)
,

(4.5.8)
1
l 1k = {[( a + ρ)2 + z 2k]1/2 − [( a − ρ)2 + z 2k]1/2},
2
1
l 2k = {[( a + ρ)2 + z 2k]1/2 + [( a − ρ)2 + z 2k]1/2},
2
z k = z /γ k , for k =1,2. (4.5.9)

The problem was first solved by Elliott (1949) by the integral transform method.
Our results are essentially in agreement with those of Elliott, who expressed them
in the form of integrals involving Bessel functions, namely,
∞
ρ
= ⌠
z
Cm x n-2cos( x ) J m( x ) exp(− x ) d x ,
⌡
n a a
0
∞
ρ
= ⌠
z
Sm x n-2sin( x ) J m( x ) exp(− x ) d x .
⌡
n a a
0
These integrals can be computed in elementary functions, and the results in our
notation are
a ( a 2 − l 21)1/2 a ( l 22 − a 2)1/2
a
S 01 = sin ( ),
-1
S 02 = , C 02 = ,
l2 l 22 − l 21 l 22 − l 21
a ( a 2 − l 21)1/2[ l 42 + a 2(ρ2 − 2 a 2 − 2 z 2)]

C 03 = ,
( l 22 − l 21)3
a ( l 22 − a 2)1/2[ a 2(2 a 2 + 2 z 2 − ρ2) − l 41]

S 03 = ,
( l 22 − l 21)3
z [ a − ( a 2 − l 21)1/2] a − ( a 2 − l 21)1/2
C 11 = , S 11 = ,
ρ( a 2 − l 21)1/2 ρ
z [ a − ( a 2 − l 21)1/2]2 a l 1( l 22 − a 2)1/2
S 21 = , S 12 = ,
ρ (a
2 2
− l 21)1/2 l 2( l 22 − l 21)
a − ( a 2 − l 21)1/2 a l 1( a 2 − l 21)1/2
C 12 = − ,
ρ l 2( l 22 − l 21)
a 2ρ( l 22 − a 2)1/2( l 22 + 3 l 21 − 4 a 2)
C 13 = ,
( l 22 − l 21)3
a 2ρ( a 2 − l 21)1/2(3 l 22 + l 21 − 4 a 2)
S 13 = ,
( l 22 − l 21)3
2 a [( l 22 − a 2)1/2 − z ] a ( l 22 − a 2)1/2
C 22 = − ,
ρ2 l 22 − l 21
2 a [ a − ( a 2 − l 21)1/2)] a ( a 2 − l 21)1/2
S 22 = − ,
ρ2 l 22 − l 21
a − ( a 2 − l 21)1/2 a l ( a 2 − l 21)1/2
C 23 =
2 a  −
1

ρ  ρ l ( l 22 − l 21) 
2
a ( a 2 − l 21)1/2[ l 42 + a 2(ρ2 − 2 a 2 − 2 z 2)]

− .
( l 22 − l 21)3
There are some misprints (or errors) in Elliott’s paper. For example, according
to his formula (4⋅2⋅5), the tangential displacement u vanishes on the plane z =0
which cannot be correct; there are some missing terms and obvious misprints in
formula (4⋅2⋅6).
In the limiting case of γ →γ →γ →1, m → m →1, H =(1−ν)/2πµ, A = A =µ,
1 2 3 1 2 44 66
formulae (4.5.3−4.5.8) give the complete solution for an isotropic body. By
using the L’Hôpital rule, one obtains
p ρei φ
a ( l 22 − a 2)1/2 2 a 2| z |( a 2 − l 21)1/2
 -1 a 
2πµ  ,
u = (1−2ν) − sin ( ) +
  l 22 l 
2
l 2 2
(
2 2l − l 2
1 )

(4.5.10)
2p 
2(1 − ν) ( a 2 − l 21)1/2 − z sin-1( )
z a
πµ 
w =
 | z| l 
2
a ( l 22 − a 2)1/2 
+ z sin-1( ) − ,
a
 l
2
l 22 − l 21 
(4.5.11)
2p 
a ( l 22 − a 2)1/2
(1 + 2ν)  2 − sin-1( )
a
σ1 =
π   l2 − l1 2 l 
 2
az 2[ l 41 + a 2(2 a 2 + 2 z 2 − 3ρ2)]
+ , (4.5.12)
( l 22 − l 21)3( l 22 − a 2)1/2

a l 21e2 i φ( l 22 − a 2)1/2 z 2[ a 2(6 l 22 − 2 l 21 + ρ2) − 5 l 42]

2p
σ2 =
π 1−2ν + ,
l 22( l 22 − l 21) ( l 22 − l 21)2( l 22 − a 2)
 
(4.5.13)
2p
a ( l 22 − a) 2 1/2
az 2
[ l 41 + a (2 a + 2 z − 3ρ )]
2 2 2 2
a
π ,
σz = − sin-1( ) −
l 22 − l 21 l ( l 22 − l 21)3( l 22 − a 2)1/2
 2 
(4.5.14)
iφ
z l e ( l 22 − a 2)1/2[ a 2(4 l 22 − 5ρ2) + l 41]
2p 1
τz = − . (4.5.15)
π l ( l 22 − l 21)3
2
This problem was first solved by Sneddon (1951), using the integral transform
method. He was seemingly unable to compute the potential function (4.5.2), so
he resorted to differentiation under the integral sign, with a subsequent
computation of various integrals involving Bessel functions. His final results are
given as elementary functions of the four parameters
r = (1 + ( z/a )2)1/2, R 2 = [(ρ/ a )2 + ( z/a )2 − 1]2 + 4( z/a )2,
φ = tan-1 2 
2 az
θ = tan-1( a/z ), 2 .
ρ + z − a 
2
This choice of parameters is not the best possible. Here is one illustration.
The expression for S 01 in Sneddon’s notation takes the form (Sneddon, 1951,
p.497)
r sinθ + √ R sin(φ/2) 
S 01 = tan-1 ,
rcosθ + √R cos(φ/2)
with the limitation ρ≠0, and no indication of what the result would be if ρ=0.
In our notation the corresponding result is sin-1( a / l ), with no limitations attached.
2
Introduction of Sneddon’s parameters r and θ seems to be unnecessary. There

exist relationships between his parameters R and φ, and our l and l , namely,
1 2
R 2 = ( l 22 − l 21)/ a 2, sin(φ/2) = ( a 2 − l 21)1/2/( l 22 − l 21)1/2.
These relationships may be used to compare the solutions, which are in good
agreement, except for some misprints: factor ζ is missing in Sneddon’s formula
(139, p. 496), and the last term in his formula (145, p. 499) should read − S 01,
rather than + S 00.
Exercise 4.5
1. Establish the result (4.5.3) −(4.5.8).
2. Derive (4.5.10) −(4.5.15).
3. Derive the expressions for the polar components of stresses and displacements
equivalent to (4.5.10) −(4.5.15).
4. Find the complete solution in the case where the penny-shaped crack is
loaded by a normal stress, whose magnitude is proportional to the x -coordinate.
4.6 Penny-shaped crack under uniform shear loading
Consider a circular crack of radius a in a transversely isotropic elastic

space, subjected to a uniform shear loading τ, where τ is a complex constant.
The solution of the integro-differential equation (4.4.14) in this case is
2( G 21 − G 22)
u = τ( a 2 − ρ2)1/2, for z =0 and ρ≤ a . (4.6.1)
G
1
Substitution of (4.6.1) in (4.4.9) leads to the integral
2π a
2( G 21 − G 22)
τ ⌠ ⌠ ( a 2 − ρ20)1/2 ln( R + z ) ρ dρ dφ ,
G
1 ⌡ ⌡ 0 0 0 0
0 0
which has been computed in Appendix A4.1, with all the necessary derivatives.
The complete solution will take the form:
2( G 21 − G 22) 
m − 1f 17( z1)τ + f 18( z1)τ
1
u =
G
1  1
1  
+ f ( z ) τ + f ( z ) τ + f ( z ) τ − f ( z ) τ ,
m − 1 17 2 18 2  17 3 18 3
2 
(4.6.2)
G 21 − G 22 2 m a ( l 22k − a) 2 1/2
(τρei φ + τρe-i φ) Σ sin-1( a ) −  1,

k
w =
2G m − 1 l l 22k  γk
1 k=1 k 2k
(4.6.3)
2( G 21 − G 22) A 2 γ23( m k + 1) − γ2k l (a 2
− l 21k)1/2
iφ -i φ
Σ
66 1k
σ1 = a ( τe + τe ) ,
G
1
γ2k( m k − 1) l ( l 22k − l 21k)
2k
k=1
(4.6.4)
2( G 21 − G 22) A  1 
m − 1f 19( z1)τ + f 20( z1)τ
66
σ2 =
G
1  1
1  
+ f ( z ) τ + f ( z ) τ + f ( z ) τ − f ( z ) τ , (4.6.5)
m − 1 19 2 20 2  19 3 20 3
2 
2γ γ β(τe-i φ + τei φ) 2 a l ( a 2 − l 21k)1/2

Σ(−1)
1 2 1k
σz = k
, (4.6.6)
π G 1(γ 1 − γ 2) l ( l 22k − l 21k)
k=1 2k
2γ γ  β
2

Σ (−1)k
1 2
τz =
πG1 γ − γ [ f ( z ) τ + f ( z ) τ] + H [ f ( z ) τ − f ( z ) τ]  .
γk 21 k 22 k 21 3 22 3
 1 2 k=1 
(4.6.7)
Here
2 a 3 − ( l 21 + 2 a 2)( a 2 − l 21)1/2
2 iφ
f ( z) = e , (4.6.8)
17 3ρ2
a
f ( z ) = z sin-1( ) − ( a 2 − l 21)1/2, (4.6.9)
18 l
2
a l 1( a 2 − l 21)1/2
3 iφ iφ
f 19( z ) = e − 4e f 17( z ), (4.6.10)
l 2( l 22 − l 21)
a l 1( a 2 − l 21)1/2
iφ
f 20( z ) = e , (4.6.11)
l 2( l 22 − l 21)
a ( l 22 − a 2)1/2
a
f 21( z ) = −sin ( ) + -1
, (4.6.12)
l2 l 22 − l 21
2 iφ
e a l 21( l 22 − a 2)1/2
f 22( z ) = , (4.6.13)
l 22( l 22 − l 21)
A complete solution to this problem for the case of isotropy can be found in
(Westmann, 1965).
In the case of isotropy, formulae (4.6.2−4.6.7) transform into

(−5 + 4ν) z sin-1( ) + 4(1 − ν)( a 2 − l 21)1/2τ
1 a
πµ(2 − ν) 
u =

l2 
za ( l 22 − a 2)1/2 l 21 
+ τ + τe2 iφ , (4.6.14)
l 22 − l 21  l 22

(τe + τe )ρ1−2ν -1 a
iφ -i φ a ( l 22 − a 2)1/2 za 2( a 2 − l 21)1/2
 +
πµ(2 − ν)  2  ,
w = sin ( ) −
l2
 l 22  l 22( l 22 − l 21)

(4.6.15)
a l 1( a − 2
l 21)1/2
2(τe + τe )
iφ -i φ
π(2 − ν) 
σ1 = −2(1 + ν)
l 2( l 22 − l 21)

z l 1( l 22 − a 2)1/2[ a 2(4 l 22 − 5ρ2) + l 41]

+ , (4.6.16)
l 2( l 22 − l 21)3


iφ a l 1( a 2 − l 21)1/2
2e
π(2 − ν) 
σ2 = − 4(1 − ν) τ
l 2( l 22 − l 21)

z l 1( l 22 − a 2)1/2 4 a 2 a 2(4 l 22 − 5ρ2) + l 41 

+  τe
2 iφ
− (τ + τe2 ),
iφ
l 2( l 22 − l 21)  l 22 ( l 22 − l 21)2 
(4.6.17)
iφ
2(τe + τe )
-i φ z l 1( l 22 − a ) [a
2 1/2 2
(4 l 22 − 5ρ ) +
2
l 41]
σz = − , (4.6.18)
π(2 − ν) l 2( l 22 − l 21)3
 
a ( l 22 − a 2)1/2
− sin-1( )
2 a
π(2 − ν) 
τz = (2 − ν)
  l2 − l12 2 l 2

z ( a 2 − l 21)1/2[ l 41 + a 2(2 a 2 + 2 z 2 − 3ρ2)]

+  τ + ν a ( l 2 − a 2)1/2
( l 22 − l 21)3   2
2 iφ
z ( a 2 − l 21)1/2[ a 2(6 l 22 − 2 l 21 + ρ2) − 5 l 42]
 1
l 2e 
l 22( l 22 − l 21) 
+ τ . (4.6.19)
( l 22 − l 21)2
Exercise 4.6
1. Investigate a penny-shaped crack of radius a in a transversely isotropic

iφ
elastic body, subjected to the shear loading τ= c ρe , where c =const.
Answer: the main potential function will be proportional to the integral
2π a
iφ
I = ⌠ ⌠ ln( R 0 + z )( a 2 − ρ20)1/2e 0
ρ20 dρ0dφ0
⌡ ⌡
0 0

ρe ( a 2 − l 21)1/2 a 2 + 7ρ2 −
π iφ 4 19 2
= l − 4 l 22
8
 3 3 1
 + z(4 a 2 − 3ρ2 + 4 z2)sin-1( a ) − 16 a .

2(8 a 4 + 4 a 2 l 21 + 3 l 41) 5
+
15ρ2  l2 15ρ2

It might be a good exercise for the reader to perform the differentiation and
write a complete solution. As an example, here are the first two derivatives
 l 21 l 21 
ρei φ a ( l 22 − a 2)1/215 2 − 12 − 2 2 + sin-1( )4 a 2 − 3ρ2 + 12 z 2,
∂I π a
=
∂z 8
  a l 2 l 
2

∂2 I i φ 2 2 1/2a  + 3 zsin-1( a ).

2
= πρe ( a − l ) 2 − 3
∂z 2
 1
 l2  l 
2
Note : ∂ I /∂ z is proportional to the main potential function for the case of linear
normal loading of a penny-shaped crack.
2. Find the expressions for the energy release rate by using a procedure similar
to the one employed by Kassir and Sih (1968).
Answer: G = 2π2 H k 21, G2 = π2( G 1 − G 2) k 22, G3 = π2( G 1 + G 2) k 23.
1
3. Find the Green’s functions for a semi-infinite plane crack in a transversely

isotropic space, subjected to a shear loading.
Hint : consider the limiting case of (4.4.36−4.4.41), when the radius a →∞, and
the coordinate origin moves from the circle centre to its boundary.
4.7 Asymptotic behavior of stresses and displacements

near the crack rim
Kassir and Sih (1975) have derived expressions relating the stress intensity
factors to the field of stresses and displacements in the immediate vicinity of the
crack edge, by using their very complicated solution for an elliptical crack.
Their derivation looks so complicated that nobody so far was capable to repeat it
and verify its accuracy. There is a notion in fracture mechanics that the
asymptotic behavior is defined completely by three stress intensity factors, and is
invariant for any crack, with a smooth boundary. If this is so, then we can
obtain the same results from the much simpler solution for a penny-shaped crack.
The results, presented here, are simpler than those of Kassir and Sih, and
obtained in a simple manner.
4.7 Asymptotic behavior of stresses and displacements 253
Asymptotic behavior for mode I loading. Though formulae (4.5.3−4.5.8)

are valid for a penny-shaped crack subjected to a uniform pressure only, they
can be used to obtain some general results which are valid for an arbitrary crack
with a smooth boundary subjected to arbitrary loading, in other words, we can
explore the asymptotic behavior of displacements and stresses near the rim of a
general crack. Introduce the local system of spherical coordinates ( r ,θ,φ), with
the coordinate origin at the crack rim. The following asymptotics are valid for
the main parameters used in (4.5.3−4.5.8):
1 1
ρ = a + r cosθ, l 1k ≈ a − rS 2k, l 2k ≈ a + rT 2k,
2 2
z = r sinθ, l 22k − l 21k ≈ 2 arQk, ( a 2 − l 21k)1/2 ≈ ( ar )1/2 S ,

k
a r
( l 22k − a 2)1/2 ≈ ( ar )1/2 T , sin-1( ) ≈ −( )1/2 T . (4.7.1)
k l2 a k
Here the notation was introduced:
Q k = [cos2θ + (1/γ2k)sin2θ]1/2 S k = [ Q k − cosθ]1/2,
T k = [ Q k + cosθ]1/2, for k = 1,2,3. (4.7.2)

Introducing the opening mode stress intensity factor
p √2 a
k1 = , (4.7.3)
π
the following asymptotic expressions can be derived by substitution of (4.7.1) and

(4.7.3) in (4.5.3−4.5.8):
γ1 T 1 γ2 T 2
u = u n = −2π Hk 1√2 r  +  + 0(1), (4.7.4)
 1
m − 1 m 2
− 1 
m S m S
w = 2π Hk 1√2 r   + 0( r),
1 1 2 2
+ (4.7.5)
 1
m − 1 m 2
− 1 
2 γ2k − ( m k + 1)γ23 T k
2
σ1 = 2π A 66 Hk 1( )1/2
r Σ k=1
γk( m k − 1) Qk
+ 0(1), (4.7.6)
2 γT
Σ(m
2 k k
σ = −2π A Hk ( )1/2 + 0(√ r ), (4.7.7)
2 66 1 r − 1) Q
k=1 k k
k γT γT
σ =  1 1 − 2 2 + 0(1),
1
(4.7.8)
z √2 r (γ − γ )  Q 1 Q 
2
1 2
k S S
τ = τ = −  1 − 2  + 0(√r),
1
(4.7.9)
z zn √2 r (γ − γ ) Q 1 Q
2
1 2
These results were computed for φ=0. This assumption allows us to avoid a
cumbersome axis transformation, without loss of generality. The parameter σ in
1
this case is interpreted as the sum σ + σ , and σ = σ − σ + 2iτ . By
n t 2 n t nt
taking the sum and the difference of (4.7.6) and (4.7.7), one can get
k T T
σ = − 1 + 2 ,
1
n √2 r (γ − γ ) γ1 Q 1 γQ
2 2
1 2
[2γ21 − ( m + 1)γ23] T [2γ22 − ( m + 1)γ23] T

σ = π A Hk ( )1/2 .
2 1 1 2 2
+
t 66 1 r  γ ( m − 1) Q
1 1 1
γ ( m − 1) Q
2 2 2

Formulae (4.7.4−4.7.9) are essentially in agreement with the results of Kassir
and Sih (1975), except for some misprints, for example, one should read √ n and
1
√ n 2 instead of n 1 and n 2 in the denominator of the terms in curly brackets of
their formula (8.94c). In order to compare our results with those of Kassir and
Sih, one should keep in mind that their definition of the stress intensity factor is
√2times greater than ours, their notation n corresponds to our γ2k; Kassir and Sih
k
seem to have not noticed the properties (4.1.10) and the relationship
S =(sinθ)/(γ T ), which in some cases can be used to simplify their results
k k k
significantly. For example, they have an expression [ A m − A γ2k]/[ A ( m +
13 k 11 44 k
1)] in formula (8.95a), without realizing that it is equal to −1 for k =1,2.
The asymptotic behavior of the displacements and stresses near the crack
edge in an isotropic body can be found from either (4.7.4−4.7.9) or
(4.5.10−4.5.15). The result is
k 1√ r
cos( ) 2(1 − ν) − cos2( ) + 0(1),
θ θ
u = un = (4.7.10)
2µ 2  2
k 1√ r
sin 2(1 − ν) − cos2( ) + 0( r ),
θ θ
w = (4.7.11)
µ 2  2
k1
cos 1 + 2ν − sin sin  + 0(1),
θ θ 3θ
σ = (4.7.12)
1 √r 2  2 2
k1
cos 1 − 2ν − sin sin  + 0(√ r ),
θ θ 3θ
σ2 = (4.7.13)
√r 2  2 2
k1
cos 1 + sin sin  + 0(1),
θ θ 3θ
σz = (4.7.14)
√r 2  2 2
k1
3θ
τz = τzn = sinθ cos + 0(√ r ), (4.7.15)
2√ r 2
which is in agreement with the results given in (Sih and Liebowitz, 1968).
Asymptotic behavior for mode II and III loading. We can derive again some
results of general nature, namely, the asymptotic behavior of the field of stresses
and displacements in the neighbourhood of the edge of a flat crack with a
smooth boundary. We recall that at φ=0 the decompositions u = u x+ iu y and
τz=τzx+ i τyz are equal to u (n)= u n+ iu t and τ(n)=τzn+ i τtz respectively; σ1 is understood
as σn+σt, and σ2=σn−σt+2 i τnt. This will allow us to avoid a cumbersome axis
transformation. The complex stress intensity factor, introduced in (4.4.57), can be
expressed through the prescribed shear loading τ as
G2
√2 a 
k = k 2 + ik 3 = τ + τ, (4.7.16)
π  G1 
and its inversion gives

π G 1( k G 1 − k G 2)
τ = . (4.7.17)
√2 a ( G 21 − G 22)
Substitution of (4.7.1) and (4.7.17) in (4.6.2−4.6.13) yields
k 2γ1γ2√2 r S S i k 3γ3√2 r
u n + iu t = − 1
+
2
 + S 3,
A 44(γ1 − γ2) m 1 + 1 m 2 +1 A 44
(4.7.18)
k 2γ1γ2√2 r m T m T
w = − 1 1
+
2 2
, (4.7.19)
A 44(γ1 − γ2) γ1( m 1 + 1) γ2( m 2 + 1)
2 γ23( m k + 1) − γ2k S k
σ1 = 2π k 2γ1γ2 H A 66
2
( )1/2
r Σ
k=1
γ2k( m k − 1) Qk
, (4.7.20)
S1 S2 ik 3√2S 3
2 1/2 
σ2 = 2π k 2γ1γ2 H A 66( ) + − ,
r ( m 1 − 1) Q 1 ( m − 1) Q 
2 2
γ 3√ r Q 3
(4.7.21)
k 2γ 1γ 2 S S
σz = −  1 − 2 , (4.7.22)
√2 r (γ1 − γ2) Q 1 Q 2
k 2γ 1γ 2 T T ik T
τzn + iτ = −  1 − 2  + 3 3. (4.7.23)
tz √2 r (γ1 − γ2) γ1 Q 1 γ 2 Q 2 √2 r Q 3
By taking the sum and the difference of (4.7.20) and (4.7.21), one gets
k 2γ 1γ 2 S S
σn =  1 − 2 , (4.7.24)
√2 r (γ1 − γ2) γ21 Q 1 γ22 Q 2
k 2γ 1γ 2 S S
(γ − γ ) 2 − 2Q −  2 − 2Q 
11 1 1 1 1 2
σt =
√2 r γ1 γ3 1 γ2 γ3 2
 1 2
S1 S2 
+ 2π H A  + , (4.7.25)
66Q Q 
1 2 
k3S3
τnt = − . (4.7.26)
√2 r γ3 Q 3
Our formulae (4.7.18−4.7.26) are in relatively good agreement with similar results
of Kassir and Sih (1975), except for formula (8.96b, p.371) for u which should
t
correspond to the imaginary part of our (4.7.18). Formula by Kassir and Sih
(8.96b) seems to be in error because it implies that u depends on k , γ , and
t 2 1
γ2, which is wrong: our result relates u t to k 3 and γ3 only, as it should be.
There are several misprints in their formulae (8.96a) and (8.96c). The remaining
formulae are in agreement, though the formulae by Kassir and Sih (1975) look
more complicated than ours, mainly because they did not notice the properties
(4.1.10) which could make some expressions much simpler.
Exercise 4.7
2. Derive (4.7.4) −(4.7.9).
3. Derive (4.7.18) −(4.7.23).
4. Find the asymptotic behavior of the stresses and displacements near the crack
rim for the case of mode II and III loading in isotropic bodies.
θ  
sin 2(1 − ν) + cos2( ) k + 2 i k ,
√r θ
Answer: u n + iu t =
µ 2  2 2 3
 
cos( ) −(1 − 2ν) + sin2( ),

√r θ θ
w = k
2 µ 2  2
k
sin( )−2(1 + ν) − cos( )cos( ),
2 θ θ 3θ
σ =
1 √r 2 2 2 
sin(θ/2)  
−2(1 − ν) − cos(2)cos( 2 )k 2 − 2 ik 3,
θ 3θ
σ2 =
√r
 
k2
3θ
σz = sinθ cos( ),
2√ r 2
cos(θ/2)  θ 3θ  
τzn + i τtz = 1 − sin( ) sin( ) k
2  2
+ ik .
√r 2 3
 
Note : compare with the results presented in Sih and Liebowitz (1968).
k S
3 3
τ = − . (4.7.26)
nt √2 r γ Q
3 3
Our formulae (4.7.18−4.7.26) are in relatively good agreement with similar results
of Kassir and Sih (1975), except for formula (8.96b, p.371) for u which should
t
correspond to the imaginary part of our (4.7.18). Formula by Kassir and Sih
(8.96b) seems to be in error because it implies that u depends on k , γ , and
t 2 1
γ , which is wrong: our result relates u to k and γ only, as it should be.
2 t 3 3
There are several misprints in their formulae (8.96a) and (8.96c). The remaining
formulae are in agreement, though the formulae by Kassir and Sih (1975) look
more complicated than ours, mainly because they did not notice the properties
(4.1.10) which could make some expressions much simpler.
Exercise 4.7
2. Derive (4.7.4) −(4.7.9).
3. Derive (4.7.18) −(4.7.23).
4. Find the asymptotic behavior of the stresses and displacements near the crack
rim for the case of mode II and III loading in isotropic bodies.
θ  
sin 2(1 − ν) + cos2( ) k + 2 i k ,
√r θ
Answer: u + iu =
n t µ 2  2 2 3
 
cos( ) −(1 − 2ν) + sin2( ),

√r θ θ
w = k2
µ 2  2
k2
sin( )−2(1 + ν) − cos( )cos( ),
θ θ 3θ
σ1 =
√r 2  2 2 
sin(θ/2)  
−2(1 − ν) − cos(2)cos( 2 )k 2 − 2 ik 3,
θ 3θ
σ2 =
√r
 
k
2 3θ
σ = sinθ cos( ),
z 2√ r 2
cos(θ/2)  
1 − sin(2) sin( 2 )k 2 + ik 3.
θ 3θ
τ + iτ =
zn tz √r
 
Note : compare with the results presented in Sih and Liebowitz (1968).
4.8 Flat crack of general shape
The general method is applied here to the analysis of an elastic space

weakened by a flat crack of arbitrary shape under the action of a uniform
normal pressure. A simple yet accurate relationship is established between the
crack face displacements and the applied pressure for an arbitrary flat crack.
Specific formulae are derived for a crack in the shape of a polygon, a rectangle,
a rhombus, a cross, a circular sector and a circular segment. All the formulae
are checked against the solutions known in the literature, and their accuracy is
confirmed. A similar approach can be used for the analysis of a crack under a
general polynomial loading. The material in this section follows the paper
(Fabrikant, 1987b).
Theory. Consider an elastic space weakened in the plane z =0 by a flat
crack occupying the domain S whose boundary is given in polar coordinates as
ρ = a (φ). (4.8.1)
Let a uniform pressure p be applied normally to the crack faces in opposite

directions. The governing integral equation in this case is given by (4.1.9).
The approach is based on the integral representation of the reciprocal of the
distance between two points established in (1.1.27). Substitution of (1.1.27) into

(4.1.9) gives, after interchanging the order of integration
x2
λ( , φ−φ )
ρ 2π a (φ0) ρρ0 0
dx
σ(ρ,φ) = −
1 ⌠ ⌠ dφ ⌠
3 ∆ w (ρ0,φ0)ρ0dρ0.
2π H ⌡ (ρ − x ) ⌡
2 2 1/2 0 ⌡ (ρ20 − x 2)1/2
0 0 x
(4.8.2)
It is noteworthy that the change of the order of integration which led to (4.8.2)
is valid inside the circle ρ≤min{ a (φ)} only, and this is one of the reasons why
the accuracy generally deteriorates for domains with the aspect ratio very far
away from unity. Nevertheless, one can obtain from (4.8.2) the exact solution
for an ellipse and sufficiently accurate formulae for various crack shapes as will
be demonstrated further on. Let the normal displacements of the crack face be
δ  2
a (φ) − ρ2 ,
1/2
w = (4.8.3)
a (φ)  
where δ is a constant to be defined. Now substituting (4.8.3) in (4.8.2), we
can verify how close to a constant will be the traction σ producing the
displacements (4.8.3). Integration with respect to ρ gives
0
∞ ρ 2 πa 2 ( φ ) − x2
Σ
δ xdx
σ(ρ,φ) = − ∆ ⌠ (x )|n| 2 ⌠ 0
8π2 H ⌡ ρ (ρ − x ) ⌡ a 2(φ0)
2 1/2
n=-∞ 0
0
× F 2− , ; 2; 1− 2
|n| 1 x 2  i(φ-φ0)
e dφ . (4.8.4)
 2 2 a (φ0) 0
Here F stands for the Gauss hypergeometric function. Further evaluation of the
normal traction can be done separately for each value of n . The zero th term
has the form
δB
σ0 = , (4.8.5)
8π H
where the notation
2π
dφ
B = ⌠ (4.8.6)
⌡ a (φ)
0
was introduced. It is clear that the value of the integral in (4.8.6) will depend
not only on the domain contour but also on the location of the coordinate
origin. The following argument might be useful for establishing certain rules in
this regard. According to (4.8.3), the coordinate origin location corresponds to
the point where the crack face displacement attains its maximum. We shall call
this point the crack centre. In the case of a crack domain with one axis of
symmetry, we may conclude from physical considerations that this point should be
located at the axis. When this domain possesses two axes of symmetry the
location of the crack center is at their intersection, i.e. at the center of gravity
of the domain. It is noteworthy that the integral (4.8.6) attains its minimum in
this case. One can extend this rule to a general crack, namely, the crack center
should be identified with the point inside S where the integral (4.8.6) reaches its
minimum. Direct computations for various domains indicate that this minimum
is, in general, sufficiently flat, so that in many cases one may locate the crack
center at the center of gravity, without any significant loss in accuracy. We
shall discuss this in more detail further on when considering the domain S in
the shape of a circular segment and sector.
It is important to note that the second harmonic is equal to zero for an
arbitrary contour, and that all the odd harmonics will be zero if the expression
for a (φ) does not contain odd harmonics. Here is the expression for the fourth
harmonic
2πcos4(φ − φ0)dφ0
4δ
σ4 = − ρ⌠ (4.8.7)
5π2 H ⌡ a 2(φ )
0 0
The investigation of the fourth and further harmonics shows that their amplitude
decreases for general domains, and they vanish in the case of an ellipse. If we
assume that p ≈σ then the remaining harmonics may be called the solution error.
0
This implies the establishment of the following relationship between the applied
traction p and the maximum displacement of the crack face
δB
p = (4.8.8)
8π H
One can verify that in the case of an ellipse, the solution given by (4.8.3) and
(4.8.8) is exact . We expect it to be reasonably accurate for a crack of general
shape. This assumption will be justified in the next Section where several
particular crack configurations are considered. We also expect (4.8.3) to be
sufficiently accurate in the neighbourhood of the crack center, though the relative
error might be quite significant close to the boundary.
The crack energy can be defined as
W = ⌠ ⌠ σ w dS. (4.8.9)
⌡⌡
S
Feeding (4.8.3) and (4.8.8) in (4.8.9) yields
16π Hp 2 A
W = , (4.8.10)
3B
where A is the crack area. Introduce the average displacement δ as

av
1⌠ ⌠
δav = w dS.
A⌡ ⌡
S
Substitution of (4.8.3) in the last expression gives
2
δ = δ.
av 3
Define the dimensionless parameter τ in the form
δ
av W
τ = = . (4.8.11)
2π Hp √ A 2π Hp 2 A 3/2
The physical meaning of τ can be defined either as a ratio of the average

displacement to a certain fixed displacement, or as a ratio of the crack energy to
a certain fixed one. Both forms (4.8.11) lead to the following expression for τ:
8
τ = . (4.8.12)
3√ A B
One can deduce that the value of τ does not depend on the size of the domain
S , and is determined by its shape only. It attains its maximum in the case of
a circle, so that 0≤τ≤4/(3π3/2)=0.2394. Tabulation of the coefficient τ for various
crack shapes might prove very useful since its knowledge allows us to find the
maximum (or average) crack face displacement and the crack energy by using
(4.8.11). It might seem more logical to define τ as the ratio of the given crack
energy to the energy of a circular crack having the same area. In this case the
value of τ would vary between zero and unity. The main reason for the
definition (4.8.12) was the desire to preserve the bridge between the crack
problems in mechanics and the mathematically equivalent problems in

electrostatics. The value of τ defined by (4.8.11) corresponds exactly to the
coefficient of electrical polarizability in the theory of wave propagation through
small apertures (Bethe 1944). We did not find in the mechanics literature any
report containing numerical data for nonelliptic cracks which could be compared
with the theory of this paper. The situation is slightly better in electrical
sciences. Cohn (1952) has measured the coefficient of electrical polarizability of
several aperture configurations experimentally. Numerical solution to the same
problem was given by De Meulenaere and Van Bladel (1977), and by Okon and
Harrington (1981). These numerical and experimental data will be used to
estimate the accuracy of the proposed theory. An empirical formula for the
coefficient of electrical polarizability was proposed by Fikhmanas and Fridberg
(1973). This formula in our notation reads
8√ A
τ = , (4.8.13)
3π L
where L stands for the perimeter of the domain S . Formula (4.8.13) is also
exact for an ellipse. It is of interest to compare its performance with our
(4.8.12). Several crack shapes are considered for this purpose. A high degree
of accuracy of formula (4.8.12) is confirmed by comparison with available
numerical solutions.
Example 1: Polygon. Consider a flat crack in the shape of a polygon
with n sides, with the only limitation that the function a (φ) describing its
boundary be continuous and single-valued. The origin of the coordinate system
is located at the crack center as it was defined earlier. Let us number the
polygon sides in a counter-clockwise direction from 1 to n , a being the length
k
of the k th side. The apex, at which the sides a and a intersect, is
k k+1
numbered k +1. It is clear that the value of the index n +1 is to be understood
as 1. Denote the distance from the crack center to the k th apex as b . Let
k
Ak be the area of the triangle formed by a , b and b , the total area A of
k k k+1
the polygon being equal to the sum of A . Then formulae (4.8.6) and (4.8.12)
k
yield the following expression for the coefficient τ:
8  1 1/2-1
n
a 2k a 2k
Σ 1 1/2
τ =  − 2 +  −  . (4.8.14)
3√ A
 k=1
4 A 2k bk   4 A 2k b 2k+1 

In the case of a regular polygon formula (4.8.14) simplifies to
4√cot(π/ n )
τ = . (4.8.15)
3 n 3/2sin(π/ n )
Formula (4.8.13) gives for a regular polygon
4  cot(π/ n ) 1/2
τ = . (4.8.16)
3π  n 
It is of interest to compare the numerical results due to (4.8.15) and (4.8.16).
Here the relevant computations are presented
n= 3 4 5 6 7 8 ∞
formula (4.8.15) τ= 0.2251 0.2357 0.2380 0.2388 0.2391 0.2392 0.2394
formula (4.8.16) τ= 0.1862 0.2122 0.2227 0.2280 0.2312 0.2331 0.2394
discrepancy (%) 17.3 10.0 6.5 4.5 3.3 2.6 0.0
While both formulae in the limiting case n →∞ give the same exact result for a
circle, their discrepancy for small n is quite significant, so it is important to
establish which one is more accurate. We have not found any data for an
equilateral triangle. If one takes the experimental result by Cohn for a square
τ=0.2274 as exact, then our formula (4.8.15) is in error by 3.6% while formula
(4.8.16) due to Fikhmanas and Fridberg is in error by 6.7%. The numerical
result due to Okon and Harrington for a square is 0.2258 which also favours our
formula. In the case of a regular hexagon, the result by Okon and Harrington
is 0.2375, so that our result differs by 0.5% only, while the error of (4.8.16) is
4%. It is noteworthy that the value of τ does not change significantly in the
whole range 3≤ n <∞.
We can also compare the normal displacements along a central line of a
hexagon perpendicular to its side, given by (4.8.3) with numerical data due to
Okon and Harrington (1981). Here are the results ( w ∗ stands for w /2π Hp √ A )
ρ/ a = 0. 0.1667 0.3333 0.5000 0.6667 0.8333

Okon et al w ∗= 0.351 0.346 0.331 0.305 0.263 0.210
formula (4.8.3) w ∗= 0.357 0.352 0.3366 0.3092 0.266 0.1973
discrepancy (%) -1.7 -1.7 -1.4 -1.4 -1.2 6.0
As we expected, the agreement is good, except for the points very close to the
boundary.
Example 2: Rectangle. Consider a rectangular crack, a and b being its
semiaxes along the axes Ox and Oy respectively. Introduce the aspect ratio
ε= b / a ≤1. Formula (4.8.15) in this case reduces to
√ε
τ = . (4.8.17)
3(1 + ε2)1/2
Formula (4.8.13) in this case gives
4√ε
τ = . (4.8.18)
3π(1 + ε)
We present below the results of computations due to (4.8.17) and (4.8.18)

compared with the experimental results of Cohn If one assumes the results of
ε= 0.1000 0.1500 0.2000 0.3000 0.5000 0.7500 1.0000

experiment τ= 0.1202 0.1411 0.1565 0.1789 0.2093 0.2251 0.2274
formula (4.8.17) τ= 0.1049 0.1277 0.1462 0.1749 0.2108 0.2309 0.2357
discrepancy (%) 12.7 9.5 6.6 2.3 -0.7 -2.6 -3.7
formula (4.8.18) τ= 0.1220 0.1429 0.1582 0.1788 0.2001 0.2100 0.2122
discrepancy (%) -1.5 -1.3 -1.1 0.1 4.4 6.7 6.7
Cohn to be exact then our formula performs better for ε≥0.5 while the formula
by Fikhmanas and Fridberg is more accurate for ε<0.5. If instead we take the
numerical results received in a personal communication from De Smedt as correct
then the conclusion might be different. For example, his value of τ for ε=0.1
is 0.1142; now our result is in error by 8% while the result by Fikhmanas and
Fridberg is in error by -7%. At this moment nobody seems to know which
estimate is correct. We can also compare the dimensionless displacements w ∗
due to (4.8.3) with the numerical results received in a personal communication
from De Smedt for a rectangle with aspect ratio ε=0.5 (as before, w ∗ stands for
w /2π Hp √ A ). Here are the data computed along the axis Ox for y/b =0.025. The
x/a = 0.0250 0.2250 0.4250 0.6250 0.8250 0.9750

De Smedt w *= 0.3161 0.3118 0.2989 0.2713 0.2107 0.0852
formula (4.8..3) w *= 0.3158 0.3081 0.2862 0.2469 0.1787 0.0703
Discrepancy (%) 0.1 1.2 4.2 9.0 15.2 17.5
agreement is not bad except for the zone x/a >0.625. Here are the data
computed along the axis Oy for x/a =0.025. We observe here a good agreement
y/b = 0.0250 0.1250 0.2250 0.3250 0.4250 0.4750

De Smedt w *= 0.3161 0.3067 0.2836 0.2424 0.1690 0.0976
formula (4.8.3) w *= 0.3158 0.3062 0.2824 0.2403 0.1666 0.0987
Discrepancy (%) 0.1 0.2 0.4 0.8 1.4 -1.2
Example 3: Rhombus. Consider the case where the domain S is a

rhombus, a and b being its semiaxes along the axes Ox and Oy respectively.
Introduce the aspect ratio ε= b / a ≤1. Formula (4.8.14) in this case reduces to
√2ε
τ = . (4.8.19)
3(1 + ε)
The result due to Fikhmanas and Fridberg is
2√2ε
τ = . (4.8.20)
3π(1 + ε2)1/2
We did not find in the mechanics literature any result related to a crack with a
rhombus planform. The coefficient of electrical polarizability for a diamond with
the aspect ratio ε=0.5 was found numerically by Okon and Harrington as
τ=0.2082. Our result is 0.2222 (discrepancy 6.7%) while formula (4.8.20) gives
0.1898 (discrepancy 8.9%). We have received two sets of data in personal
communications from De Smedt and Lee. Here are the data received as
compared to formulae (4.8.19) and (4.8.20) The data received from Lee is given
ε= 0.100 0.200 0.333 0.500 0.800 1.000

De Smedt τ= 0.111 0.151 0.182 0.204 0.219 0.221
formula (4.8.19) τ= 0.136 0.176 0.204 0.222 0.234 0.236
Discrepancy (%) -21.9 -16.4 -12.0 -9.0 -6.8 -6.6
formula (4.8.20) τ= 0.094 0.132 0.164 0.190 0.210 0.212
Discrepancy (%) 15.1 12.8 9.8 6.9 4.4 4.1
as a function of the angle α=tan-1ε We have presented both sets of data in
α(deg.)= 10. 15. 20. 25. 30. 40. 45.

Lee τ= 0.147 0.174 0.193 0.207 0.216 0.226 0.228
formula (4.8.19) τ= 0.168 0.192 0.209 0.220 0.227 0.235 0.236
Discrepancy (%) -14.2 -10.6 -8.1 -6.3 -5.2 -3.8 -3.6
formula (4.8.20) τ= 0.124 0.150 0.170 0.186 0.197 0.211 0.212
Discrepancy (%) 15.8 13.7 11.8 10.1 8.5 6.9 6.7
order to underline the fact that there is no really reliable data as yet. The first
set of data suggests that the formula by Fikhmanas and Fridberg is the more
accurate, while the second set favours ours. It is noteworthy that formula
(4.8.19) seems to give the upper bound, and formula (4.8.20) provides the lower
bound, their average being very close to the numerical data.
We can also compare the normal displacements due to our (4.8.3) with a
ρ/ a = 0. 0.3333 0.6667
Okon et al. w ∗= 0.335 0.304 0.257
formula (4.8.3) w ∗= 0.3333 0.3142 0.2484
discrepancy (%) 0.5 -3.4 3.3
mainly due to the assumption of a square root singularity in (4.8.3) which does
not hold for domains with sharp angles.
Example 4: Circular segment. Let the radius r and the angle 2α be the
segment parameters. Direct numerical computations show that the crack center
can be identified with the center of gravity, with an error comparable with the
accuracy of the theory presented. The location of the center of gravity is
defined by x = kr, where
c
2 sin3α
k = .
1
3(α − sin 2α)
2
The equation of the segment boundary with respect to its center of gravity takes
the form
a (φ) = r [− k cosφ + (1 − k 2sin2φ)1/2], for 0≤φ≤π−γ or π+γ≤φ<2π,

and
k − cosα
a (φ) = r , for π−γ≤φ≤π+γ, (4.8.21)
cos(π − φ)
where γ = tan-1(sinα/( k − cosα)). Feeding of (4.8.21) in (4.8.6) and (4.8.12)

gives
τ =
4  k sinγ + E(π−γ, k ) + sinγ −1. (4.8.22)
3(α − sin2α)1/2
1 1 − k 2 k − cosα 
2
where E(⋅,⋅) stands for the incomplete elliptic integral of the second kind. The
formula due to Fikhmanas and Fridberg gives
1
4(α − sin2α)1/2
2
τ = . (4.8.23)
3π(α + sinα)
The coefficient of electrical polarizability for a semi-circle was computed by

Okon and Harrington as τ=0.2161. Our result due to (4.8.22) is τ=0.2163 which
is practically identical to the previously mentioned one. The result due to
(4.8.23) is τ=0.2069 (discrepancy 4.3%). An additional confirmation of

correctness of the new method can be obtained by observing the plot of the
electrical polarizability density distribution for a semi-circle presented by Okon
and Harrington (1981). Its maximum is located at a distance ≈0.47 r from the
circle’s center. Our definition of the crack center requiring the minimization of
the integral (4.8.6) gives its coordinate at 0.48 r which is very close. The center
of gravity of the semi-circle is located at 0.42 r .
Example 5: Circular sector. Let r and 2α be its radius and the polar
angle. The crack center is assumed to be located on the axis of symmetry at a
distance kr from the circle’s center. Numerical computations show that the crack
center may be located at the center of gravity for 0.1π<α<0.6π. In this case
the value of k is defined by k = 2sinα/(3α). In the range α<0.1π or α>0.6π,
the value of k should be found from the minimum condition for the integral
(4.8.6). Repetition of the procedure described in the previous paragraph leads to
the following result
4  k sinγ + E(γ, k ) cosα + cos(α − γ) −1

τ = + . (4.8.24)
3√α  1 − k2 k sinγ 
Here, γ = tan-1(sinα/(cosα − k )). The formula due to Fikhmanas and Fridberg

reads
4√α
τ = . (4.8.25)
3π(1 + α)
Note that neither (4.8.24) nor (4.8.25) reduce to the exact value for a circle
when α=π. This is due to the fact that we do not really have the case of a
penny-shaped crack when α approaches π: we have a circular crack which has
its faces bonded along the radius φ=π. This case has not been considered by
other authors so we cannot say which formula is more accurate. Okon and
Harrington obtained in the case of a quadrant τ=0.2269, formula (4.8.24) gives
τ=0.2308 (discrepancy 1.7%), and formula (4.8.25) gives τ=0.2107 (discrepancy
7%). It is noteworthy that the value of τ for a quadrant is greater than that
for a semi-circle. The general impression is that our theory in the particular
cases of a circular sector and segment provides the upper bound for τ while the
formula due Fikhmanas and Fridberg gives the lower bound.
Example 6: Cross. Consider a crack configuration obtained by an
orthogonal intersection of two equal rectangles with sides 2 a and 2 b . Introduce
the aspect ratio as ε= b / a ≤1. The area can be expressed as
A = 4 a 2ε(2 − ε) ,
The following expression can be obtained for τ:

√2ε
τ = (4.8.26)
6√2 − ε{[2(1 + ε2)]1/2 − 1}
The formula due to Fikhmanas and Fridberg is
2√ε(2 − ε)
τ = . (4.8.27)
3π
Here, we present the results given by formulae (4.8.26) and (4.8.27) compared to
the experimental results of Cohn and the numerical results by De Meulenaere and
Van Bladel (1977), and those received in personal communication from De Smedt
ε= 0.1000 0.2000 0.3000 0.4000 0.6000 0.8000 1.0000

experimental τ= 0.0942 0.1333 0.1609 – – – 0.2274
De Meulenaere τ= – – – 0.19 0.22 0.23 0.238
De Smedt τ= 0.0835 0.1183 – 0.1767 0.2084 0.2193 0.2212
formula (4.8.26) τ= 0.1284 0.1777 0.2078 0.2252 0.2376 0.2372 0.2357
formula (4.8.27) τ= 0.0925 0.1273 0.1515 0.1698 0.1944 0.2079 0.2122
We did not compute the discrepancy since the data disagreement is too large
thus making all the data not very reliable. The general impression is that our
(4.8.26) gives the upper bound for τ while the formula due to Fikhmanas and
Fridberg provides the lower bound. This conclusion might be wrong if the
numerical results received in the personal communication from De Smedt are
correct. For example, his result for ε=0.1 is τ=0.08347 which differs from the
experimental result by 11%. All this proves one point: the existing numerical
methods are too crude and there is a need to develop some new and more
reliable numerical methods.
It should be noted that the function defined by (4.8.26) is not monotonic: a
relatively flat maximum is observed for ε≈0.7. The remaining data are
monotonic. We have no rigorous proof to claim that the quantitative behavior of
(4.8.26) is correct while the other data behavior is not, but we can indicate that
the value of τ for a quadrant is also greater than that for a semi-circle, and this
is mainly due to the fact that the shape of a quadrant is more close to the
shape of a circle than that of a semi-circle. A similar statement can be made
about a cross with the aspect ratio ε≈0.7 as compared to a square.
Discussion. The majority of the examples considered indicate that the exact
result is sandwiched between the results given by our (4.8.12) and by the
formula due to Fikhmanas and Fridberg (4.8.13). In this sense the formulae act
as upper and lower bounds respectively, which leads to a conjecture: for an
arbitrary contour one of the inequalities holds, namely, either τ12≤τexact≤τ13, or
τ12≥τexact≥τ13. We can indicate one way to disprove the conjecture. A look at
the table related to a rectangle in the previous Section indicates that our formula
(4.8.17) should give the lower bound for small aspect ratio ε, and it should give
the upper bound for ε close to unity, and vice-versa for the formula (4.8.18).
This means that there should be a value of ε for which both formulae (4.8.17)
and (4.8.18) are exact , and give the same result. By equating (4.8.17) and
(4.8.18), one gets the value of the aspect ratio
π2 ± 4( 2π2 − 16 )1/2
ε1,2 = ,
16 − π2
which yields ε =0.3482 and ε =2.8716, with an obvious property ε ε =1. The
1 2 1 2
corresponding value of τ is 0.18576. From the table of the previous Section,
one can see that for ε=0.3 the exact value of τ should be between 0.1749 and
0.1788. If we could be sure that the experimental value 0.1789 is exact then
this would disprove the conjecture, but at the moment none of the existing
experimental or numerical techniques can offer such an accuracy. This can be
achieved on the basis of the method for accurate evaluation of singular
two-dimensional integrals presented in (Fabrikant, 1986e). A significant effort is
required to prove or to disprove the conjecture, and is left to the interested
reader.
The accuracy of formula (4.8.12) can be improved by taking into
consideration the fourth harmonic (4.8.7) in combination with the variational
approach (Noble, 1960). The following functional is stationary at the exact
solution of (4.1.9)
I ( w ) = 2⌠ ⌠ σ( M ) w ( M )d S +
1 ⌠⌠
w ( M ) ∆ ⌠ ⌠ w( N) d S d S . (4.8.28)
⌡⌡ M 4π2 H⌡ ⌡  ⌡ ⌡ R ( M,N) N M
S S S
Taking
−
1
∆ ⌠ ⌠ w( N) dS ≈ σ + σ ,
4π2 H ⌡ ⌡ R ( M,N) N 0 4
S
where σ and σ are defined by (4.8.5) and (4.8.7) respectively, and substituting
0 4
them in (4.8.28), we obtain a functional which can be considered as a function
of δ. From the extremum condition
∂I
= 0
∂δ
one finally gets

8
τ = , (4.8.29)
3 B √ A (1 − c )
where
3( F E + F E )
c c s s
c = ,
5 AB
and the following geometrical characteristics were introduced
2π 2π
cos4φ dφ sin4φ dφ
F = ⌠ , F = ⌠ ,
c
⌡ a 2(φ) s
⌡ a 2(φ)
0 0
2π 2π
E = ⌠ a 3(φ)cos4φ dφ, E = ⌠ a 3(φ)sin4φ dφ.

c
⌡ s
⌡
0 0
The results of computations due to (4.8.29) for a rectangle are presented below
against the experimental results of Cohn Comparison of this table with a
ε= 0.1000 0.1500 0.2000 0.3000 0.5000 0.7500 1.0000

Cohn τ= 0.1202 0.1411 0.1565 0.1789 0.2093 0.2251 0.2274
formula (4.8.29) τ= 0.1054 0.1290 0.1484 0.1785 0.2125 0.2257 0.2278
discrepancy (%) 12.3 8.6 5.2 0.2 -1.5 -0.3 -0.2
corresponding one presented earlier indicates that the variational approach does
improve the accuracy, though the improvement is still not sufficient for small ε.
There is no proof that the variational approach will always improve the accuracy.
On the contrary, one can find quite a few examples when the accuracy
deteriorates. This can usually be observed for domains with a very small aspect
ratio ε. It is up to the user to decide whether the more cumbersome
computations are worth somewhat better accuracy.
In this section we have considered in detail only the case of a uniform
crack pressure. Some considerations can be presented for a general case. It is
known (see, for example, 4.1.13) that in the case of a penny-shaped crack the
following relationship can be established between the displacements w and the
internal pressure σ
2π a
[( a 2 − ρ2)( a 2 − ρ20)]1/2 σ(ρ0,φ0)
2 ⌠ ⌠ -1 
w (ρ,φ) = H tan ρ dρ dφ ,
π ⌡ ⌡  aR  R 0 0 0
0 0
where R =ρ2 + ρ20 − 2ρρ cos(φ−φ ). The following generalization of the last
0 0
formula seems to be natural
2π a (φ0) [( a 2(φ) − ρ2)( a 2(φ ) − ρ20)]1/2 σ(ρ0,φ0)

w (ρ,φ) = H ⌠ ⌠ tan-1 
2 0
ρ0dρ0dφ0.
π ⌡ ⌡  a (φ) R  R
0 0
(4.8.30)
Though a complete investigation of (4.8.30) is beyond the scope of this book,
there is reason to believe that it will be sufficiently accurate for a general
loading of cracks whose aspect ratio is not far away from unity. Here is an
example. Let us compute the displacement w at the centre of an elliptical
0
crack ( a and b are the semiaxes of the ellipse, a > b ) under a uniform pressure
p . The result due to (4.8.30) is w 0=(8/π) HpbK(k) , where K(k) is the complete
elliptic integral of the first kind and k =(1− b 2/ a 2)1/2. The exact result is
w 0=2π Hpb/E(k) . Both results are close to each other for small k , coinciding in
the case of a circle ( k =0). Direct computation shows that the error of the
approximate expression does not exceed 7% for an ellipse with aspect ratio
b/a ≥0.5. In the case of a square the dimensionless displacement w * at its centre
can be computed from (4.8.30) as (4/π2)ln(1+√2 )=0.3572, while a similar result
due to the experimental value of Cohn is (3/2)τ=0.3411, and the discrepancy does
not exceed 5%. Of course, these examples do not prove anything conclusively,
but they make it quite clear that expression (4.8.30) is worth to investigate
further. A similar statement can be made about the following expression
2πa (φ0) [ a 2(φ ) − ρ20]1/2σ(ρ ,φ )ρ dρ dφ

σ(ρ,φ) = −
2 ⌠ ⌠ 0 0 0 0 0 0
,
π[ρ − a 2(φ)]1/2⌡ ⌡
2
ρ2 + ρ20 − 2ρρ0cos(φ−φ0)
0 0
which gives the normal stress distribution outside a circular crack directly through
its values inside the crack. The future research will show whether the last
expression is useful for a general crack.
Investigation of the stress intensity factors was beyond the main scope of
this section, but we can show that some simple formulae may be derived to give
results which are close to (Weaver, 1977). Make use of the asymptotic
relationship between the limiting values of the crack opening displacements and
the stress intensity factor, namely,
w = 2 H (2π r )1/2 k 1.
Consider a rectangular crack of dimensions 2 a and 2 b ( a>b ), subjected to a

4.9 General crack under uniform shear 273
uniform pressure p . By substituting (4.8.3) in the last expression, the following

formulae are obtained for the dimensionless stress intensity factor κ= k 1/[ p (π b )1/2]:
along the shorter side κ=3τ; along the longer side κ=3τ( a/b )1/2. In the case of
a square our formulae give κ=3×0.2357=0.7071. The result due to Weaver is
about 0.73, with the discrepancy of about 3%. In the case of b/a =0.3, our
result at the middle of the longer side is κ=0.98 which is very close to the
result of Weaver. Expression (4.8.17) indicates that the limiting value of the
stress intensity factor as ( b/a )→0 is κ→1, as it should be in the case of an
infinite strip slit. The only discrepancy with the results of Weaver (1977) is the
value of the dimensionless stress intensity factor along the shorter side: according
to our formula it should decrease with b/a , tending to zero as ( b/a )1/2; in the
paper by Weaver (1977) its value increases. The reader is referred also to the
paper (Fabrikant, 1987f) where some inconsistencies of (Kassir and Sih, 1975) in
defining the stress intensity factor for elliptical cracks are pointed out.
The mathematically equivalent problem of sound penetration through an
aperture of general shape in a soft screen is solved in (Fabrikant, 1988d). The
same apparatus is used in the investigation of electrical polarizability of small
apertures of general shape (Fabrikant, 1987c).
Exercise 4.8
1. Derive (4.8.5).
3. Try to prove or disprove the conjecture, expressed in the section ’Discussion’

above.
4. Find the domain of validity of formula (4.8.30) for an elliptical crack.

(Find the ratio of ellipse semiaxes, for which the error does not exceed 5%.)
5. Solve the problem of a general flat crack subjected to normal loading, with
its magnitude proportional to the x -coordinate. (Bending of an elastic space, with
a flat crack of general shape).
4.9 General crack under uniform shear
Let the crack boundary be described in polar coordinates by the equation

ρ= a (φ), where a (φ) is a single-valued continuous function. We can always
choose the coordinate axis orientation so that the first harmonic will vanish from
the Fourier expansion of a (φ). An approximate analytical solution of (4.4.14) for
a general crack can be obtained by the method used in previous section. The
method uses the following representation
x2
λ( , φ−φ )
ρ 2π a (φ0) ρρ0 0
⌠ ⌠ u( N) dS = ⌠ dx
⌠ dφ ⌠ u (ρ0,φ0)ρ0dρ0.
⌡ ⌡ R ( N,N0) N ⌡ ( ρ2
− x 2 1/2
) ⌡ 0 ⌡ (ρ20 − x 2)1/2
S 0 0 x
(4.9.1)
Consider the case of a uniform shear loading. Let the tangential displacements
have the form
u (ρ,φ) = u 0[ a 2(φ) − ρ2]1/2/ a (φ) , (4.9.2)
where u is an as yet unknown complex constant. Substitution of (4.9.2) in

0
(4.9.1) yields, after integration and retaining the first two harmonics only,
2π
 u0
⌠ ⌠ u( N) dS ≈ ⌠ 2 a ( φ )
⌡ ⌡ R ( N,N0) N 16π( G 21 − G 22) ⌡

0
S 0
ρ2  
− 1 + 3cos2(φ−φ )dφ .
a (φ0) 0  0

(4.9.3)
By substituting (4.9.3) in (4.4.14) and performing necessary differentiation, we
obtain the relationship between the shear loading and the amplitude of the
tangential displacements, namely,
τ =
1  u G B + 3u G B  , (4.9.4)
4π( G 21 − G 22)  0 1 1 0 2 2
where
2π 2π
dφ e2 i φdφ
B1 = ⌠ , B2 = ⌠ . (4.9.5)
⌡ a (φ) ⌡ a (φ)
0 0
Equation (4.9.4) can be solved, to give
4π( G 21 − G 22)
u0 = [ G B τ − 3 G B τ]. (4.9.6)
G 21 B 21 − 9 G 22 B 22 1 1 2 2
The integrals in (4.9.5) can be computed easily for various crack shapes. For
example, a rectangular crack with sides 2 a and 2 a is characterized by the
1 2
values
4( a 21 + a 22)1/2 4( a 22 − a 21)
B = , B = . (4.9.7)
1 aa
1 2
2 3 a a ( a 21 + a 22)1/2
1 2
It is noteworthy that despite the fact that the integral representation (4.9.1) is
valid inside a circle ρ≤min{ a (φ)} only, and despite the approximate nature of
(4.9.3), the solution given by (4.9.4−4.9.5) is exact for an ellipse. We did not
find in the literature any reliable data related to a non-elliptical crack under
shear loading, therefore it is difficult to say how accurate the solution is for
various crack shapes.
Exercise 4.9
1. Derive (4.9.3).
3. Consider the case of semicircular crack.
4. Solve the problem for a cross-shaped crack
5. Solve the problem of a general flat crack subjected to torsion.
4.10 Close interaction of pressurized coplanar circular cracks
The general method is applied here for the stress analysis of an elastic
space weakened by several arbitrarily located coplanar circular cracks under the
action of an arbitrary normal pressure. The governing integral equations are
derived, which have definite advantages over other methods: the equations are
non-singular, the iteration procedure is rapidly convergent even for very close
interactions; there is no need to solve the integral equations if one is interested
only in obtaining the upper and the lower bounds for the quantities of interest.
In the case of the cracks which are far apart, these bounds are so close that
they provide, in fact, a sufficiently accurate solution to the problem. The
method allows us also to obtain a practically exact numerical solution to the
problem of very close interactions. Several illustrative examples are considered.
Theory. Consider an elastic space weakened in the plane z =0 by n

arbitrarily located circular cracks. The cracks do not intersect. Let the centre
of the k th crack be located at the point with Cartesian coordinates x and y , its
k k
radius be denoted by a . Let an arbitrary pressure σ be applied normally to
k k
the crack faces in opposite directions. The set of governing integral equations
can be written, due to (4.1.9), in the form
n w (M )
σ (M ) = −
i i
1
4π2 H
∆ Σ ⌠ ⌠ k k d S , for i = 1,2, ...
⌡ ⌡ R ( M i , M k) k
,n,
k=1 Sk
(4.10.1)
where ∆ is the two-dimensional Laplace operator, S is the k th crack domain,

k
R(M ,M ) stands for the distance between the points M and M , ( M ∈ S and
i k i k i i
M ∈ S ); w denotes the normal displacements of the crack face (an unknown
k k k
function), σ stands for the normal traction acting inside the i th crack (a known
i
function). We can single out, without loss of generality, the crack number 1,
and consider the set of cracks in the local polar system of coordinates with the
origin coinciding with the centre of the first crack. By using the integral
representation (4.1.11), the first equation from the set (4.10.1) can be rewritten as
ρ
1 1 d ⌠ xdx
σ (ρ,φ) = − 2 L ( ) 2
2 1/2 L( x )
1 π Hρ ρ dρ ⌡ (ρ − x )
2
0
1 ρ dρ
a n w (M )
Σ ∆⌠ ⌠
d 0 0 1 1 k k
× ⌠ L( ) w 1(ρ0,φ) − dS .
d x (ρ0 − x 2)1/2
2 ρ 4π2 H ⌡ ⌡ R(M ,M )
⌡
k
0 k=2 Sk 1 k
x
(4.10.2)
Since the integrals under the summation sign in (4.10.2) are non-singular, the
differentiation can be performed under the integral sign, with the result
ρ
1 1 d ⌠ xdx
σ (ρ,φ) = − 2 L( ) L( x 2 )
1 π Hρ ρ dρ ⌡ (ρ − x )
2 2 1/2
0
1 ρ dρ
a n w (M )
Σ ⌠⌠ k k
d 0 0 1 1
× ⌠ L( ) w 1(ρ0,φ) − dS .
d x (ρ0 − x )1/2
⌡
2 2 ρ0 4π2 H ⌡ ⌡ R 3( M , M )
1 k
k
k=2 Sk
x
(4.10.3)
Formula (4.1.12) allows us to express w 1 from (4.10.3) by constructing an
inverse operator, namely,
a
1 x ρ0dρ0 ρρ0
w 1(ρ,φ) = 4 H ⌠ d x ⌠ 2 1/2 L( 2 ) σ1(ρ0,φ)
⌡ ( x − ρ ) ⌡ ( x − ρ0)
2 2 1/2 2
x
ρ 0
( a 21 − ρ2)1/2 n w ( ρ , φ ) ρ dρ dφ
Σ ⌠⌠ k 0 0 0 0 0
+ .
π2 ⌡ ⌡ (ρ0 − a 1) [ρ + ρ0 − 2ρρ0cos(φ−φ0)]
2 2 1/2 2 2
k=2 Sk
(4.10.4)
3
Here we used the following integral representation for 1/ R
x2
λ( , φ−φ0) x 2d x
ρ ρ0
1 2 1 d⌠
3/2 = πρL(ρ) dρ 2 3/2,
(ρ + ρ0 − 2ρρ0cos(φ−φ0)) ⌡ (ρ − x ) (ρ0 − x )
2 2 2 2 1/2 2
0
for ρ >ρ. (4.10.5)
0
The representation (4.10.5) allows us to compute various integrals involving the
Abel and the L-operators. For example, the following integral is an immediate
consequence of (4.10.5)
x
⌠ 2 ρ dρ2 1/2 L(ρ)[ρ2 + ρ20 − 2ρρ cos(φ−φ )]-3/2
⌡ (x − ρ ) 0 0
0
x x2
= λ( , φ−φ0).
(ρ20 − x 2)3/2 ρ0
A similar procedure can be applied to the remaining n −1 cracks, and the

additional n −1 equations of the type (4.10.4) can be obtained. Note that each
such equation is valid in a local system of polar coordinates related to a certain
crack. The set of equations (4.10.4) can be solved numerically by iteration.
Here we show that one can obtain the upper bound, the lower bound and
a reasonably accurate central estimation for all the quantities of interest without
solving the integral equations (4.10.4). Since w does not change sign in S , we
k k
can apply the mean value theorem to the second integral in (4.10.4), with the
result
a
1 x ρ0dρ0 ρρ0
w 1(ρ,φ) = 4 H ⌠ d x ⌠   σ (ρ ,φ)
2 1/2 L
⌡ ( x − ρ ) ⌡ ( x − ρ0)
2 2 1/2 2
x  1 0
2
ρ 0
( a 21 − ρ2)1/2 n Vk
+
π2 Σ(r
k=2
2
1k − a 21)1/2 [ρ2 + r 21k − 2ρ r 1kcos(φ−φ1k)]
.
(4.10.6)
Here V denotes half of the volume of the opened k th crack
k
V k = ⌠ ⌠ w kd S k,
⌡⌡
Sk
and though the exact location of the point R with the polar coordinates
1k
( r ,φ ) is unknown, the fact of belonging to the domain S limits the possible
1k 1k k
variation of the quantities of interest and allows us to obtain the upper and the
lower bounds as well as a sufficiently accurate central estimation which will be
discussed further. The symmetry considerations can also be used to sharpen the
estimations. It will be shown further that these bounds can be so close in the
case of cracks remote from each other, that they provide, in fact, a sufficiently
accurate solution to the problem.
The crack volume 2 V can be estimated by integration of (4.10.4) over the
1
domain S 1. The result is
2π a
1
V 1 = 4 H⌠ ⌠ σ(ρ,φ)( a 21 − ρ2)1/2ρdρdφ
⌡ ⌡
0 0
n a a
Σ ⌠⌠  w (ρ,φ)ρdρdφ.
2 1 1
2 1/2 − sin
-1
+
π 
⌡ ⌡ (ρ − a 1)
2 ρ  k
k=2 Sk
(4.10.7)
We can use again the mean value theorem, with the result
2π a n a a
1
V 1 = 4 H⌠ ⌠ σ1(ρ,φ) ( a 21 − ρ2)1/2ρdρdφ +
2 
Σ V .
1 1
− sin-1
⌡ ⌡ π (ρ1k − a 1)
2 2 1/2 ρ1k
 k
0 k=2
0
(4.10.8)
Again, the point with the polar coordinate ρ1k belongs to S k thus limiting the
possible variation. Integration of the remaining n −1 equations over the area of
each crack provides finally a system of n linear algebraic equations which can be
solved with respect to the unknowns V k. Their feeding back into (4.10.6) gives
the complete solution to the problem. Although the exact values of the
coordinates r ik , φik and ρik are not known, we can always use their maximum
and minimum values in order to obtain the upper and the lower bounds for all
the quantities of interest.
Defining the stress intensity factor as
k k(φ) = lim {(ρ − a k)1/2σk(ρ,φ)} ,

ρ→ a
k
one can get an equivalent expression through the crack face displacements
w k(ρ,φ)
1
k k(φ) = lim 1/2 . (4.10.9)
4π H ( a − ρ)
ρ→ a k
k
2π a ( a 21 − ρ2)1/2σ (ρ,φ ) ρdρdφ

1
k (φ) =
1 ⌠ ⌠ 1 0 0
1 π √2 a 1 ⌡ ⌡
2
ρ2 + a 21 − 2ρ a 1cos(φ − φ0)
0 0
√a1 n w (ρ, φ ) ρdρdφ

Σ ⌠⌠ k 0 0
+ .
2√2π3 H ⌡ ⌡ (ρ2 − a 21)1/2[ρ2 + a 21 − 2ρa 1cos(φ−φ0)]
k=2 Sk
By using again the mean value theorem, the following expression for the stress
intensity factor can be obtained
2π a ( a 21 − ρ2)1/2σ (ρ,φ ) ρdρdφ

1
k (φ) =
1 ⌠ ⌠ 1 0 0
1 π √2 a ⌡ ⌡
2
ρ2 + a 21 − 2ρ a cos(φ−φ )
1 0 1 0
0
√a n Vk
Σ
1
+ .
2√2π3H ( r 21k − a 21)1/2 [ r 21k + a 21 − 2 r a cos(φ − φ )]
k=2 1k 1 1k
(4.10.10)
Similar expressions can be derived for the remaining n −1 cracks. The first term
in (4.10.10) presents the stress intensity factor of a solitary crack opened up by
an arbitrary normal pressure σ , the remaining terms display the influence of
1
interacting cracks. It is clear from (4.10.10) that the stress intensity factor of
interacting coplanar cracks is always greater than the stress intensity factor of a
solitary crack under similar pressure. This proves the results of Mastrojannis and
Mura (1983) to be incorrect, since they report a decrease in the stress intensity
factor along part of the boundary.
Example 1: Two cracks. Consider the case of two coplanar circular
cracks with the radii a and a under the action of a uniform normal pressure
1 2
σ =p and σ = p respectively. Let l be the distance between their centres.
1 1 2 2
Equations (4.10.8) in this case will take the form
a a
π a 1 Hp + V  2 ,
8 3 2 1 1
V1 = − sin-1
3 1 π 2(ρ12 − a 1)
2 1/2 ρ12
a2 a2
8 3 2  .
V2 = π a 2 Hp 2 + V 1 2 − sin-1
3 π (ρ21 − a 22)1/2 ρ 
21
The solution is
a 31 p + c a 32 p c a 31 p + a 32 p
8 1 12 2 8 21 1 2
V = πH , V = πH , (4.10.11)
1 3 1 − c 12 c 21 2 3 1 − c 12 c 21
where
a a a a
2 1 1
, 2 2 2
.
c = − sin-1
c 21 = − sin-1
12 π(ρ12 − a 1)
2 2 1/2 ρ  π(ρ21 − a 2)
2 2 1/2 ρ 
12 21
(4.10.12)
Here ρ12 varies between l − a 2 and l + a 2, and l − a 1≤ρ21≤ l + a 1. The upper bound
for V k corresponds to ρ12= l − a 2 and ρ21= l − a 1; the lower bound is achieved at
ρ12= l + a 2 and ρ21= l + a 1. The estimations can be sharpened by using the
reciprocal theorem which implies the identity c 12 a 32= c 21 a 31. This identity narrows
the range of admissible variation for ρ (in the case of unequal cracks only) thus
making the estimations sharper. In this vein the case of two equal cracks
should be considered as the least accurate. We shall also consider the central
estimation which corresponds to ρ12=ρ21= l . It will be shown that the central
estimation gives a sufficiently accurate solution even for relatively close crack
interactions.
Formulae (4.10.11−4.9.12) simplify in the case of equal cracks as a 1= a 2= a ,
and if p 1= p 2= p , then
V0
2 
− sin-1 .
a a
V 1 = V 2 =V = , c = (4.10.13)
1 − c π (ρ − a )
2 2 1/2 ρ
Here V 0=(8/3)π Ha 3 p stands for a half of the volume of a solitary crack, and ρ
varies between l − a and l + a . Note that in the case of a uniform pressure, the
crack energy W is proportional to its volume, namely, W = pV . It is clear from
(4.10.13) that the crack interaction increases their energy. Substitution of
(4.10.13) in (4.10.10) and use of the mean value theorem yield the following
expression for the stress intensity factor
 2ε3 
K (φ) = K 01 + ,
{3π(1 − ε2)1/2 − 6[ε − (1 − ε2)1/2sin-1ε]}[1+ε2−2εcosφ]
 
(4.10.14)
where K 0= p √2 a /π corresponds to the stress intensity factor for an isolated crack
under the action of a uniform normal pressure p . Here the upper bound for the
stress intensity factor is given by ε= a /( l − a ), the lower bound corresponds to
ε= a /( l + a ), and the central estimation is defined by ε= a / l . Now we need an
accurate numerical solution in order to estimate the accuracy of the approximate
formulae derived. Assume the crack face displacements in the form
w (ρ,φ) = 4 Hp ( a 2 − ρ2)1/2 f (ρ,φ), (4.10.15)
where f is as yet unknown function. It may be called the interaction function

since it is equal to the ratio of the crack opening displacements to those of a

solitary crack. The values of f ( a ,φ) are equal to the ratio of the stress intensity
factor of interacting cracks to the stress intensity factor of a solitary crack. We
shall call f ( a , φ) the interaction factor . Substitution of (4.10.15) in (4.10.4)
gives a convenient expression for the procedure of iteration

w (ρ, φ) = 4 Hp ( a 2 − ρ2)1/2 1

2π a ( a 2 − r 20)1/2 f(r0, ψ 0) r0 d r0 dψ0
1⌠ ⌠ 
+ .
π2⌡ ⌡ ( l 2 + r 20 + 2 lr 0cosψ0 − a 2)1/2[ r 2 + r 20 + 2 rr0cos(ψ+ψ0)]
0 0 
(4.10.16)
Here we have introduced the new variables r =(ρ + l −2 l ρcosφ) , ψ=sin [(ρ/ r )sinφ].
2 2 1/2 -1
The integral in (4.10.16) has a logarithmic singularity for r = l − a , ψ=0, as l →2 a ,

therefore the procedure of iteration might not be convergent for l very close to
2 a . The limiting value of l can be found by analyzing the integral operator
2π a ( a 2 − r 20)1/2 f(r0, ψ 0) r0 d r0 dψ0

1⌠ ⌠
Z(f) = .
π2⌡ ⌡ ( l 2 + r 20 + 2 lr 0cosψ0 − a 2)1/2[( l − a )2 + r 20 +2( l − a ) r 0cosψ0]
0 0
(4.10.17)
According to the Banach’s theorem, it is sufficient to prove that the integral
operator (4.10.17) is a contraction operator. We define the distance in the class
of continuous functions by
δ( f,f 1) = max| f (ρ,φ) − f 1(ρ,φ)|.
We assess the value of
| Z ( f ) − Z ( f 1)|
2π a ( a 2 − r 2)1/2 | f(r, ψ ) − f1(r, ψ ) | r d r dψ

1⌠ ⌠
=
π2⌡ ⌡ ( l 2 + r 2 + 2 lr cosψ − a 2)1/2[( l − a )2 + r 2 + 2( l − a ) r cosψ]
0 0
δ( f,f 1) 2π a
( a 2 − r 2)1/2 r d r dψ
≤ ⌠ ⌠
π2 ⌡ ⌡ ( l 2 + r 2 + 2 lr cosψ − a 2)1/2[( l − a )2 + r 2 + 2( l − a ) r cosψ]
0 0
2δ( f,f ) a
( a 2 − r 2)1/2 r d r 2 a δ( f,f )
⌠
1 1
< < .
π ⌡ [( l − r) − a ] [( l − a ) − r ]
2 2 1/2 2 2
π[( l − a )2 − a 2]1/2
0
The integral operator (4.10.17) will be a contraction operator if
2a
< 1 ,
π[( l − a )2 − a 2]1/2
indicating that the iteration procedure will be convergent for l >2.18 a which
corresponds to a fairly close interaction. The estimation given above is crude.
Direct computations show that the iteration procedure converges even for
l =2.0005 a (which corresponds to the case when the shortest distance between the
cracks is equal to 0.0005 of its radius), and converges rapidly: the first iteration
with f ≡1 has the maximum relative error less than 2%, and the sixth iteration
may be considered practically as an exact solution since the maximum relative
error becomes less than 10-7. The accuracy of the first iteration improves as the
distance between cracks increases. For example, the first iteration for l =10 is
practically exact with maximum relative error less than 10-7. We could not go
closer than l =2.0005 a , not because of non-convergence, but because the standard
subroutine DBLIN from IMSL library, which was used to compute the integrals,
failed giving terminal errors. Though we do not have a rigorous proof, it seems
probable that the iteration procedure is theoretically convergent for an arbitrarily
small distance between cracks.
The values of the interaction function f (ρ,φ) are presented in (Fabrikant,
1987g) for various ratio of l/a . We limit ourselves here to just one abbreviated
table, related to the closest interaction considered, with l =2.0005 a . The reader is
referred to the original paper for additional data. The first line in Table 4.10.1
gives the ratio of the stress intensity factor of the interacting cracks to the stress
intensity factor of a solitary crack under the same uniform load. All the
-6
computations were made with the relative error not exceeding 10 . It was
established that Collins’ (1963) formulae are accurate within 1% for l >2.5 a . The
relative error of the central estimation corresponding to (4.10.14) does not exceed
2% for l >2.5 a . One can also notice that the central estimation is always
slightly below the exact value, thus giving in the case of two cracks a very
close lower bound for the quantities of interest. The same can be said about
the formulae by Collins (1963). The accuracy of the central estimation
deteriorates rapidly as l decreases, for example, the maximum error in the stress
intensity factor for l =2.2 (the distance between cracks is 0.2 of its radius) is
about 10%. The accuracy of the central estimation of the crack energy,
corresponding to (4.10.13) is much better, and is discussed in more detail in the
Table 4.10.1. The interaction function for l =2.0005 a
ρ φ 0. 30. 60. 90. 120. 150. 180.
1.00000 2.77577 1.18634 1.06686 1.03605 1.02471 1.02009 1.01881

0.91667 1.42135 1.16479 1.06729 1.03761 1.02614 1.02137 1.02003
0.83333 1.27366 1.14560 1.06728 1.03916 1.02766 1.02277 1.02139
0.75000 1.20176 1.12870 1.06682 1.04067 1.02928 1.02431 1.02289
0.66667 1.15812 1.11396 1.06593 1.04214 1.03102 1.02601 1.02455
0.58333 1.12865 1.10117 1.06464 1.04353 1.03287 1.02788 1.02642
0.50000 1.10743 1.09012 1.06299 1.04482 1.03483 1.02996 1.02850
0.41667 1.09146 1.08058 1.06103 1.04597 1.03692 1.03228 1.03086
0.33333 1.07903 1.07234 1.05883 1.04696 1.03911 1.03486 1.03353
0.25000 1.06913 1.06519 1.05646 1.04777 1.04142 1.03775 1.03657
0.16667 1.06108 1.05899 1.05396 1.04836 1.04382 1.04101 1.04007
0.08333 1.05442 1.05358 1.05141 1.04872 1.04630 1.04468 1.04411
0.00000 1.04884 1.04884 1.04884 1.04884 1.04884 1.04884 1.04884
next section. In the examples to follow we shall consider only the central
estimation for all the quantities.
Example 2: Infinite row of equal cracks. Let the crack radius be a , and
the distance between the adjacent crack centres be l . The cracks are opened by
a uniform pressure p . The central estimation for the crack opening volume 2 V
can be defined, according to (4.10.8), by a single equation
∞
V =
8
3
π Ha3p +
4
π
V Σ
k=1
a
2 1/2 − sin
(k l − a )
2 2
-1 a 
kl 
,
with the result
V0
V = , (4.10.18)
∞
1 −
4
π Σ
k=1
a
2 1/2 − sin
(k l − a )
2 2
-1 a 
kl 
where V 0=(8/3)π Hp is the corresponding result for the case of a solitary crack.
The crack opening displacement will take the form, according to (4.10.6)
w (ρ,φ) = ( a 2 − ρ2)1/24 pH

∞
ρ2 + ( kl )2
2V
+ 2
π Σ(k l
k=1
2 2 2 1/2 4 4
,
− a ) [ρ + ( kl ) − 2(ρ kl ) cos2φ] 2
and, as an immediate consequence of the previous expression, the stress intensity

factor will be defined by
∞
K (φ) = K 01 +

V
Σ a 2 + ( kl )2 ,
2π pH ( k l − a ) [ a + ( kl ) − 2( akl ) cos2φ]
2
k=1
2 2 2 1/2 4 4 2
where K stands, as before, for the stress intensity factor of a solitary crack.
0
Example 3: Polygonal configuration. Consider a circular crack of radius

b , its centre coinciding with the centre of a regular polygon, surrounded by n
cracks of radius a , with their centres located at the apices of the polygon. Let
l be the distance from the polygon centre to its apex. Let a uniform pressure
p c open up the central crack, and a uniform pressure p act inside the cracks
located at the polygon apices. Due to the system symmetry, the crack opening
volume can be defined by just two equations
nV  2
8 2 b b 
Vc = π Hb3p c + − sin-1 ,
3 π  ( l − b 2)1/2 l 
Vc  2
8 2 a a 
V = π Ha3p + − sin-1
3 π  (l − a )
2 1/2 l 
+
2
π
V Σ
k=2
 a
2 1/2 − sin
 (lk − a )
2
-1 a 
lk 
,
where V and V denote half of the volume of the central crack and the apex
c
crack respectively, and l =2 l sin(π k/n ). The solution is
k
b 3 p c c 22 + a3p c 12 b 3 p c c 21 + a3p c 11
8 8
Vc = πH , V = πH ,
3 c 11 c 22 − c 12 c 21 3 c 11 c 22 − c 12 c 21
(4.10.19)
where
c
11
= 1, c
22
= 1 −
2
π Σ(l
k=2
2
k
a
2 1/2 − sin
− a)
-1 a 
l
k
,
2  a  2 
− sin-1 .
a b b
c = − sin-1 , c = n 2
21 π  ( l 2 − a 2)1/2 l  12 π ( l − b )
2 1/2 l
The central crack opening displacements are, according to (4.10.6),
w (ρ, φ) = 4 Hp ( b 2 − ρ2)1/21
c c 
n
+
V
4π Hp ( l 2 − b 2)1/2
2
c
Σ k=1
ρ + l
2 2
1
k

− 2ρ l cos(φ − φ ) 
(4.10.20)
Assuming φ =2π k/n , the summation in (4.10.20) can be performed, with the result
k
w (ρ,φ) = 4 Hp ( b 2 − ρ2)1/21
c c 
+
nV l 2n − ρ2n .
4π Hp ( l − b ) ( l − ρ ) (ρ + l − 2ρ l cosn φ)
2 2 2 1/2 2 2 2n 2n n n
c
The stress intensity factor for the central crack takes the form

K = K 1 +
nV l 2n − b 2n .
c 0 4π Hp ( l − b ) ( b + l − 2 b l cosn φ) 
2 2 2 3/2 2n 2n n n
c
Due to symmetry of the system, all the apex located cracks will have the same
characteristics. The following expressions are valid in a local system of polar
coordinates, with its origin at the crack centre and the polar axis coinciding with
the line connecting the polygon centre with its apex. The crack opening
displacements are
 1 
V
c
w (ρ,φ) = 4 Hp ( a − ρ ) 1 + 2 2 1/2
4π Hp ( l − a ) (ρ + l 2 + 2 l ρcosφ)
2 2 2 1/2 2

,
n
+ Σ
k=2
( l 2k − a 2)1/2
V
[ρ2 + l 2k + 2 l kρcos(φ + ψk)]

where V and V c are defined by (4.10.19); l k=2 l sin[π( k −1)/ n ], and ψk=2π( k −1)/ n .
The stress intensity factor can be written as
 1 
Vc
K = K0  1 +
4π2 Hp  ( l 2 − a 2)1/2 ( a 2 + l 2 + 2 la cosφ)

.
n
+ Σ
k=2
( l 2k − a 2)1/2
V
[ a 2 + l 2k + 2 al kcos(φ + ψk)] 

Discussion. It is of interest to compare our results to those available in the
literature. The paper by Collins (1963), though published 23 years ago, seems to
be still the most reliable source. He did not give the stress intensity factor
explicitly but it could be derived easily for the case of two equal cracks, and it
reads in our notation
 2ε 3 8ε 5 4ε 6 30ε7 12ε8
K = K 0 1 + + + + +
3π 5π 9π2 7π 5π2

4ε 4 
+ 9ε4cosφ +
2ε 3 4ε 5  18ε2 2ε 3 
+ 1 + 3ε 2 + 1 + + cos2φ
3π  3π  3π  5 3π 
4ε 6  21ε2 4ε 7 4ε 8 
+ 1 + cos3φ + cos4φ + cos5φ. (4.10.21)
3π  5  3π 3π

Our result (4.10.14), if expanded in series, reads
∞
K = K 01 +

2ε 3
3π
(1 + 2 Σ
k=1
εkcosk φ)(1 +
3 2
2
ε +
2ε 3
3π
+
4
ε + ... ).
7 4

(4.10.22)
Comparison of (4.10.21) with (4.10.22) reveals quite a few common terms.

Though Collins himself assumed that his results are valid only for the cracks
whose radius is much smaller than the distance between their centres, the
numerical results indicate that (4.10.21) is accurate within 1% for l ≥2.5 a which
corresponds to the shortest distance reported in the literature. Direct computations
show that the central estimation corresponding to (4.10.14) does not differ from
Collins’ (4.10.21) by more than 3% in the whole range 2 a < l <∞, and differs by
less than 0.9% for l >2.5 a . The stress intensity factor, due to Andreikiv and
Panasiuk (1971), is
2ε3 2ε4
K = K (1 + + cosφ).
0 3π 3π
One should notice that a factor 2 is missing in the third term of their result.
Collins (1963) gave the following expression for the crack energy of two
equal cracks
W = W 1 + ,
2ε 3 6ε 5 4ε 6 18ε7 32ε8
+ + + + + ...
0 3π 5π 9π2 7π 15π2 
where W =(8/3)π Ha3p2 is the energy of a solitary crack. Our expression for the
0
crack energy is
W
0
W = , (4.10.23)
1 − c
where c is defined according to (4.10.13) as
2 
− sin-1ε .
ε
c =
π  (1 − ε )
2 1/2

Series expansion of (4.10.23) results in
W = W 1 + .
2ε 3 3ε 5 4ε 6 15ε7
+ + + + ...
0 3π 5π 9π2 28π 
which is very close to Collins’ result, ours being slightly lower. Table 4.10.2
displays the values of W/W . It confirms that the error of the central estimation
0
(4.10.23) is under 3% even for a close interaction when the shortest distance
between the cracks is equal to 0.001 of their radius. Collins’ results are
presented in order to emphasize the accuracy of our numerical solution. The
energy increase due to the crack interaction is relatively small even for very
close interactions; this is mainly due to the sharp localization of interaction
effects (see, for example, Table 4.10.1.)
Table 4.10.2. Crack energy increase (two equal cracks).
l/a upper lower central result of exact error

bound bound estimation Collins result (%)
2.001 – 1.008800 1.035366 1.046308 1.062976 2.6

2.005 – 1.008762 1.035105 1.045921 1.061694 2.5
2.010 – 1.008714 1.034783 1.045444 1.060378 2.4
2.020 – 1.008621 1.034152 1.044510 1.058066 2.3
2.050 – 1.008349 1.032352 1.041861 1.052411 1.9
2.100 2.965366 1.007921 1.029637 1.037908 1.045276 1.5
2.200 1.498350 1.007150 1.025091 1.031413 1.035399 1.0
2.500 1.117132 1.005370 1.016120 1.019160 1.020067 0.4
3.000 1.035432 1.003526 1.008809 1.009900 1.010035 0.1
5.000 1.003526 1.001009 1.001764 1.001834 1.001835 0.0
10.000 1.000294 1.000161 1.000214 1.000216 1.000216 0.0
It is also of interest to compare the upper bound for the stress intensity
factor defined by (4.10.14) with the upper bound derived by Ioakimidis (1982).
His result for two equal cracks reads in our notation
K
0
K = (4.10.24)
3
2a
1 −
3π( l − 2 a )
If we expand (4.10.13) in power series, retaining the first term only, the result
is
V0
V = (4.10.25)
3
2a
1 −
3π( l − a )
Comparison of (4.10.24) with (4.10.25) explains why our estimation is sharper:

we have in the denominator l − a while Ioakimidis has l −2 a . Here is a
numerical example. For l =3 a the exact result reads K =1.0234 K 0. Our upper
bound gives K =1.127 K 0 with an error of 10%, while the result of (4.10.24) is
K =1.269 K 0, with an error of about 25%.
Collins (1963) gave the following expression for the crack energy in the
case of an infinite row of equal cracks
W = W (1 + 0.5102ε3 + 0.7921ε5 + 0.2603ε6 + 1.6507ε7 + 0.8083ε8 + ...)

0
(4.10.26)
Our expression due to (4.10.18) is
W
0
W = (4.10.27)
∞
1 −
4
π Σ(k
k=1
2
ε
2 1/2 − sin
− ε)
-1 ε 
k 
The value of W/W , computed on the basis of (4.10.26) and (4.10.27), can be
0
found in (Fabrikant, 1987g). Our (4.10.27) does not differ from (4.10.26) by
more than 2% in the whole range of 2.001 a ≤ l <∞. Of course, this does not
mean that the central estimation is so accurate; it means only that our simple
approximate solution is almost as accurate as a very complicated one by Collins.
We are not aware of any result in the mechanics literature to compare with our
results for the polygonal configuration.
Some well known results can be simplified significantly by using the
method of computation of various integrals involving distances between several
points. For example, here is the set of governing integral equations derived by
Panasiuk et al (1986, p. 83) for the problem of interaction of N coplanar thin
spheroidal inclusions:
N
w (x , y ) + HΓw (x , y ) −
n n n n n n
1
4π2 Σ ⌠ ⌠ w ( x , y )Γ[( x − x )2
⌡⌡ k k k n k
k=1 Sk
k≠n
+ ( y n − y k)2]-3/2 d x kd y k = f n( x n, y n), for n = 1,2, ... , N.

(4.10.28)
Here f n is a known function, w n is the normal displacement at the boundary of
the n th inclusion, S n is its median crossection, ( x n, y n)∈ S n and the integral
operator Γ is defined by
Φ(ξ,η) dξdη
ΓΦ( x k, y k) = ⌠ ⌠
⌡ ⌡ [( x k − ξ)2 + ( y k − η)2]1/2
Sk
1 dξ dη
−
π ⌠ ⌠ {[ξ + η − a k][( x k − ξ)2 + ( y k − η)2]}1/2
2 2 2 2
⌡⌡
Sk
Φ(ξ1,η1)( a 2k − ξ21 − η21)1/2

×⌠ ⌠ dξ1dη1. (4.10.29)
⌡ ⌡ (ξ − ξ )2 + (η − η )2
Sk 1 1
Here a k is the radius of median crossection of the k th inclusion, and S k denotes

the area outside S k. The double and quadruple integrals in (4.10.29), which is a
kernel of the integral equation (4.10.28), make any numerical solution next to
impossible. Panasiuk et al. (1986) have managed to give an approximate
solution for the case when the inclusions are far apart, which is of little
practical value, since there is almost no interaction at such distances between the
inclusions. Let us show that (4.10.28) can be simplified so that its kernel be
presented in elementary functions. Making use of the integral
2π ∞
( a 2 − ρ20)1/2 1 r d r dψ
⌠ ⌠ 2
⌡ ⌡ ( r − a 2)1/2 ρ20 + r2 − 2ρ0 rcos(φ0−ψ) [ρ2 + r2 − 2ρrcos(φ−ψ)]1/2
0 a
2π  π ( a 2 − ρ2)1/2( a 2 − ρ20)1/2
-1  ,
= − tan
R  2  aR 
one can change the order of integration in the second integral of (4.10.29),
perform the integration in S k, and the integral operator Γ simplifies in polar
coordinates significantly, namely,
2 ⌠⌠ 1 η
ΓΦ(ρ,φ) = tan-1( ) Φ(ρ0,φ0) ρ0dρ0dφ0,
π ⌡⌡ R R
Sk
which is much simpler than (4.10.29). It is reminded that R and η are defined
by (4.1.14). We can also compute Γ[( x n − x k)2 + ( y n − y k)2]-3/2 in elementary
functions. Indeed, one may obtain from (1.6.19)
2π a ρ dρ dφ
⌠ ⌠ 1 tan-1(η ) 0 0 0
⌡ ⌡ R R [ρ20 + r 2 − 2 r ρ cos(φ −ψ)]3/2

0 0
0 0
2π( a 2 − ρ2)1/2
= , for r > a .
( r 2 − a 2)1/2[ r 2 + ρ2 − 2 r ρcos(φ−ψ)]
The above results simplify (4.10.28) so significantly that now it can be easily
solved by iteration numerically or analytically.
Exercise 4.10
1. Derive (4.10.4).
2. Rederive (4.10.4) by using an alternative approach: the results of section 4.2

combined with the reciprocity theorem.
3. Investigate the convergence of an iterative process, applied to the integral

equation (4.10.16).
4. Consider the interaction of a crack with a microcrack.
5. Consider the interaction of two cracks due to the bending of an elastic

space. Assume the space stretched so that cracks do not close due to the
bending.
4.11 Close interaction of coplanar circular cracks

under shear loading
In this section, the general method is applied to the stress analysis of an

elastic space weakened by several arbitrarily located coplanar circular cracks
subjected to an arbitrary shear loading. The problem is reduced to a set of
Fredholm integral equations. The number of equations is equal to the number of
cracks, and can be reduced in the case of a symmetrical configuration. The
equations are non-singular. It is shown that the iteration procedure is convergent,
and the convergence is so rapid that a practically exact numerical solution can be
obtained even for very closely located cracks. One can get an approximate
analytical solution without having solved the integral equations. It provides
sufficiently accurate estimations for the quantities of interest, like the stress
intensity factor, the crack energy, the crack face displacement, etc. The cases of
two cracks and an infinite row of equal cracks are considered as illustrative
4.11 Close interaction of coplanar circular cracks 293
examples.
Theory. Consider an elastic space weakened in the plane z =0 by n
arbitrarily located circular cracks. The cracks do not intersect. Let the centre
of the k th crack be located at the point with Cartesian coordinates x and y ,
k k
and its radius be denoted by a k. We introduce the complex tangential
displacement u = u x+ iu y, and the complex shear stress τ=τzx+ i τyz. Let an arbitrary
skew-symmetric shear traction τk be applied to the k -th crack faces. We can
single out, without loss of generality, the crack number 1, and consider the set
of cracks in the local polar system of coordinates with the origin coinciding with
the centre of the first crack. In order to be able to use the reciprocal theorem,
we need to consider the second set of tractions applied to the same crack
configuration. We apply two unit concentrated forces T to both faces of the
x
first crack in opposite directions at the point with polar coordinates (ρ,φ), and
parallel to the axis Ox . We also apply shear tractions q and q to the
kx ky
remaining cracks. These tractions are chosen so as to provide zero displacement
discontinuity at the crack faces, so that the whole system would behave as a
single crack (number one) in an infinite body. This choice will allow us to use
the Green’s functions for an isolated circular crack derived in section 4.4. The
following integral equation can be obtained by using the reciprocal theorem:
n n
u 1x + Σ ⌠ ⌠ q u dS +
⌡ ⌡ kx kx k Σ ⌠ ⌠ q u d S = ⌠ ⌠ (τ u + τ u )d S .
⌡ ⌡ ky ky k ⌡ ⌡ 1x xT x 1y yT
x
1
k=2 Sk k=2 Sk S1
(4.11.1)
Here q kx and q ky stand for the shear tractions in the k th crack domain due to a
pair of unit concentrated forces applied at an arbitrary point of the first crack in
the direction parallel to the Ox axis; u and u are the tangential
xT yT
x x
displacements of the first crack face due to the same unit forces; and u 1x, u kx,
and u ky are the as yet unknown tangential displacements of the first and the k th
crack faces respectively. Similar considerations, with the unit concentrated forces
T y applied parallel to the Oy direction, yield the second integral equation
n n
u 1y + Σ ⌠ ⌠ s u dS +
⌡ ⌡ kx kx k Σ ⌠⌡ ⌠⌡ s u d S k = ⌠ ⌠ (τ1x u xT
ky ky ⌡⌡ y
+ τ u )d S .
1y yT
y
1
k=2 Sk k=2 Sk S1
(4.11.2)
The meaning of the notation in (4.11.2) is similar to that in (4.11.1). All the
integrals in (4.11.1) and (4.11.2) are evaluated on one side of the relevant crack.
Now we need the explicit expressions for the quantities q kx, q ky, s kx, s ky, u xT ,
x
u yT , u xT , u yT . These expressions were derived in section 4.4, and are
x y y
q kx= −ζ − ℜ Z , q ky = −ℑ Z = s kx , s ky = −ζ + ℜ Z , (4.11.3)
u xT = ζ1 − ℜ Z 1 + ζ2 + ℜ Z 2, u yT = −ℑ Z 1 + ℑ Z 2 ,
x x
u xT = ℑZ1 + ℑZ2 , u yT = ζ1 −ℜ Z 1 − ζ2 − ℜ Z 2 , (4.11.4)

y y
where ℜ and ℑ stand for the real and the imaginary part, and
iφ -i φ
-i φ
( a 21 − ρ2)1/2 G2 ( a 21 − ρ2)1/2 e 0(3ρ0e 0
− ρe )
ζ = , Z = -i φ
,
π2(ρ20 − a 21)1/2 R 2 G 1 π2(ρ20 − a 21)1/2 -i φ
ρ0(ρe − ρ0e 0 2
)
G1 η( a 1) G 22 (3 − t 1) η( a 1)
ζ1 = tan -1
, Z1 = ,
πR R πG1 a 21(1 − t 1)2
2 iφ
G 2ξ η( a 1) G 2 η( a 1)(ξ − t 1e 0
)
ζ2 = tan -1
, Z2 = . (4.11.5)
πR R π a 21 (1 − t 1)(1 − t 1)
Here
R =[ρ2 + ρ20 − 2ρρ0cos(φ−φ0)]1/2, η( x ) = ( x 2 − ρ2)1/2( x 2 − ρ20)1/2/ x ,

(4.11.6)
iφ -i φ
i(φ-φ0) iφ -i φ
t 1=(ρρ0/ a 21)e , ξ = (ρe − ρ0e 0)/(ρe − ρ0e 0
). (4.11.7)
We multiply equation (4.11.2) by the imaginary unit i , add the result to (4.11.1)
and, after substitution of (4.11.3), (4.11.4) and (4.11.5), obtain
G 1 2π a
1 η( a 1) G 22 (3 − t 1) η( a 1)
u 1(ρ,φ) = ⌠ ⌠  1 tan-1 − τ (ρ ,φ )ρ dρ dφ
π⌡ R R G 21 a 21(1 − t 1)  1 0 0 0 0 0
2
0 ⌡ 0
2 iφ
G 2 2π a
1 η( a 1) η( a 1) (ξ − t 1e 0
)
+ ⌠ ⌠  ξ tan-1  +  τ ( ρ , φ ) ρ dρ dφ
π⌡
⌡ R  R  a 1(1 − t 1)(1 − t 1)  1 0 0 0 0 0
2
0 0
-i φ iφ
-i φ
( a 21−ρ2)1/2 n u k(ρ0,φ0) G 2 u k(ρ0,φ0) (3ρ0e 0
− ρe )e 0
ρ dρ dφ
+
π2 Σ ⌠⌠
⌡⌡ R2
+
G1 -i φ
-i φ
 0 0 0.
(ρ20−a 21)1/2
k=2 Sk ρ0(ρe − ρ0e 0 2
)
(4.11.8)
The first two integrals in (4.11.8), though looking formidable, can be computed
exactly and expressed in elementary functions for any polynomial loading. They
give the tangential displacements of the first crack, as if it were an isolated
crack, under the prescribed loading τ . The remaining integrals represent the
1
influence of the other cracks. A similar procedure can be applied to the
remaining n −1 cracks, and the additional n −1 equations of the type (4.11.8) can
be obtained. Note that each such equation is valid in a local system of polar
coordinates related to a certain crack. The equations are non-singular. They can
be solved numerically by iteration. As will be shown further, the convergence is
so rapid, that the first iteration has an error less than 2.5% even for a very
close interaction when two cracks are separated only by 0.01 of their radius.
In the case of a uniform shear loading τ=τ = const, the set of equations
0
(4.11.8) simplifies as follows
G 21 − G 22 ( a 21 − ρ2)1/2 n u k(ρ0,φ0)
u 1 = 2τ0
G1
(a 2
− ρ) 2 1/2
+
π2 Σ ⌠⌠
⌡⌡ R2
k=2 Sk
-i φ iφ
G 2 u k(ρ0,φ0) (3ρ0e 0
− ρe-i φ) e 0
ρ dρ dφ
+  0 0 0
2 1/2 . (4.11.9)
G1 -i φ
-i φ  (ρ0 − a 1)
2
ρ0(ρe − ρ0e 0 2
)
In some cases we can obtain a sufficiently accurate analytical solution of the set
(4.11.8) by applying the mean value theorem. The result is:
2π a η( a 1)
G 1 G 22 (3 − t 1) η( a 1)
u (ρ,φ) =
1
⌠ ⌠  1 tan-1  − τ (ρ ,φ )ρ dρ dφ
π ⌡ R  R  G 21 a 21(1 − t 1)  1 0 0 0 0 0
2
0 ⌡
1
0
2 iφ
G2 2π a η( a 1) η( a 1) (ξ − t 1e 0
)
1
+ ⌠ ⌠  ξ tan-1 +  τ (ρ ,φ ) ρ dρ dφ
π ⌡  R R a 21(1 − t 1)(1 − t 1) 
0 ⌡
1 0 0 0 0 0
0
-i φ iφ
n
( a 21 − ρ2)1/2 Uk G 2 U k(3ρke k
− ρe-i φ)e k
+ Σπ (ρ 2 2
− a 21)1/2

ρ2 + ρ2k − 2 ρ ρkcos(φ − φ k)
+
G1 -i φ
.

k=2 k
ρk(ρe-i φ − ρke k 2
)
(4.11.10)
Here
U k = ⌠ ⌠ u kdSk ,
⌡⌡
Sk
and ρ and φ are the polar coordinates of a point inside S . Though the exact
k k k
location of the point is unknown, the fact of belonging to the domain S limits
k
the possible variation of the quantities of interest and allows the construction of
upper and lower bounds as well as a sufficiently accurate central estimation
which corresponds to the assumption of ρk and φk being located at the centre of
the k th crack. The symmetry considerations can also be used to sharpen the
estimates. It will be shown further that the central estimation provides in some
cases a sufficiently accurate solution to the problem.
The value of U 1 can be estimated by integration of (4.11.8) over the
domain S . The result is
1
G 21 − G 22 2π 1
a n
 a1
U1 = 2
G1 ⌡
⌠ ⌠ τ(ρ,φ)( a 2 − ρ2)1/2ρdρdφ + 2
1
⌡ π Σ ⌠ ⌠ 
⌡ ⌡ (ρ2 − a 21)1/2
0 k=2 Sk
0
a G u (ρ,φ)e2 i φ 
− sin-1
1
 u (ρ,φ) + 2
a 31
k
ρdρdφ. (4.11.11)
ρ k G ρ2(ρ2 − a 21)1/2
1 
We can use again the mean value theorem, with the result for the central
estimation
2π a
G 21 − G 22 1
U1 = 2 ⌠ ⌠ τ(ρ,φ)( a 21 − ρ2)1/2ρdρdφ
G
1 ⌡ ⌡
0 0
2 iφ
1k
n a1 a1 2G n U ke
Σ  U +
Σ
2 2
+ − sin-1 a3 .
π  ( l 21k − a 21)1/2 l 1k k πG1 1 l 21k( l 21k − a 21)1/2
k=2 k=2
(4.11.12)
Here ( l ,φ ) are the polar coordinates of the k th crack centre, with respect to
1k 1k
the system of coordinates having its origin at the centre of the first crack.
Integration of the remaining n -1 equations over the area of each crack provides
finally a system of n linear algebraic equations which can be solved with respect
to the unknowns U k. Their feeding back into (4.11.10) gives the complete
solution to the problem.
Define the stress intensity factor at the edge of the first crack as
K (φ) = lim {(ρ − a )1/2τ (ρ,φ)}.

1 1 1
ρ→ a
1
An important feature of the present method is the possibility to compute the

stress intensity factor directly through the displacements (see the derivation of
(2.8.46) for details):
a1 G 1 u 1(ρ,φ) + G 2e2 i φ u 1(ρ,φ)

K 1(φ) = − lim  . (4.11.13)
π( G 21 − G 22)√2 a 1 ρ→ a
 ( a 21 − ρ)
2 1/2

1
The stress intensity factor for the remaining cracks can be defined in a similar
manner. Substitution of (4.11.8) in (4.11.13) yields, after simplification,
2π a ( a 21 − ρ20)1/2 τ(ρ0,φ0) ρ0dρ0dφ0

 1
 ⌠ ⌠
1
K 1(φ) = − 2
π √2 a 1
 ⌡
0 0⌡
a 21 + ρ20 − 2 a 1ρ0cos(φ − φ0)
-i φ
2π a -i φ
G 2 ei φ 1 3 a 1e − ρ0e 0

+ ⌠ ⌠ ( a 21 − ρ20)1/2 τ(ρ0,φ0) ρ0dρ0dφ0
G1 a1 ⌡ -i φ
0 ⌡
-i φ
0
[ a 1e − ρ0e 0 2
] 
a1 n
G G 2
−
π3( G 21 − G 22) √2 a 1 Σ ⌠ ⌠  1 + 2 Θ e2 i φ u (ρ ,φ )
⌡ ⌡ R 21 G1 1  k 0 0
k=2 Sk 
 e
2 iφ
  ρ0dρ0dφ0
+ G2 + Θ1 u k(ρ0,φ0) . (4.11.14)
 R 21   (ρ20 − a 21)1/2
Here
-i φ
-i φ
3ρ0e 0
− a 1e
Θ1 = -i φ -i φ
, R 1 = [ a 21 + ρ20 − 2ρ0 a 1cos(φ − φ0)]1/2.
-i φ
ρ0e 0
( a 1e − ρ0e 0 2
)
The first two integrals in (4.11.14) give the stress intensity factor for an isolated
crack, while the remaining integrals represent the influence of the other cracks.
We shall see next how these general expressions can be applied to some specific
problems.
Example: Two cracks. Consider the case of two coplanar circular cracks
with radii a 1 and a 2 under the action of a uniform shear loading τ1 and τ2
respectively. Let l be the distance between their centres. As was established in
the previous Section, the problem is reduced to evaluating the integral
characteristics U 1 and U 2. We consider the central estimation only. Equations
(4.11.12) in this case will take the form
G 21 − G 22 a1 a1 2G2 a 31 U 2
4 3 2   +
U1 = πa τ + U − sin-1 ,
3 1 1 G1 π 2( l − a 21)1/2 l π G 1 l 2( l 2 − a 21)1/2
G 21 − G 22 a2 a2 2G2 a 32 U 1
4 3 2   +
U2 = πa τ U + − sin-1 .
3 2 2 G1 π 1( l − a 22)1/2 l π G 1 l 2( l 2 − a 22)1/2
(4.11.15)
Strictly speaking, the mean value theorem is not applicable in this case, since the
imaginary part of u does change sign inside the crack. A numerical evidence
will be presented later which justifies neglect of the imaginary part. Under this
assumption, the solution is
G 21 − G 22 a 1 τ1 + c 12 a 2 τ2
3 3
4
U1 = π ,
3 G1 1 − c c
12 21
G 21 − G 22 c 21 a 1 τ1 + a 2 τ2
3 3
4
U2 = π , (4.11.16)
3 G1 1 − c c
12 21
where
a1 a1 G2 a 31
2   ,
c 12 = − sin-1 +
π ( l 2 − a 21)1/2 l G 1 l 2( l 2 − a 21)1/2
a2 a2 G2 a 32
2  .
c 21 = − sin-1 + (4.11.17)
π ( l 2 − a 22)1/2 l G 1 l 2( l 2 − a 22)1/2
It will be shown that the central estimation gives a sufficiently accurate solution
even for relatively close crack interactions.
Formulae (4.11.16−4.11.17) simplify in the case of equal cracks as a = a = a ,
1 2
and if τ =τ =τ then
1 2 0
U
0
U = U =U = ,
1 2 1 − c
G2 a3
2  a -1 a .
c = − sin + (4.11.18)
π ( l − a )
2 2 1/2 l G1 l (l − a ) 
2 2 2 1/2
Here U 0=(4/3)π a 3τ0( G 21− G 22)/ G 1. Note that in the case of a uniform loading, the
crack energy W is proportional to U , namely, W =τ0 U . It is clear from
(4.11.18) that the crack interaction increases their energy when the applied
loadings act in the same direction, otherwise, their energy decreases. The crack
face displacements will take the form, according to (4.11.9) and (4.11.10),
 G 21 − G 22
u 1(ρ,φ) = ( a 21 − ρ2)1/2  2τ1 G1

U2 G 2 3 l − ρe-i φ 
+  1
+ ,
π2( l 2 − a 21)1/2 ρ + l − 2ρl cosφ
2 2 G 1 l ( l − ρe ) 
-i φ 2

 G 21 − G 22
u 2(ρ,φ) = ( a 22 − ρ)  2τ2
2 1/2
G1

U1 G 2 3 l + ρe-i φ 
+  1
+ . (4.11.19)
π2( l 2 − a 22)1/2 ρ + l + 2ρl cosφ
2 2 G φ
1 l ( l + ρe ) 
-i 2 
We recall that each expression in (4.11.19) is valid in a local system of polar

coordinates, with the coordinate origin located at the centre of the respective
crack. Substitution of (4.11.19) into (4.11.13) yields the expression for the stress
intensity factor.
Now we need an accurate numerical solution in order to estimate the
accuracy of the approximate formulae derived. For the sake of simplicity,
consider the case of two equal cracks. Assume the crack face displacements in
the form
G 21 − G 22
u (ρ,φ) = 2τ0 ( a 2 − ρ2)1/2 f (ρ,φ), (4.11.20)
G1
where f is an as yet unknown complex function. It may be called the

interaction function since it is equal to the ratio of the interacting crack face
displacements to those of an isolated crack. The values of f(a , φ) are related to
the stress intensity factor of interacting cracks through
√2 a
K (φ) = − [ G τ f ( a ,φ) + G τ f ( a ,φ)e2 i φ]. (4.11.21)
πG 1 0 2 0
1
In the case when τ is a real constant, we can neglect the imaginary part of f ,
0
and the stress intensity factor can be expressed as follows:
K (φ) = K (φ) f ( a ,φ), (4.11.22)

0
where
√2 a
K (φ) = − τ [ G + G e2 i φ]
0 πG1 0 1 2
is the stress intensity factor for an isolated crack. We shall call f(a ,φ) the
interaction factor , since its value is approximately equal to the ratio of the stress
intensity factor of an interacting crack to that of an isolated crack. Substitution
of (4.11.20) into (4.11.9) gives a convenient expression for the procedure of
iteration
G 21 − G 22 
u (ρ,φ) = 2τ ( a 2 − ρ2)1/2 1
0 G
1 
2π a ( a 2 − r 20)1/2 r 0d r 0dψ0
f(r , ψ )
1⌠ ⌠  0 0
+ 2 
π ⌡ ⌡ ( l2 + r20 + 2lr cosψ0 − a 2)1/2 r2 + r20 +2rr cos(ψ − ψ0)
0 0 0  0
-i ψ iψ
G f ( r ,ψ ) (2 l + r e 0
− r e-i ψ)( l + r 0e 0
) 
2 0 0 0
+
G -i ψ
. (4.11.23)
-i ψ 2
1
( r 20 + l 2
+ 2 lr cosψ ) ( r e 0
− re ) 
0 0 0
Here we have introduced the new variables r =(ρ2+ l 2−2 l ρcosφ)1/2,

ψ=π−sin-1[(ρ/ r )sinφ]. The integral in (4.11.23) has a logarithmic singularity for
r = l − a , ψ=0, as l →2 a , therefore the procedure of iteration might not be
convergent for l very close to 2 a . Direct computations show that the iteration
procedure converges for l =2.01 a ( which corresponds to the case when the
shortest distance between the cracks is equal to 0.01 of its radius), and converges
rapidly: the first iteration with f ≡1 has the maximum relative error less than
2.5%, and the sixth iteration may be considered practically as an exact solution,
since the error becomes less than 10-7. The accuracy of the first iteration
improves as the distance between cracks increases. For example, the first
iteration for l =10 is practically exact with maximum relative error less than 10-7.
We could not go closer than l =2.01 a , not because of non-convergence, but
because the standard subroutine DBLIN from the IMSL library, which was used
to compute the integrals, failed, giving terminal errors. Though we do not have
a rigorous proof, it seems probable that the iteration procedure is theoretically
convergent for arbitrarily small distance between cracks.
In the case when G =0 (for an isotropic body this condition is equivalent
2
to the Poisson ratio ν=0), the crack interaction due to a shear loading is the
same as the crack interaction due to a normal loading (compare (4.1.9) with
(4.4.14)). The maximum value of the ratio G / G for an isotropic body is 1/3,
2 1
and this value was taken in all the numerical computations, in order to expose
the maximum possible difference between the crack interaction due to a normal
loading, and the crack interaction due to a shear loading. Some values of the
interaction function f (ρ,φ) are presented in Tables 4.11.1 and 4.11.2, for the
closest interaction considered, corresponding to l/a =2.01. The reader is referred
to the original paper (Fabrikant, 1989) for the complete data.
Table 4.11.1. The interaction function (real part) for l =2.01 a
ρ φ 0 15 30 45 90 135 180
1.00 2.14189 1.52935 1.24944 1.14303 1.06042 1.04512 1.04195

0.75 1.32715 1.28240 1.20537 1.14660 1.07271 1.05393 1.04976
0.50 1.19220 1.18283 1.16050 1.13539 1.08451 1.06514 1.06027
0.25 1.13108 1.12916 1.12389 1.11649 1.09335 1.07930 1.07499
0.00 1.09668 1.09668 1.09668 1.09668 1.09668 1.09668 1.09668
Table 4.11.2. The interaction function (imaginary part) for l =2.01 a ]
ρ φ 0 15 30 45 90 135 180
1.00 0.00000 -0.11448 -0.09508 -0.06835 -0.02738 -0.01092 0.00000

0.75 0.00000 -0.04250 -0.05420 -0.04922 -0.02518 -0.01053 0.00000
0.50 0.00000 -0.01526 -0.02474 -0.02774 -0.01983 -0.00918 0.00000
0.25 0.00000 -0.00458 -0.00834 -0.01082 -0.01107 -0.00609 0.00000
0.00 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000
All the computations were made with a relative error not exceeding 10-6.
The first line in Table 4.11.1 is approximately equal to the ratio of the stress
intensity factor of interacting cracks to the stress intensity factor of an isolated
crack under the same uniform load. The data in Table 4.11.2 justify our neglect
of the imaginary part in the interaction function when deriving the approximate
analytical solutions: the maximum value of the imaginary part is less than 8% of
the corresponding real part, and it reduces significantly, when the distance
between the cracks increases. The data in the tables are presented for 0≤φ≤π.
The following rules should be applied if one is interested in the value of the
interaction function for φ>π: ℜ f (ρ,φ)=ℜ f (ρ,2π−φ) and ℑ f (ρ,φ)=−ℑ f (ρ,2π−φ).
Comparison with the results given in section 4.10 shows that the crack interaction
due to a shear loading (when acting on both cracks in the same direction) is
stronger than the interaction due to a normal loading. For example, the
maximum value of the interaction factor for the case l =2.01 a and G /G =1/3 is
2 1
2.1419 (Table 4.11.1) while the corresponding value in the case of a normal
loading is 1.8613.
We can now assess the accuracy of the analytical solution given by
(4.11.18−4.11.19). We can compute the exact value of the crack energy W by
using the computed data and compare it with the approximate values due to
(4.11.18). The ratio W/W ( W is the energy of an isolated crack) is given in
0 0
Table 4.11.3. The approximate value was computed as W/W =1/(1− c ), where c
0
Table 4.11.3. The ratio W / W 0 due to exact and approximate solutions.
l/a 10.0 3.0 2.5 2.1 2.05 2.02 2.01
exact 1.00043 1.01940 1.03817 1.08367 1.09621 1.10603 1.11000

approx. 1.00043 1.01736 1.03165 1.05802 1.06333 1.06684 1.06808
error (%) 0.0 0.2 0.6 2.4 3.0 3.5 3.8
is defined by (4.11.18). The agreement is very good for l/a =≥2.5. Even for a
very close interaction ( l/a =2.01) the relative error does not exceed 4%; of course,
this is mainly due to the fact that the increase in the crack interaction energy is
rather small. This should be attributed to a sharp localization of the interaction
effects (see Table 4.11.1).
The analytical expression for the interaction function can be written,
according to (4.11.19), in the form
G1 U  1
f (ρ,φ) = 1 +  2
2τ ( G 21 − G 22) π2( l 2 − a 2)1/2
ρ + l − 2ρl cosφ
2
0
G2 3 l − ρe-i φ 
G 1 l ( l − ρe-i φ)2 
+ .

We have computed only the interaction factor f ( a ,φ) due to the last formula and
compared it with the exact values in Table 4.11.4. The relative error of the
Table 4.11.4. Comparison of exact and approximate solutions

for the interaction factor.
l/a φ(deg.)= 0 30 60 90 120 150 180
Real exact 1.11020 1.07479 1.04136 1.02764 1.02198 1.01959 1.01892

approximate 1.07927 1.05793 1.03465 1.02393 1.01925 1.01722 1.01664
2.50 error (%) 2.8 1.6 0.6 0.4 0.3 0.2 0.2
Imag. exact 0.00000 -0.01875 -0.01485 -0.00936 -0.00547 -0.00254 0.00000
approximate 0.00000 -0.01439 -0.01210 -0.00782 -0.00461 -0.00215 0.00000
error (%) 0.0 23.3 18.5 16.5 15.7 15.4 0.0
Real exact 1.63484 1.21826 1.08826 1.05542 1.04397 1.03945 1.03821

approximate 1.21014 1.12652 1.06405 1.04226 1.03379 1.03029 1.02931
2.05 error (%) 26.0 7.5 2.2 1.2 1.0 0.9 0.9
Imag. exact 0.00000 -0.07834 -0.04299 -0.02416 -0.01359 -0.00622 0.00000
approximate 0.00000 -0.04268 -0.02834 -0.01667 -0.00951 -0.00437 0.00000
error (%) 0.0 45.5 34.1 31.0 30.0 29.7 0.0
central estimation of the real part of the interaction factor does not exceed 3%
for l >2.5 a . Though the relative error of the imaginary part is large, this is due
to the fact that the imaginary part constitutes a small percentage of the real part;
the absolute error is very small, and we can consider the analytical solution
(4.11.18−4.11.19) sufficiently accurate when the distance between interacting cracks
is not less than half of their radius. The accuracy of the central estimation
deteriorates rapidly as l decreases. One can also notice that the central
estimation is always slightly below the exact value thus giving a very close
lower bound for the quantities of interest in the case of two interacting cracks.
Infinite row of equal cracks. Let the crack radius be a , and the distance
between the adjacent crack centres be l . The cracks are subjected to a uniform
shear loading τ. The central estimation for the integral characteristic U can be
defined, from (4.11.12), by a single equation
G 21 − G 22 ∞

4
U = π a 3τ
3 G
1
+
4
π Σ  2 2
k=1 
a
2 1/2 − sin
(k l − a )
-1 a 
kl 
U
G
2 a3U 
G k2l2 ( k2l2 − a 2)1/2
+ .
1 
If we neglect the imaginary part of U then the solution is
U
0
U = ,
∞
 G a3 
Σ -1 a 
4 a 2
1 −
π  22 − sin
kl 
+
G k2l2 ( k2l2 − a 2)1/2
( k l − a 2)1/2
k=1  1 
(4.11.24)
where U =(4/3)π a 3
τ( G 21− G 22)/ G corresponds to the case of an isolated crack. The
0 1
crack face displacement will take the form, according to (4.11.9) and (4.11.10),
 G 21 − G 22
u (ρ,φ) = ( a 2 − ρ2)1/2  2τ G
 1
∞ G 3 k2l2 − ρ2e-2 i φ 
Σ  2(ρ2 + k2l2) 
U 2 1
G ( k2l2 − ρ2e-2 i φ)2  ( k2l2 − a 2)1/2 
+ 2 + ,
π  ρ4 + k4l4 − 2ρ2 k2l2 cos2φ 1 
k=1
(4.11.25)
and substitution of (4.11.25) in (4.11.13) will give the expression for the stress
intensity factor.
Discussion. It is of interest to compare our results with those available in
the literature. We have found only one paper (Fu and Keer, 1969) where the
problem of two interacting coplanar circular cracks was considered by a method
similar to that of Collins (1963). Only the case when the distance between the
crack centres l is much greater than the crack radius a (ε= a/l << 1) was
considered. Fu and Keer considered in detail two equal cracks subjected to a
uniform shear loading, acting on both cracks in the same direction horizontally
(Case a ), and acting in opposite directions (Case b ). There are several points in
their solution which seem to be incorrect. One of the results (Fu and Keer,
1969, p. 371) states that the absolute value of the m th harmonic ( m =1,2,3, ...)
of the vertical displacement u is equal to the corresponding harmonic of the
y
horizontal displacements u x. This cannot be true, since the vertical displacements

depend on the ratio G2/G1 (in the isotropic case this ratio is equal to ν/(2−ν),
where ν is the Poisson coefficient), and the vertical displacements vanish when
G 2=0 (ν=0)).
It is also possible to compare the expressions for the increase in the strain
energy of deformation W per crack. The expression, given by Fu and Keer
(1969), reads in our notation:
W = W 0 1
4 3 3 2 16 6
ε ( + ) ε5 + ε , (4.11.26)
 3π 5 5π 9π2 
where ε= a/l ; the sign corresponds to the cases (a) and (b) respectively, and
W =(4/3)π a τ ( G 1− G 2)/ G stands for the energy of an isolated crack. Note an
0
3 2 2 2
1
obvious misprint in (4.11.26): the plus sign should correspond to the case (a)
and minus to the case (b). Our expression for the crack energy is
W0
W = , (4.11.27)
1 − c
where c is defined according to (4.11.18) as
G2 a3
2  a -1 a .
c = − sin + (4.11.28)
π ( l 2 − a 2)1/2 l G 1 l 2( l 2 − a 2)1/2
Series expansion of (4.11.27) results in
G G
W = W 0 1 +
2ε3 1 2 ε5 3 2
( + ) + ( + )
 π 3 G1 π 5 G1
G2 G2
4ε6
+ 2 (
1
+ ) +
2 3ε7
(
5
+ ) + ... . (4.11.29)
π 3 G1 4π 7 G1 
There is a definite disagreement between (4.11.26) and (4.11.29): each term in

(4.11.29) depends on the ratio G2/G1 while each term in (4.11.26) is not
dependent of the elastic constants. In the case of an isotropic body
G2/G1 =ν/(2−ν) (as it should), and we can observe an agreement between (4.11.26)
and (4.11.29) only for ν=1/2 which is just a coincidence. Collins (1963) gave
Appendix A4.1 307
the following expression for the case of two cracks subjected to a normal
loading:
W = W 0 1 + .
2ε3 6ε5 4ε6 18ε7 32ε8
+ + + + + ...
 3π 5π 9π2 7π 15π2 
(4.11.30)
As was noticed before, in the case when G 2=0 (ν=0) the interaction of cracks
subjected to a shear loading is mathematically equivalent to the interaction under
a normal loading, which means that both (4.11.26) and (4.11.29) should be in
agreement with (4.11.30) for ν=0. One can see that this is not the case for
expression (4.11.26).
Exercise 4.11
1. Derive (4.11.8).
2. Establish (4.11.11).
3. Investigate convergence of the procedure of iteration, applied to (4.11.8).
4. Consider the case of a polygonal configuration of identical cracks.
Appendix A4.1
Here the main potential function is given, together with selected partial
derivatives. We define the potential function by
2π a
Ψ = ⌠ ⌠ ( a 2 − ρ20)1/2 ln( R 0 + z ) ρ0dρ0dφ0 ,

⌡ ⌡
0 0
where R 0=[ρ2 + ρ20 − 2ρρ0cos(φ−φ0) + z 2]1/2. The integral can be computed in

elementary functions:
π 2 a 1 10 2
Ψ = z (2 a 2 − ρ2 + z 2)sin-1( ) + ( a 2 − l 21)1/2(5ρ2 − a
2 3 l2 3 3
a ln[ l 2 + ( l 22 − ρ2)1/2]
11 2 4 3
− 2 l 22 − l) + (A4.1.1)
3 1 3 
The following derivatives may be computed:
l 21 + 2 a 2
= π x  − z sin-1
∂Ψ a 2 a 3
+ ( a 2 − l 21)1/2(1 − ) + (A4.1.2)
∂x  l
2
3ρ2 3ρ2
l 21 + 2 a 2
= π y  − z sin-1
∂Ψ a 2 a 3
+ ( a 2 − l 21)1/2(1 − ) + (A4.1.3)
∂y  l
2
3ρ2 3ρ2
2 a 2 − 3 l 21
π 
( l 22 − a 2)1/2 
∂Ψ a
= (2 a 2 + 2 z 2 − ρ2) sin-1 − (A4.1.4)
∂z 2  l a 
2
l 21 + 2 a 2
ΛΨ = πρe  iφ
− z sin-1 a
+ ( a − l 1) (1 −
2 2 1/2
) +
2 a 3
(A4.1.5)
 l
2
3ρ2 3ρ2
= π− z sin-1
∂2Ψ a
+ ( a 2 − l 21)1/2
∂x2  l
2
2 a 3 − ( l 21 + 2 a 2) ( a 2 − l 21)1/2
2x2
+ (1 − 2 )  (A4.1.6)
ρ 3ρ2 
∂2Ψ  -1 a
2 = π − z sin + ( a 2 − l 21)1/2
∂y  l
2
2 a 3 − ( l 21 + 2 a 2) ( a 2 − l 21)1/2
2y2
+ (1 − 2 )  (A4.1.7)
ρ 3ρ 2

2 a 3 − ( l 21 + 2 a 2) ( a 2 − l 21)1/2
∂2Ψ
= −2π xy (A4.1.8)
∂x∂y 3ρ4
2 a 3 − ( l 21 + 2 a 2) ( a 2 − l 21)1/2
2 iφ
Λ2Ψ = −2πe (A4.1.9)
3ρ2
∆Ψ = 2π  − z sin-1 + ( a 2 − l 21)1/2 
a
(A4.1.10)
 l
2

Appendix A4.1 309
∂2Ψ  zsin-1 a − ( a 2 − l 2)1/2 

2 = 2π (A4.1.11)
∂z  l
2
1

a ( l 22 − a 2)1/2
ΛΨ = πρei φ  −sin-1 
∂ a
+ (A4.1.12)
∂z  l
2
l 22 
∂3Ψ  -1 a
a ( l 22 − a 2)1/2
1 + 2 a x
2 2

= π −sin l +  
(A4.1.13)
∂x ∂z
2
l 22 l 22( l 22 − l 21)
 2
∂Ψ
3  -1 a
a ( l 22 − a 2)1/2
 2a2y2 
= π −sin + 1 + 2 2 (A4.1.14)
∂ y 2∂ z

l
2
l 2
2
 l 2( l 2 − l 21) 
∂3Ψ  -1 a
a ( l 22 − a 2)1/2 
= 2 π sin l −  (A4.1.15)
∂z 3
l 22 − l 21
 2 
a 3 xy ( l 22 − a 2)1/2
∂3Ψ
= 2π (A4.1.16)
∂x∂y∂z l 42( l 22 − l 21)
a 2 x ( a 2 − l 21)1/2
∂3Ψ
= 2π (A4.1.17)
∂ x∂ z2 l 22( l 22 − l 21)
a 2 y ( a 2 − l 21)1/2
∂3Ψ
= 2π (A4.1.18)
∂ y∂ z2 l 22( l 22 − l 21)
a 3( l 22 − a 2)1/2
∂
Λ2Ψ = 2πρ2e2 i φ (A4.1.19)
∂z l 42( l 22 − l 21)
a ( l 22 − a 2)1/2
∆Ψ = 2π  −sin-1 
∂ a
+ (A4.1.20)
∂z  l
2
l 22 − l 21 
( a 2 − l 21)1/2
∂2
2 iφ
2 ΛΨ = 2π a ρe (A4.1.21)
∂z l 22( l 22 − l 21)
( a 2 − l 21)1/2
iφ
Λ∆Ψ = − 2π a ρe 2
(A4.1.22)
l 22( l 22 − l 21)
 4[( l 21 + 2 a 2)( a 2 − l 21)1/2 − 2 a 3] a l ( a 2 − l 21)1/2 

Λ3Ψ = −2πe3 i φ
1
+ 
3ρ 3
l ( l 22 − l 21)
 2 
(A4.1.23)
za [ l 41 + a (2 a + 2 z − 3ρ )]
2 2 2 2
∂4Ψ
= −2π (A4.1.24)
∂ z4 ( l 22 − l 21)3( l 22 − a 2)1/2
l ( l 22 − a 2)1/2
∂Ψ 4 1
3 = −2π [ a 2(4 l 42 − 5ρ2) + l 41] (A4.1.25)
∂ρ∂ z l ( l 22 − l 21)3
2
a 2(6 l 22 − 2 l 21 + ρ2) − 5 l 42
∂2 2 2 iφ 3
2 Λ Ψ = 2πρ e
2
a z (A4.1.26)
∂z l 42( l 22 − l 21)3( l 22 − a 2)1/2
The following identities were used in the derivation of (A4.1.1−A4.1.26):
l l = a ρ, l 21 + l 22 = a 2 + ρ2 + z 2, (A4.1.27)
1 2
( l 22 − ρ2)1/2( l 22 − a 2)1/2 = z l , ( a 2 − l 21)1/2(ρ2 − l 21)1/2 = z l ,

2 1
( a 2 − l 21)1/2( l 22 − a 2)1/2 = za , ( l 22 − ρ2)1/2(ρ2 − l 21)1/2 = z ρ.

(A4.1.28)
∂l zl ∂l zl
1 1 2 2
= −2 , = ,
∂z l 2 − l 21 ∂z l 22 − l 21
∂l al − ρl ρ( a 2 − l 21) ∂l ρl − al ρ( l 22 − a 2)
1 2 1 2 2 1
= = , = = .
∂ρ l 22 − l 21 l ( l 22 − l 21) ∂ρ l 22 − l 21 l ( l 22 − l 21)
1 2
(A4.1.29)
Appendix A4.2 311
Appendix A4.2
Here we present some indefinite integrals of expressions containing l 1 and

l 2.
l 22 − 2 a 2
⌠ ( l 2 − a 2)1/2d z = ( a 2 − l 2)1/2 ρ2
+ ln[ l 2 + ( l 22 − ρ2)1/2], (A4.2.1)
⌡ 2 1 2a 2
l 21 + 2 a 2
⌠ ( l 2 − a 2)1/2 l 2d z = −a ( a 2 − l 2)1/2 + a 2ρ2ln[ l 2 + ( l 22 − ρ2)1/2],
⌡ 2 1 1 3
(A4.2.2)
2a −
2
l 21
⌠ ( a 2 − l 2)1/2d z = ρ2 a
( l 22 − a 2)1/2 + sin-1( ), (A4.2.3)
⌡ 1 2a 2 l2
l 21(2 l 21 + 3ρ2)
⌠ ( a 2 − l 2)1/2 l 21d z = −
3 a
( l 22 − a 2)1/2 + ρ2( ρ2 − a 2) sin-1( ),
⌡ 1 8a 8 l2
(A4.2.4)
l 21 2 2
4a + 3 l 21
− l 21)1/2 1 − ,
8a
⌠ ( l 2 − a 2)1/2 2
−
2 dz = a(a (A4.2.5)
⌡ 2
l2  15ρ2 15 l 22 
( a 2 − l 21)1/2
⌠ a
d z = −sin-1( ), (A4.2.6)
⌡ l 22 − l 21 l2
( a 2 − l 21)1/2 a ( l 22 − a 2)1/2
1 
⌠ − sin-1( ) ,
a
dz = (A4.2.7)
⌡ l 22( l 22 − l 1)
2
2a2  l 22 l2 
⌠ sin-1( a ) d z = zsin-1( a ) − ( a 2 − l 2)1/2 + a ln[ l + ( l 2 − ρ2)1/2], (A4.2.8)

⌡ l2 l2 1 2 2
2 a 2 + l 21
⌠ zsin-1( a ) d z = 1 (2 a 2 + 2 z2 + ρ2) sin-1( a ) + ( l 2 − a 2)1/2 ,
⌡ l2 4 l2 2 4a
(A4.2.9)
⌠ z sin ( a ) d z = 1 z3sin-1( a ) + 1 ( a 2 − l 2)1/2(3 l 2 + 6ρ2 + 8 a 2 − 2 l 2)
2 -1
⌡ l2 3 l2 18 1 2 1
1
− a (3ρ2 + 2 a 2) ln[ l 2 + ( l 22 − ρ2)1/2]. (A4.2.10)
6
The integration in (A4.2.1−A4.2.10) was performed by parts, with a consequent

change of variables: z =( a 2 − l 21)1/2(ρ2 − l 21)1/2/ l 1 or z =( l 22 − a 2)1/2( l 22 − ρ2)1/2/ l 2.
z (2 a 2 − l 2)
⌠ ρ sin-1( a ) dρ = ρ sin-1( a ) +
2 1
, (A4.2.11)
⌡ l2 2 l2 2( a − l 1)1/2
2 2
z ρ(2 a 2 − l 2)
⌠ ρ2sin-1( a )dρ = ρ sin-1( a ) +
3 1
⌡ l2 3 l2 6( a − l 1)
2 2 1/2
l2 l1
1 1
+ a ( a − 3 z ) cosh 2
2 2 -1
− z (3 a − z )sin 2
2 2 -1
,
6 ( a + z 2)1/2 6 ( a + z 2)1/2
(A4.2.12)
⌠ a sin-1( a ) d a = 1(2 a 2 + 2 z2 − ρ2)sin-1( a )
⌡ l2 4 l2
+ l 1(ρ2 − l 21)1/2 − 2 z ( a 2 − l 21)1/2. (A4.2.13)


The integration in (A4.2.11−A4.2.12) was performed by parts, with a consequent
change of variables: ρ= y [1+ z 2/( a 2− y 2)]1/2, which corresponds to the substitution
l 2= y . A similar remark is valid for formula (A4.2.13).
Appendix A4.3
Two important integrals are computed here. The first integral is
2π a iφ iψ ( a 2 − r 2)1/2( a 2 − ρ20)1/2
1
⌠ ⌠ ρe − re tan 
-1  r d r dψ . (A4.3.1)
2π ⌡ ⌡ R 3( M,N)  a R ( N,N0)  R ( N,N0)
0 0
The integral (A4.3.1) can be computed indirectly by using (40), which leads to
an equivalent expression:
Appendix A4.3 313
⌠ Λ 1
tan-1
h  d z (A4.3.2)
⌡  R ( M,N 0
)  R ( M,N 0
) 
∞
Let us make use of the following identities:
φ
h ρei
Λh = − , (A4.3.3)
l 22 − l 21
Λ tan-1 
1 h
 R ( M,N0)  R ( M,N0) 
iφ iφ
iφ iφ
ρe − ρ0e 0
iφ ρe − ρ0e 0
= − tan-1  − 2
h h  ρe + .
R 30 R 0 R 0 + h 2 l 22 − l 21 R 20 
(A4.3.4)
The notation R in this Appendix is used as a contraction for R ( M,N ). The
0 0
substitution of (A4.3.4) in (A4.3.2) yields, after integration by parts:
⌠ Λ 1
tan-1
h  d z
⌡  R ( M,N
0
)  R ( M,N
0
) 
∞
= −
1  z tan-1 h − ⌠ z ∂ tan-1 h  d z
φ
-i φ  R0 R ⌡ R 0 ∂ z R 
ρe-i − ρ0e 0 0 0
∞
iφ
φ
z
ρe iφ ρei − ρ0e 0
− ⌠  +  hdz . (A4.3.5)
⌡  l2 − l1  R 20 + h 2
2 2
R 20
∞
The following identities are to be used now:
∂h h (ρ2 − l 21)
= , (A4.3.6)
∂z z ( l 22 − l 21)
∂ hR 0 ρ2 − l 21 z2 
-1  
h
tan = − 2 . (A4.3.7)
∂z R 0 z ( R 20 + h 2)  l 22 − l 21 R0 
The substitution of (A4.3.7) in (A4.3.5) allows us to proceed:
 z -1 
h
z
h 
ρ2 − l 21 z2  
−
1
 R tan R  − ⌠ − d z 
-i φ
-i φ
⌡ R 20 + h 2  l 22 − l 21 R 20 
ρe − ρ0e 0  0 0
∞ 
iφ
iφ
z iφ
ρe ρe − ρ0e 0
− ⌠   hdz tan-1 
z h
+ = −
⌡  l2 − l1  R 20 + h 2 -i φ R 0
2 2
R 20 -i φ
∞ (ρe − ρe 0
)R
0 0
z
ρ2 − l 21
+ ⌠
h dz  − ρe
iφ 
⌡ ( R 0 + h )( l 2 − l 1)  -i φ 
2 2 2 2
-i φ
∞ ρe − ρe 0
0
z
z2 iφ
− ⌠  hdz
+ ρe
iφ
− ρ e 0 
⌡ R 0( R 0 + h )  ρe-iφ − ρ e 0 φ 
2 2 2 -i 0
∞
0
i(φ-φ )
 z z ρρ e 0 − l 21 hdz 
-1   
1 h
⌠
0
=  R
− tan
R 0
+ − 1
 R 20 + h 2
ρe
-i φ
− ρe
-i φ
0  0 ⌡  l 22 − l 21
∞
0
i(φ-φ )
 z ρρ e 0 − l 22
 h d z 
z
− R tan R  + ⌠ 
1 -1
h 0
=
-i φ
-i φ
 ⌡ l 22 − l 21  R 20 + h 2
ρe − ρe 0 0 0
∞
0
(A4.3.8)
Taking into consideration the identity
i(φ-φ0) -i(φ-φ0)
R 20 + h 2 = ( l 22 − ρρ e ) ( l 22 − ρρ e )/ l 22 , (A4.3.9)
0 0
the integral in (A4.3.8) can be transformed as follows:

Appendix A4.3 315
h l 22 d z ( a 2 − ρ20)1/2 ( a 2 − l 21)1/2 l 22 d l
⌠ = ⌠ 2
⌡ ( l 2 − l 2)( l 2 − ρρ e-i(φ-φ0) ) a ⌡ z l ( l 2 − ρρ e-i(φ-φ0) )

2 1 2 0 2 2 0
l dl ( a 2 − ρ20)1/2 ( l 22 − a 2)1/2
= (a 2
− ρ20)1/2 ⌠ 2 2
= -1
tan ,
⌡ ( l 2 − a 2)1/2( l 2 − ρρ e-i(φ-φ0) ) s s
2 2 0
(A4.3.10)
-i(φ-φ0) 1/2
where s =( a 2 − ρρ e ) . Finally, formulae (A4.3.8) and (A4.3.10) allow us
0
to compute the original integral (A4.3.1):
2π a φ ψ
ρe − r e ( a 2 − r 2)1/2( a 2 − ρ20)1/2
 r d r dψ
i i
tan 
1
⌠ ⌠ -1
2π ⌡ ⌡ R 3( M,N )  a R ( N,N )
0
 R ( N,N0)
0 0
( a 2 − ρ20)1/2
=
1  tan-1
s
−
z
tan-1
h  .
φ
-i φ  s ( l 22 − a 2)1/2 R ( M,N ) R ( M,N )
ρe − ρe
-i 0 0 0
0
(A4.3.11)
It is reminded that h is defined by (4.1.18), and R ( M,N )=[ρ 2
+ ρ20 −
0
2ρρ cos(φ−φ ) + z 2]1/2. The right-hand side in (A4.3.11) simplifies in the
0 0
limiting case of ρ →ρ and φ →φ, namely,
0 0
φ ( l 22 − a 2)1/2
ρei  1 ( a 2
− ρ2 1/2
)  .
tan-1 2 2 1/2 −
2( a 2
− ρ) 2 1/2
 ( a 2 − ρ2)1/2 (l2 − a ) l2 − ρ
2 2

(A4.3.12)
The second integral to be computed is:
z  0 R
I2 = ⌠ + tan-1  d z .
h
(A4.3.13)
⌡ R0  h
3
R 0
We proceed with integration by parts. The result is
I2 = ⌠ tan-1  + ⌠ tan-1  .
zdz 1 h dz d h
− (A4.3.14)
⌡ R0h
2 R0 R 0 ⌡ R0 dz R 0
We modify (A4.3.7) as follows:
R0 h (ρ2 − l 21)
-1 h  
d z z
tan = 2 + −
dz R 0 R 0 + h 2  z ( l 22 − l 21) h hR
0
R 0( l 22 − a 2)1/2( l 42 − ρ2ρ20) z
= − . (A4.3.15)
(a −
2
ρ20)1/2( l 22 − l 21) l 22( R 20 + h)2 hR 0
By substituting (A4.3.15) in (A4.3.14), and taking into consideration (A4.3.9) and

(A4.1.29), we obtain
1 -1 h 
I = − tan
2 R
0
R  0
( l 42 − ρ2ρ20) d l
+
1 ⌠ 2
( a 2 − ρ20)1/2 ⌡ ( l 2 − ρ2)1/2( l 2 − ρρ ei(φ-φ0) )( l 2 − ρρ e-i(φ-φ0) )

2 2 0 2 0
(A4.3.16)
The integral in (A4.3.16) is elementary, i.e.

tan-1  +
1 h 1
= −  ln[ l 2 + ( l 2 − ρ ) ]
2 2 1/2
I
2 R0 R 0 ( a 2 − ρ20)1/2

1  a (ζ − 1)1/2 1 -1
a (ζ − 1)1/2 ρ i(φ-φ0)
− − 2 1/2 , ζ = ρ e
-1
1/2tan 1/2tan
(ζ − 1) ( a − l 1) 
2 2 1/2
(ζ − 1) ( a − l 1) 
2
0
(A4.3.17)
Since the integration was indefinite, we might have lost a function of the
variables, other than z . This function can be found from the condition that the
result of integration should not have a logarithmic singularity at ρ=0 or at q =0.
The functions eliminating such a singularity are tan-1[(ζ − 1)1/2] and tan-1[(ζ −
1)1/2]. The final result can now be represented in the form
R
⌠ z  0 + tan-1 h  d z
⌡ R 30  h R 0
Appendix A4.4 317

tan-1  +
1 h 1
= −  ln[ l 2 + ( l 2 − ρ ) ]
2 2 1/2
R0 R 0 ( a 2 − ρ20)1/2

 1  -1
a (ζ − 1)1/2 1/2
 ρ i(φ-φ0)
− 2ℜ  tan − tan-1
( ζ − 1) , ζ = e
(ζ − 1)1/2 ( a 2 − l 21)1/2  ρ0
 
The last expression proves the correctness of formula (5.1.13).
Appendix A4.4
Some integrals related to the problem of a penny-shaped crack under shear
loading are presented here, without derivation. The first integral, which can be
computed directly, is:
i (ψ-φ0)
2π a (3 a 2 − r ρ0e ) ( a 2 − r 2)1/2( a 2 − ρ20)1/2
⌠ ⌠ r d r dψ
⌡ ⌡ ( a 2 − rρ ei (ψ-φ0))2 a R ( M,N )
0 0 0
( a 2 − ρ20)1/2  2 ( l 22 − a 2)1/2[ l 21(4 − t ) − 3 a 2]

ρ -1 a
= π  t sin (l ) +
a3 a (1 − t )2
 2
1/2 
1  3z2 ρ2  -1 a (1 − t ) 
( l 22 − a 2)1/2
+ + a (3 − 2 t ) −
2
tan .
(1 − t )3/2  1 − t t 

(A4.4.1)
Here t is defined by (4.4.16). Application of the operator Λ to both sides of
(A4.4.1) yields
i (ψ-φ0)
2π a (3 a 2 − r ρ0e ) ( a 2 − r 2)1/2( a 2 − ρ20)1/2
⌠ ⌠ (ρei φ − r ei ψ) r d r dψ
⌡ ⌡ ( a 2 − rρ ei (ψ-φ0))2 3
a R ( M,N)
0 0 0
( a 2 − ρ20)1/2  a ( l 22 − a 2)1/2
iφ 1 -1 a
= −2πρe  t sin (l ) + i(φ-φ )
a3 (1 − t )( l 22 − ρρ0e 0 )
 2
1 -1 a (1 − t ) 
1/2 
( l 22 − a 2)1/2 
− tan . (A4.4.2)
t (1 − t )3/2

Another application of Λ to both sides of (A4.4.2) gives
i (ψ-φ0)
2π a (3 a 2 − r ρ e ) ( a 2 − r 2)1/2( a 2 − ρ2)1/2
⌠ ⌠ 0 0
3(ρei φ − r ei ψ)2 r d r dψ
⌡ ⌡ ( a 2 − rρ ei (ψ-φ0))2 5
a R ( M,N )
0 0 0
ρ2e2 i φ( a 2 − ρ20)1/2( l 22 − a 2)1/2(3 l 22 − ρρ e

i(φ-φ0)
)
0
= 2π i(φ-φ0) 2 . (A4.4.3)
l 22( l 22 − l 21)( l 22 − ρρ e )
0
Differentiation with respect to z of both sides of (A4.4.1) results in
i (ψ-φ0)
2π a (3 a 2 − r ρ e ) z ( a 2 − r 2)1/2( a 2 − ρ2)1/2
⌠ ⌠ 0 0
r d r dψ
⌡ ⌡ ( a 2 − rρ ei (ψ-φ0))2 a R 3( M,N )
0 0 0
ha 2 3 3( l 22 − a 2)1/2
tan-1 2 .
t s
= 2π − − (A4.4.4)
s  s
2 2
l 22 − a t 2
s 3
( l 2 − a 2)1/2
Here h is defined by (4.1.18). Another differentiation of both sides of (A4.4.4)

with respect to z yields
i (ψ-φ0)
2π a (3 a 2 − r ρ e ) ( a 2 − r 2)1/2( a 2 − ρ2)1/2
⌠ ⌠ 0 0
1 − 3 z2  rd rdψ
⌡ ⌡ ( a 2 − rρ ei (ψ-φ0))2 a R 3( M,N )  R 2( M,N )
0 0 0
( a 2 − ρ20)1/2  a ( l 22 − a 2)1/2 3( l 22 − l 21 t )
= 2π  2 
 ( l 2 − l 1)( l 2 − ρρ0e ) 
a (1 − t )
3 2 2 i(φ-φ0) 1 − t
i(φ-φ0)
ρρ e (2 l 22 + l 21 t − 3ρ2) 
0
 − 3 -1 a (1 − t ) 
1/2
( l 22 − a 2)1/2 
+ tan .
l 22 − ρρ e
i(φ-φ0)
 (1 − t )3/2
0
Appendix A4.4 319
(A4.4.5)
Application of the operator Λ to both sides of (A4.4.4) yields
i (ψ-φ0)
2π a (3 a 2 − r ρ0e ) z ( a 2 − r 2)1/2( a 2 − ρ20)1/2
⌠ ⌠ 3(ρe
iφ
− r ei ψ) r d r dψ
⌡ ⌡ ( a 2 − rρ ei (ψ-φ0))2 5
a R ( M,N )
0 0 0
h ρei φ(3 l 22 − a 2 t )
= 2π . (A4.4.6)
( l 22 − l 21)( l 22 − a 2 t )2
A different result is obtained if Λ is applied to a complex conjugate of

expression (A4.4.4), namely,
- i (ψ-φ0)
2π a (3 a 2 − r ρ0e ) z ( a 2 − r 2)1/2( a 2 − ρ20)1/2
⌠ ⌠ 3(ρei φ − r ei ψ) r d r dψ
⌡ ⌡ ( a 2 − rρ e- i (ψ-φ0))2 a R 5( M,N )
0 0 0
a 2 i φ 15( l 2 − a )
0
2 2 1/2
-1
s  − 15
= 2π h  2 ρ e tan
s
0  s5 ( l 22 − a 2)1/2 s4
ρei φ(3 l 22 − a 2 t ) 
+
5
+
2t  +  . (A4.4.7)
s 2( l 22 − a 2 t ) ( l 22 − a 2 t )2 ( l 22 − l 21)( l 22 − a 2 t )2

Integration of both sides of (A4.4.1) with respect to z gives
i (ψ-φ0)
2π a (3 a 2 − r ρ0e ) ( a 2 − r 2)1/2( a 2 − ρ20)1/2
⌠ ⌠ ln[ R ( M,N ) + z ] r d r dψ
⌡ ⌡ ( a 2 − rρ ei (ψ-φ0))2 a
0 0 0
 ( a 2 − l 21)1/2
 z2
= π( a − ρ20)1/2  − ρ − − + ρ2
2 2 2 1/2
2ln[ l + ( l ) ] 2 +
2 2
a (1 − t ) 
3 1 − t

( l 1 + 2 a 2) +  z2 + a 2(3 − 2 t − ζ)

1 2 zζ a z
− sin-1( ) + 3
3  a l2 a (1 − t )3/2  1 − t 
( a 2 − l 21)1/2 
×tan-1 2  + 2(ζ − 1)1/2tan-1  − tan-1  .
a (1 − t ) 1/2 1
( l 2 − a 2)1/2  (ζ − 1)1/2 a (ζ − 1)1/2 
(A4.4.8)
Here ζ is defined by (A4.3.17), and the bar, as usual, indicates the complex
conjugate value. A similar integration with respect to z of (A4.4.3) yields
i (ψ-φ0)
2π a (3 a 2 − r ρ0e ) ( a 2 − r 2)1/2( a 2 − ρ2)1/2
⌠ ⌠ 0
Λ2{ln[ R ( M,N) + z ]} r d r dψ
⌡ ⌡ ( a 2 − rρ ei (ψ-φ0))2 a
0 0 0
2π  (ζ − 1)1/2  -1 1  − tan 

( a 2 − l 21)1/2

= ( a − ρ0)  −
2 2 1/2
tan -1
q q  (ζ − 1)1/2 a (ζ − 1)1/2
ei φ ( a − l 1)   − 1  .
2 2 1/2
ρ2
+ 1 + (A4.4.9)
ρ a  l 22 − ρρ0e 0 
i(φ-φ )

It is reminded that q is defined by (4.1.28), and
2 R ( M,N) + z
Λ2ln[ R ( M,N) + z ] = −(ρei φ − r ei ψ)2
R 3( M,N) [ R ( M,N) + z ]2
(A4.4.10)
Yet another application of the operator Λ to (A4.4.9) gives
i (ψ-φ0)
2π a (3 a 2 − r ρ e ) ( a 2 − r 2)1/2( a 2 − ρ20)1/2
⌠ ⌠ 0
Λ3{ln[ R ( M,N ) + z ]} r d r dψ
⌡ ⌡ ( a 2 − rρ ei (ψ-φ0))2 a
0 0 0
2π  3(ζ − 1)1/2  -1 1  -1

( a 2 − l 21)1/2

= ( a − ρ0) 
2 2 1/2
tan − tan
q q 2

 (ζ − 1) 
1/2
a (ζ − 1)  1/2
e i φ( a 2 − l 21)1/2 l 22 + ρ2 2ρ2( l 22 − a 2)
2
−  + 
i(φ-φ0) 2 + 1
a ( l 22 − l 21)  l 22 − ρρ ei(φ-φ0) ( l 2 − ρρ0e
2
) 
0
Appendix A4.4 321
ei φ 3 2ei φ ( a 2 − l 21)1/2

l 22 + 2ρ2
1 ei φ 
ρ 
+ + − + 2( + ) .
ρ q ρ a  q ( l 22 − ρρ ei(φ-φ0) ) q
0 
(A4.4.11)
Here
iφ iψ 3
8 R 2( M,N ) + 9 R ( M,N ) z + 3 z 2
Λ ln[ R ( M,N ) + z ] = (ρe
3
− re ) .
R 5( M,N ) [ R ( M,N ) + z ]3
(A4.4.12)
Formula (4.1.27) can be used to obtain some additional results. Integration of
(4.1.27) with respect to z gives
2π a
ρei φ − r ei ψ ( a 2 − r 2)1/2( a 2 − ρ20)1/2 r d r dψ
⌠ ⌠ tan-1 
⌡ ⌡ R ( M,N )[ R ( M,N ) + z ]  a R ( N,N
0
)  R ( N,N0)
0 0
2π
 R 0tan (R ) − ( a − ρ0)  tan  2 
-1 h 2 2 1/2 z -1 s
=
q
 0
s ( l 2 − a 2 1/2
) 
( a 2 − l 21)1/2 
tan-1 1 
− (ζ − 1) 1/2 − tan 1/2  .
1/2 -1
(A4.4.13)
 (ζ − 1) a (ζ − 1) 

The following indefinite integrals were used here
⌠ tan-1 s  d z = z tan-1 s 
⌡ ( l 2 − a ) 
2 2 1/2
( l 2 − a 2)1/2
2
+ s ln[ l + ( l 22 − ρ2)1/2] + (ζ − 1)1/2 tan-1 2

a (ζ − 1)1/2
,
 2 ( a − l 21)1/2
⌠ z tan-1 h  d z = R tan-1 h  − ( a 2 − ρ2)1/2−ln[ l + ( l 2 − ρ2)1/2]

⌡ R0 R 0 0 R 0 0
 2 2
( a 2 − l 2)1/2 ( a 2 − l 2)1/2
+ (ζ − 1) tan   + (ζ − 1)1/2tan-1  .
1 1
1/2 -1
a (ζ − 1) 
1/2
a (ζ − 1) 
1/2
(A4.4.14)
By applying the operator Λ to both sides of (A4.4.13), one gets
2π a iφ iψ
(ρe − r e )2[2 R ( M,N ) + z ] ( a 2 − r 2)1/2( a 2 − ρ20)1/2
⌠ ⌠ tan -1  rd rdψ
⌡ ⌡ 3
R ( M,N ) [ R ( M,N ) + z ] 2
 a R ( N,N0)  R ( N,N0)
0 0
iφ
z ρ0e
0
2π 0
R 2 + z2
2 1/2
-1 h
2  -1 s 
 tan ( ) + ( a − ρ0) −
2
= tan
q R0q

R0  s  s2 q  ( l 2 − a ) 
2 2 1/2
e ha 2 
(ζ − 1)1/2  -1 ( a 2 − l 21)1/2 iφ iφ
+ tan
1
− tan-1  +
e 
− ,
q  (ζ − 1)1/2 a (ζ − 1)1/2 ρ ρs2

(A4.4.15)
and yet another application of Λ to (A4.4.15) yields
2π a
iφ iψ 3
⌠ ⌠ (ρe − re ) [8 R ( M,N) + 9 R ( M,N) z + 3 z ]
2 2
⌡ ⌡ R 5( M,N ) [ R ( M,N ) + z ]3
0 0
( a 2 − r 2)1/2( a 2 − ρ20)1/2
× tan -1  rd rdψ
 a R ( N,N0)  R ( N,N0)
iφ iφ
2π
3 R 40 + 6 R 20 z 2 − z 4 e 2e 3
h
 tan ( ) + ( a 2 − ρ20)1/2
-1
= +
q R 30 q 2 R0 ρ ρ q

iφ 2 iφ
4ρ0e 0
3ρ20e 0
− ( a 2 − ρ20)1/2
z 8  tan-1 s 
− +
 s q 2 2
s q s 4
 ( l 2 − a 2)1/2
2
3(ζ − 1)1/2  -1 ( a 2 − l 21)1/2

−
1 
1/2 − tan
-1
tan
q 2
 (ζ − 1) a (ζ − 1)  1/2
iφ iφ
2 iφ 2ρ0e 0
iφ ρ0e 0
2 ( l 2 − a ) t 
2 2
+
ha e  −
2e
− +  2
2
−
ρs2  s2 ρ q  s q  l 22 − a 2 t 
Appendix A4.4 323
3 iφ ei φ( l 22 − ρ2) 
− 2
h  q ρe + −
z2 q
+ 2e2 i φ.
R 0 + h l 22 − l 21
2
ρq R 20 q 
(A4.4.16)
CHAPTER 5
APPLICATION TO CONTACT PROBLEMS
The various elastic contact problems, solved in Chapters 2 and 3, deal

primarily with the stresses and displacements in the plane z =0. We concentrate
here on the complete solutions. The solution is called complete when explicit
expressions are given for the field of stresses and displacements in the whole
half-space. The complete solution, combined with the reciprocal theorem, enables
us to solve more complicated problems of punch interactions, influence of
external loads on punches, etc. An approximate analytical solution is given to
the non-classical punch problem. Significant part of the material presented here
is as yet unpublished. The rest follows the papers (Fabrikant, 1974a, 1986a,
1986c, 1986i, 1987h).
5.1 Contact problem for a smooth punch.
A smooth punch is pressed against a transversely isotropic elastic half-space

z ≥0 by a normal force P . The term ‘smooth’ is used to identify a punch
which does not exert any shear traction at its base. Let S denote the domain
of contact. The mixed boundary conditions on the plane z =0 are:
σz = 0, for (ρ,φ)∉ S ; w = ω(ρ,φ), for (ρ,φ)∈ S ,
τz = 0, for −∞<( x,y)<∞. (5.1.1)

As in the previous chapter, we may assume again that
F 1( z ) = c 1 F ( z 1), F 2( z ) = c 2 F ( z 2), F 3( z ) = 0. (5.1.2)
Substitution of (5.1.2) and (2.1.12) in the third condition (5.1.1) yields:
324
c = − c γ 1/ m 1γ 2. (5.1.3)
1 2
Now we need to define the main potential function F so that its second
z -derivative vanishes at z =0 outside the domain of contact. Comparison with
(4.1.4) gives
F (ρ,φ, z ) = F ( M ) = ⌠ ⌠ ln[ R ( M,N ) + z ] σ( N ) d S . (5.1.4)

⌡⌡ N
S
The simple layer potential property yields inside the domain of contact:
∂2 F
= −2πσ, (5.1.5)
∂ z 2 z=0
which, combined with (2.1.12), gives the second equation for the constants c and
1
c
2
2π A 44[(1 + m 1) c + (1 + m 2) c ] = 1. (5.1.6)
1 2
Equations (5.1.3) and (5.1.6) determine the constants
c = H γ1/( m 1 − 1), c = H γ2/( m 2 − 1). (5.1.7)

1 2
The simplifications made in (5.1.7) are due to the properties (4.1.10).

Finally, substitution of (5.1.2), (5.1.4) and (5.1.7) in the second of equations
(2.1.6) yields for z =0 the well known governing integral equation of elastic
contact problem for a smooth punch:
ω( N 0) = H ⌠ ⌠
σ( N )
dS . (5.1.8)
⌡ ⌡ R ( N,N0) N
S
An approximate analytical solution of (5.1.8) for a punch of arbitrary shape will

be given later. We consider in more detail the case of a circular punch. The
integral equation (5.1.8) can be rewritten for a circular domain of contact of
radius a as follows (c.f. 1.4.5)
ρ a
ρ dρ 2
4H ⌠ 2
dx
⌠
0 0
L  x  σ(ρ ,φ) = ω(ρ,φ)
⌡ (ρ − x ) ⌡ (ρ0 − x )
2 1/2 2 2 1/2
ρρ0 0
0 x
326 CHAPTER 5 APPLICATION TO CONTACT PROBLEMS
Its exact closed form solution is (c.f. 1.4.10)
a
1 d ⌠ xdx
σ(ρ,φ) = − 2 L (ρ)
π Hρ dρ ⌡ ( x − ρ2)1/2
2
ρ0dρ0
x
× L  2
1 d⌠
L (ρ0) ω(ρ0,φ).
x  d x⌡ ( x 2 − ρ20)1/2
0
(5.1.9)
The potential function F can be found in two stages. First of all, one has
from (5.1.4):
2π a
σ( r ,ψ) r d r dψ
= ⌠⌠
∂F
(5.1.10)
∂z ⌡ ⌡ [ρ2 + r2 − 2 rρcos(φ−ψ) + z2]1/2
0 0
Substitution of (5.1.9) in (5.1.10) yields, after integration (c.f. 1.4.21)
2π a
R0
= 2 ⌠⌠ + tan-1  3 ω(ρ0,φ0)ρ0dρ0dφ0.
∂F 1 h z
(5.1.11)
∂z π H ⌡ ⌡ h R 0 R 0
0 0
Here R 0=[ρ2 + ρ20 − 2ρρ0cos(φ−φ0) + z 2]1/2, and h is defined by (4.1.18). The

next integration of (5.1.11) with respect to z gives the expression for the
potential function directly in terms of the prescribed displacement under the
punch, namely,
2π a
1 ⌠⌠
F (ρ,φ, z ) = K(ρ,φ, z ; ρ0,φ0) ω(ρ0,φ0)ρ0dρ0dφ0, (5.1.12)
πH ⌡⌡2
0 0
where
1 -1 h  1 
K(ρ,φ, z ; ρ0,φ0) = − tan
R 0
+  ln[ l 2 + ( l 22 − ρ2)1/2]
R0 ( a − ρ0)
2 2 1/2

 1  -1
a (ζ − 1)1/2 1/2
 ρ i(φ-φ0)
− 2ℜ  tan − tan-1
( ζ − 1) , ζ = e
(ζ − 1)1/2 ( a 2 − l 21)1/2  ρ0
 
(5.1.13)
The details of integration are presented in Appendix A4.3. Now all the Green’s
functions related to the field of displacements and stresses can be obtained in
elementary functions from (5.1.13) by differentiation
R0
= 3  + tan-1  ,
∂K z h
(5.1.14)
∂z R0  h R 0
1 
q  R0 ( a 2 − l 21)1/2
-1 
h
ΛK = 3 + tan − 1 −
R0  h R 0 hq a

(a −
2
l 21)1/2
 -1
a (ζ − 1)1/2 1/2


− tan − tan-1
( ζ − 1) , (5.1.15)
a (ζ − 1)1/2 ( a 2 − l 21)1/2
1 3 z 2R 0 z2 ρ2 − l 21
-1   ,
∂2K h 1
2 = 3 1 − + tan + − 2 (5.1.16)
∂z R 0 R 0  h
2
R 0 h ( R 0 + h )R 0
2 2 2
l2 − l1 
2
∂K 3 zq R 0 -1 
h z  ρei φ q 
Λ = − 5 + tan + + , (5.1.17)
∂z R0  h R 0 h ( R 20 + h 2)  l 22 − l 21 R 20 
R l 22 − ρ2
3q2  0 -1   − e2 i φ
h 1 q 2
ΛK = − 5
2
+ tan +
R0  h R 0 h ( R 0 + h )R 0
2 2 2
l2 − l1
2 2

3 1 1  1  -1
a (ζ − 1)1/2 1/2

+ 2 −  
1 + tan − tan-1
( ζ − 1)
q h ( a 2 − ρ20)1/2 (ζ − 1)1/2 ( a 2 − l 21)1/2
  
ei φ 1 − a 3ζ .
+ (5.1.18)
ρ q ( a 2 − ρ20)1/2 ( a 2 − l 21)1/2( a 2ζ − l 21)
The following identities were used in the simplification of (5.1.14−5.1.18):
( l 22 − ρ20)( a 2 − l 21) + a 2 qq = a 2( R 20 + h 2),
( a 2ζ − l 21)( a 2ζ − l 21) = ( R 20 + h 2) l 21 a 2/ρ20,

( l 22 − ρ20)( qq + z 2) − qq( a 2 − ρ20) = ( l 22 − a 2)( R 20 + h 2).

(5.1.19)
Formulae (5.1.13−5.1.18) are the main new results of this section. They give a
complete solution for the contact problem under consideration when combined
with formulae (2.1.6), (2.1.12), (5.1.2) and (5.1.7). When the prescribed
displacement can be expressed as a sum of powers in x and y , the complete
solution is elementary. Some particular examples will be considered further.
Exercise 5.1
1. Derive 5.1.13.
2. Verify (5.1.14) −(5.1.18).
3. Prove the identities (5.1.19).
4. Derive the general solution for the case of isotropy.
5.2 Flat centrally loaded circular punch.
We consider a transversely isotropic elastic half-space z ≥0, indented by a

rigid circular punch of radius a . The punch loading is statically equivalent to a
centrally applied normal force P . Denote the punch settlement by ω=const.
Solution of the integral equation (5.1.8) is given by (5.1.9), and in this particular
case takes the form
ω
σ(ρ,φ) = . (5.2.1)
π H(a
2 2
− ρ2)1/2
Integration of the last expression over the circle ρ≤ a relates the punch settlement
ω to the total force P as
2ω a
P = .
πH
The substitution of (5.2.1) in (5.1.4) leads to an integral which can be evaluated

by differentiation of the expression for the main potential function (A4.1.1) with
respect to a . The result is
2ω 
z sin-1( ) − ( a 2 − l 21)1/2 + a ln[ l 2 + ( l 22 − ρ2)1/2].
a
F = (5.2.2)
πH  l2 
5.2 Flat centrally loaded circular punch. 329
The appropriate differentiation of the potential function (5.2.2) gives the complete
solution:
2ω a ei φ γk ( a 2 − l 21k)1/2
2
u =
πρ Σm k=1
k
1 −
− 1 a
,

(5.2.3)
2
mk
w =
2ω
π Σm
k=1
k − 1
sin-1(
a
l 2k
), (5.2.4)
4ω A 66 2
γ2k − ( m k + 1)γ23 ( a 2 − l 21k)1/2
σ1 = −
π Σ k=1
γk( m k − 1) l 22k − l 21k
, (5.2.5)
4ω A 66e2 i φ 2
( a 2 − l 21k)1/2
γk
σ2 =
π Σk=1
( m k − 1)  l 22k − l 21k

2a ( a 2 − l 21k)1/2 
− 2 1 − ,
ρ  a  
(5.2.6)
ω
2
(a −2
l 21k)1/2
σz =
π2 H (γ1 − γ2) Σ k=1
(−1)kγk
l 22k − l 21k
, (5.2.7)
ωei φ l 1k( l 22k − a 2)1/2

2
τz =
π2 H (γ1 − γ2) Σ(−1)
k=1
k
l 2k( l 22k − l 21k)
. (5.2.8)
This problem was first solved by Elliott (1949). His solution is essentially in
agreement with ours. The fact that he uses the notation ρ as an elastic
parameter and as a polar radius simultaneously, somewhat complicates the
comparison.
k = lim{( a − ρ)1/2σz}, for z =0. (5.2.9)

1
ρ→ a

ω P a 1/2
k = − = − ( ) . (5.2.10)
1 π H √2 a
2 2π2
The asymptotic behavior of the field of stresses and displacements near the punch
edge can be derived by substitution of (4.7.1) and (5.2.10) in (5.2.3−5.2.8). The
result is:
γ1 S 1 γ2 S 2
u = u n = 2π Hk 1√2 r  +  + 0(1), (5.2.11)
m 1 − 1 m 2 − 1
m1T1 m2T2
w = −2π Hk 1√2 r  +  + 0(1), (5.2.12)
m 1 − 1 m 2 − 1
2
γ2k − ( m k + 1)γ23 S k
2
σ1 = 2π A 66 Hk 1( )1/2
r Σ
k=1
γk( m k − 1) Qk
+ 0(1), (5.2.13)
2
γk S k
2
σ2 = −2π A 66 Hk 1( )1/2
r Σ(m
k=1
k − 1) Q k
+ 0(√ r ), (5.2.14)
k1 γS γS
σz =  1 1 − 2 2 + 0(1), (5.2.15)
√2 r (γ1 − γ2)  Q 1 Q2 
k1 T T
τz = τzn = −  1 − 2  + 0(√r), (5.2.16)
√2 r (γ1 − γ2) Q 1 Q 2
Comparison of (5.2.11−5.2.16) with (4.7.4−4.7.9) indicates that they become

identical if one substitutes formally θ by π−θ. This is easy to explain. The
punch problem is mathematically equivalent to an external crack problem. There
exist a notion that the asymptotic behavior of stresses and displacements near the
edge of an arbitrary flat crack with a smooth boundary is completely defined by
three stress intensity factors. This means that the asymptotics of an internal
crack and an external one should be the same. The system of local axes was
introduced in previous section so that the angle θ was measured from the
direction outside the penny-shaped crack. In the case of the punch problem the
axis On should be directed into the circle, this will make the expressions
(5.2.11−5.2.16) identical to (4.7.4−4.7.9).
5.2 Flat centrally loaded circular punch. 331
Exercise 5.2
1. Find the complete solution to the problem of a smooth flat centrally loaded
circular punch in the case of an isotropic half-space.
Answer:
ωei φ  
a − ( a 2 − l 21)1/2

z l 1( l 22 − a 2)1/2
π(1 − ν)  ,
u = −(1 − 2ν) +
  ρ  l 2( l 22 − l 21)

z ( a 2 − l 21)1/2
2ω  -1 a ,
w = sin ( ) +
π  l2 2(1 − ν)( l 22 − l 21)
2ωµ ( a 2 − l 21)1/2 z 2[ l 41 − a 2(2 a 2 − ρ2 + 2 z 2)]

σ1 = − (1+2ν) + ,
π(1 − ν)  l 22 − l 21 ( a 2 − l 21)1/2( l 22 − l 21)3 
2ωµe2 i φ  ( a 2 − l 21)1/2 2[ a − ( a 2 − l 21)1/2]

−(1 − 2ν) 2 
π(1 − ν) 
σ2 = −
  l 2 − l 21 ρ2 
az ( l 22 − a 2)1/2[2 l 41 + ρ2( l 21 + 3 l 22 − 6 a 2)]

− ,
l 22( l 22 − l 21)3

2ωµ  ( a 2 − l 21)1/2 z 2[ l 41 + a 2(ρ2 − 2 a 2 − 2 z 2)]

π(1 − ν)  ,
σz = − +
l 22 − l 21 ( a 2 − l 21)1/2( l 22 − l 21)3
 
2ωµ z ρei φ( a 2 − l 21)1/2(3 l 22 + l 21 − 4 a 2)

τz = − .
π(1 − ν) ( l 22 − l 21)3
Hint : use formulae (5.2.3−5.2.8).
2. Rewrite the result of the Exercise 1 in polar coordinates.
Answer:
ω  a − ( a 2 − l 21)1/2 z l 1( l 22 − a 2)1/2
−(1 − 2ν)  +
π(1 − ν)  ,
u =
ρ
  ρ  l 2( l 22 − l 21)

2ωµ  a − ( a 2 − l 21)1/2 ( a 2 − l 21)1/2

π(1 − ν) 
σ = (1 − 2ν) −
ρ ρ2 l 22 − l 21

az ( l 22 − a 2)1/2[ l 41 − l 42 + ρ2( l 21 + 3 l 22 − 4 a 2)]

− ,
l 22( l 22 − l 21)3

2ωµ  a − ( a 2 − l 21)1/2
π(1 − ν) 
σ = − (1 − 2ν)
φ ρ2

( a 2 − l 21)1/2[ a 2 − (1 − 2ν) l 22]

+ ,
l 22( l 22 − l 21)

2ωµ z ρ( a 2 − l 21)1/2(3 l 22 + l 21 − 4 a 2)
τ = − .
ρz π(1 − ν) ( l 22 − l 21)3
5.3 Inclined circular punch on an elastic half-space
The case of a flat circular punch pressed against a transversely isotropic

elastic half-space by a non-centrally applied force P can be considered as a
superposition of two problems: that of a centrally loaded punch, which was
considered earlier, and a punch subjected to a tilting moment M which will be
considered here. Let the displacements under the punch be
ω = bxy − byx. (5.3.1)
Here b x and b y are the tilting angles about the axes Ox and Oy respectively.
Introduce the complex tilting angle
b = b x + ib y. (5.3.2)
Expression (5.3.1) can now be rewritten as
ω(ρ,φ) = ℑ{ b ρei φ}. (5.3.3)

Here ℑ indicates the imaginary part. The substitution of (5.3.3) in (5.1.9) yields
2ℑ{ b ρei φ}
σ(ρ,φ) = . (5.3.4)
π2 H ( a 2 − ρ2)1/2
Evaluation of the potential function (5.1.4) leads to the integral
2π a i φ0
ρ0e
⌠⌠ ln( R 0 + z ) ρ0dρ0dφ0. (5.3.5)
⌡ ⌡ ( a 2 − ρ20)1/2
0 0
Taking into consideration that
iφ
ρe
= −Λ( a 2 − ρ2)1/2,
(a 2
− ρ) 2 1/2
the integral in (5.3.5) can be evaluated by parts, with the result given by
(A4.1.5). The potential function takes the form:
l 21 + 2 a 2
2ℑ{ b ρei φ} a 2 a 3
F = z sin-1 − ( a 2 − l 21)1/2(1 − ) − .
πH  l2 3ρ2 3ρ2
(5.3.6)
All the necessary derivatives are readily available from Appendix A4.1, and we
can write the complete solution:
2
 γk
u =
2i
π Σ
k=1

mk − 1 

b z ksin-1( ) − ( a 2 − l 21k)1/2
a
l 2k 
2 a 3 − ( l 21k + 2 a 2)( a 2 − l 21k)1/2

2 iφ
− be , (5.3.7)
3ρ2

2
mk a ( l 22k − a 2)1/2
2
w = ( bxy − byx)
π Σ
k=1
sin-1( a ) −
m k − 1 l 2k l 22k
,

(5.3.8)
8 A 66 2
γ2k − ( m k + 1)γ23 ( a 2 − l 21k)1/2
σ1 = −
π
( bxy − byx) a2 Σ
k=1
γk( m k −1) l 22k( l 22k − l 21k)
,
(5.3.9)
iφ
2 A 66 i e 2
 a l 1k( a 2 − l 21k)1/2
γk
σ2 =
π Σk=1
m k − 1

b
l 2k( l 22k − l 21k)
4[( l 21k + 2 a 2)( a 2 − l 21k)1/2 − 2 a 3] a l 1k( a 2 − l 21k)1/2 

− be
2 iφ  + , (5.3.10)
 3ρ3
l 2k( l 22k − l 21k) 
2( b x y − b y x ) 2
a 2( a 2 − l 21k)1/2
σz =
π2 H (γ1 − γ2) Σ(−1) γ
k=1
k
k
l 22k( l 22k − l 21k)
, (5.3.11)
  -1 a 2
a ( l 22k − a 2)1/2
τz = 2
i
π H (γ 1 − γ 2)
k=1 
Σ
 b sin (l ) −
2k l 2k − l 1k 
2 2

a l 21k( l 22k − a 2)1/2

2 iφ
+ be . (5.3.12)
l 22k( l 22k − l 21k)

Note. Some of the integrals involving special functions can now be

computed simply by comparison of the solutions obtained by the integral
transform method with the corresponding solution given by the present method.
For example, comparison of (4.1.24) with formulae (1.42) and (1.43) of (Kassir
and Sih, 1975) leads to
∞ l1
⌠ J ( as) J (ρs) e-sz d s =  2  ( a ρ)-n⌠ 2 x d x2 1/2.

2n 1/2
⌡ n+1/2 n √s πa  ⌡ (ρ − x )
0 0
We have verified numerically that the last formula is correct for non-integer n as
well. An enormous amount of material on Bessel functions exists in the
literature, so it is difficult to claim that the last result is new, but there is some
chance that it is new, since our notation l 1 and l 2 does not seem to have been
used before. Here is another example of just how useful this notation is.
Suppose that we need to compute the integral
⌠ sinax J 1(ρx ) e-zx d x2 . (5.3.13)

⌡ x
0
We could not find this integral in the tables, but we have found another one
(Gradshtein and Ryzhik, 1963, formula 6.752.2)
⌠ sinax J 1(ρx ) e-zx d x = a (1 − r), (5.3.14)

⌡ x ρ
0
where the parameter r is a positive root of the equation
ρ2 z2
a2 = − . (5.3.15)
1 − r2 r2
The original integral (5.3.13) can be computed by integration of both sides of

(5.3.14) with respect to z . We need to get an explicit expression for r from
the fourth order algebraic equation (5.3.15), substitute the result in (5.3.14) and
integrate the result with respect to z which does not seem possible at first. The
introduction of the parameters l 1 and l 2 allows us to find the positive root of
(5.3.15) in a very simple form, namely, r =( a 2 − l 21)1/2/ a . The z -integration can
be performed using (A4.2.3), and the final result is
∞
(2 a 2 − l 21)( l 22 − a 2)1/2 − 2 a 2 z
⌠ sinax J 1(ρx )e dx ρ a
-zx
= + sin-1( ).
⌡ x2 2aρ 2 l2
0
(5.3.16)
Exercise 5.3
1. Consider the interaction of a concentrated load P 0, applied at an arbitrary

point (ρ,φ, z ) in the z -direction, with a flat circular punch of radius a .
Solution :
One can deduce from (5.2.4) that the normal displacement w at the point (ρ,φ, z ),
due to a unit force applied to the punch, is
2
mk
w =
H
a Σm k=1
k − 1
sin-1(
a
l 2k
).
Application of the reciprocal theorem immediately gives the punch settlement ω

due to the load P 0
2
mk
ω =
H
P
a 0 Σm
k=1
k − 1
sin-1(
a
l 2k
).
The tilting angle δ of the punch can be obtained in a similar manner from
(5.3.8). The result is
2
mk l 1k( l 22k − a 2)1/2
Σ ρsin-1( a ) − .
3H
δ = P0
2a 3 m k − 1 l 2k l 2k 
k=1
All the other parameters of interest can be found in a similar manner.
2. Investigate the problem of interaction between a flat circular punch and an

arbitrarily located tangential force T = T x+i T y.
5.4 Flat punch of arbitrary planform under the action

of a normal centrally applied force
The general method is applied here to the non-classical punch problem. A

simple yet accurate relationship is established between the punch settlement and
the applied force for an arbitrary flat punch. Specific formulae are derived for
a punch whose planform has the shape of a polygon, a triangle, a rectangle, a
rhombus, a circular sector and a circular segment. All the formulae are checked
against the solutions known in the literature, and a good accuracy is confirmed.
Theory. The presentation in this section will be made in terms of the
elastic contact problem but one should keep in mind that all the results will be
applicable in other branches of engineering science. Here, we outline the idea of
the analytical treatment of the elastic contact problems which allows us to derive
simple yet accurate formulae for various punch shapes. The governing integral
equation is given by (5.1.8). The analytical approach is based on the integral
representation for the reciprocal distance established in (1.1.27).
Substitution of (1.1.27) into (5.1.8) gives, after changing the order of integration
5.4 Flat punch of arbitrary planform under a normal force 337
x2
λ( , φ−φ0)
ρ 2π a (φ0)
ρρ0
dx
w (ρ,φ) =
2 ⌠
H ⌠ dφ ⌠ σ(ρ0,φ0)ρ0dρ0.
π ⌡ (ρ − x )
2 2 1/2
⌡ 0 ⌡ (ρ20 − x 2)1/2
0 0 x
(5.4.1)
Consider a flat-ended punch with a planform S whose boundary is defined

in polar coordinates as
ρ = a (φ). (5.4.2)
Let the normal pressure distribution under the punch be
ca (φ)
σ = (5.4.3)
a 2(φ) − ρ2 1/2
 
where c is a constant which can be determined from the condition that the
integral of σ over S should give the total force P .
2π a (φ) 2π
⌠ dφ ⌠ ca (φ)
ρdρ = c ⌠ a 2(φ)dφ = 2 Ac = P ,
⌡ ⌡ ⌡
0 0 a 2(φ) − ρ21/2 0
 
(5.4.4)
where A is the area of S . It is noteworthy that the total force does not depend
on the location of the coordinate system origin. This location can be determined
from the condition that the stress distribution (5.4.3) should not produce any
tilting moment about the origin, which leads to the two equations
2π 2π
⌠ a 3(φ) cosφ dφ = 0 , ⌠ a 3(φ) sinφ dφ = 0.

⌡ ⌡
0 0
The left hand sides of both equations are proportional to the x and y coordinates
of the center of gravity which means that the origin of the system of polar
coordinates should be located at the center of gravity of the domain of contact
S . One gets immediately from (5.4.4) that
P a (φ)
σ = . (5.4.5)
2 A  a 2(φ) − ρ2
1/2
 
For the case of a flat punch w = δ = const. Now substituting (5.4.5) in
(5.4.1), we can verify how close to a constant will be the displacements,
produced by the traction distribution (5.4.5). Integration with respect to ρ0 gives
∞ ρ
Σ
xdx
w (ρ,φ) =
HP ⌠ (x )|n| 2
⌡ ρ (ρ − x )
2A 2 1/2
n=-∞
0
2π
F 1 −
x2 
×⌠ e
in(φ-φ0) |n| 1
, ; 1; 1 − 2 dφ . (5.4.6)
⌡  2 2 a (φ0) 0
0
Here F stands for the Gauss hypergeometric function. Further evaluation of the
normal displacements can be carried out separately for each harmonic. The
zero th harmonic has the form
2π
HP π ⌠
w0 = a (φ)dφ. (5.4.7)
2A 2 ⌡
0
It is important to note that the second harmonic is equal to zero for an arbitrary
contour, and that all the odd harmonics will be zero if the expression for a (φ)
does not contain odd harmonics. The expression for the fourth harmonic is
2π 4 i (φ-φ0)
e dφ0
HP 8 3⌠
w4 = ρ . (5.4.8)
2 A 35 ⌡ a 2(φ0)
0
The investigation of further harmonics shows that their amplitude decreases.

Now consider in more detail the case of a square of side 2 l . The
equation of the boundary in this case is a (φ)= l /cosφ for −π/4<φ<π/4, and the
pattern is repeated outside this range. We can evaluate several non-zero
harmonics:
HP HP 32ρ3cos4φ
w0 = 4π l ln(1 + √2), w4 = , (5.4.9)
2A 2A 105 l 2
ρcos8φ( )2 +
HP 64 ρ 12 ρ 4 20 ρ 6
w8 = − ( ) + ( ) ,
2 A 3465 l 13 l 13 l 
9 7
6 Γ( ) 144 Γ( )
ρcos12φ
HP 2 ρ2 2 ρ4
w 12 = ( ) + ( )
2A  Γ(17) l 19
Γ( )
l
2 2
5 7 7 11
1200 Γ( ) Γ( ) 1680 Γ( ) 3780 Γ( )
( )10.
2 2 ρ6 2 ρ8 2 ρ
+ ( ) + ( ) +
1 21 l 23 l 25 l 
Γ( ) Γ( ) Γ( ) Γ( )
2 2 2 2
If we assume that the punch settlement δ≈ w 0 then the remaining harmonics may
be called the solution error. Direct computations show that the error is less than
3% inside the circle ρ≤ l . The error is reasonably small outside the circle
reaching 20% at the apex, and decreasing very rapidly with the distance from the
apex. Taking into consideration that the error sign fluctuation will result in even
smaller error in the total force value, we may assume (5.4.7) being the
relationship between the punch settlement and the total force which can be
rewritten in the form
HP
δ ≈ , (5.4.10)
g√A
where A is the punch base area, and g is a dimensionless coefficient depending

on the punch geometry only.
2√ A
g = 2 , (5.4.11)
π ra
where
2π
1 ⌠
r = a (φ) dφ (5.4.12)
a 2π ⌡
0
can be called the average radius with respect to the center of gravity If the
correct pressure distribution required to yield w =δ over the punch could be found
then the total force P would be related to δ by a formula
HP
δ = .
g 1√ A
Our basic assumption is that the g of equation (5.4.10), which we can calculate,
is likely to be a good approximation to g 1. The problem now is to find the
value of g for various punch planforms. One can compute the coefficient g for
the square from (5.4.9) as
1
g = = 0.3611,
π ln(1 + √2)
which is very close to the approximate value 0.3607 given by Maxwell for the
capacity of the square. Using the electrostatic analogy, one can easily deduce
that our coefficient g is related to the capacity C of a flat lamina by
C
g = . (5.4.13)
√A
Of course, closeness to the result of Maxwell does not mean that ours is so
accurate. The value of g which seems to be accurate was obtained in (Noble,
1960) and (Solomon, 1964a), and is 0.367, so that our result is in error by
1.6% which is not bad. Now it seems reasonable to assume that formulae
(5.4.10−5.4.12) are valid for an arbitrary flat punch, and we need to verify how
good they really are for each specific case.
We have found in the literature only one general formula of the type
(5.4.10) suggested by Solomon (1964b). His result expressed through the
coefficient g reads
1/8
29/8 I 0
g = . (5.4.14)
π11/8 A 1/4
where I 0 stands for the polar moment of inertia. One can easily verify that
formula (5.4.14) is exact for a circle, so one should expect it to be sufficiently
accurate for domains with the aspect ratio not far away from unity, but the error
might be quite significant for oblong domains. For example, in the case of an
ellipse with semi-axes a and b formula (5.4.14) gives
27/8 a b 1/8
g = 3/2 ( b + a ) .
π
The error of this formula can be quite significant for large aspect ratio ε= a / b .
Our formulae (5.4.10−5.4.12) in the case of an ellipse are exact.
Example 1: Polygon. Consider a flat punch, shaped as a polygon with n

sides, with the only limitation that the function a (φ) describing its boundary be
continuous and single-valued. The origin of the coordinate system is located at
the center of gravity, as before. Let us number the polygon sides in a
counter-clockwise direction from 1 to n , a k being the length of the k th side.
The apex at which the sides a k and a k+1 intersect is numbered k +1. It is clear
that the value of index equal to n +1 is understood as 1. We denote the
distance from the center of gravity to the k th apex as b k. Let A k be the area
of the triangle formed by a k, b k and b k+1, the total area A of the polygon being
equal to the sum of A k. Then formulae (5.4.11) and (5.4.12) yield the
following expression for the coefficient g
2√ A
g = . (5.4.15)
n
Ak b k + b k+1 + a k
π Σ k=1
ak
ln
b k + b k+1 − a k
The formula due to Solomon (5.4.14) in this case gives
29/8 
n
2 A 3k a 4k + 3( b 2k+1 − b 2k)2 1/8
g = 11/8 
π

Σ
k=1
2 2
 1 +
A ak  48 A 2k
 .

(5.4.16)
In the case of a regular polygon formulae (5.4.15) and (5.4.16) simplify to
4√tan(π/ n )
g = , (5.4.17)
1 + sin(π/ n )
π√ n ln
1 − sin(π/ n )
and
29/8  2cos2(π/ n ) + 1 1/8
g = (5.4.18)
π11/8  3 n sin(2π/ n ) 
respectively. Formulae (5.4.17) and (5.4.18), though looking different, give

practically the same results in the whole range 3≤ n <∞. Consider several
particular values of n . For an equilateral triangle ( n =3) formula (5.4.17) gives g
= 0.3673. The value of g , which seems likely to be accurate, can be computed
from (Solomon 1964a), and is equal to 0.3829, so that our result is in error by
4.1%. As we have seen earlier, the error of (5.4.17) for a square is 1.6%.
Since formula (5.4.17) in the limiting case n →∞ gives the exact result for a
circle g =2/π3/2=0.35917, we should expect that the error of (5.4.17) will decrease
with n . For a regular pentagon g =0.3599. We did not find in the literature
anything to compare with this result. The value of g for a regular hexagon is
0.3595 which is within the bounds given by Pólya and Szegö (1951)
0.35917< g <0.3635, and it is quite clear that the maximum possible error indeed
decreases with n . It is noteworthy that the value of g does not change
significantly in the whole range 3≤ n <∞.
Example 2: Triangle. In the case of a triangular punch with the sides a 1,
a2 and a 3, formula (5.4.15) simplifies as follows:
b1 + b2 + a1 b2 + b3 + a2
6 1 1
g = ln + ln
π√ A a 1 b 1 + b 2 − a 1 a2 b2 + b3 − a2
b 3 + b 1 + a 3 -1
1
+ ln  . (5.4.19)
a 3 b 3 + b 1 − a 3
The parameters in (5.4.19) can be determined from the well known geometric
formulae
A = [ p ( p − a 1)( p − a 2)( p − a 3)]1/2, p = ( a 1 + a 2 + a 3)/2,
1 1
b1 = [2( a 21 + a 23) − a 22]1/2 , b2 = [2( a 22 + a 21) − a 23]1/2,
3 3
1
b3 = [2( a 23 + a 22) − a 21]1/2.
3
We are unaware of any report treating a triangle of general type but certain
particular cases have been considered, so we can compare the results. When
a 1= a 2= l , and the angle between these two sides is equal to α, the formula for
the coefficient g can be rewritten in the form
√tan(α/2) 2sin( )ln(cot cot ) + ln tan( + ) ,

6 α 2γ − α α π γ -1
g =
π  2 4 4 4 2
(5.4.22)
where γ = tan (3tan(α/2)).
-1
The isosceles triangle was considered by Rvachev et al (1977) who gave an

approximate expression for the stress distribution but, for some reason, did not
give the relationship between the total force and the punch settlement. They
have though presented a graph depicting the location of the point of application
of the force P as a function of the angle α. Their graph indicates a small
variation (within 0.01 of the height of the triangle) of the coordinate about the
center of gravity. We think that Rvachev et al did not realize that this
fluctuation should be attributed to the approximate nature of their method, and
that the exact location of the point is at the center of gravity. Of course, the
previous statement should be understood as a conjecture since our method is also
approximate.
The case of α=π/2 was considered by Pólya and Szegö (1951) who gave
the following bounds for g : 0.35917< g <0.4282. Okon and Harrington (1970)
obtained g = 0.3867 as the most probable result. Our result is g = 0.374
which is well inside the admissible region and differs by 3.3% from the result
of Okon and Harrington. The following bounds were established by Pólya and
Szegö for a triangle with the sides a , a /2 and a √3 /2: 0.35917< g <0.517. Our
result g = 0.3822 is within this interval. There seems to be no other source to
compare with this result.
Example 3: Rectangle. Consider a punch with a rectangular base, a and
b being its semiaxes. Introduce the aspect ratio ε= a / b . Formula (5.4.15) in this
case reduces to
2
g = . (5.4.21)
π √ε sinh-1( ) + sinh-1ε 
1 1
 ε √ε 
Howe (1920) suggested an approximate formula for the capacitance of a rectangle

which in terms of the coefficient g reads
1
g = . (5.4.22)
2√ε sinh-1ε + sinh-1( )+
1 1 ε 1 (ε + 1) 
2 3/2
+ 2 −
 ε ε 3 3ε 3ε2 
The result due to Solomon (5.4.16) in this case takes the form
29/8  ε 1 1/8
g = + . (5.4.23)
π11/8  12 12ε 
We have found in the literature some numerical results which seem to be more
or less accurate. Borodachev and Galin (1974) have considered the case of a
narrow rectangular punch, and Noble (1960) investigated an equivalent problem of
the electric charge distribution on a rectangular lamina. Their data, expressed in
terms of the coefficient g , are presented below and compared with our result
(5.4.21) and those due to Howe (5.4.22) and Solomon (5.4.23). The following
relationship can be established between Borodachev-Galin’s coefficient γ and our
g : γ=1/2π g √ε . Several useful conclusions can be drawn from the data presented.
It seems logical to assume that the error of an approximate formula should
change monotonically (or to have only one extremum) with respect to a certain
ε = 0.020 0.050 0.100 0.125 0.150 0.200 0.250 0.500 1.000

Borodachev
and Galin 0.7375 0.5661 0.4819 – 0.4458 0.4259 – – –
Noble – – – 0.4543 – – 0.4047 0.3762 0.3670
Formula (5.4.21) 0.8031 0.6072 0.5037 0.4771 0.4576 0.4306 0.4128 0.3742 0.3612
Howe (5.4.22) 0.6916 0.5317 0.4481 0.4268 0.4112 0.3899 0.3759 0.3462 0.3363
Solomon (5.4.23) 0.5402 0.4819 0.4423 0.4304 0.4211 0.4071 0.3969 0.3715 0.3613
Discrepancy (%)
Formula (5.4.21) -8.9 -7.3 -4.5 -5.0 -2.6 -1.1 -2.0 0.5 1.6
Howe (5.4.22) 6.2 6.1 7.0 6.1 7.8 8.5 7.1 8.0 8.4
Solomon (5.4.23) 26.7 14.9 8.2 5.3 5.5 4.4 1.9 1.3 1.6
parameter. The fact that the discrepancy due to each formula jumps when
moving from the data due to Noble to those by Borodachev and Galin, indicates
that the results of at least one author are not exact. Our formula seems to
perform better then the other two in a sufficiently wide range of aspect ratio.
As expected, the formula due to Solomon performs well only when the aspect
ratio is not far away from unity. There seems to be little change in the error
of Howe’s formula (5.4.22). If this is really so, then its accuracy can be
improved dramatically just by multiplication by a constant factor, say, 1.07.
Example 4: Rhombus. Let α be the angle at one of the rhombus apices.
Formula (5.4.15) in this case yields
2
g = . (5.4.24)
cos(α/2) + sin(α/2) + 1
π√sinα ln
cos(α/2) + sin(α/2) − 1
The same formula in terms of the rhombus semiaxes a and b and the aspect
ratio ε = a / b has the form
[2(ε + 1/ε)]1/2
g = . (5.4.25)
1 + ε + (1 + ε ) 2 1/2
π ln
1 + ε − (1 + ε2)1/2
We did not find in mechanics literature any result related to a punch with a
rhombus planform. In electrical sciences, the related problem of the capacity of
a diamond was solved numerically by Okon and Harrington (1970). Their result,
expressed in terms of the coefficient g , for a diamond with the aspect ratio
a : b =0.7:1.65 is g =0.3855. Formula (5.4.25) gives g =0.3744 which differs by 3%
from the result of Okon and Harrington. They also considered a rhombus with
the aspect ratio 1:2. Their result, g =0.3705, almost coincides with ours which is
g =0.3698.
segment parameters. The location of its center of gravity is defined by x = kr,
c
where
2 sin3α
k = . (5.4.26)
1
3(α − sin2α)
2
The equation of the segment boundary with respect to its center of gravity takes
the form
a (φ) = r [− k cosφ + (1 − k 2sin2φ)1/2], for 0≤φ≤π−γ or π+γ≤φ<2π;

and
k − cosα
a (φ) = r for π−γ≤φ≤π+γ. (5.4.27)
cos(π − φ)
Substitution of (5.4.27) into (5.4.11−5.4.12) gives
1
2( α − sin 2α)1/2
2
g = . (5.4.28)
π2E( k ) − E(γ, k ) − k sinγ + ( k − cosα) ln tan( + )
π γ
 4 2 
where γ = tan-1(sinα/( k − cosα)). We have found only one numerical example

to verify the accuracy of (5.4.28): Okon and Harrington (1970) have computed
the capacity of a semicircle. Their result expressed through the coefficient g is
0.3724. Formula (5.4.28) gives 0.3714 with the discrepancy of 0.3%.
Example 6: Circular sector. Repetition of the procedure, described in the
previous paragraph, leads to the following result for a circular sector of
subtended angle 2α:
2√α
g = . (5.4.29)
πE(γ, k ) − k sinγ + k sinα ln[cot(α/2)cot((γ − α)/2)]
 
Here, k =2sinα/(3α), and γ=tan-1(sinα/(cosα− k )). Okon and Harrington (1970)

obtained g =0.3668 for the case of a quadrant. Formula (5.4.29) for α=π/4 gives
g =0.3639, with the discrepancy of 0.8%.
Discussion. We might enquire whether there exists any contour, other than
an ellipse, for which an expression of the type (5.4.5) would be an exact
solution to the integral equation (5.1.8). Expression (5.4.6) can provide the
sufficient conditions:
2π
⌠ ein(φ-φ0) F 1 − | n |, 1; 1; 1 − x dφ
2
(5.4.30)
⌡  2 2 a 2(φ0) 0
0
should be equal to zero for n ≠0. The integral (5.4.30) will vanish for all odd
n if a (φ) contains even harmonics only. In the case of even n , the
hypergeometric function in (5.4.30) represents a finite polynomial in x / a (φ) of
degree not greater than n −2 which means that the integral will vanish if ( a (φ))-2
contains harmonics not higher than the second, which corresponds to an ellipse.
The question as to whether these conditions are also necessary requires an
additional investigation.
Solomon’s formula (5.4.14) can be considered as a particular case of a
more general one, namely
2π 2π
g =
2  1 ⌠ ( a (φ))mdφ -1/2m
 1 ⌠ ( a (φ))ndφ1/2n. (5.4.31)
π√ A 2π ⌡  2π ⌡ 
0 0
for m =2 and n =4. One may ask now whether this choice of the parameters m
and n is in any sense optimal. Direct computations show that for a regular
polygon, m =1 and n =9 give much better accuracy than (5.4.18). In the case of
a rectangle m =1 and n =7 are more accurate than (5.4.23). It is clear that the
formula chosen by Solomon (5.4.14) is not the best particular case of (5.4.31).
One can suggest many formulae of the type (5.4.31) but this would be just an
exercise in curve-fitting, which is outside the scope of this book.
Mossakovskii (1972) considered the case of a flat punch of nearly circular
cross-section. He assumed a solution for the contact tractions in the form
F 0(ρ,φ) + α F 1(ρ,φ) + α2 F 2(ρ,φ) + . . .

σ(ρ,φ) = , (5.4.32)
a 2(φ) − ρ2 1/2
 
where α is a small parameter. Expression (5.4.32) has two essential
disadvantages as compared with (5.4.3): ( i )even in the case of an ellipse
expression (5.4.32) gives an infinite series rather than the closed form exact
solution; ( ii )the solution (5.4.32) at the point ρ=0 is a function of φ which is
physically meaningless. Though Mossakovskii’s method can give reasonably
accurate estimations for the relationship between the punch settlement and applied
force, his claim of being able to evaluate the stress distribution exerted by the
punch seems to be wrong because of incorrect assumption of a square root
singularity at the punch boundary (for example, in the case of a square punch).
The mathematically similar problem of sound penetration through an aperture
of general shape in a rigid flat screen was considered in (Fabrikant, 1986b).
The same method was used to find the electrical capacity of flat laminae
(Fabrikant, 1986k).
Exercise 5.4
2. Derive (5.4.7).
3. Verify (5.4.15).
4. Derive (5.4.19).
5. Generalize the theory for the case of a non-homogeneous half-space, with the
modulus of elasticity being proportional to a power function of the depth.
5.5 Inclined flat punch of general shape
An approximate analytical solution is given here to the contact problem of

a flat inclined punch of arbitrary planform under the action of a normal
non-centrally applied force. Some accurate relationships are established between
the tilting moments and the angles of inclination of an arbitrary flat punch.
Specific formulae are derived for a punch whose planform has the shape of a
polygon, a triangle, a rectangle, a rhombus, a circular sector and a circular
segment. All the formulae are checked against the solutions known in the
literature, and their accuracy is confirmed.
Theory. Consider a flat-ended punch with a planform S whose boundary is
given in polar coordinates as
ρ = a (φ).
where the function a (φ) is bounded and single-valued. The punch is pressed
against an elastic half-space by a normal force P applied at the point with
cartesian coordinates x 0 and y 0. This loading is statically equivalent to a
centrally applied force P and two tilting moments M x= Py 0 and M y=− Px 0. The
case of a centrally applied force was considered in the previous section. It
remains here to consider the punch under the action of the tilting moments, and
to superpose the results. Repeating the procedure of the section 5.4, we come
to the same governing integral equation, namely,
x2
λ( , φ−φ0)
ρ 2π a (φ0)
ρρ0
dx
w (ρ,φ) =
2 ⌠
H ⌠ dφ ⌠ σ(ρ0,φ0)ρ0dρ0.
π ⌡ (ρ2 − x 2)1/2 ⌡ 0 ⌡ (ρ20 − x 2)1/2
0 0 x
(5.5.1)
Let the normal displacements under the punch be
w = αx y − αy x , (5.5.2)
where αx and αy are the tilting angles about the axes Ox and Oy respectively.
It is necessary to relate these angles to the tilting moments.
Present the normal stress distribution under the punch as
a (φ)ρ( p 1cosφ + p 2sinφ)

σ = , (5.5.3)
a 2(φ) − ρ21/2
 
where p 1 and p 2 are the as yet unknown constants. Make use of the condition
that the integral of σ over S should be equal to zero. Since p 1 and p 2 are
independent, this leads to the two equations
2π 2π
⌠ ( a (φ))3cosφ dφ = 0 , ⌠ ( a (φ))3sinφ dφ = 0.
⌡ ⌡
0 0
(5.5.4)
It follows that (5.5.4) will be satisfied if and only if the origin of coordinates is
located at the center of gravity of the domain of contact. The axis orientation
will be discussed later.
The relationships between the tilting moments and the parameters p 1 and p 2
can be established from the statics conditions
M x = Py 0 = ⌠ ⌠ σ y dS , M y = − Px 0 = −⌠ ⌠ σ x dS,
⌡⌡ ⌡⌡
S S
which leads to
8 8
Mx = ( p I + p 2 I x), M y = − ( p 1 I y + p 2 I xy), (5.5.5)
3 1 xy 3
where I x, I y and I xy are the well known quantities of the moments of inertia
and the product of inertia respectively. Now it is necessary to relate p 1 and p 2
to the angles αx and αy. This can be done by substitution of (5.5.3) into
(5.5.1) which, after integration with respect to ρ0, yields
∞ ρ
Σ
2
w (ρ, φ) = H ⌠ (x )|n| 2 x d x 2 1/2
n=-∞
⌡ ρ (ρ − x )
0
2π
F
x2 
×⌠ e
in(φ-φ0) 3 − |n| 1
, ; 1; 1 − 2 ( p cosφ0 + p 2sinφ0)dφ0. (5.5.6)
⌡  2 2 a (φ0) 1
0
Here F denotes the Gauss hypergeometric function. Further evaluation of the

normal displacements can be carried out separately for each harmonic. Note that
the zero th and all the even harmonics of w will be zero if a (φ) contains only
even harmonics. The first harmonic will take the form
2π
π ⌠
w 1(ρ, φ) = H ρ cos(φ − φ0)( p 1cosφ0 + p 2sinφ0) a (φ0) dφ0,
2 ⌡
0
which can be simplified as
π
w 1(ρ,φ) = H ρ[( p 1 J y + p 2 J xy)cosφ + ( p 1 J xy + p 2 J x)sinφ], (5.5.7)
2
where the following quantities were introduced
2π 2π
J x = ⌠ a (φ) sin2φ dφ , J y = ⌠ a (φ) cos2φ dφ ,

⌡ ⌡
0 0
2π
J xy = ⌠ a (φ) sinφ cosφ dφ (5.5.8)

⌡
0
These quantities do not seem to have been used before in engineering practice so
they do not have an accepted name. Since their tensor properties are similar to
those of the moments of inertia, we shall call J x and J y the linear moments of
a two-dimensional domain about the axes Ox and Oy respectively, while J xy will
be called the linear product of a two-dimensional domain about the axes Ox and
Oy
It is important to note that the third harmonic is equal to zero for an
arbitrary contour. Here is the expression for the fifth harmonic
2π
cos5(φ − φ0)
128 4⌠
w 5(ρ,φ) = Hρ ( p 1cosφ0 + p 2sinφ0)dφ0,
315 ⌡ a 2(φ0)
0
which can be modified as
64
w 5(ρ,φ) = H ρ4{[( A + A ) p 1 + ( A − A ) p 2]cos5φ
315 c6 c4 s6 s4
+ [( A
s6
+A ) p1 + ( A
s4 c4
− A c6) p 2]sin5φ}. (5.5.9)
Here, the following geometrical characteristics of the domain of contact were

introduced
2π 2π
= ⌠ = ⌠
cos4φ dφ cos6φ dφ
A , A ,
c4 ⌡ ( a (φ))2 c6 ⌡ ( a (φ))2
0 0
2π 2π
sin4φ dφ sin6φ dφ
A = ⌠ , A = ⌠ . (5.5.10)
s4 ⌡ ( a (φ))2 s6 ⌡ ( a (φ))2
0 0
Investigation of further harmonics shows that their amplitude decreases.

Now consider in more detail the case of a square of side 2 l . The
equation of the boundary in this case is a (φ) = l /cosφ for −π/4<φ<π/4, and the
pattern is repeated outside this range. We can evaluate the first two non-zero
harmonics:
w 1 = π H l ρln(1 + √2)( p 1cosφ + p 2sinφ),
128 H ρ4
w5 = ( p 1cos5φ + p 2sin5φ).
945 l 2
Since the amplitude of w 5 is significantly less than that of w 1, it seems natural

to assume w ≈ w 1, and the remaining harmonics may be called the solution
error. Direct computations show that the error is less than 3% inside the circle
ρ≤ l . The error is reasonably small outside the circle, reaching about 20% at the
apex and decreasing very rapidly with the distance from the apex. Taking into
consideration that the error sign fluctuation will result in even smaller error in
the integral characteristics sought, a direct comparison of (5.5.2) and (5.5.7) leads
to
π π
αx = H ( p 1 J xy + p 2 J x) , αy = − H ( p 1 J y + p 2 J xy). (5.5.11)
2 2
The inversion of (5.5.11) gives
2 J xyαx + J xαy 2 J yαx + J xyαy

p1 = − , p2 = .
πH J x J y − J xy2 πH J x J y − J xy2
(5.5.12)
Substitution of (5.5.12) in (5.5.5) finally gives the required relationship
16 16
Mx = ( m α + m 12αy) , My = ( m α + m 22αy)
3π H 11 x 3π H 21 x
(5.5.13)
where
J y I x − J xy I xy J xy I x − J x I xy
m 11 = , m 12 = ,
J x J y − J xy2 J x J y − J xy2
J xy I y − J y I xy J x I y − J xy I xy
m 21 = , m 22 = .
J x J y − J xy2 J x J y − J xy2
It is clear that all these results can be rewritten in a matrix or a tensor form.
One can verify that formulae (5.5.13) are invariant with respect to an arbitrary
rotation of the axes. The same property holds for m 11+ m 22 and m 12− m 21.
Strictly speaking, according to the reciprocal theorem, m 12 should be equal to
m 21, so that formulae (5.5.13) generally do not satisfy this theorem. But we
may state that this theorem is satisfied ’approximately’. We mean by this the
following property which has been verified by several direct computations, namely,
| m 12− m 21|/ m 11 <<1 and | m 12− m 21|/ m 22 <<1. This theorem will be satisfied exactly for
any domain which has at least one axis of symmetry because in this case
m 12= m 21=0, provided that the coordinate axes coincide with the central principal
axes of the domain of contact. Since we have no numerical data for
non-symmetrical domains which could be used to verify the accuracy of (5.5.13),
we shall consider further only the case when the domain of contact has an axis
of symmetry. In this case formulae (5.5.5), (5.5.11) and (5.5.13) simplify
significantly, namely,
8 8
Mx = I p, My = − I p, (5.5.14)
3 x 2 3 y 1
π π
αx = HJ p , α y = − H J y p 1, (5.5.15)
2 x 2 2
I I
16 x 16 y
Mx = α , My = α. (5.5.16)
3π H J x x 3π H J y y
Returning back to the problem of a non-centrally loaded flat punch and
using the results of section 5.4, we can write the following expression for the
traction distribution under the punch in terms of the applied force P and the
coordinates of its point of application x 0 and y 0
P a (φ)
1 + 3 A  0 + 0,
xx yy
σ = (5.5.17)
 4  Iy I x 

2 A a (φ) − ρ
2 2
1/2
 
where A is the area of the domain S . An expression equivalent to (5.5.17) can
be written in terms of the normal displacement δ and the tilting angles αx and
αy
2 a (φ) αy αx
σ =  δ + x − y , (5.5.18)
 J0 Jx Jy 
π H  a 2(φ) − ρ2
1/2
 
where
2π
J 0 = ⌠ a (φ) dφ
⌡
0
The quantity J 0 may be called the polar linear moment due to the analogy with
the moments of inertia and the property J 0= J x+ J y. One can note also that J 0 is
proportional to the average polar radius. Expressions (5.5.17) and (5.5.18) are
exact for an ellipse. We expect them to be reasonably accurate in the
neighbourhood of the coordinate origin for an arbitrary punch planform with at
least one axis of symmetry, while the error might become quite significant close
to the boundary of the domain S .
Let us rewrite formulae (5.5.16) in the form
A 3/2 A 3/2
Mx = hα , My = hα, (5.5.19)
2π H x x 2π H y y
where
32 I x 32 I y
hx = , hy = . (5.5.20)
3 A 3/2 J x 3 A 3/2 J y
We introduced the coefficients h x and h y for two reasons: since they are
dimensionless they characterize the shape of S and do not depend on its size;
both h x and h y are equal to the corresponding coefficients of magnetic
polarizability which will simplify the comparison of our results with the numerical
data available. There is an advantage of formulae (5.5.19) over the equivalent
(5.5.16): the factors depending on the shape of S are separated from those
depending on its size. One can draw from (5.5.19) an immediate conclusion that
in the case when a domain S is magnified so that its linear dimensions double,
its area quadruples, and the tilting moment should be multiplied by 8 in order to
produce the same tilting angle. This conclusion is not so clear from (5.5.16).
The remaining part of this section will be devoted to the evaluation of the
coefficients h x and h y for various punch planforms. Several punch planforms are
considered. Each configuration is related to its central principal axes and
assumed to have at least one axis of symmetry coinciding with the axis Ox .
The high degree of accuracy of formulae (5.5.20) is confirmed by comparison
with available numerical solutions.
Example 1: Polygon. Consider a polygon with n sides. The function
a (φ) describing its boundary is bounded and single-valued. The origin of the
coordinate system is located at the center of gravity, as before. Let us number
the polygon sides in a counter-clockwise direction from 1 to n , a k being the
length of the k th side. The apex at which the sides a k and a k+1 intersect is
numbered k +1. It is clear that the value of index n +1 is to be understood as
1. We denote the distance from the center of gravity to the k th apex as b k;
ψk stands for the angle between the axis Ox and the perpendicular to the side
a k. Let A k be the area of the triangle formed by a k, b k and b k+1, the total
area A of the polygon being equal to the sum of A k. The following expressions
can be obtained for the moments of inertia
n
2 A 3k b 2k+1 − b 2k a 4k + 3( b 2k+1 − b 2k)2
Ix = Σ
k=1
 2
2 sin ψk +
ak  4Ak
sin2ψk +
48 A 2k
cos2ψk,

(5.5.21)
n
2 A 3k b 2k+1 − b 2k a 4k + 3( b 2k+1 − b 2k)2
Iy = Σk=1
 2
2 cos ψk −
ak  4Ak
sin2ψk +
48 A 2k
sin2ψk.

(5.5.22)
The linear moments can be computed in the form
n
Ak
Jx = Σ
k=1
a 2k
−( 1 + 1 )[ a 2 − ( b − b )2]cos2ψ
 bk b k+1 k k k+1 k
b k + b k+1 + a k
cos2ψk,
1 1
+ 4 A k( − ) sin2ψk + 2 a k ln
bk b k+1 b k + b k+1 − a k 
(5.5.23)
n
Ak
Jy = Σ
k=1
a 2k
( 1 + 1 )[ a 2 − ( b − b )2]cos2ψ
 bk b k+1 k k k+1 k
b k + b k+1 + a k
sin2ψk.
1 1
− 4 A k( − )sin2ψk +2 a k ln
bk b k+1 b k + b k+1 − a k 
(5.5.24)
Substitution of (5.5.21−5.5.24) into (5.5.20) gives the coefficients h x and h y for
an arbitrary polygon. In the case of a regular polygon a k= a , b k= b = a /[2sin(π/ n )],
ψk=2π( k -1)/ n , Ak =[ a 2cot(π/n)]/4=[ b 2sin(2π/ n )]/2, A = nA k, and formulae
(5.5.21−5.5.24) simplify to
cot cot2( ) +  = sin 2 + cos ,

na 4 π π 1 nb 4 2π 2π
Ix = Iy =
64 n n 3 24 n n
(5.5.25)
na π 1 + sin(π/ n ) nb π 1 + sin(π/ n )
Jx = Jy = cot ln = cos ln .
4 n 1 − sin(π/ n ) 2 n 1 − sin(π/ n )
(5.5.26)
Substituting (5.5.25) and (5.5.26) in (5.5.20) leads to
2π
16(2 + cos )
n
hx = hy = .
9 n 3sin cos3( ) ln
π π 1/21 + sin(π/ n )
 n n 1 − sin(π/ n )
(5.5.27)
Consider several particular values of n . For an equilateral triangle ( n =3) formula
(5.5.27) gives h x= h y=31/416/[27ln(2+√3 )]=0.5922. We did not find any numerical

data to compare with this result. In the case of a square n =4, and
h x= h y=4/[9ln(1+√2 )]=0.5043 which is inside the interval from 0.4973 to 0.5162
given by Okon and Harrington (1981) and within 3% from the result of de
Smedt (1979) which is 0.5193. Since formula (5.5.27) gives the exact result for
a circle in the limiting case of n →∞, namely, h x= h y=8/(3π3/2)=0.4789, we should
expect that the error of (5.5.27) will decrease with n . The value of the
coefficients for a regular hexagon is h x= h y=40√2 /(3 81 ln3)=0.4830 which differs
1/4
by 1.4% from the result 0.49 due to Okon and Harrington (1981), and it is
quite clear that the maximum possible error indeed decreases with n . It is
noteworthy that the values of the coefficients do not change significantly in the
whole range 3≤ n <∞.
Example 2: Isosceles triangle. In the case of a triangle with sides
a 1= a 2= l , the angle between them being equal to α, formulae (5.5.20−5.5.23) give
1 4 1 4
Ix = l sinα sin2(α/2), Iy = l sinα cos2(α/2),
12 36
l cos sinα + sin(α + γ) − 2sinγ

2 α
Jx =
3 2
cot ) + ln tan( + ),

α 2γ − α α π γ
+ 2sin3( )ln(cot
2 4 4 4 2
l cos −sinα − sin(α + γ) + 2sinγ

2 α
Jy =
3 2
cot ),
α 2γ − α α
+ sinαcos ln(cot
2 4 4 
with the result for the coefficients

h x = 8(tan(α/2))3/23sinα + sin(α + γ) − 2sinγ
 
γ -1
cot ) + ln tan( + ) ,
α 2γ − α α π
+ 2sin3 ln(cot
2 4 4 4 2

(5.5.28)

h y = 8√cot(α/2) 9−sinα − sin(α + γ) + 2sinγ
 
α -1
cot ) ,
α 2γ − α
+ sinαcos ln(cot
2 4 4 

where γ = tan-1(3tan(α/2)).
The isosceles right triangle was considered by Okon and Harrington (1981)
who gave the interval between 0.9829 and 1.021 for only one of the coefficients,
which in our notation is h x. Our result for h x is 0.9255 which differs by less
than 10% from theirs. The second formula (5.5.28) gives h y = 0.3995, and we
found nothing in the literature to compare with this result.
Example 3: Rectangle. Consider a punch with a rectangular base, a 1 and
a 2 being its semiaxes. Introduce the aspect ratio ε= a 2/ a 1. Formulae
(5.5.21−5.5.24) in this case reduce to
4 4
Ix = a a 3, Iy = a 3a ,
3 1 2 3 1 2
J x = 4 a 1sinh-1ε , J y = 4 a 2sinh-1(1/ε).
and formulae (5.5.20) yield
4ε3/2 4ε-3/2
hx = , hy = . (5.5.29)
9sinh-1ε 9sinh-1(1/ε)
The coefficients of magnetic polarizability were computed by de Smedt (1979) for

a rectangle with various aspect ratios ε. Here, we present his results along with
those given by (5.5.29) Our formula (5.5.29) seems to perform satisfactorily in a
sufficiently wide range of aspect ratio. The approximate expression for the stress
distribution under the punch according to (5.5.17) takes the form
P a (φ) xx
1 + 9 0 + 0.
yy
σ = (5.5.30)
 4 a 21 a 22 

8 a 1 a 2 a (φ) − ρ
2 2
1/2
 
Expression (5.5.30) can be used for analyzing the process of movement of the
applied force P , say, along the axis Ox . This analysis can be done by
ε= 0.1000 0.2000 0.3333 0.5000 0.7500 0.8000 1.0000

de Smedt h x= 0.1287 0.1881 0.2531 0.3249 0.4240 0.4436 0.5193
Formula (32) h x= 0.1408 0.2001 0.2612 0.3265 0.4165 0.4341 0.5043
de Smedt h y= 4.1070 2.0260 1.2600 0.8892 0.6426 0.6130 0.5193
Formula (32) h y= 4.6876 2.1488 1.2701 0.8708 0.6228 0.5929 0.5043
Discrepancy in h x (%) -9.4 -6.4 -3.2 -0.5 1.8 2.2 2.9
Discrepancy in h y (%) -14.1 -6.1 -0.8 2.1 3.1 3.3 2.9
requiring that the contact traction vanish at the edge. One can conclude from
(5.5.17) that the boundary at which this occurs will always be a straight line.
It is clear from (5.5.30) that the punch will be in contact with the half-space as
long as x 0≤4 a 1/9, after which the punch will start separating from the half-space.
Assuming that the new domain of contact is also a rectangle (of course, with a
different aspect ratio), one can again apply the formulae of this section to
analyze the process further. If we denote the width of the zone of separation
by c , the following relationship holds
2 4
c = (9 x − 4 a 1) , for x 0≥ a 1.
5 0 9
The last formula states, for example, that when the force P is applied at
x 0=13 a 1/18 only a half of the punch will be in contact with the half-space.
Unfortunately, there is no data to verify these relationships. Further analysis
reveals that the core inside which the force can be applied without causing any
separation is a rhombus with semiaxes 4 a 1/9 and 4 a 2/9 respectively. As one
knows, in the case of a circular punch the core is a circle of radius equal to
one third of the radius of the punch. The results due to (5.5.30) can be
compared with the numerical data received in a personal communication from de
Smedt. In order to make the comparison possible, we must set P =0, M x=0 in
(5.5.30) and replace M y by (5.5.19), with the result
9√ε a (φ) h y x
σH = . (5.5.31)
4 a 1 a 2(φ) − ρ2
1/2
 
Computations due to (5.5.31) were made for ε=0.5 along the axis Ox , the value
h y was taken 0.8708 (see the preceding Table). Here are the results compared
with those communicated by de Smedt Since the numerical method of De Smedt
is also approximate, we are using the word discrepancy rather than the word
error in the tables throughout this and adjacent sections. We can also compare
x/a 1= 0.0833 0.2500 0.3333 0.5000 0.5833 0.6667 0.7500 0.8333 0.9167
de Smedt σ H = 0.1143 0.3501 0.4759 0.7523 0.9367 1.1460 1.4304 1.8303 2.8182
Formula (5.5.31) σ H = 0.1159 0.3577 0.4898 0.7999 0.9950 1.2392 1.5709 2.0886 3.1777
Discrepancy (%) -1.3 -2.2 -2.9 -6.3 -6.2 -8.1 -9.8 -14.1 -12.8
the same values along the axis Oy . One can use a formula similar to (5.5.31)
replacing all x by y and interchanging a 1 and a 2, the value of h x was taken to
be 0.3265. The results are:
y/a 2= 0.1667 0.3333 0.5000 0.6667 0.8333
de Smedt σ H = 0.1756 0.3663 0.6011 0.9014 1.6413
our result σ H = 0.1756 0.3673 0.5998 0.9292 1.5662
Discrepancy (%) 0.0 -0.3 0.2 -3.1 4.6
The agreement is satisfactory.

Example 4: Rhombus. Let α be the angle at one of the rhombus apices,
and l be its side. Formulae (5.5.21−5.5.24) in this case yield
14 α 14 α
Ix = l sinα sin2( ), Iy = l sinα cos2( ), A = l 2sinα,
6 2 6 2
J x = 2 l sinαcos − sin + sin2( )ln

α α α cos(α/2) + sin(α/2) + 1
,
 2 2 2 cos(α/2) + sin(α/2) − 1
J y = 2 l sinα−cos + sin + cos2( )ln

α α α cos(α/2) + sin(α/2) + 1
.
 2 2 2 cos(α/2) + sin(α/2) − 1
The coefficients are determined as
α
8sin2
2
hx = ,
9(sinα)3/2cos − sin + sin2( )ln
α α α cos(α/2) + sin(α/2) + 1
 2 2 2 cos(α/2) + sin(α/2) − 1
α
8cos2
2
hy = .
9(sinα)3/2−cos + sin + cos2( )ln
α α α cos(α/2) + sin(α/2) + 1 
 2 2 2 cos(α/2) + sin(α/2) − 1 
(5.5.32)
The same formulae can be expressed in terms of the rhombus semiaxes a and b
and the aspect ratio ε= b / a , giving
2√2ε (1 + ε2)
hx = ,
91 − ε +
ε2
1 + ε + (1 + ε )  2 1/2
2 1/2 ln
 (1 + ε ) 1 + ε − (1 + ε2)1/2 
2√2 (1 + ε2)
hy = . (5.5.33)
9ε3/2ε − 1 +
1 1 + ε + (1 + ε )  2 1/2
ln
 (1 + ε ) 2 1/2
1 + ε − (1 + ε2)1/2 
We did not find in mechanics literature any result related to a punch with a
rhombus planform. In the electrical sciences, the mathematically equivalent
problem of the coefficients of magnetic polarizability of a diamond was solved
numerically by de Smedt (1979). Here, we present his results compared with
those given by formula (5.5.33)
ε= 0.1000 0.2000 0.3333 0.5000 0.7500 0.8000 1.0000
de Smedt h x= 0.1181 0.1729 0.2341 0.3052 0.4101 0.4323 0.5193
Formula (5.5.33) h x= 0.1078 0.1627 0.2258 0.2986 0.4026 0.4230 0.5043
de Smedt h y= 6.1820 2.7060 1.5240 0.9946 0.6703 0.6323 0.5193
Formula (5.5.33) h y= 4.5987 2.1982 1.3254 0.9095 0.6388 0.6052 0.5043
Discrepancy of h x (%) 8.7 5.9 3.6 2.2 1.8 2.1 2.9
Discrepancy of h y (%) 25.6 18.8 13.0 8.6 4.7 4.3 2.9
The deterioration of the accuracy of (5.5.33) for small values of ε can be

attributed to the erroneous assumption of a square root singularity in (5.5.3)
which is grossly incorrect for domains with sharp angles.
The traction distribution under the punch can be expressed according to
(5.5.17) as
xx yy
σ =
P a (φ) 1 + 9 0 + 0.
 2 a 2 b 2 

2 A a (φ) − ρ
2 2
1/2
 
Further analysis of the last expression reveals that the core inside which the
force can be applied without causing any separation is a rectangle with semiaxes
2 a /9 and 2 b /9 respectively. In the case of ε=1 the rhombus transforms into a
square, and all the results are in agreement with those of the Example 3.
segment parameters. The location of its center of gravity is defined by x = kr,
c
where k is given by (5.4.26). The equation of the segment boundary with
respect to its center of gravity takes the form (5.4.27).
Computation of the area and moments (5.5.21−5.5.24) yields
1 1 2 1 2
A = r 2(α − sin2α), Ix = Ar (1 − k cosα), Iy = Ar (1+3 k cosα−4 k 2),
2 4 4
(5.5.34)
2  1 − k2
Jx = r − k sin3γ + (1 − k 2sin2γ)1/2sinγcosγ + F(π−γ, k )
3 k2

+
2k2 − 1
E( π−γ , k ) + 3( k − cosα ) −sinγ + ln tan(π + γ ),
k2  4 2

2 
r sinγ k sin2γ − 3cosα − (1 − k 2sin2γ)1/2 cosγ
3 
Jy =
  
1 − k2 1 + k2 
− F( π−γ , k ) + E( π−γ , k ) , (5.5.35)
k2 k2

where γ = tan-1(sinα/( k − cosα)), and the functions F(⋅,⋅) and E(⋅,⋅) stand for the
incomplete elliptic integrals of the first and the second kind respectively.
Substituting in (5.5.20) leads to
4(1 − k cosα)  1− k 2
−k sin γ + (1 − k sin γ) sinγcosγ + F(π−γ, k )
3 2 2 1/2
hx =
k2
α − 1sin2α1/2
 2 
+
2 k 2−1
E( π−γ , k ) + 3( k − cosα ) −sinγ + ln tan(π + γ )-1,
k2  4 2

4(1 + 3 k cosα − 4 k 2) 
sinγk sin γ − 3cosα − (1 − k sin γ) cosγ
2 2 2 1/2
hy =
α − 1sin2α1/2 
 2 
1 − k2 1 + k2 -1
− F( π−γ , k ) + E( π−γ , k )  . (5.5.36)
k2 k2

We can use this result to investigate the case of a circular punch under the
action of a normal force P applied at x 0> r /3. From the classical theory we
know that there should be a separation between the punch and the half-space.
Assuming that the domain of contact after separation is a circular segment, one
can get the following relationship between the coordinate x 0 and the size of the
segment characterized by the angle α
(1 − k )(1 + 4 k )
x0 = r . (5.5.37)
3( k − cosα)
The last expression is exact in two limiting cases: the complete circle α=π gives
x 0= r /3, and α→0 results in x 0= r . The problem of an inclined circular punch
was considered numerically in the book by Rvachev and Protsenko (1977). Here,
we compare the results
α (deg)= 158.4 108.1 102.0
Rvachev et al. x 0= 0.3583 0.5833 0.6250
Formula (5.5.37) x 0= 0.3543 0.5418 0.5750
Discrepancy (%) 1.1 7.1 8.0
The agreement should be considered as surprisingly good, especially taken into

consideration that Rvachev and Protsenko considered the domain of contact not in
the form of a segment but having a more complicated shape.
Example 6: Cross. Consider a punch configuration obtained by the
orthogonal intersection of two equal rectangles with sides 2 a and 2 b . Introduce
the aspect ratio as ε= b / a The area and the moments will take the form
4 4
A = 4 a 2ε(2 − ε) , Ix = Iy = a ε(1 + ε2 − ε3),
3
J x = J y = 4 a ln[ε + (1 + ε2)1/2] + ε ln
1 + (1 + ε2)1/2
.
 (1 + √2)ε 
(5.5.38))
The coefficients will be determined as
4ε(1 + ε2 − ε3) 1 + (1+ε2)1/2-1

hx = hy = ln[ ε + (1+ ε 2 1/2
) ] + ε ln .
9ε(2 − ε)3/2  (1 + √2)ε 
(5.5.39)
The values of h = h x= h y were computed from (5.5.39) and compared with those
given by de Smedt (1979). The results are presented below
ε= 0.1000 0.2000 0.3333 0.4000 0.5000 0.6000 0.7500 0.8000 1.0000
de Smedt h = 1.5910 0.8720 0.6255 0.5725 0.5267 0.5069 0.4985 0.4997 0.5193
Formula (5.5.39) h = 1.7382 0.8758 0.6006 0.5465 0.5049 0.4890 0.4893 0.4926 0.5043
Discrepancy (%) -9.3 -0.4 4.0 4.5 4.1 3.5 1.9 1.4 2.9
Taking into consideration the shape complexity, we should consider the results
agreement as surprisingly good, not only quantitatively but qualitatively as well:
both sets of data display a relatively flat minimum around ε=0.75.
Discussion. It is noteworthy that the change of the order of integration
which led to (5.5.1) is valid inside the circle ρ≤ min { a (φ)} only. Nevertheless,
one can obtain from (5.5.1) the exact solution for an ellipse and sufficiently
accurate formulae for various punch planforms as it was demonstrated in the
preceding Examples.
The accuracy of formulae (5.5.20) can be improved by taking into
consideration the fifth harmonic (5.5.9) in combination with the variational
approach (Noble, 1960). The following functional assumes its maximum value at
the exact solution of (5.1.8)
σ( N )
σ( M ) w ( M )d S M − ⌠ ⌠ σ( M )⌠ ⌠ d S Nd S M.
2⌠ ⌠
I (σ) = (5.5.40)
H⌡ ⌡ ⌡⌡ ⌡ ⌡ R ( M,N ) 
S S S
Taking
σ( N )
H⌠ ⌠ d S ≈ w 1 + w 5, (5.5.41)
⌡ ⌡ R ( M,N) N
S
and substituting (5.5.3), (5.5.7), (5.5.9), and (5.5.41) in (5.5.40), one gets, after
integration with respect to ρ
2π

I = ⌠ ( a (φ))4( p 1cosφ + p 2sinφ) (αxsinφ − αycosφ)
4
⌡ 3 H
0 
π π 4
− ( p J + p 2 J xy)cosφ − ( p 1 J xy + p 2 J x)sinφ − ( a (φ))3([ p 1( A c6 + A c4)
3 1 y 3 63

+ p 2( A s6 − A s4)]cos5φ + [ p 1( A s6 + A s4) + p 2( A c4 − A c6)]sin5φ)dφ.

(5.5.42)
Considering now the functional I as a function of p 1 and p 2, the extremum
conditions
∂I ∂I
= 0 , = 0,
∂p1 ∂p2
give two linear algebraic equations with respect to the unknowns p 1 and p 2.
The complete solution is rather cumbersome. Here, we present the final result
for the coefficients h x and h y which are valid only for domains having at least
one axis of symmetry, and the central principal axes taken as the coordinate axes
32 I x 32 I y
hx = , hy = , (5.5.43)
3 A 3/2 J x(1 + ηx) 3 A 3/2 J y(1 + ηy)
where the correction terms
( B c4 − B c6)( A c4 − A c6) ( B c4 + B c6)( A c4 + A c6)

ηx = , ηy = ,
42π I x J x 42π I y J y
(5.5.44)
the parameters A c4 and A c6 are defined by (5.5.10), and
2π 2π
B c6 = ⌠ ( a (φ)) cos6φ dφ,

7
B c4 = ⌠ ( a (φ))7cos4φ dφ.
⌡ ⌡
0 0
Since expression (5.5.41) is approximate, there is no guarantee that (5.5.43) will

be more accurate than (5.5.20). We performed the necessary computations for a
rectangle. Here are the results compared with those of de Smedt (1979)
ε= 0.1000 0.2000 0.3333 0.5000 0.7500 0.8000 1.0000

de Smedt h x= 0.1287 0.1881 0.2531 0.3249 0.4240 0.4436 0.5193
Formula (5.5.43) h x= 0.1405 0.1988 0.2577 0.3207 0.4165 0.4376 0.5331
de Smedt h y= 4.1070 2.0260 1.2600 0.8892 0.6426 0.6130 0.5193
Formula (5.5.43) h y= 4.5856 2.0985 1.2479 0.8714 0.6463 0.6190 0.5331
Discrepancy in h x (%) -9.2 -5.7 -1.8 1.3 1.8 1.3 -2.7
Discrepancy in h y (%) -11.7 -3.6 1.0 2.0 -0.6 -1.0 -2.7
Comparison with similar data computed on the basis of formula (5.5.29) shows
that the correction terms ηx and ηy in this particular case resulted in decreasing
the value of discrepancy, positive as well as negative. We caution again that
there is no guarantee that this will be valid for an arbitrary domain. For
example, here are the data computed for a rhombus
ε= 0.1000 0.2000 0.3333 0.5000 0.7500 0.8000 1.0000
de Smedt h x= 0.1181 0.1729 0.2341 0.3052 0.4101 0.4323 0.5193
Formula (5.5.43) h x= 0.2268 0.1860 0.2351 0.3031 0.4058 0.4264 0.5091
de Smedt h y= 6.1820 2.7060 1.5240 0.9946 0.6703 0.6323 0.5193
Formula (5.5.43) h y= 8.5600 2.5916 1.4196 0.9408 0.6490 0.6138 0.5091
Discrepancy of h x (%) -92.0 -7.6 -0.4 0.7 1.0 1.4 2.0
Discrepancy of h y (%) -38.5 4.2 6.8 5.4 3.2 2.9 2.0
Comparison with the data computed due to (5.5.33) indicates that the discrepancy
decreased for ε≥0.2 while for ε=0.1 it has jumped in the opposite direction to
−92%. The main reason for this is a jump in the value of the coefficients ηx
and ηy when ε is very small. The following rule of thumb may be suggested
for the user wishing to improve the accuracy: when the value of the correction
coefficients ηx and ηy does not exceed small percentage of unity, their inclusion
generally results in an improvement in accuracy, otherwise one should not use
formulae (5.5.43).
It is worthwhile to give the solution due to (5.5.42) for the case when the
domain of contact has no axis of symmetry, and only the first harmonic of the
displacements w 1 is taken into consideration. The result is
αx( c 22 I xy − c 12 I x) + αy( c 12 I xy − c 22 I y)
p1 =
c 11 c 22 − c 122
(5.5.45)
αx( c 11 I x − c 12 I xy) + αy( c 12 I y − c 11 I xy)
p2 =
c 11 c 22 − c 122
where
πH πH
c 11 = ( J y I y + J xy I xy) , c 22 = ( J x I x + J xy I xy) ,
2 2
πH
c 12 = ( J xy( I x + I y) + I xy( J x + J y))
4
Formulae (5.5.45) are different from the equivalent set (5.5.12) derived earlier.
In the absence of any numerical data for a general domain, it is impossible to
say whether formulae (5.5.45) are more accurate than (5.5.12), but they are
definitely more complicated. It is noteworthy that in the case of a domain with
an axis of symmetry both (5.5.45) and (5.5.12) simplify to the same equations
(5.5.15).
One can notice a certain similarity between the formulae derived and those
related to the Saint-Venant theory of bending. This similarity will become more
evident if, for example, we rewrite equation (5.5.17) in the form
a (φ) M y M x
 P + 3 x − y .
σ =
 A 4 I x I y 
2 a 2(φ) − ρ2
1/2
 
We think that this similarity is not a pure coincidence since the method used in
this section can also be called semi-inverse. The method can be developed
further into a complete Saint-Venant type theory of elastic contact problems which
could combine the simplicity and accuracy sufficient for a practical engineer.
The mathematically identical problem of the magnetic polarizability of small
apertures was solved in (Fabrikant, 1987j).
Exercise 5.5
1. Find the moments of inertia and the J -moments for a circular sector
characterized by the radius r and the polar angle 2α.
Answer:
1 4 1 9α2 + 9αsinαcosα − 16sin2α

Ix = r (α − sin2α), I y = r4 ,
4 2 36α
2 2 k 2−1 1− k 2
Jx = r  2 E(γ, k ) − k sin3γ − (1 − k 2sin2γ )1/2sinγcosγ + F(γ, k )
3 k k2

α γ − α 
+ 3 k sinαcosα + cos(α + γ) + sin2α ln(cot cot ) ,
 2 2  
2  1 − k2
Jy = r  k sinγ(sin2γ − 3) + (1 − k 2sin2γ)1/2sinγcosγ − F(γ, k )
3 k2

+
1+ k 2
E( γ , k ) + 3 k sinα cos2α ln(cotαcotγ−α) − cosα − cos(α+γ).
k2  2 2  
Here k = 2sinα/(3α), and γ = tan-1(sinα/(cosα − k )).
2. Find the shape coefficients h x and h y for the circular sector described in the
Exercise 1.
Answer:
2(2α − sin2α) 1 − k 2
hx =  F(γ, k ) − k sin3γ − (1 − k 2sin2γ)1/2sinγcosγ
α 3/2
k 2

+
2 k 2−1
E( γ , k ) + 3 k sinα cosα + cos(α+γ) + sin2α ln(cotαcotγ−α)-1,
k2  2 2 

4(9α2 + 9αsinαcosα − 16sin2α)

 k sinγ(sin γ − 3)
2
hy =
9α 5/2

1 − k2 1 + k2
+ (1 − k 2sin2γ)1/2sinγcosγ − F( γ , k ) + E(γ, k )
k2 k2
γ − α -1
+ 3 k sinα−cosα − cos(α + γ) + cos2α ln(cot cot
α
)  .
 2 2  
3. Verify (5.5.4).
4. Derive (5.5.7).
5. Establish (5.5.14) −(5.5.16)).

5.6 Curved punch of general planform
In this section, we analyze the elastic contact problem for a curved punch
of non-elliptic planform under the action of a normal force. The punch base is
assumed to be a quadratic surface. Some general relationships are established
between the applied force and the punch settlement. Specific formulae are
derived for a punch whose planform has the shape of a polygon, a rectangle, a
rhombus and a cross. An example of a finite rigid cylinder lying on its
generator and pressed against an elastic half-space is considered in detail. The
method allows us to have singular tractions at the cylinder edges and zero
tractions at the rest of the contact domain boundary. The last condition serves
to determine the width of the domain of contact. All the formulae are checked
against the previously published solutions, and a good accuracy is confirmed for
a sufficiently wide range of values of aspect ratio.
Theory. Consider a punch with a curved base and arbitrary planform. The
punch is pressed against an elastic half-space by a normal force P . Let the
boundary of the domain of contact S be given in the polar coordinates as
ρ = a (φ),
where the function a (φ) is bounded and single-valued. Here we assume that the
domain of contact is prescribed. The case when the domain of contact is
unknown (or partially unknown) will be discussed further. The punch base is
assumed to be a quadratic surface. This limitation is not essential. The method
can be applied to higher order surfaces as well. The transformed governing
integral equation is given by (5.5.1).
Let the normal displacements under the punch be
w = g 0 + g x y 2 + g xy xy + g y x 2, (5.6.1)
where g 0 is the punch penetration, and g x, g y and g xy are the known constants
defined by the punch base geometry. Let the pressure distribution under the
punch be
a (φ)[α0 + ρ2(αxsin2φ + αxysinφ cosφ + αycos2φ)]

σ(ρ,φ) = , (5.6.2)
a 2(φ) − ρ2
1/2
 
where α0, αx, αy and αxy are the as yet unknown constants. We make use of
the condition that the integral of σ over S should be equal to P . This leads to
the expression
8
P = 2α0 A + (αx Ix + αxy Ixy + αy Iy ), (5.6.3)
3
where A is the area of the domain of contact S ; and Ix , Iy and Ixy are the
axial moments of inertia and the product of inertia respectively. The location of
coordinate origin can be defined from the condition that the tilting moments
produced by the tractions (5.6.2) should vanish. This condition yields the
following set of equations:
2π
⌠ a 3(φ)[α + 3 a 2(φ)(α sin2φ + α sinφ cosφ + α cos2φ)]sinφdφ = 0,
⌡ 0 4 x xy y
0
2π
⌠ a 3(φ)[α + 3 a 2(φ)(α sin2φ + α sinφ cosφ + α cos2φ)]cosφdφ = 0. (5.6.4)

⌡ 0 4 x xy y
0
The point whose coordinates satisfy (5.6.4) will be called the punch centre.
When the domain of contact has one axis of symmetry the punch centre is
located on this axis. In the case of two axes of symmetry, the punch centre
coincides with the centre of gravity. Now it is necessary to relate α0, αx, αy,
and αxy to the parameters g 0, g x, g y, and g xy. This can be done by substitution
of (5.6.2) into (5.5.1) which yields, after integration with respect to ρ0,
∞ ρ 2π
Σ ⌠ ein(φ-φ0) F 2−| n |, 1; 1; 1 − x dφ

xdx 2
w (ρ,φ) = H α0 ⌠ (x )|n| 2
⌡ ρ (ρ − x ) 0⌡
2 1/2
 2 2 a 2(φ0) 0
n=-∞
0
∞ ρ 2π
x 3d x
+ H Σ
n=-∞
⌠ (x )|n| 2
⌡ ρ (ρ − x ) ⌡
2 1/2
⌠ ein(φ-φ0) F (4−| n |, 1; 1;
2 2
0 0
x2
1− )(αxsin2φ0 + αxysinφ0cosφ0 + αycos2φ0)dφ0.
a (φ0)
2
Here F denotes the Gauss hypergeometric function. Further evaluation of the

normal displacements can be carried out separately for each harmonic. Note that
all the odd harmonics of w will be zero if a (φ) contains only even harmonics.
The zero th harmonic will take the form
2π
w 0(ρ,φ) = H⌠ 2α0 + ( a 2(φ0) +

π 1 2
ρ)
4 ⌡ 2
0
×(αxsin2φ0 + αxysinφ0cosφ0 + αycos2φ0) a (φ0)dφ0, (5.6.5)


which can be simplified as
π 
H 2α0 J0 + αx B x + αxy B xy + αy B y + (αx Jx + αxy Jxy + αy Jy ),
ρ2
w 0(ρ,φ) =
4  2 
(5.6.6)
where the J -moments were defined in the previous section, and the following
additional quantities were introduced:
2π 2π
B x = ⌠ a (φ)sin φ dφ ,
3 2
B y = ⌠ a 3(φ)cos2φ dφ ,
⌡ ⌡
0 0
2π
B xy = ⌠ a 3(φ)sinφ cosφ dφ. (5.6.7)

⌡
0
Since the tensor properties of the B -moments are similar to those of the
moments of inertia, we shall call B x and B y the cubic moments of a
two-dimensional domain about the axes Ox and Oy respectively, B xy will be
called the cubic product of a two-dimensional domain about the axes Ox and
Oy .
Here is the expression for the second harmonic
2π
π H ρ2 ⌠ (αxsin2φ0 + αxysinφ0cosφ0 + αycos2φ0) a (φ0)cos2(φ−φ0)dφ0,

3
w 2(ρ,φ) =
8 ⌡
0
which can be modified as
3
w 2(ρ,φ) = π H ρ2{−αx[( C − C ) cos2φ + 2 C sin2φ] + αy[( C
8 xxxx xxyy xxxy yyyy
− C )cos2φ + 2 C sin2φ] + αxy[( C − C )cos2φ + 2 C sin2φ]}.

xxyy xyyy xyyy xxxy xxyy
(5.6.8)
Here, the following geometrical characteristics of the domain of contact have been
introduced:
2π 2π
Cxxxx = ⌠ a (φ)sin4φdφ, Cxxxy = ⌠ a (φ)sin3φcosφdφ,

⌡ ⌡
0 0
2π 2π
Cxxyy = ⌠ a (φ)sin2φcos2φdφ, Cxyyy = ⌠ a (φ)sinφcos3φdφ,

⌡ ⌡
0 0
2π
Cyyyy = ⌠ a (φ)cos4φdφ. (5.6.9)

⌡
0
The C -moments will be called the linear moments of the fourth order. Their
relationships with the J -moments are easy to establish, for example, Jx = Cxxxx + Cxxyy ,
Jxy = Cxyyy + Cxxxy , etc . It is important to note that the parameter α0 did not appear
in (5.6.8), and the parameters αx, αxy and αy will not appear in the expression
for the fourth harmonic. The investigation of further harmonics shows that their
amplitude decreases. In the case of an ellipse they vanish thus making the
solution exact. It seems natural to assume w ≈ w 0+ w 2, and the remaining
harmonics may be called the solution error. We should also take into
consideration that the error sign fluctuation will result in even smaller error in
the integral characteristics sought, like the total force P . Now we have an
approximate expression for the displacements under the punch
π 
H 2α J + αx B x + αxy B xy + αy B y + x 2−αx( Cxxxx − 2 Cxxyy )
4  0 0
w =
 
+ αxy(2 Cxyyy − Cxxxy ) + αy(2 Cyyyy − Cxxyy ) + y 2αx(2 Cxxxx − Cxxyy )

 

+ αxy(2 Cxxxy − Cxyyy ) − αy( Cyyyy − 2 Cxxyy ) + 6 xy αx Cxxxy + αxy Cxxyy + αy Cxyyy .
  
(5.6.10)
Comparison of (5.6.1) and (5.6.10) leads to the following set of equations
π
H (2α0 J0 + αx B x + αxy B xy + αy B y) = g 0,
4
π 
H α (2 C − C ) + αxy(2 C − C ) − α y( C − 2 C ) = g x
4  x xxxx xxyy xxxy xyyy yyyy xxyy 
π 
H −α ( C − 2 C ) + αxy(2 C − C ) + α y(2 C − C ) = g y,
4  x xxxx xxyy xyyy xxxy yyyy xxyy 
3π 
H αx C + αxy C +αy C  = g xy.
2  xxxy xxyy xyyy 
(5.6.11)
The last three equations of (5.6.11) can be solved with respect to αx, αy,
and αxy. In the case of elastic contact problem when g 0 is prescribed, the value
of α0 can be found from the first equation (5.6.11), after which the total force
P is defined by (5.6.3), and the solution is completed. When the total force P
is given, the value of α0 can be determined from (5.6.3) after which the punch
penetration g 0 is given by the first equation (5.6.11).
A significant simplification occurs when the domain of contact S has at
least one axis of symmetry. In this case C = C = B xy=0. The last equation
xxxy xyyy
(5.6.11) becomes decoupled from the previous three. The solutions can be then
written explicitly
4[ g x(2 C − C ) + g y( C − 2C )]
yyyy xxyy yyyy xxyy
αx = ,
3π H ( C C − C 2xxyy)
xxxx yyyy
4[ g x( C − 2 Cxxyy ) + g y(2 Cxxxx − Cxxyy )]

xxxx
αy = ,
3π H ( C C − C 2xxyy)
xxxx yyyy
2 g xy
αxy = . (5.6.12)
3π H C
xxyy
Substitution of (5.6.12) into the first equation (5.6.11) gives
g xβx + g yβy
2  ,
α0 = g − (5.6.13)
π H J0  0 3( C C − C 2xxyy) 
xxxx yyyy
where
βx = B x(2 C − C ) + B y( C − 2C ),
yyyy xxyy xxxx xxyy
βy = B x( Cyyyy − 2 Cxxyy ) + B y(2 Cxxxx − Cxxyy ).
Expressions (5.6.2), (5.6.3), (5.6.12), and (5.6.13) give a complete and exact
solution for an ellipse. We shall prove them to perform well for an arbitrary
punch planform. We expect (5.6.2) to be reasonably accurate in the
neighbourhood of the coordinate origin, while the error might become quite
significant close to the boundary of the domain S .
It is convenient to discuss the following particular cases: g 0=2π√ A ,
g x= g y= g xy=0; g y=2π/√ A , g x= g xy= g 0=0; and the case g x=2π/√ A , g y= g xy= g 0=0. In
each case we compute the integral
H⌠ ⌠
p = σd S ,
A⌡ ⌡
S
which is proportional to the average value of σ, and is dimensionless, thus

characterizing the shape of S and being independent of its size. We shall
denote these parameters by p 0, p y and p x for each case respectively. There are
two reasons for these definitions: i ) they correspond exactly to the parameters
used in the theory of sound penetration through holes, so it would be easy to
compare the numerical results; ii ) tabulation of these parameters for various
shapes will simplify significantly the solution of any specific contact problem.
Indeed, the relationship between the applied force P and the punch penetration g 0
can be rewritten as
p
A  0
P = g + √ A ( p y g y + p x g x).
2π H √ A 0 
The last expression indicates that the knowledge of the shape coefficients and the
punch area is sufficient for establishing the relationship between the applied force
and the punch penetration. Formulae (5.6.3), (5.6.12) and (5.6.13) lead to the
following expressions for the parameters p 0, p y and p x
8{8 J0 [ Ix ( Cyyyy − 2 Cxxyy ) + Iy (2 Cxxxx − Cxxyy )] − 3 A B y}

8√ A
p0 = , py = .
J0 9 A 3/2 J0 ( Cxxxx Cyyyy − C 2xxyy)
8{8 J0 [ Ix (2 Cyyyy − Cxxyy ) + Iy ( Cxxxx − 2 Cxxyy )] − 3 A B x}

px = . (5.6.14)
9 A 3/2 J0 ( Cxxxx Cyyyy − C 2xxyy)
Some further simplifications take place when the domain S possesses such a
symmetry that all the moments about the axis Ox are equal to the similar
moments about the axis Oy . In this case
8(8 I 0 J0 − 3 A B 0)
py = px = , (5.6.15)
3 A 3/2 J 20
where the moments with the subindex 0 indicate corresponding polar moments.
Formula (5.6.13) also simplifies as follows:
B0
2 
α0 = g0 − ( g + g y). (5.6.16)
π H J0  J0 x 
Formulae (5.6.2), (5.6.3), (5.6.12) and (5.6.13) are the main results of this
section. The elastic contact problem for a wide variety of planforms can now
be solved by a relatively simple computation of the geometrical characteristics
(moments) of the domain of contact.
Several examples are considered further. We present only the necessary
computations of the moments involved. The domain of contact is assumed to be
prescribed. The more complicated case when the domain of contact is partially
unknown is discussed in the context of the problem of a rigid roller on an
elastic half-space.
Example 1: Polygon. Consider a polygon with n sides. The general

notation of its parameters is the same as in the previous section, where the
moments of inertia and the J -moments are computed. Here the remaining
moments are presented.
The C -moments can be computed by the following formulae

n
C
xxxx
= Σ−q cos2ψ
k=1
k k − u kcos4ψk + v ksin4ψk + 4 s ksinψkcos3ψk + 2 t kcos4ψk,
C
xxxy
= Σv cos4ψ
k=1
k k + u ksin4ψk + s kcos2ψk(1 − 4sin2ψk)
1
+ q ksin2ψk − 2 t ksinψkcos3ψk, (5.6.17)
2
Cxxyy = Σu cos4ψ
k=1
k k − v ksin4ψk −
1
s sin4ψk + 2 t ksin2ψkcos2ψk,
2 k
(5.6.18)
Cxyyy = Σ−v cos4ψ

k=1
k k − u ksin4ψk − s ksin2ψk(1 − 4cos2ψk)
1
+ q sin2ψk − 2 t ksin3ψkcosψk, (5.6.19)
2 k
n
Cyyyy = Σq cos2ψ
k=1
k k − u kcos4ψk + v ksin4ψk − 4 s ksin3ψkcosψk + 2 t ksin4ψk,
(5.6.20)
where
Ak b k + b k+1 + a k A 2k
1 1
tk = ln , sk = 4 2 ( − ),
ak b k + b k+1 − a k ak k b b k+1
Ak
1 1
qk = ( + )[ a 2k + ( b k − b k+1)2]. (5.6.21)
a 2k bk b k+1
 b 2k+1 − b 2k + a 2k3
Ak b 2k − b 2k+1 + a 2k 3
uk =  +   ,
12 a 4k  b k+1   bk 

16 A 4k
vk =  1 − 1 . (5.6.22)
3 a 4k b 3k+1 b 3k
The following formulae can be derived for the cubic moments

n
Bx = Σ−j cos2ψ
k=1
k k + r ksin2ψk + 2 f kcos2ψk,
By = Σj cos2ψ
k=1
k k − r ksin2ψk + 2 f ksin2ψk,
B xy = Σ(j
k=1
k − f k)sin2ψk + r kcos2ψk, (5.6.23)
where
2 A k 3 b k + b k+1 + a k 2Ak 2
jk =   ln , rk =   ( b − b ),
 a k  b k + b k+1 − a k  a k  k+1 k
b k+1 − b k 2
( b k + b k+1) A k1 +   + 1 j.
1
fk = (5.6.24)
4   ak   4 k
Substitution of (5.6.17−5.6.24) into (5.6.2, 5.6.3, 5.6.12 and 5.6.13) gives the
complete solution for an arbitrary polygon. In the case of a regular polygon
significant simplifications occur. The moments of inertia and the J -moments take
the form (5.5.25) and (5.5.26), and the remaining moments are determined by
nb 3 2π π 1 + sin(π/ n )
B x =B y = sin + cos3( )ln , (5.6.25)
4  n n 1 − sin(π/ n )
3 nb π 1 + sin(π/ n )
C = C = cos ln
xxxx yyyy 8 n 1 − sin(π/ n )
nb π 1 + sin(π/ n )
C = cos ln (5.6.26)
xxyy 8 n 1 − sin(π/ n )
Note that formulae (5.6.26) are valid for any regular polygon except the square,
due to the fact that the trigonometric series summation
n n
Σ
k=1
sin ( k −1)
4 2π
n
= Σcos (k−1)2nπ
k=1
4
=
3n
8
, (5.6.27)
is not valid for a square. The C -moments for a square of side 2 l can be
expressed as
= l 4ln(1 + √2) −
2√2 2√2
C = C , C = l.
xxxx yyyy  3  xxyy 3
(5.6.28)
Formulae (5.6.3, 5.6.12, and 5.6.13) simplify for a regular polygon
4(5 g x + g y) 4( g x + 5 g y) 16 g xy
αx = , αy = , αxy = . (5.6.29)
3π H J0 3π H J0 3π H J0
Again, one should note that formulae (5.6.29) are not valid for a square. The
formulae to follow are valid for any regular polygon including the square.
B0
2 
α0 = g0 − ( g x + g y),
π H J0  J0 
A B0
4 
Ag +  I 0 − ( g + g ).
8
P = (5.6.30)
π H J0  0 3 J0  x y

The dimensionless coefficients p y and p x take the form
2π π
23+7cos sin
8√2  n n .
py = px = −
2π π 1+sin(π/ n )  36 1+sin(π/ n )
( n sin )1/2 n cos ln ln
n n 1−sin(π/ n ) 1−sin(π/ n )
(5.6.31)
Consider several particular values of n . For an equilateral triangle ( n =3)

formula (5.6.31) gives p y= p x=0.3782. We are not aware of any numerical data
to compare with this result. In the case of a square n =4, and p y= p x=0.2697.
The result due to De Smedt is 0.2645, with the discrepancy less than 2%.
Since formula (5.6.31) in the limiting case n →∞ gives the exact result for a
circle p y= p x=4/(3π3/2)=0.2394, we should expect that the error of (5.6.31) will
decrease with n . The value of the coefficients for a regular hexagon is 0.2443,
and again, we did not find in the literature anything to compare with this result.
a 2 being its semiaxes along the axis Ox and Oy respectively. Introduce the
aspect ratio ε= a 2/ a 1. Formulae (5.6.17−5.6.24) in this case reduce to
4 4 3
Ix = a 1 a 23, Iy = a a ,
3 3 1 2
Jx = 4 a 1sinh-1ε , Jy = 4 a 2sinh-1(1/ε). (5.6.32)
ε ε
C = 4 a 1(sinh-1ε − ), C = 4a1 ,
xxxx 3(1 + ε2)1/2 xxyy 3(1 + ε2)1/2
1 1
C = 4 a 2(sinh-1 − ). (5.6.33)
yyyy ε 3(1 + ε2)1/2
B x = 2 a 31ε(1 + ε2)1/2 − sinh-1ε + 2ε3sinh-1( ),

1
 ε
B y = 2 a 31ε(1 + ε2)1/2 − ε3sinh-1( ) + 2sinh-1 ε.

1
(5.6.34)
 ε 
The coefficients p y and p x were computed by De Smedt (1979) for a rectangle
with various aspect ratios ε. Here, we present his results along with those given
by our method
ε= 0.1000 0.2000 0.3330 0.5000 0.7500 1.0000
De Smedt p y= 2.9980 1.3730 0.7942 0.5229 0.3491 0.2645
our method p y= 3.2809 1.3959 0.7782 0.5100 0.3485 0.2697
Discrepancy in p y (%) -9.4 -1.7 2.0 2.5 0.2 -2.0
De Smedt p x= 0.0376 0.0639 0.0982 0.1399 0.2022 0.2645
our method p x= 0.0284 0.0577 0.0963 0.1431 0.2086 0.2697
Discrepancy in p x(%) 24.6 9.7 1.9 -2.3 -3.2 -2.0
Our formulae seem to perform satisfactorily over a sufficiently wide range of

aspect ratio. The traction distribution due to (5.6.3) can be compared with the
numerical data received in a personal communication from De Smedt.
Computations were made for ε=0.5, g y=2π/√ A , H =1, g 0= g x=0. Here are the
results along the axis Ox , compared with those communicated by De Smedt
x/a 1= 0.0000 0.2500 0.3333 0.5000 0.5833 0.6667 0.7500 0.8333 0.9167
De Smedt σ= -0.4715 -0.3933 -0.3249 -0.1238 0.0452 0.2515 0.5456 0.9462 2.0580
our method σ= -0.4731 -0.3953 -0.3314 -0.1290 0.0232 0.2273 0.5141 0.9602 1.8556
Discrepancy(%) -0.3 -0.5 -2.0 -4.2 48.8 9.6 5.8 -1.5 9.8
The relevant results along the axis Oy are given below

y/a 2= 0.0000 0.1667 0.3333 .5000 .6667 .8333
De Smedt σ= -0.4715 -0.4765 -0.4774 -0.4837 -0.5063 -0.5311
our method σ= -0.4731 -0.4744 -0.4791 -0.4907 -0.5198 -0.6138
Discrepancy (%) -0.3 0.5 -0.3 -1.4 -2.7 -15.6
As we predicted, the discrepancy becomes quite significant close to the boundary.

Example 3: Rhombus. Let a 1 and a 2 be its semiaxes along Ox and Oy
respectively. We denote its side l =( a 21+ a 22)1/2, and introduce the aspect ratio
ε= a 2/ a 1. Formulae (5.6.17−5.6.24) in this case yield
l 4 ε3 l 4ε 2 l 2ε
Ix = , Iy = , A = .
3(1 + ε2)2 3(1 + ε2)2 (1 + ε2)
Jx =
4lε  1 − ε + ε2
ln
1 + ε + (1 + ε2)1/2 
,
(1 + ε2)  (1 + ε2)1/2 (1 + ε2) 1 + ε − (1 + ε2)1/2 
Jy =
4lε  − 1 − ε + 1
ln
1 + ε + (1 + ε2)1/2 
.
(1 + ε2)  (1 + ε2)1/2 (1 + ε2) 1 + ε − (1 + ε2)1/2 
(5.6.35)
Bx =
2l ε3 3
 ε + 4ε − 3 + 2 − ε ln1 + ε + (1 + ε ) ,
3 2 2 1/2
(1 + ε )  (1 + ε2)1/2
2 3
(1 + ε2) 1 + ε − (1 + ε2)1/2
By =
2 l 3ε  1 + 4ε2 − 3ε3 + ε2(2ε2 − 1)ln1 + ε + (1 + ε2)1/2.
(1 + ε2)3  (1 + ε2)1/2 (1 + ε2) 1 + ε − (1 + ε2)1/2
(5.6.36)
C =
4lε  2 − ε + 5ε − 4ε + 2
ε 1 + ε + (1 + ε ) 
3 4 2 1/2
2 ln ,
xxxx (1 + ε ) 
2 2
3(1 + ε )
2 1/2
(1 + ε ) 1 + ε − (1 + ε2)1/2
C =
4lε  −4 + 5ε − ε2 + 2ε3
+
1
ln
1 + ε + (1 + ε2)1/2
,
yyyy (1 + ε2)2  3(1 + ε2)1/2 (1 + ε2) 1 + ε − (1 + ε2)1/2
C =
4lε  1 − 2ε − 2ε 2 + ε 3 + ε2
ln
1 + ε + (1 + ε2)1/2
.
xxyy (1 + ε2)2  3(1 + ε2)1/2 (1 + ε2) 1 + ε − (1 + ε2)1/2
(5.6.37)
We did not find in the mechanics literature any result related to a punch with a
rhombus planform. The mathematically equivalent problem of sound penetration
through an aperture in the shape of a diamond was solved numerically by De
Smedt (1979). Here, we present his results compared with those given by our
method
ε= 0.1000 0.2000 0.3333 0.5000 0.7500 1.0000
De Smedt p y= 4.6520 1.8890 0.9844 0.5933 0.3655 0.2631
our method p y= 3.7425 1.6605 0.9192 0.5770 0.3661 0.2697
Discrepancy (%) 19.6 12.1 6.6 2.7 -0.2 -2.5
De Smedt p x= 0.0314 0.0549 0.0862 0.1270 0.1923 0.2631

our method p x= 0.1944 0.1435 0.1345 0.1532 0.2050 0.2697
Discrepancy (%) -518.4 -161.3 -56.0 -20.6 -6.6 -2.5
Though our results are satisfactory for p y, they are unacceptable for p x when
ε≤0.5. Despite the large relative errors of p y and p x for small ε, there is
reason to believe that the accuracy of solution of various contact problems will
be quite satisfactory due to the following: different signs in discrepancy will
partially compensate the total error; the term with p 0 generally dominates in real
contact problems, and it will generally determine the total error of the solution.
We shall see further that in the case of a rigid cylindrical roller our theory
works well for the aspect ratios ε very far away from unity. An alternative
approach which uses the variational principle and somewhat improves the
accuracy, is discussed further on.
Example 4: Cross. Consider a punch with a configuration obtained by the
orthogonal intersection of two equal rectangles with sides 2 a and 2 b , ( a ≥ b ).
Introduce the aspect ratio as ε= b / a . The area and some of the moments are
given in the previous section. The remaining moments are
 3
1 + (1+ε2)1/2 
B x = B y = 2 a 2ε(1+ε ) + ln[ε + (1+ε ) ] + ε ln
3 2 1/2 2 1/2
− √2.
  ε(1 + √2) 
(5.6.38)
The comparison between our results and those given by De Smedt (1979) are
presented below
ε= 0.1000 0.2000 0.3333 0.5000 0.7500 1.0000
De Smedt p y= p x= 0.9675 0.4854 0.3271 0.2671 0.2523 0.2645
our method p y= p x= 1.6943 0.6765 0.3716 0.2683 0.2517 0.2697
Discrepancy(%) -75.1 -39.4 -13.6 -0.5 0.3 -2.0
Taking into consideration the shape complexity, we should consider the results
agreement as surprisingly good, not only quantitatively but qualitatively as well:
both data display a relatively flat minimum around ε=0.75. The discrepancy
becomes unacceptably large for ε≤0.3. It will be shown later on that the
variational approach slightly improves the results.
Example 5: Rigid roller. Consider a rigid right cylinder of length 2 l and
radius r 0 lying on its generator and pressed against an elastic half-space by a
normal centrally applied force P . This case corresponds to g x= g xy=0 and
g y=−1/(2 r 0). All the previously derived formulae are valid here. Formulae
(5.6.12) yield
2( C − 2 Cxxyy )
yyyy
αx = − ,
3π r 0 H ( C C − C 2xxyy)
xxxx yyyy
2(2 Cxxxx − Cxxyy )

αy = − , αxy = 0. (5.6.39)
3π r 0 H ( Cxxxx Cyyyy − C 2xxyy)
Assuming that the contact region is close to a rectangle, we can use formulae
(5.6.32−5.6.34) for all the moments involved. We denote the width of the
contact region by 2 b . In the previously considered contact problems the contact
region was prescribed. Here, we have the domain of contact partially unknown.
Its length 2 l is prescribed, and the tractions should be singular at y =± l , while its
width 2 b is as yet unknown and is to be found from the condition that the
tractions vanish at x =± b .
We define α0 from the condition
α0 = −αy b 2. (5.6.40)
Returning back to (5.6.2), one can see that the condition (5.6.40) will cause σ
Fig. 5.6.1. The geometry pertinent to the rigid roller problem.]
to vanish along a part of an ellipse (see Fig. 5.6.1)
2 a (φ)pha/x y 2
x
+ = 1, (5.6.41)
b2 αy b 2
while the stresses will remain singular at y =± l . Direct computations show that
the ellipse semiaxis along Oy is at least several times greater then the other
semiaxis, thus making the part of the ellipse (5.6.41) very close to a straight
line and validating our assumption of a rectangular domain of contact. The
ellipse semiaxis ratio is smallest for a square where it is equal to 2.4612. In
this sense the case l = b is the least accurate and will be considered in more
detail further.
Formulae (5.6.3), (5.6.13) and (5.6.39) now give the following expressions
for the total force P and the punch penetration g 0
8
P = ( α x I x + α y I y ) − 2α y b 2 A , (5.6.42)
3
1
g0 = π H [αx B x + αy( B y − 2 J0 b 2)], (5.6.43)
4
where αx and αy are defined by (5.6.39) and the moments by (5.6.32−5.6.34),

and this completes the solution. We have found only one report (Kalker, 1972)
where some formulae were presented for the case of a very narrow domain of
contact with the aspect ratio ε= l / b >>1. These formulae in our notation take the
form
l3  1 − ln2 
P = 1 + , (5.6.44)
2ε r 0 H 
2 2ln(4ε) + 1
l2  
g0 = 2 2ln(4ε) + 1 . (5.6.45)
4 r 0ε  
We have received in private communication from Kalker some data generated by

his numerical method. Although his computer program does not give the value
of aspect ratio, we managed to compare the results by the following procedure.
Assuming equality of his punch settlement with ours (5.6.43), we can find the
value of aspect ratio ε which, being substituted into (5.6.42) allows us to
compare the total forces. The results of comparison are given in Table 5.6.1.
The agreement over a wide range of aspect ratios should be considered as very
good, taking into consideration the approximate nature of our theory, and also the
fact that Kalker’s program ignores the stress singularity at the roller edges. Such
a good agreement allows us to claim that formulae (5.6.42−5.6.43) give a
sufficiently accurate analytical solution to the problem of a rigid roller on an
elastic half space. Surprisingly enough, our method seems to be the most
accurate around ε=1, despite the fact that our assumption of a rectangular domain
of contact is the least accurate in this case. We have compared the traction
Table 5.6.1. Comparison of our method with the numerical results of Kalker
aspect punch total force P* total force P* discrepancy

ratio ε settlement g * (Kalker) formula (5.6.42) (%)
23.31 0.2501E-03 0.5379E-04 0.6251E-04 -16.20

10.77 0.1000E-02 0.2590E-03 0.2891E-03 -11.65
6.799 0.2250E-02 0.6618E-03 0.7174E-03 -8.409
4.879 0.4000E-02 0.1298E-02 0.1377E-02 -6.118
3.757 0.6251E-02 0.2200E-02 0.2296E-02 -4.389
2.511 0.1225E-01 0.4934E-02 0.5019E-02 -1.729
1.843 0.2026E-01 0.9110E-02 0.9114E-02 -0.4892E-01
1.282 0.3603E-01 0.1853E-01 0.1835E-01 0.9399
0.9653 0.5633E-01 0.3240E-01 0.3208E-01 1.015
0.6741 0.1003 0.6748E-01 0.6719E-01 0.4326
0.4174 0.2263 0.1925 0.1952 -1.393
0.3018 0.4041 0.4122 0.4216 -2.279
0.2359 0.6351 0.7499 0.7717 -2.908
0.1933 0.9212 1.231 1.272 -3.335
0.1633 1.265 1.882 1.952 -3.703
0.1411 1.670 2.734 2.844 -4.043
0.1239 2.139 3.817 3.982 -4.323
0.1101 2.679 5.173 5.411 -4.612
0.9884E-01 3.297 6.847 7.185 -4.942
0.8939E-01 4.000 8.890 9.362 -5.306
0.6806E-01 6.771 18.40 19.30 -4.901
distribution due to (5.6.2) with the numerical results of Kalker. The typical
picture is presented below for r 0/ l =20, ε=0.9653, H =1, y/l =0.1.
x/l = 0.1200 0.3600 0.6000 0.8400
Kalker σ= 0.7967E-02 0.7420E-02 0.6226E-02 0.3352E-02
our method σ= 0.7960E-02 0.7513E-02 0.6528E-02 0.4675E-02
discrepancy(%) 0.09 -1.25 -4.84 -39.49
As predicted, the relative error becomes quite significant close to the boundary of
the contact region. A similar behavior is observed along a line parallel to the
axis Ox :
y/l = 0.1000 0.3000 0.5000 0.7000 0.9000
Kalker σ= 0.7967E-02 0.8146E-02 0.8623E-02 0.9498E-02 0.1819E-01
our method σ= 0.7960E-02 0.8177E-02 0.8762E-02 0.1018E-01 0.1570E-01
discrepancy (%) 0.09 -0.38 -1.61 -7.17 13.67
It is also of interest to compare the numerical results by Kalker with his

approximate formulae (5.6.44−5.6.45). This comparison is given in Table 5.6.2.
We expected Kalker’s formulae to be the most accurate for large ε with the
Table 5.6.2. Comparison of formulae due to Kalker with his numerical results
aspect punch total force P* total force P* discrepancy

ratio ε settlement g * Kalker formula (5.6.44) (%)
22.34 0.2501E-03 0.5379E-04 0.5163E-04 4.018

10.27 0.1000E-02 0.2590E-03 0.2459E-03 5.054
6.456 0.2250E-02 0.6618E-03 0.6243E-03 5.658
4.621 0.4000E-02 0.1298E-02 0.1223E-02 5.743
3.551 0.6251E-02 0.2200E-02 0.2079E-02 5.511
2.368 0.1225E-01 0.4934E-02 0.4706E-02 4.621
1.734 0.2026E-01 0.9110E-02 0.8838E-02 2.987
1.197 0.3603E-01 0.1853E-01 0.1873E-01 -1.103
0.8848 0.5633E-01 0.3240E-01 0.3471E-01 -7.126
0.5772 0.1003 0.6748E-01 0.8364E-01 -23.95
accuracy decreasing with ε. This is not the case: the error is almost constant in
the interval 2<ε<22, it decreases to zero around ε=1, after which the error
changes sign and increases rapidly. This means either that the formulae of
Kalker (5.6.44−5.6.45) are not exact asymptotically, or that his numerical
procedure has an error of about 5%.
Discussion. An alternative method can be suggested by using the variational
approach (Noble 1960). We can use again the functional (5.5.40), which assumes
its stationary value at the exact solution of (5.1.8). We take
σ(N)
H⌠ ⌠ dS ≈ w 0 + w 2, (5.6.46)
⌡ ⌡ R(M,N) N
S
where σ is defined by (5.6.2) and w 0+ w 2 is given by (5.6.10). Substitution of

(5.6.1), (5.6.2), (5.6.10) and (5.6.46) into (5.5.40) makes it possible to consider
the functional I as a function of α0, αx, αy and αxy. The extremum conditions
∂I ∂I ∂I ∂I
= 0, = 0, = 0, = 0,
∂α 0 ∂α x ∂α y ∂αxy
give four linear algebraic equations with respect to the unknowns α0, αx, αy, and
αxy. The complete solution is rather cumbersome. Here, we present the set of
equations for the coefficients α0, αx, and αy which are valid only for domains
having at least one axis of symmetry.
4
c 11α0 + c 12αx + c 13αy = 4[ A g 0 + ( I g + Iy g y)],
3 x x
16 1
c 12α0 + c 22αx + c 23αy = [ Ix g 0 + ( Dxxxx g x + Dxxyy g y)],
3 5
16 1
c 13α0 + c 23αx + c 33αy = [ I g + ( Dxxyy g x + Dyyyy g y)]. (5.6.47)
3 y 0 5
Here,
c 11 = 2π H J0 A ,
1 4 4
c 12 = πH[ B x A + Ix (2 J0 + 2 C − C ) − I (C − 2 C )],
2 3 xxxx xxyy 3 y xxxx xxyy
1 4 4
c 13 = πH[ B y A + I (2 J + 2 C − C ) − I (C − 2 C )],
2 3 y 0 yyyy xxyy 3 x yyyy xxyy
4
c = π H [5 B x Ix + D (2 C − C ) − D (C − 2 C )],
22 15 xxxx xxxx xxyy xxyy xxxx xxyy
2
c 23 = π H [5( B x Iy + B y Ix ) − D ( C − 2C )
15 xxxx yyyy xxyy
+ 2D (C + C − Cxxyy ) − Dyyyy ( Cxxxx − 2 Cxxyy )],

xxyy xxxx yyyy
4
c 33 = π H [5 B y Iy + D (2 C − C ) − D (C − 2 C )]. (5.6.48)
15 yyyy yyyy xxyy xxyy yyyy xxyy
The D -moments are introduced similar to (5.6.9) as
2π 2π
Dxxxx = ⌠ a 6(φ)sin4φdφ, Dxxyy = ⌠ a 6(φ)sin2φcos2φdφ,

⌡ ⌡
0 0
2π
D = ⌠ a 6(φ)cos4φdφ. (5.6.49)
yyyy ⌡
0
It is clear that the variational approach solution is more cumbersome than the
one introduced earlier. It remains to be seen whether it will be more accurate.
One advantage should be noted: the matrix of (5.6.48) is symmetric (as it is
required by the reciprocal theorem) while the matrix of (5.6.11) is generally not
symmetric.
Let us compare the results for several particular configurations. First of
all, consider a regular polygon. In the tables hereafter the word simple refers to
the method introduced earlier, while the word variational refers to the solution of
the set of equations (5.6.47). In the case of a regular polygon we shall need
the polar D -moment only
b n sin 1 + cos4( ).

1 6 2π 4 π 8 π
D0 = cos2( ) +
5 n 3 n 3 n
Here are the results of computations for a regular polygon with n sides
n= 3 4 5 6 7 9 ∞
simple p y= p x= 0.3782 0.2697 0.2502 0.2443 0.2420 0.2403 0.2394
variational p y= p x= 0.3409 0.2612 0.2472 0.2429 0.2412 0.2401 0.2394
Discrepancy (%) 9.9 3.2 1.2 0.6 0.3 0.1 0.0
Both methods seem to work well. If one considers the result by De Smedt for
a square, 0.2645, as exact then this might be an indication that the variational
approach is somewhat more accurate. In the limiting case of n →∞ both
methods give the exact result for a circle.
The D -moments for the rectangle will take the form
24 24 5 8 3 3
Dxxxx = a a 5, Dyyyy = a a, Dxxyy = a a.
5 1 2 5 1 2 3 1 2
Here are the numerical results computed for a rectangle

ε= 0.1000 0.2000 0.3330 0.5000 0.7500 1.0000
De Smedt p y= 2.9980 1.3730 0.7942 0.5229 0.3491 0.2645
variational p y= 3.4239 1.4523 0.8023 0.5166 0.3437 0.2612
Discrepancy (%) -14.2 -5.8 -1.0 1.2 1.5 1.2
De Smedt p x= 0.0376 0.0639 0.0982 0.1399 0.2022 0.2645
variational p x= 0.0316 0.0588 0.0939 0.1370 0.1997 0.2612
Discrepancy (%) 15.9 7.9 4.3 2.1 1.2 1.2
Again, the general impression is that the variational approach is more accurate
but not in all cases. For example, the discrepancy in p y for ε=0.1 increased as
compared with the result from the simple method. It is up to the user to
decide whether a somewhat better accuracy of the variational approach is worth
more cumbersome computations.
The mathematically equivalent problem of the analytical determination of the
quadratic term in the low-frequency expansion, related to the sound transmission
through an aperture in a rigid screen, is considered in (Fabrikant, 1986d).
Exercise 5.6
1. Find the D -moments for a rhombus, with semiaxes a 1 and a 2.

Answer:
4 4 5 2 3 3
Dxxxx = a a 5, Dyyyy = a a, Dxxyy = a a.
5 1 2 5 1 2 15 1 2
2. Find the variational solution for a rhombus and compare it with the data of
De Smedt, and with the simple solution.
Answer: see the Table below
ε= 0.1000 0.2000 0.3333 0.5000 0.7500 1.0000
De Smedt p y= 4.6520 1.8890 0.9844 0.5933 0.3655 0.2631
variational p y= -0.5952 3.3549 0.9465 0.5580 0.3534 0.2612
Discrepancy (%) 112.8 -77.6 3.8 6.0 3.3 0.7
De Smedt p x= 0.0314 0.0549 0.0862 0.1270 0.1923 0.2631
variational p x= 0.0090 0.1464 0.1110 0.1400 0.1971 0.2612
Discrepancy (%) 71.5 -166.7 -28.8 -10.2 -2.5 0.7
Conclusion: the discrepancy decreased for ε>0.33, but the results are unacceptable
for ε<0.33.
3. Find the D -moments for a cross-shaped punch.

Answer:
24 6 8 6 3
D = D = a ε(1 + ε4 − ε5), D = a ε (2 − ε3).
xxxx yyyy 5 xxyy 3
4. Find the variational solution for a cross and compare it with the data of De
Smedt, and with the simple solution.
Answer: see the Table below
ε= 0.1000 0.2000 0.3333 0.5000 0.7500 1.0000
De Smedt p y= p x= 0.9675 0.4854 0.3271 0.2671 0.2523 0.2645
variational p y= p x= 1.4346 0.5822 0.3397 0.2606 0.2482 0.2612
Discrepancy (%) -48.3 -19.9 -3.9 2.4 1.6 1.2
Conclusion: Comparison of this table with the data from the simple solution leads
to the same conclusion: the results become valid in a wider range of the aspect
ratio ε, but the theory fails for very small ε.
5.7 Flat flexible punch of general planform 387
5. Apply the theory above to the case of a roller, with the aspect ratio ε=1.
Answer:
a (φ) l ln(1 + √2) −

1  x2 √2ln(1 + √2) − 1 y 2
1 − 2 −
 2√2 l 2√2ln(1 + √2) − 1 l 2 
σ =
π Hr0ln(1 + √2)[3ln(1 + √2) − √2] a 2(φ) − ρ2
1/2
 
0.4869 a (φ) l 1 − 2 − 
x2 y2
 l (2.4612 l )2 
≈
π Hr0 a 2(φ) − ρ2
1/2
 
2 l 3[12ln(1 + √2) − √2] 1.878 l 3

P = ≈ ,
9π Hr0 ln(1 + √2) [3ln(1 + √2) − √2] π Hr0
l 2[13ln2(1 + √2) − 6√2 ln(1 + √2) + 2] 1.06558 l 2

g0 = ≈
4 r 0(1 + √2) [3ln(1 + √2) − √2] r0
5.7 Flat flexible punch of general planform

under shifting load
The following mixed boundary value problem for a transversely isotropic

elastic half-space is considered: a uniform tangential displacement is prescribed
over a finite domain of general shape, while the remaining part of the boundary
is traction-free. The problem may be interpreted as an external crack problem,
with a remote shear loading, or as a contact problem, with a flexible punch,
subjected to a tangential displacement. We use the term flexible to indicate a
special kind of a punch which does not exert any normal pressure. A general
relationship is established between the resultant shifting force and the
displacement. A favorable comparison is made with the available numerical
results.
Theory. Consider a transversely isotropic elastic half-space z ≥0, with the

following mixed boundary condition prescribed on the plane z =0:
σz = 0, for −∞<( x , y )<∞; τ = 0, for ( x , y )∉ S ;
u = u ( x , y ), for ( x , y )∈ S . (5.7.1)
Here S denotes the domain of contact in the elastic contact problem, or for the
crossection of the connection between two elastic half-spaces in the external crack
problem. Let the boundary of the domain S be expressed in polar coordinates
as
ρ = a (φ), (5.7.2)
where a (φ) is a single-valued bounded function. The governing integral equation

can we rewritten from (2.6.2) as follows:
2π a (φ)0 2π a (φ)0
G1 τ(ρ0,φ0) G2 q τ(ρ0,φ0)
⌠⌠ ρ0dρ0dφ0 + ⌠⌠ ρ0dρ0dφ0 = u (ρ,φ).
2 ⌡⌡ R 2 ⌡⌡ q R
0 0 0 0
(5.7.3)
Here the overbar indicates the complex conjugate value, and
iφ
q = ρei φ − ρ0e 0, R2 = q q. (5.7.4)
The approach is based on the integral representation for the reciprocal

distance established in (1.1.27). We also need an integral representation for the
kernel in the second term of expression (5.7.3). It was established in (2.5.6).
We consider further only the case where the right hand side in (5.7.2) u =const.
Substitution of (1.1.27) and (2.5.6) into (5.7.2) gives, after changing the order of
integration and retaining the zero harmonic only
2π ρ a (φ0)
G1 τ(ρ ,φ )ρ dρ
⌠ dφ ⌠ ⌠ 0 0 0 dx
0
π ⌡ 0
⌡ (ρ2 − x 2)1/2 x⌡ (ρ20 − x 2)1/2
0 0
2π ρ a (φ0)
G2 2 i φ0
ρ20 − 2 x 2
+ ⌠ e dφ0 ⌠
dx
⌠ τ(ρ0,φ0)dρ0 = u .
π ⌡ ⌡ (ρ − x )
2 2 1/2 ⌡ ρ0(ρ20 − x 2)1/2
0 0 x
(5.7.5)
We assume the shear stress distribution in the form
c a (φ)
τ = , (5.7.6)
[ a (φ) − ρ2]1/2
2
where c is a complex constant. This loading is statically equivalent to a

resultant force T = T x+i T y. Integration of (5.7.6) over S yields
T = 2 Ac , (5.7.7)
where A is the area of the domain S . It is of interest to note that the

relationship (5.7.7) does not depend on the location of origin of the system of
coordinates. This location can be determined from the condition that the shear
tractions should not produce any torque. This leads to two equations
2π 2π
⌠ ( a (φ))3cosφ dφ = 0, ⌠ ( a (φ))3sinφ dφ = 0. (5.7.8)

⌡ ⌡
0 0
One can note that the left-hand side of each equation (5.7.8) is proportional to
the x or y coordinates of the center of gravity. This means that the origin of
the system of coordinates should be located at the center of gravity of the
domain S . The axis orientation will be discussed later.
It is now necessary to relate the tangential force T to the displacement u .
This can be done by substitution of (5.7.6-5.7.7) into (5.7.5) which yields, after
integration with respect to ρ0
2π 2π
π 2 iφ
u = G 1 T ⌠ a (φ0)dφ0 + G 2 T ⌠ e 0 a (φ0)dφ0. (5.7.9)
8A ⌡ ⌡ 
0 0
The complex expression (5.7.9) is equivalent to two real ones, namely,
π
ux = { T x[ G 1 J 0 − G 2( J x − J y)] + 2 T y G 2 J xy},
8A
π
uy = {2 T x G 2 J xy + T y[ G 1 J 0 − G 2( J y − J x)]}, (5.7.10)
8A
The J -moments were introduced in (5.5.8), and some of them were computed in
paragraph 5.6. One can now derive specific formulae for a variety of punch
shapes. We leave this exercise to the reader.
Note that the change of the order of integration in (5.7.5) is valid inside
the circle ρ=min( a (φ)) only, nevertheless, the solution given by (5.7.9) is exact in
the case of an ellipse. This is easy to explain. It is known (Willis, 1970) that
a traction distribution over an ellipse, given by a polynomial multiplied by
a (φ)/[ a 2(φ) − ρ2]1/2, produces a polynomial displacement. In our case the
displacement is constant. Since expression (5.7.9) gives the exact value of the
integrals in (5.7.2) for ρ=0, the result is exact all over the ellipse. Indeed,
formula (5.7.10) yields for an ellipse with semiaxes a and b ( a ≥ b )
1
G 1K + G 2
(2 − k 2)K − 2E
ux = T,
2a  k2  x
1
G 1K − G 2
(2 − k 2)K − 2E
uy = T,
2a  k2  y
where K and E are complete elliptic integrals of the first and second kind, with
the argument k =(1−( b / a )2)1/2. In the isotropic case G 1=(2−ν)/(2πµ), G 2=ν/(2πµ),
and the last result coincides with Mindlin’s (1949).
We can always choose the coordinate axes orientation so as to make J xy
vanish. In this case formulae (5.7.9) simplify as follows
π π
ux = [ G 1 J 0 − G 2( J x − J y)] T x, [ G 1 J 0 − G 2( J y − J x)] T y.
uy =
8A 8A
(5.7.11)
Since expressions (5.7.10) and (5.7.11) are exact for an ellipse, we may expect
them to be reasonably accurate for an arbitrary domain S . This assumption was
justified in previous paragraphs when compared with various numerical results
available in the literature.
In order to verify the accuracy of our method, some numerical computations
were performed by Kalker. Equations (5.7.11) may be rewritten as
T x = C x u x, T y = C y u y,
with
8A 8A
Cx = , Cy = .
π[ G 1( J x + J y) − G 2( J x − J y)] π[ G 1( J x + J y) + G 2( J x − J y)]
(5.7.12)
The values of C x and C y, due to formulae (5.7.12), are presented in the table
below, and compared with the numerical results of Kalker.
Computations were made for an isotropic body, with the shear modulus µ=0.5,
and the Poisson coefficient ν=0.3. The accuracy of the numerical procedure was
ε= 0.1 0.2 0.3 0.5 0.7 1.0

C x our method 5.16 6.12 6.87 8.1 9.19 10.67
C x Kalker 4.644 5.650 6.445 7.766 8.904 10.43
Discrepancy (%) 10. 7.7 6.2 4.1 3.1 2.3
C y our method 4.32 5.32 6.13 7.56 8.85 10.67

C y Kalker 3.991 5.044 5.904 7.373 8.667 10.43
Discrepancy (%) 7.6 5.2 3.7 2.5 2.1 2.3
assessed by computation of C x and C y for an ellipse, for which the exact

solution is well known. The exact solution was about 4% above the numerical
result by Kalker. All our results are also above the numerical ones. If we
assume that error pattern for a rectangle is the same as for an ellipse (this
means, for example, that our 10% discrepancy translates into 10−4=6(%) error),
then our formulae must be considered as surprisingly accurate over a wide range
of aspect ratios. We expect the error of our method to increase monotonically
with decreasing ε, since the assumed traction distribution (5.7.6) is less realistic
for a narrow rectangle than for a square. The fact that the discrepancy of C y
with the numerical results does not change monotonically, indicates some flaws in
the Kalker’s numerical procedure. We have also compared the shear traction
distribution due to (5.7.6), (5.7.7) and (5.7.12) for a square with a 1=4, along the
line y =0.5. The tangential displacement u x was assumed equal to unity, and
u y=0. The comparison is given in the table below.
Taking into consideration the approximate nature of both methods, the agreement
x= 0.5 1.5 2.5 3.5

τ our method 0.084 0.090 0.107 0.172
τ Kalker 0.0863 0.0923 0.105 0.222
Discrepancy (%) −2.6 −2.5 1.7 −29.
should be considered surprisingly good, except for the point close to the
boundary, where neither method may claim to be accurate.
Kalker has made computation for a cross, with a =4 and ε=0.5. The remaining
parameters were taken the same as for rectangle. His result C x= C y=9.033, our
result is 9.26, with the discrepancy of 2.4%. We have also compared the shear
stress distribution along the line y =0.5. The tangential displacement u x was
assumed equal to unity, and u y=0. The comparison is given in the table below.
τ our method 0.0996 0.104 0.124 0.199

τ Kalker 0.0997 0.106 0.120 0.254
Discrepancy (%) −0.08 −1.9 2.9 −27.
The agreement is good, except for the point close to the boundary.
Discussion. Some qualitative analysis of the general solution is given here. Let
us rewrite equations (5.7.11) as follows
πG1 J0 πG1 J0 G 2( J x − J y)
ux = [1 − κ] T x, uy = [1 + κ] T y, κ = .
8A 8A G1J0
(5.7.13)
First of all, consider the case when J x= J y. We have from (5.7.13) that κ=0,
and hence the elastic compliance will be the same in any direction. Due to the
fact that a circle has the greatest value of J 0 of all domains with the same area
A , one may come to the conclusion that a circular domain is the easiest to
move. A numerical comparison with a square shows that the ratio of the linear
polar moments of a square to that of a circle is equal to (2/√π )ln(1+√2 )=0.9945
which is very close to unity. Taking into consideration that our theory is
approximate, it is surprising that the difference can still be observed.
Comparison of a cross with a circle shows that its compliance may become
arbitrarily small as ε→0.
Now let us deform the domain S so that J x is no longer equal to J y
while keeping its area constant. Formulae (5.7.13) show that it will become
easier to move the domain in the oblong direction, and a greater force will be
required for a displacement in the perpendicular direction. In the extreme case
of a needle, the limiting value for κ is G 2/ G 1; in the case of isotropy
( G 2/ G 1)=ν/(2-ν)≤1/3. It is noteworthy that for G 2=0, the elastic compliance does
not depend on the direction of the shift. In the case of isotropy, this
corresponds to the Poisson ratio being equal to zero.
Exercise 5.7
1. Verify the invariance of formulae (5.7.10) with respect to the rotation of

axes.
2. Find the relationship between the shifting force and the translational
displacement of an elliptical punch, when its semiaxis a < b .
(2 − k 2)K( k 1) − 2E( k 1)
1   T ,
Answer: u x = G K( k 1) − G 2
2a 1  k12
 x
(2 − k 2)K( k 1) − 2E( k 1)
1   T .
uy = G 1K( k 1) + G 2
2a  k12
 y
3. Find the tangential compliance of a cross-shaped punch, with the longer side
2 a and the aspect ratio ε.
Hint : use (5.7.12) and (5.5.38)
4. Find the tangential compliance in the direction of the axes of symmetry for
a rhombus-shaped punch.
5.8 Reissner-Sagoci problem for general domains
The problem of torsion of a transversely isotropic elastic half-space by a

punch of general planform is considered. An approximate analytical solution is
obtained by the general method. A general relationship is established between
the torque and the torsion angle. Some specific formulae are derived for
punches having the planform of a polygon, a rectangle, and a cross.
Theory. Consider a transversely isotropic elastic half-space z ≥0. A flexible
punch of general planform S is attached to the half-space surface, and a torque
M z is applied, producing the torsion angle ω. We need to relate the torsion
angle to the torque. The mathematical formulation of the problem leads to the
following mixed boundary conditions on the plane z =0:
σz = 0, for −∞<( x , y )<∞; τzx = 0 and τyz = 0, for ( x , y )∉ S ;
u x = −ω y and uy = ωx, for ( x , y )∈ S . (5.8.1)
Here S denotes the domain subjected to torsion, and ω is the torsion angle.
Introduce the complex tangential displacements u = u x+i u y, and the complex
shear tractions τ=τzx+iτyz. Let the boundary of the domain S be expressed in
polar coordinates as
ρ = a (φ), (5.8.2)
where a (φ) is a single-valued bounded function. The governing integral equation

is given by (5.7.3). The approach is based on the integral representations
established in (1.1.27) and (2.5.6). Substitution of (1.1.27) and (2.5.6) into
(5.7.3) gives, after changing the order of integration and retaining the first
harmonic only
ρ 2π a (φ0)
x 2d x τ(ρ0, φ0) dρ0
u =
2
G ⌠ ⌠ cos(φ−φ ) dφ ⌠
πρ 1 ⌡ (ρ2 − x 2)1/2 ⌡ 0 0
⌡ (ρ20 − x 2)1/2
0 0 x
ρ 2π a (φ )
 x 2
d x i φ0
0
τ(ρ0, φ0) dρ0
+
1
G 2 −e
i φ⌠ ⌠ e dφ ⌠
πρ ⌡ (ρ2 − x 2)1/2 0⌡ ⌡ (ρ20 − x 2)1/2
0
 0 x
ρ 2π a (φ0)
x 2d x 3ρ20 − 4 x 2 
+ e ⌠ -i φ ⌠e 3 i φ0
dφ0 ⌠ τ(ρ0,φ0)dρ0.
⌡ (ρ2 − x 2)1/2 ⌡ ⌡ ρ20(ρ20 − x 2)1/2 
0 0 x
(5.8.3)
Note that the change of the order of integration in (5.8.3) is valid only inside
the circle ρ=min[ a (φ)], and the fact that we have ignored all the harmonics but
the first one. Nevertheless, it will be shown further that the results are exact
for an ellipse, and seem to be reasonable for a wide variety of nonelliptical
shapes.
Let the shear traction distribution under the punch be
a (φ) ρ(i q ycosφ − q xsinφ)

τ = τzx + iτyz = , (5.8.4)
[ a 2(φ) − ρ2]1/2
where q y and q x are the as yet unknown constants. Make use of the condition
that in the case of pure torsion the resulting force should be equal to zero,
which means that the integral of τ over S should vanish. Since q y and q x are
independent, this leads to two equations, namely,
2π 2π
⌠ ( a (φ))3cosφ dφ = 0, ⌠ ( a (φ))3sinφ dφ = 0 .
⌡ ⌡
0 0
(5.8.5)
One can note that the left-hand side of each equation (5.8.5) is proportional to
the x or y coordinates of the center of gravity. This means that the origin of
the system of coordinates should be located at the center of gravity of the
domain of contact. The axis orientation will be discussed later.
The relationships between the torque M z and the parameters q y and q x can
be established from the statics conditions
M z = ⌠ ⌠ τyz x dS − ⌠ ⌠ τzx y dS,

⌡⌡ ⌡⌡
S S
which leads to
8
Mz = ( q I + q x I x). (5.8.6)
3 y y
where I x and I y are the moments of inertia.

Next, it is necessary to relate q y and q x to the torsion angle ω. This can
be done by substitution of (5.8.4) in (5.8.3), which yields, after integration with
respect to ρ0
2π
G ρ⌠ cos(φ−φ0)(i q ycosφ0 − q xsinφ0 )dφ0

π
u =
4 1 ⌡
0
2π
π iφ
− G 2ρei φ⌠ a (φ0)(−i q ycosφ0 − q xsinφ0)e 0 dφ0
8 ⌡
0
2π
G 2ρe-i ⌠ a (φ0)(−i q ycosφ0 − q xsinφ0)e 0 dφ0.

3π φ 3iφ
+ (5.8.7)
8 ⌡
0
The overbar everywhere denotes the complex conjugate value. Expression (5.8.7)
can also be rewritten as follows:
π
u = G ρ[(i q y J y − q x J xy)cosφ + (i q y J xy − q x J x)sinφ]
4 1
π φ
− G 2ρei [− q y(i J y − J xy) − q x( J xy + i J x)]
8
3π φ
+ G 2ρe-i {−i q y[4 C − 3 J y + i(3 J xy − 4 C xxxy)]
8 yyyy
− q x[4 C xyyy − 3 J xy + i(3 J x − 4 C xxxx)]} (5.8.8)
The J -moments were introduced by (5.5.8), and the C -moments are defined by
(5.6.9). Substitution of the last two conditions (5.8.1) in (5.8.8) yields
π
ω(− y + i x ) = G [(i q y J y − q x J xy) x + (i q y J xy − q x J x) y ]
4 1
π
− G ( x + i y )[− q y(i J y − J xy) − q x( J xy + i J x)]
8 2
3π
+ G 2( x − i y ){−i q y[4 C yyyy − 3 J y + i(3 J xy − 4 C xxxy)]
8
− q x[4 C xyyy − 3 J xy + i(3 J x − 4 C xxxx)]} (5.8.9)
Separation in (5.8.9) of the terms related to x and y will lead to two linear
algebraic equations for the unknowns q y and q x. In the general case, the
parameters q y and q x are complex, and one has to solve four linear algebraic
equations. These equations may be simplified by the assumption that J xy=0.
One can always achieve this by appropriately choosing the coordinate axis
directions. In the case of symmetry, these axes will coincide with the principal
axes of inertia. For the sake of simplicity, we assume also that C = C =0,
xyyy xxxy
which will always be the case if domain S has at least one axis of symmetry.
Under these assumptions, expression (5.8.9) yields just two equations with real
coefficients, namely,
π 5 3 3
G q J + π G 2[ q y( J y − C yyyy) + q x( C xxxx − J x)] = ω,
4 1 y y 4 2 2
π 3 5 3
G 1 q x J x + π G 2[ q y( C yyyy − J y) + q x( J x − C xxxx)] = ω.
4 2 4 2
(5.8.10)
The solution of (5.8.10) is
4ω[ G 1 + 3 G 2(3 − 4 c x)]

qy = ,
π( G 1 + G 2) J y[ G 1 + 3 G 2(3 − 2 c x − 2 c y)]
4ω[ G 1 + 3 G 2(3 − 4 c y)]

qx = .
π( G 1 + G 2) J x[ G 1 + 3 G 2(3 − 2 c x − 2 c y)]
(5.8.11)
Here the parameters were introduced
c x = C xxxx/ J x, c y = C yyyy/ J y. (5.8.12)
By substituting (5.8.11) in (5.8.6), one can find the relationship between the
torque and the angle of torsion in the form
32ω[( G 1 + 9 G 2)( I y/ J y + I x/ J x) − 12 G 2( c x I y/ J y + c y I x/ J x]
Mz = .
3π( G 1 + G 2)[ G 1 + 3 G 2(3 − 2 c x − 2 c y)]
(5.8.13)
One can verify that the solution (5.8.11) and (5.8.13) is exact for an ellipse.
Indeed, consider an ellipse with semiaxes a and b ( a ≥ b ). The necessary
geometrical characteristics are
J x = 4 b [E( k ) − (1 − k 2)K( k )]/ k 2, J y = 4 b [K( k ) − E( k )]/ k 2,
C xxxx = 4 b [2(2 k 2 − 1)E( k ) + (1 − k 2)(2 − 3 k 2)K( k )]/(3 k 4),
C yyyy = 4 b [(2 + k 2)K( k ) − 2(1 + k 2)E( k )]/(3 k 4), k = [1−( b / a )2]1/2.

(5.8.14)
Here K( k ) and E( k ) are the complete elliptic integrals of the first and the
second kind respectively. Substitution of (5.8.14) in (5.8.11) and (5.8.13) yields
ω{ G 1 k 2[E − (1− k 2)K] + G 2[(8−7 k 2)E − (1− k 2)(8−3 k 2)K]}

qy = ,
π b ( G 1+ G 2){ G 1(K−E)[E −(1− k 2)K] + G 2[ k 2K(K+E)−(K−E)(K+3E)]}
ω{( G 1 k 2 − 8 G 2)[K − E] + G 2 k 2[5K − E]

qx = ,
π b ( G 1+ G 2){ G 1(K−E)[E −(1− k 2)K] + G 2[ k 2K(K+E)−(K−E)(K+3E)]}
(5.8.15)
2ω a { G 1 k E − G 2[8(1− k )(2− k )K − ( k −16 k +16)E]}
3 2 2 2 4 2
Mz = .
3( G 1+ G 2){ G 1(K−E)[E −(1− k 2)K] + G 2[ k 2K(K+E)−(K−E)(K+3E)]}
(5.8.16)
The abbreviations E and K in (5.8.15) and (5.8.16) stand for E( k ) and K( k )
respectively. The same results can be obtained by direct substitution of (5.8.4)
in (5.7.3), with an exact computation of the integrals involved, by using relevant
formulae from Appendix A5.1. In the case of isotropy, G 1=(2−ν)/(2πµ),
G 2=ν/(2πµ), and formulae (5.8.15) and (5.8.16) simplify as follows:
µω{ k 2[E − (1− k 2)K] + 2ν(1− k 2)[2E − (2− k 2)K]}

qy = ,
b {[K − E][E − (1− k 2)K] + νE[2E − (2− k 2)K]}
µω{ k 2[K − E] + 2ν[2E − (2 − k 2)K]}

qx = , (5.8.17)
b {[K − E][E − (1− k 2)K] + νE[2E − (2− k 2)K]}
2πµω a 3{ k 4E + 4ν(1 − k 2)[2E − (2 − k 2)K]}

Mz = . (5.8.18)
3{[K − E][E − (1− k 2)K] + νE[2E − (2− k 2)K]}
Formulae (5.8.17) and (5.8.18) are in agreement with the corresponding results of
Mindlin (1949). Some misprints are noticed in the relevant formula of (Kassir
and Sih, 1968). The result by Willis (1970) seems to be in error, since it
indicates that the parameters q y and q x do not depend on elastic constants, which
is incorrect.
In the case when J x= J y and c x= c y, formulae (5.8.11) and (5.8.13) simplify
significantly, namely,
64 I 0ω
8ω
qy = qx = , Mz = (5.8.19)
π J 0( G 1 + G 2) 3π J 0( G 1 + G 2)
Formulae (5.8.11) and (5.8.13) are the main results of this section. They are
exact for an ellipse, and there is reason to believe that they will be sufficiently
accurate for a general shape. The derivation of specific formulae for any
particular shape is left to the reader.
Example 1: Regular polygon. Consider a punch shaped as a regular
polygon with n sides. The necessary moments were computed in paragraph 5.6.
We can use expressions (5.8.19), with the result
8ω
qy = qx = ,
π 1 + sin(π/ n )
π nb ( G 1 + G 2)cos ln
n 1 − sin(π/ n )
32ω b 3sin(π/ n )[2 + cos(2π/ n )]

Mz = . (5.8.20)
1 + sin(π/ n )
9π( G 1 + G 2) ln
1 − sin(π/ n )
In the limiting case of n →∞, formulae (5.8.20) give the results for a circular
punch
4ω 16ω b 3
qy = qx = , Mz = (5.8.21)
π b ( G 1 + G 2)
2 3π( G 1 + G 2)
Introduce the stiffness parameter

D = M z( G 1 + G 2)/ω = 64 I 0/(3π J 0), (5.8.22)
which is a geometric characteristic of domain S . If we take the ratio of the

stiffness of a regular polygon D to the stiffness D of a circle, having the
p c
same area, the result is:
Dp 2sin(π/ n )[2 + cos(2π/ n )]

= . (5.8.23)
Dc 1 + sin(π/ n )
3ln
1 − sin(π/ n )
Elementary analysis of (5.8.23) shows that an equilateral triangle has stiffness

1.24 times greater than that of a circle having an equal area. The corresponding
ratio for a square is 1.05, and further increase in n makes the polygon stiffness
practically indistinguishable from that of a circle. One may prove a theorem
stating that of all simply connected domains, having the same area, the circle has
the lowest stiffness. This is opposite to the relevant theorem in the Saint-Venant
theory of torsion of rods, where a circle has the greatest stiffness. This is easy
to explain since in the torsion of a half-space, one has to twist not only an
imaginary rod, but its surroundings as well.
a 2 being its semiaxes along the axis Ox and Oy respectively. Introduce the
aspect ratio ε= a 2/ a 1. The area and the necessary moments were computed in
paragraph 5.6. Formulae (5.8.11) and (5.8.13) for a rectangle take the form
ω[( G 1 − 3 G 2)(1 + ε2)1/2sinh-1ε + 4 G 2ε]

qy = ,
π a 1ε( G 1+ G 2){( G 1−3 G 2)(1+ε2)1/2sinh-1εsinh-1(1/ε)+2 G 2[εsinh-1(1/ε)+sinh-1ε]}
ω[( G 1 − 3 G 2)(1 + ε2)1/2sinh-1(1/ε) + 4 G 2]

qx = ,
π a 1( G 1+ G 2){( G 1−3 G 2)(1+ε2)1/2sinh-1εsinh-1(1/ε)+2 G 2[εsinh-1(1/ε)+sinh-1ε]}
(5.8.24)
32ω a 31{( G 1 − 3 G 2)(1 + ε ) [ε sinh (1/ε) + sinh ε] + 4 G 2ε(1 + ε )}
2 1/2 3 -1 -1 2
Mz = .
9( G 1+ G 2){( G 1−3 G 2)(1+ε2)1/2sinh-1ε sinh-1(1/ε)+2 G 2[εsinh-1(1/ε)+sinh-1ε]}
(5.8.25)
In order to verify the accuracy of our theory, some computations were made by
Kalker for a rectangle with a 1=4 and various aspect ratios. The half-space was
assumed to be isotropic, with the shear modulus µ=0.5, and the Poisson
coefficient ν=0.3. The quantity C z= M z/ω was computed by using the universal
software developed by Kalker. Our results due to (5.8.25), compared with
numerical results by Kalker, are presented in the table below.
ε= 0.1 0.2 0.3 0.5 0.7 1.0

C z our method 48.3 62.9 77.9 113.7 160.6 258.
C z Kalker 36.74 51.29 66.47 102.4 148.4 241.
Discrepancy (%) 23.9 17.5 14.7 9.9 7.6 6.6
The accuracy of the numerical procedure was assessed by computation of C z for

an ellipse, where the exact solution is well known. The exact solution was
about 7% above the numerical result by Kalker. All our results are also above
the numerical ones. If we assume that error pattern for a rectangle is the same,
as for an ellipse, then our formulae must be considered as surprisingly accurate
over a wide range of aspect ratio.
We have also compared the shear stress distribution due to (5.8.4), and
(5.8.24) for a rectangle, with a 1=4 and ε=0.3, along the line x =0.5. The torsion
angle was assumed equal to unity. The shear traction is presented in polar
form. The comparison for the modulus |τ| and its argument arg (τ) in degrees is
given in the table below
y= 0.15 0.45 0.75 1.05
|τ| our method 0.139 0.219 0.375 0.806
|τ| Kalker 0.145 0.228 0.366 1.11
Discrepancy (%) −1.9 −4.2 2.3 −37.7
arg (τ) our method 112.57 141.27 154.3 161.03

arg (τ) Kalker 112.14 140.72 152.94 161.15
The agreement of the argument of τ is very close. The agreement in modulus

is satisfactory, except for the point close to the boundary where neither method
may claim to be accurate. The fact that the discrepancy does not change
monotonically suggests some flaws in the Kalker’s numerical procedure.
Example 3: Cross. Consider a punch with a configuration obtained by the

orthogonal intersection of two equal rectangles with sides 2 a and 2 b ( a ≥ b ).
Introduce the aspect ratio as ε = b / a . The area and the moments are given in
paragraph 5.5.
Formulae (5.8.11) and (5.8.13) in this case yield
ω
qy = qx = ,
π a ( G 1+ G 2) ln(ε + (1+ε2)1/2) + εln 
1 + (1+ε ) 2 1/2
 (1 + √2)ε 
64ω a 3ε(1 + ε2 − ε3)

Mz = .
9π( G 1 + G 2)ln[ε + (1+ε2)1/2] + εln
1 + (1+ε ) 
2 1/2
 (1 + √2)ε 
Kalker has made computation for a cross, with a =4 and ε=0.5. The remaining
numerical data was taken the same as for rectangle. His result C z=154.7, our
result is 167.9, with the discrepancy of 7.9%. Taking into consideration the
assumed error of Kalker’s software being 7%, our result might be considered
very accurate.
We have also compared the shear traction distribution along the line y =0.5.
The torsion angle ω was assumed equal to unity. The comparison is given in
the table below.
x= 0.5 1.5 2.5 3.5
|τ| our method 0.120 0.280 0.535 1.12
|τ| Kalker 0.118 0.287 0.512 1.47
Discrepancy (%) 1.5 −2.5 4.4 −23.
arg (τ) our method 135. 108.4 101.3 98.13

arg (τ) Kalker 135. 108.0 101.1 98.72
The agreement is good, except for the point close to the boundary.
Exercise 5.8
2. Derive (5.8.8).
3. Verify (5.8.11) and (5.8.13).
4. Consider the torsion of a half-space by a rhomboidal punch.
5. Compare the torsional rigidity of a rectangular punch with the torsional

rigidity of a rhomboidal one, having the same area and the same aspect ratio.
5.9 Interaction between punches subjected to normal pressure
A general theorem is established which relates the resultant forces, acting on

a set of arbitrary punches, with their generalized displacements through a system
of linear algebraic equations. The theorem is applied to the case of arbitrarily
located elliptical punches. Several specific examples are considered.
Theory. Consider a set of N arbitrary punches penetrating an elastic
half-space z ≥0. Let Sn be the domain of contact for the n th punch, and P n be
the normal force acting on the n th punch. The friction forces between the
punches and the half-space are neglected. The problem is to find the
relationships between the generalized displacements of the punches and the acting
forces. The boundary conditions for the problem are
w = w n( M ) for M ∈ Sn ,
σz( M ) = 0 for M ∉ Sn , n =1,2,..., N , (5.9.1)
where w denotes the normal displacement of a point at the boundary z =0, and σ
stands for the pressure distribution. The prescribed function w n is defined by the
punch face. By using the known solution of the Boussinesq problem and the
principle of superposition, we can write
N
σ (T )
w( Q) = H Σ ⌠ ⌠ n n dS .
⌡ ⌡ R ( T n, Q ) n
(5.9.2)
n=1 Sn
Substitution of the boundary conditions (5.9.1) in (5.9.2) leads to a set of N

integral equations. The exact solution of these equations is not known at the
present time, even for the case of several circles. Here, we are to show that
we do not really need to know these solutions if we are interested in the
integral characteristics only. We can single out, without loss of generality, the
first punch, and consider the related integral equation
σ (T ) N
σ (T )
w 1( Q 1) = H ⌠ ⌠ dS1 + H ⌠ ⌠ Σ
1 1 n n
dS . (5.9.3)
⌡ ⌡ R ( T 1, Q 1) ⌡ ⌡ R ( T n, Q 1) n
S1 n=2 Sn
Suppose that the functions σ0, σx and σy are known, satisfying respectively the
following integral equations inside S 1
σ (Q )
⌠ ⌠ 0 1 d S = 1, (5.9.4)
⌡ ⌡ R ( T 1, Q 1) 1
S1
σ (Q )
⌠ ⌠ x 1 dS = x, (5.9.5)
⌡ ⌡ R ( T 1, Q 1) 1
S1
σ (Q )
⌠ ⌠ y 1 dS = y. (5.9.6)
⌡ ⌡ R ( T 1, Q 1) 1
S1
Multiplication of both sides of (5.9.3) by σ0( Q 1) and integration over the area S 1
yields
σ (T )
⌠ ⌠ σ ( Q ) w ( Q ) d S = H ⌠ ⌠ σ ( Q )d S ⌠ ⌠ 1 1 d S +
⌡⌡ 0 1 1 1 1
⌡ ⌡ 0 1 1 ⌡ ⌡ R ( T 1, Q 1) 1
S1 S1 S1
N
σ (T )
+H Σ ⌠ ⌠ σ ( Q )d S ⌠ ⌠ n n d S .
⌡ ⌡ 0 1 1 ⌡ ⌡ R ( T n, Q 1) n
(5.9.7)
n=2 S1 Sn
By interchanging the order of integration in (5.9.7) and taking into consideration

the fact that σ0 satisfies (5.9.4), the following result can be obtained:
⌠ ⌠ σ ( Q ) w ( Q )d S = H P +
⌡⌡ 0 1 1 1 1  1 Σ⌠⌡ ⌠⌡ w 1n
( T n)σn( T n)d Sn ,

S1 n=2 Sn
(5.9.8)
where P 1 is the total force acting on the first punch, and
σ (Q )
w 1n( T n) = ⌠ ⌠
0 1
dS , (5.9.9)
⌡ ⌡ R ( T n, Q 1) 1
S1
which is proportional to the normal displacement in the domain Sn due to a flat

punch in S 1 under the action of a unit force. By invoking the mean value
theorem, which is valid as long as σn does not change sign, we obtain the
linear algebraic equation
⌠ ⌠ σ ( Q ) w ( Q )d S = H P +
⌡⌡ 0 1 1 1 1  1 Σw 1n
( C n) P n

(5.9.10)
S1 n=2
The exact location of the point C n is not known but the fact that C n∈ Sn allows
only a limited variation within Sn , and in many cases provides sufficiently close
upper and lower bounds for the parameters sought. By using the same
procedure, N −1 additional linear algebraic equations can be derived for the
remaining punches. This set of equations provides the necessary relationships
between the normal displacements of the punches and the applied forces.
Let us derive similar relationships for the angular displacements.
Multiplication of both sides of (5.9.3) by σx( Q 1) and integration over the area S 1
yields
σ (T )
⌠ ⌠ σ ( Q )w ( Q )d S = H⌠ ⌠ σ ( Q )d S ⌠ ⌠ 1 1 d S +
⌡⌡ x 1 1 1 1 ⌡ ⌡ x 1 1 ⌡ ⌡ R ( T 1, Q 1) 1
S1 S1 S1
N
σ (T )
+ H Σ ⌠ ⌠ σ ( Q )d S ⌠ ⌠ n n d S .
⌡ ⌡ x 1 1 ⌡ ⌡ R ( T n, Q 1) n
(5.9.11)
n=2 S1 Sn
By changing the order of integration in (5.9.11) and taking into consideration that
σx satisfies (5.9.5), the following result can be obtained
⌠ ⌠ σ ( Q ) w ( Q )d S = H  − M +
⌡⌡ x 1 1 1 1  1y Σ ⌠ ⌠ x α ( T )σ ( T )d S ,
⌡ ⌡ 1n n n n n 
S1 n=2 Sn
(5.9.12)
where M 1y is the tilting moment acting on the first punch about the axis Oy ,
and
σ (Q )
1⌠ ⌠ x 1
α1n( T n) = d S 1. (5.9.13)
x ⌡ ⌡ R ( T n, Q 1)
S1
If σn does not change sign we can use again the mean value theorem to yield
⌠ ⌠ σ ( Q ) w ( Q )d S = H  −M +
⌡⌡ x 1 1 1 1  1y Σα 1n
( A n) x n P n

(5.9.14)
S1 n=2
where An(xn,yn) ∈ Sn . The additional N −1 equations relating the punch rotations

about the axis Oy with the corresponding tilting moments can be obtained in a
similar way.
Multiply both sides of (5.9.3) by σy( Q 1) and integrate over S 1. The
procedure leads to the equation
N
⌠ ⌠ σ ( Q ) w ( Q )d S = H M +
⌡⌡ y 1 1 1 1  1x Σ ⌠ ⌠ y β ( T )σ ( T )d S ,
⌡ ⌡ 1n n n n n 
S1 n=2 Sn
(5.9.15)
where M is the tilting moment about the axis Ox acting on the first punch,
1x
and
σ (Q )
1⌠ ⌠ y 1
β1n( T n) = d S 1. (5.9.16)
y ⌡ ⌡ R ( T n, Q 1)
S1
As before, application of the mean value theorem gives

N
⌠ ⌠ σ ( Q ) w ( Q )d S = H M +
⌡⌡ y 1 1 1 1  1x Σβ 1n
( B n) y n P n.

S1 n=2
(5.9.17)
Three sets of linear algebraic equations of the type (5.9.10), (5.9.14) and (5.9.17)
are the main results of this paragraph. It is clear that each equation can be
interpreted in terms of the reciprocal work. In order to use them, one need to
know the normal displacements outside every punch in the system due to three
types of loading which, at the moment, is available for the elliptical punches
only. This particular case is considered below.
Application to elliptical punches. Consider the interaction of a set of N
flat elliptical punches arbitrarily located on a transversely isotropic elastic
half-space. Let a n and b n be the major and the minor semiaxes of the n th
ellipse; X n and Y n define its center, and θn be the angle between the axis Ox
and the major semiaxis a n; and P n be the normal force acting upon the n th
punch.
The functions σ0, σx and σy have the form (Lur’e, 1955):
σ0 =
1 1 − x 2 − y 2 -1/2
2π b 1K( k 1)  a 21 b 21
σx =
x x2 y 2 -1/2
1 − 2 − 2
2π b 1D( k 1)  a1 b 1
σy =
y 1 − x 2 − y 2 -1/2
2π b 1B( k 1)  a 21 b 21
The boundary conditions (5.9.1) in this case will take the form
w n = δn − αn x + βn y , for n =1,2,..., N . (5.9.18)
Substitution of (5.9.18) into (5.9.10), (5.9.14) and (5.9.17) yields respectively
a1 N F (φ1n, k 1)
δ = HP 1 +
K( k 1) 1  Σ n=2
K( k 1)
P n,

(5.9.19)
a 31 N
α = HM −
3D( k 1) 1  1y Σα
n=2
x P
1n n n
,

(5.9.20)
a 1 b 21 N
3 B (k1)
β1 = H M +

1x Σβ
n=2
y P
1n n n
,

(5.9.21)
where
F (φ1n, k 1) − E (φ1n, k 1)
α1n = , (5.9.22)
K( k 1) − E( k 1)
E (φ1n, k 1) − (1 − k 21) F (φ1n, k 1) − k 21(ρ21n − 1)1/2/ρ1n(ρ21n − k 21)1/2

β = ,
1n E( k 1) − (1 − k 21)K( k 1)
(5.9.23)
E( k 1) − (1 − k 21) K( k 1)
B( k 1) = , (5.9.24)
k 21
K( k 1) − E( k 1)
D( k 1) = , (5.9.25)
k 21
K( k 1), F (φ1n, k 1), E( k 1), E (φ1n, k 1) stand for the complete and incomplete elliptic
integrals of the first and second kind respectively; and ρ1n, φ1n are defined by
1
φ1n = sin-1( ), ρ1n = [ L + ( L 2 − k 21 x 2n/ a 21)1/2]1/2,
ρ1n
1 2
L = [ k + ( x 2n + y 2n)/ a 21],
2 1
where x n and y n are coordinates of a certain point inside S n, and k 1 is the

eccentricity of the first ellipse
k 1 = [1 − b 21/ a 21]1/2.
Each of the equations (5.9.19), (5.9.20) and (5.9.21) represent the first of a set
of N equations. When the acting forces are known, the three sets of equations
define the normal and the angular displacements of the punches. In the case
where the displacements are known, the three sets of linear algebraic equations
have to be solved for P n, M nx and M ny. It is also important to notice that
each equation in the set is valid in the system of coordinates located at the
center of the ellipse.
Example 1: Two equal elliptical punches. Consider the case where N =2,
a 1= a 2= a , b 1= b 2= b , X 1= Y 1=0, X 2= l , Y 2=0, θ1=θ2=0. If we also have P 1= P 2= P
then, due to the symmetry of the system, the set of equations, equivalent to
(5.9.19), reduces to just one equation, namely,
δ = HP +
a F (φ, k ) 
P , (5.9.26)
K( k )  K( k ) 
with the immediate result
P0
P = , (5.9.27)
F (φ, k )
1 +
K( k )
where P 0=δ a / H K( k ) denotes the force which produces a punch settlement equal to
δ when acting on a solitary punch. Equation (5.9.27) shows that the interaction
between the punches decreases the value of the force necessary to produce the
required settlement. The upper and lower bounds for P can be obtained from
(5.9.27) by taking
φ = sin-1
a 
φ = sin-1
a 
, and . (5.9.28)
l − a  l + a 
respectively. We shall also consider the central estimation for P defined by
a
φ = sin-1( ). (5.9.29)
l
Figure 5.9.1 plots the ratio P/P0 versus l/a for a =2, b =1. The solid line gives
the upper bound, the dashed line gives the lower one and the circles represent
the central estimation. As one can see, the maximum possible error of the
Fig. 5.9.1. Interaction between two elliptical punches.
central estimation is less than 9% for l/a >3.5, it is less than 5% for l/a >5, it is
less than 2% for l/a >8, and it is less than 1% for l/a >12. Since there is no
accurate solution available for this case, it is difficult to say how great is the
real error of the central estimation, but there is a reason to believe that it is
much less than that indicated above. This belief is founded in a comparison of
the central estimation for two equal circular punches with the numerical solution
of Kobayashi (1939). The values of the ratio P/P0 for various d = l/a are given
in Table 5.9.1.
If one takes Kobayashi’s solution as exact then the maximum error of the
central estimation does not exceed 0.4% in the whole range of 2≤ d <∞. Even if
one assumes the accuracy in the case of two elliptic punches to be ten times
Table 5.9.1 Comparison of two solutions
upper lower central result of error

d bound bound estimation Kobayashi (%)
2.0 0.82213 0.50000 0.75000 0.75272 0.36

2.2 0.83172 0.61457 0.76900 0.77014 0.15
2.4 0.84030 0.66379 0.78517 0.78545 0.04
2.6 0.84804 0.69940 0.79915 0.79898 -0.02
2.8 0.85505 0.72728 0.81136 0.81096 -0.05
3.0 0.86143 0.75000 0.82213 0.82162 -0.06
3.5 0.87515 0.79241 0.84427 0.84370 -0.07
4.0 0.88638 0.82213 0.86143 0.86093 -0.06
5.0 0.90367 0.86143 0.88638 0.88602 -0.04
7.0 0.92611 0.90367 0.91637 0.91619 -0.02
10.0 0.94522 0.93381 0.94005 0.93999 -0.007
∞ 1. 1. 1. 1. 0.0
worse than the accuracy of the central estimation for two circular punches, this
would still give the maximum error of 4% which is not bad. Bearing this in
mind, we shall evaluate the central estimation only in the examples to follow.
If the normal forces are applied centrally then the angles of inclination of
the punches will be defined by (5.9.20) and (5.9.21) as
3D( k )
α1 = −α2 = − H α12 lP , β1 = β2 =0, (5.9.30)
a 31
where
F (φ, k ) − E(φ, k )
α12 = , k = (1 − b 2/ a 2)1/2, (5.9.31)
K( k ) − E( k )
and φ is defined by (5.9.29).
In the case α1=α2=0 there should be tilting moments applied to the punches
whose value can be determined from (5.9.20) as
M = −M = α12 lP . (5.9.32)
1y 2y
Example 2: Four equal elliptical punches. Consider the configuration

shown in Fig. 5.9.2. Let equal vertical forces P be applied to each punch.
The punches are numbered in the clockwise direction starting from the one at the
coordinate system origin. Due to the symmetry of the system, it is sufficient to
consider just one equation of each of the sets (5.9.19−5.9.21). The result is
Fig. 5.9.2. Geometry of four punches interaction
F (φ12, k ) F (φ13, k ) F (φ , k )
δ = HP 1 + ,
a 14
+ + (5.9.33)
K( k )  K( k ) K( k ) K( k ) 
α = H  M − P l (α13 + α ),
a3
(5.9.34)
3D( k ) 1  1y 14 
β = H M1x + Pc (β12 + β13),

ab 2
(5.9.35)
3B( k ) 1  
where α1n and β1n are defined by (5.9.22) and (5.9.23) respectively, and
1
φ1n = sin-1( ), for n =2,3,4; (5.9.36)
ρ1n
ρ12 = 1 + ( c 2 − b 2)/ a 21/2,

l
ρ14 =
  a
ρ13 =  L + ( L 2 − k 2 l 2/ a 2)1/2 ,
1 2
k + ( l 2 + c 2)/ a 2,
1/2
L =
  2 
(5.9.37)
When the forces and the tilting moments are known, equations (5.9.33−5.9.35)
define the normal and the angular displacements of the punches.
The mathematically equivalent problem of diffusion through perforated

membranes was solved in (Fabrikant, 1985a, 1987k).
Exercise 5.9
1. Consider the interaction between the four elliptical punches depicted in

Fig. 5.9.2. Find the tilting moments if the punches are not allowed to tilt.
Answer: M = − Pc (α12 + α13) , M = Pl (α13 + α ).
1x 1y 14
2. Find the limiting case of formulae (5.9.22−5.9.25) for the case of circular
punches.
2 -1 1 
a a1 a 21 1/2
Answer: α →β → sin − 1 − 2  ,
1n 1n π r 

r 
1n
r 1n 1n 
F (φ , k 1)
a1
-1 ,
2 1n
B( k 1)→D( k 1)→π/4 , → sin
K( k 1) π  1n
r
and the equations (5.9.19−5.9.21) will take the form
N
a
a δ = HP 1 + Σ P nsin  ,
2 2 1
-1
π 1 1  π r1n
n=2
4 a 31 N
3π
α1 = H  M
 1y
− Σα
n=2
xP
1n n n
,

4 a 31 N
β = H  M 1x +
3π 1  Σβ
n=2
yP
1n n n
,

where a 1 is the radius of punch one and r is the distance between the first
1n
punch centre and a point inside S n.
3. Find the generalized displacements of an elliptical punch due to the action of

several concentrated loads, applied normally outside the punch.
Hint : Consider the procedure of shrinking of the areas Sn , n =2,3,..., N . The
accuracy of equations (5.9.19−5.9.21) will increase. In the limiting case Sn →0
formulae (5.9.19−5.9.21) give an exact solution to the problem of several
concentrated forces acting outside an elliptical punch.
4. Consider the interaction of several punches on a non-homogeneous elastic

half-space, with the modulus of elasticity being proportional to a power function

of the depth.
5.10 Interaction between flexible punches under shifting loading
The mixed boundary value problem for a transversely isotropic elastic

half-space is considered for the case where uniform tangential displacements are
prescribed over several domains of arbitrary shape and the rest of the half-space
boundary is stress free. The problem can be interpreted either as an interaction
between two elastic half-spaces interconnected through several areas and subjected
to remote shear loading, or as a contact problem of several flexible punches,
connected to the half-space, with different tangential displacements prescribed. A
general theorem is established which relates the resulting tangential forces, acting
on each domain, with their generalized displacements through a system of linear
algebraic equations. The theorem is applied to the case of arbitrarily located
elliptical domains subjected to uniform tangential displacements. Several specific
examples are considered.
Theory. Consider a transversely isotropic elastic half-space z ≥0. Let the
tangential displacements be prescribed over several simply connected domains S n
while the rest of the half-space boundary is stress free. The mathematical
formulation of the boundary conditions on the plane z =0 is
u x = u x( x , y ), u y = u y( x , y ), for ( x , y )∈ S n,
σz = τzx = τyz = 0, for ( x , y )∉ S n, n =1,2, ..., N . (5.10.1)
The boundary value problem (5.10.1) can be interpreted either as an interaction

between several flexible punches S n bonded to the half-space and subjected to
tangential displacements, or as the case of two elastic half-spaces connected in S n
and subjected to a remote shear loading, the first interpretation being more
general since the tangential displacements each punch can be treated as
independent. Hereafter we shall call S n the domain of contact for the n th
punch. The punches are assumed to be flexible, so that they do not exert any
normal tractions. Let T and T be the components of the resulting tangential
nx ny
force applied to the n -th punch. The problem is to find the relationships
between the generalized displacements of the punches and the acting forces.
Introduce the complex tangential displacements u = u x+i u y, and the complex
tangential tractions τ=τzx+iτyz. In view of the principle of superposition, the
governing integral equation can we rewritten from (5.7.5) as follows:
N
G 1 
⌠ τ(ξ,η) dξdη + 2 ⌠ ⌠ q τ(ξ,η) dξdη  = u ( x , y ).
G
Σ 2 ⌠
n=1 
⌡⌡ R 2 ⌡⌡ q R

Sn Sn
(5.10.2)
Here
q = x − ξ + i( y − η), q = x − ξ − i( y − η), R 2 = q(5.10.3)

q.
Substitution of the boundary conditions (5.10.1) in (5.10.2) leads to a set of N

integral equations. The exact solution of these equations is not known at
present, even for the case of several circles. Here it will be shown that we do
not really need to know these solutions if we are interested only in the
relationship between the applied forces and the punch displacements. We can
single out, without loss of generality, the first punch, and consider the related
integral equation
G1 τ (ξ,η) G2 q 11 τ1(ξ,η)
u 1( x 1, y 1) = ⌠⌠ 1 dξdη + ⌠⌠ dξdη
2 ⌡ ⌡ R 11 2 ⌡⌡ q11 R11
S1 S1
G 1
N
τn(ξ,η) G2 q τn(ξ,η) 
Σ ⌠ ⌠ ⌠ ⌠ 1n
+ 2 dξdη + dξdη. (5.10.4)
n=2 
⌡ ⌡ R 1n 2 ⌡⌡ q R
1n 1n 
Sn Sn
Here u n stands for tangential displacement of the n -th punch, τn is the tangential
tractions exerted by the punch, the point ( x n, y n)∈ S n, and
q = x k − ξn + i( y k − ηn), q = x k − ξn − i( y k − ηn),
kn kn
R = ( q q )1/2, (ξn,ηn)∈ S n. (5.10.5)

kn kn kn
We have dropped for simplicity the subscripts of ξn and ηn in (5.10.4) and

thereafter. It should not produce any confusion for the reader.
Equation (5.10.4) can be rewritten as
G1 G
∆⌠ ⌠ R 11τ1(ξ,η) dξdη − Λ2⌠ ⌠ R 11τ1(ξ,η) dξdη
2
u 1( x 1, y 1) =
2 ⌡⌡ 2 ⌡⌡
S1 S1
N
G 1 G 
+ Σ  2 ∆⌠
n=2 
⌠ R τ (ξ,η) dξdη − 2 Λ2⌠ ⌠ R τ (ξ,η) dξdη.
⌡ ⌡ 1n n 2 ⌡ ⌡ 1n n

Sn Sn
(5.10.6)
Suppose that the function τ0, is known, satisfying the following integral equation
inside S 1
⌠ ⌠ τ (ξ,η) R dξdη = a x 2 + a y 2 + a xy + a x + a y + a ,
⌡⌡ 0 1 2 3 4 5 6
S1
(5.10.7)
where a 1, ... , a 6 are complex constants. One can verify that
∆ ⌠ ⌠ τ0(ξ,η) R 11 dξdη = 2 a 1 + 2 a 2 = d 1 = const,

⌡⌡
S1
Λ2⌠ ⌠ τ0(ξ,η) R 11 dξdη = 2 a 1 − 2 a 2 + 2i a 3 = - c 1 = const, (5.10.8)

⌡⌡
S1
which means that the function τ0 renders constant the first two integrals in
expression (5.10.4). This important property will be used in the derivation to
follow. Multiplication of both sides of (5.10.4) by τ0 and integration over the
area S 1 yields
G τ1(ξ,η)
⌠ ⌠ u (x ,y ) τ (x ,y ) dx dy = 1 ⌠ ⌠ τ (x ,y ) dx dy ⌠ ⌠ dξdη
⌡⌡ 1 1 1 0 1 1 1 1 2 ⌡⌡ 0 1 1 1 1
⌡⌡ R 11
S1 S1 S1
G2 q 11 τ1(ξ,η)
+ ⌠ ⌠ τ (x ,y ) dx dy ⌠ ⌠ dξdη
2 ⌡⌡ 0 1 1 1 1
⌡⌡ q11 R11
S1 S1
N
G 1 τn(ξ,η)
+ Σ n=2 
⌠ ⌠
 2 ⌡⌡ 0 1 1
τ ( x , y ) d x 1 d y 1
⌠ ⌠
⌡ ⌡ R 1n
dξdη
S1 Sn
G2 q 1n τn(ξ,η) 
+ ⌠ ⌠ τ (x ,y ) dx dy ⌠ ⌠ dξdη. (5.10.9)
2 ⌡⌡ 0 1 1 1 1
⌡⌡ q1n R1n
S1 Sn 
By changing the order of integration in (5.10.9) and taking into consideration the
property (5.10.8), the following result can be obtained
G G
⌠ ⌠ u (x ,y ) τ (x ,y ) dx dy = 1 d T + 2 c T
⌡⌡ 1 1 1 0 1 1 1 1 2 1 1 2 1 1
S1
N
G 1 τ0 ( x 1 , y 1 )
+ Σ 2
n=2 
⌠
⌡
⌠
⌡
τ n ( ξ , η) dξ dη ⌠
⌡
⌠
⌡ R 1n
d x 1d y 1
Sn S1
G2 q 1n τ0( x 1, y 1) 
+ ⌠ ⌠ τ (ξ,η) dξdη ⌠ ⌠ d x 1 d y 1 . (5.10.10)
2 ⌡⌡ n ⌡⌡ q1n R1n
Sn S1 
where T 1 is the complex representation for the total tangential force acting on
the first punch. Introduce the notation
τ0 ( x 1 , y 1 ) q 1n τ0( x 1, y 1)
Φ(ξ,η) = ⌠ ⌠ d x 1d y 1, Ψ(ξ,η) = ⌠ ⌠ d x 1d y 1 .
⌡⌡ R 1n ⌡⌡ q1n R1n
S1 S1
(5.10.11)
Now expression (5.10.10) can be rewritten as
G G
⌠ ⌠ u (x ,y ) τ (x ,y ) dx dy = 1 d T + 2 c T
⌡⌡ 1 1 1 0 1 1 1 1 2 1 1 2 1 1
S1
N
G 1 G2 
+ Σ  2 ⌠ ⌠ Φ(ξ,η) τn(ξ,η) dξdη + 2 ⌠ ⌠ Ψ(ξ,η) τn(ξ,η) dξdη.
n=2 
⌡⌡ ⌡⌡ 
Sn Sn
(5.10.12)
By evoking the mean value theorem, we come to the linear algebraic equation,
for the forces T n applied to each punch.
G G
⌠ ⌠ u (x ,y ) τ (x ,y ) dx dy = 1 d T + 2 c T
⌡⌡ 1 1 1 0 1 1 1 1 2 1 1 2 1 1
S1
N
G 1 G2 
+ Σ  2 Φ( x n, y n) T n + 2 Ψ(ξn,ηn) T n .
n=2  
(5.10.13)
The exact location of the points ( x n, y n) and (ξn,ηn) is not known but the fact
that they belong to S n allows only a limited variation, and in many cases
provides sufficiently close upper and lower bounds for the parameters sought.
By a similar argument, N −1 additional linear algebraic equations can be derived
for the remaining punches. This set of equations provides the necessary
relationships between the tangential displacements of the punches and the applied
forces, and represents the main result of this paragraph. It is clear that each
equation can be interpreted in terms of the reciprocal work. In order to use
equations (5.10.13), one needs to know the explicit expression for τ0 which, at
present, is available for the elliptical punch only.
Application to elliptical punches. Consider the interaction of a set of N
flexible elliptical punches arbitrarily located on an elastic half-space, with uniform
tangential displacements u n=const prescribed in S n. Let a n and b n be the major
and the minor semiaxes of the n th ellipse; X n and Y n define its center, θn be
the angle between the axis Ox and the major semiaxis a n, and T n be the
tangential force acting upon the n th punch. The function τ0 for the first punch
is
τ0( x , y ) = 1 − 2 − 2  .
x2 y 2 -1/2
(5.10.14)
 a1 b1 
Substitution of (5.10.14) in (5.10.10) yields, after computation of the integrals

involved (see Appendix A5.1 for details)
2 a 1 u 1 = G 1K( k 1) T 1 + G 2[(2 − k 21)K( k 1) − 2 E( k 1)] T 1/ k 21
N
 
+
1
2 Σ 
G ⌠ ⌠
 1⌡ ⌡ 1n n n
D τ d S + ( G /
2 1k 2 ⌠⌠
) L τ d
⌡ ⌡ 1n n n
S − i( G /
2 1k 2 ⌠⌠
)
⌡⌡
M τ
1n n
d S n ,

n=2 Sn Sn Sn
(5.10.15)
where
D 1n = [ F (φ1n, k 1) − F (ψ1n, k 1)],
L 1n = {(2 − k 21)[ F (φ1n, k 1) − F (ψ1n, k 1)] − 2[E(φ1n, k 1) − E(ψ1n, k 1)]},
M 1n = [(1 − k 21sin2φ1n)1/2 − (1 − k 21sin2ψ1n)1/2]. (5.10.16)
Here k 1=(1− b 21/ a 21)1/2, K( k 1), F(φ1n, k 1), E( k 1), E(φ1n, k 1) denote the complete and
incomplete elliptic integrals of the first and second kind respectively; ψ1n and φ1n
are defined according to formulae (A5.1.20−A5.1.21) from Appendix A5.1, namely,
x n y n + ( a 21 y 2n + b 21 x 2n − a 21 b 21 )1/2
φ1n = tan -1
,
y 2n − b 21
x n y n − ( a 21 y 2n + b 21 x 2n − a 21 b 21 )1/2
ψ1n = tan -1
. (5.10.17)
y 2n − b 21
In the case when φ <ψ , it should be replaced by π+φ . We recall that the
1n 1n 1n
point ( x n, y n)∈ S n. If the mean value theorem is applicable, equation (5.10.15)
transforms into
2 a 1 u 1 = G 1K( k 1) T 1 + G 2[(2 − k 21)K( k 1) − 2E( k 1)] T 1/ k 21
N

+
1
2 Σ
n=2
 G 1[ F (φ1n, k 1) − F (ψ1n, k 1)] T n + ( G 2/ k 1)[(2 − k 1)( F (φ1n, k 1)

2 2
− F (ψ1n, k 1)) − 2(E(φ1n, k 1) − E(ψ1n, k 1))] T n

− i( G 2/ k 21)[(1 − k 21sin2φ1n)1/2 − (1 − k 21sin2ψ1n)1/2] T n. (5.10.18)

It is important to note that in the case of symmetric configuration, some of the
terms in (5.10.18) vanish; the parameters φ and ψ do not necessarily
1n 1n
correspond to the same point ( x n, y n); they should be interpreted as fuzzy
quantities which can assume any value corresponding to ( x n, y n)∈ S n. Equation
(5.10.18) represents the first of the set of N linear algebraic equations with fuzzy
coefficients, which can be obtained in a similar manner. The solution can also
be interpreted as a fuzzy domain. Analysis of this domain will allow us to find

the upper and lower bounds for the quantities of interest. As we shall see
further, in some cases the variation between the upper and the lower bounds
becomes so small that allows us to obtain a reasonably accurate solution to the
problem.
When the acting forces are known, the set of equations define the
tangential displacements of the punches. In the case when the displacements are
known, the set of linear algebraic equations has to be solved for the forces T n.
It is also important to notice that each equation in the set is valid in a system
of coordinates located at the center of the ellipse.
Example: Two equal elliptical punches. Consider the case where N =2,
a 1= a 2= a , b 1= b 2= b , X 1= Y 1=0, X 2= l , Y 2=0, θ1=θ2=0. Let the tangential
displacements be prescribed as u 1= u 2= u where u is a real quantity. Due to the
symmetry of the system, we may also assume T 1= T 2= T , and ℑ[ T ]=0 (i.e. the
resulting force is directed along the axis Ox ). Taking into consideration the
properties
D 1n( x , y ) = D 1n(− x , y ) = D 1n( x ,− y ) = D 1n(− x ,− y ),
L 1n( x , y ) = L 1n(− x , y ) = L 1n( x ,− y ) = L 1n(− x ,− y ),
M 1n( x , y ) = − M 1n(− x , y ) = − M 1n( x ,− y ) = M 1n(− x ,− y ), (5.10.19)
we have
⌠ ⌠ M τd S = 0, (5.10.20)
⌡ ⌡ 1n n
Sn
and the set of equations, equivalent to (5.10.15), reduces to just one equation,
namely,
2 au = 2{ G 1K( k ) T + ( G 2/ k 2)[(2 − k 2)K( k ) − 2E( k )] T }
− { G 1F( k ,φ) T + ( G 2/ k 2)[(2 − k 2)F( k ,φ) − 2E( k ,φ)] T }, (5.10.21)
where, due to symmetry, φ is defined as
φ = cos-1 2 
b
(5.10.22)
( x − a + b ) 
2 2 1/2
Introduce the notation
2 au
T0 = (5.10.23)
G 1K( k ) + ( G 2/ k )[(2 − k 2)K( k ) − 2E( k )]
2
where T 0 can be interpreted as the force needed to produce the displacement u

when acting on an isolated punch. Equation (5.10.21) can be rewritten as
follows:
T0 G 1F( k ,φ) + ( G 2/ k 2)[(2 − k 2)F( k ,φ) − 2E( k ,φ)]

T = , B =
2 − B G 1K( k ) + ( G 2/ k 2)[(2 − k 2)K( k ) − 2E( k )]
(5.10.24)
Since B ≥0, one may conclude that the punch interaction reduces the force needed
to produce the displacement u , as compared to an isolated punch. The upper
and lower bounds for T can be obtained from (5.10.24) by taking
 b 
φ = cos-1  ,
[( l + a )2 − a 2 + b 2]1/2
 
and
 b 
φ = cos-1  . (5.10.25)
[( l − a )2 − a 2 + b 2]1/2
 
respectively. We shall also consider the central estimation for T defined by
 b 
φ = cos-1 .
( l 2 − a 2 + b 2)1/2
 
We can consider in a similar manner the case when the displacement is
prescribed in the Oy direction by the formal substitution of u by i u , and T by
i T in expression (5.10.21). The result is
T0 G 1F( k ,φ) − ( G 2/ k 2)[(2 − k 2)F( k ,φ) − 2E( k ,φ)]

T = , C = .
2 − C G 1K( k ) − ( G 2/ k 2)[(2 − k 2)K( k ) − 2E( k )]
(5.10.26)
Figure 5.10.1 plots the ratio T/T0 versus l/a for a =2, b =1, ( G 2/ G 1)=1/3, with the
tangential displacements being directed along the axis Ox . The upper and lower
bounds are given by circles, while the central estimation is given by a solid line.
Fig. 5.10.1. Interaction between two elliptical punches,

with shifting load in the Ox direction
As one can see, the maximum possible error of the central estimation is less
than 15% for l/a ≥2.5, it is less than 10% for l/a ≥3, it is less than 7% for
l/a >3.5, and it is less than 5% for l/a ≥4. Since there is no accurate solution
available for this case, it is difficult to say how great is the real error of the
central estimation, but there is a reason to believe that it is much less than
indicated above. This belief is supported by the same argument as in section
5.9.
Figure 5.10.2 gives the same plot for the case when the tangential
displacements are prescribed in the Oy direction. The respective errors are 12%,
7%, 5%, and 3.5%. One can see that the accuracy of the central estimation is
significantly better than in the first case. This should be expected since the
interaction in the case of displacements prescribed along the axis Ox is stronger
then in the second case.
Discussion. It is appropriate to consider certain limiting cases. In the case

of a circular punch the eccentricity k1→0, and formula (5.10.15) will simplify
Fig. 5.10.2. Interaction between two elliptical punches,

with shifting load in the Oy direction
N
 a1
2a1u1 =
π
G T +
2 1 1 Σ  1
n=2 
G ⌠
⌡
⌠
⌡
sin-1

 τ ( x, y) dxdy
( x 2 + y 2)1/2 n
Sn
a 1( x 2 + y 2 − a 21)1/2 
+ G 2⌠ ⌠ τn ( x , y ) d x d y  . (5.10.27)
⌡⌡ ( x − i y )2

Sn
Here we used the integrals (A5.1.22−A5.1.23). The interaction of two equal

circular punches subjected to loading T in the Ox direction leads to the
following expression for the central estimation
G 1 T +  G 1 T sin-1( ) + G 2
π a a ( l 2 − a 2)1/2 
2 au = T . (5.10.28)
2  l l2 
Introducing again the notation
4 au
T0 = ,
πG1
the following result can be obtained from (5.10.28)
T0
T = (5.10.29)
G 2 a ( l 2 − a 2)1/2
2 -1 a 
1 + sin ( ) +
π l G1 l 2

If the tangential displacements are prescribed in the Oy direction, we obtain
T0
T = (5.10.30)
G 2 a ( l 2 − a 2)1/2
2 -1 a 
1 + sin ( ) −
π l G1 l 2

All the results are valid for the case of an isotropic body, provided that
G 1=(2−ν)/(2πµ) and G 2=ν/(2πµ), where µ is the shear modulus and ν is the
Poisson coefficient.
Exercise 5.10
1. Verify (5.10.6).
2. Derive (5.10.12).
3. Consider the interaction of two equal elliptical punches, when their major
semiaxes are orthogonal to each other.
4. Find the generalized displacements of an elliptical punch due to the action of

several concentrated loads, applied tangentially outside the punch.
Hint : Consider the procedure of shrinking the areas S n, n =2,3,..., N , to a point.
The accuracy of equation (5.10.18) will increase. In the limiting case S n→0
formula (5.10.18) gives an exact solution to the problem of several concentrated
forces acting outside an elliptical punch.
5.11 Contact problem for a rough punch
The term rough punch is used here as the opposite to the term smooth
punch, indicating a punch which does produce shear traction at its base.
Consider a circular punch, with general base, penetrating transversely isotropic
elastic half-space. The punch is subjected to the action of an axial force P and
a tangential force T = kP , acting in the Ox direction, with k being the friction
5.11 Contact problem for a rough punch 423
coefficient. Let the domain of contact be a circle of radius a . The shear

traction τ is assumed to be proportional to the pressure σ, namely, τ = k σ.
zx zx
The governing integral equation will take the form (due to (2.2.13)):
2π a 2π a
σ(ρ0,φ0)ρ0dρ0dφ0 σ(ρ0,φ0)ρ0dρ0dφ0
⌠⌠ =
w (ρ,φ)
− α kℜ⌠ ⌠ . (5.11.1)
⌡⌡ R H ⌡⌡ ρeiφ i φ0
− ρ0e
0 0 0 0
Here w (ρ,φ) is the punch settlement. Let us rewrite (5.11.1) in the operator
form
w
Nσ = − α k Mσ, (5.11.2)
H
where N and M are the integral operators in the left- and the right-hand sides
of (5.11.1) respectively.
The values of α and k for real bodies are less that unity, and hence their
product is convenient to use as a small parameter. Let the following expansion
hold
∞
σ(ρ,φ) = Σ(αk) σ (ρ,φ).

n=0
n
n (5.11.3)
By substitution of (5.11.3) in (5.11.2) and equating the terms with equal powers
of the small parameter, the following infinite system of integral equations can be
obtained:
w
Nσ0 = , Nσ1 = −Mσ0, ..., Nσn = −Mσn-1, ... (5.11.4)
H
The operator N −1, inverse to N is known from (1.4.10), so we can write the
formal solution to the problem as follows:
∞
σ = Σ
n=0
(α k )n(−N −1 M)nN −1 .
w
 
H
(5.11.5)
Example. Consider the case of a flat circular punch. Let ω be the

settlement of its centre, and δ be its tilting angle about the Oy axis. Then
w (ρ,φ) = ω + δρcosφ. (5.11.6)
Substitution of (5.11.6) in the first equation of (5.11.4) yields

1 ω + 2δρcosφ
σ0(ρ,φ) = , (5.11.7)
π H ( a 2 − ρ2)1/2
2
which is, in fact, the solution for an inclined smooth punch. Substitution of
(5.11.7) in the second equation of (5.11.4) yields the second term of the
expansion
2 ω a ln[( a 2 − ρ2)1/2/ a ]
a
σ1(ρ,φ) = 3  cosφ + δ 1 d ⌠ x 2d x
ln a + x 1/2 +
πH
 ρ( a 2 − ρ2)1/2 ρ dρ⌡ ( x 2 − ρ2)1/2 a − x 
ρ
a
+ a 3ln[( a 2 − ρ2)1/2/ a ] + a cos2φ. (5.11.8)
( a 2 − ρ2)1/2  ρ2( a 2 − ρ2)1/2 2( a 2 − ρ2)1/2  
The procedure may be continued, and further terms of the expansion may be
obtained. It is left to the interested reader. The resultant force P and the
tilting moment M can be obtained by integration of the sum of (5.11.7) and
(5.11.8) as follows:
2a 
a δ,
αk 2 a 22 αk 
P = ω + M = aδ − ω . (5.11.9)
πH π  π H 3 π 
If the point, where the normal load P is applied, is shifted along the axis Ox
by the distance b from the punch centre, then M = Pb , and the following
relationship may be established between the punch settlement ω and the tilting
angle δ:
b + (α k /π) a ω
δ = . (5.11.10)
(2/3) a − (α k /π) b a
Formula (5.11.10) shows that the punch tilts even in the case of a central
loading. A similar effect can be observed in the case of a bonded punch
(Chapter 3) and in two-dimensional contact problems (Muskhelishvili, 1946). In
order to eliminate tilting, the normal load should be applied at the point
x =−α ka /π.
Exercise 5.11
1. Verify (5.11.5).
2. Derive (5.11.7) −(5.11.8).
Appendix A5.1 425
4. Find the corrective term in the expression for P (5.11.9) due to σ2.
2a  
ω 1 − α 2 k 2 + a δ.
1 αk
Answer: P =
πH  6  π 
5. Find the normal displacement w outside the punch, taking into consideration
the first two terms in the solution expansion.
Answer: w = ωsin-1( ) + δρsin-1( ) − (ρ2 − a 2)1/2cosφ
2 a a a
π ρ  ρ ρ  
4α k  a -1 a (ρ2 − a 2)1/2 ρ cosφ

+  ω
ρ
cos ( ) + ln 2
π2 ρ ρ (ρ − a 2)1/2

ρ δ 2 3 -1 a
+ δ(ρ2 − a 2)1/2ln 2 1/2 + a cos ( )
(ρ − a )
2
ρ23 ρ

+ (ρ2 − a 2)1/2
ρ2 + 2 a 2 ρ a 2

ln 2 −
6 
cos2φ.
3 (ρ − a 2)1/2

6. Try to find an exact closed form solution to the problem.
Appendix A5.1
Here the method of computation of several integrals over an elliptical

domain is given. We introduce the notation
R = [( x − x 0)2 + ( y − y 0)2]1/2
x 20 y 20
Z 0 = 1 − − 1/2
 a 2
b2
The following integral is to be computed
f ( x 0, y 0)d x 0d y 0
I = ⌠⌠ , (A5.1.1)
⌡⌡ RZ 0
S
where S is an ellipse with semiaxes a and b ( a ≥ b ), and f is a general function.

We use the method of Rostovtsev (1961), slightly modified in order to simplify
the computations. Introduce a new system of polar coordinates ( r ,θ), with the
origin at the point ( x , y ), namely,
x 0 = x + r cosθ, y 0 = y + r sinθ. (A5.1.2)
First of all, consider the case when ( x , y )∈ S . Substitution of (A5.1.2) in

(A5.1.1) yields
2π r1
ab dθ f ( x + r cosθ, y + r sinθ) d r
I = ⌠ ⌠ ,
⌡ ( b cos θ + a 2sin2θ)1/2 ⌡
2 2
[( r 1 − r )( r + r 2)]1/2
0 0
(A5.1.3)
where r 1 and − r 2 are the roots of algebraic equation
( x + r cosθ)2 ( y + r sinθ)2
+ − 1 = 0,
a2 b2
namely,
( Q 2 + PZ )1/2 Q x cosθ y sinθ

r 1,2 = , Q = 2 + ,
P a b2
cos2θ sin2θ x2 y2
P = + , Z = 1 − − . (A5.1.4)
a2 b2 a2 b2
Now, we may split the integral (A5.1.3) in two, according to the scheme
2π r1 π r1 2π r1
⌠ dθ ⌠ d r = ⌠ dθ ⌠ d r + ⌠ dθ ⌠ d r. (A5.1.5)
⌡ ⌡ ⌡ ⌡ ⌡ ⌡
0 0 0 π 0 0
Make a formal replacement of θ by π+θ in the second integral of (A5.1.5).

Taking into consideration that r 1(θ)= r 2(π+θ), transformation of (A5.1.5) may
proceed as follows
2π r1 π r1 π r2 π r1 π 0
⌠ dθ ⌠ d r = ⌠ dθ ⌠ d r + ⌠ dθ ⌠ d r = ⌠ dθ ⌠ d r + ⌠ dθ ⌠ d r.
⌡ ⌡ ⌡ ⌡ ⌡ ⌡ ⌡ ⌡ ⌡ ⌡
0 0 0 0 00 0 -r 0 0 2
(A5.1.6)
In the last term of (A5.1.6) we have made a formal replacement of r by − r .
Appendix A5.1 427
Taking into consideration that the expression f ( x + r cosθ, y + r sinθ) remains

unchanged with the replacements of θ by π+θ and r by − r , the two integrals in
(A5.1.6) can be combined, and the final result will read
π r1
ab dθ ⌠ f ( x + r cosθ, y + r sinθ)d r
I = ⌠ , (A5.1.7)
⌡ ( b cos θ + a sin θ) ⌡ [( r 1 − r )( r + r 2)]1/2
2 2 2
2 1/2
0 -r 2
We recall the integral
1 1
r1 n Γ( n + − k ) Γ( + k ) r
Σ − 2k .
n 2 2
⌠ r dr n
1/2 = r 1
⌡ [( r1 − r)( r + r2)] Γ( n + 1 − k ) Γ( k + 1)  r 1
-r 2 k=0
(A5.1.8)
Now various integrals can be computed fairly easily. Here are the results which
are used in the main text of this paper
π
d x 0d y 0
⌠⌠ ab dθ
= π⌠ 2 1/2 = 2π b K( k ), (A5.1.9)
⌡⌡ RZ0 ⌡ ( b cos θ + a sin θ)
2 2 2
S 0
π
ab e2iθ dθ
q d x 0d y 0
⌠⌠ = π⌠ 2 1/2 = 2π b [(2 − k )K( k ) − 2E( k )]/ k .
2 2
⌡⌡ q RZ0 ⌡ ( b 2
cos2
θ + a 2
sin θ)
S 0
(A5.1.10)
Now consider the case when the point ( x , y )∉ S . In this case substitution
of (A5.1.2) in (A5.1.1) yields
α2 r1
ab dθ ⌠ f ( x + rcosθ, y + rsin1/2 θ) d r
I = ⌠ ,
⌡ ( b cos θ + a sin θ) ⌡
2 2 2 1/2 2
[( r 1 − r )( r + r 2)]
α1 -r 2
(A5.1.11)
where α1 and α2 are the angles between the axis Ox and the two lines passing
through the point ( x , y ) tangentially to the ellipse. We can find these angles by
using classical formulae of differential geometry. The parametric equation of the
ellipse can be written
X = a cost , Y = b sin t . (A5.1.12)
The equation of the lines passing through the point ( x , y ) tangentially to the
ellipse may be written as
y − b sin t x − a cost
= . (A5.1.13)
b cost − a sin t
Equation (A5.1.13) can be solved for sin t , and the two solutions are
b[a2y x ( x 2 b 2 + y 2 a 2 − a 2 b 2)1/2]
(sin t )1,2 = . (A5.1.14)
x2b2 + y2a2
This gives the following two expressions for cost
a [ b 2 x ± y ( x 2 b 2 + y 2 a 2 − a 2 b 2)1/2]
(cos t )1,2 = . (A5.1.15)
x2b2 + y2a2
Now we can write
y − b sin t
tanα = . (A5.1.16)
1,2 x − a cost
Substitution of (A5.1.14−A5.1.15) in (A5.1.16) yields
xy ( a 2 y 2 + b 2 x 2 − a 2 b 2)1/2
tanα1,2 = . (A5.1.17)
x2 − a2
We can now compute the integrals
α2
d x 0d y 0
⌠⌠ ab dθ
= π⌠ = π b [F(φ, k ) − F(ψ, k )], (A5.1.18)
⌡⌡ RZ0 ⌡ ( b cos θ + a 2sin2θ)1/2
2 2
S α1
α2
q d x 0d y 0 2iθ
⌠⌠ dθ
= π⌠
ab e
⌡⌡ q RZ0 ⌡ ( b cos θ + a 2sin2θ)1/2
2 2
S α1
= π b {(2 − k 2)[F(φ, k ) − F(ψ, k )] − 2[E(φ, k ) − E(ψ, k )]

Appendix A5.1 429
− 2i[(1 − k 2sin2φ)1/2 − (1 − k 2sin2ψ)1/2}/ k 2. (A5.1.19)
Here
xy + ( a 2 y 2 + b 2 x 2 − a 2 b 2)1/2
tanφ = cotα1 = . (A5.1.20)
y2 − b2
xy − ( a 2 y 2 + b 2 x 2 − a 2 b 2)1/2
tanψ = cotα2 = . (A5.1.21)
y2 − b2
In the limiting case of a circle, b → a , k →0, and formulae (A5.1.18−A5.1.19) will

take the form
d x 0d y 0
⌠⌠ = 2π a sin-1
a
, (A5.1.22)
⌡⌡ RZ0 ( x + y 2)1/2
2
S
q d x 0d y 0
⌠⌠ = 2π a 2
( x 2 + y 2 − a 2)1/2
. (A5.1.23)
⌡⌡ q RZ0 ( x − i y )2
S
The more complicated integrals can be computed in the same manner.

Here are some additional integrals, where x and y are inside the domain S .
2
⌠ ⌠ x 0d x 0d y 0 = π b 3 1  1 E − K + π b x 2 1 2K − (2 + k 2)E + π b y 2 1 2E − (2 − k 2)K
⌡ ⌡ RZ 0 k 2 1 − k 2  k4   k4  
S
⌠ ⌠ x 0 y 0d x 0d y 0 = 2πb xy1 − k 2 − k E − 2K

2 2
⌡⌡ RZ 0 k 4 1 − k 2 
S
2
⌠ ⌠ y 0d x 0d y 0 = π b 3 1 K − E + π b x 2 1 − k 2E − (2 − k 2)K
2
⌡ ⌡ RZ 0 k2   k4  
S
1 
+ π b y2 4 2(1 − k ) K − (2 − 3 k )E
2 2 2
k  
3
⌠ ⌠ x 0d x 0d y 0 = π b 3 x 1 2 − k E − 2K + π b xy2 1 ( − 8 + 5 k 2)K + (8 − k 2)E
2
⌡ ⌡ RZ 0 k 4 1 − k 2  k6  
S
1  
+ π b x3 6 (8 − 3 k + k )K − (8 + k + 2 k )E
2 4 2 4
3k  
2
⌠ ⌠ x 0 y 0d x 0d y 0 = π b 3 y 1 (2 − k 2)K − 2E
⌡⌡ k  
RZ 0 4
S
+ π b x 2 y 6 ( − 8 + 9 k 2 − k 4)K + (8 − 5 k 2 − k 4)E
1
k  
1  
+ π b y3 6 (8 − 11 k + 3 k )K − (8 − 7 k )E
2 4 2
3k  
2
⌠ ⌠ x 0 y 0d x 0d y 0 = π b 3 x 1 (2 − k 2)K − 2E
⌡⌡ k  
RZ 0 4
S
1 
+ π b xy 2 6 (8 − 15 k + 7 k )K + ( − 8 + 11 k − 2 k )E
2 4 2 4
k  
1  
+ π b x3 6 (8 − 9 k + k )E − (8 − 13 k + 5 k )K
2 4 2 4
3k  
3
⌠ ⌠ y 0d x 0d y 0 = π b 3 y 1 (2 − k 2)E − 2(1 − k 2)K
⌡ ⌡ RZ 0 k  
4
S
1  6 
+ π b y3 6 (8 − 17 k + 11 k )E + ( − 8 + 21 k − 19 k + 6 k )K
2 4 2 4
3k  
1 
+ π b x2y 6 (8 − 19 k + 14 k − 3 k )K − (8 − 15 k + 7 k )E
2 4 6 2 4
k  
4
⌠ ⌠ x 0d x 0d y 0 = πbx 4 1 ( − 48 + 8 k 2 − 4 k 4 − 6 k 6)E + (48 − 32 k 2 + 5 k 4 + 3 k 6)K
⌡ ⌡ RZ 0 12 k  
8
S
1  
+ π bx 2 y 2 8 (48 − 16 k − k )E − (48 − 40 k + 4 k )K
2 4 2 4
2k  
Appendix A5.1 431
1  2 
+ π by 4 8 (16 − 16 k + 3 k )K − (16 − 8 k )E
2 4
4k  
1 8 − 5 k 2 − k 4  + πb 3 y 2 1 (8 − 3 k 2)K − 8 − 7 k 2E

+ πb3 x2 E − (8 − k 2
)K
2k  1 − k  2k  1− k 
6 2 6 2
+ πb5
1  − 2 + 4 k 2E + (2 − 3 k 2)K
4 k (1 − k )  1 − k 
4 2 2
3
⌠ ⌠ x 0 y 0d x 0d y 0 = πbx 3 y 1 (48 − 40 k 2 − 2 k 6)E − (1 − k 2)(48 − 16 k 2 + k 4)K
⌡ ⌡ RZ 0 3k 
8

S
+ π bx y 3 8 (1 − k 2)(16 − 8 k 2)K − (16 − 16 k 2 + k 4)E

1
k  
+ π b 3 xy 6 (8 − 5 k 2)K − (8 − k 2)E

1
k  
2 2
⌠ ⌠ x 0 y 0d x 0d y 0 = πbx 4(1 − k ) (48 − 16 k 2 − k 4)E − (48 − 40 k 2 + 4 k 4)K
2
⌡ ⌡ RZ 0 12 k 8  
S
1 
+ π bx 2 y 2 ( − 48 + 72 k 2 − 20 k 4 − 2 k 6)E + (1 − k 2)(48 − 48 k 2 + 5 k 4)K
2k8  
1  
+ π by 4 8 (48 − 80 k + 31 k )E − (1 − k )(48 − 56 k + 12 k )K
2 4 2 2 4
12 k  
1  2 
+ πb3 x2 6 (8 − 7 k + k )K − (8 − 3 k )E
2 4
2k  
1   + πb 5 1 2 − k 2E − 2K
+ πb3 y2 (8 − 5 k 2
)E − (8 − 9 k 2
+ 2 k 4
)K
2k6   4 k 4 1 − k 2 
3
⌠ ⌠ x 0 y 0d x 0d y 0 = πbx 3 y (1 − k ) (1 − k 2)(16 − 8 k 2)K − (16 − 16 k 2 + k 4)E
2
⌡ ⌡ RZ 0 k8  
S
1  
+ π bx y 3 8 (48 − 104 k + 64 k − 6 k )E − (1 − k )(48 − 80 k + 33 k )K
2 4 6 2 2 4
3k  
+ π b 3 xy 6 (8 − 7 k 2)E − (1 − k 2)(8 − 3 k 2)K

1
k  
4
⌠ ⌠ y 0d x 0d y 0 = πbx 4(1 − 2 k + k ) (16 − 16 k 2 + 3 k 4)K − (16 − 8 k 2)E
2 4
⌡ ⌡ RZ 0 4k
8
 
S
(1 − k 2) 
+ π bx 2 y 2 (48 − 80 k 2 + 31 k 4)E − (1 − k 2)(48 − 56 k 2 + 12 k 4)K
2k
8
 
1  
+ π by 4 8 ( − 48 + 136 k − 132 k + 50 k )E + (1 − k )(48 − 112 k + 85 k − 24 k )K
2 4 6 2 2 4 6
12 k
 
(1 − k 2) 
+ πb3 x2 (8 − k 2)E − (8 − 5 k 2)K
2k  
6
1   5 1  2 
+ πb3 y2 6 ( − 8 + 11 k − 2 k )E + (1 − k )(8 − 7 k )K + π b
2 4 2 2
4 (2 + k )K − (2 + 2 k )E
2
2k
  4k  
The case where x and y are outside the domain S :
⌠ ⌠ x 0d x 0d y 0 = πbx 1 [F(φ) − F(ψ)] − [E(φ) − E(ψ)]

⌡ ⌡ RZ 0 k2  
S
+ π bx 
sin φ cos φ sin ψ cos ψ
+ π by 2 
1 1 1 
− −
 ∆(φ) ∆(ψ)  k  ∆(ψ) ∆(φ)
⌠ ⌠ y 0d x 0d y 0 = πby 1 [E(φ) − E(ψ)] − (1 − k 2)[F(φ) − F(ψ)]

⌡ ⌡ RZ 0 k2  
S
− π by 
sin φ cos φ sin ψ cos ψ
+ π bx 2 
1 − k2 1 1 
− −
 ∆(φ) ∆(ψ)  k  ∆(ψ) ∆(φ)
Appendix A5.1 433
2
⌠ ⌠ x 0d x 0d y 0 = πbx 2 1  − (2 + k 2)[E(φ) − E(ψ)] + 2[F(φ) − F(ψ)]
⌡ ⌡ RZ 0 2k4  
S
1  2  sin φ cos φ sin ψ cos ψ

− (1 − k 2)
sin φ cos φ sin ψ cos ψ
+ π bx 2 2 (2 + k ) − −
2k   ∆(φ) ∆(ψ)   ∆3(φ) ∆3(ψ) 
1 − k 2  1 1  3 − k 2 1 1 
+ π bxy − − 2 ∆(φ) − ∆(ψ)
k  ∆ (φ) ∆ (ψ) 1 − k 
4 3 3

1  
+ π by 2 4 2[E( φ) − E(ψ)] − (2 − k )[F( φ) − F(ψ)]
2
2k  
1   sin φ cos φ sin ψ cos ψ sin φ cos φ sin ψ cos ψ

+ π by 2 −2 − + −
2 k 2   ∆(φ) ∆(ψ)  ∆3(φ) ∆3(ψ) 
− πb3
1  sin φ cos φ − sin ψ cos ψ
2(1 − k )  ∆(φ)
2 ∆(ψ) 
⌠ ⌠ x 0 y 0d x 0d y 0 = πbx 2 1 − k  − 3 1 − 1  + (1 − k 2) 1 − 1 
2
⌡⌡ RZ 0 2 k 4   ∆(φ) ∆(ψ)  ∆3(φ) ∆3(ψ)

S
1 − k 2  k2  
+ π bxy 2 + 2 [E( φ) − E(ψ)] − 2[F(φ) − F(ψ)]
k 
4
1−k  
1 − k 2  sin φ cos φ sin ψ cos ψ 2 − k 2 sin φ cos φ sin ψ cos ψ

+ π bxy − − −
k 2  ∆3(φ) ∆3(ψ) 1 − k 2 ∆(φ) ∆(ψ) 
1 − k2   1 1 
− 3 + 3  + πb3 2 
1 1 1 1 1 
+ π by 2 3 − −
2 k   ∆(φ) ∆(ψ) ∆ (φ) ∆ (ψ)
4
2 k  ∆(φ) ∆(ψ)
2
⌠ ⌠ y 0d x 0d y 0 = πbx 2 1 − k 2[E(φ) − E(ψ)] − (2 − k 2)[F(φ) − F(ψ)]
2
⌡ ⌡ RZ 0 2k4  
S
1 − k 2   sin φ cos φ sin ψ cos ψ

+ (1 − k 2)
sin φ cos φ sin ψ cos ψ
+ π bx 2 −2 − −
2k 2
  ∆ ( φ) ∆ ( ψ)   ∆3(φ) ∆3(ψ) 
1 − k2  2
( k − 1)( 3 − 3 ) + (3 − 2 k 2)
1 1 1 1 
+ π bxy −
k 4
 ∆ (φ) ∆ (φ)  ∆ ( φ) ∆ ( ψ)
1  2 
+ π by 2 4 (3 k − 2)[E( φ) − E(ψ)] + 2(1 − k ) [F(φ) − F(ψ)]
2 2
2k  
1  2  sin φ cos φ sin ψ cos ψ sin φ cos φ sin ψ cos ψ

+ π by 2 2 (2 − 3 k ) − − (1 − k 2) −
2k   ∆ ( φ) ∆ ( ψ)  ∆3(φ) ∆3(ψ) 
− πb3 
1 sin φ cos φ sin ψ cos ψ
−
2  ∆(φ) ∆(ψ) 
3
⌠ ⌠ x 0d x 0d y 0 = πbx 3 1 (8 − 3 k 2 + k 4)[F(φ) − F(ψ)] − (8 + k 2 + 2 k 4)[E(φ) − E(ψ)] +
⌡ ⌡ RZ 0 6k6  
S
1  4 sin φ cos φ sin ψcos ψ

+ π bx 3 2
4 (8 + k + 2 k ) − +
6k   ∆(φ) ∆(ψ) 
+(− 7 + 6 k 2 + k 4) 3
sin φcos φ sin ψcos ψ
+ 3(1 − k 2)2 5
sin φcos φ sin ψcos ψ
− −
 ∆ (φ) ∆ (ψ) 
3
 ∆ (φ) ∆5(ψ) 
1  2  1 1 
+ π bx 2 y 6 (9 k − 15) ∆(φ) − ∆(ψ) +
2k   
+(1 − k 2)(10 − 3 k 2) 3 − 3  − 3(1 − k 2)2 5 − 5 

1 1 1 1
∆ (φ) ∆ (ψ) ∆ (φ) ∆ (ψ)
1  
+ π bxy 2 6 ( − 8 + 5 k )[F( φ) − F(ψ)] − (8 − k )[E( φ) − E(ψ)]
2 2
2k  
+ π bxy 2
1  4 sin φcos φ sin ψcos ψ
2 ( − 8 + 3k + 2k ) −
2
+
2 k (1 − k )
4
  ∆(φ) ∆(ψ) 
+(7 − k 2) 3
− 3(1 − k 2) 5
− − +
 ∆ (φ) ∆ (ψ) 
3
 ∆ (φ) ∆5(ψ) 
1  2  1 1  2  1
− 3  + 3(1 − k 2) 5 − 5 
1 1 1
+ π by 3 6 3(5 − 3 k ) ∆(φ) − ∆(ψ) + (10 − 8 k )
6k    ∆ (φ) ∆ (ψ)
3
∆ (φ) ∆ (ψ)
Appendix A5.1 435
+ π xb 3
1  − 2[F(φ) − F(ψ)] + (2 − k 2)[E(φ) − E(ψ)]
2 k (1 − k ) 
4 2

+ π xb 3
+ (1 − k 2) 3
2 (−2+k ) − −
2 k (1 − k )   ∆(φ) ∆(ψ)   ∆ (φ) ∆3(ψ) 
2
1   1 1   1
− 3 
1
+ π yb 3 4 3 ∆(φ) − ∆(ψ) −
2k    ∆ (φ) ∆ (ψ)
3
2
⌠ ⌠ x 0 y 0d x 0d y 0 = πbx 3 1 − 3(1 − k 2)(5 − k 2) 1 − 1  +
⌡ ⌡ RZ 0 6k6  ∆(φ) ∆(ψ)
S
+(1 − k 2)2(10 − k 2) 3 − 3  − 3(1 − k 2)3 5 − 5 

1 1 1 1
∆ (φ) ∆ (ψ) ∆ (φ) ∆ (ψ)
1  
+π bx 2 y 6 −(8 − k )(1 − k )[F( φ) − F(ψ)] + (8 − 5 k − k )[E( φ) − E(ψ)]
2 2 2 4
2k  

+ π bx 2 y 4 (8 − 9 k + 3 k )
2
− +
2k   ∆(φ) ∆(ψ) 
+(7 − k 2)(1 − k 2) 3

− 3(1 − k 2)2 5
− −
 ∆ (φ) ∆ (ψ) 
3
 ∆ (φ) ∆5(ψ) 
1 − k2 
(15 − 3 k 2)
1 
− (10 − 6 k 2) 3 − 3 +
1 1 1
+ π bxy 2 −
2k 6
 ∆ ( φ) ∆ ( ψ)  ∆ (φ) ∆ (ψ)
+ 3(1 − k 2) 5 − 5 
1 1
∆ (φ) ∆ (ψ)
1  
+π by 3 6 (8 − 3 k )(1 − k )[F( φ) − F(ψ)] − (8 − 7 k )[E( φ) − E(ψ)]
2 2 2
6k  

− 7(1 − k 2) 3
+ π by 3 4 (8 − 7 k ) − − +
6k   ∆ ( φ) ∆ ( ψ)   ∆ (φ) ∆3(ψ) 
+ 3(1 − k 2) 5
−
 ∆ (φ) ∆5(ψ) 
1  2  1 1  2  1
− 3 
1
+ π xb 3 4 (3 − k ) ∆(φ) − ∆(ψ) − (1 − k )
2k    ∆ (φ) ∆ (ψ)
3
1  
+ π yb 3 4 (2 − k )[F( φ) − F(ψ)] − 2[E( φ) − E(ψ)]
2
2k  
1  sin φcos φ sin ψcos ψ sin φcos φ sin ψcos ψ

+ π yb 3 2 − − −
2 k 2   ∆(φ) ∆(ψ)   ∆3(φ) ∆3(ψ) 
2
⌠ ⌠ x 0 y 0d x 0d y 0 = πbx 31 − k  − (8 − 5 k 2)[F(φ) − F(ψ)] + (8 − k 2)[E(φ) − E(ψ)]
2
⌡ ⌡ RZ 0 6k6  
S
1 − k2 
− (8 − k 2)
+ 7(1 − k 2) 3
sin φcos φ
+ π bx 3 − −
6k 
4
 ∆(φ) ∆(ψ)   ∆ (φ)
sin ψcos ψ
− 3(1 − k 2)2 5
− −
∆ (ψ) 
3
 ∆ (φ) ∆5(ψ) 
1 − k2 
(15 − 12 k 2)
1 
− (1 − k 2)(10 − 4 k 2) 3 −
1 1
+ π bx 2 y −
2k6  ∆(φ) ∆(ψ) ∆ (φ)
1 
+ 3(1 − k 2)2 5 − 5 
1 1
−
∆ (ψ)
3
∆ (φ) ∆ (ψ)
1  
+ π bxy 2 6 (8 − 7 k )(1 − k )[F( φ) − F(ψ)] + ( − 8 + 11 k − 2 k )[E( φ) − E(ψ)]
2 2 2 4
2k  

+ π bxy 2 4 (8 − 11 k + 2 k )
2
− −
2k   ∆(φ) ∆(ψ) 
− (7 − 5 k 2)(1 − k 2) 3
+ 3(1 − k 2)2 5
− −
 ∆ (φ) ∆ (ψ) 
3
 ∆ (φ) ∆5(ψ) 
1 − k2 
− 3(5 − 4 k 2)
1 
+ (10 − 9 k 2) 3 − 3  − 3(1 − k 2) 5 − 5 
1 1 1 1 1
+ π by 3 −
6k 6
 ∆ ( φ) ∆ ( ψ)  ∆ (φ) ∆ (ψ)  ∆ (φ) ∆ (ψ)
1  
+ π xb 3 4 (2 − k )[F( φ) − F(ψ)] − 2[E( φ) − E(ψ)]
2
2k  
Appendix A5.1 437
1  sin φcos φ sin ψcos ψ

− (1 − k 2) 3
+ π xb 3 2 2 − −
2k   ∆ ( φ) ∆ ( ψ)   ∆ (φ) ∆3(ψ) 
1  2  1 1  2  1
− 3 
1
+ π yb 3 4 − (3 − k ) ∆(φ) − ∆(ψ) + (1 − k )
2k    ∆ (φ) ∆ (ψ)
3
3
⌠ ⌠ y 0d x 0d y 0 = πbx 3(1 − k ) 3(5 − 2 k 2) 1 − 1  − (1 − k 2)(10 − 2 k 2) 1 −
2 2
⌡ ⌡ RZ 0 6k6  ∆(φ) ∆(ψ) ∆3(φ)

S
1 
+ 3(1 − k 2)2 5 − 5 
1 1
−
∆ (ψ)
3
 ∆ (φ) ∆ (ψ)
1 − k2 
+ π bx 2 y (8 − k 2)(1 − k 2)[F(φ) − F(ψ)] − (8 − 7 k 2)[E( φ) − E(ψ)]
2k 
6

1 − k2 
(8 − 7 k 2)
− (7 − 3 k 2)(1 − k 2) 3
+ π bx y
2
− −
2k 
4
 ∆ ( φ) ∆ ( ψ)   ∆ (φ) ∆3(ψ) 
+ 3(1 − k 2)2 5
−
 ∆ (φ) ∆5(ψ) 
(1 − k 2)2 
− (15 − 6 k 2)
1 
+ (10 − 7 k 2) 3 − 3  − 3(1 − k 2) 5 − 5 
1 1 1 1 1
+ π bxy 2 −
2k 
6
∆(φ) ∆(ψ) ∆ (φ) ∆ (ψ) ∆ (φ) ∆ (ψ)
1  
+ π by 3 6 − (8 − 13 k + 6 k )(1 − k )[F( φ) − F(ψ)] + (8 − 17 k + 11 k )[E( φ) − E(ψ)]
2 4 2 2 4
6k  

+ (1 − k 2)(7 − 8 k 2) 3
+ π by 3 4 (− 8 + 17 k − 11 k )
2
− − −
6k   ∆(φ) ∆(ψ)   ∆ (φ) ∆3(ψ) 
− 3(1 − k 2)2 5
−
 ∆ (φ) ∆5(ψ) 
1 − k2   1 1 
+ (1 − k 2) 3 − 3 
1 1
+ π xb 3 −3 −
2k 4
 ∆ ( φ) ∆ ( ψ)  ∆ (φ) ∆ (ψ)
1  
+ π yb 3 4 − 2(1 − k )[F( φ) − F(ψ)] + (2 − k )[E( φ) − E(ψ)]
2 2
2k  

+ (1 − k 2) 3
+ π yb 3 2 − (2 − k ) − −
2k   ∆ ( φ) ∆ ( ψ)   ∆ (φ) ∆3(ψ) 
In the integrals above, the abbreviations F(φ), E(φ), F(ψ), E(ψ) were used to
denote F(φ, k ), E(φ, k ), F(ψ, k ), E(ψ, k ) respectively. The notation ∆(φ) stands for
(1 − k 2sin2φ)1/2
The case of a circle of radius a and x, y inside the domain S :

2
⌠ ⌠ x 0d x 0d y 0 = π2 a 1 4 a 2 + 5 x 2 − y 2
⌡ ⌡ RZ 0 16  
S
⌠ ⌠ x 0 y 0d x 0d y 0 = π2 a 3 xy
⌡⌡ RZ 0 8
S
2
⌠ ⌠ y 0d x 0d y 0 = π2 a 1 4 a 2 − x 2 + 5 y 2
⌡ ⌡ RZ 0 16  
S
3
⌠ ⌠ x 0d x 0d y 0 = π2 ax 1 6 a 2 + 7 x 2 − 3 y 2
⌡ ⌡ RZ 0 32  
S
2
⌠ ⌠ x 0 y 0d x 0d y 0 = π2 ay 1 2 a 2 + 9 x 2 − y 2
⌡⌡ RZ 0 32  
S
2
⌠ ⌠ x 0 y 0d x 0d y 0 = π2 ax 1 2 a 2 − x 2 + 9 y 2
⌡⌡ RZ 0 32  
S
3
⌠ ⌠ y 0d x 0d y 0 = π2 ay 1 6 a 2 − 3 x 2 + 7 y 2
⌡ ⌡ RZ 0 32  
S
References
A.E. Andreikiv and V.V. Panasiuk, Elastic equilibrium of a body weakened by a

set of circular cracks situated along a plane. Soviet Physics – Doklady , Vol. 16,
1971, pp. 255-256.
H. Bateman, A. Erdélyi, (Ed.), Higher Transcendental Functions . (3 Vols.). New

York: McGraw-Hill, 1955.
E. Beltrami, Sulla teoria delle funczioni potenziali simmetriche.

Mem. Accad. Sci. Inst. Bologna , Ser. 4, Vol. 2, 1881, pp. 461-480.
H.A. Bethe, Theory of diffraction by small holes. Physical Reviews, Vol. 66,
1944, pp. 163-182.
N.M. Borodachev and L.A. Galin, Contact problem for a stamp with narrow
rectangular base. Journal of Applied Mathematics and Mechanics (PMM) ,
Vol. 38, 1974, pp. 108-113.
I.W. Busbridge, Dual integral equations. London Mathematical Society.

Proceedings., Ser. 2, Vol. 44, 1938, pp. 115-129.
W.T. Chen, Some aspects of a flat elliptical crack under shear stress. Journal
of Mathematics and Physics, Vol. 45, 1966, p.213-223.
G.P. Cherepanov, Mechanics of Brittle Fracture. Nauka, Moscow, 1974 (in

Russian). English translation: McGraw-Hill, 1979.
S.B. Cohn, The electric polarizability of apertures of arbitrary shape. Proc. of

the IRE , Vol. 40, 1952, pp. 1069-1071.
W.D. Collins, Some coplanar punch and crack problems in three-dimensional

elastostatics. Proc. Roy. Soc., Series A , Vol. 274, 1963, pp. 507-528.
W.D. Collins, Some axially symmetric stress distributions in elastic solids

containing penny-shaped cracks. I. Cracks in an infinite solid and a thick plate.
Proc. Roy. Soc., Series A, Vol. 266, 1962, pp. 359-386.
E.T. Copson, On the problem of the electrified disk. Edinburgh Mathematical

Society. Proceedings., Ser. 2, Vol. 8, 1947, pp. 14-19.
H.A. Elliott, Three-dimensional stress distributions in hexagonal aeolotropic

crystals. Proc. Cambridge Phil. Soc., Vol. 44, 1948, pp. 522-533.
439
440 References
H.A. Elliott, Axial symmetric stress distributions in aeolotropic hexagonal

crystals. The problem of the plane and related problems. Proc. Cambridge
Phil. Soc., Vol. 45, 1949, pp. 621-630.
V.I. Fabrikant, Close interaction of coplanar circular cracks under shear loading.
Computational Mechanics, Vol. 4, 1989, pp. 181-197.
V.I. Fabrikant, Potential of several arbitrarily located discs. Journal of the

Australian Mathematical Society, Series B , Vol. 29, 1988a, pp. 342-351.
V.I. Fabrikant, Green’s functions for a penny-shaped crack under normal loading.
Engineering Fracture Mechanics, Vol. 30, 1988b, pp. 87-104.
V.I. Fabrikant, Sound penetration through an arbitrarily shaped aperture in a soft

screen: Analytical approach. Journal of Sound and Vibration , Vol. 121, 1988d,
pp. 1-12.
V.I. Fabrikant, Penny-shaped crack revisited: Closed form solutions. Philosophical

Magazine A , Vol. 56, 1987a, pp. 191-207.
V.I. Fabrikant, Flat crack of arbitrary shape in an elastic space. Philosophical

Magazine A , Vol. 56, 1987b, pp. 175-189.
V.I. Fabrikant, Electrical polarizability of small apertures: Analytical approach.

International Journal of Electronics, Vol. 62, 1987c, pp. 533-545.
V.I. Fabrikant, Mixed problems of Potential Theory in spherical coordinates.

Zeitschrift fu¨ r Angewandte Mathematik und Mechanik , Vol. 67, 1987d,
pp. 507-518.
V.I. Fabrikant, Electrostatic problem of several arbitrarily charged unequal coaxial

discs. Journal of Computational and Applied Mathematics , Vol. 18, 1987e,
pp. 129-147.
V.I. Fabrikant, The stress intensity factor for an external elliptical crack.
International Journal of Solids and Structures, Vol. 23, 1987f, pp. 465-467.
V.I. Fabrikant, Close interaction of coplanar circular cracks in an elastic

medium. Acta Mechanica , Vol. 67, 1987g, pp. 39-59.
V.I. Fabrikant, Frictionless elastic contact problem for a curved rigid punch of
arbitrary shape. Acta Mechanica , Vol. 67, 1987h, pp. 1-25.
References 441
V.I. Fabrikant, Closed form solution to some mixed boundary value problems for
a charged sphere. Journal of the Australian Mathematical Society, Series B ,
Vol. 28, 1987i, pp. 296-309.
V.I. Fabrikant, Magnetic polarizability of small apertures: Analytical approach.

Journal of Physics A: Mathematical and General, Vol. 20, 1987j, pp. 323-338.
V.I. Fabrikant, Diffusion through perforated membranes. Journal of Applied

Physics, Vol. 61, 1987k, pp. 813-816.
V.I. Fabrikant, Flat punch of arbitrary shape on an elastic half-space.

International Journal of Engineering Science, Vol. 24, 1986a, pp. 1731-1740.
V.I. Fabrikant, Sound transmission through an arbitrarily shaped aperture in a

rigid screen: Analytical approach. Journal of Sound and Vibration , Vol. 111,
1986b, pp. 489-498.
V.I. Fabrikant, Inclined flat punch of arbitrary shape on an elastic half-space.

ASME Journal of Applied Mechanics, Vol. 53, 1986c, pp. 798-806.
V.I. Fabrikant, Sound penetration through an arbitrarily shaped aperture in a rigid

screen: Analytical determination of the quadratic terms in low-frequency
expansion. Journal of the Acoustical Society of America , Vol. 50, 1986d, pp.
1438-1446.
V.I. Fabrikant, Computer evaluation of singular integrals and their applications.

International Journal for Numerical Methods in Engineering , Vol. 23, 1986e,
pp. 1439-1453.
V.I. Fabrikant, Exact solutions to some external mixed problems in potential

theory. Aplikace Matematiky , Vol. 31, 1986f, pp. 224-246.
V.I. Fabrikant, Inverse crack problem in elasticity. Acta Mechanica , Vol. 61,
1986g, pp. 29-36.
V.I. Fabrikant, A new approach to some problems in Potential Theory.

Zeitschrift fu¨ r Angewandte Mathematik und Mechanik , Vol. 66, 1986h,
pp. 363-368.
V.I. Fabrikant, Several elliptical punches on an elastic half-space. ASME Journal

of Applied Mechanics, Vol. 53, 1986i, pp. 390-394.
V.I. Fabrikant, Four types of exact solutions to the problem of an axisymmetric

punch bonded to a transversely isotropic half-space. International Journal of
Engineering Science, Vol. 24, 1986j, pp. 785-801.
442 References
V.I. Fabrikant, On the capacity of flat laminae. Electromagnetics, Vol. 6, 1986k,

pp. 117-128.
V.I. Fabrikant, On the potential flow through membranes. Zeitschrift fu¨ r

Angewandte Mathematik und Physik , Vol. 36, 1985a, pp. 616-623.
V.I. Fabrikant, External crack in non-homogeneous elasticity. Engineering

Fracture Mechanics, Vol. 22, #5, pp. 855-858, 1985b.
V.I. Fabrikant, Closed form solution of one integral equation. Trans. of Academy
of Sciences of ArmSSR, Mekhanika , No. 2, pp. 16-19, 1976, (in Russian).
V.I. Fabrikant, Interior fundamental mixed problem for a transversely Isotropic

Half-Space. Mekhanika Tverdogo Tela , No. 1, pp. 27-33, 1975, (in
Russian). English translation: Mechanics of Solids , Vol. 10, pp. 22-26.
V.I. Fabrikant, The three-dimensional contact problem for a rough die.

Prikladnaia Mekhanika , Vol. 10, No. 7, pp. 106-109, 1974a, (in Russian). English
translation: Soviet Applied Mechanics, pp. 772-774.
V.I. Fabrikant, Exact Solution of the exterior fundamental mixed problem.

Prikladnaia Matematika i Mekhanika , Vol. 38, No. 3, pp. 534-538, 1974b, (in
Russian). English translation: Journal of Applied Mathematics and Mechanics,
pp. 494-499.
V.I. Fabrikant, Exterior axisymmetric mixed problem for a transversely isotropic

half-space. Prikladnaia Matematika i Mekhanika , Vol. 36, pp. 947-951, 1972, (in
Russian). English translation: Journal of Applied Mathematics and Mechanics,
pp. 897-901.
V.I. Fabrikant, Axially symmetric problem of a stamp on a transversely isotropic

half-space. Mekhanika Tverdogo Tela , No. 6, pp. 141-146, 1971a, (in
Russian). English translation: Mechanics of Solids , pp. 121-125.
V.I. Fabrikant, Solution of a non-axisymmetric boundary-value problem with

circular line of division of boundary conditions for a transversely-isotropic
half-space. Trans. of Academy of Sciences of UzSSR, Series of Technical
Sciences, No. 3, pp. 93-96, 1971b, (in Russian).
V.I. Fabrikant, One non-axisymmetric mixed problem for a transversely isotropic

half-space. Prikladnaia Mekhanika , Vol. 7, No 3, pp. 36-40, 1971c, (in
Russian). English translation: Soviet Applied Mechanics, pp. 264-267.
References 443
V.I. Fabrikant, Effect of shearing force and tilting moment on a cylindrical punch
attached to a transversely isotropic half-space. [Prikladnaia Matematika i
Mekhanika], Vol. 35, No. 1, pp. 178-182, 1971d, (in Russian). English translation:
Journal of Applied Mathematics and Mechanics, pp. 147-151.
V.I. Fabrikant, Closed form solution of a two-dimensional integral equation.

Izvestiia Vysshikh Uchebnykh Zavedenii, Matematika , No. 2, pp. 102-104, 1971e,
(in Russian).
V.I. Fabrikant, Effect of concentrated force on a transversely isotropic elastic

body. Izvestiia Vysshikh Uchebnykh Zavedenii, Mashinostroenie No. 3, 1970,
pp. 9-12 (in Russian).
R.F. Fikhmanas and P.Sh. Fridberg, Theory of diffraction at small

apertures. Computation of upper and lower boundaries of the polarizability
coefficients. Radio Eng. Electron Phys. Vol. 18, 1973, pp. 824-829.
W.S. Fu and L.M. Keer, Coplanar circular cracks under shear loading.
International Journal of Engineering Science, Vol. 7, 1969, pp. 361-372.
L.A. Galin, Contact Problems in the Theory of Elasticity (in Russian),

Gostekhteorizdat, 1953. Translated by H. Moss, Dept. of Math., North Carolina
State College, Raleigh, 1961.
I.S. Gradshtein and I.M. Ryzhik, Tables of Integrals, Series and

Products. Moscow, 1963. English translation: Academic Press Inc., New York,
1965.
E.W. Hobson, On Green’s Function for a Circular Disc, with applications to

Electrostatic Problems. Trans. Cambr. Phil. Soc., Vol. 18, 1900,
pp. 277-291.
G.W.O. Howe, The capacity of rectangular plates and a suggested formula for the
capacity of aerials. The Radio Review , Vol. 1, Oct. 1919 – June 1920,
pp. 710-714.
N.I. Ioakimidis, Upper bounds for the stress intensity factors along the boundaries
of interacting coplanar cracks in three-dimensional elasticity. Engineering Fracture
Mechanics, Vol. 16, 1982, pp. 821-826.
J.J. Kalker, On elastic line contact. ASME Journal of Applied Mechanics ,

Vol. 39, 1972, pp. 1125-1132.
A.A. Kapshivyi and G.F. Masliuk, The solution of the mixed axisymmetric
problem from the theory of elasticity for a half-space by the method of
444 References
p -analytic functions. Prikladnaia Mekhanika , Vol. 3, No. 7, 1967, pp. 21-27,

English translation: Soviet Applied Mechanics, pp. 13-16.
M.K. Kassir and G. Sih, Three-dimensional stresses around elliptical cracks in
transversely isotropic solids. Engineering Fracture Mechanics, Vol. 1, 1968,
pp. 327-345.
M.K. Kassir and G. Sih, Three-dimensional Crack Problems, Noordhoff

International Publishing, Leyden, 1975.
I. Kobayashi, Das electrostatische Potential um zwei auf derselben Ebene liegende

und sich nicht schneidende gleichgroße Kreisscheiben. Sc. Rep. Toˆ hoku
Imp. Univ., Vol. 27, 1939, pp. 365-391.
M.Ia. Leonov, General problem of a circular punch pressed against an elastic

half-space. Prikladnaia Matematika i Mekhanika , Vol. 17, 1953, pp. 87-98 (in
Russian).
A.I. Lur’e, Three-Dimensional Problems of the Theory of Elasticity . Moscow:

1955. English translation, J.R.M. Radok (Ed.). New York: Interscience, 1964.
E.N. Mastrojannis and T. Mura, On the problem of two coplanar cracks inside
an infinite isotropic elastic solid. International Journal for Numerical Methods in
Engineering , Vol. 19, 1983, pp. 27-35.
F. De Meulenaere and J. Van Bladel, Polarizability of some small apertures.

IEEE Trans. on Antennas and Propagation , Vol. AP-25, 1977, pp. 198-205.
R.D. Mindlin, Compliance of elastic bodies in contact. Journal of Applied

Mechanics., 1949, Vol. 16, pp. 259-268.
V.I. Mossakovskii, The fundamental mixed problem of the theory of elasticity for
a half-space with a circular line separating the boundary conditions. Prikladnaia
Matematika i Mekhanika , Vol. 18, 1954, pp. 187-196. (in Russian).
V.I. Mossakovskii, Relationship between the force and settlement for a plane
stamp of nearly circular cross-section. Gidroaeromekhanika i Teoriia Uprugosti,
No. 14, Izd. Dnepropetrovsk. Univ., 1972 (in Russian).
V.I. Mossakovskii, N.E. Kachalovskaia, and S.S. Golikova, Contact Problems of

the Mathematical Theory of Elasticity , ”Naukova Dumka”, Kiev, 1985 (in
Russian).
N.I. Muskhelishvili, Singular Integral Equations . (in Russian). Second ed.,

Moscow 1946. English Translation, J.R.M. Radok (Ed.). Groningen: Noordhoff,
1953.
References 445
B. Noble, The numerical solution of the singular integral equation for the charge
distribution on a flat rectangular lamina. Sympos. Numerical Treatment of
Ordinary Differential Equations, Integral and Integrodifferential Equations.
(Proc. Rome Sympos. 20-24 September 1960) Birkhauser, Berlin-Stuttgart, 1960.
E.E. Okon and R.F. Harrington, The polarizabilities of electrically small apertures
of arbitrary shape. IEEE Trans. on Electromagnetic Compatibility , Vol. EMC-23,
1981, pp.359-366.
E.E. Okon and R.F. Harrington, The capacitance of discs of arbitrary

shape. Tech. Rep. No.10, Contract No. N00014-76-C-0225 . US Dep. Navy, Office
of Naval Research, Rep. No. TR-79-3, April, 1970.
V.V. Panasiuk, M.M. Stadnik, and V.P. Silovaniuk, Stress Concentration in

Three-dimensional Bodies with Thin Inclusions. ”Naukova Dumka”, Kiev, 1986
(in Russian).
G. Pólya and G. Szegö, Isoperimetric Inequalities in Mathematical Physics. Princeton

University Press, 1951.
G.Ia. Popov, Axisymmetric contact problem for an elastic inhomogeneous

half-space in the presence of cohesion. Prikladnaia Matematika i Mekhanika.,
Vol. 37, 1973, pp. 1109-1116, English Translation: Journal of Applied
Mathematics and Mechanics, pp. 1052-1059.
N.A. Rostovtsev, An integral equation encountered in the problem of a rigid

foundation bearing on nonhomogeneous soil. Prikladnaia Matematika i
Mekhanika , Vol. 25, 1961, pp. 164-168, English Translation: Journal of Applied ,
Mathematics and Mechanics, pp. 238-246.
N.A. Rostovtsev, On certain solutions of an integral equation of the theory of

linearly deformable foundation. Prikladnaia Matematika i Mekhanika , Vol. 28,
1964, pp. 111-127, English Translation: Journal of Applied Mathematics and
Mechanics, pp. 127-145.
V.L. Rvachev and V.S. Protsenko, Contact Problems of the Theory of Elasticity
for Non-classical Domains. Kiev, ”Naukova Dumka”, 1977 (in Russian).
G.C. Sih and H. Liebowitz, Mathematical theories of brittle fracture. Fracture,

Vol. 2. Academic Press, New York, 1968.
R. De Smedt, Low-frequency Penetration through Apertures: Results for the

Integral Equations. Laboratorium voor Electromagnetisme en Acustica, Internal
report 79-9, University of Ghent, 1979.
446 References
I.N. Sneddon, Fourier Transforms. McGraw-Hill, New York, 1951.
I.N. Sneddon, A note on the problem of the penny-shaped crack. Proc.

Cambridge Philos. Soc., Vol. 61, 1965, pp. 609-611.
I.N. Sneddon, Mixed Boundary Value Problems in Potential , Theory.

Amsterdam: North Holland, 1966.
L. Solomon, Upon the geometrical punch-penetration rigidity.

Lincei-Rend. Sc. fis. mat. e nat. , Vol. 36, 1964a, pp. 832-835.
L. Solomon, Une solution approchée du problème du poinçon rigide à base plane

bornée convexe non elliptique. Compt. Rend. Acad. Sc. Paris, Vol. 258, 1964b,
pp. 64-66.
Ia.S. Ufliand, The contact problem of the theory of elasticity for a die, circular
in its plane, in the presence of adhesion. Prikladnaia Matematika i Mekhanika ,
Vol. 20, 1956, pp. 578-587, (in Russian).
Ia.S. Ufliand, Integral Transforms in the Theory of Elasticity . Second edition:

Nauka, Leningrad, 1967. English translation of the first edition: Survey of Articles
on the Application of Integral Transforms in the Theory of Elasticity . North
Carolina State University, Applied Mathematics Research Group, File
No. PSR-24/6, 1965.
Ia.S. Ufliand, Method of Dual Equations in Mathematical Physics

Problems. Academy of Sciences of the USSR, ”Nauka”, Leningrad, 1977 (in
Russian).
J. Weaver, Three-dimensional crack analysis. International Journal of Solids and

Structures, Vol. 13, 1977, pp. 321-330.
H. Weber, Ueber die Besselschen Functionen und ihre Anwendung auf die
Theorie der elektrischen Strome. Journal fu¨ r die reine und angewandte
Mathematik , Vol. 75, 1873, pp. 75-105.
R.A. Westmann, Asymmetric mixed boundary-value problem of the elastic

half-space. ASME Journal of Applied Mechanics., Vol. 32, 1965, pp. 411-417.
J.R. Willis, The distribution of stress in anisotropic elastic body containing an

exterior crack. International Journal of Engineering Science, 1970, Vol. 8,
No. 7, pp. 559-574.
Subject Index 447
Subject Index
The page numbering in this Index is approximate because present edition of the
book does not conserve the pages from original publication.
Abel operator, 27-30, 42, 61, 65, 82, Boundary value problems,
92, 94, 183, 287 see also Axisymmetric problems
Abel integral, 168 internal mixed of type I, 80-91
Abel generalized integral equation, internal mixed of type II,
182-187 105-111
Annulus, uniform pressure distributed external mixed of type I,
over, 90 91-101
Approach, 2, 5, 6, 36, 182, 268, external mixed of type II,
280, 303, 350, 379, 402, 407, 413 111-131
Axisymmetric Boussinesq problem, 422
pressure, 5 see also Concentrated load, point
problems force solutions
internal mixed-mixed,
148-176 Circular
external mixed-mixed, crack, 7, 83, 85, 89, 90, 109,
176-182 237, 285-319
crack, 227, 239 disk, 1, 3-4, 38
punch, 2, 7, 26, 88, 145, 148,
Bessel functions, 4,5 163, 172-176, 186-193,
integrals of, 253-256, 348 349 341-350, 418, 432, 445-448
Bonded contact, 7, 145, 148-176 sector, 277, 360, 382-383
asymmetric punch problem, segment, 276, 359, 376
187-193 Complex
concentrated load, influence of, conjugate, 38
200-206 constants, 114
flexible punches, 434 functions, 107
non-homogeneous half-space, stress intensity factor, 120
186 stresses, 74
two-dimensional problems, 447 tangential displacements, 72
Boundary conditions, 3, 27, 39, 44, Compliance, 411
50-59, 80, 91, 105, 111, 131, 145, Concentrated load
188, 239, 337, 406, 412, 422, 433 outside a crack, 237-239
outside a punch, 175, 200,
211, 349, 433
point force solutions, 75-80
Contact, adhesive
see Bonded contact
Contact problems
see also Boundary value problems
448 Subject Index
for a circular punch, 341-350 shear loading, 239-252

for a flat punch of general point force loading of a
shape, 350-383 penny-shaped crack, 234-237
for a flexible punch under rectangular, 274, 281, 285
shifting load, 406-412 rhomboidal, 275
for a general curved punch,
383-406 Deformation, energy of, 318
for a rough punch, 445-448 Dirac delta-function, 109
for a smooth punch, 337-341 Dirichlet
interaction between punches, conditions, 7
422-445 problem, 17
Reissner-Sagoci problem, Displacement
412-422 complex tangential, 72
Contact, torsional normal, 2, 7
see Contact problems, Reissner-Sagoci outside a penny-shaped crack,
problem 255, 258-260
Contour, 269, 279, 352, 360 polynomial, 3
Crack problems, under a bonded punch, 148-176
see also Boundary value problems, under a circular punch,
asymptotic behavior of stresses 341-346
and displacements near the under a point load, 75-80
crack rim, 262-268 Dual integral equations, 3-4
circular sector shape, 277
circular segment shape, 276 Elastic constants, 71, 73, 75, 241,
close interaction of pressurized 319, 418
coplanar circular cracks, Elastic half-space, 2, 7, 8, 26
285-304 see also Elastic constants
close interaction of coplanar non-homogeneous, 81
circular cracks under shear transversely isotropic, 71, 77
loading, 304-319 Electrical capacity, 362
concentrated load outside a Electrical charge distribution, 358
circular crack, 237-239 Electrical polarizability, 272
cross-shaped, 278 Electrified disk, 1
flat crack of general shape, Elliptic integrals, 134, 377, 409,
268-284 417, 439
flat crack under arbitrary Elliptic punch, see Punch elliptic
normal loading, 227-234 Equilibrium
general crack under uniform conditions, 191
shear, 284-285 equations, 72
hexagon-shaped, 273
penny-shaped crack under Force
uniform pressure, 252-257 see also Concentrated load;
penny-shaped crack under complex tangential, 76
uniform shear loading, 257-262 friction, 8, 191
plane crack under arbitrary normal, 148
Subject Index 449
resultant, 83, 134, 153 for a smooth punch problem,

shearing, 187, 191 338
unit, 305 for mixed boundary value
Fourier series, 7, 9, 17, 37, 81, problems of potential theory,
105, 112, 193, 284 27, 40
Fracture, see Crack problems for non-homogeneous half-space,
Fredholm integral equations, 304 80, 91
Friction set of, 406-444
coefficient, 445 Integral representation
forces 8, 191 for the reciprocal of the
Function distance between two points,
see also Bessel functions 9-17
complex stress, 152-172 involving inverse cube of the
Green’s, 1, 6, 234, 245, 340 distance, 19-25, 101-105
hypergeometric, 36 of more general type, 60-70
potential, 5, 26, 232, 339 Integro-differential equation, 229,
241, 257
Green’s functions see functions, Isotropic medium, 238, 255, 259-260,
Green’s 344-345
Half-space see also Elastic Jacobi polynomial expansion, 182

half-space; Axisymmetric problem;
Bonded contact; Boundary value Kernel of the integral equation, 302,
problems; Contact problems 407
Boussinesq problem, 75-80 singular part of, 198, 209
loading over annulus, 173-174 degenerate part of, 202,211
non-homogeneous, 186 Klein-Gordon equation, 17
torsion of, 412-422
transversely isotropic, 71 Laplace
Harmonic function, 3, 7, 30, 41, 44, equation, 29
73, 228, 239 operator, 3, 41, 134, 286
L-operator, 17-19
Inclusion problem, 131-144, 302 Loading,
Indentation, axisymmetric, 5
see Bonded contact; Boundary value distributed over annulus, 90
problems; Contact problems point force, 200-206
Integral equations shear, 239-252
see also Dual integral equations uniform, 252-257
approximate solutions for
general domains, 341-411 Magnetic polarizability, 369-383
for a bonded punch problem, Mehler-Fok transform, 200
193, 206 Moment
for a half-space under shear complex, 135
loading, 105, 111 inertia, 366,382
for a rough punch, 446 linear, 365
450 Subject Index
linear of fourth order, 387 Rhombus, see Punch, rhomboidal;

tilting, 83, 94, 187-193, 367 Crack problems, rhomboidal
Neumann boundary conditions, 7, 32 Saint-Venant theory

Normal stress, 71 of bending, 382
of torsion, 419
Penetration, see Indentation Separation in contact problems, 373,
Poisson 377
coefficient, 75, 166, 171, 318 Shear modulus, 238
operator, 17 Shear stress, 136, 304, 407, 420
Polynomial Singular integral, 153
displacements, 3, 408 Singular integral equation, 152, 198
kernel, 198, 209 Sphere, 15, 39
loading, 6, 231, 232, 268, 307 Spherical bowl, 1
profile, 8 Spherical coordinates, 262
Punch Spheroidal inclusions, 302
see also Contact problems Strain energy, 318
arbitrary planform, 350-421 Stress
bonded, 148-176, 187-193 complex, 74
circular, 341-350 components of in cylindrical
circular sector base, 360, polar coordinates, 345-346
382-383 function, 152-172
circular segment base, 359, 376 intensity factor, 86, 90, 97,
cross-shaped, 378, 397, 405, 120, 143, 237, 251, 263-267,
421 289, 309
curved, 383-406 intensity function, 86, 91, 97
elliptical, 409, 417, 426-433, normal, 71
438-445 Stress-strain relations, 71
inclined, 346-350, 362-383 Surface
interaction between, 422-445 displacements, 165
polygonal base, 355, 369, 391, quadratic, 383
418 tractions, 164
rectangular, 358, 372, 380,
394, 403, 410, 419 Toroidal coordinates, 1, 15
rigid roller, 398-402 Traction, see Stress
rhomboidal, 359, 374, 380, 395 Transforms, see under particular
rough, 445-448 name
settlement, 89, 165, 172, 341, Transversely isotropic material, 71-74
350, 357, 400
smooth, 337-341 Uniform
triangular base, 356, 371 electrical charge distribution, 38
pressure, 90, 180, 252, 268,
Rectangle, see Punch, rectangular; 290
Crack problems, rectangular shear tractions, 130, 257, 284,
Reissner-Sagoci problem, 412-422 307
Subject Index 451
tangential displacements, 406, Work, 85, 95, 426, 438

433
Variational approach, 280, 379, 402

Applications of Potential Theory in Mechanics: Selection of New Results

Enviado por

Dados do documento

Descrição original:

Título original

Direitos autorais

Formatos disponíveis

Compartilhar este documento

Compartilhar ou incorporar documento

Opções de compartilhamento

Você considera este documento útil?

Este conteúdo é inapropriado?

Direitos autorais:

Formatos disponíveis

Applications of Potential Theory in Mechanics: Selection of New Results

Enviado por

Direitos autorais:

Formatos disponíveis

APPLICATIONS

It is not easy to find something new in mathematics. It is difficult to

The book is addressed to a wide audience ranging from engineers, involved

CHAPTER 1. DESCRIPTION OF THE NEW METHOD

CHAPTER 2. MIXED BOUNDARY VALUE PROBLEMS

CHAPTER 3. MIXED-MIXED BOUNDARY VALUE PROBLEMS

CHAPTER 4. APPLICATIONS IN FRACTURE MECHANICS

CHAPTER 5. APPLICATION TO CONTACT PROBLEMS

SUBJECT INDEX 447

A short survey of methods for solving mixed boundary value problems of

a sinhτ cosφ a sinhτ sinφ a sinσ

(1 + coshτ) cos(σ /2)

Here a is the punch radius, H is an elastic constant, ω is the normal

The solution is presented in the form

⌠ A ( t ) J ( t ρ)d t = 0, for ρ> a . (0.6)

By using the discontinuous Weber-Schafheitlin integrals, one can deduce that

The solution (0.5) can now be rewritten as

and the charge density σ over the disc is given by

f (ρ, z ) = ⌠ A ( s ) J 0(ρ s ) exp(− sz)

⌠ sinst J (ρs) exp(−sz) d s = sin-1 t , (0.13)

where the notation

was introduced. Formulae (0.11) and (0.12) can be combined to give

A general solution in terms of one harmonic function is given in Chapter 5

DESCRIPTION OF THE NEW METHOD

Some integral representations, which are of fundamental value to the

1.1 Integral representation for the reciprocal

The author has decided to start with a derivation of a new integral

expression (1.1.2) can be transformed into

After one cumbersome derivation is finished, we can always find a way to do it

η( x ) = [(ρ2 − x 2)(ρ20 − x 2)]1/2/ x , (1.1.6)

expression (1.1.4) may be rewritten as

The integral in (1.1.7) can be evaluated by using formula (3.241.4) from

ρ2 + ρ20 − 2ρρ0cos(φ−φ0) + z 2 = l 21 + l 22 − 2 l 1 l 2cos(φ−φ0). (1.1.10)

This leads to two equations

l 21 + l 22 = ρ2 + ρ20 + z 2, l 1 l 2 = ρρ0. (1.1.11)

The solution will take the form

Hereafter the following abbreviations will be used:

l 1( x ) ≡ l 1( x ,ρ, z ), l 2( x ) ≡ l 2( x ,ρ, z ), (1.1.14)

l 1 ≡ l 1( a ,ρ, z ), l 2 ≡ l 2( a ,ρ, z ). (1.1.15)

Note the limiting properties

lim l 1( x ) = min( x ,ρ), lim l 2( x ) = max( x ,ρ). (1.1.16)

In view of the properties above, the representation (1.1.4) can be generalized

Formula (1.1.17) simplifies when u =0

g ( x ) = x [1 + z 2/(ρ2 − x 2)]1/2. (1.1.25)

It is important to notice that the function g ( x ) is inverse to l 1 for x 2<ρ2, and is

(ρ20 − y 2)1/2[ l 22( y ) − y 2]1/2

A particular case of (1.1.18), when z =0, reads

The same result takes another form due to (1.1.22)

2. Prove the identity ( R is defined by the first line of 1.1.27)

3. Prove the identity

4. Prove the identities

( l 22 − ρ2)1/2( l 22 − a 2)1/2 = z l , ( a 2 − l 21)1/2(ρ2 − l 21)1/2 = z l ,

( a 2 − l 21)1/2( l 22 − a 2)1/2 = za , ( l 22 − ρ2)1/2(ρ2 − l 21)1/2 = z ρ.

5. Prove that g ( x ) is inverse to both l and l , namely, prove that g ( l )= a , and

6. Prove the identities

∂l1 a l 2 − ρl 1 ρ( a 2 − l 21) ∂l2 ρl 2 − a l 1 ρ( l 22 − a 2)

7. Evaluate the integral

(ρ20 − x 2)1/2[ l 22( x ) − x 2]1/2

8. Evaluate the integral

( x 2 − l 21( x ))1/2( x 2 − ρ20)1/2

9. Establish the integral representation

1.2 Properties of the L-operators