Filonenko-Borodich - Theory of Elasticty

M.
FILONENKO-BORODICH
TEXT FLY WITHIN
THE BOOK ONLY
<f)> 00 ^
OU 164095
PEACE
PUBLISHERS
M. M. 4>HJIOHEHKO BOPOHH4
ynpyrociH
POCyÂPCTBEHHOE H3#ATEJIbCTBO
H3MKO-MATEMATMMECKOft JIMTEPATypbl
MocKea
of
ELASTICITY
by
M. FILONENKO-BORODICH
Translated from the Rust
by
MARINA KONYAEVA
PEACE PUBLISHERS MOSCOW

First published 1963
Second printing
CONTENTS
Introduction 9
Chapter I. Theory of Stress 13
1. State of Stress in a Body 13

2. Differential Equations of Equilibrium . 16
3. Stresses on Areas 'Incli'ned to" the Co-ordinate Planes. Surface
Conditions ..:..:....; 22
4. Analysis of the State of Stress at a Given Point in a Body.
Principal Areas and Principal Stresses 25
5. Stfess Distribution Given Point. Cauchy's Stress Surface;
at a
Invariants of the Stress Tensor. Lame's Ellipsoid 29

6. Maximum Shearing Stresses 35
*7. Octahedral Areas and Octahedral Stresses 39
*8. Spherical Tensor and Stress Deviator 39
9. Generalisation of the Law of Reciprocity of Stresses. Examples 42
Chapter II. Geometrical Theory of Strain 45
10. Displacement Components and Strain Components, and Rela-

tion Between Them 1 45
11. Compatibility Equations 53
*12. tensor Character of the Strain at a Given Point in a Body. . 58
*13. Dilatational Strain. Invariants of the Strain Tensor 6i
*14. Strain Deviator and Its Invariants 65
*15. Finite Strain 67
Chapter III. Generalised Hopke's Law 72
16. General 72
17. Stfaihs' Expressed in Terms of Stresses 75
18. Stresses Expressed in Terms of Strains . .78
*19. Work Done by Elastic Forces in a Solid . 82
*20. Potential of Elastic Forces, . , 83
CONTENTS
*21. Stress-strain Relations; Hypothesis of the Natural State of

a Body 84
*22. Elastic Constants; Reduction in Their Number Due to the
Existence of the Potential of Elastic Forces . . . . '. 88
*23. Isotropic Body 89
Chapter IV. Solution of the Elasticity Problem in Terms of Displacements 95
24. Compendium of Basic Equations of the Theory of Elasticity .

96
25. Lame's Equations 99
26. Longitudinal and Transverse Vibrations in an Unbounded
Elastic Medium 102
27. General Solution of the Equation of Vibrations 106
28. Longitudinal Vibrations of a Bar. Fourier's Method 109
Chapter V. Solution of the Elasticity Problem in Terms of Stresses . . . 115
29. The Simplest Problems 115

30. Torsion of a Circular Bar 116
31. Saint-Venant's Principle 118
32. The Problem of Torsion of a Circular Bar (Continued) .... 122
33. Pure Bending of a Prismatical Bar 126
34. Prism Stretched by Its Own Weight 132
35. Uniqueness of Solution of Elasticity Equations 136
36. Beltrami-Michell Equations 139
*37. Three Kinds of Problems of the Theory of Elasticity.
Uniqueness Theorem 142
Chapter VI. Plane Problem in Cartesian Co-ordinates 147
38. Plane Strain 147

39. Generalised Plane Stress. Maurice Levy's Equation. Stress
Function 151
40. Solution of the Plane Problem by Means of Polynomials .... 162
41. Bending of a Cantilever 163
42. Beam on Two Supports 171
43. Triangular and Rectangular Retaining Walls (M. Levy's Solu-
tions) 177
44. Bending of a Rectangular Strip; Filon's and Ribiere's Solutions 181
*45. One Modification of Filon's Method : 189
*46. Strip of Infinite Length 196
Chapter VII. Plane Problem in Polar Co-ordinates 200
47. General Equations of the Plane Problem in Polar Co-ordinates 200

48. Problems in Which Stresses Are Independent of the Polar
Angle ,. 205
CONTENTS
49. Effect of a Concentrated Force (Flamant-Boussinesq Problem) 211

50. Wedge Loaded at the Vertex 217
*51. General Solution of the Plane Problem in Polar Co-ordinates 222
Chapter VIII. Torsion of Prismatical Bars and Bending 231
52. Torsion of Prismatical Bars 231

53. Saint-Venant's Method. Special Cases 238
54. Solution of the Torsion Problem in Terms of Stresses.
Prandtl's Analogy 250
55. Case of Transverse Bending 258
*
Chapter IX. More General Methods of Solving Elasticity Problems ... 265
56. General Solution of Differential Equations of Equilibrium in
Terms of Stresses. Stress Functions 265

57. Equations of Equilibrium in Cylindrical Co-ordinates. Their
General Solution 270
58. Harmonic and Biharmomc Functions 273
59. Biharmonic Equation 278
0. Reduction of Lame's and Beltrami's Equations to Bihar-
monic Equations 282
61. Boussinesq's Application of Harmonic Functions to
Method;
Seeking of Particular Solutions of Lame's Equations 284
62. Effect of a Load on a Medium Bounded by a Plane (Bous-
sinesq's Problem) 290
63. Effect of a Concentrated Force Normal to the Boundary and
Applied at the Origin 294
64. Solution of the Plane Problem of Elasticity by Means of
Functions of a Complex Variable 301
65. Filon's Method * 303
66. Wave Equations 310
67. Some Particular Solutions of the Wave Equation 313
Chapter X. Bending of a Piate 317
68. General 317

69. Basic Equations of Bending arid Torsion of a Plate 319
70. Analysis of the Results Obtained 323
71. Boundary Conditions for a Plate 328
72. Elliptic Plate Clamped at the Edge 331
73. Rectangular Plate. Navier's Solution 333
74. Rectangular Plate. Levy's Solution 339
75. Circular Plate 344
76. Membrane Analogy. Marcus's Method 347
CONTENTS
*
Chapter XI. Variational Methods of the Theory of Elasticity '. . 350
77. Variational Principles of the Theory of Elasticity. Fundamen-
tal Integral Identity 350
78. Lagrange's Variational Equation 352
79. Ritz-Timoshenko Method 358
80. Castigliano's Variational Equation 364
81. Application of
Castigliano's Variational Equation to the
Problem of Torsion of a Prismatical Rod 368
82. First Problem of the Theory of Elasticity; Second Theorem
of Minimum Energy 373
83. Approximate Method Based on Variational Equation (11.61) . .375
84. Lame's Problem for an Elastic Rectangular Prism 379
References 387
Name Index 388
Subject Index 390

Introduction
The theoryof elasticity developed into an important branch

of mathematical physics in the first half of the 19th century.
Before that the scientists of the 17th and 18th centuries Ga-
lileo Galilei, E. Mariotte, R. Hooke, J. Bernoulli, L. Euler,
C. A. Coulomb and others had worked out, in some detail, the
theory of the bending of thin elastic bars. At the beginning of
the 19th century J. L. Lagrange and Sophie Germain solved the
problem of the bending and vibrations of thin elastic plates.
Some characteristics of such thin elastic bodies made it possible
to simplify considerably the formulation and solution of prob-
lems concerning deformation under the action of external forces
without going too deep into the essence of the phenomena taking
place in a material. The beginning of the 19th century was
marked by vast progress in mathematical analysis, partly due to
a variety of new important problems arising in physics and
requiring the application of the complex mathematical apparatus
and its further development. This led to the emergence of a spe-
cial trend in physics, called mathematical physics. Among the
great number of problems confronting this new branch of science
it is necessary to mention the need for a profound
investiga-
tion of the properties of elastic materials and for the con-
struction of a mathematical theory which would permit study-
ing as completely as possible the internal forces occurring in
an elastic body under the action of external forces, as well as
the deformation of a body, i.e., the change of its shape. Inves-
tigations of this kind were also urgently needed to meet the
demands of rapid technological progress entailed by railway
building and engineering. These demands ensued from the neces-
sity of devising the theoretical methods for designing strong
structural members and machines. In 1825 the prominent French
engineer and scientist C.L.M.H. Navier published Resume des
tefons sur âpplication de la mecanique. P. I. Contenant les tegons
sur la resistance des materiaux (Paris, F. Didot), based on the
experimental data available at that time and the approximate
10 INTRODUCTION
theories indicated above. In Russia an analogous treatise by

N. F. Jastrzgbski appeared "in 1837.
The fundamentals of the theory of elasticity were worked out
by the French mathematicians and mechanicians A. L. Cauchy,
C.L.M.H. Navier, and S.D. Poisson, who derived the basic
differential equations of this theory. One must mention the great
difficulty encountered in its development. Mathematical analysis
in those days was built on the concept of continuous geometri-
cal space in which it was possible to consider infinitesimal seg-
ments and to introduce the processes of differentiation and inte-
gration on this basis. The universally recognised Newtonian molec-
ular theory of structure of bodies, on the other hand, represented
them as discrete media composed of individual particles that
are connected with each other by the forces of mutual attraction
and repulsion. It proved to be very difficult to justify the appli-
cability to such media of the apparatus of mathematical analy-
sis, which was essentially connected with the concept of contin-
uous functions capable of receiving indefinitely small (i.e.,
infinitesimal) increments, and with the possibility of passing to
the limit in their summation, i.e., in their integration. On account
of this, the first works on the mathematical theory of elasticity
gave rise to much discussion; their validity was questioned. How-
ever, the fact that even an extremely small volume, presumedly
isolated from a body, contains a great number of molecules
prompted investigators to appeal to the law of large numbers and
to apply the method which was subsequently called statistical;
this made it possible to bridge the gap between the continuous
space of mathematical analysis and the solid body as a discrete
medium. It became possible to apply the powerful apparatus of
mathematics to the development of the new branch of physics.
The importance of the application of the theory of elasticity
in physics and technical mechanics and the tremendous complex-
ity involved in the stated problems from the viewpoint of mathe-
matical analysis attracted the attention of the greatest investiga-
tors of the 19th and 20th centuries to this new branch of science.
Besides the founders of the theory of elasticity named above,
Cauchy, Navier and Poisson, we might mention such outstanding
scientists as M. V. Ostrogradsky, G. Lame (who published the
first course of lectures on the theory of elasticity in 1852), B.P.E.
Clapeyron, B. de Saint-Venant, G. Green, J. C. Maxwell, Lord

Kelvin (Sir W. Thomson), Lord Rayleigh, J. H. Michell, E. Ma^
thieu, F. S. Jasinsky, S. P. Timoshenko, G. V. Kolosoff, N. I. Mus-
khelishviliand many others.
In treating statics in theoretical mechanics, it is stated that
the conditions of equilibrium of a body or a system of bodies do
INTRODUCTION
not involve internal forces since they are pairwise mutually equi-
librated on the basis of Newton's third law of equality of action
and counteraction. The theory of elasticity sets forth the problem
of determining internal forces in a solid body. These forces re-
present interaction between molecules; they ensure the existence of
a solid body as such, its strength. They also act when no external
forces are applied to the body; these forces are not in themselves
the object of study in the theory of elasticity; under the action of
external forces the body deforms, the mutual position of molecules
changes and so do the distances between
them; the action of the external forces that
produce deformation gives rise to addi-
tional internal forces. Statics deals with
external forces applied to a body; it is
necessary therefore first to convert the
internal forces in question into external
Forces. This is achieved by the so-called
method of sections which consists in the
following (Fig. 1).
Let us imagine a body cut into two
l
Darts by a plane this plane intersects
\
the lines of forces of interaction between

molecules situated on the two sides of it.
If we now imagine that one of the parts of the body has been re-
moved, for example B, the system of forces of interaction applied

to molecules of the part A, located near the plane of section, will
be unbalanced. The remaining part /4, however, is in equilibrium;
hence, the system of internal forces brought about by the method
of sections and converted now into forces external relative to the
part A must be balanced by the system of the remaining external
forces PI, /*2, acting on this part of the body.
We now isolate in the plane of section an elementary area A/7
that is very small compared with the dimensions of the section,
but very great compared with the distances between the individual
molecules of the body; this area is intersected by a large number
3f lines of action of the discovered internal forces applied to mo-
lecules of the part A and exerted by the removed part B. Let AP
denote their resultant force vector; the ratio
AP
A/7
s called the average stress of internal forces in the body across

the area AF. In order to apply subsequently the apparatus of
1
A curved surface is often taken instead of a plane.
12 INTRODUCTION
mathematical analysis, we shall contract the contour of the ele-

ment AF around any of its points, M\ then its area A/7 and the
resultant force vector will diminish indefinitely; the limit of the
ratio
lim -r- =P
is called stress of internal forces or simply the stress of the point
M of the body on the area lying in the plane of the section made.
The notion of stress in the theory of elasticity is a fundamental
measure of the intensity of internal forces brought about by the
method of sections. The stress is characterised by its magnitude
2
having the dimension force/length 1, direction and position, i.e.,
|
it has a vector character; it is represented, therefore, as a vector p
applied at the point M

to which it corresponds. The vector p is
termed the total stress in distinction of the components into which
it can be resolved. The basic properties of stresses will be dealt
with later, in Chapter I.
The passage, made above, from the average stress across the
area to the stress at the point is connected with the imaginary
7
process of reducing the dimensions of the area A/ to zero, which
is necessary for the application of the analysis of infinitesimals.
The validity and justification of such a formal process, as already
stated above, had long been questioned and discussed by scientists;
however, the application of the resulting basic equations of the
theory of elasticity to the solution of problems of physics demon-
strated fairly soon the efficiency of the methods worked out and
brought a series of remarkable results confirmed by experience;
this pertains first and foremost to the study of vibrations and
propagation of waves (e.g., sound waves) in elastic bodies; some
of the simpler problems of this kind are treated in Chapters IV
and IX. The middle of the 19th century was especially rich in
achievements as regards the development of the theory of elasticity
and the solution of problems of importance for physics and
technical mechanics; the work of the eminent French investigator
Saint-Venant and his pupils played a dominant part here. Under
these circumstances, doubts concerning the physical justification
of the method of elasticity dealing, as it were, with a continuous
solid medium gradually disappeared. From this point of view, it
is sometimes said that the theory of elasticity is based on the
hypothesis of the continuous structure of solids. It must be borne

in mind, of course, that this hypothesis is but a working hypothesis;
it is dictated by the adopted mathematical method of
investigation
and does not intrude into the branches of physics that are directly
concerned with the problems of body structure.
I
Theory of Stress
1. STATE OF STRESS IN A BODY
Beginning the study of the state of stress in a body, let us

firstintroduce a system of notation of stresses which would permit
us to distinguish easily between stresses on different areas passing
through a given point and be as mnemonic as possible.
There exist several systems of notation; here we shall use the
system employed in many courses on the theory of elasticity.
Fig. 3
Imagine (Fig. 2) a solid cut by a plane into two parts A and B.

The part B is removed thereafter; we isolate an element of area dF
in the plane of the section; the orientation of this area will be
characterised by the outward normal to it (outward relative to the
partA which has remained).
With the aid of the outward normal we indicate tersely and
clearly not only the orientation of the area but also the part of
the body (B or A) which is removed after the section has been
made and the action of which is subsequently replaced by forces.
Let p be the total stress at a given point on the area dF*
Consider any axis x (i.e., a straight line with a specified direc*
14 THEORY OF STRESS [Ch T
tion on it); we shall denote the projection of the stress p on

this axis by X v and read this notation thus: "The projection on
the x axis of the stress p on an area with outward normal v".
Thus, the main letter X indicates the direction of the axis on
which the stress is being projected; the subscript v characterises
the orientation of the area on which the stress is being con-
sidered.
We somewhere in space an arbitrary system of rectangular
select
co-ordinates Oxyz with the aid of which we shall determine the
position of the points of the body under consideration. Stresses
will be determined by their projections on the axes of this system;
we shall go on to use it in the analysis of strain.
The projections on the coordinate axes of the total stress p
acting on the area dF will be denoted by X V9 Y v Z v , . If we remove
Fig. 4
the part A after the section has been made, the outward normal to
the same area dF will be directed opposite; we denote it, therefore,
by v. The projections of the total stress on it (Fig. 3) have to
be denoted as: X. Vl Y~ v Z- v it is obvious that
, \
y -v
s+ YV
'* Y -v Y V* 7
** v
7
z"z/
since stresses X v Y v Z v express the action of the part B on the

, ,
part A, while stresses X~ v Y- v Z- v express the action of the part A

, ,
on the part B\ these actions are equal in magnitude, but opposite

in sense.
Subsequently we shall frequently make sections perpendicular
to one of the co-ordinate axes. Let us make, for instance, a section
perpendicular to the axis Oy\ we then obtain the following projec-
tions of the total stress (Fig. 4):
Vy y*
The projection Y y is normal to the area and is called the normal

stress-, the projections X v and Z v , lying in the plane of the area,
are called the shearing stresses.
Sec. 1] STATE OF STRESS IN A BODY 15
Fig. 5a, b shows the notation of stresses on the areas normal to

the other two axes. Bringing these results together, we obtain the
Fig. 5
following system of stresses on the areas parallel to the co-

ordinate planes
Xx ,
Y x Zx
, \ X y,
Yr Zy \ X
z, Y z Zz
,
.
One can easily see that the stresses
AY*> YV yi ^z
7
in whose notation both letters are the same, are normal stresses.
The remaining six stresses are
shearing stresses.
In order to gain a better orienta-
tion in these new notations the reader
is recommended to do the following
exercises.
Exercises.
1. Fig. 6 represents the cross sec-

tion yOA of the wall of a dam which
is subjected to water pressure against
the face Oy\ the face OA does not
carry any load.
Designate the normal and shearing
stresseson the face Oy and write
down what they are equal to (pay Fig. 6
attention to the direction of pressure,
the direction of the axis Ox and the sign of the projection). Designate also the
components of stress on the oblique face OA (the outward normal v) and write
down therefrom the conditions that the face OA is free of load. Designate the
stresses at the sections aa and bb when any part of the wall (left or right,
upper or lower) is removed after the cut has been made.
16 THEORY OF STRESS [Ch.
<
?. A cylindrical body(Fig. 7) is twisted by forces applied to the end cross
sections. Designate the stresses at any point k of the cross section In and
write down the condition that the normal stress is absent. Designate the stress
at any point m
of the lateral surface and write
down the conditions trjat this surface is not
loaded.
3. A
cylindrical rod (Fig. 8) is bent by
applied to the end cross section pq.
forces
n Write down the conditions that the lateral sur-
face is free of load (see the preceding exercise),
designate those stresses at the section mn (nor-
mal and shearing) which are determined in
strength of materials.
Some authors now use the following

notations which are also employed in
technical literature: the normal stresses
ig. 7 Fig. 8
on the areas normal to the axes OX, OY and OZ are denoted,

respectively, by
the shearing stresses on the same areas are denoted by
so that
= T^y > *
2
==T y2 ^Z == ^XZ"
2. pIFFERENTIAL EQUATIONS OF EQUILIBRIUM
Imagine that we isolate in a solid an infinitesimal element in

the form of a parallelepiped with three pairs of faces parallel to
the co-ordinate planes. Let the edges of this parallelepiped be
$x\ dy, dz (Fig. 9). Its volume is dr=dxdydz. The action of
the removed parts of the body upon the isolated element is
replaced by forces. The stress on each face is resolved into three
Sec. 2) DIFFERENTIAL EQUATIONS OF EQUILIBRIUM 17
components. Altogether there will be 6 X 3

= 18 stress compo-
nents acting on all faces. These will be the external forces acting
on our parallelepiped. In addition we shall assume that there exist
the so called body forces in the given body. These forces will
be considered to be applied to the mass of the body; such is, for
instance, the gravity force which will be dealt with later.
Let some body forces, which we also resolve into three com-
ponents, X, y, Z, be applied to unit mass of the body. It follows
that these quantities have the dimension of acceleration. The body
forces acting on the mass of the parallelepiped pdr (p is the
density of the body at a given point representing the mass per
Fig. 9
unit of volume and having the dimension 2 4

kg sec /on ) will
then be
= A'p dx dy dz\
i = Kp dx dy dz\
Zp rf-c = Zp dx dy dz.
The stresses set up in a solid by external forces vary from
point to point, and they are therefore, generally speaking, func-
tions of the co-ordinates of points
z)\
(1-1)
z).
If .the area abed of the element (Fig. 9) is acted upon

by the
stress X^x Xx the area of We'd' will be acted upon by the
,
2-1013
18 THEORY OF STRESS [Ch f
stress Xx + ~jjj-dx since, in passing to the area a'b'c'd', we

change only one of the co-ordinates, namely x, in the first equa-
lity of (1.1). Thus, we can easily designate the stresses on all
the areas bounding the parallelepiped.
Let us assume that the body under consideration is in

equilib-
rium; the six equations of equilibrium of statics must then be
satisfied for each particular element
= 0, (1.2)
To begin with, examine the first group of these equations and,

in particular, consider the first
equation
E*=-0. (1.3)
Here we must take into account only those forces which

give
projections on the axis Ox\ these will be the normal stresses on
the lateral faces abed and a'b'c'd'
(Fig. 10) and the shearing
stresses parallel to the axis Ox,
acting on the other faces.
Sec. 2] DIFFERENTIAL EQUATIONS OF EQUILIBRIUM 19
Writing out the equation of equilibrium (1.3) we have
(1.4)
If we consider the case of motion of particles of an elastic

body (elastic vibrations), then %X does not vanish. According
to Newton's second law, it is equal to the product of the mass
of the element and the projection of its acceleration on the axis
Ox\ if the projections of the path (displacements) traversed by a
given particle on the co-ordinate axes are denoted by u, v and w,
dû
the acceleration along the axis Ox is on the right-hand side
-372*1'
of equation (1.4) there appears, therefore, the expression

, d2u , , , d2u
as indicated in parentheses.
After cancelling out like terms in (1.4) and dividing the re-
sulting equation by dx dy dz = dr we obtain it in the final form
dXx dXy dXz
In the same way we write out the other two equations
In the case of motion, the product of the mass of the element

and acceleration along the respective axes should be substituted
for zero on the right-hand side
dY* ,
dyy dY* , , ^_n
dZ x dZy dZz
dx *
dy
'
dz
We now proceed to write out the last three equations of

equilibrium (1.2). Take, for instance, the equation 2M* = 0.
Accordingly, we retain in the drawing only the forces which can
produce moments about the axis Ox, i.e., the ones that are normal
to it. To simplify calculation we place the origin of co-ordinates
at one of the vertices of the parallelepiped (Fig. 11). We note
2*
20 THEORY OF STRESS [Ch.
that the moments of some of the forces shown in the drawing

will be small quantities of the third order, others of the fourth
order. For instance, for the normal forces over the left-hand and
right-hand faces we have the moment
This moment thus turns out to be a small quantity of the fourth

order; of the same order will be the moments of the body forces
+ Kp dx dy dz -~ ;
Zp dx dy dz '- .
Examining Fig. 11, we find only two forces producing the mo-
ment of the third order; they are indicated by thicker lines m
'S^l**
the drawing. Equating to zero the sum of the moments of these

forces about the axis Ox, we obtain
dropping here the small quantities of the fourth order, we get

Sec. 2] DIFFERENTIAL EQUATIONS OF EQUILIBRIUM
or cancelling out, we have
This is the law of reciprocity (pairing, conjugation) of shearing'

stresses. Accordingly, the other two equations give
Yx = Zx X z.
Thus, the equilibrium conditions of statics have led us to three

differential equations
dx dy
~dT
-
d H-pJ" =o = P ^ (1.5)-
dZ r
dy
which contain nine functions of the co-ordinates of the point in

question: X x X y X z Yx Y y> Y z Z x Z y and Z 2 however, the law
, , \ , \ , ;
of conjugation of shearing stresses

Y z =Z
'
^y>
(1.6)
Xy
shows that three of these nine functions are pairwise equal to
three others. Consequently, we have three differential equations
(1.5) containing six unknown functions
y, z),
(1.7).
z).
In the case of elastic vibrations the right-hand members of these

equations depend on the time t as well.
Since the -number of unknowns (1.7) exceeds the number of
equations (1.5), we conclude that the problem of the theory of
elasticity is statically indeterminate. The lacking equations can
be obtained only by studying the conditions of deformation .and
by taking into consideration the physical properties of a given-
elastic body. In fact, up to this point we have not been con-
22 THEORY OF STRESS [Ch. I
cerned with these properties when isolating in a body an element

for which equations (1.5) have been derived; the state of stress
in a body, however, depends undoubtedly on these properties.
The method by which we have derived equations (1.5) is often
employed in the elementary theory of bending of bars, for instance,
to obtain the relations between bending moment, shearing force
and load:
dx dx
or to calculate shearing stresses in bending !

.
For that purpose we isolate in a bar and balance an infinitely

thin layer between two cross sections, i.e., an element which is
infinitesimal in one direction; here, however, we obtain the condi-
tions of equilibrium of an element which is infinitesimal in all
three dimensions.
3. STRESSES ON AREAS INCLINED TO THE
CO-ORDINATE PLANES. SURFACE CONDITIONS
The equations of equilibrium (1.5) correlate stresses over areas

parallel to the co-ordinate planes.
A
^
Fig. 12 Fig. 13
It is sometimes necessary, however, to know stresses over areas

oriented in a different manner; on the other hand, if the whole
elastic body is divided into elements by planes parallel to the
co-ordinate planes, it will generally be impossible to isolate ele-
ments in the form of parallelepipeds at the surface of the body
(Fig. 12). The inclined faces thus obtained at the surface will
These relations are derived in all courses on strength of materials.

Sec. 3] STRESSES ON AREAS INCLINED TO CO-ORDINATE PLANES 2S
obviously be acted upon by the external forces (loads) applied to

the given body. The foregoing considerations also induce us to
derive relations between stresses over three areas parallel to the
co-ordinate planes and stresses over an area inclined arbitrarily
to these planes.
To obtain such relations, consider (Fig. 13) an infinitesimal
element isolated in a given solid by three planes parallel to the
co-ordinate planes and by the fourth plane intersecting all the
three axes, i.e., consider an element Oabc in the form of a
tetrahedron, or a triangular pyramid. Let the area of its face
abc be denoted by dS and the outward normal to it by v\ then
the areas of the other faces will be found as the projections of
the area of the face abc on the co-ordinate planes
area abc dS\

area Oab dScos(vz);
area Obc = dScos(vx);
area Oac = dScos(vy) 1
.
Let X v Y v Z v be the projections on

, ,
the co-ordinate axes of
the total stress P v acting on the area abc. The first condition
of equilibrium of the isolated tetrahedron I>X = gives then
Cancelling out dS, we have
Xv = ^cos (vx) + X y
cos (vy) -f- J^ cos (vz).
Similarly, the other two conditions of equilibrium give
Y v = Yx cos (vx) + Yy cos (vy) + Yz cos (us);

Z,,
= Z x cos (UK) + Z y
cos (vy) +Z 2 cos (vz).
Let us introduce an abbreviated notation of the direction cosines

of the outward normal v to the area
vz)
= n. (1.9)
Here x, y, z in parentheses denote the directions parallel to the co-ordi-

1
nate axes Ox, Oy, Oz, respectively.

Then equations (1.8) will be written down in the form
Xv = Xx l -f-
Xytn -f- X n\
z
which will be frequently used in what follows.

The body forces are not involved in these equations since they
are small quantities of the third order, for instance:
>pXdt = pX *
I
*
,
while all terms of the
equation of (1.8)
first
in its initial form are small quantities of the second order.

We assume that the area of the face abc is made to approach
zero as a limit; equations (1.8) give then the relation between
the stresses at a point O on an oblique area with outward nor-
mal v and on three areas parallel to the co-ordinate planes. If
we cut out a tetrahedron Oabc at the surface with the face abc
belonging to the surface, then AV, K v Z v are the stress components
,
due to the external forces (Loads on the given body) applied on

the surface. Equations (1.8) give then the relation between the
external load and internal forces. In this case they are called
the conditions on the surface of a body and appear to be very
closely connected with the differential equations of equilibrium
(1.5); in fact, if functions (1.7) are such that equations (1.5) and
surface conditions (1.8) are satisfied at all points of the body
and on its surface, this guarantees the equilibrium of all elements
(parallelepipeds and tetrahedrons) into which the given body has
been divided; hence, the equilibrium of the body as a whole
will be secured. The mathematical meaning of this conclusion
is that equations (1.5) and boundary conditions (1.8) must be
considered simultaneously, for equations (1.5) cannot have any
definite meaning unless conditions (1.8) involving the external
load on a body are given.
Thus, the body as a whole will be in equilibrium if the equations
of equilibrium (1.5) are satisfied and the surface loads are speci-
fied; in other words, we have proved the sufficiency of equations
(1.5) and boundary conditions (1.8)* for the equilibrium of a body
under the action of the given external forces. One can prove the
necessity of equations (1.5). Indeed, if a body is in equilibrium, six
conditions of equilibrium (1.2) are satisfied for any part isolated
in it by a closed surface S. The first of them is written out as
J
Xv dS + f XP d* = 0.
(S) (t)
Sec. 4] ANALYSIS OF STATE OF STRESS AT GIVEN POINT IN A BODY 25-
Here the first integral is extended over the surface of the body,
and the second integral over its volume. However, the first integral
can also be transformed into a volume integral by the use of the
Green-Ostrogradsky formula
l
Xv dS = cos (vx) X cos (vy) +X cos (vz)] dS =

f J \XX -f- y 2
(S) (S)
'
dX r dXv d>
(0
On this basis the preceding equation of equilibrium may be

written down as
The integral on the left-hand side must vanish for any volume r
isolated in the body, but this is possible only if the integrand is
identically equal to zero at all points of the volume t:
This proves the necessity of the first of equations (1.5). In the

same way one can prove the necessity of the other two equations
of (1.5) and equations (1.6).
4. ANALYSIS OF THE STATE OF STRESS AT A GIVEN POINT IN A BODY.

PRINCIPAL AREAS AND PRINCIPAL STRESSES
*x
Let us suppose that an element Oabc has been cut out within
an elastic body and that X v Y v Z v are the projections of the
, ,
total stress P v over the oblique area abc on the axes of a random
co-ordinate system Oxyz (Fig. 13). Knowing the projections of
the total stress (X v Y v Z v ), we readily find the normal and
, ,
shearing stresses on the same area abc.
1
The Green-Ostrogradsky formula has the form
dQ . d
[P cos (vx) Q cos (vy) ^ cos
f (4r -j- -f-
(S)
where P, Q, R are functions of x, y, z, being continuous on the closed sur-

face S; their partial derivatives must exist and must be continuous throughout
the volume T bounded by the surface S; v is the outward normal to the sur-
face 5 at a certain point (x y, z). t
THEORY OF STRESS [Ch. I
For this purpose we take the normal v to the area to be one of

new co-ordinate axes. The other two axes u and w will be chosen
in the plane abc (Fig. 14). We denote the cosines of the angles
Fig. 14
between the old and new axes according to the scheme
Now equations (1.8), similarly to (1.8a), may be rewritten more

compactly
Xv = -f-
(1.11)
Obviously (Fig. 14), the normal stress on the area abc will be
obtained as the sum of the projections of stresses X v Yv Z v on
, ,
the normal v.
Designating the normal stress, according to the system accepted

above, by V V9 we get
Vv =
(1.12)
Sec. 4] ANALYSIS OF STATE OF STRESS AT GIVEN POINT IN A BODY 27
Similarly we obtain the shearing stresses projecting Xv ,

Yv ,
on the axes u and w, respectively:

Uv = i
2 +Z n n
z } 2 -+
+ Yz (
X
(1.13)
(/yi 3
/l
2 + ^2^3) + ^ (^3^2 + ^2^) +
Formulas (1.12) and (1.13) are considerably simplified in the case
of the so-called state of plane stress when all forces are parallel
to a single plane and distributed uniformly in a direction perpen-
dicular to this plane.
Fig. 15
In fact, are located in the plane Oxy, one can

if all forces
isolate a triangular prism Oab in place of the tetrahedron Oabc
(Fig. 15). In formulas (1.12) and (1.13) X 2 Yz Z 2 disappear; , ,
these formulas then become
U = (1.14)
The angles (ux), (uy) (vx) and (vy) are indicated

y
in scheme
(a), and the respective cosines in scheme (b)
(a) (b)
;28 THEORY OF STRESS [Ch. I
On the basis of scheme (b) equations (1.14) take the form
Vv = Xx sin a + Y
2
y
cos 2 a + 2X y
sin a cos a, }
2
sin a). J
They are known in this form from strength of materials; equations

(1.15) can, of course, be derived directly from the consideration
of Fig. 15.
Formulas (1.12) and (1.13) state the important proposition
that the stresses over three mutually perpendicular elementary
if
areas at a point O are known
X, y,
Y
1
X">
Y V Y Z>
' *
(1.16)
we can calculate all three components of stress
V ifv* W
v v> u wv
over any fourth area (at the point 0) determined by its outward
normal v\ in view of .this the quantities (1.16) are sometimes
called elements of the state of stress in a body. One can proceed
further and construct two more areas with outward normals u
and w at the point O\ the stresses over these areas will be
expressed by formulas of the same form as (1.12) and (1.13),
for instance
1
/
1
+ 2X y
l
1
m l
.
After this has been done, we obtain nine formulas [three of the
type (1.12) and six of the type (1.13)] expressing the stress
components over the areas of the trihedral Ouvw
UU V
-
tt
W
U
Uv Vv v W
u wv (1.17)
Uw Vv w W
^ ww
in terms of stresses (1.16) over the areas of the former trihedral
Oxyz.
Nine components (1.16) or (1.17) define the stress tensor and
from this point of view they are called components of the stress
tensor. The formulas of the type (1.12) and (1.13) (as we have
said, there are nine of them) determine the transformation of
the tensor from one co-ordinate system to another. The stress
tensor (1.16) is symmetric, since the components symmetric with
Sec. 5] STRESS DISTRIBUTION AT GIVEN POINT 29
respect to the principal diagonal (X x Y yj Z 2 ) are equal to each ,
other by virtue of (1.16); this property is obviously maintained

for other systems of co-ordinates as well
l
.
5. STRESS DISTRIBUTION AT A GIVEN POINT.
CAUCHY'S STRESS SURFACE; INVARIANTS OF THE STRESS TENSOR.

LAME'S ELLIPSOID
Let us now turn to the very important question of how stresses

are distributed over all the possible areas passing through a given
point in a body; for this purpose we shall apply a very'^legant
and visual method devised by the founder of the theory of elastic-
ity, A. L. Cauchy. We place the origin of co-ordinates/at some
chosen point M
in a body, pass an elementary area with outward
normal v, and along this normal construct a vector of length p
which as yet is left undefined. The co-ordinates of the end of this
vector are
C = /w, (1.18)
where /, m, n are the cosines of the angles that the normal makes
with the co-ordinate axes. The normal stress on the area in ques-
tion according to (1.12) is
Vv = X x l 2
+ y m + Z,n + 2y mn + 2Z nl + 2X lm.
y
2 2
2 x y (1.19)
We determine /, m, n from the preceding equalities and substitute

2
them in (1.19); multiplied by p , the equation reduces to
Xx ? + Y tf + Zp + 2Y rt + 2Zx
y z tt H- 2X y ty = p Vv
2
.
(1 .20)
We now adjust the length of the vector p so that the right-hand

member of this equation will always remain constant
the +
sign should be taken on the right-hand side if Vv > 0, i;e M
if the normal stress on the area is tensile, and the is
sign if it
compressive. Then
or ^=' (1-21)
Introducing the notation

,
(1.22)
For further details in regard to this tensor see N. I. Muskhelishvili,

1
Some Basic Problems of the Mathematical Theory of Elasticity, Groningen

P. Noordhoff, 1953.
30 THEORY OF STRESS [Ch I
we write down equation (1.20) as
O(5, TJ, )= c2 .
(1.23)
It represents a surface of the second order which is called the

stress surface or Cauchy's quadric; this surface has the centre at
the origin and it may, therefore, be (a) an ellipsoid, (b) a hyper-
!
boloid of one sheet, or (c) a hyperbo-

loid of two sheets; the ends of the vec-
tors p lie on this surface.
Let us suppose, for the sake of defi-
niteness, that the quadric is a hyperbo-
loid of one sheet and that it has been
constructed; to simplify the drawing,
Q Fig. 16 shows its section by one of the
principal planes on the assumption
that the outward normal to the Mv
area lies in this plane. Taking the
area and drawing the outward normal
Mv
to it we find the length of the MMi
vector p; knowing it, we determine the
normal stress V V = MN by the second of
formulas (1.21).
now show that we can easily
Let us
Fig. 16 find the total stress P on the given
area as well. In fact, we take partial
derivatives of function (1.22) and make use of (1.18) and (1.8a).
We obtain
= (X x + X m + A =
2/? l
y
= 2p (Yx + Y m + Y n) =
l
y z
= 2(Z, ,C)
= (Z
2/? ,n)
= 2pZ v .
It is known that these partial derivatives are proportional to

the cosines of the angles that the normal MiQ to the surface
makes with the co-ordinate axes; on this basis the last equalities
show that the cosines of the direction angles of the normal to the
surface are proportional to the projections X v Y v Z v of the total , ,
stress over the area concerned on the co-ordinate axes. Hence the
conclusion: the total stress MP
on the area is perpendicular to the
tangent plane 5Sj to the surface; knowing its direction, we obtain
1
Since there are no terms of the first degree in |, r).
in equation (1.23),
Sec. 5] STRESS DISTRIBUTION AT GIVEN POINT 31
its magnitude P v = MP
by drawing NP _L MN. Now, of course,
we can easily find the shearing stress MT. Thus, Cauchy's quadric
enables the stress distribution at the given point of the body to M
be fully investigated.
It is known from analytic geometry that by rotating the co-ordi-
nate axes the equation of the surface of the second order (1.23)
can be transformed so as to eliminate the terms containing prod-
ucts of co-ordinates in a new co-ordinate system (u, v, w). Then,
we shall obviously have
Hence, the shearing stresses will vanish on three mutually perpen-

dicular areas normal to the new axes (a, v, w). These three areas
are called principal] the normal stresses on them f/ u V v lv (they , ,
W
will also be the total stresses) are called principal stresses at the
given point M
in the body. We denote them more compactly
Then the equation of Cauchy's quadric referred to the new axes is
4- Nff =
2
c .
(1 .24)
The final form of this surface depends on the signs of the prin-
cipal stresses N\, 2 N N
3 and on the associated sign of the right-
,
hand member of the equation. Consider two fundamental cases:

(1) all three principal stresses are of the same sign; (2) the signs
of the principal stresses are different.
Case 1. Let us assume that N\ 0, 2 >
0, A/ 3 N >
0, i.e., all the >
principal stresses are tensile. Then the plus sign should apparently
be chosen in the right-hand member of (1.24), and it can be re-
written as
2 2
^2
\N 3
This is the equation of an ellipsoid with the semi-axes
One sees from formula (1.21) that in this case !/ =

v~
i.e., the normal stresses on all areas passing through the point in
question are positive and consequently tensile. If all three prin-

cipal stresses are compress! ve: A/, 0, 2 0, <
3 N <
0, the minus N <
2
sign should be taken before c in (1.24); this equation once again
32 THEORY OF STRESS [Ch. 1
gives an ellipsoid, but according to formula (1.21) we get
and the normal stresses on all areas are compressive.

Case 2. Let us assume that N > 0, N 2 > 0, Af 3 < 0, i.e., two of
{
the principal stresses are tensile and the third is compressive. In

this case equation (1.24) has to be written in two ways:
where Af 3 denotes the absolute value of the compressive prin-

I 1
cipal stress. Just as in the first case, we write these equations in

the form
2 2 2
i) C
e c c* (1.26)
( *\ ( *\ (
W7 UVi/ \Wl
The first of these equations gives a hyperboloid of one sheet, and

the second a hyperboloid of two sheets. We add an intermediate
case, setting c
2 = 0:
A^ 2
+ A^ 2
-|Af 3 |t
2 = 0. (1.28)
This equation gives the asymptotic cone separating the above two
hyperboloids and tending to approach them at infinity (Fig. 16).
If the end of the vector representing the normal stress on the
area appears to be on the hyperboloid of one sheet (1.26), this
stress is positive, i.e., tensile; if, on the other hand, it appears
to be on the hyperboloid of two sheets, the stress is compressive.
In the intermediate case it may be directed along the generator
of the asymptotic cone (1.28); in this case the length of the vec-
tor becomes infinite and according to (1.21) Vv = 0. Hence, on
areas normal to the generators of the asymptotic cone there act
only shearing stresses.
If a given area abc (Fig. 14) is a principal area, the total
stress P v acting on it is directed along the normal v and is a
principal stress: we denote it by /V; its projections on the co-
ordinate axes are
Sec 5] STRESS DISTRIBUTION AT GIVEN POINT 33
We substitute these values in equations (1.21) and write them

down as
(X JC N)l+X m + X n=0 y g 9
N) m + Y> =
]
Yx l 4- (Y y 0, (
1
.29)
Zx l + Z m + (Z N)n =
y z Q. t
Adding here the basic relation among the direction cosines of the
normal v
>2_|_^2 =li (1.30)
we have four equations for the determination of the principal

stress N
and the corresponding principal area, i.e., the direction
cosines of its normal, /, m, n. The procedure of solution will be
as follows: the system of homogeneous equations (1.29) does not
admit a trivial solution
because it is contrary to (1.30); however, for the existence of

other solutions of this system (in which at least one of the co-
sines /, m, n has a value different from zero) it is necessary that
its determinant should be equal to zero. Thus, we come to the
condition
X
= 0. (1.31)
Writing out the determinant in the left-hand member, we obtain

the cubic equation
,V3__eA/ 2 + HjV-- A = 0, (1.32)
in which the coefficients have the following values:
YyY z
H=
' 1
xY y y
1 7
Y z^z
(1.33)
Y x 1Y y 1Y z
1
z x z y zz
All three roots of equation (1.32) are real they give the val-
1
;
ues of the three principal stresses N\, /V 2 N*. Substituting any ,
1
This is proved in analytic geometry in investigating the principal axes
of the central surface of the second order.
3-1013
of these values of N in equations (1.29) and using two of them

1
,
we find the quantities //, m'i, n't proportional to the direction co-
sines of the normal to the unknown principal area
Substituting this in (1.30), we find the factor K and, consequently,

the cosines themselves. Thus, the directions of the principal areas
will be determined.
Another point: the roots of equation (1.32) must not depend on
the system of co-ordinates x, y, z\ consequently, the coefficients
of this equation do not depend on the choice of a system of co-
ordinates either. Hence we conclude that formulas (1.33) give
three functions of the components of the stress tensor (1.16) which
are invariants under a transformation of co-ordinates. Of particular
significance is the first of them, the linear invariant
e = xx +r y +z,, (1.34)
as weshall see in Chapter III.

the principal areas at a given point have been found, then,
If
along with Cauchy's quadric, it is possible to indicate another

geometrical representation of stress distribution, the one proposed
by G. Lame.
Let us assume that the co-ordinate planes Oxy, Oyz and Ozx
coincide with the principal areas at a given point and, conse-
quently, on these areas: Y z
= Z y = 0; Zx = X z = 0; X y = Yx = 0;
X x = N^Y y = N Z =2t
'
z N,.
Then equations (1.8a) are simplified and take the form
(1.35)
Relationships (1.35) may be represented geometrically. To this

end, we lay off from the given point O (Fig. 17) a vector OP
equal to the total stress on the chosen area with outward normal
v\ the co-ordinates of the end of this vector are
(1.36)
1
The third equation will be a consequence of Jhe other two, according to
condition (1.31).
Sec. 6] MAXIMUM SHEARING STRESSES 35
As the inclination of the area is changed, the point P describes

a certain surface that appears to be an ellipsoid; indeed, on the
basis of (1.35) and (1.36) we have
x = A/!/; y = N m; 2 z = A/ 3 /t;
hence
=1; (L37)
this is the equation of an ellipsoid referred to the principal axes.

This ellipsoid is called the stress ellipsoid or Lame's ellipsoid.
One of its three semi-axes is the longest, another the shortest,
the third is half-way between them; consequently, the three
principal stresses are maximum, minimum and minimax,
respectively.
Fig. 17
If two of the principal stresses are equal (for instance, Ni=N 2 ),

Lame's ellipsoid is an ellipsoid of revolution and the st^te of
stress at a given point is symmetrical with respect to the third
principal axis Oz. If all the principal stresses are equal, 2 Ni=N =
= A/3, Lame's ellipsoid becomes a sphere and all the areas at the
given point are principal and the stresses on them are equal;
this will be the case, for instance, under all-round compression
or tension.
6. MAXIMUM SHEARING STRESSES

Let us now turn to the determination of the maximum shearing
stresses.To simplify the investigation, we take again the axes of
co-ordinates Oxyz in the direction of the normals to the princi-
pal areas (i.e., along the principal stresses).
Wechoose an arbitrary area with outward normal v determined
by the cosines /, m, n\ the total stress on it P v according to ,
3*
36 _ _
(1.35),
The normal
If we
has the projections
X v = N,l\
THEORY OF STRESS
stress on this area

Y v = Njn\
denote by T v the total shearing stress on the area con-

we have
Z v = N,n.
V w> according to (1.19), is
[Ch. I
cerned,
Pl=Vl+Tl
whence we find
+ Nln - (M/ + N^ + A^
2
)
2 2 2
)
.
(1 .38)
With rotating area the stress T v varies, being a function of

two variables, say / and m, since according to (1.30)
taking this into account, the preceding expression can be

written as
(1.39)
To obtain - Tv we equate to zero the partial derivatives of Tv

with respect to / and m\ it is more cortvenient, however, to set
up other conditions
dl ~v dl
*" * w '
dm dm
It may be well to note that, in a'ddition to the required solu-
tions, we shall then obtain the extraneous solution 7^ which =
will obviously give a principal area; this solution will easily be
found and dropped later. Writing out the preceding conditions
and using (1.39), we obtain two equations for the determination
of / and m
-N,)m = 0.
Consider the most general case when Afi, N% and Af 3 are all
different;then the preceding equations can be divided through
by the differences of the principal stresses entering in them to
Sec.
We
_
6]
represent these equations, after
have two equations

MAXIMUM SHEARING STRESSES
some manipulation, as
of -the third degree in / and m; accord-

_ 37
ingly, we shall obtain three solutions that are easy to find* The
first and the simplest, in which
/ = =
/tt 0, n= 1,
must be dropped since it gives a principal area lying in the

plane Oxy. It will therefore be necessary to consider three cases:
(1) /=0; m = 0; (2) / = 0; /rc=0; (3) /=0; m + Q.
The third case is, however, impossible since then, cancelling /
and m out of equations (1.40), respectively, and subtracting the
resulting equations one from the other, we immediately obtain
which is contrary to the initial assumption. In the first case

(/ =h 0; m==0) the second of equations (1.40) is satisfied, and the
first equation, after the cancelling out of /, reduces to
(N l
N 3 ) (1 2/ 2 ) = 0, (1 .41)
whence
/== m==0: n=
In the second case, from the second equation of (1.40) we obtain
similarly
/ = 0; m=~\ J/2
/i=4=-
J^2
If, at the outset, we had eliminated, say, m instead of n from

(1.38) and repeated the analysis, we would have obtained addi-
tionally one more solution
Each of these three solutions determines two areas passing through

one of the co-ordinate axes and inclined to the other two at
angles of 45 and 135.
Substituting the first solution in (1.38), we obtain the required
maximum (or minimum) value of the shearing stress
replacing the notation Tv by T2 we , find
T
1 2
+
the remaining two solutions give similarly
(1.42)
7*,= "'-";
_4
'a- .
We draw the final conclusion from the analysis carried out.

The three mutually perpendicular principal areas at a point M
Fig. 18
form a trihedral. If (Fig. 18) the axes MX, My, Mz

represent
the lines of their intersection, three pairs of areas on which the
maximum shearing stresses occur pass through these axes and
bisect the dihedral angles between the principal areas.
The maximum shearing stresses, according to (1.42), are equal
to half the difference between the principal stresses over two
areas intersecting along that of the axes #, y, z through which a
given area of max T passes.
If
N >N >W
l 2 3,
the greatest of stresses (1.42) will be
T
'2
+
~
If, for instance,
Sec. 8] SPHERICAL TENSOR AND STRESS DEVIATOR 39
Cauchy's quadric and Lame's ellipsoid will be surfaces of revolu-

tion about the axis Mz\ all the areas passing through the axis Mz
(there is an infinite number of them) will be principal. The areas
on which max T = T 2 are also infinite in number; they are tan-
gent to a circular cone with the axis Mz and the vertex angle
of 90.
*7. OCTAHEDRAL AREAS AND OCTAHLDRAL STRESSES 1
Formulas (1.19) and (1.38) make it possible to determine the

normal and shearing stresses over any area at a given point if
the principal areas and principal stresses are known. Keeping this
in mind, let us find stresses over areas equally inclined to the
principal areas. In eight octants of the co-ordinate system we can
thus construct eight areas which will form an octahedron; they
are called octahedral; the stresses on them are also called octa-
hedral.
The direction cosines of the normals to these areas are
Substituting these values in (1.19) and (1.38), we have

Vv = N =^(N l + N + N^ = ^(X + y + Z
2 Jf
y 9 ); (1.43)
the square of the shearing octahedral stress is
Tl = 1 (2/V? + 2N\ + <2N\ 2NiN 2 2N 2N3 2Nrffi) =

2
î) l- (1 -44)
*8. SPHERICAL TENSOR AND STRESS DEVIATOR

The octahedral normal stress may be interpreted as the mean
normal stress at a given point in a body; using it, we construct
the state of stress at a given point represented by the following
stress tensor
(1.45)
1
Here and subsequently an asterisk indicates the chapters, sections and
parts thereof which the reader may easily omit in his first acquaintance with
ihe subject-matter without impairing continuity.
in this case the three principal stresses are the same, as are there-
fore the stresses over all the areas at the given point; Cauchy's
stress surface and Lame's stress ellipsoid appear to be spheres and
the tensor (1.45) is thus called spherical. For the sake of conven-
ience we introduce the concept of addition (subtraction) of ten-
sors. The sum (difference) of two tensors is defined as a tensor
whose components are equal to the sums (differences) of the
respective components of these tensors.
Let us take an arbitrary stress tensor
(1.46)
We add to it and then subtract from it the spherical tensor (1.45)

where A/ has the value (1.43):
(1.47)
The stress tensor can thus be represented, in general, as the

sum of two tensors. The first of these tensors is called the stress
deviator; briefly we denote it by (D s ).
The representation of the stress tensor in the form (1.47) is of
cardinal importance in the investigation of the behaviour of
elastic and plastic bodies under loading; but even now it is easy
to perceive the importance of resolving the state of stress (1.45)
into two component states (1.47). The spherical tensor (1.45)
isolates from an arbitrary state of stress an all-round tension
(compression), uniform in all directions, involving only the change
of volume of the given element of a body but no change in its
shape; therefore, the stress deviator (D s ) represents such a state
of an element in which the shape of the element distorts without
change of volume,
Consider the invariants of the stress deviator. Its first
(linear) invariant is equal to zero, since an all-round tension-com-
pression is excluded; irtdeed, by using formulas (1.33), we obtain
(1.48)
Sec 8] SPHERICAL TENSOR AND STRESS DEVIATOR 41
The second (quadratic) invariant is
H = (Y y - N ) (Z,
- N + (Xx - N (Z - N +
) ) z )
+ (Xx - N ) (Y y
- N - Y Z - ZX X - Y X = Y Z + X XZ +
) z y Z x y y z Z
-f- XX Y y 3A/o Y Zx X =
X XZ Z -f- z y
- -I [K,Z, + XxZ + XxY - Xl -Yl- Zl - 3Yl - 3ZX -3X$\ =

t y
2
(1.49)
If we take the principal axes for co-ordinate axes, i.e., if we set
X x = Nt r y = Af 2 ; Zz = ^ X y = Y = Zx = Q<z
this expression is simplified and made to assume the form
II = --J- [(î - ^ + (^2 - â) + (^3 - 2)

2 2
Aî)
2
L (1 -50)
i.e.,except for a numerical factor, it coincides with the square

of the octahedral shearing stress (1.44).
The third (cubic) invariant is
-N* Yx Zx
A= x y
ry -N zy
X, Y2
For the sake of simplicity, we take here too the principal axes
for co-ordinate axes; then, writing out the determinant, we obtain
By direct calculation one can prove that this expression may be

represented in the alternative form
A = y [(N, - N^ + (N, - Af + (A _ A^ )3 f
3 o) 3j. (1 >52)
Then it is evident that
fA = ^
[(Ni
~ ô
represents the average cubic deviation of a given state of stress
(Ni, Af 2 MB) from the mean hydrostatic stress
,
N = 3- (^ + ^+ ^3). (1.53)
We revert to the quadratic invariant (1.50) and rewrite it as
H= 1 (W? + U\ + N\ N,N 2 NN2 3 Af 3Afi). (1 .54)
By analogy with (1.52) we now form the square of the average

quadratic deviation of a given state of stress (N it N N
2, 3) from
the mean hydrostatic stress (1.53):
Substituting here the value of N from (1.43), we find that

2/222 \
and hence, comparing with (1.54),
A* = -|H. d.55)
Now comparing (1.48), (1.52) and (1.55) we come to the conclu-

sion that the linear invariant of the stress deviator indicates the
absence of
compression-tension "on the average"; the quadratic
and cubic invariants characterise, respectively, the average quad-
ratic and cubic deviations of the state of stress i9 2 3 from N N N ,
the mean hydrostatic stress corresponding to the stress tensor at

a given point in a body l
.
9. GENERALISATION OF THE LAW OF RECIPROCITY

OF STRESSES. EXAMPLES
To conclude this chapter let us generalise the law of reci-
procity of shearing stresses (1.6).
Consider (Fig. 19) two areas, with normals x and v, which
pass through a point Af, and denote the total stresses over them
by Px and P v respectively. According to the notation here ac-
,
cepted
Furthermore,
*, = />, cos (/>,*); K jr
=P jr cos(P jr y);
1
The interpretation of the quadratic invariant as the average quadratic
deviation was first given by V. V. Novozhilov (1952) and S. D. Ponomarev
(1953).
Sec. 9] GENERALISATION OF LAW OF RECIPROCITY OF STRESSES 43
Introducing this in the first of equations (1.8) and taking into

account (1.6), we obtain
Pv cos (Pvx) = X V = PX [cos (ox) cos (Px x) +
+ cos (oy) cos (P + cos (vz) cos (P
x y) x z)\ Px cos (Px v),
but once again, according to our notation,
thus,
XV =VX \
(1.56)
hence, the projections of the total stresses P v and P x over tfoe

areas concerned, respectively, on the normals x and v to these
Fig. 19
areas are equal to each other (see Fig. 19). The law of reci-
procity (1.6) now follows as a special case if the areas are
mutually perpendicular.
Exercises to Sections 2-4.
1. In the wall of the dam represented in Fig. 6 there are induced the
following stresses:
x
y
y
__ AVz Z*7 r = Z7z = AU.
Substitute these expressions in equations (1.5) and check in what conditions
they will be satisfied. Find the load acting on the face Oy (i.e., for *=0);
by using conditions (1.8) find also the load on the face OA (for *=*/tanp).
2. In the circular bar of diameter 2r (Fig. 7), there occur the following
stresses:
0; :
GTAT.
Check whether they satisfy equations (1.5) and in what conditions. Prove
by the use of equations (1.8) that the lateral surface is free of load (i.e.,
/,-y._Z.-0).
Hint: for a point on the lateral surface
cos (vx) = ; cos (vy ) = -~ .

-j-
3. In the straight bar under pure bending (with the system of co-ordinate
axes as shown in Fig. 8) there occur the stresses
Xy = Yg - = 0.
Do the same as in the preceding exercise.
Fig. 20
4. In one of the cases of bendnig a rod of rectangular cross section the
following stresses are set up (Fig. 20):

12
h3
*y- r(
J
X =2 =t 2T =0
Check whether these stresses satisfy equations (1.5); find the loads acting
on the faces of the rod.
II
Geometrical Theory
of Strain
10. DISPLACEMENT COMPONENTS AND STRAI1
COMPONENTS, AND RELATION BETWEEN THEM

Let us take an elastic body and fix it in such a way as to
prevent its displacement as an absolutely rigid body. Then displace-
ments of each of its points will be caused only by deformations.
Consider any point M(x, y, z) in the body fixed as indicated
above (Fig. 21). The point M will be displaced into a new posi-
tign M' as a result of the deformation produced. We designate
the projections of the displacement MM' on the co-ordinate axes by
u, v, w\ (2.1)
since displacements vary from point to point, the projections of

displacements (2.1) are functions of position
y, *); (2.2)
y, z). .
Let us now pass from displacements to deformations. We isolate

in the elastic body an infinitesimal parallelepiped (Fig. 22) with
edges dx, dy, dz. During the deformation of the body it will dis-
place and deform: the length of its edges will change and the
initially right angles between the faces will distort.
To estimate the deformation of the elastic body at the given
point M it is necessary to examine the elongations (linear defor-
mations) of the edges dx, dy, dz of the isolated parallelepiped
and the distortions of the angles 1M2, IMS, 2M3 (shears or an-
gular deformations). For this purpose consider the projections of
the parallelepiped on the co-ordinate planes; obviously, the defor-
mation of the parallelepiped itself can be deduced from the defor-
mation of these three projections. In future we shall restrict our-
46 GEOMETRICAL THEORY OF STRAIN [Ch. II
selves to very small deformations, in which case the subsequent

treatment may be greatly simplified.
Take, for instance, the projection of the element M123 on the
plane Oxy. Before deformation, the lengths of the edges are
ig. 21 Fig. 22
(Fig. 23): AB = dx, AC^dy. After deformation, they occupy the

position A'B' and A'C'. We now fix attention on the projection of
c'
B'
-M u j?
Fig. 23
AB. If the displacement of the point A along the axis Ox is u,

the corresponding displacement of the point B is
where 6u is the increment; since the point B differs from the
point Aonly by the co-ordinate x, the small increment 6w in the

last formula is replaced, to infinitesimal quantities of the second
order, by the partial differential of the function u with respect
Sec. 10] DISPLACEMENT COMPONENTS AND STRAIN COMPONENTS 47
to the variable x. Likewise, if the displacement of the point A

along the axis Oy is v, the displacement of the point B along
the same axis is expressed as
The projection of the absolute elongation of the segment AB on

the axis Ox is
the unit elongation of this edge is
~
e * x __
6 (dx) _
~ du
dx dx'
By reasoning analogously, we obtain for the unit elongation

of the edge AC directed along the axis Oy
dv
in the same way, for the edge parallel to the axis Oz (i.e., the
edge M3 in Fig. 22) we have
dw
We have obtained the formulas of linear deformations (elonga-

tions) at the given point M of the body in the direction of the
three co-ordinate axes.
Let us now turn to the analysis of angular deformations. We
easily find the angle of rotation a yx of the edge = dx in the AB
plane Oxy:
dv , dv
~ B B"
'
A B"
'
~&x _ ~dx
'
dx + ^-dx
dx
1
1 + 4^-
]
dx
Since we have confined ourselves to the case of very small de-
formations, we may omit the

quantity-^-
=e xx in the denominator
of the last expression as negligibly small compared with unity;
we get
a* x *
/o <o
v^ /
3\
dx
Similarly, we obtain the angle of rotation of the edge AC=dy

In the plane Oxy:
^..
(2.4)
We now can easily find the shearing strain, i.e., the distortion
of the right angle BAG:
\
dv , du
Formula (2.5) gives the expression of the shear angle occurring

in the plane Oxy.
Similarly, we obtain the expressions of shearing strains in the
other two co-ordinate planes by a cyclic change between the let-
ters according to the scheme shown in Fig. 24.
Fig. 24
Collecting together the above results, we get six basic relations

characterising deformation:
(a) unit elongations (extensional strains or extensions)
du dv
e
dx yy
(b) shearing strains

dv du
'
dx *
dy
(2.6)
dw dv
dz
du dw
These equations were derived by A. L. Cauchy.

Sometimes the halves of the shear angles are introduced in
formulas (2.6):
p
c p _ p
yz 2 y 2 zx<>
then the last three of them can be written down as
"tydZ"
'
'
dy
dv
<T
2e
e y* *=?-.
dy "dT 1
(2.6a)
rT~ dU dw
'
dx
They are occasionally found to be more convenient in this form,

As regards the signs of the strains e xx e mh e zz e yz e zxt e xy we , , ,
should note the following. If the function u grows with increas-
ing JC, i.e., e xx ~Y~

= 0, we >
are obviously dealing with an
increase in the length dx (the displacement of the point B to the

right, according to Fig. 23, is greater than that of the point A).
Thus, if e xx >0, the result is elongation, and if e xx <0, the result
is contraction; the same rule holds true for e yy and e zz Further- .
more, if the function v grows with increasing x, then [formula

(2.3)] a
== in this case (Fig. 23) the segment AB rotates
y* "5J>0 ;
from the +x axis to the +y axis; similarly, for a^ y =-^-->Othe

segment AC rotates from the +y axis to the +x axis. Hence, it
follows, according to (2.5), that the shearing strain e xy >Q the t if
right angle between the 4-x and +y axes decreases. The same
rule applies to the shearing strains in the other two planes. Sum-
marising what has been said above, we obtain the following sign
rule for deformations:
(a) elongations along the co-ordinate axes correspond to posi-
tive linear deformations e xx e yy e zz and contractions to negative;
, , ,
(b) a decrease in the angles between the positive directions of

the axes corresponds to positive angular deformations (shear de-
formations) e yzt e zx e xy and an increase in the same angles to
, ,
negative.
Formulas (2.6) show that six functions e xx e yv e zzy e yx e zx e xy , , , , ,
which are called strain components or components of strain, are

linearly expressed in terms of nine partial derivatives of the
displacement components u, v, w:
du du du t
'
dx dy
'
dz
'
dv dv dv
dx
'
dy
'
dz
'
(2.7)
dw dw dw
' '
dx dy
'
dz
The components of the principal diagonal of this matrix are
elongations [see the first three relations of (2.6)]. The remaining
components represent the angles of rotation of the edges of the
elementary parallelepiped M123, shown in Fig. 22, about the x,
y, z axes; this can be seen, for instance, from formulas (2.3) and
(2.4); in the notation accepted therein matrix (2.7) is written as
e xx* a
y*'
a ^ ;
ajry ^yy
a
zy' (2.8)
a w e*z-
41013
These notations are found

to be convenient; the second subscript
of the letter a
indicates the elementary segment (dx, dy, dz) the
rotation of which is being considered; the first subscript denotes
the axis in the direction of which the rotation takes place. For
instance, 0^ = -^- designates the angle of rotation of the element
dy from the y axis to the z axis.

in the direction
Weshould note the important fact that it is impossible to set
up a system of equations inverse to equations (2.6), i.e., to
express nine components of matrix (2.7) in terms of six com-
ponents of strain (2.6); equations (2.6) are inadequate. This
Fig. 25
is due to the fact that our geometrical representation of defor-

mations at a given point is as yet incomplete; to complete it and
to obtain symmetry in calculations let us introduce three more
components; let the element M123 in Fig. 22 be a cube (dx = dy=*
= dz) and consider the angles of rotation of its diagonals about
the x, y, z axes in the case when elongations e xx = e yy = e 2Z = 0\
Fig. 25 a shows the projection ABCD of the cube under consid-
eration on the plane Axy.
Obviously, the angle of rotation of the diagonal of the cube
about the z axis is equal to the angle of rotation of the projec-
tion AD of this diagonal about the point A. First consider the
particular case of "simple shear" parallel to the x axis (Fig. 25 a);
for the angle of rotation p' z of the diagonal AD we get then
This can be proved if we construct a circle ADDi with centre

at B and assume, disregarding small quantities of the second or-
der, that the point D lies on the circumference; the angles DADi
{
and DBDi are then subtended by the common arc DD^ hence
there follows the foregoing approximate equality which implies

that
' l du
The minus sign is taken because in the case under consideration
and p;<0 1
.
-g->0,
In the case of Simple shear along the y axis (Fig. 25 6) we find
in a similar way
ft '/_ 1 dv
?* 2"^T'
In the general case of shear we have

o
P
= P+ K"
o'
P*
i
dv - du\
= T (-5F 37)
= ? (*y ~ I f 1 , v
**)
Reasoning in the same way with regard to the yz and zx planes,

we obtain analogous formulas for the angles of rotation of the
diagonal of the elementary cube about the x and y axes. It is
usual practice to introduce double angles of rotation; they are
denoted by 2co with appropriate subscripts. Thus, we get
dw dv
"3jT "dJ ;
du dw
-ted^'' (2.9)
dv du
-5r 57-
The quantities CD*, co y , co z are called components of rotation; to

complete the picture of deformations at a given point equations
(2.9) should be added to equations (2.6). Let us examine the
deformation in a simple but very important case when displace-
ments u, v, w are linear functions of the co-ordinates of a point,
namely
u = u + c u x+ c y + c^z;
Q l2 \
(2.10)
We assume, as is customary in kinematics, that counterclockwise rotation

1
about an axis is positive.
4*
52 GEOMETRICAL THEORY OF STRAIN [Ch II
The components of strain (2.6) and the components of rotation

(2.9) then be constants; such a deformation is defined as
will
homogeneous. During the deformation a given body is converted
into another body; each point (*, y, z) moves to another point:
It evident that in the case of homogeneous deformation

is
(2.10) each plane (or straight line) is transformed into a plane

(or a straight line): two parallel planes (or straight lines) are
likewise transformed into two parallel planes (or straight lines);
a rectangular parallelepiped is transformed, generally speaking,
into an oblique parallelepiped. Inasmuch as in deriving equations
(2.6) we have disregarded small quantities of higher order, the
deformation defined by these equations will be homogeneous,
generally speaking, only in a very small region isolated in the
body; hence, the elementary parallelepiped M123 (Fig. 22) will
be converted into an oblique parallelepiped in the general case
as well; its opposite faces will remain plane and mutually parallel.
Consider also equations (2.10) in the case when all strain com-
ponents (2.6) are equal to zero' and the body does not deform;
we easily find then
Introducing more compact notations
CM=P\ c^ =q
we write down (2.10) as
ry\
i
pz; 1
(2.11)
These formulas coincide with the formulas of kinematics for

infinitesimal displacements of an absolutely rigid body; u v WQ , ,
are the translatory displacements of the body; p, q, r are the

angles of rotation about the co-ordinate axes. Furthermore, we
have
similarly (2.1 la)

Sec. 11] COMPATIBILITY EQUATIONS 5$
hence, the components CD*, co^, co 2 introduced here represent the

angles of rotation of the given element about the axes.
In the theory of elasticity it is the custom to say that for-
mulas (2.11) express a "rigid
body displacement".
Nowconsider the inverse case when the components of strain
(2.6) are different from zero but the components of rotation (2.9).
vanish; then
dw == dv da _ dw dv
~~
du _ m _
"57 17 ~dz"~"dx' "dJ ;
IhT
'
but this is the condition that the expression
it dx 4- v dy + w dz
is the total differential of a certain function
<(*, y, z),
i. e.,
It is said in such a case that there occurs "pure deformation'"

and that displacements have a potential (D(;c, y, z).
11. COMPATIBILITY EQUATIONS
Summarising the foregoing conclusions we see that the displace-

ment of a given point in an elastic body is determined by three
functions (see 2.2):
u, v, w\
the deformation at the given point is determined by six functions

c c c eê e .
(2 12V
However, equations (2.6) show that if three functions (2.2) are

specified, all six components of strain (2.12) will be determined
thereby, being expressible in terms of the first derivatives of
functions (2.2); hence, it may be foreseen that six strain compo-
nents (2.12) cannot be prescribed arbitrarily; there must exist,
certain relations among them to the derivation of which we now
proceed. These relations number six, and they fall into two groups.
First group. Let us differentiate the first two of equations (2.6);
as follows:
^ 2 yy 3
d^^XX d* U
~~dxT
= dy^ dx^ 2 '
GEOMETRICAL THEORY OF STRAIN [Ch. II
adding up these equations by members, we get

2
d 2e x .
d Ida , dv\ o exy
/}2/>
'
dy
2
dx 2 dx dy \dy *
dx / dxdy
This is one of the required relations among strains. By a cyclic

change between its letters, we get the other two equalities. This
leads us to the first group of relations
2
o-e yy o exy
d e
dy
2 '+ Ar2
= ~dx~dy*
x '
^_
d 2 e yy d 2 e zz d 2
ev
[yf_. (2.13)
dz 2 ^ v22 '
'dy dydz
2
d e fr
dx 2 *
dz 2 dz dx
Second group. We differentiate the last three of equations (2.6)

.as follows:
We change the signs in the last of the resulting equations and

then add them all up by members:
~ oZ dxd wdy
_ *
.
de*y
'
dx '
dy dz
Differentiating this equation once more with respect to z and ob-

serving that
dx dy dz ''Wdy'
we get
d fozx d**z.
dz
(
\
deyz
dx
,
dy
s^ =2 dxdy
'
(2.14)
This is one of the required relations of the second group. By a

cyclic change between the letters in equation (2.14), we obtain
.two more equations of the same form. Adding these relations to
Sec. 11] COMPATIBILITY EQUATIONS
relations (2.13), we obtain the following system of equations:
_i/î-t
de*x de
*\_ o
<te (
~^~
dx dy dz 1~ cÂrc^y
'
d*zx
'
*T( dy c)^ dx dydz

*x g
dyl
i
^
*
^
|
^JC
y^
dy dz dx
We have justified the necessity of the existence of these rela-

tions analytically, by comparing the number of functions (2.2) to
the number of functions (2.12) and taking into account the rela-
tions between them (2.6); this same purpose can be achieved by
a geometrical approach. To this end, let us imagine an elastic
body cut into small parallelepipeds and assign each of them the
deformation defined by six quantities (2.12); it is easy to conceive
that if the components of strain (2.12) are not connected by cer-
tain relations, it is impossible to make a continuous (deformed)
solid from individual deformed parallelepipeds. Equations (2.15)
furnish us with these relations; if we do not take them into consid-
eration in specifying strains (2.12), there will occur infinitesimal
discontinuities at each point of the body after deformation. These
considerations lend a new meaning to equations (2.15), and from
this new point of view equations (2.15) are called the equations
of continuity or compatibility of strain components. They were de-
rived by Saint-Venant and are often called Saint-Venant's equa-
tions. We shall make another important observation.
If, given the loads acting on a body, we are in a position to
find directly the displacements of its points u, v, w, strains (2.12)
can be calculated thereupon according to equations (2.6); the com-
patibility conditions will then be satisfied automatically, since
they have been derived from equations (2.6) and follow from them
as a consequence l
.
If, however, given the loads acting on a body, we find stresses

and then strains (2.12), it will be necessary to satisfy simulta-
1
From this point of view equations (2.15) are sometimes called Saint-
Venant's identities.
GEOMETRICAL THEORY OF STRAIN [Ch. II
neously the equations of compatibility (2.15) as well; otherwise

deformations will be incompatible and we shall not be able to
find displacements (2.2) from equations (2.6) as the latter will
involve intrinsic contradictions.
Consider this question in greater detail. By introducing the
components of rotation we find possible to write down a system
it
of equations inverse to (2.6), to solve a set of simultane-

i.e.,
ous equations (2.6) and (2.9) with respect to derivatives (2.7) and
to express them in terms of components of strain and rotation:
du du / v du ,
,
v
dv dv dv
(2.16)
dw dw dw
The matrix of components (2.7) can now be represented conven-

tionally as the sum of two matrices, so that each component of
matrix (2.7) is equal to the sum of the respective components of
the added up matrices:

du du du
dx dy dz
dv
dx
dv dv
dz
- (2.17)
dy
dw dw dw
dx dy dz
The first matrix in the right-hand member is symmetric: it deter-

mines a pure deformation (without rotation). The second matrix
is antisymmetric: we see that it determines a rigid body rotation
(without deformation). Our treatment can be correlated with the

theory of tensors thus making it possible to formulate the last
!
,
result as follows: the tensor of small deformation (2.7) can be re-

solved into a symmetric tensor of pure deformation and an anti-
symmetric tensor of rigid body rotation. Tensor (2.7) is some-
times called the tensor of relative displacements. In fact, consider
an infinitesimal parallelepiped with edges dx=l, dy=l, dz=l-,
it is obvious that
dw
~dx
^
" ^L-e
dy
^
e yr
'
represent the displacements u, v, w of the vertices of this paral-

lelepiped (along the co-ordinate axes) relative to one of them.
1
See the above-cited N. I. Muskhelishvili's book.
Sec. 11] COMPATIBILITY EQUATIONS
The quantities
dv du_
dx dy
are the angles of rotation of its edges with respect to their orig-
inal positions.
Equalities (2.16) may be regarded as differential equations in a,
v, w. If the components of strain and rotation are given at all
points of a body, the displacements themselves of the points of the
body u, v, w can be found by integrating equations (2.16). Car-
rying out this integration, we shall, in effect, convince ourselves-
of the necessity of satisfying the compatibility conditions (2.6).
Let us turn to the equations in the first line of (2.16) from which
we shall seek the function u(x, y, z}\ we note that, according to
the meaning of these equations, their right-hand members must be
partial derivatives of one and the same function; hence, the fol-
lowing three conditions should be satisfied:
dexx
dy dz
which result from eliminating the function u from equations (2.16).

Using (2.18), we find
d<*>z de K y de xx
dx dy
'
dx dz
de K y de v
===
i
A-
dz A*
dz
The last equation can be simplified on the basis of the following;

identity, which is obtained by differentiating equalities (2.9):
~
dz (2.19)
and we have then
de xx
dx
(2.20)
'
dz
.58 GEOMETRICAL THEORY OF STRAIN [Ch. II
By performing identical operations upon the second and third

lines of equations (2.16), we eliminate v and w
from them and
obtain six more equations in addition to (2.20); in this way the
nine partial derivatives of the components of rotation will be ex-
pressed in terms of partial derivatives of the components of strain;
in other words, we obtain a system of equations for co*, o) ?/
(o z
,
analogous to equations (2.16) for u, v, w. Hence, it is permissible

to apply the foregoing procedure to the equations thus obtained
and to eliminate o> x co^, co z from them; we shall again obtain nine
,
equalities but they will contain only the second derivatives of the
strain components. Only six of these equalities are different and
they will coincide with equations (2.15). Thus, Saint-Venant's
compatibility equations (2.15) are, in effect, the necessary condi-
tions for displacements to be determined from (2.6) according to
the given strain components.
Saint-Venant's conditions are simultaneously sufficient for this
purpose, if we consider a simply connected body that has no
through cavities. In the case of a multiply connected body, Saint-
Venant's conditions also permit the determination of displacements
u, v, w by integrating Cauchy's equations (2.6); these displace-
ments, however, may now be represented by multiple-valued func-
tions of x, y, z; besides Saint-Venant's conditions, it is necessary
to introduce some additional conditions to make displacements
single-valued, as it is required by the physical meaning of the
problem.
The case of deformation of a multiply connected body is treated
in detail in Chapter I and Appendix II of the above-cited
N. I. Muskhelishvili's book.
*12. TENSOR CHARACTER OF THE STRAIN AT A GIVEN

POINT IN A BODY
In Chapter I we introduced the concept of the stress tensor;

from the viewpoint of the general theory of tensors the formal
and at the same time basic feature of the tensor character of a
state of stress at a given point is that, as one passes from the co-
prdinate areas to an arbitrary area with outward normal v, the
stress components X Vy Y v Z v are expressed by formulas (1.8a),
,
which are linear with respect to the original components (1.16)

as well as to the direction cosines /, m, n. Under a complete
transformation involving the passage from the axes x, y, z to new-
axes u, v, w, the components of the state of stress are expressed
in terms of the original components according to formulas of the
form of (1.12) and (1.13), which are linear with respect to the
Sec. 12] TENSOR CHARACTER OF STRAIN AT GIVEN POINT IN A BODY 50
original components (1.16) and quadratic (or so-called bilinear)

with respect to the direction cosines of the new system (1.10).
In this section we shall show that the deformation of a body
at a given point, determined by nine components of matrix (2.7) r
is also a tensor from the above point of view.
The deformation at a given point in a body will be completely
determined if we can calculate the unit elongation of any infini-
tesimal segment drawn from the given point. Therefore, consider
(Fig. 26) such a segment
= p. We designate its projections on
AB
the co-ordinate axes by g, r), The drawing shows the co-ordi-
.
nates of the initial point A of the segment and those of the ter-
minal point B. In the process of deformation these points will
Fig. 26
occupy new positions AI and fli; their co-ordinates are also shown
in Fig. 26. The projections deformed segment A Bi on the
of the {
co-ordinate axes will obviously be obtained by subtracting the

original co-ordinates from the new ones; for instance,
(2.21 a)
in the same way we find
7)!
= + ij 8<y; d = C+8w. (2.215)
Obviously, the quantities
gj g = 8g; Tj! 7]
= 8i|; ^ = 8t (2.22)
represent the increments of the projections of the segment p in

the process of deformation of the body while the quantities 6w,
6t>, 8w are the increments of the functions u v, w as the point A t
moves to the point B, i.e., as the independent variables x, y, z

are changed by the amounts
dx = 5; dy = TJ; dz = C.
<60 GEOMETRICAL THEORY OF STRAIN [Ch. II
These quantities are infinitely small because of the infinitesimal

magnitude of the segment AB= p; therefore, the infinitesimal in-
crements 6w, dv, 6w of the functions u, v, w may be replaced by
their differentials (omitting small quantities of higher order):
. du * du du ,
.
dx ^ dz' '
'
dy
* dv * dv dv *
* .
formulas (2.21a, b) then become

du du du - m
dx ~dz^
^
dw
dv
dw
dv
dw
..
(2.23)
~dx~
We divide these equalities by the original length of the segment

p and introduce the notations
(2.24)
Here /, m, n are the direction cosines of the initial segment AB\

xr e yo e.p are the elongations of the projections of this segment
-e ,
per unit of its length. Formulas (2.23) then become
e^ n
du
dx
= du du
*
= dv
' dv_ r
dv
(2.25)
dw dw dw
~w "~dz"
These relations are analogous in structure to relations (1.8a), in

which the quantities
^JT *
y ^2> *
z* ^*xi ^y
represent the components of the stress tensor; hence, the compo-
nents of matrix (2.7) entering in relations (2.25) are the compo-
nents of a tensor which we have called the tensor of relative
displacements; however, unlike (1.16), this tensor is asymmetric
Sec 12] TENSOR CHARACTER OF STRAIN AT GIVEN POINT IN A BODY 61
since, in general,
dv du dw dv du dw
dx "37" "aF T5T
These inequalities are transformed into equalities only in the ab-
sence of rotation at a given point in a body (co x = 0)^ = 0)2 = 0), i.e.,
in the case of pure deformation (see end of Sec. 10).
The left-hand members of equalities (1.8a) and (2.25) are also
identical in meaning; they are projections on the
co-ordijjate
axes: in the first case the projections of the total stress on the
area with outward normal v(l, m, n)\ in the second case the
quantities proportional to the elongations of the projections of
a segment having the direction cosines /, m, n.
Now consider the elongation of the segment AB=p itself;
since
the elongation, to small quantities of the second order, will be

obtained from the equality
By dividing both members of this equality by p

2
, we get
L
P
= p
4- .il .4-
p p
"
p p ?
The left-hand member represents the unit elongation of the seg-

ment p:
by using notations (2.24) and equalities (2.25), we find

dw 2
= dx
du
-^ r+
/2
.
'
dv
-3
dy
m *
z
H ,
'
?
dz
/ dv
n* +- ,
'
du \
\
-r
dx
,
H^
'
dy )
or on the basis of Cauchy's equations (2.6)
m 2
+ en 2
#/. (2.27)
Equalities (2.26) and

(2.27) confirm definitely the tensor character
of the quantities involved in the matrices
*xx ex
and ~e (2.28)
for, knowing the components of these matrices, we can find the

unit elongation of an arbitrary segment p drawn from a given
point; this, obviously, determines completely the deformation of
a body at a given point in the same way as the knowledge of
nine components of the stress tensor (1.16) fully determines the
state of stress at a given point in a body, i.e., allows the stresses
to be determined over any area passing through this point. We
have introduced the halves of the shear angles e xy e yz e zx in rela- , ,
tion*^ (2.27) and in the second of matrices (2.28) (cf. 2.6a); rela-
tion (2.27) has then assumed a form quite similar to (1.19). The
first of matrices (2.28) was called the tensor of relative displace-
ments; it can now be established that the second matrix defines

the strain tensor at a given point. Relation (1.19) permitted
us to give a geometrical representation of the stress distribution
at a given point with the aid of Cauchy's stress surface; relation
(2.27) makes it possible to give a similar picture to describe the
deformation of a body at a given point. Indeed, we multiply both
members of equality (2.27) by p 2 and take into consideration
the first line of notation (2.24); it can then be rewritten as
P
2
*
P
= exx $ + e yym? + e, n? + 2^ z y fy + 2e ^ -+ 2,.
yz (2.27')
We take the point under investigation to be the origin of co-ordi-

nates (|, T), t,) and construct at it a vector equal in magnitude
to the segment p; then g, TJ, t, will be the co-ordinates of the
end of this vector; we shall adjust the length of the vector so as
to have at all times
P 2e
= p, i.e., p
= |/^
_ -
(2.29)
p
Then (2.27') will be transformed into the equation of a surface of

the second order
where
F&\ C)=6 2
,
(2.30)
F& ^ 9= ^ 2
+ * 1 + *t + 2^
yy
2 2
(2.31)
Surface (2.30) is quite analogous to Cauchy's stress surface (1.23);

it possesses the same properties and is called the strain surface.
It is a central surface of the second order with centre at the
point under investigation and it may be either an ellipsoid or

the combination of a hyperboloid of one sheet and a hyperboloid
of two sheets with the common asymptotic cone. If we draw ra-
dius-vectors p from its centre up to the intersection with the sur-
face, we shall have from (2.29)
Sec. 12] TENSOR CHARACTER OF STRAIN AT GIVEN POINT IN A BODY 63
It follows that the unit elongation at the point in question

measured in the direction of the vector p is inversely proportional
to the square of the length of this vector. The whole series of
such vectors characterises completely the deformation of the body
at a given point and may, therefore, be regarded as a geometrical
representation of the strain tensor (just as a straight line segment
of given length and direction is a geometrical representation of
a vector).
Surface (2.30), like stress surface (1.23), has three mutually
orthogonal principal axes; if it is referred to these axes, the
terms containing the products of co-ordinates disappear in equa-
tion (2.30), i.e., their coefficients become zero:
hence, deformations are absent in the principal planes.

shear
To determine the principal axes one must, according to the rule
of analytic geometry, equate to zero the determinant
-xy *
-yy
e *y* = 0,
-yz
which gives for the determination of e the cubic equation
similar to equation (1.32); its coefficients
(2.32a)
^~^ -yz
Ti
(2.32b)
'xy
(2.32 c)
represent invariants under a transformation of co-ordinates. Its

three roots are real and represent principal elongations (elonga-
tions along the principal axes); denoting them by e it e 2 e^ we ,
can write down the equation of strain surface (2.30) in the prin-
cipal axes
> = 62 .
(2.33)
The form of this surface, just as that of the stress surface, depends
on the signs of the principal elongations e it e 2t e 3 If all three .
elongations are of the same sign, the surface is an ellipsoid; in

this case, in all directions at a given point there occurs either
extension (if the principal elongations are positive) or compression
(if the principal elongations are negative). If, however, the prin-
cipal elongations are of different signs, surface (2.33) should be

represented as a combination of hyperboloids of one and two
sheets, with the asymptotic cone separating them. Let us construct
a vector in the desired direction from the given point of the body
from the centre of the surface); if the vector intersects the
(i.e.,
hyperboloid of one sheet, extension occurs in this direction; in the

directions which intersect the hyperboloid of two sheets, compres-
sion occurs; in the direction of the generators of the asymptotic
cone the length of the vector becomes infinite; in these directions
elongations are equal to zero.
*13. DILATATIONAL STRAIN.

INVARIANTS OF THE STRAIN TENSOR
In addition to strain components (2.12), let us also find the

dilatational Consider an infinitesimal element of a body
strain.
of volume d~c = dxdydz. It is easy to conceive that if shear
deformations alone are produced in the given element without
elongation of its edges, the unit change of volume of the element
will be a small quantity of higher order compared to the shear
deformations. Consequently, we may assume, neglecting small
quantities of higher order, that the change of volume dr will
depend only on the elongations of the edges dx, dy, dz.
The new volume of the element after deformation will be
Opening the parentheses in the right-hand member we get

di + 5 (di) dxdydz(\ + e xx 4- e yy + +
e zz
i
e xx e yy i
&xx e zz \
e yy e zz ~f~ ^xx^yy^zzr
The four terms in parentheses, which are small quantities
last
of the second and third order, are disregarded; then, taking into
account that dxdydz = di, we have
Hence we obtain the dilatational strain (the cubical dilatation or

simply the dilatation)
,
*x '
yy '
*
Sec. 14] STRAIN DEVIATOR AND ITS INVARIANTS 65
Let this quantity be denoted by 0; in addition to equations (2.6)

we write down the expression of the dilatational strain
+ *, (2-34)
or
(2.35)
Thus, the dilatation coincides exactly with the first, linear, invari-
ant of the strain tensor (2.32a). It is constructed, just as the
other two invariants, similarly to the respective invariant of the
stress tensor; therefore, all we have said in Sec. 5 concerning the
invariants of the stress tensor may apply to the invariants of the
strain tensor; this will reduce to the formal replacement of the
notations
Xx i
Yy ,
y,
/ z,
by the appropriate notations
*14. STRAIN DEVIATOR AND ITS INVARIANTS
The strain tensor
-yz (2.36)
is resolved, similarly to (2.28), into two tensors
-yy
e e yz (2.37)
tyz
where
1
(2.38)
represents themean elongation at a given point. The deformation

corresponding to the second term in (2.37) is such that an ele-
ment isolated at a given point in a body receives equal elonga-
tions in all directions (the strain surface is a sphere); hence, the
element remains similar to itself, only its volume being changed,
5-1013
and so the tensor
(2.39)
is spherical. The first term
&xx ^ ^xy
&
f> pv
c (2.40)
represents the strain deviator characterising the change of shape

(shearing distortion or simply distortion) of the element isolated
at a given point in a body without change in its volume, since
in this case the volume expansion becomes zero
(2.41)
The resolution of (2.37) is not merely a formal operation; it
reflects the physical properties of the phenomenon of deformation,
since real materials behave differently, in the sense of resistance
to the change of volume and to the change of shape.
Consider the invariants of the strain deviator; they are con-
structed in the same way as the invariants of the stress and
strain tensors, with an appropriate replacement of notations.
The first, linear, invariant is zero, according to (2.41). The
second, quadratic, invariant, by analogy with (1.49), has the form
+e zx )*}\ (2.42)
it represents the first characteristic of distortion corresponding

to the deformation defined by the strain tensor (2.36); if the
principal axes of strain are taken for co-ordinate axes, then,
according to (1.50), it will assume a simpler form
1 r/
(2.420
The third, cubic, invariant
-e e xy
-xy
e yy -e (2.43)
Sec
is
or,
15]
_
simpler form
X'
according to (1.52),
= (e,
FINITE STRAIN
the second characteristic of distortion; if

principal axes of strain, it will assume,
e) (e 2 e) (e^
_ it
e)
is referred to the
similarly to (1.51),
*)
3
1-
67
(2-44)
Thus, it may be said (just as in the case of the stress deviator)

that the second invariant of the strain deviator characterises the
average quadratic deviation, and the third invariant the average
cubic deviation of the given deformation from the volumetric
deformation defined by tensor .(2.39), where
and in which no distortion occurs. In this connection the quad-

ratic invariants of the stress and strain deviators play an impor-
tant role in the present theory of plasticity, since in the plastic
state of a body the forces applied to it produce mainly a change
in its shape and only an insignificant change in its volume.
In ChapterI we derived two basic groups of equations of the
theory of elasticity: (1.5) and (1.8). These equations are of a static
character. In the present chapter we have obtained new basic
groups of equations of a geometrical character: (2.6), (2.15) and
(2.35). It is to be noted that these equations are approximate
since in deriving the basic relations
xx
du
xx === -3
dx
a
dv
-3
J
dx
= '
'
Q
P
r
= du
-3
dy
we have disregarded small but finite quantities. One must take
it into consideration in .evaluating all the subsequent conclusions;
they will be sufficiently accurate in the case of small deforma-
tions (when all the strains e xx e yy e zx e xy are very small as
, , . . . , ,
compared to unity); in the general case strains are expressed in

terms of displacements in the form of far more complicated non-
linear relations which will be dealt with in the next section.
*15. FINITE STRAIN
In the preceding sections, we dealt, strictly speaking, with

infinitesimal deformations; indeed, basic relations (2.6) are abso-
lutely exact in this case since only small quantities of higher
order have been neglected in deriving these relations, but this
5*
68
_ _
more exact
GEOMETRICAL THEORY OF STRAIN
does not affect their accuracy. It will now be shown how to obtain
relations corresponding to finite deformations in place
of relations (2.6).
Starting with formulas (2.23), we replace their left-hand
bers according to (2.2 la, b) and rewrite them as
[Ch. II
mem-
,
dv
+-3
dw ..
,
dw ,
We divide both members of these equalities by the altered length

of the segment PJ; furthermore, we multiply and divide the
right-hand members by the original length of this segment p. Then,
taking into account that
represent the direction cosines of the segment p before deforma-

tion and
!L
=/I; L
JL = mi L ;
JL = ni
!
pi PI PI
are the direction cosines of this segment after deformation, we

arrive at the following equatities:
p_
J pi
(2.45)
We eliminate from these equalities the direction cosines /i, m^ n^

of the segment pj, which are obtained in the process of defor-
mation and are not known in advance; for this purpose we square
all three equalities and add them up by members; then, taking
into account that
;we obtain the relation

2
(2 46)
.
Sec. 15] FINITE STRAIN 69
where A, B, C represent the bracketed expressions in equalities

<2.45). Writing out the values of A 2 B 2 C2 we
, , , get
dw dw
We add up these three expressions together and substitute them

in (2.46) we begin with the first terms oJF the expressions and
;
note that
we transpose this result to the left-hand side; adding up the

remaining terms of the expressions, we collect together the coeffi-
cients of the squares and products of the cosines /, m, n. We
transform the left-hand member of the equality as
~"
here e= pl p
represents the unit elongation of the 'segment p.
Hence we obtain the following equality:

~ "
JC /*/), (2.47)
where for brevity we have introduced the notation
*
2
dv 2
du
P
xx ~~ dx
^L ^ 1
\j \ ~
r \^)
i ( \
"""
i
t
2 [ \dx ) (dx
dv , du . du du . dv dv . dw dw (2.48)
dy dz dz dy
du ,
dw ,
du dw ,
"
dv dv ,
dw dw
fry n AY '
"i
d^ dx f\y dx
fry '
dz dx*
/^^ /) I
/}y
dz dx
/)v-
Comparing relation (2.47) with (1.19) and (2.27), we see that

quantities (2.48) are the components of the tensor
(2.49)
To distinguish this tensor from (2.36), it is called the finite

strain tensor. If its components are known, we can calculate the
unit elongations at a given point in a body in any direction
defined by the cosines /, m, n\ indeed, denoting for brevity the
right-hand member of (2.47) by 2/(/, m, n), we obtain the quad-
ratic equation for the determination of e
e* + 2e 2/=0; e = V\+2fl; (2.50)
the second root is dropped, since it gives only a negative value

of e. The terms in'formulas (2.48) coincide with the right-hand
first
members of Cauchy's formulas (2.6) which are obtainable from
those formulas if* we disregard the squares and products of the
components of the tensor of relative displacements (2.17) in the
case of their smallness. It should be noted, however, that compo-
nents (2.48) do not by themselves express strains as contrasted
to the case of small deformations. Indeed, let us take the segment p
in the direction of the x axis, setting /=!, m = n = 0; then from
(2.47) we find the elongation along the x axis
If components (2.7) represent proper fractions considerably less

than unity, then assuming approximately
/l+2e jrjr
1 + e,,,
we find that they approximate to strains, for instance,
Finally, if all the derivatives entering in (2.48) are very small,

then, disregarding small quantities of the second order in (2.48),
we return to Cauchy's formulas.
Exercises to Sections 10 and 11.
1. Given a rod subjected to bending in the plane Oxz (Fig. 8), it is
required:
(a) to write the equation of its axis (i.e., axis Ox) before bending and
to express the deflection of any point by the use of equations (2.2) of Sec. 10;
(b) assuming that the rod is built-in at the origin and using equations
(2.3), and (2.4), to write down the conditions that the element dx of the axis
Sec. 151 FINITE STRAIN 71
of the rod at the origin is fixed (does not rotate in the plane Oxz)\ the same
for the element dz in the plane of the cross section.
2. In the case of the pure bending of the rod considered in the preceding
exercise, displacements (2.2) of Sec. 10 are expressed as
(2.5i;
2P
It is
requirea:
(a) to find the general expressions of strains according to equations (2.6)
and (2.35) and to check whether they satisfy the compatibility equations (2.15);
(b) to write the equation of the deflected axis of the rod; for that substi-
tute the co-ordinates of a point of the axis before bending in expressions (2.51)
(point "a" of the preceding exercise);
(c) to check whether the fastening conditions at the left-hand end are sa-
tisfied (point "b" of the preceding exercise).
3. In an investigation of the torsion of a circular bar (Fig. 7) one obtains
the displacements
u tyz -f- #y + bz -f- c\ v = ixz ax -\- ez -f- /; w= bx ey-\- k.
It is required:
to select the coefficients a, b f c, e, f, k so that the end section 2=0
(a)
will be fixed in the following manner: the point O
must have no displacements;
the element of the axis dz rotates neither in the plane xOz nor in the plane
yOz\ the element of the axis dy does not rotate in the plane xOy\
(b) to find the magnitudes of the strains (2.6) and (2.35);
(c) to check whether equations (2.15) are satisfied.
Ill
Generalised Hooke's Law
16. GENERAL
In the preceding chapters we presented the theory of stresses
elucidating the static aspect of the problem, and also the theory
of displacements and strains elucidating the problem from the
geometrical point of view. These two theories alone cannot be
instrumental in solving the physical problems of the theory of
elasticity concerning deformations that are produced in an elastic
body under the action of external forces applied to it until stresses
and strains have been connected by a physical law. The physical
nature of this law makes it connect dissimilar features of the phe-
nomenon under investigation stresses and strains.
The theory of elasticity treats this law in the most general form.
Its analytic form, i.e., the general form of functions relating
stresses to strains, is ascertained in the first place:
===
-^x J\ \Pxxi ^yy ^zz> ^yz zx , ^ x y)t
' :=r::
y
, e yy , (3.1)
===
/6
it is readily established that in the case of small deformations

the simplest and most rational form of relations (3.1) is a linear
form. After that the values of the coefficients of functions (3.1)
are determined for various cases of structure of an elastic body.
Denoting the coefficients by a mn we represent the first function
,
of (3.1) as follows:
Xx = a u e xx -+ a l2 e yy + aê + a zz }4 e yz +a l5
e zjc -f a l6e xy .
(3.2)
The linear form of relations (3.1) formulates in its very essence

the principle of superposition (the law of independence of effects
Sec. 16] GENERAL 73
of implying, for example, that the stress Xx developed

forces)
in the presence of several strain components e xxi e yy e xy is
,
. . .
,
equal to the sum of stresses caused by each of these components

separately.
According to their structure, physical bodies are divided, first
of all, homogeneous and nonhomogeneous. A homogeneous
into
body is one whose structure and composition are the same at
all its points. The theory of elasticity deals almost exclusively
with homogeneous bodies; however, even among homogeneous
bodies one has to distinguish between tsotropic bodies whose prop-
erties are the same in all directions and nonisotropic (aniso-
tropic) bodies. Many crystals appear to be nonisotropic; while
1.
a b
Fig. 27
the structure of such crystals is homogeneous, their elastic and

optical properties are different in different directions. detailed A
analysis of relations such as (3.2) for nonisotropic bodies shows
that the numerical values of the coefficients a mn are closely
related to the elastic properties of a given body in different di-
rections.
It follows from this analysis that relations (3.2) assume the
simplest form for an isotropic elastic body, i.e., for a body whose
physical properties are identical in all directions; they can be
derived on the basis of Hooke's law for elastic bars in tension and
compression, which is known from physics, and also on the basis
of the above-formulated principle of superposition.
Fig. 27a, b shows the so-called tensile test diagram expressing
the relation between the tensile (normal) stress X
x developed in
the bar and its unit elongation e xx The shape of this diagram is
.
different for different materials and is substantially affected by

their chemical composition and structure. A diagram of the a type
is characteristic of metals possessing ductile properties, for in-
stance, of mild steels with low content of carbon. The initial por-
tion OA of this diagram is straight, provided the test is conducted
thoroughly, an indication of the proportionality between stress and
74 GENERALISED HOOKE'S LAW [Ch. Ill
unit elongation
Xx = Eexx \ (3.3a>
this relation formulates Hooke's law of elasticity. The terminal

point A of the portion corresponds to the stress a p which is called
the proportional limit. At the same time, experiments show that
for nearly all materials the longitudinal elongation of the bar
c xx in simple tension is accompanied by equal lateral deforma-
tions e yy = e zz of opposite sign (i.e., by contractions), these defor-
mations being proportional to the primary elongation e xx :
e yy =e =
zz oe x
-
(3.3b>
the factor of proportionality a represents a constant for each ma-

terial but it is different for different materials. Thus, Hooke's law
is formulated by two relations: (3.3a) and (3.3b); they contain

two numbers characterising the elastic properties of a material:
the longitudinal modulus of elasticity or Young's modulus E (also
called the tensile or
compressive modulus
of elasticity or simply
the modulus of elasticity) having the dimension of stress
2
(force/length and the dimensionless number a called Poisson's
!
ratio\ sometimes the reciprocal of Poisson's ratio. ra = is em-

,
ployed. It is called Poisson's number.

The numerical value of the modulus of elasticity E varies over
a very wide range for different materials; for example, for steels
we have approximately = 2.1X10 6 kg/cm 2 for wood
, =1X
5
XlO kg/cm 2
Poisson's
. ratio a is always expressed by a proper
fraction smaller than 0.5; this latter circumstance can be estab-
lished a priori from physical considerations, as will be shown in
Sec. 18. In the case of materials which possess ductile properties
to a small extent only or not at all, i.e., in the case of brittle
materials, such, for instance, as hard alloy steels, cast iron, stone,
the tensiletest diagram has no initial straight-line portion
(Fig. 276); but in most cases its initial portion deviates but slightly
from a straight line; to simplify the theory this portion is replaced
approximately by a straight line and thus Hooke's law is some-
times applied conditionally to materials which behave as brittle.
Experiments show that as long as the material acts under con-
ditions of elastic properties (the straight-line portion of the dia-
gram in Fig. 27a), there is proportionality between shearing
stresses on the faces of an elementary parallelepiped and shearing
strains:
XyÔe^ Y2 = Qe^ Zx = Ge zx .
This relation formulates Hooke's law in shear-, the factor of pro-

portionality G is called the modulus of elasticity in shear .or the
Sec. 17] STRAINS EXPRESSED IN TERMS OF STRESSES 75
transverse modulus of elasticity (also called the shearing, modu-

lus of elasticity, the shear modulus or modulus of rigidity); its
dimension, just as the dimension of the tensile (compressive)
modulus of elasticity, coincides with the dimension of stress.
The modulus of elasticity in shear is also a characteristic of the
elastic properties of a material but, as we shall see at the end
of Sec. 23, there exists a relation between the three elastic con-
stants E, G and a
G= (3 ' 4 >
ÎT^
resulting from Hooke's law; therefore, only two of these constants
are independent and they must be found by experiment; the third
constant is determined from formula (3.4). For instance, it is most
convenient and reliable to determine the moduli E and G by
experiments; then Poisson's ratio is calculated according to for-
mula (3.4):
In addition to considerations presented here, let us assume, as

yet without proof, that in an elastic homogeneous isotropic mate-
rial normal stresses do not cause shear deformations and, vice
versa, shearing stresses do not produce elongations in the direc-
tion of their action. On the basis of this assumption we may
consider separately the case of normal stresses and the case of
shearing stresses to obtain the most general relations between
stresses and strains of an elastic homogeneous isotropic material.
A better substantiated derivation of these relations and their
extension to homogeneous but anisotropic materials will be given
in Sections 19-23.
17. STRAINS EXPRESSED IN TERMS OF STRESSES
Let us begin with normal stresses and consider (Fig. 28)

a parallelepiped with edges equal to unity, subjected to the action
of normal forces. If the stress Xx were acting alone, the unit
elongation e xx , according to Hooke's law, would be e'

xx
= -X .
Un-
der the action of the stress Y y alone, the elongation along the axis
Y
Ox would be^ = "g-
where a is Poisson's ratio. Likewise,
under the action of the stress Z z we would have e'" = o-^-.

Assuming without proof the principle of superposition, we obtain

the total elongation along the axis Ox:
or
1
Similarly, for elongations along the other two axes we have
e yy = -r\Y -^Z,+ X )Y
y JC
e zz = [Zz -a(Xx +Y y )}. (3.5)
Formulas (3.5) express the generalised Hooke's law for normal
>-
Fig. 28
stresses; they can be represented in the alternate form adding

to the right-hand members respectively
we then obtain
(3.6)
where
Hooke's law is more convenient in this form for some compu-

tations.
Sec. 17] STRAINS EXPRESSED IN TERMS OF STRESSES 77
Relations between shearing strains and stresses are taken in the

above-indicated form:
-yz G
(3.8)
G
Formulas (3.5) or (3.6) in conjunction with formulas (3.8) give
the generalised Hooke's law for a homogeneous and isotropic
elastic body, i.e., a body whose elastic properties are identical in
all directions.
This leads us to a new basic group of equations of the theory
of elasticity
I r ** / J
_ (j
l Z
_ (3.9)
e zx * x
G E "
P J- y 2 + g)
y
Let us revert to equalities (3.6) and add them up by members:
yy
= \(\ + a) (Xx + Y +Z -
y 2)
Since on the basis of equations (3.7) and (2.34)
the last equation can finally be written down as

G 1 2a ^
(3.10)
i.e., the unit volume expansion is proportional to the sum of

three normal stresses 0. This is Hooke's law in the volumetric
form.
From equality (3.10) we can draw the following conclusion:
the dilatation as such must not depend on the directions of the
co-ordinate axes. Consequently, it does not vary with the rotation
of the co-ordinate system and, therefore, according to (3.10), the
sum of normal stresses on three mutually perpendicular areas at

a given point is a constant quantity and does not depend on the
inclination of these areas. In other words, the quantity
is an invariant under a transformation of co-ordinates, as already

stated in Chapter I.
18. STRESSES EXPRESSED IN TERMS OF STRAINS
Equalities (3.9) and (3.10) give Hooke's law in the form

solved with respect to strains. It is often necessary to have* the
same relations in the inverse form, i.e., in the form solved with
respect to stresses. For this purpose we take the first of equations
6=
n>
(3.6) and introduce in it the value

-y ^-6 from equation
(3.10); we get
hence we find
or
y* _
~ __ (1 + 2a)
a) (1
~r
1 -f. a xx*
XX = M -\-2v4w (3.11)
where
X== *= 3 12 )
(l + a)(l-2a) ;
2<l + o)' <

'
X and are called Lame's coefficients. They characterise, as do

jut
the moduli of elasticity E and G, the elastic properties of a solid,

being connected with the moduli of elasticity by formulas (3.12).
It can easily be observed that the coefficient p, coincides with
the modulus == ^'

of shear
2/14-0) p.
= E>
From equation (3.11) we obtain two other equations by a cyclic

change; adding to them the last three equations of group (3.9),
we have Hooke's law in the desired form solved with respect to
stresses
Xy = Kry
Sec. 18] STRESSES EXPRESSED IN TERMS OF STRAINS 79
Summing up the first three equations of (3.13) by members, we

again get the volumetric Hooke's law but expressed now in terms
of Lame's coefficients:
Xx +Y y + Z = 3X9 + 2p
z (exx +eyy + *).
or
e= + (3X 2tx)8, (3.14)
Dividing both members of this equality by 3 and taking into

account that (Sections 7 and 13)
represents the mean or octahedral !

stress and
is the mean elongation at a given point, we represent relation

(3.14) as
.
(3.15)
Formulas (3.12) express Lame's coefficients in terms of the

modulus E and Poisson's ratio or; we can easily obtain inverse
relationships; for that we eliminate the modulus E from formu-
las (3.12) by dividing them by members; we get
X _
~~
23
'
[A 1 2a
hence
__
G
and further
Substituting this in the second of formulas (3.12), we find
c-
= ar
Thus, we have the formulas expressing E and a in terms of Lame's
coefficients:
We have considered two possible ways of expressing the gen-

eralised Hooke's law: in terms of the moduli E and a and in terms
of Lame's coefficients K and ji.
1
See Sec. 7.
Itpossible to introduce two other constants for an isotropic

is
body different ways; it is sometimes convenient, for example,

in
to introduce the following two constants:
2a (3.17)
2a
then equations (3.13) are written down as
(3.18)
If the element under consideration undergoes extension in the

directions of the three axes #, y, z, i.e., if
#,>0; Ky >0; Zz > 0; e > 0,

its volume must not decrease and we must have
0>0.
On this basis we obtain from (3.10)
but then from (3.12) and (3.17) we have

X>0; t*>0; k>0.
Relations (3.9) and (3.13) developed in this chapter constitute
the last group of basic equations of the theory of elasticity, which
completes all the necessary premises for the general solution of
the elasticity problem.
Wehave assumed without proof the equations of the generalised
Hooke's law in the form they are given in strength of materials.
Sections 18-23 below present considerations to prove that these
equations give the most general relation between stresses and
strains in an isotropic elastic body.
In conclusion it may be noted that the concepts of the stress
and strain tensors and deviators introduced in Sections 4, 8, 12
and 14 permit us to express the generalised Hooke's law in more
compact tensor form. Indeed, let us develop the expressions of the
Sec. 18J STRESSES EXPRESSED IN TERMS OF STRAINS 81
components of the stress deviator (1.47) in terms of strains by

using relations (3.13). Taking into account relationship (3.15),
we get
Xx - A/o = 3X* + 21*,, - (3X + 2p) e = 2|i (*,,
- e)\
in the same way we find the other two normal stresses:
The shearing stresses (expressed through the halves of the shear

angles) are
Zx =
Examining these relations, we see that all the components of the

stress deviator are proportional to the corresponding components
of the strain deviator with one and the same factor of propor-
tionality 2(i. This fact can be expressed in tensor form: the stress
deviator is proportional to the strain deviator
e yz
e zz e
or in the compact notation of the deviators accepted by us
(3.19)
In the left-hand member of equality (3.14) expressing the volu-

metric Hooke's law, we introduce the mean (hydrostatic) stress in
place of 0:
e = 3/v ;
we then obtain
N Q
= W, (3.20)
where
is termed the volumetric modulus of elasticity or the bulk modulus

(also called the modulus of volume expansion or the modulus of
compressibility of the material). We
see, thus, that the generalised
Hooke's law is expressible by two equalities scalar (3.20) and
tensor (3.19) and contains, as above, two elastic constants 2u
and K.
82 GENERALISED HOOKE'S LAW [Ch. II I
Exe-cises.
The reader is recommended to carry out the following derivations, as they

are useful for what follows.
1. A thin rectangular rubber plate rests closely, but without pressure,
between two steel plates which can be assumed to be absolutely rigid as
compared to the rubber plate. Friction between the plates is eliminated. If the
co-ordinate system is chosen so that the axis Oz is normal to the planes of
contact of the plates, then, according to the condition of the problem, one
may assume that there are no elongations along the axis Oz, i.e., zz 0. The
rubber plate is compressed by the forces applied to the faces normal to the
axes Ox and Oy.
It is required to find the relation between normal stresses and elongations,
and also the relation among stresses A" x Y v Z z Solve the problem in two
, ,
.
variants, starting with equations (3.9) and (3.13).

2. A rubber cube is inserted tightly in a steel box of the same form; the
cube is closely covered at the top with a steel plate, to which pressure
2
p kg/cm is applied. Considering the steel to be absolutely hard and assuming
that there is no friction between steel and rubber, find:
(a) the pressure of rubber against the box walls;
(b) the maximum shearing stresses in rubber [starting with equations (1.42)
of Sec. 6J.
Solve the problem in two variants starting with equations (3.9) and (3 13),
carry out calculations for cases = and and analyse both of

o*=-p-,find
these cases.
*I9. WORK DONE BY ELASTIC FORCES IN A SOLID
Let us isolate an infinitesimal parallelepiped of dimensions dx r

dy, dz in a body and calculate the work done by the elastic
forces to it, if we give some virtual displacements to points
applied
of the given body.
First consider the case of tension-compression of the element
in the direction of the axis OK. Let the stresses on the faces normal
to Ox be
Xx and Xx + ^dx (3.21)
with the resulting unit elongation e xx and, consequently, absolute

elongation e xx dx. By giving some virtual displacements to the body
this elongation will be changed by a certain amount 8e xx dx.
*
Disregarding in (3.21) the infinitesimal
d
dx
compared as
with the finite quantity X x we
, find that two equal and opposite
normal forces X x dydz applied to the faces of the parallelepiped
do the work X x 8e xx dx dy dz. In the same way we obtain the work
done by the remaining normal forces acting on the faces of the
parallelepiped:
Yy e yy dy dx dz\ Zz e zz dz dx dy.
Sec. 20] POTENTIAL OF ELASTIC FORCES 83
We proceed to the tangential components. The upper and lower

faces of the parallelepiped (Fig. 10) are acted upon by the tan-
gential tractions X z dxdy (the infinitesimal difference between them
is again neglected) forming a couple with the moment
Xz dxdydz\
to the work done by this couple we must multiply its
obtain
moment by the angle of rotation; this angle of rotation will
obviously be given by the increment
which the shear angle e xz receives on the virtual displacement

undergone by the body. On this basis we obtain the work done
by three pairs of tangential tractions applied to the parallelepiped:
X 2 t>e xz dx dy dz\ Yz e yz dz dx dy, X
dy dx dz.
y
e xy
Collecting together the results obtained, we write down the work

done by all the forces applied to the parallelepiped on the virtual
displacement:
= (Xx *exx + Y y
*e
yy +Z z le zz +Y z *e yz +
The work per unit volume of the body at the point where the
parallelepiped has been isolated is
.
(3.22)
*20. POTENTIAL OF ELASTIC FORCES

Wenow introduce a hypothesis which is exceedingly important
in the theory of elasticity, stating that internal elastic forces have
a potential, i.e., the work done by elastic forces is transformed
completely into the potential energy accumulated in a body when
it undergoes elastic deformations and given up by the body in the
form of work done by the forces when the deformations disappear >.
The potential energy is due to deformations and is caused by them
is only approximately true, for a part of the work done

This hypothesis
1
by transformed into other forms of energy (thermal and elec-

elastic forces is
tromagnetic), which are lost by the body and are not recovered in the form
of work done by elastic forces; these additional amounts of energy, however,
are not great and may be ignored in most practical cases.
alone; therefore, if we define by the energy per unit volume of W

a body at a given point, it must be a function of the components
of strain:
W = F(exx , e yr e zz , e yz , e zx , e xy ). (3.23)
If we
give elastic virtual displacements to a body, then, accord-
ing to the hypothesis introduced, the work (3.22) done by internal
forces on these displacements must be transformed completely into
elastic energy and give an increment >W:
%A = 8 W. (3.24)
The increment bW of function

(3.23) replaced by is its first total
differential, to small quantities of the second order,
Expressions (3.22) and (3.25), according to equation (3.24),

must be coincident for any values of virtual deformations
8^, *e yr *e zz , ..., *e xr (3.26)
By virtue of this, the coefficients of quantities (3.22) in equations

(3.22) and (3.25) must be identically equal:
v dW vV dW 7 6W *
y\ Jf
"
-N > * ^ \ /^ y ~~V.
\ x yy zz
dW . .
dW (3.27)
dexy
Thus, if internal elastic forces have a potential, the stress com-

ponents
**x*
Y 7 Y z" **7 xi Y
Vy *"z
'
**y
* 1
are expressed as partial derivatives of the potential energy (3.23)

with respect to corresponding deformations. Equalities (3.27) in
conjunction with (3.23) are an analytic expression of the assump-
tion concerning the existence of the potential of elastic forces.
*21. STRESS-STRAIN RELATIONS;

HYPOTHESIS OF THE NATURAL STATE OF A BODY
Since we are seeking a relation of a physical nature, the
external mathematical form of functions /i, /2, /e in equations . . .
,
(3.1) (Sec. 16), expressing this relation, is in no way restricted

in advance and we have the right to choose it as we please,
it corresponds to the physical conditions of the problem,
provided
i.e., provided it reflects correctly the physical phenomenon under
Sec. 21] STRESS-STRAIN RELATIONS 85-
investigation. Taking advantage of this, we select the simplest

form of relations (3.1), viz., a linear form:
a l4 e yz
Yy = aêxx + a 22 e yy
Yz = aê xx + a 4<2
e yy
(3.28)
-f a 6 ^y .
The considerations supporting the reasonableness of our choice

are as follows.
1. The linear form of the relations is in good agreement with
experiments for many materials under simple tension or compres-
sion, where it is observed as direct proportionality between stress
and unit elongation (Hooke's law).
2. To given strains
at a given point there should correspond just one system of

stresses
Xx ,
Yr Z z Yz Z x
, , , X y (3 30)
and, conversely, given stresses (3.30) entail exactly one system

of strains (3.29).
These conditions are satisfied by a linear form
of the relations . J
3. If we
limit ourselves, as before, to a case of very small
deformations, we shall show that whatever the analytic form of
relations (3.1) may be, they can be replaced approximately by
linear functions; in fact, expanding the right-hand members of
equations (3.1) in a Maclaurin series, we obtain, to small quanti-
ties of the second order,
, (3-31).
where the zero index in the function / t and its derivatives shows
that one must put
C xx = 6yy ==...= 6x = y
U.
1
If we solve equations (3.28) with respect to strains, we shall also obtain,
linear expressions for them in terms of stresses.
GENERALISED HOOKE'S LAW [Ch. Ill
The coefficients of the strains in (3.31) are constant numbers and,

therefore, function (3.31) is linear with respect to strains (3.29)
!
.
It should also be noted that in the right-hand members of

equations (3.28) we have omitted constant terms that are not
dependent on strains. If we retain them and set all the strains
(3.29) equal to zero, we shall find that stresses (3.28) are differ-
ent from zero (stresses produce no deformations); conversely,
setting all the stresses (3.30) equal to zero, we convert equations
(3.28) into a system of six linear equations with six unknowns
(3.29) for which we find nonzero values (deformations are not
âccompanied by stresses).
Fig. 29
Both circumstances may occur, since in studying deformations

caused in a body by external forces, it may sometimes be expected
that the body has undergone deformation prior to the application
of forces. For instance, take a piece of iron in the form of a sphere
(Fig. 29a) and, making a narrow sectorial cut-out ACBD
in it,
join again the cut planes ACBand ADB

and weld them. We
obtain
a solid body (Fig. 296): after this operation there will undoubtedly
be induced stresses in it, although no external forces act upon it;
such stresses are called initial] they exist, for instance, in cast
iron and steel castings on account of nonuniform shrinkage in
the process of cooling. Let us take state b of the body as initial
when measuring deformations under the action of some forces; we
say, then, that there are no deformations in this state but the
stresses are not equal to zero. If we cut the body again through the
plane ACB, it will return to its former state a\ deformations will
occur then, but the stresses will become zero.
1
This reasoning holds true unless all the coefficients of the strains in
equation (3.31) vanish, in which case one must take into consideration small
terms of the second order in the Maclaurin series; no such case is possible,
however, since in the case of simple tension the existence of first order terms
is readily ascertained (Hooke's law).
Sec. 21) STRESS-STRAIN RELATIONS 87
The nature and fhagnitude of initial stresses are usually unknown

because they depend upon the prior history of a body henceforth J
;
we shall exclude them from consideration, introducing the hypo-

thesis of the natural state of a body, i.e., assuming that in the
absence of deformations in a body stresses in it are equal to zero.
It is clear from the above discussion that there must be no con-
stant terms in equations (3.28) in this case, i.e., they must be
homogeneous, as has been assumed at the outset.
Having established the form of relations (3.28), we can develop
the expression for the potential elastic energy W. Comparing
relations (3.27) and (3.28), we conclude that the partial deriva-
tives of the function W
represent linear homogeneous functions
of the strain components
thus, the function W

itself will be a homogeneous function of the
second degree in these arguments. It can be obtained by inte-
grating equations (3.27), but it is much easier to make use of
Euler's known theorem on homogeneous functions which states
that if F(x, {/, z, .) is a homogeneous function of degree n, then
. .
= _A: + _y + _
"
x
dF dF dF . .
nF(x, y, z, ...)
Applying it to the function W, we find that

dW ,
dW .
,
d
or, using (3.27), we get

.+ (3.32>
This expression can further be handled in two ways:
1. To replace the stress components by their
expressions in terms
of the strain components according to (3.28); then we obtain the
required expression of W
in the form of a homogeneous function
of the second degree in the strain components.
2. To solve system (3.28) with respect to
and substitute these expressions in the preceding equality; then W

will be represented as a homogeneous function of the second
degree in the stress components
Y x* y v >
Y
** v*
1
In the case considered, for instance, the magnitude of stresses depends
upon the size of the sectorial cut-out ACBD.
$8
_ _ GENERALISED HOOKE'S LAW
In the case of an isotropic body, we use rotations (3.13) and

inverse relations (3.9) instead of (3.28); then we obtain the
following expressions for the potential elastic energy:
(1) as a function of the strain components
f- 4 + el) + n (4 + e\x + ely\

[Ch
(3.33)
III
or
e zz
(2) as a function of the components of stress
(3.34)
From (3.33) we draw a very important conclusion that always
since X>0 and |ii>0, as has been shown at the end of Sec. 18.
*22. ELASTIC CONSTANTS; REDUCTION IN THEIR

NUMBER DUE TO THE EXISTENCE OF THE POTENTIAL
OF ELASTIC FORCES
Equations (3.28) contain 36 constant coefficients a^ n called the

elastic constants. They characterise the elastic properties of a
body
and by their dimension they are quite analogous to the moduli of
elasticity E
and G. The number of elastic constants of a body in
general, as we see, is very great; however, it is reduced consider-
ably when the potential of elastic forces exists.
Indeed, take a partial derivative of the potential energy with
respect to any of the strains, for example, with respect to e yy on \
the basis of equations (3.27) and (3.28) we have
z + aê yz -f- aê
taking a partial derivative of both members once more, for example,

with respect to e zxj we get
*
=a '
(3 - 35 )
Sec. 23]
Now
order
find the same second
dW
ISOTROPIC BODY
_
dezx de yy
== de yy
_
derivative by differentiating in reverse
52<
Since the magnitude of a derivative does not depend on the

89*
order of differentiation, we compare equations (3.35) and (3.36)

and find a 25 = a 52 the same can be proved, of course, for any two
;
coefficients a mn and a nm
amn = anm- (3-37>
Relation (3.37) resulting from the existence of the potential of

elastic forces shows that in equations (3.24) the coefficients which
are symmetric in respect to the diagonal passing from the upper
left corner to the lower right corner, are pairwise equal to each
other. Consequently, in the most general case, 36 elastic constants
will include 6 different constants
Qfi A
which are located along the diagonal and ?p
=15 among the
remaining constants, making altogether 6+15 = 21 constants.

Only "the most anisotropic" body exhibiting entirely different
elastic properties in different directions can possess such a large
number of elastic constants.
*23. ISOTROPIC BODY
Physical nonisotropic bodies (crystals) usually reveal greater or

smaller symmetry of structure, as a result of which the number
of elastic constants is considerably reduced; we shall consider
here only the case of an isotropic body whose elastic properties
are the same in all directions. For such a body equations (3.28)
must not alter under any transformations of co-ordinates what-
soever. Hence we can easily reduce the number of elastic constants
to 9 if we take into account the sign rule for shearing strains
(Sec. 10); it follows from this rule that a shearing strain (e. g.,
e xy ) preserves its magnitude but changes the sign, if we reverse
the direction of one of the axes in the plane of which the shearing
strain occurs (Fig. 30).
Take, for instance, the first of equations (3.28) and reverse the
direction of the axis Oy. This will obviously have no effect on
GENERAEISED HOOKE'S LAW [Ch. HI
the left-hand member of the equation; the first three terms in the
right-hand member (containing elongations) will remain unaltered;
the fifth term will not change either; the fourth and sixth terms
preserving the magnitude will change the sign; hence, the
equality between the left-hand and right-hand members will be
violated; this will not happen only when 0i4 = 0, 0ie = 0, i.e., these
two coefficients must vanish. If we reverse the direction of the
axis 02, we shall find in the same way that 0u = 0i5 = 0; thus, we
draw the general conclusion that au = a 15 = a 16 = 0, i.e., the nor-
mal stress Xx is not connected with shear deformations. In
view of the complete equality of the co-ordinate axes in an
isotropic body this conclusion also applies to the other two normal
y
j
Fig. 30
-stresses, Yy and Z z i.e., 024 025

, 026 =
0; 034 035 =
036 0. the= = = = On
basis of equation (3.37) we note now that the symmetric coefficients
in the last three of equations (3.28) become zero:
a41 =a =a =
42 43 0,
a 51 = #53 =
i. e., shearing stresses are not connected with elongations.

The number of elastic constants has been reduced to 12 and
equations (3.28) have fallen into two independent groups:
= ae
\\xx Yz =
= an exx -f- aê y Z ==
Let us turn to the last three equations. If the direction of the

Ox is reversed, the sign of the last two terms in the fourth
equation will be changed to its opposite; the left-hand member
of the equation will remain unaltered and, as before, we, therefore,
Sec. 23] ISOTROPIC BODY
conclude that
and further
The number of elastic constants has been reduced to 12 3=9,.

as stated before,
Xx = a u exx -f 12 yy +a }3e zz ,
tr - i
Y = (3.38)
The transformations of co-ordinates which we have so far

employed came to a rotation of particular axes through 180; now
we employ a rotation of axes through 90 by replacing Oy by Oz+
Oz by Ox, Ox by Oy, etc. We take the first of equations (3.38)
and interchange the axes Oy and Oz. This will have no effect on
the left-hand member of the equation in view of the isotropy of
the body; in the right-hand member, however, the elongations e m
and e zz will be interchanged; consequently, the equality will not
be violated only when 012 = 013; likewise, in the second and third
= =
equations we must have 021 023, 031 032 and finally on the basis
of equality (3.37): 012 013=023 021
= 031 =
032. Now successively
= =
replacing one axis by another and keeping in mind that the form
of equations (3.38) must not be altered thereby, we come to the
conclusion that
11 22 33* 44
hence, equations (3.38) become

Xx = &\\ exx r ^12 (^yy
Y y = a u e yy +a l2 (e zz
z 11 zz\ 12 V xx i
~yyn /Q QQ\
Y =a e (o-oy)
Zx = Q>\t zx ,
Y j
Thus, the number of elastic constants has been reduced to three
an a, *44'
We have achieved this result by fulfilling the requirement that

the form of the relations should not change with the rotation of
the co-ordinate axes through 90 and 180; however, this condi-
tion is not sufficient to guarantee the isotropy of a body, i.e.,
its complete homogeneity in all directions. Therefore, we make
one more transformation of co-ordinates associated with the rota-
tion of the axes through an arbitrary angle.
For simplicity we take the plane-strain situation when
p
^ p
c p
^ u *
Then from equations (3.39) we have
/
y (3.40)
X = y
Consider the case of simple tension in the direction of the x

axis when
Xy
= 0,446x3
= 0.
Hence
(3.41)
-and the first of equations (3.40) yields

2 2
^n ^ 14
X X
&\\
e XX"
Next we calculate the shearing stress U v on an arbitrary area

inclined to the plane Oxz at an angle a; this can be done in
two ways. On the one hand, according to the second- of formulas
(1.15) which gives
Uv = X x sin a cos a = -~ sin 2a = n

^
12
e xx sin 2a.
On the other hand, use can be made of the last of equations

(3.40), since the body is assumed to be isotropic and this equation
js therefore valid for any direction of the co-ordinate axes:
uv aê .
Comparing both expressions of U v we

, get
(
an a ia) exx sin 2a = 2an aue uv (3 - 42)
Sec. 23] ISOTROPIC BODY
We now establish a relation between the strain components e xx

and e uv which appear in (3.42). For this purpose (Fig. 31) we
take a square ABCD with the side equal to unity in the plane Oxy.
It is known from experiment that in the case of
simple extension
in the direction of the axis Ox horizontal segments elongate while
vertical segments shorten and the square takes the shape AiB^^^
Let us calculate the angles of rotation of the sides of the square
and then the shear angle e uv as the distortion of the right angle
ABC. We have
OA = 1 sin a; OB 1 cos a,
and
AA l
1 sin a (
e yy ), BB l
= 1 cos a e xx .
Assuming the deformation to be small we obtain, to small quan-

tities of the second order,
A^ = AA 2 l
=
cos a 1 sin a cos a ( yy ),
BBc, = BB l
sin a = 1 sin a cos a e xx .
We calculate the angle of rotation p of the side AB with the

same degree of accuracy
r vy )smacosa,
or
e xx
-sin2a.
94 GENERALISED HOOKE'S LAW [Ch. Iir
The angle of rotation pi of the other face BC will be obtained

from the same formula by replacing a by a ~.
.
,
The required shear angle is
*uv
= P Pi
= for* e yy ) sin 2<x,
or on the basis of (3.41)

_ 011 + 012
This is the desired relation between e xx and e uv substituting this \
value of e uv in (3.42) we find after a simple transformation
au a l2 = 2a. (3.43)
Thus, the three elastic constants are governed by the linear

relation (3.43) indicating that only two of them are independent;
the third constant will be expressed through them by means of
equation (3.43). Taking a i2 and a 4 4 as independent constants, we
find
an = 012 + 2044-
Consequently, an isotropic elastic body is characterised by only

two elastic constants: a i2 and a 44 These are Lame's coefficients .
that we denoted by K and \i in Sec. 18. Returning to this nota-

tion, 0i2
= A, 44
=
an = A,+2|i, and substituting the same in
M',
equations (3.39), we immediately reduce these equations to the

form (3.13) given in Sec. 18. Hence, we come to the conclusion
that equations (3.13) follow directly from the most general equa-
tions (3.28), if one accepts the hypothesis of the existence of the
potential of elastic forces and if a given body is assumed to be
isotropic.
In Sec. 16 we presented without proof the relation among three
elastic constants , \i,
G:
It can now be shown that this relation is obtainable as a con-

sequence of the generalised Hooke's law (3.13), containing only
two elastic constants K and ji:
(3.44)
Sec. 23] ISOTROPIC BODY 95
By adding up the first three equalities, we obtained Hooke's law

in the volumetric form in Sec. 18:
or 6 =
Substituting this value of in the first of equalities (3.44), we
find the expression of e xx in terms of stresses
e ;
let us compare this expression with the first of equalities (3.6)

written down in the form
Their left-hand members are identical; by comparing the right-

hand members, we obtain
We have taken into account here that from comparison of the

last of equalities (3.9) and (3.13) we get !ti=G; the first of equal-
ities (3.45) gives the required relation (3.4); this together with
the second equality yields, as in (3.16),
IV
Solution
of the Elasticity Problem
in Terms of Displacements
24. COMPENDIUM OF BASIC EQUATIONS

OF THE THEORY OF ELASTICITY
We have completed the derivation of all basic groups of equa-
tions of the theory of elasticity with the generalised Hooke's law;
let us present them once more in the general list and for con-
venience in making references in future furnish them with special
numeration (R-an numerals).
A. Static equations
1. Differential conditions of equilibrium (Navier's equations)

(1.5):
dX x
dx
dYx
dx (I)
dZ x
~dx~
2. Surface conditions (1.8):
Xv = Xx cos (vx) + X y
cos (vy) + X cos (vz)\
t
Yv = Y x cos (vx) + Y y
cos (vy) + Y cos
z (II)
Z& = Zx cos +Z
(I;A:) y
cos (^y) + Z^ cos (vz).
'Sec. 24] COMPENDIUM OF BASIC EQUATIONS OF THEORY OF ELASTICITY 97
B. Geometrical equations
3. Relation between displacements and strains (Cauchy's equa-
tions) (2.6):
du dw , dv
xx X yz i
fi z
'
(jy
l/V ^
fM-
i
dw
t/UC/
yy dv
*
^x /"I i
/) v
The
*
(III)
c^w dv . du
zz dz
'
*y d* '
dy
4. Compatibility equations (Saint-Venant's equations) (2.15):
e xx d 2 yy d 2 exy d /de yz de zx de xy \
(
' '
dy dx 2 dx dy dz \ dx '
dy dz / dxdy
d 2 e yy
_
2
_
L ~~~ ^yz
'
dx dx '
(IV)
dy dz dz dy dz
'
dy \ dy
2
de yz
****
dz 2 dz
'
dz ~dx~~~
^ dezx \
d e yy
dz dx
d;t dy \
C. Physical equations
5. Generalised Hooke's law (3.9), (3.13):
yy
= îy y (Z,+ A,)]; ^ = ^-Z r jr ;
[ (V)
! e*y
= ir X r
(Va)
(V)
(V'a)
On the basis of these groups of equations it is possible to
proceed directly to the solution of the general problem of the
theory of elasticity for the stresses and strains produced in an
isotropic elastic body under the action of external forces.
Itshould be noted once again that all the foregoing equations
have been derived on the assumption of* very small deformations;
7-1013
98 SOLUTION OF ELASTICITY PROBLEM IN TERMS OF DISPLACEMENTS [Ch IV
this was already mentioned in Sec. 10 relative to equations (III)

and, consequently, to (IV); equations (I) have been developed
for the unstrained state of a body and may be considered true
only in the case of very small deformations.
Equations (I)-(V) involve a large number of unknown func-
tions representing stresses, 'displacements and strains. It is,
therefore, necessary first to choose those quantities which we shall
take as basic unknowns to be determined in the first instance
and with which it will subsequently be possible to find all the
remaining factors characterising the state of stress and deforma-
tion of a body.
The history of the development of the theory of elasticity shows
that here one can proceed in two different ways.
(1) To take displacements of points in an elastic body as basic
unknowns; then we have three unknowns at each point (x, y, z)
of the body:
n(x, y, z), v(x, y, z), w(x, y, z}, (4.1)
and the problem be reduced to the determination of three

will
functions (4.1) with three conditions of equilibrium (I) to be
satisfied; at the same time it will be necessary to satisfy condi-
tions (II) on the surface which contain external forces (load)
'
**v* v* ^v'
(2) To take stresses as basic unknowns; then we have six

unknowns at each point of a body
X x (x, y, z), Y y (x, y, z), Zz (x, y, z),
J
Y z (x* y, z\ Zx (x, y, z), X y (x, y, z). J
'
'
The problem is reduced to the determination of six functions

(4.2); they must satisfy three equations of equilibrium (I); these
equations, however, are inadequate, and it will be necessary to
make appeal to six compatibility conditions (IV). At the same
time conditions (II) must be satisfied on the surface, just as in
the first method.
appears from these arguments that the first procedure ("the
It
solution of the problem in terms of displacements") is simpler
from a mathematical point of view, for it involves a smaller
number of unknowns and one has to deal with a smaller number
of equations. After determining displacements (4.1), we readily
find strains e xx e yyy e zz e yz e zx e xy from equations (III) and,
, , , ,
finally, substituting them in the equations of Hooke's law (V')> we

obtain stresses
A* *
y> ^2 *
2' A* Xy.
Sec 25] LAME'S EQUATIONS 99
In the present chapter we shall consider the method of solving

the problem of the theory of elasticity in terms of displacements
and apply it to several particular cases.
25. LAME'S EQUATIONS
Let us write down the differential equations of equilibrium (I)
/ d 2u \ \
dx dy
+ '
+ '
dz
'
.
dY v d2
h dzL + r =
i
y i
~~dx~ ~T~~d~y o(^ P P (I)
dZ x dZ
~~dx~ ^ J?L
.
dy
+ + Z = (^P~^/'
^ z
P
\
and the conditions on the surface (II)
Xv = cos y
cos cos (vz)\
= ^T cos cos (II)
Zv = cos cos cos
All these equations have to be transformed to express stresses

in terms of displacements (4.1) which we take as unknowns. For
this purpose we express stresses in terms of strains
7
according to
Hooke's law (V ), and the strains in terms of displacements
according to (III). Selecting in (V") the stresses entering in the
first of equations (I), we get
dv
(4.3)
du
By differentiating them, we obtain
dX x , dO ,
d 2u d*u
100 SOLUTION OF ELASTICITY PROBLEM IN TERMS OF DISPLACEMENTS [Ch. IV
Substituting this in the first of equations (I), we get

,
-
d 2
w \ ,
( d*u . d*u . d*u \

.
==
We note that
d 2u ,
d*v d*w _
~ d Ida .
~^
dv ,
"^ a^
do; \
_
~ dQ
'
()A: (dx dy J dx
and introduce a compact notation
This expression is called Laplacian operator or Laplace's opera-

tor on the function u(x, y, z). Hence, equation (4.4) becomes
(4.5)
In similar manner we transform the other two of equations

a
(I); are obtained directly from (4.5) by a cyclic change
they
between the letters (x, y, z) and (u, v, w)\ we arrive at the fol-
lowing system of basic equations of the theory of elasticity for the
determination of displacements:
(VI)
the body forces pX, pK, pZ are absent, equations (VI) will
If
be homogeneous. In the case of equilibrium, their right-hand
members are equal to zero; dividing all the equations by [i and
introducing the elastic constants M, and k (3.17), we reduce them
to
, <?0
(VI')
Equations (VI) are known as Lame's equations. They involve

a synthesis of the theories of stresses, deformations and the rela-
tion between them the theories elucidated in the preceding three
chapters. Consequently, Lame's equations include all the premises
Sec. 25] LAME'S EQUATIONS 101
of mechanical, geometrical and purely physical character on

a
which the theory of elasticity is based. In fact,
(1) they express the conditions of equilibrium of each element
of a body (if their right-hand members are equal to zero) or
represent the equations of motion of this element;
(2) they contain the geometrical characteristics of deforma-
tions u, v, w and 0;
(3) they contain the physical factors A,, (i and p characterising
the elastic properties and density of a body.
From these considerations alone we can estimate a priori the
enormous role equations (VI) play in the theory of elasticity by
making it possible to approach the solution of a number of very
important problems.
In the same way we transform the surface conditions (II); re-
placing in the first of them the stresses by their expressions (4.3),
we have
x,, = XO cos (vx) + [~ cos (vx) + ~ cos (vy + -^ cos (vz)]
[i )
The right-hand member of this equation can be written in a

simpler form if we bear in mind that the first bracketed expression
represents the derivative of the function u(x, y, z) along the normal
to the surface of the body:
da _ du dx da dy . da dz _ da
~ ,
COS ^
>
x'
W~~dx dV "dy' dV ~^~
~dF ~dV~ ~~dx
da da
4-
,
jf
COS (Vy)
v
+ -gj
,
COS
.
(VZ).
.
Reasoning in the same way with regard to the other two equa-
tions (II), we obtain the definitive form of the conditions on the
surface:
cos
Yv =
(Via)
cos fr -
cos ( <y2:
17
IF
In conjunction with the surface conditions (Via), Lame's equa-

tions (VI) permit us to proceed directly to the solution of prob-
lems of the theory of elasticity. If we are in a position to integrate
equations (VI) and find the functions u, v and w which satisfy the
conditions on the surface in the form (II) or (Via), then, by
introducing them in equations (III), we determine strains e xx ,
e x]l by introducing the latter in the equations of Hooke's law (V')>

\
we obtain the stresses

yx^ v y y
-*y
Integration of equations (VI), subject to the surface conditions

(Via), is a very difficult task. In solving many problems of prac-
tical importance, however, it is convenient to employ the inverse
method by prescribing displacements as functipns of the co-ordi-
nates of a point (x, y, z) and seeking, on the basis of conditions
(Via), the external forces on the surface of a body (load) to
which the given displacements correspond. Saint-Venant's "semi-
inverse" method has proved to be very fruitful too; according
to this method, it is necessary to prescribe a part of external forces
and a part of displacements, while the remaining factors are
sought from the condition that the basic equations (VI) and (Via)
are satisfied. By way of illustration, we shall apply both methods
to problems encountered in practice.
26. LONGITUDINAL AND TRANSVERSE VIBRATIONS

IN AN UNBOUNDED ELASTIC MEDIUM
Let us apply the inverse method to this problem, i.e., prescribe

displacements and verify whether they are possible in a homo-
geneous elastic medium, in other words, whether they satisfy
Lame's equations (VI). Since we are studying the case of motion,
displacements (4.1) (Sec. 17) should depend not only on the
co-ordinates of a point, but also on the time t.
A. Let us choose the following expressions of displacements:
(4.7)
and, moreover, suppose that the body forces are absent, i.e.,
Since v = w = the displacements of all points take place parallel

y
to the axis Ox; furthermore, the displacement u does not depend on

Sec. 26] VIBRATIONS IN UNBOUNDED MEDIUM 103
y, z\ consequently, if we
consider points which in the
(Fig. 32)
absence of motion are located plane P normal to the axis Ox,
in the
all these points will be displaced equally and simultaneously; in
other words, the plane P will move in the direction of the axis
Ox without deforming.
Indeed, the equation of the plane P at rest will be X = X Q It .
will have the form X = X Q +U or X = XQ + U(XQ, t) at any instant

during the motion. This is again a
plane parallel to the plane yOz but lo-
cated at a time-dependent variable dis-
tance from the plane yOz. If we select
a number of such planes P, PI, P^
in our elastic medium, they will all dis-
place normally to Ox, coming together
or moving away from each other In 1
.
<
this case the motion described by equa-

fions (4.7) is called uniform longitudi-
nal vibration along the axis Ox.
The surface conditions (Via) need
not be satisfied, since the medium has
been assumed to be unbounded.
In order to check the possibility of vibrations governed by
equations (4.7) we substitute them in equations (VI). First we
calculate
6 = du ae
dx
ao _ d 2u a9 _ d 2u
dy dx dy dz aâ^
d2w
dt 2
= 0.
We now observe that the second and third of equations (VI)

are satisfied identically; the first equation is transformed as
d2u d2 u
or
d2u
dt 2 (4.8)
where
(4.9)
1
Since the distance d between any two planes XXQ and x**x\ depends
on the time t
during the motion,
<>, t)] *=*X } ) u (x0t t).

104 SOLUTION OF ELASTICITY PROBLEM IN TERMS OF DISPLACEMENTS [Ch. IV
Hence we draw the basic conclusion: the longitudinal vibratory

motion governed by equations (4.7) is possible if the function u(x, t)
satisfies differential equation (4.8).
B. Let us now choose the following expressions of displacements:
(4.10)
w w(x, t).
Let the body forces be absent as before:
In this case all displacements take place parallel to the axis Oz.
By reasoning as above, we easily deduce that all points of any
plane P (Fig. 32) displace equally and simultaneously, remaining
at a constant distance from the plane zOy. If several such parallel
planes are considered, they will move in a vertical direction. In
the case of a periodic motion we shall have uniform transverse
vibration along the axis Oz. Let us check the possibility of such
vibrations; proceeding from equations (4.10), we have
u ~~ *!L
dx ^ dy _i_^
+*L ^ dz u
o-'
thus, there is no dilatation. Furthermore,
dx~ dy ~ dz U
^_^L__^L_o- '
V W ~
ma)- dx*
-
'
~ ~U
^L-^L-Q
dt* dt*
'
In these conditions, the first and second of equations (VI) are

satisfied identically; the third equation becomes
or
dt*
__ dx*
'
where
2
(4.12)
Vibration (4.10) is thus possible if the function w(x, t) satisfies

differential equation (4.11). We
observe that equations (4.8) and
(4.11) are alike in form and differ only by the value of the con-
stant factor [cf. formulas (4.9) and (4.12)].
C. Consider the particular case of harmonic vibrations. Let us
Sec. 26] VIBRATIONS IN UNBOUNDED MEDIUM 105
[cf. equations (4.7)]:
(4.13)
Substituting this expression in equation (4.8) and cancelling,

we obtain
J__
T 2
Jfl2_
2 '
/
or
(4.14)
Consequently, vibration (4.13) is possible in an unbounded elastic

medium if the parameters / and T satisfy relationship (4.14).
The parameter A (the amplitude of vibration) remains arbitrary.

The parameter T is the period of vibration; in fact, if x is held
constant (i.e., if we consider one and the same plane P in Fig. 32
at all times) and the time t is given an increment T, the deviation
u will not change, according to equation (4.13).
The parameter / is the wave length. Its geometrical meaning
can best be clarified by considering the unit elongation
du
(4.15)
Let t be kept constant, i.e., let us consider all the possible

planes P (Fig. 32) at a certain instant of time. As e xx is repre-
sented by a periodic function of x, formula (4.15) shows that by
giving the abscissa x an increment /, we shall obtain a new
plane P at whose points e xx has the same value; let us consider,
for example, min e xx this will be the case at those places where
\
the thickness of the planes P is greatest at the given instant of

time t (Fig. 33); the distance between these points, as we have
pointed out, is equal to /, and accordingly / is called the wave
length. We shall obtain the magnitude of the maximum contrac-
tion (extension), according to equation (4.15), if we put
or
x t tn / A 1 s*\
7 T^T' (4 ' l6 >
where m is an integer.
Equation (4.16) shows that the abscissa x of this point moves
uniformly as time passes; the velocity of this motion is
i/
dx /
V
~~~dt~ T
*
This is the velocity of wave propagation. On the basis of equa-

tion (4.14) we have
V=a= y ^^. (4.17)
We conclude from the foregoing analysis that the harmonic

vibrationgiven by (4.13) possible in a homogeneous elastic
is
medium. The amplitude A is arbitrary; the wave length / and
the period of vibration T may also be variable, but the velocity
of propagation of longitudinal wave V y, according to equation =
(4.17), is a constant which depends on the elastic coefficients
A, and p, and the density p of a medium.
The longitudinal vibrations under consideration are the so-
called sonic vibrations, since for
-jg-
sec > T > 50)000
sec
these vibrations are heard as a sound. Formula (4.17) is impor-

tant because it gives the velocity of propagation of sound in a
solid of infinite dimensions. Since formula (4.17) simultaneously
gives the velocity of propagation of longitudinal strain e xx [cf.
formula (4.15)], it follows that longitudinal strains are propagat-
ed through an elastic medium with a quite definite velocity equal
to the velocity of sound.
27. GENERAL SOLUTION OF THE EQUATION

OF VIBRATIONS
We have considered only the case of harmonic longitudinal
vibration according to law (4.13); however, there can be various
types of vibrations in an elastic body. Indeed, the most general
-Sec 27] GENERAL SOLUTION OF EQUATION OF VIBRATIONS 107
solution of equation (4.8) is
tt = <f(x at) + y(x + at), (4.18)
where cp and
\|)
are arbitrary functions; substituting this expres-
sion in equation (4.8), we see that it is satisfied identically. It
can easily be shown that the right-hand member of equation
(4.18) represents two vibrations travelling with the velocity a in
the positive and negative directions of the axis Ox.
Take, for example, the particular solution
u = y(x at). (4.19)
If the independent variables x and / are connected by the con-

dition
x at =C (4.20)
Id. equality (4.16)], we obtain: Equationu = y(C) = constant.

(4.20) indicates that a point with a given constant value of the
deviation u moves uniformly with the velocity
dx , r
It can be seen that equation (4.13) is a particular case of equa-

tion (4.19); in fact, it may be rewritten as
u = As\n-^-(x at], where a = y.

Similarly, the particular solution u=ty(x+at) corresponds to
vibration travelling with the same velocity in the negative direc-
tion of the axis Ox.
The ultimate conclusion: the velocity of propagation of lon-
gitudinal vibration in a homogeneous elastic body is constant and
is given by formula (4.17). The law of vibration may be arbitrary
on account of the arbitrariness of functions cp and \|> in the gen-
eral solution (4.18) of equation (4.8).
These considerations are valid, of course, with respect to equa-
tion (4.11) of transverse vibrations as well.
Its general solution is
Here b is the velocity of propagation of transverse vibrations.

Consider a special case of harmonic transverse vibration
w = B sin 2ic y-
= B sin (* bt\
j^
where
* = 7- (4-21)
Here /i is the wave length of transverse vibrations (Fig. 34); 7\

is the period of vibration. As in the case of longitudinal vibra-
tions, we find that the velocity of propagation of vibration [cf.
formulas (4.21) and (4.12)] is
(4.22)
Comparing (4.17) and (4.22), we see that
\V \<\V\, l (4.23)
i.e., the velocity of propagation of transverse vibrations is less

than that of longitudinal vibrations. Their ratio is
f*
(4.24)
v ^
+ 2f*
it depends only on the elastic constants of a medium.

2
Fig. 34
On the basis of equations (III) of Sec. 24 and (4.10) we con-

clude that transverse vibrations are accompanied by shear
dw
xz dx
This shear is propagated through an elastic medium with the ve-
locity given by formula (4.22). From formulas (4.23) and (4.24)

it follows that transverse strains (shearing strains) are propa-
gated through an isotropic elastic medium far more slowly than
longitudinal strains. Indeed, if we set, for example, a= in
(4.24), we obtain
Sec. 28) LONGITUDINAL VIBRATIONS OF A BAR 109
The velocities V and Vi themselves, as is known, are very great;

let us calculate, for instance, the velocity of propagation of lon-
gitudinal vibrations for steel; according to (4.14), we have
'
r r
p ,p
= ,/> 2(l-) ,/" r^

!/ y-rrsr-^V 7
For steel = 2X10 6 I
kg/cm
/ 2
, p
= 7
7.85
^9 kg/cm
1Q33 X981
10 81
3
,
a = 0.3; substitut-
ing this in the preceding formula, we get

2 x 1Q9 X 981 - Q-3) 1 04 T q-gEQ
1Q4 /- 3.350
V =
7.85 (1+0.3) (1-0.6)
= 580,000 cm/sec = 5,800 m/sec.
It should be noted, however, that with such a great velocity of

elastic waves involved, the velocities of the particles of the body
du dv dw .
~
remam ver y
28. LONGITUDINAL VIBRATIONS OF A BAR.

FOURIER'S METHOD
The preceding sections dealt with an unbounded elastic medium,
i.e., with a medium extending unlirnitedly in the direction of
Fig. 35
the three co-ordinate axes. If we consider a thin cylindrical bar

(Fig. 35 a), its vibrations will occur in a manner somewhat differ-
ent from that for an unbounded medium; for instance, when
subjected to longitudinal vibrations, its lateral dimensions will
vary along with longitudinal elongations e xx consequently, in ad- \
dition to primary displacements u along the axis Ox there will

appear secondary transverse displacements v and w, and the

phenomenon will become much more complicated. When it is a
question of a thin bar, however, transverse strains may be neg-
lected without introducing serious error, and that will make it
easy to derive the differential equation of longitudinal vibrations
of such a bar.
Let us isolate (Fig. 35 b] from the bar an element of length
dx and replace the action of the removed parts by the forces
where F is the cross-sectional area of the bar. The mass of the

isolated element is pf dx and the equation of its motion is
pF dx -^ =
d 2u
-~
dX
dxF, (4.25)
or
r
~~ '
dt 2 dx
Since we assume that the bar is subjected to simple extension or

compression only, in accordance with Hooke's law, we have
xx
'x Ee
^^xx E
*-*
fix
and, consequently,
dXx ~~ F
L _ d 2u
*
dx dx 2
Substituting this expression in equation (4.25), we get

^L = u cC~
d 2"
,
' (4 26}
{t.AV)
()f2 \
fl x2
where
?
= f- (4-27)
Equation(4.26) differs from equation (4.8) for an unbounded

medium only by the constant factor [cf. formulas (4.9) and (4.27)];
the difference is due to the fact that in the case of an unbound-
ed medium we assumed that there were no transverse displace-
ments v and w and respective strains, and this, in turn, sets
up transverse stresses Y v and Z 2 in the case of the bar, how-
;
ever, these stresses do not occur. Equation (4.27) makes it possible

to calculate the modulus of elasticity of the bar E if its density
p and the velocity of propagation of sound in it a\ are known.
1
This equality coincides, obviously, in the problem under consideration with
the first ofNavier's equations (I).
Sec. 28] LONGITUDINAL VIBRATIONS OF A BAR HI
The general solution of the differential equation (4.26) has the

previous form (4.18) but in solving specific problems one most
!
frequently uses integration of equation (4.26) in Fourier series.

For this purpose we write the equation in the general form:
d 2n
-3?r
=a 3F-
2
u ,
A OR
(4-28)
.
where a equals a, 6, i depending on the character of the problem

being solved. Let
us attempt to find a particular solution of
equation (4.28) in the form
a = XT, (4.29)
where X is a function of one variable x, T is a function of one

variable /.
Substituting this in equation (4.28), we find
d 2X
Xv d2T
,,,
dt 2
= a9T 2
/
,
dx 92
or, separating the variables,
a2 T dt 2
~ X dx 2
'
Vu;
The left-hand member of this equation depends only on t, while
the right-hand member depends only on x\ since both of these
variables are independent, equation (4.30) can be satisfied identi-
cally only in the case if its left-hand and right-hand members
are equal separately to one and the same constant number; desig-
2
nate it by A, then equation (4.30) is decomposed into two or-
;
dinary differential equations
'
dt 2 dx 2
Their general solutions are
T = A cos Xa/ B sin -f- Xa/,
X = C cos D sin X;c -f- Xjc.
Substituting this in equation (4.29), we obtain a particular solu-

tion of equation (4.28):
u = (A cos Xctf -f B sin Xotf) (C cos X* +D sin Xjc). (4.31)
Varying A, B, C, D, A, here, we obtain any number of particular

solutions. Since equation (4.28) is linear, the sum of all such par-
It was derived by D'Alembert for a string, the equation of transverse vi-

1
bration of which has the same form (4.8}.

ticular solutions also satisfies it and we get its solution in the

form of a series
u = Z(A cos Xotf +B sin Xotf) (C cos \x + D sin Xjc); (4.32)
A, B
C, D, A, may have their particular values in each term of
y
this sum; the number of terms of the series is not limited.

The indefinite constants entering here should be sought from
the boundary and initial conditions of the bar. This procedure
will now be illustrated by a simple example. Let us consider a
bar of length / with the lower end built in, performing longitu-
dinal vibrations. The fastening condition at the lower end is:
#=*0 at any t and at x = 0. To satisfy this condition it is obvio-
usly necessary to set in equation (4.32)
C = 0;
then
u = Z(A cos Xctf -f B sin Xorf) sin X* '.
(4.33)
Let us suppose that the upper end of the bar is free and the elon-
gation is therefore equal to ** = =
at any instant
for x=l
-^~
of time t.
Keeping this in mind, we obtain cos>J=0 from equa-
tion (4.33). Hence
M = -g"
,
t
in
or
.
\ = jj-,in
where /
may be equal to any odd number: i=l, 3, 5, . . .
,
oo and,
therefore,
00
a=
2 (A- cos -g- a/ + .
sin
-g-a/)
sin -- jc.
(4.34)
To determine the arbitrary constants A and B it it is neces- l
sary specify the

to state of the bar at the initial moment, for
example, at i = 0, i.e., to indicate displacements u and velocities

f> it
-n- of all points of the bar at this instant; let
(4.35)
We can obviously omit the constant D by setting it equal to unity, with-

1
out loss of generality of the solution, since by introducing it in parentheses

we again obtain only the two con/tants AD and BD.
Sec. 28] LONGITUDINAL VIBRATIONS OF A BAR 113
But
,; cos sin -*. (4.36)

7 = 1,3, 5, ...
On the basis of equations (4.34) and (4.36), conditions (4.35)

give
// V^ A
f( x >=
x
A, A- sin
(4.37)
The problem has been reduced to the expansion of the func-

tions f(x)and CP(A;) into Fourier series appearing in the right-
hand members of equations (4.37), i.e., to the determination of
P-ffh 28m 35s S- 11h 32m 24s
Fig. 36
coefficients Ai and Bi in these series. These coefficients, as is

known, are expressed by Euler's formulas and have the following
values in this case:
sin
-i
(4.38)
-I
The foregoing method is very effectual in the case of free vi-

brations of bars. Investigation of forced vibrations, as well as
transverse vibrations of bars subjected to a periodically varying
or moving load, makes the problem far more complicated.
The general theory of vibrations of an elastic body is widely
used in seismology to study vibrations of the earthcrust; without
going deep into question, we shall note only that the ob-
this
served constancy of the velocities of longitudinal and transverse
vibrations makes it possible to determine from the record of
8-1013
vibrations at the seismological station (seismogram) the distance

between the epicentre and the station; transverse vibrations, as
we have seen, travel more slowly than longitudinal vibrations
and reach the station later; Fig. 36 represents a section of the
seismogram of the earthquake in Asia Minor on February 9, 1909,
recorded by the Pulkovo Observatory; here the points P and S
show the moments of the arrival of the first longitudinal and the
first transverse waves. Knowing the interval of time PS and the
velocities of both kinds of waves, we can find the required dis-
tance. Some additional information on the propagation of elastic
waves is given in Sections 56 and 57 of Chapter IX.
V
Solution
of the Elasticity Problem
in Terms of Stresses
29. THE SIMPLEST PROBLEMS

It was pointed out in Sec. 24 that if we wish directly to find
stresses in an elastic body according to prescribed external forces
V
-A^-,
V
/ Z7z Vz
/ ^7X V /t^1\
("*/
y, , , t
-Ay,
the conditions of equilibrium (I)
dx
dx '
dy
'
^ ' v' '
1
W
dZ x t
dZ y ,
dZ z
dx
will be inadequate for this purpose, since the number of un-
knowns exceeds the number of equations (I); it will, there-
(5.1)
fore, be necessary to resort to compatibility conditions (IV).
With that end in view we shall further transform equations (IV),
substituting stresses for strains [in accordance with equations (V)
of Sec. 24].
Let us now consider the particular case when stresses (5.1) are
expressed by functions of the first degree (linear functions) of the
co-ordinates of a point or when stresses are constant. By proceed-
ing from equations (V), we can easily show that the second de-
rivatives of strains are always expressed by linear functions of
the second derivatives of stresses (5.1), for instance,
d*Z
dy
2
~ d*Yy
+ *MI
.
dy* I J
'
etc.
11) SOLUTION OF ELASTICITY PROBLEM IN TERMS OF STRESSES [Ch. V
Since the stresses are linear functions of x, y, z in our case,

all the second derivatives of strains vanish; the compatibility
conditions (IV) will thus all be satisfied. It remains only to
satisfy equations (I) and the conditions on the surface of a
body (II):
Xv = Xx cos (vx) + Xy cos (vy) + X 2 cos (vz)\
'
Y v = Y x cos (vx) + Y y cos (vy) +Y z cos (vz)\ (II)
Zv Zx cos (vx) + Z cos (vy) + Z

y z cos (vz). .
Problems of this kind are called the simplest problems of the

theory of elasticity. We shallnow consider three such problems.
30. TORSION OF A CIRCULAR BAR

This problem deals with the state of stress and deformation
of a bar having the form of a circular cylinder (Fig. 37) whose
Fig. 37
bases are acted upon by external forces that produce only shearing
stresses and reduce *to two opposite couples; the moment of such
a couple is called the twisting moment (torque). According to the
theory of torsion elaborated by C. A. Coulomb at the end of the
18th century, the deformation of a bar consists in rotations of
the plane cross sections of the bar with respect to one another,
involving no distortion (warping) of cross sections; at all points
of the cross section there occur only shearing stresses T z directed
normally to radius-vectors of the points. These shearing stresses
are assumed to be distributed similarly over all sections (includ-
ing the ends of the cylinder); the system of forces acting at the
Sec. 30] TORSION OF A CIRCULAR BAR 117
section reduces to the twisting moment
M = fT rdF
z z (5.2>
F
where dF is an element of area isolated in the section; the inte-
gral is extended over the entire cross-sectional area. Let t denote

the angle of mutual rotation of two sections a distance equal to
unity apart; from elementary geometrical considerations it is easy
to find now the shear angle which occurs at a point of the section
distant r from the axis of the section and which corresponds to
the mutual angle of rotation t:
Hence on the basis of Hooke's law in shear we obtain the magni-

tude of shearing stress Tz at the given point
(5.3)
Its projections on the x and y axes are (Fig. 37)
X=
z
T2 sina. = T2 2r ; Y 2 = Tz cos a = T z ~
or, according to (5.3),
Assuming that the remaining stress components (5.1) vanish,

we arrive at the following system of stresses:
= 0; \
= (5>4)
Let us verify whether this system of stresses is possible from

the viewpoint of the theory of elasticity. Since stresses (5.4) are
functions of not higher than the first degree in the co-ordinates
of a point, the problem under consideration is one of the simplest;
it will, therefore, be necessary only to satisfy equations (I) and
to check, as required by the surface conditions (II), whether
stresses (5.4) correspond to the conditions of the torsion problem.
Substituting the data from equations (5.4) in equations (I), we
find
* = 0; K"0; Z = 0,
consequently, stresses (5.4) are possible provided body forces are
absent (for instance, the own weight of a bar).
Let us now turn to the conditions on the surface of the bar.
It can easily be seen (Fig. 37) that everywhere on the lateral
118 SOLUTION OF ELASTICITY PROBLEM IN TERMS OF STRESSES [Ch. V
surface cos (vz) =0;

cos (vx) = cos a = ~ ; cos (vy) = sin a = y
- .
Substituting these values and those from equations (5.4) in equa-

tions (II), we find that
Xv = 0\ K w = 0; Zv = 0.
i.e., the lateral surface is free of load.
At the end cross section we have
cos (vx) cos (vy) = 0,
cos (vz) 1 .
Introducing these values and those from (54) in (II), we have
< 5 5>
'
Consequently, only tangential tractions are applied here (as in

any other cross section). Their resultant has the following projec-
tions on the axes Ox and Oy:
ff
X dF=
g
ff
since the origin of co-ordinates is placed at the centroid of the
section. Consequently, the forces applied to the end section reduce
to a couple; we thus really have the problem of the torsion of
a bar.
The moment of the torsional couple about the axis Oz is
M, = / / (X y - Y x)dF = - Gt / f (x* + y dF = - G*JP (5.6)

z z
2
) ,
where / = -g
TlR 4
p the polar moment of inertia of a circular sec-
is
tion.
31. SAINT-VENANT'S PRINCIPLE
It is important to make the following remark. We have shown

that the system of stresses (5.4) corresponds to the torsion of a
circular bar; we have no right, however, to draw the converse con-
clusion that the torsion of a circular bar always results in the sys-
tem of stresses (5.4).
Indeed, when formulating the problem of torsion, we indicate
only that there is a couple applied to the end cross section; this
Sec. 31] SAINT-VENANT'S PRINCIPLE 119'
indication, however, is inadequate because the couple can be

applied in a number of different ways, i.e., the stresses reducing
to the couple can be distributed differently over the points of
the section. The system of stresses (5.4) results when the torsional
couple applied in the specific manner indicated by formulas
is
(5.5); the torsional couple is applied in a different manner,

if
another system of stresses will be set up in the bar.

These considerations also apply to many other problems; they
show that we often formulate problems schematically, sometimes
prescribing not the forces applied to a body but only their result-
ants (for instance, "compression of a bar by two forces" or "bend-
ing by a couple", etc.). Each of these formulations leads to a
variety of quite different problems when a more exact method of
solution is used.
Fig. 38
Saint-Venant, however, propounded a very important general'

principle indicating that there is much in common in the infi-
nite set of problems covered by such a general schematic formula-
tion; this principle states that at points in the solid which are at
a sufficient distance from the surfaces of application of external
loads, stresses depend very slightly on the particular manner in
which these loads are applied-, for instance, in the case of the
problem of torsion just solved Saint-Venant's principle states that
at points in the bar sufficiently far remote from the end cross
sections the stresses set up by twisting depend to a very little
extent on the manner in which torques are applied and vary
slightly when this manner is changed. But at the ends of the bar
the mode of application of torques affects the nature and magni-
tude of stresses substantially.
Saint-Venant's principle results from the following postulate
which is now sufficiently well proved both in the solution of
many specific problems and in the general form.
a balanced system of forces is applied to any part of a body
//
A it will induce stresses in this body which diminish
(Fig. 38),
very rapidly with the distance from the part A.
Fig. 38 illustrates this postulate in the particular case of a bar
which is compressed by pliers producing a balanced system of.
forces. However
great these forces may be (e.g., in cutting
a with pliers), they will obviously cause
wire no appreciable
stresses outside the small region A outlined by the dotted
line.
This postulate being accepted, we can pass on to Saint-Venant's
principle formulated above by means of a simple static transfor-
mation. Our reasoning will again be illustrated by a concrete
example (Fig. 39 a, b) of a rod bent in one case by a weight P
suspended from the end of the rod from below, and in the other
by the same force but pressing from above.
It is evident that the stresses in the region A will be essentially
different in these two cases, but it can easily be seen that the
-difference will be insignificant outside the region A.
: / :
A ,
Fig. 39
Indeed, let us consider an auxiliary third case (Fig. 39c), add-

ing two equal and opposite forces P and P 2 to case a, the three
{
forces being the same in modulus: Pi = P 2 = P.

We observe that case c differs
from case a by the balanced system of forces P it P2
from case b by the balanced system of forces P, P2 .
These balanced systems, however, will cause no appreciable

stresses outside the small region A, in accordance with the above-
stated principle; consequently, case c will give a very slight differ-
ence in stresses outside this region in comparison with cases a
and ft; thus, cases a and b will also be almost identical
with regard to stresses outside the region A, which was to be
proved.
It should be pointed out that the objective of the foregoing
considerations is to give a general idea of Saint-Venant's prin-
ciple in view of its exceptional importance for the statement of
problems of the theory of elasticity and for their solution. How-
ever, the above schematic formulation of the postulate concerning
Sec. 31] SAINT-VENANT'S PRINCIPLE 121
a balanced system of forces applied to a part of a body may lead

to erroneous application of Saint-Venant's principle, unless we
introduce an essential amendment; taking this into consider-
ation, we shall give a more complete formulation of the condi-
tions in which Saint-Venant's principle is valid.
Let us take on the surface of a body a small region whose
dimensions are not great compared with the overall dimensions of
the body, let a balanced load (i.e., a load whose resultant and
moment are equal to zero) be distributed over this region; Saint-
Venant's principle states that under the action of such a load
stresses will be appreciable only in the part of the body in the
immediate vicinity of the loaded region; they diminish so quickly

that at points whose distance from the loaded region exceeds the
greatest dimension of this region they can practically be neg-
lected.
It is essential in this formulation that the loaded region be
small in comparison with the overall dimensions of a body; in the
limiting case it may be of the same order as the smallest dimen-
sion of a body. If this requirement is not fulfilled, the conditions
in which Saint-Venant's principle is valid are not satisfied and
it cannot be applied. This consideration must be kept in mind in
the case of thin shells and the so-called thin-walled bars, such
as bars of I-section (Fig. 40ft). Here the dimensions characterising
the shape of a body may differ widely; for instance, the overall
dimensions AB and BC of the cross section may be small as com-
pared with the length of the bar AE, but considerably greater
than the web thickness 6. Take two bars (Fig. 40) built into a
wall in the same manner, one of them with a rectangular sec-
tion and the other with an I-section, the dimensions AB, BC and
AE being equal in both cases. Let identical balanced systems of

forces consisting of two couples with equal but opposite moments
+M and M
be applied to the free end of each bar. If the cross-
sectional dimensions AB and BC are small compared to the length
of the bar AE, then, according to Saint-Venant's principle, we
can stale that in the case represented in Fig. 40a the stresses
developed in the bar will be practically negligible except for the
small portion adjacent to the free end; the length of this portion
will be of the same order as the cross-sectional dimensions AB
and BC.
In the case represented in Fig. 406 this cannot be affirmed,
since the web thickness 6 of the cross section is small in compar-
ison with the dimensions of the loaded region ABCD. This is
evident from the following considerations: let us imagine that the
web thickness 6 is made to approach zero; in the limit the web
will disappear and the bar will be converted into two separate
strips (flanges of the I-bar), each loaded by an unbalanced system
of forces, i.e., by a couple; each strip will undergo pure bend-
ing to be considered in the next section; the stresses will be the
same in all cross sections of each strip and will extend up to its
fixed end, no matter how great its length may be. The web thick-
ness 6 is not in reality zero and there will be no such sharp
-effect; however, the smaller the 6, the greater the region of exten-
sion of stresses along the length of the bar. In spite of this, it
cannot be said that in the case represented in Fig. 406 Saint-
Venant's principle is not true: there are simply no conditions
here for the principle to be applicable. Such conditions will be
fulfilled, if we draw the couples +M and M nearer to each
other and apply them to the web of the I-bar at some point K so
that the distance between the planes of the couples does not
exceed 6, the smallest dimension of the bar. Then Saint-Venant's
principle will be valid and it may be stated that the stresses will
be negligible everywhere except for a small region in the neigh-
bourhood of the point /(; the dimensions of this region are of the
same order as the web thickness 6.
32. THE PROBLEM OF TORSION OF A CIRCULAR BAR

(CONTINUED)
Let us finish the problem of the torsion of a circular bar and
find the displacements of its points.
Introducing stresses (5.4) in Hooke's law (V) (Sec. 24), we
obtain strains; we then replace the strains by their expressions
in terms of displacements in accordance with equations (III)
Sec. 32] THE PROBLEM OF TORSION OF CIRCULAR BAR
(Sec. 24) and obtain the following system of equations:
du n dv n dw n
~T~~ \Jy
"
-x U V U (5.7)
dw
w
da
dv
dz
dw
(5.8)
dz
Integrating equations (5.7), we have

u=f(y, z), v = , z),
w= ty(x, y),
(5.9)
where /, cp, are arbitrary functions; their form

\|)
will be deter-
mined from equations (5.8) which, on the basis of (5.9), become
dy
dz (5.10)
Let us attempt to eliminate functions /, cp, \f> in succession from

these equations by means of differentiation. This process can be
carried out systematically; the succession of the required compu-
tations will be given in detail as it will be employed in the same
form in other problems treated below in Sections 33 and 34. Let
us differentiate equations (5.10) successively, as indicated in the
first, second and third columns to the left of these equations; we
obtain the following nine equations:
(5.11)
dy dz
n^ dx dy
124 SOLUTION OF ELASTICITY PROBLEM IN TERMS OF STRESSES (Ch. V
We add up the last three of these equations by members thrice,

changing the signs each time, as indicated in three columns to
the left; we obtain the following three equalities:
- - .
(5.12)
v '
dydz dxdy
Thus, equations (5.11) and (5.12) give the following table of the
.second derivatives of functions (5.9):
() y
dx*
'
dxdz
= T .' y
2
==n .
(5.13)
The object of the preceding computations was to obtain pre-

cisely this, simpler system of equations in place of (5.10); based
on it, the general form of functions /, can easily be
established now; in fact, the last three equations show that the
function \|> will be of the first degree in x and y\ the first and
second lines show that / and cp must be of the second degree (in
*/, z and in x, 2, respectively) but must not contain the squares
of the variables; therefore, on the basis of equations (5.13) and
(5.9) we can easily write
(5.14)
*
= g* + hy + k. )
Functions (5.14) are the general solution of equations (5.13);

these latter have been obtained as a consequence of equations
(5.10) by means of differentiation. Their order, therefore, is higher
than the order of the latter; thus, it is not every solution of
equations (5.13) which will satisfy equations (5.10). For instance,
by differentiating the equation
\ve obtain
The general solution of the latter equation

Sec. 32] THE PROBLEM OF TORSION OF CIRCULAR BAR 125
will satisfy the first equation only if Ci = 0. It is, therefore,

necessary to ascertain which of the solutions (5.14) satisfy equa-
tions (5.10).
Substituting functions (5.14) in equations (5.10), we establish
the following relationships:
h -f- ix +- e = ix,
or
and finally we have
uf=
v = y = txz ax^-ez-\-f, I
(5.15)
w = = bx
The obtained expressions of displacements (5.15) involve six
arbitrary constants whose presence here can easily be explained.
Indeed, in Sec. 10 we agreed to fix the elastic body under inves-
tigation in order to eliminate its motion dealt with in mechanics
of absolutely rigid bodies and defined above, at the end of Sec. 10,
as a rigid body displacement; in the present problem, however,
we have not yet realised such fixing; the linear trinomials in
formulas (5.15) indicate that it is still possible for the rod to
rotate as a whole about the co-ordinate axes through arbitrary
small angles e, 6, a and to have translatory displacements
c, /,k along these axes [cf. formulas (2.11)].
Let the fixing conditions be realised at the upper end of the
bar: fix the centroid of the section 2 = 0, requiring that
for x=y=zQ
This condition, however, is inadequate as, in spite of it, the bar
can rotate about the origin (Fig. 41); in order to eliminate the
possibility of rotation let us require that two of three elements
dx, dy and dz near the origin of co-ordinates O remain immov-
able. In Sec. 10 [see formulas (2.3) and (2.4) and matrix (2.8)]
we already obtained the expressions of the angles of rotation of
elementary segments; by using them, we can write down the
required conditions.
(a) The segment dz is immovable

for x = y = z = 0,
(b) The segment dy is immovable in the plane Oxy

for x=--y =z= 0, = 0. (5.18)
It is easy to conceive that conditions (5.16), (5.17) and (5.18)

are sufficient to prevent a rigid body displacement of the bar.
Fig. 41
Applying these conditions to equations (5.15), we obtain

a b c = = = e=f=k = 0\
and therefore
u =
v= (5.19)
w= 0.
The last equation supports the hypothesis of plane cross sec-

tions of a circular bar remaining plane in torsion.
33. PURE BENDING OF A PRISMATICAL BAR

We consider a bar of prismatical cross section (Fig. 42) with
equal but opposite couples applied to its ends. Let the z axis be
directed along the axis of the bar; the xz plane coincides with
the plane of action of the applied couples. This case is commonly
referred to as pure bending; its elementary theory was elaborated
by J. Bernoulli and L. Euler in the 18th century; it is based on
1
Thus all components of a rigid body displacement in formulas (2.11)
vanish.
Sec 33] PURE BENDING OF PRISMATICAL BAR 127
the assumption that the axis of the bar OB will bend along a curve
lying in the xz plane and that plane cross sections of the bar
will remain plane and normal to the deflected axis. From simple
geometrical considerations (as set forth in the courses on strength
of materials) it may be concluded that longitudinal "fibres" of
the bar parallel to its axis receive unit elongations proportional
to their distances from the yz plane containing the axis of the
bar, which is called the neutral plane
where the radius of curvature of the deflected axis of the bar.

p is
It is assumed
that all components of the stress tensor are zero
except for the component Z 2 causing simple tension or compres-
sion of the fibres of the bar; then, in accordance with Hooke's
law we find that
Ex
where E is Young's modulus. On the basis of the above we arrive

at the following system of stresses:
Xx = Y, = X = Y = Z x = ^
y z Z2 = .
(5.20)
Let us see whether the system of stresses (5.20) is possible

from the viewpoint of the theory of elasticity and whether these
stresses correspond to pure bending.
substituting functions (5.20) in the equations of equilibrium
By
(I), we
see that they are satisfied in the absence of body forces.
The equations of continuity (IV) (Sec. 24) will be satisfied. As
regards the conditions on the lateral surface of the bar we have
everywhere
cos (vz) = 0.
Substituting this and (5.20) in equations (II), we find that the

lateral surface is free of stresses.
The forces on the end section B reduce to a couple with the !
moment
1
The reader is recommended to confirm for himself that the resultant oi
forces Z z dF is equal to zero.
128 SOLUTION OF ELASTICITY PROBLEM IN TERMS OF STRESSES [Ch V
Consequently, stresses (5.20) are possible and correspond to pure

bending We proceed to the determination of displacements.
1
.
Substituting stresses (5.20) in Hooke's law (V) and the strains

thus obtained in equations (III), we arrive at the system of equa-
tions:
du ax dv sx dw x_ m
'
t
'
~~~ (^ 9.]}
dx ~~jT "dy" p dz p
dw i
dv ~. du . dw UV I uli r\
'
dx dx (5.22)
dv '
dz dz ' '
dy
They are integrated in the same manner as in the preceding

section. From equations (5.21) we find
V = -
+ cp (X, Z), (5.23)
The form of functions /, cp and \|)

is determined by the use of
equations (5.22):
dz
' (5.24)
dz '
dx p
dy df <iy
"dx '
~3y~ p
These equations have the same form as equations (5.10) in the

problem of torsion; consequently, to determine
functions /, cp, tp
we can apply the procedure given above; in this way we find the
second derivatives:
= 0;
dy
2 dz 2 dy dz
dz* dxdz -0;
dx 2 dy
2
dxdy
= 0.
With 'these expressions we readily determine the form of func-

tions /, cp, x|> and, substituting the result in equations (5.23), we
1
It is to be noted that the foregoing theory is valid only when the plane xz
of action of the couples is a principal plane of the bar, i.e., it contains one of
the principal axes of inertia of each section of the b0"
Sec. 33] PURE BENDING OF PRISMATICAL BAR 129
obtain these in the following form:
(5.25)
= xz ,
bx ey +, ,
k.
The linear trinomials included here represent the rigid body dis-
placement of the bar, just as in the problem of torsion. To deter-
mine them let us fix the bar, for example, at its left-hand end
JZ?
Fig. 42
.(Fig. 42). In the first instance we fix the centroid of the section,
requiring that
for x=y =z=
on this basis from equations (5.25) we find
We fix further the element dz of the bar axis eliminating its rota-
tions in the planes Oxy and Oyz, i.e., requiring that
for x=y=z=Q = 0, --^-

= 0.
Finally, it remains to eliminate the rotation of the bar about

its axis; for that we fix, for instance, the element dx of the axis
Ox to prevent it from rotating in the plane Oxy, i.e., we require
that
for ,c = y = *-0 4^ = 0.
91013
From all these conditions we find that a =b=e=Q and we obtain

the final expressions for displacements:
(5.26)
w=
Each point (x\, t/i, of the bar moves to a new position after
deformation:
z=z + w = z l l
-
r
For points of the axis of the bar (x y= ( we obtain
(5.27)
this is the equation of the deflected axis of the bar.

Let us take some plane section of the bar Z = Z Q The . new co-
ordinate z of any of its points after deformation will be
but in accordance with equations (5.26) we have
Consequently,
(5.28)
This the equation of a plane parallel to the axis Oy (Fig. 43);

is
it is by all points of the section 2=20 after deformation.
satisfied
Consequently, a plane cross section remains plane after deforma-
tion.
We rewrite equation (5.28) as
*= -4-* + p. (5.29)
Sec. 33] PURE BENDING OF PRISMATICAL BAR 131
The angular coefficient will be tan (J

= -. The slope of the
tangent to axis (5.27) at the point Z = ZQ is
tan a = / du \
(-du_
dz )z=z
,
zn
(5.30)
Thus we have (Fig. 43)

tan a tan = 1.
Consequently, the plane section z = z Q remains normal to the axis

after bending and the hypothesis of plane cross sections is fully
justified in the case of pure bending.
Fig. 44
Consider another particular case of a bar of rectangular cross

section (Fig. 44 a) and investigate the distortion of the contour
of its cross section. This question is of interest for what follows.
Its lateral side before bending is defined by the equations
*=z ;
y=Y;
after deformation these equations become
There result two equations of the first degree in x, y, z; conse-

quently, the lateral sides remain straight after bending. From
the equation
we see that (Fig. 44 b)

,
tan =-. (5.31)

where \ is the angle between the projections of the original and

deformed positions of the lateral side on the plane Oxy. Now con-
sider the upper and lower sides given by the equations
After bending, their equations will be
~2+u =
h h
x= , . ,
TT +
,
The first of these equations is of the second degree, while the

second equation is of the first degree in x, y, z. Consequently,
the upper and lower sides become plane, curves. The first equation
gives the projection of these curves on the plane Oxy:
2
^ a/.
(5.32)
This the equation of two parabolas (the upper and lower sides).
is
The radius of curvature of the deflected axis (5.27) and the ra-
dius of curvature of parabolas (5.32) are approximately equal to
pand-, respectively.
Equations and (5.32) are approximate, since we have

(5.27)
disregarded small quantities of higher order in the expres-
finite
sions of strains [cf. Sec. 10, formulas (2.3) and (2.4)]. A more
exact solution shows curves (5.27) and (5.32) to be circles with
radii
p and .
The distorted form of the section contour is represented in

Fig. 44 b. The upper broadened part corresponds to compressed
longitudinal fibres, the lower part to stretched ones.
34. PRISM STRETCHED BY ITS OWN WEIGHT

Let us have (Fig. 45 a) a prism or a cylinder of arbitrary cross
section suspended from the upper end and stretched by its own
weight. It will immediately be seen that the following system of
stresses corresponds to the given problem:
Yz = Zx = X = ^ y (5.33)
Sec. 34] PRISM STRETCHED BY ITS OWN WEIGHT 133
Substituting these expressions in equations (I), we find that

X Y = Q, Z g", i.e., we actually have the problem involving
the action of the own weight. Furthermore, at 2 = we get
i.e., the lower end

is free of load; introducing values (5.33) in
equations (II), we
see that the lateral surface of the prism is also
free of lo^d. Displacements are sought in exactly the same way
as in the two previous problems.
Fig. 45
On the basis of (III) and (V) we obtain the system of differen-

tial equations
du ap^ dv ap-
*'
'dx E dy
(5.34)
~~ *'
dz E
da ,
dv ~ dv ,
dw ^
' "~
dy dx dz dy
'
(5.35)
By integrating equations (5.34), we get
u = -^-zx + f(y, 2),
, z), (5.36)
=^-- + ^(x, y).

Introducing this in (5.35), we obtain the equations for the deter-

mination of /, q> and ij?:
'
dy dx
dz (5.37)
:
+4r=o.
The method of determining them remains similar to that used
in Sections 32 and 33 and, therefore, without repeating it, we
present the final result:
GO or
v ~- zy ax + ez + /rc, (5.38)
w= -ij;\z + c(x +y 2 2 2
)]
bx ey + k.
The fixing conditions of the prism are realised in the same

cvay as in the two preceding problems, but according to the con-
ditions of thisproblem we fix the centroid of the upper section
ather than the co-ordinate origin; we require, therefore, that
for x= Q, y = 0, z =
Jiere must be
dv
dz Jx~
Hence we shall easily find that a = b = c = e = m = 2

|~/ and
displacements (5.38) will finally take the form
u
// -.- .
OOP-
'o
-p-
yy
xz, v
fj\
>o
(JpjOr
-p- yz,
w
(5.39)
The points on the axis of the bar (itsequations: x = Q, y = 0)'

lispl-ace in a vertical direction (a
= 0; v=0); all other points have
lorizontal displacements too.
du
1
One may require instead that = 0.
Sec. 34] PRISM STRETCHED BY ITS OWN WEIGHT 135
If take a point Jt f/o, o n ^ ie unstrained state,

we ,
' its co-ordi-
nates, as a result of deformation, will be
1
(1
_
a P^
ZQ
^
j
,
(5.40)
Let us use them to investigate the character of deformation of

the prism.
1. Consider its cross section
z = z = const.
Q
By squaring the first two of equations (5.40) and adding them

up,we obtain
or
ao/r \2
2
where r2 = x +y
2 2
,
rg=x + yg; thus each circle of radius r at the
cross section with centre on the axis of the prism is transformed
into a circle of radius r with r<r The surface into which the .
section z = z is transformed will be found by eliminating the

initial co-ordinates X Q and r/ from (5.40); taking into account
(5.41), we obtain from the third equation of (5.40)
z z -
-1-.UL f z 2 P -\ - - r2 \
This equation gives a paraboloid of revolution in cylindrical

co-ordinates r, <p since the polar angle cp is not involved here.
2,
2. Imagine a circular cylindrical surface of radius r isolated
in the prism
*g+y = r. (5.43)
As a result of deformation it will be transformed into a surface

of revolution; this could be predicted on the basis of the above
considerations concerning the deformation of circles described by
equation (5.41). To obtain the equation of this surface it is
necessary to eliminate the initial co-ordinates # yo, ZQ from four ,
equations (5.40) and (5.43); the first two of equations (5.40) t

however, can be replaced by equation (5.41):

o it1 QQr&
r
-y
I
/-2
B o
% _L
-Z ^ _. Iz _ ^ôr
2 I
2
o
/
*
2
-I- a ( 2 -4-V 2M
iôJJ
r 2o*
Hence we eliminate x +y at once; there remain two equations
r _(i
r=r n --_2r~ P \
j/-
r0>
2
z ZA _J_ IfL
^01^ 2
f
\
Zo2 _ 72 I
i
0/-2\
or
oj'
(5.44)
To obtain the the required surface it remains to

equation of
eliminate z from them (or to treat them as parametric equations
of a surface with the parameter Z Q these equations give the gen- \
erator of our surface in co-ordinates r, z). The first of equations

(5.44) gives
Substituting this in the second equation, we get
This is a surface obtained by rotating a parabola about the z

axis.
35. UNIQUENESS OF SOLUTION OF ELASTICITY

EQUATIONS
In the solution of the preceding problems we used the inverse
method prescribing stresses and determining what forces acting
on the surface cause the selected system of stresses; then each
time the question may arise whether or not it is possible to obtain
the same forces on the surface for some other system of stresses.
If so, the solution of equations of the theory of elasticity will be
multiple-valued: several systems of stresses will correspond to

given forces on the surface and it will be necessary to clarify
which of the systems actually take place. In this case, using
either the inverse or semi-inverse method of solution, we shall
not be certain that we have chosen the system of stresses which
corresponds to reality. Accordingly, the question of uniqueness
of solution of elasticityequations acquires great importance.
It is to be noted that if initial stresses are possible in a
body,
then under the action of external forces on this body the total
stresses developed in it may differ very widely; take, for instance,
the iron sphere with the initial stresses spoken of in Sec. 2U
Sec. 35] UNIQUENESS OF SOLUTION OF ELASTICITY EQUATIONS 137
If we subject it to the action of forces, the final overall stresses

in it depend on the dimension of the sectorial cut-out which
will
has been previously made. In cases like these, the equations of
the theory of elasticity must have a multiple-valued solution.
Let us omit the case of initial stresses, i.e., accept the hypoth-
esis of the natural state of a body, and also accept the principle
of superposition. It is easy to prove, then, that the solution
will be single-valued and, therefore, unique.
Indeed, let us assume the converse, namely, that under the
action of given surface tractions
v , Y Zv (5.46)
and body forces
X, Y, Z (5.47)
two different systems of stresses are possible:
X'x ,
Y'y , Z;, Y' Z'x , X'y (5.480
and
x\ V2
* ^Y y (5.48")
Both of .these systems must satisfy basic equations (I) and (II):
dxx dx' dX'z
~~dx~
dYi dY'z
(5.490
~JT ~dz~
dZ' dZ' dZ'
dx ^ dy
"^ dz -i-^ ^,
X = X'x cos
v (vx) + Xy cos (vy) + X' cos z (vz),
Yv = Y'x cos (vx) + Yy cos (vy) + Y' cos z (1^2:), (5.500

Zv = Zx cos + ^y cos (^y) + Z cos
(T;A:) z (^2:);
(?A: dy
<)K
_ar; (5.49")
~5F
az: dZ"
dAT
Xv + Xy cos (vy) + X" cos

X"x cos (vx) z (vz),
Y = Y" cos (vx) + Y" cos (vy) + cos
v x y Y", (vz), (5.50")
Zv = Zx cos (vx) Zy cos (vy) + Z cos

-\- z (ue).
By subtracting the respective equations of systems (5.49) and

(5.50) by members, we obtain the following new system of equa-
tions:
dx dy dz
dx dy dz
= 0,
dy
= (X X"x] cos ^) cos (vy) + (5.51)
=(Z'X Z"X } cos (vx) + (Zy Zy) cos (vy) +
Proceeding from the principle of superposition, we can take the

differences of stresses entering in equations (5.51) as a new sys-
tem of stresses. Equations (5.51) show, however, that these
stresses exist in the absence of surface and body forces and must
therefore, be all zero on the basis of the hypothesis of the natural
state of a body, i.e.,
Xx = X =X
y y , etc.
7
Hence, both systems of stresses (5.48 ) and (5.48") are coinci-
dent, which was to be proved.
If the principle of superposition does not hold in a given
problem, we again encounter a multiple-valued solution. This
can be illustrated by Euler's problem of the buckling of a bar.
Indeed, let us take two loads, PI and P 2 each of which is slightly
,
less than the critical P CT When subjected to each of the loads

.
Pi and P 2 separately the bar undergoes simple compression. Un-

der the action of the sum of loads Pi + P 2 however, colossal,
bending stresses are added to the total compressive stress; the

principle of superposition is not, therefore, applicable here and
the action of the sum of the forces is not equal to the sum of their
separate actions. At the same time the solution appears to be
multiple-valued, since several configurations of equilibrium are
Sec. 36] BELTRAMI-MICHELL EQUATIONS 139
possible with a particular system of stresses corresponding to

each of them.
The question of uniqueness of solution will be discussed in
more general form in Sec. 37.
36. BELTRAMI-MICHELL EQUATIONS

Let us now
turn to the general case of solving the elasticity
problem terms of stresses. It has already been pointed out
in
more than once (Sections 24 and 29) that if stresses are chosen
as basic unknowns, the number of these unknowns will be equal
to six:
Xx ,
Y y, Z z Yz Z X , , y
Xy. (5.52)
Hence three equations of equilibrium (I)
dX,
dx ^ OXy
dy
^ dX,
dz n-K"
dY x dYy dYz
~dT ~^~ ~d^~
~T~
~d~z
hp/ ~~ (I)
dZ x dZy dZ z
dx ' '
dz
:=o
dy
are inadequate, and to solve the problem one has also to make
appeal to the compatibility conditions (IV). In addition, the
conditions (II) must, of course, be satisfied on the surface:
Xv Xx cos (vx) + X y
cos (vy) +X z cos (vz)\
Yv = Yx cos (vx) + Yy cos (vy) ~\- Y z cos (vz)\
Zv = Zx cos (vx) + Z cos (vy) y -f- Z z cos (<vz).
Thus the problem is reduced to the integration of nine equations
(I) and (IV) with six unknown functions. Arbitrary functions
entering in the general solution of these equations must be deter-
mined from the surface conditions (II).
Since the conditions of continuity (IV) connect the components
of strain
we have to transform these conditions expressing strains in terms

of stresses (5.52) by means of Hooke's law (V). Carrying out
this substitution and simultaneously using the equations of equi-
librium (I), we transform equations (IV) to the following form
absence of body forces or these forces are constant

1
in the if
,
d*e . )
+ <
(VII)
+ dydz -0;
: +
0+' =-0.
Thus to solve the problem it will be necessary to integrate nine

equations (I) and (VII) and to satisfy the surface conditions (II).
E. Beltrami deduced equations (VII) in a somewhat different
manner, starting with Lame's equations (VI). The line of reason-
ing is then as follows. Let us suppose, as above, that there are
no body forces or that they are constant at all points of a body,
i.e.,
dX dY
= = dZ
=~ /r ^ ov
-^ =? -i . . .
-^ 0. (5.o3);
dx dy dz v
First of all we prove an auxiliary proposition that the signs of

Laplace's operator and partial derivative may be interchanged,
for instance,
_
dx 1
dx 2 \dx /
r
dy'
2
\ dx ,
dx
Thus, if t denotes any of variables x, y, z, then
"dt (5.54)
We prove further that under conditions (5.53) the dilatation

strain 6 satisfies Laplace's equation
V'9^0. (5.55)
For that we differentiate the first of equations (VI) with respect

to x, the second with respect to y and the third with respect to z;
1
This derivation is omitted here; it may be found in S. P. Timoshenko,
Theory of Elasticity, New York, McGraw-Hill Book Co., 1934.
Sec 36] BELTRAMI-MICHELL EQUATIONS 141
adding up the obtained results and using (5.53) and (5.54), we get
whence we arrive at equation (5.55).

Now by differentiating Lame's equations (VI), we shall derive
six relations between strains. Differentiating the first
differential
of equations (VI) with respect to x and using equations (5.53)
and (5.54), we obtain
(^ + ^5 + ^^ = 0. (5.56)
Hence by a cyclic change we obtain two more analogous rela-

tions:
(5.57)
We now differentiate the second of equations (VI) with respect

to z and the third with respect to y, and, adding up the results,
we get
v ,
dw
or
(5-58)
A cyclic change yields two mor.e relations:
(5.59)
In our case (when there are no body forces or they are constant)
six equations (5.56) -(5.59) connecting the second derivatives of
strains of an isotropic elastic body are equivalent to six Saint-
Venant's equations (VI) and can replace them. Now it remains
to convert equations (5.56) -(5.59) to stresses on the basis of
Hooke's law (V) and (V'a) we have
From (V) we obtain

-and further, using (5.55), we determine
Substituting this in (5.56), we find that
= 0. (5.60)
But (see Sec. 18)

1 ! 11 ^ 3A-f2(J. /K/M\
Keeping this in mind, we note that (5.60) coincides with the

first ofequations (VII). The second and third of equations (VII)
will be obtained from (5.61) by a cyclic change.
7
Let us turn to equations (5.58) -(5.59); since, according to (V ),
we have
e yz ~ -"
substituting this and from (V'a) in (5.58), we obtain
2<x+ rt 2
() e
"T"
2
u ~
3X + z '
2(x dy dz
which, according to (5.61), coincides with the fourth of equations

(VII); the other two will be obtained from this by a cyclic change.
J. H. Michell derived equations (VII) for the general case when
body forces are not constant. Then the relations become more
complicated and, for instance, the first of equations (VII) is
replaced by
v A -1-
1
~~ _1
d^__ /* dY , dZ\
The reader is recommended to derive these equations on his

own; for that he should repeat the derivation given in this section
but retain the body forces pA', p7, pZ in Lame's equations.
*37. THREE KINDS OF PROBLEMS OF THE THEORY

OF ELASTICITY. UNIQUENESS THEOREM
The Beltrami-Michell equations complete the system of equa-
tions of the theory of elasticity, permitting us to solve the nec-
essary problems in terms of displacements or stresses.
Up to this point we have assumed that there are prescribed
loads on the surface of an elastic body, as well as body forces.
The problem thus formulated will be called the first basic problem
Sec 37] THREE KINDS OF PROBLEMS OF THEORY OF ELASTICITY 143
of the theory of elasticity. In applied elasticity another case is

encountered when displacements u, v, w are prescribed on the
surface of an elastic body at all its points; let us define this
case as the second basic problem of the theory of elasticity, there
also occur stresses on the surface, together with the given dis-
placements of its points, but they are not known beforehand.
A mixed problem of the theory of elasticity is also possible when
displacements are prescribed on one part of the surface, and loads
(stresses) on the other; in all these problems there may also
exist body forces specified in advance.
The above three basic problems constitute a certain schematisa-
tion of real physical problems; the so-called contact problems
are closer to actual conditions; these problems deal with the
mutual pressure of two bodies: of two elastic bodies or of an
absolutely rigid body against an elastic body; one has to deal
with a contact problem, for instance, in the design of bearings
(simple and ball bearings), rollers and plates, movable supports
of trusses and beams as well as in the problems of pressure of
a punch on a plane surface of an elastic body.
Given in Sec. 35 was the proof of the uniqueness of solution
of the first basic problem of the theory of elasticity; we shall
now extend it to the second and mixed problems; the proof,
presented below, was given by G. Kirchhoff; it is based on the
properties of the work of forces producing deformation of an
elastic body.
Let
v X
Y v Z v be the projections of an external load applied
, ,
at the surface of a body, which are related to stresses near the

surface by equations (II); let a, v, w be the displacements of
the respective point of the surface. Let us set up the following
double integral extended over the surface of the body:
J = f(X v Z v w) dS.
(S)
For the subsequent derivation the physical meaning of this

integral need not be given, but it can easily be observed that
it represents the doubled work of external loads in the process
of deformation of a body if these loads increase very slowly
from the initial natural state of the body. This follows from
Clapeyron's theorem
l
.
Let us replace X v Y v Z v by their expressions in accordance

, ,
with formulas (II) and collect together the coefficients of the

cosines of the angles between the outward normal v and the
This theorem expressed below in the general form by equation

1
is (5.62).
co-ordinate axes. Our integral will then become

J = J [P cos (vx) + Q
cos (vy) + /? cos (vz)\ dS,
(S)
where
On
the basis of the Green-Ostrogradsky formula presented in
Sec. 3 (footnote on p. 25) this integral can be transformed into
an integral extended over the volume of the body t:
'-/(!? (t)
++)*
We calculate the derivatives here
dx
~
^_ dx
__
dx r- dx --
~T-X dx
-- Y
Kx
dx
--
~x dx
^^~^ tt + ^^~^^ + ^^+ ry^

dR dXz ,
dY2 . dZ z . du ,
dv . ^ dw
Let us add up these equalities by members, taking into account

that in summing up the first three columns on the right-hand
sidewe get on the basis of equations (I)
dXx dXy dXz
~AT"~I 57"+"dr
=~ p ^' etc *
In summing up the last three columns, we take into account

the law of reciprocity of shearing stresses (1.6) and formulas
(III); hence we obtain
dP dQ dR ,
+X x e. + Y e + Z e + Y e + Zx e
y yy z zz z yz z
or on the basis of (3.32)

dP
- dQ . dR '
By using this relation in the transformation of the integral / in

question, we arrive at the following equality:
J (Xv u -f Yv v + Z v w) dS + J (fXu + fYv + ?Zw) d*

= 2 f W dt.
(T) (T)
(5.62)
Sec. 37) THREE KINDS OF PROBLEMS OF THEORY OF ELASTICITY 145
The first integral in the left-hand member as has already been

stated, represents the doubled work of surface tractions done in
the process of deformation; the second integral represents the
doubled work of body forces; the right-hand member involves the
doubled potential elastic energy accumulated in the body. Obvi-
ously, relationship (5.62) formulates the assumption made at the
beginning of Sec. 20 in Chapter III concerning the existence of
the potential of elastic forces; according to this hypothesis, the
work done by surface and body 'forces must be completely stored
in the form of elastic potential energy.
Equality (5.62) makes it easy to prove the theorem of unique-
ness of solution for the three basic problems of the theory of
elasticity; for this purpose let us continue to reason as in Sec. 35
and assume that we have obtained two different systems of
stresses, displacements and strains, as designated in (5.48),
subject to the same conditions on the surface and the same body
forces.
If we accept the principle of superposition, the "difference of
these solutions"
XV _
- y'V y"V ** """""
* >
vV _
* - v'
*
V
"~"~ v"V
' 7^V_
- 7' 7" -"U ~"U i
XX = X X - X X \
Yy = Yy
- Yy \
Xy = X - X y y ,
(5.63)
it a' u!'\ v v' v"\
xy xy
can also be taken as a solution of a certain problem of the theory

of elasticity; consequently, one may apply equality (5.62) to this
solution in its simplified form with the second integral in the
left-hand member vanishing as the body forces are identical in
both solutions and, therefore,
pX = 9 X' =
//
P Jf 0,
etc.
Thus, we obtain
J J (Xv u + Y v v + Zv w) dS
= 2 J / J W fa, (5.64)
(S) (T)
where X V9 Y v Z Vy
u, v, w are taken from the first and third lines
,
of (5.63); the function

if W
is represented by (3.33), its arguments
should be taken from the fourth line of (5.63); if, however, it is
given in the form (3.34), its arguments should be taken from the
second line of (5.63). Turning now to the integral in the left-hand
member of (5.64), we note that
10- 1013
(1) case of the

in the first basic problem we shall have
everywhere on the surface
Xv = 0, Y v = 0, 7V = 0; (5.65)
(2) in the case of the second basic problem we shall have

everywhere on the surface
u = 0, v = Q, w = 0\ (5.66)
(3) the case of the mixed problem conditions (5.65) are

in
satisfiedon the portions of the surface where stresses are given,
while on the remainder where displacements are given conditions
(5.66) are fulfilled.
It is obvious that in all three cases the integral in the left-hand
member of (5.64) vanishes and there results the equality
CO
At the end of Sec. 21, however, it was shown that
W>0
at all points of the body; the last equality, therefore, is possible
only if all arguments of the function are zero, i.e.,W
*- =O f
etc.
Hence, follows that both systems of stresses assumed at the

it
start must coincide

at all points of the body; the same applies
to strains as well.
This proves the uniqueness theorem. It does not follow, how-
ever, from the foregoing proof that displacements will also be
identical in both of the assumed solutions; in the case of the
first basic problem, where stresses are prescribed on the surface,
this will not be true; in fact, from the conditions
*jrjr=^-C = ' etC "
according to what has been said in Sec. 10 concerning formulas

(2.11), it follows that one may give any small additional rigid
body displacement to an elastic body having certain stresses
and strains under the action of given forces. Obviously, in the
second and mixed problems displacements will also be the same
in both solutions, since they are specified over the whole or a
part of the surface of a body.

VI
Plane Problem in Cartesian
Co-ordinates
38. PLANE STRAIN

We now proceed to a large category of problems of the theory
of which are important for practical application and
elasticity
at the same time admit considerable simplification in the mathe-
matical aspect of solution.
Simplification implies that one may disregard one of the co-ordi-
nate axes in these problems, for instance Oz, and consider that
the whole phenomenon takes place in one plane Oxy. It is very
difficult to realise such a case in pure form in practice, but we
encounter something like it in many problems. These problems
fall into two groups opposite in a sense but common by the mathe-
matical form of solution.
The first of these groups corresponds to the case when one of
the displacements, for instance w, is zero everywhere, while the
other two, u and v, are not dependent on the co-ordinate z
corresponding to the displacement w\ this case is thus character-
ised by the following conditions which are valid throughout the
body:
In these conditions equations (III) give
* = 0; * y , = 0; e zx =Q\
* J rjr
= <Plt*. y)5 *yy
=? 2 (*. ^ j
(6-2)
Equations (6.1) and (6.2) "show that all displacements and

deformations take place exclusively in the directions parallel to
the plane Oxy, the pattern of displacements and strains being
the same in all sections of the body parallel to the plane Oxy
10*
148 PLANE PROBLEM IN CARTESIAN CO-ORDINATES [Ch. VI
(z=z 0t where z is any number). Such deformation is called

plane strain.
A case similar to this is encountered in problems dealing
with a long prismatical or cylindrical body with the axis parallel
to the axis Oz, under a lateral load normal to the axis Oz and
constant along the axis (though perhaps varying in the directions
normal to the axis). Such, for instance, are the problems of
a long dam (Fig. 46a), a long roller (Fig. 466), a long dome
(Fig. 46c), a long plate (Fig. 46d) with an axis parallel to the
Fig. 46
axis Oz; let us suppose that the load does not vary along the
axis Oz in all these cases.
If we isolate a thin element with a load acting upon it in such
a long prismatical body far from its ends by two sections a short
distance apart and parallel to the plane Oxy and imagine that
it works as a separate elastic body, we note that there must occur
elongations in it along the axis Oz\. these elongations result from

the transverse action of the load parallel to the plane Oxy.
Actually, however, this element is adjacent to two other elements
having the same elongation but in the opposite direction; as a
result of the interaction of these elements there will be neither
elongations nor displacements along the axis Oz, but there will
ocxur forces of interaction between the adjacent elements repre-
sented by normal stresses Z 2 developed due to the elimination of
the strain e zz .
Sec 38] PLANE STRAIN
Indeed, take the third of the equations of Hooke's law (V);

the elimination of the strain e zz will lead to the equation
or
(6.3)
Consequently, Z'z =/= 0, but it is a function of primary stresses Xx

and Y v produced by the load.
These considerations will now be illustrated by an example.
Let us have a blind flooring composed of a number of rectangu-
lar beams subjected to bending (Fig. 47a); let the axis Ox be
m ii
37
P q
m n
Fig. 47
directed along their span. The cross sections of the beams distort
in bending (Fig. 476), as has been shown previously (Sec. 33),
and there result elongations and contractions e zz If, however, .
we have a solid wide slab of width AA' in place of a beam floor-

ing, no distortion of its individual elements mnpq isolated far
from the ends A and A' can occur under flexure; there will
appear stresses Z z instead, since at the top points n (Fig. 476)
the elements will press on one another while those at the bottom
points q will stretch. These stresses Z z are expressed by formula
(6.3). Furthermore, assuming that the neighbouring elements of
the prismatical body under consideration (Figs 46 and 47) carry
the same load and work in exactly the same conditions, there
is no reason to expect any shear deformations between them;
hence we can easily see that everywhere e yz = e zx = 0. Thus 1
conditions (6.1), (6.2), (6.3) of plane strain are satisfied approxi-

mately for the elements sufficiently remote from the ends of the
long prismatical body under consideration. One can easily see
that in these conditions all our basic groups of equations are
considerably simplified. Let us begin with Hooke's law in the
form (V); the first of these equations on the basis of (6.3)
becomes
e xx = ^[XX - oY - y
<? (Xx + Y = 11 - Xx - oK
y }\ [(1 a) y| f
the second equation becomes
These equations may conveniently be put in the form
__ .... (6 4) '
where
.
= T^ :
i=Tb- < 6 5)
-
The fourth and fifth equations on the basis of (6.2) give
X, = 0; Z, = 0. (6.6)
The sixth equation becomes
But, using (6.5), we find
and therefore
e xy = 2(1 + gl)
x .
>1
y
y (6.40
.Furthermore, on the basis of (6.2), equations (6.4) show that
Xy = Yx *h(x> y), (6.7)
on this basis according to equation (6.3)
i.e., nonvanishing stresses are independent of the co-ordi-

all
nate which is, of course, clear all along.
z,
The following conclusions can be drawn from the above. There
remain only the first two equations of the group of equilibrium
Sec 39] GENERALISED PLANE STRESS
equations (I); these become

dX x dX y
dYx dY,
"""
dx dy
The surface conditions (II) are simplified as follows:
Xv = X x cos (vx) + X y
cos (vy);
\
(Hp>
Relations (III) reduce to

du dv t
'
dy
e = du
dy
'
As can easily be seen, there remains only the first of the six
conditions of continuity (IV)
""
.
2 "^
dx 2 (IVp)
dy dxdy
Hooke's law (V) becomes
e
^xx -( p \'
"yy (Vp)
v
y
Equations (Ip)-(Vp) make it possible to proceed to the solution
of the problem of plane strain.
39. GENERALISED PLANE STRESS. MAURICE LEVY'S

EQUATION. STRESS FUNCTION
Let us now turn to the other case, which is analogous to the

preceding one but opposite in the sense of extension along the
axis Oz, i.e., let us consider again the problems illustrated in
Fig. 46. Suppose, however, that the length of the prismatical
body along the axis Oz is very small; we shall thus have a
thin plate (of width h) loaded over the lateral surface (edge)
by forces parallel to its faces. The equilibrium of the plate, so<
long as the load does not exceed a certain limit (critical value) ,.
will be stable and the plate will not deflect in the direction of
the axis Oz. This case is called the generalised plane stress.
The faces of the plate (parallel to the plane Oxy) are assumed
to be free of stress, i.e., everywhere on them
Zz== 0; *z = 0; Yz = 0.
But since the distance between these faces (the width of the
plate) is very small, the stresses can be considered to be vanish-
ingly small within the plate as well The remaining stress

!
.
components Xx Yy X y will vary very slightly along the axis Oz

, ,
for the same reason (small thickness along the axis Oz) and,
therefore, we shall assume them to be independent of the co-or-
dinate z. Thus in the problem under consideration the following
-conditions are satisfied approximately for stresses:
(6.8)
They differ from the respective conditions (6.6) and (6.7) for
plane strain only by the condition Z = 0.
2
As for strains, this problem differs from the problem of plane

strain by the fact that here e zz is not zero [cf. formulas (6.2)]. It
will represent the transverse deformation along the axis Oz,
produced by primary stresses X x Y y lying in the planes paral-
,
lel to Oxy.
Thestrain e zz will entail some distortion of the plane faces
of the plate. In view of the small thickness of the plate, how-
ever, this distortion will be very slight. Let us take, for instance,
the case of the pure bending of a rectangular beam considered
in Sec. 33. The load-free lateral edges of the cross section incline
through an angle y (Fig- 446); one can see, however, from for-
mula (5.31) of Sec. 33 that this angle is proportional to the
width of the beam b\ therefore, if b is small, i.e., if the beam
has the shape of a thin plate, angles y WJ"U a s be very small '
and the distortion of its lateral surfaces will be negligible.

Proceeding from equations (6.8), we draw the following con-
clusions: the basic groups of equations (I) and (II) have the same
form (Ip) and (Up) in this case as for plane strain; it is also
1
an assumption is justifiable with regard to normal stresses Z z
Such \
as stresses X z and Y z the assumption is true if the load on the lateral

to ,
surface is distributed uniformly along the generator (as was supposed in the
case of plane strain). Let us generalise the problem and assume that the load
is distributed symmetrically with respect to the middle plane of the plate
O 0, but nonuniformly (Fig. 48a); then shearing stresses Z x or Z y may
Sec. 39] GENERALISED PLANE STRESS 153
sufficient to retain only equations (Hip) of group (III); equally,

we may limit ourselves to one equation (IVp) in group (IV); as
regards Hooke's law (V), on the basis of the last of conditions
(6.8), there will be a distinction from equations (Vp) of the
problem of plane strain; indeed, in our case we have
1 / -*v \r \
P ( Y (
yy E ^ y (V'P)
x y
.
Summarising the foregoing considerations we conclude that in

the solution of problem's both of plane strain and generalised
plane stress, we may use the basic groups of equations: (Ip),
be considerable in the section mn\ however, the curve of their distribution

(diagram) will be antisymmetrical and its area will be equal to zero; this
can be expressed in the form of equalities
- {' =
0;
1 ( Z y dz
h J h J
It may thus be said that shearing stresses Z x and Z y

will be equal to zero
on the average across the thickness of the Therefore, in a more strict
plate.
presentation of the theory of generalised plane stress all the components of
Fig. 48
the stress tensor are replaced by their average values across the thickness of
the plate. For simplicity we shall assume in future that surface loads are
distributed uniformly along generators and, therefore, we shall assume that
throughout the plate
Xz = 0.
154 PLANE PROBLEM IN CARTESIAN CO-ORDINATES [Ch VI
(Up), (IIIp) and (IVp). Hooke's law, however, has different ex-
pressions for these problems: equations (Vp) for plane strain, and
equations (V'p) for plane stress. It is important, however, to
note that the form of these equations is the same in both cases;
the difference is only in the values of the elastic constants which
are expressed in terms of E and a by formulas (6.5) in the case
of plane strain.
Henceforth we choose stresses X Xl Yy ,
X y =Yx as unknowns;
it is, necessary to transform the equation of compat-
therefore,
ibility (IVp) replacing strains by their expressions from equa-
tions (Vp) or (V'p), depending on whether we have plane strain
or plane stress.
Let us differentiate equations (Vp) as required by equation
(IVp) 2 2
d exx 2
1 / d X, d Yy
~~
dy
2
E l \ dy
2 l
dy
2
d 2e
yy
1 /d 2 Y y d2 X (6.9)
U y2
f) A, p
L, |
\
\
f) y'i
1/./V
\
/)
L/^V y%
d*exy 2(1 + a,) d*X y

Ox dy dx dy (6.10)
j
We transform the right-hand member of equation (6.10) express-

ing shearing stress X y in terms of normal stresses X x and Y y
from equations (Ip); differentiating them and transposing some
terms to the right-hand side, we have
d2 xy d2 xx dX
dx 2 P-
v
dx
dydx
d 2
Yx d 2
Y dY
^
dx dy dy
2
dy
Adding up by members, we obtain

d2
_ d2 Xx d 2 Yy idX dY
2
Xy
~dx~dy
= Jx 2
dy~
2 p
\dx~
~^~
~d^
In the sequel we shall restrict ourselves to the case when body
forces are constant throughout the volume of the body; then
dX
dx
_
~ dY _
~~ U *
dy
In particular, we shall deal with the gravity force and, if we

direct the axis Oy downward, we shall have
Sec. 39] GENERALISED PLANE STRESS 15J>
where p is the weight per unit volume of the body; now equa-
tion (6.11) becomes
Y
A* A
_ Ayx fity
~~ ^
/ /)2 Y
y y
* ""T 2
dx dy \ dx 2
dy
and equations (6.9) and (6.10) are rewritten as
2
dy E>
dx 2 ~"T
~~
^jc dy E l
'
Adding them up by members, we obtain the left-hand member

of (IVp) in the form converted to stresses:
d'*yy ^jry ~~ 1 _ [
^2 ^ ^2
^y &** ^y 1 ~
_
'
dx 2 dx dy E l L dx 2 ""
();c
2 '
r)>;
2 '
dy
2
\
', L dx 2 '
dy
2
Here the bracketed expression represents the action of the Lap-

lacian operator of the second order on the function
functio (X x +Y y ):
d2 (Xx +Y y)
T
v* >
'
y> dx 2 ]
dy
2
Introducing this in the equation of compatibility (IVp), we ob-

tain it in the final form, converted to stresses:
This is the so-called M. Levy's condition. It may be obtained

in a different way if we proceed from the fact that in the ab-
sence of body forces the volume expansion is a harmonic func-
tion [cf. formula (5.55)]:
?2 = 0,
and, consequently, according to (3.10) we have

V 6=
2
0, where Q = Xx -\-
But in the case of plane strain

Zz = c(Xx +Yy ),
and in the case of generalised plane stress

Hence we obtain
Thus the solution of the plane problem is reduced to integra-

tion of three differential equations:
dXx dX v dYx dY
-f J%) = 0, (IV'p)
and satisfaction of the conditions on the surface:

X = Xx cos (vx) + X
v y
cos (vy),
Y = Y x cos (vx) + Y
v y
cos (vy). J
(U P)
If in this way we -find stresses

Xx , Yy ,
Xr (6.13)
then, introducingthem in (Vp) for plane strain or in (V'p) for

plane stress, we determine strains e XXl e yy e xy as functions of ,
x and y\ substituting them further in equations (Hip), we get

*!L -
e *x* ^L e
Cyr ^
*L-4-*L
dx dy dx dy
and, integrating the last equations, we obtain displacements u, v.

This completes the solution of the problem.
It is very important to note the following. If it is only stresses
that are to be found in a given problem, one may confine oneself

to equations (Ip), (IV'p) and (Up) which do not involve the
elastic constants. Hence one might expect that the state of stress
does not depend on the material; this conclusion is true for a body
bounded by a simply connected surface (which is represented by
a simply connected contour in the plane of action of forces). In
the case of a multiply connected contour, the state of stress does
not depend on the material if the external loads are balanced at
each of the contours separately These conclusions underlie the
l
.
practical application of the photoelastic method of stress analysis

in polarised light; a given material is replaced then by another
one a transparent and photoelastic material from which a plate
is made and tested as a model.
G. B. Airy indicated the possibility of further simplification in
the solution of the problem. This simplification is based on the
fact that we can easily find the general solution of the system of
equations (Ip) so as not to deal with them later. The system (Ip)
1
To be exact, if the resultant of the loads is equal to zero at each contour
separately.
is nonhomogeneous and, therefore, its general solution represents

the sum of the general solution of the homogeneous system
dX dX y
(6.14)
dx
and a particular solution of the system (Ip); this particular solu-

tion can easily be found if we assume, for instance [see formula
(6.12)1
Xx = Y y = Q\ Xy =-px, (6.15)
or
Xx = X = Q\
y
Yy = -py. (6.150
Let us now turn to system (6.14). Its general solution contains

one arbitrary function <p(#, y) of the independent variables x and y
and has a simple form
If wesubstitute these expressions of stresses in equations (6.14),

we shall see that they are satisfied identically, whatever func-
tion y(x, y) may be, provided its partial derivatives up to the
fourth order inclusive exist and are continuous. This function is
called the stress function or Airy's function.
The general solution (6.16) can readily be obtained as follows:
we satisfy the first equation of the system (6.14), assuming
'
where ty( x > y) ls an arbitrary function. The second equation will

be satisfied if
X =-dy
y
' Yy = -L
dx
'
/
6 IQX
10.18)
where %(x, y) is also an arbitrary function. Both equations (6.14),

however, will be satisfied simultaneously only if the expressions
of X y in (6.16) and (6.17) coincide, so as to have
dx
or --4--
dx ^
dy dy
this equation will be satisfied if we assume
T
,
dv
^=-^1-;
dy
' y
A = --dx
dv
-I-;'
introducing these values of ty and x in (6.17) and (6.18), we ob-

tain formulas (6.16). The general solution of equations (Ip) will
be found by adding their partial solution, for instance, in the
form (6.15') to (6.16); we thus have finally
px '
Substituting these expressions in equations (Ip), we can easily

see that the latter will be satisfied identically and we shall not
have to deal with them in future. There remains one equation
(IV'p) which is expressible in terms of the function cp; indeed,
on the basis of (VIII)
-
y dx 2 -r dy2
and equation (IV'p) will be represented symbolically as
+ |J)=0. (IX')
Writing out the symbol of Laplace's operator, we get
Thus solution of the plane problem in terms of stresses is

the
reduced the integration of one partial differential equation
to
(IX) of the fourth order; if we determine the function <p(x, y)
from this equation, we shall find stresses at any point of a body
by formulas (6.16) or (VIII). Boundary conditions corresponding
to each specific problem must, of course, be added to equation
(IX). Here we shall dwell on the first basic problem of the theory
of elasticity (Sec. 37); in this case we must prescribe stresses, i.e.,
loads, on the boundary of a body, i.e., on the contour of the
cross section (if plane strain is considered) or on the contour of
a plate (in a state of generalised plane stress). The boundary
conditions will have the simplest form, if the contour of a body
is a rectangle with the sides parallel to the co-ordinate axes Ox
and Oy. We must then prescribe stresses Y y and X y on the sides
parallel to the axis Ox, and stresses X x and Y x on the sides par-
allel to the axis Oy\ according to (6.16), these stresses are equal
to the corresponding values of the second derivatives of the func-
tion (p(#, y) which have to be specified. Several problems of this
kind are treated in the following sections.
*
Consider the boundary conditions in the case of a contour of
arbitrary configuration; for this purpose turn to the general con-
ditions on the surface (Up); by the use of the stress function
y) they will be written down as
= -- cos (**)
(X)
dxdy
Thus the solution of the plane problem is as follows: one must

find a function y(x, y) satisfying equation (IX) at all points of
the cross section of the body in question (Fig. 46), and equations
(X) on the contour of this section, where X v and Y v are the pro-
jections of the external load on the co-ordinate axes. Once the
function cp is found, we determine stresses by equations (VIII);
further determination of strains and displacements is carried out
according to equations (Vp) or (V'p) and (Hip), as indicated
above.
Remark. If there are no body forces (i.e., if p = 0), the con-
ditions on the surface (X) can be written in a more compact
form, introducing the concept of the derivative of a function with
respect to arc length of the contour of the cross section. Indeed,
we have (Fig. 486)
cos (vx) = sin a = -~- , cos (vy) = cos a = --
ds
.
Introducing this in equations (X), we have
v
_
~ ~*
d 2(
P dy
~
. p dx _
d / dcp \
'
'
dx dy ds ~~~ds (~dy)
d 2 y dx d (6.19)
dy
dxdy ds ~dx*~ds 'ds dx
These equalities permit us to represent the conditions on the

contour in the alternative form which is of interest for the gen-
eral statement of the question concerning the integration of the
equation of the plane problem (IX) with loads prescribed on
the contour. We multiply (6.19) by ds and integrate with respect
to ,s along the contour starting from an arbitrary point S Q taken
as the origin of arcs:
A
-^-
dx
~ A
" Y v as
f r
J
ds
(6.20)
Here A and B are arbitrary constants; they express the values

of the derivatives
dy dy
'
dx dy
at the point 5 of the contour.

To make further computations more clear, let us introduce the
following analogy: we replace the contour of the body under
Fig. 49
investigation by a bar of the same form, which is cut at the point

SQ (Fig. 49), where we apply the forces
A parallel to the y axis,
B parallel to the x axis.
Keeping in mind that the surface loads X v and Y v act per unit
arc length of the contour, we note that the quantities and YW X&
in the right-hand members of equalities (6.20) represent the sums
of the projections of forces applied to the part 5 5 of the bar on
the axes Ox and Oy. If we take new axes ON
and OT (Fig. 49)
directed parallel to the normal and the tangent to the contour
at the point 5 in place of the axes Ox and Oy, formulas (6.20)
in the new co-ordinates n and / will be written as
(6.21)
(6.22)
where 7>> is the longitudinal force at the point S of the bar and
is the transverse force at the point S of the bar.
In formula (6.21) the quantity
represents the derivative of the stress function q> along the nor-
mal to the contour; similarly
is the derivative along the tangent to the contour or with re-
spect to arc length of the contour
Comparing equality (6.22) with the well known theorem on the

derivative of the bending moment in a bar
dt
~~'
we may consider that
cp
= Af (5)
, (0.23)
where MW is the moment of the

applied S5
forces to the part
of bar with respect to the point S; in calculating
the by M&
integration of equation (6.22) an arbitrary constant will be added
which can be prescribed by applying a couple with any moment
C at the initial point S (Fig. 49).
All the foregoing considerations enable us to present the con-
ditions on the contour (when loads are prescribed on it) in the
required new form; for that we must calculate the values of the
stress function y(x y) and its normal derivative
f
dy
~dn
at each point of the contour on the basis of the loads prescribed

on the contour, according to formulas (6.21) and (6.23) as the
longitudinal force and the bending moment due to the loads
prescribed on the contour. The expressions of cp and -p- will in-
volve three arbitrary constants A, B and C (initial parameters);

they do not, obviously, affect the prescribed contour loads X v
and Yv and the stresses produced by them.
Functions satisfying equation (IX) are called biharmonic; the
plane problem of the theory of' elasticity, when loads are pre-
scribed on the contour, can now be treated in the following math-
11-1013
ematical form: it is required to find a biharmonic function

y)
at all points of the region bounded by the given contour, if

the values of the function cp itself and its normal derivative are
prescribed on the contour.
40. SOLUTION OF THE PLANE PROBLEM

BY MEANS OF POLYNOMIALS
In many problems it is often found convenient to employ the

semi-inverse method specifying in advance the analytic form of
a stress function cp(x, y) and selecting its parameters (for in-
stance, coefficients) so as to satisfy the surface conditions (X)
and the basic equation (IX). Let us consider several problems in
which <p(x, y) can be prescribed in the form of an integral function
(polynomial). If the function <p represents a polynomial of the
second degree
cp (
X ,
y) = Xi + bxy + ~ y\ (6.24)
equation (IX) will obviously be satisfied everywhere for any val-

ues of a, b and c. The stresses, in accordance with equations
,(VIII), will be expressed as
J
Xx = c\ Y y = a\ Xy = bpx. (6.25)
If p = 0,stresses are constant; we

the have the case of the
homogeneous state of stress.
If q> is a polynomial of the third degree
(6.26)
equation (IX) again be satisfied for arbitrary values of the

will
coefficients; the stresses, according to equations (VIII), will be
represented as
(6.27)
i.e., they will be linear functions of the co-ordinates.
From equations (6.25) one can see that terms of the first degree need not
1
be included in function (6.24) as they will not affect the magnitude of stresses.
Sec. 41] BENDING OF A CANTILEVER 163
If 9 is given as a function of the fourth or higher degree, its
derivatives entering in equation (IX) will be different from zero,

in general; the coefficients must, therefore, be chosen in such a
way as to satisfy the compatibility condition (IX) for arbitrary
Fig. 50
values of x and y. The stresses will then be functions of the sec-

ond and higher degree.
Pure bending. Consider the particular case of the stress func-
tion (6.26):
? = iry
k ~
3
-
Neglecting the influence of the own weight, we obtain the follow-

ing stresses from equations (6.27):
X x = ky\ K v = 0; Xy = Y jr
= Q. (6.28)
If we take a plate of width h (Fig. 50) and assume that

stresses (6.28) are set up in it, the following stresses will act on
its surface. On the upper and lower faces ( for //= -rH
Yy = / _
y
= Xy = X _ = y
0;
on the lateral faces (for x=. -g-
The diagrams of the stresses X x are shown in the drawing; ob-

viously, we have the case of pure bending investigated in the
general form in Sec. 33.
41. BENDING OF A CANTILEVER

Consider a plate built-in at the left-hand end and subjected
to a loadQ distributed over the right-hand end section (Fig. 51).
The elementary theory of bending gives the following system of
11*
stresses for this case:
(6.29)
y y
1 *
yd
^y jb 2J
Here M is the bending moment, Q is the shearing force, / is

the moment of inertia of the cross section; in the case of gener-
alised plane stress
/=-jg-; in the case of plane strain /= ;
6 is the thickness of the plate in the direction perpendicular to

the plane of the drawing.
0\
Fig. 51
Let us check whether stresses (6.29) are possible from the

viewpoint of the plane problem and whether they correspond to
the case of bending represented in Fig. 51. To answer the first
question it is sufficient to ascertain whether stresses (6.29) can
be obtained from some stress function y(x, y) satisfying equa-
tion (IX).
The general form of functions (6.29) expressing stresses is
dy*
Y
*
y
=0u = -i
dx*
'
(6.30)
Integrating the first of these equations twice, we have
(6.31)
Sec. 41) BENDING OF A CANTILEVER 1$5
Differentiating this function with respect to x and substituting

the results in the last two of equations (6.30), we obtain
>''(*)
= o, (6.32)
- *'(*) = C+Dy. (6.33)
Both of these conditions should be satisfied for any values of x

and y\ therefore, we draw the conclusions:
substituting this in equation (6.33), we have
Hence
E=-C; D = f--
The stress function (6.31) becomes
(6.310
The terms of the first degree Hx + Fy + K may be omitted because

they do not apparently affect the stresses. One can easily see
that function (6. 31') satisfies equation (IX) (Sec. 39) for any
values of the coefficients, i.e., this function is its solution.
The stresses will be expressed as
(6.29a)
a> ? B
dxdy 2
This shows that stresses of the form (6.29) or of the more gener-
al form (6.30) satisfy the equations of the theory of elasticity.
We shall attempt now to select the indefinite coefficients A, B
and E so as to satisfy the conditions on the surface (Fig. 51).
1. On the upper and lower faces
2. On the right-hand end section
Since the law of distribution of shearing stresses over the sec-

tion x = l is in no way stipulated in the conditions of the problem,
there remains only one condition
for x = l
~*"2
requiring that the forces at the end cross section reduce to the
given load Q.
The reader is recommended to carry out the necessary compu-
tations as an exercise and to confirm for himself that the stresses
will be represented by expressions (6.29).
Thus, in this problem the expressions for stresses (6.29) obtained
on the basis of the hypothesis of plane sections are supported
from the viewpoint of the theory of elasticity as well, provided
the load Q is distributed over the end section in accordance
with the law indicated by the last of formulas (6.29). When the
load Q is applied in a different manner, the expressions for stresses
will also be different, but a considerable numerical difference will
be only near the loaded right-hand end of the plate, on the basis
of Saint-Venant's principle (Sec. 31, Fig. 39).
In this problem it is of interest to investigate strains and
displacements and to compare the result with that given by the
hypothesis of plane sections in strength of materials. Since we.
shall now make use of Hooke's law, let us first decide which of
the two cases (plane strain or plane stress) we shall dwell on;
to obtain full agreement with an analogous problem in strength
of materials let us treat a state of plane stress and use Hooke's
law in (V'p) (Sec. 39). Then from (IIIp), (V'p) and
the form
(6.29) we
obtain the following system of equations for determin-
ing displacements:
..
du_ Q_
dx EJ ^ y
dv (6.34)
dy
dv ,
,
du (l-f-a)Q / h2 9
2
\ ,- orv
6 35 >
-57+17^-^7^11-- y )- <
-
Sec 41] BENDING OF A CANTILEVER 167
Integrating equations (6.34), we get
^
W /
\~2
ly
2
__ xy
2
2~~)
\
,
~r"Ej
_Q_
(6.36)
where /i and/ 2 are arbitrary functions; for convenience in carry-
ing out further computations, we have furnished them with the

coefficient
-gj.
To determine the form of these functions we sub-
stitute expressions (6.36) in equation (6.35); first we find
dv a
~~
EJ
- 2 EJ i'
dy \ / )
Substituting this in equation (6.35) and cancelling out

-gy,
we
obtain
or after transposition of the terms
(6.38)
The bracketed functions depend: the first one on x alone, the
second one on y alone; but since x and y are arbitrary and
independent of each other, equality (6.38) can exist only if the
bracketed expressions are equal to constant numbers m and n, so
that
= (l+a)l~. (6.39)
On this basis we obtain from equation (6.38)
Integrating these equations, we have
A (x) = ~2
tx*
g-
(6.40)
168
We
_ PLANE PROBLEM IN CARTESIAN CO-ORDINATES
introduce this in equations (6.36):
lya-My + p.
_
I
[Ch. VI
^ ^
To determine the arbitrary constants m, n, a and p let us fix

the left-hand end; in strength of materials all considerations are
referred to the axis of a rod; therefore, here too we first fix the
initial point of the axis, i.e., we specify the conditions
for x=y= u = 0; v = 0.
Now from equations (6.41) we immediately have a
= p = 0.
Itis necessary next to fix (Fig. 51) the supporting section
(against rotation about the point 0); for this purpose the condi-
tion that the initial tangent be horizontal is most frequently
specified:
( X=Q dv
A TT^
y=
for ;
\
ox
in this case we find from equations (6.41)

/it = 0,
and from equation (6.39)
n
Equations (6.41) give
From equation (6.43), setting r/

= 0, we find the equation of the
deflected axis of the rod:
which coincides with the solution obtained in strength of mate-

rials. Now investigate the deformation of plane cross
sections;
let the equation of such a section before deformation be
and let its equation after deformation be
1
Cf. this with formula (2.3) in Sec. 10.
Sec 41] BENDING OF A CANTILEVER
or
(6.45)
y]
Consequently, the section does not remain plane but distorts

along a parabola of the third order (6.45). In the above fixing
conditions at the left-hand end (Fig. 52) the left-hand end sec-
tion (x = 0) also distorts along the curve
+ q)A y
(2 + q)
y
,
4 6
If we take a linear element dy on the axis of the rod at the

= =
point of fixing (at * 0, y 0), its angle of rotation [formula (2.4)
in Sec. 10] will be
du 3Q
4EJ
Consequently, the element rotates (Sec. 10) in the direction from

the positive axis Oy to the positive axis Ox\ thus, plane cross
Fig. 52 Fig. 53
sections distort after deformation and do not remain normal to

the axis '. This shows that in our problem the formula of the nor-
mal stress
M
derived from the hypothesis of plane sections, holds valid when
sections distort.
One can easily find the general formula for the inclination of the section
1
with respect to the axis at all its points; this inclination is obviously equal
(Fig. 52) to the shear
dv .du
It can most easily be calculated according to (6.35).

In Sec. 31 we pointed to the schematisation encountered some-

times in the formulation of problems with regard to load; it is
useful to note here the same schematisation in the formulation
of the "fixing conditions at the support'*. Indeed, fixing can be
realised in an infinite number of manners: one may, for instance,
require that all points of the section remain immovable or one
may fix only several points of the section in the same sense;
finally, one may add to these conditions the requirement that
some linear elements should not rotate due to deformation. It
is obvious that the stress distribution near the fixed end of the
cantilever will depend in such conditions on the mode of fixing.
The solution given above, however, has fully predetermined the
stress distribution in the whole cantilever. To realise the fixing
there remains at our disposal only the rigid body displacement
of the cantilever, which does not affect the stress distribution;
therefore, in this problem the concept of fixing is of a kinematic
character only and the modes of fixing are limited.
In the preceding discussion we have fixed the left-hand end
requiring that the element of the axis dx at the supporting end
must be horizontal
Let us now realise the same fixing in a different manner

(Fig. 53) specifying the condition
which requires [formula (2.4) in Sec. 10] that the element dy of

the supporting section must remain vertical.
Then the first of equations (6.41) gives n = Q while from equa-
tion (6.39)
from equations (6.41) we obtain the following equations in place

of (6.42) and (6.43):
~EJ
Thus, all displacements have altered. The equation of the deflected

axis will be [cf. formula (6.44)]:
Sec 42] BEAM ON TWO SUPPORTS 171
The deflection of the right-hand end in the first case [according

to formula (6.44)1 is
~ 3EJ '
for the second mode of fixing, however, it is given (Fig. 53) as
J
f ~~
3EJ
|
"t"
4EJ
~~
_ Q/
3
|
3Q/ ,
fid7 v
The last term

-^^ accounts for the effect of shear deformations
(in other words, the effect of shearing force) on the deflection.

From
the preceding discussion it is clear that the concept of
the deflection of right-hand end as such is indefinite in this
problem; the deflection may be different depending on the manner
in which the left-hand end is "fixed".
In addition to the above two modes of fixing one can imagine
an infinite number of other modes; corresponding to each of them,
there will be certain specific displacements and certain deflection.
Saint-Venant's principle states that the stresses vary substantially
only in the neighbourhood of the fixed section; the displacements
will be different, however, throughout the length of the rod.
42. BEAM ON TWO SUPPORTS

Consider the bending of a beam supported at the ends and
subjected to a uniformly distributed load q. The supporting
reactions are assumed to be the shearing forces distributed over
the end sections. If the co-ordinate axes are located as in Fig. 54
the elementary solution of the problem leads to the following
stresses:
"y 'x 27
Let us write these formulas in a more general form:
(6.48)
and check whether they satisfy all the equations of the plane
problem. The stresses Y v are usually neglected; this circumstance
permits us to state a priori that the system of stresses (6.48) in
172 PLANE PROBLEM IN CARTESIAN CO-ORDINATES (Ch. VI
conjunction with the assumption Y y =0 apparently will not satisfy

the equations of the theory of elasticity as on the
upper surface
y = -- 2"
we have the equality
We shall, therefore, attempt to satisfy the equations of the theory

of elasticity prescribing stresses (6.48) but
rejecting the condition
4 == U.
y
T
h -*-,r
Fig. 54
First we select the general form of the stress function for stresses
(6.48); according to equations (VIII) (Sec. 39) we have
-
dx dy
By integrating the first of these equations we get
x), (6.50)
B
-JT-
x 2 y-, + f l (x) y -f/2 (x) ,
(6.51)
where and
/ 2 are arbitrary functions of x. Differentiating equa-
/t
tion with respect to x and substituting the
(6.50) resulting value
in the second equation of
(6.49), we have
hence we obtain
Introducing this value in equality (6.51), we have

Sec. 42] BEAM ON TWO SUPPORTS 173
Substituting this expression in equation (IX) (Sec. 39), we see

that is not satisfied thus, + * (x

3
> y) -
(6 52)
-
'
The function ty( x > y) ls selected so as to satisfy the compatibility

condition (IX). From equality (6.52) we have
V
and equation (IX) becomes (assuming /2 (x) =0)
(6. 5 3>
Looking at this equation we can easily see that its simplest

solution will be represented by an integral function of the fifth
degree:
Introducing this in equation (6.53), we have
(6.54)
2 3
The term -yj
JC y in the function 0(3 may be omitted since equality
D
2
(6.52) already contains the term of a similar form -g-.x: y
3
; then-
from equation (6.54) we have

ff= 4B F.
After that the stress function (6.52) takes the final form
(6>55)
which satisfies condition (IX). Hence we obtain stresses
(6.56)
It will become 4Bi/=0.

1
1
The term Ey, which does not affect the stresses, is omitted.
To complete the solution it remains to satisfy the conditions on

the surface; first consider the conditions on the upper and lower
faces.
1. On the upper face y= -
(6.57)
2. On the lower face */

= + -*
(6.58)
The first of conditions (6.58) gives

B *>
24
3
L
2
.
r
is2\-'i 4
Hence we obtain
and stresses (6.56) will be
(6.59)
X = Y x = - Bxf + Cx.
y
From conditions (6.57) and (6.58) for normal stresses we get
__ 24 '
2
_ ,
'
,
^' 24 2
_,,_Q
"
Adding up these equations, we find
= -f; (6.60)
subtracting them, we obtain
^-
/,3
Ch = q. (6.61)
Conditions (6.57) and (6.58) for shearing stresses yield
^p-+-C
= 0; (6.62)
solving equations (6.61) and (6.62) simultaneously, we find
B =-> C =-ff (6-63)

Sec 42] BEAM ON TWO SUPPORTS 175
Substituting the values of L, B and C from (6.60) and (6.63)

in formulas (6.59), we obtain
"
yy
r ^L\JL
h* \ 3 4
y
> ^
I
12 (6.64)
We proceed to the conditions on the end sections of the rod;
for*= Y
(6.65)
/
" A
2
The last condition requires that the shearing stresses on the end
section reduce to the supporting reaction ( ; one can easily
-^-j
see that it is already satisfied, since the last of equations (6.64)
coincides exactly with the prescribed expression of the shearing
stress [formulas (6.48)]. At the same time we note that the first
of conditions cannot be satisfied; indeed, for

(6.65) x=-^-
< 6 66 >
-
and, consequently, X x does not vanish; but since the right-hand

member of equality (6.66) is an odd function of y, it follows
that
h
+
T
/ (6.67)
i.e., the normal stresses over the end section reduce to a couple
with the moment
4
I==
^TT S~ ^"90"' ( 6<68 )
Thus, solution (6.64) corresponds to the bending of a plate

under the action of a uniformly distributed load q and moments
M at the supports (Fig. 55).
By a suitable adjustment of the coefficient A we can assign any
value to the supporting moments (6.68). Expressing the coefficient
3
HHH I Hill M
Fig. 55
A through the supporting moment M according to equation (6.68),

we have
and the expressions of stresses (6.64) will be written finally as
20
(6.69)
If we set here
(6.70)
the supporting moments will be eliminated and we shall return

to the bending of a plate under a continuous load alone, but in
the presence of normal stresses (6.66) on the supporting sections;
on the basis of equations (6.67) and (6.70), however, these stresses
reduce to a balanced system of forces and, therefore, their
influence, by virtue of Saint-Venant's principle (Sec. 31), will be
appreciable only near the ends of the beam. For = the first M
of equations (6.69) can be written down as
Comparing this with the first of equations (6.48), we note that

the difference in the expressions of the stress
Sec 43] TRIANGULAR AND RECTANGULAR RETAINING WALLS 177
will have only a little effect on the magnitude of stresses in the

middle part of the beam (for small x) when the height of the
section h is small compared with the span /.
Summarising the foregoing conclusions, we can say that:
1. The expression of shearing stresses given by the elementary
solution is confirmed by the present more exact solution (6.69).

2. .For the normal stress X X9 the elementary
solution gives values close to reality far from JL j
the ends of a slab or a plate (in the case when the f

height h is small compared with the span), where _A
these stresses play a dominant part. 2
3. The second of equations (6.69) gives the 1
_
expression of the normal stress Y y which is ne-
j
glected in the elementary solution. The diagram of fy
the distribution of these stresses across the depth "o"
of the section (cubic parabola) is shown in Fig. 56. I

J
By using equations (6.69), one can obviously
solve the problem of a beam with the ends built \ < 9 x
into walls; for that the magnitude of the sup- 2
porting moments M must

be chosen in such
Fig 56
a way as to satisfy the conditions of the form
_JL =or
-^-==0 (cf.
Sec. 41) at the ends of the axis of the beam.
First it is necessary, of course, to find the displacements of the
beam, as done in Sec. 41.
43. TRIANGULAR AND RECTANGULAR RETAINING WALLS

(M. LEVY'S SOLUTIONS)
Consider a problem of a dam wall or a retaining wall subjected

to the pressure of water or a loose material according to the
hydrostatic law in proportion to the depth of a given area
(Fig. 57). The conditions on the surface will obviously be ex-
pressed as:
1. On the face Oy for jc =
2. On the face OA for x=y tan p

X V = Q; Y v = 0. (6.73)
Since the number of conditions on the surface is four, for the

solution of the problem it is sufficient to take a stress function
with four indefinite coefficients. Let us make use of the function
12-1013
of the third degree (6.26) retaining the first four terms in it, i.e.,
setting
a =b=c= 0.
According to equations (6.27) (Sec. 40) the stresses will be linear

functions and will be expressed as
Y y = dx + ey\
= X = - ex y fy - ?gx. (6.74>
Here pg is the weight per unit volume of the wall; for brevity
designate it by p, i.e.,
Applying conditions (6.72), we have: /# = 0, ky= yy. Hence

/
= 0, k= y and stresses (6.74) will be written as
/
.
(6.74 >
We v/rite out conditions (6.73) according to equations (Up)
Fig. 57
(Sec. 38) and introduce in them the values of the cosines of the
angles between the outward normal v and the axes (Fig. 57):
Xx cos p X y
sin p = 0; Y x cos p Y y sin p
= 0.
We introduce here the values of stresses (6.74') setting, in accord-

ance with conditions (6.73), x = yian p
^y cos p + (e +/?) y tan p sin p

= 0,
(e + p)y tan p cos p (dy tan p + ey) sin p = 0.
Cancelling y out of both equations and dividing by cos p, we

obtain
(e -t P) tan
2
p
== 7; (d tan p -f- *) tan p = - -
(e +p) tan p.
Sec 43] TRIANGULAR AND RECTANGULAR RETAINING WALLS 179
Hence we easily find
*
~-
tan 3
and stresses (6.74) assume the final form:
(6.75)
_ *V _ .
tan 2 a
The diagrams of stresses Y y and X y over the horizontal section

y = yo are shown in Fig. 58. Comparing this with the result of
the elementary calculation according to the formulas
,
x -
/Y
we find that the values of the normal stresses Y, coincide in bothy
solutions; the distribution of the shearing stresses X however, t/ ,
appears to be essentially different.

Hitherto we have selected the stress function cp(#, y) in the
form of a polynomial of not higher than the fifth degree [cf. for-
mula (6.55) in Sec. 42].
By increasing the degree of a polynomial, it possible to find is
the solution of more complicated problems. For instance, prescrib-
ing cp as a polynomial of the sixth degree, one can obtain the
solution given by M. Levy for the case of a rectangular dam or
retaining wall. This solution has the following form:
(6.76,
The reader can easily prove that these stresses satisfy the follow-
ing conditions on the surface (Fig. 59):
1. Forjc =
2. For x = a
x_ x = w r_ = o.
3. Fory =
^= 0; K,=0.
12*
180 PLANE PROBLEM IN CARTESIAN CO-ORDINATES [Ch. VB
The last necessary condition

f or y =
is not satisfied completely and on the upper face of the wall there
remain shearing stresses
These stresses, however, reduce to a balanced system of forces

since their resultant is equal to zero:
they have therefore only local significance and the region of

their influence, according to Saint-Venant's principle, is small, the
Yy diagram
Xy diagram
Fig. 58
more so as they act in the upper, less stressed part of the wall
where the stress analysis is not usually required.
Solving this problem by the elementary method, we would
obtain the folloVing expression for the stresses Y y :
_N -/T-T-
,
Af /
\x
a\_
Tj
IZiZ
2 3 a
which corresponds to the first two terms of the second of equa-

tions (6.76).
Sec. 44] BENDING OF RECTANGULAR STRIP 181
The reader is recommended to compare, in a similar manner,

the expressions of the stress X y = Yx in the elementary solution
and according to equations (6.76).
44. BENDING OF A RECTANGULAR STRIP;

FILON'S AND RIBIERE'S SOLUTIONS
The method of solution of the plane problem by means of
algebraic polynomials, considered in the preceding sections, affords
limited possibilities for practical application because it is very
difficult to select a polynomial which gives the solution corres-
ponding to a prescribed load which is more or less complicated.
The method of trigonometric polynomials, proposed by C. Ribiere
and L.N.G. Filon for the case of bending a rectangular strip
whose length is considerably greater than the height h (Fig. 60),
M
'
V
I
Fig. 60
was found to be far more effective. In such cases it is most impor-

tant to satisfy as exactly as possible the conditions on the long
sides of the strip where a prescribed load producing the bending^
of the strip acts; the conditions on the short sides (ends) can be
satisfied only by prescribing the resultant factors Qo, NI, # ,
M ,
Qi, MI, which characterise the reactions of the supports or the

loads at the ends of the strip (without regard to the law of stress
distribution across the height h of the initial and terminal sec-
tions); when the ratio is small this is permissible from the
point of view of Saint-Venant's principle (Sec. 31).

As indicated in Sec. 39, the solution of the plane problem by
means of the stress function y(x, y) is reduced to the integration
of equation (IX)
-=0 (6.77)
with the prescribed boundary conditions; these conditions will be

dealt with later. Now we shall attempt to obtain as general a solu-
tion of equation (6.77) as possible. Such a solution can easily be
found by the method of separation of variables used above in
Sec. 28 with respect to the equation of vibrations (4.28); Ribiere's
and Filon's methods are particular cases.
Let us seek a particular solution of equation (6.77) in the form
? (x, y)
= XY, (6.78)
where X function of x only, and

is a Y is a function of y only.
Substituting (6.78) in (-6.77), we write it down as
A'
(IV)
K + 2X"Y" + AT = (IV)
0. (6.79)
We shall eliminate the function X from here; for that it is suffi-

cient to require that ^< IV and X" be proportional to X:
>
X (IV}
= a ^; 4
X" =- X
2
*, (6.80)
where a 4 and K 2 are certain constants. Thus, the function X must

satisfy simultaneously two differential equations (6.80). Let us
examine in what conditions they will be consistent; for this pur-
pose we differentiate the second equation twice and, comparing
the result with the first equation, we obtain
The first equation (6.80) can then be replaced by
.and instead of (6.80) we consider the system
WX" = <*X\ X" =

It follows that
-^r =X 2
1 i.e., we may assume a = k and the system
(6.80) will be replaced by
* [IV1
= X *; 4
X" = -\ X. 2
(6.81)
Substituting this in (6.79), we eliminate X from it and obtain

-theequation of the fourth order for determining Y:
K fiv] _ 2X 2
Sec. 44] BENDING OF RECTANGULAR STRIP 183-
The function X will be found from the second equation of (6.81):
X" + tfX=Q. (6.83)
Its general solution is well known:
X= KI cos \x +- K 2 sin \x. (6.84).
We now concentrate attention on equation (6.82). Seeking its

particular solutions in the form
y = es \
we obtain the characteristic equation
(
5 2__X2)2^Q, (G.82a)
having two double roots

c
,s
_m 4-
Hence we obtain the general solution of equation (6.82) in the

usual way in the form
Y = A cosh Xy -j- B sinh Xy -j- Cy cosh Xy + Dy sinh Xy, (6.85V
where A, B, C, D are arbitrary constants.

Substituting the values (6.83) and (6.85) in (6.78), we find
the following particular solution of the equation of the plane
problem (6.77):
cp (x, y) (K cos \x
l + KZ sin \x) (A cosh \y -f- B sinh \y +
4-CycoshXy-f-DysinhXy). (6.8 )
This function is a solution of equation (6.77) for arbitrary values

of the constants
K K A, B, C, D and X.
Consequently, it is possible to construct any number of solu-

tions of the type (6.86); the sum of such solutions will also be
a solution of equation (6.77) (since this equation is linear).
Taking a sufficiently large number of terms of this sum, we
shall have many arbitrary constants at our disposal; later we
shall try to select them so as to satisfy the conditions prescribed
at the boundaries of the strip as fully as possible.
J84 PLANE PROBLEM IN CARTESIAN CO-ORDINATES (Ch. VI
First consider the conditions at the ends of the strip having

length / (Fig. 60). Let us require that at these sections, i.e.,
: = and x=l
no normal stresses occur; in other words, we require that
for x = (
There will still remain shearing stresses at the ends, and the
loads on the long sides of the strip will be balanced by the
forces Qo and Qi applied at the ends of the beam; these forces
*can be considered to be the reactions of the supports of a simple
beam of length /.
Conditions (6.87) will be satisfied if we set in the particular
solution (6.86)
where m is zero or a positive integer: w = 0, 1, 2, 3, . . .

, oo. Let
us form the sum of particular solutions taking
m=l, 2, 3, ..., /i
1
:
+ Cm y cosh ^L + Dm y sinh ^-] .

(6.88)
There appear here 4n arbitrary constants A m B m Cm D m

, , , which,
for sufficiently great number n, will enable us to satisfy the
a
conditions on the long sides of the strip if loads are prescribed
there. The solution in the form (6.88) was first proposed by
L.N.G. Filon.
We proceed to the conditions on the long sides. Let an arbitrary
normal load <7i = <7i (X) and a tangential load t i ti
(x) be given
=
on the upper side (y = h)\ let a normal load ?2 <72 (X) and a =
=
tangential load ^2 ^2 (x) act on the lower side (# = 0). As regards
the tangential loads, we assume that the sum of each of them
separately is equal to zero, i.e.,
t
, (x) dx = 0, f
t2 (x) dx = 0. (6.89)
1
The value m= (J is omitted for the time being and will be considered later.
Then we obtain the following boundary conditions for the stress

function:
u Y
e IL
Y ___.
Q
(6.90>
dxdy
or, substituting here the expression of y(x, y) from (6.88), we

get
mnx mnh
sin-
/2
Ji cosh 2^ + DJisinh 22* )==/,<*);

"i cos m
mn
_. + D~ m smh m7r/z
.
-y- 4-
\
J
.
t. . <
cosh smh (6.90a>
D n A cosh
2^.]
= /,(*);
-
mcos + C^ J =
Considering the second and fourth of these equations, we see
the necessity of the restriction imposed on the loads ti and t 2
by conditions (6.89); indeed, the corresponding integrals of the
left-hand members of these equalities taken between the limits
(0, /) vanish. In any case, the above equalities show that the
functions appearing in the right-hand members must be repre-
sented approximately in the interval (0, /) by Fourier series in
imtx , tnnx ,. f _,, , ,,
sin and cos The greater the number of
j -y- respectively.
terms in the stress function (6.88), the more exact this repre-
sentation. To determine the coefficients of the series we follow
the general rule: we multiply both members of the equality by
b-jry kllJC
sin
y
or
j- respectively
and integrate from
cos to /; in
this way we obtain the following four equations for determining

the coefficients A h B h Ck k appearing in the /eth term of the
, , ,
D
stress function (6.88) :
A k cosh kith + Bn k smh -y- + C k h cosh kith + Dr\ k h smh -p

- , , 1 knfi
=
knh .
s^ . i ,
. *
27 .
sin
Knrt knh
KT\* .
smh -j- + B cosh -y- ,

r>
k
/c
H- A smh
/ i- i i L.
/
(6.90b)
.
mh-^
,
h cosh
/JTC/l
-y-
= |-
. , ^7T/Z\
1
2
J
C 2
t{
/ \
(x) cos
k^X
-y-
~J
rf^:;
"Setting here fe=l, 2, 3, . . .

, ft, we obtain equations for deter-
mining all the coefficients of the stress function (6.88). With
Fig. 61
the stress function we can find stresses at any point of the strip
according to the formulas
*y x ~ _. dy*
'
vy _
~ jl- dx*
J
dxdy
'
The foregoing method of investigating the bending of a strip

proved to be very convenient since, when using it, we can intro-
duce sufficiently arbitrary loads on the long sides. In the courses
on the theory of elasticity by S. P. Timoshenko and P. F. Papko-
vitch one can find many examples of application of this method
to problems of practical importance.
If the tangential loads /i and t2 we have the problem

are absent,
of the transverse bending of the
strip by the usual transverse
loads q\ and q%. Consider a particular case when the load qi=f\(x)
is prescribed arbitrarily, while the load <72=/2(*) is distributed
over two portions (Fig. 61); its intensities q'2 and q"^ are chosen
so that all these loads are mutually balanced. Though there still
will remain tangential tractions at the ends of the beam, their
resultants Q and Q/ (Fig. 60) will vanish; the influence of these

balanced tangential tractions Yx according to Saint-Venant's
,
principle, will be appreciable only near the ends of the beam.

In this way we shall obtain the solution of the problem of a beam
resting on two supports, provided the reactions of the supports
#2 anc* 02 are distributed to a
according law. A num-
preassigned
ber of problems of this kind were solved by G. N. Maslov for
simple and continuous beams.
Another interesting particular case will result if we take the
loads qi and q 2 (Fig. 62) equal but opposite in sense (^ and t 2
are absent as before). The forces Qo and Q/ will be equal to zero
and we shall obtain the compression of the strip in the lateral
direction. If the loads qi and q 2 are distributed over a small
portion (Fig. 62a), then by contracting it we shall obtain in the
limit a scheme close to Fig. 38 which was used to illustrate Saint-
Venant's principle. In the present problem this principle is con-
188 _ PLANE PROBLEM IN CARTESIAN CO-ORDINATES
firmed by the shape of the diagram of stresses Y u over the middle

section ab, shown in Fig. 626 for the limiting case of compression
_ {Ch. VI
of a strip by concentrated loads. The solution of this problem was

given by Filon
l
.
Ribiere applied to some problems a solution analogous to

(6.88) and obtainable in the same way from (6.86) if we set
# = 0, A-
2 -21; thus we have
.
(6.91)
A series of applications of this solution are given in the course

on the theory of elasticity by P. F. Papkovitch.
In order to lift the restriction imposed on the tangential loads
by conditions (6.89) in Filon's solution and to broaden the limits
of its applicability in general, it is necessary to add to the solu-
tion (6.88) one more particular solution, resulting from the adopt-
ed method of separation of variables but so far omitted. Indeed,
in addition to the values ra= 1, 2, 3, ... used in (6.88) to construct
the function q>, let us set m = 0, i.e., X

-^-=0; then equations
(6.83) and (6.82) become
*" = (),
K (1V)
=O
their general solutions are
and according to (6.78) we obtain 2 the corresponding stress

function
(6.92)
Thus, the solution (6.88) complemented according to the

is
above scheme by the solution the form of a polynomial of
in
the fourth degree (6.92); the components of the state of stress
E. G Coker, L. N. G. Filon, A Treatise on Photoelasticity, Cambridge,

1
Univ. Press, 1931; S. P. Timoshenko, Theory of Elasticity, New York, McGraw-

Hill Book Co., 1934.
2
Henceforth we assume D=0.
Sec 45] ONE MODIFICATION OF FILON'S METHOD 189
will now be
(6.93)
We obtain the following stresses on the long sides of the strip
for y =
(6 ' 94>
Consequently, by a suitable adjustment of the arbitrary con-

stants a,p, 4, B, C we can:
1. Obtain constant tangential loads X v of arbitrary intensity
according to (6.94) and thereby get rid of the restriction (6.89)
on the long sides (y = 0, y = h).
2. Obtain normal stresses at the ends of the strip:
for x= Xx
for x=I X x = (a/ + p) (Ay + 5), )
(6 95)
'
distributed according to a linear law; this permits us to add two

more factors of arbitrary intensity to solution (6.88): compression-
tension of the strip along its axis and bending moments at the
ends.
*45. ONE MODIFICATION OF FILON'S METHOD 1
The modification presented here may be found useful in cases

where a given strip is to be designed under various loads and
also in cases where it is necessary to determine the displace-
ments of its upper or lower boundaries; the latter is required in
the solution of contact problems-, for instance, in the case of
a strip resting on an absolutely rigid foundation and loaded over
the upper face; sometimes it is required to calculate the displace-
ments v (deflections) of the lower and upper faces af a strip
supported at the ends.
To solve the problem stated in Sec. 44 it was necessary to
represent the stress function (p(x, y) in the form (6.88); the first
See the author's paper On the

1
Bending of a Strip, Vestn. Voen.-inzh. akad.

im. V. V. Kuibysheva, 1936, No. 20.
190 PLANE PROBLEM IN CARTESIAN CO-ORDINATES [Ch. V!
factor under the summation sign sin

j
is a particular solution
of equation (6.83); we shall leave it unchanged; the second factor

(in parentheses) represents the general solution of equation (6.82)
for X = j-
It is composed of four particular solutions
cosh^; sinh^; ycosh^; ysinh^ (6.96)
obtained by the use of the characteristic equation (6.82a); how-

ever, instead of these we can take some other four particular,
linearly independent solutions representing linear functions of
solutions (6.96); let us write them down in the form
*im (*)
= a \m cosh z + b }m sinh z + c }m z cosh z + d lm z sinh 2,
(*)
= a cosh z + 6 2m sinh z + c^z cosh z + d2m z sinh
<2m z,
= a 3m cosh z-f 63m sinh z + c3m z cosh z + d^z sinh z,
(6.97)
hm (*) = a *m cosh z + b4m sinh z+ cAm z coshz + d^z sinh z,
where for brevity we have introduced the notation
(6.98)
the quantities a\ m , &i m ,..., d 4m are so far arbitrary; we shall

choose them later in such a manner as to satisfy the boundary
conditions as simply as possible.
The general solution of equation (6.82), given by formula (6.85),
is now replaced by
(z), (6.99)
where A m Bm
, , Cm, Dm are new arbitrary constants. The stress
function (6.88) becomes
n
? = 2 sln T L}
2 -
'ni' (6.100)
In selecting the coefficients in formulas (6.97) the following

considerations will be used as a guide. The components of the
state of stress acting on the upper and lower faces of the strip
are represented by
WTC mnx *,'

cos Ym .
There appear here functions (6.99) and their first derivatives;

let us adjust functions (6.97) so that for */=0 and y~h the
expressions of Y m and Y'm will be monomials; for this purpose

we determine the coefficients of functions (6.97) from the follow-
ing boundary conditions (where p

= TC
-j-\
:
*i*()=l: *'(0) = 0: *?) = 0; *
(6.101)
1Ierc **m() and ^m^P) denote the derivatives of functions

(6.97) with respect to the argument, with the substitution of its
boundary values z= andz = m$ = ^^-. Introducingthe expres-

sions of functions (6.97) in these conditions, we obtain four
systems of equations for determining the coefficients
a km">
where the index k is given the values 1, 2, 3, 4 in succession.
These equations can readily be solved in general form; the fol-
lowing table presents the expressions of the coefficients thus
obtained:
These coefficients are independent of the load and can be cal-

culated once and for ever for strips with a
given ratio of the
dimensions h:l entering in the parameter p=r<ii:~-; calculations
are simplified by a pairwise equality of a series of coefficients
'1m
192 PLANE PROBLEM IN CARTESIAN CO-ORDINATES [Ch. V
The particular solutions (6.97) of equation (6.82) thus construct

ed permitus to obtain the following simple boundary valuer
of the functions Y m (y) and Y' m (y) at y = and y = h (in differen
tiating we take into account that ~-~ = H :
(6.102
This, in turn, results

a considerable simplification in the
in
boundary conditions a strip in comparison with those pre-
for
viously obtained (6.89). In fact, if the lower face (y 0) is
=
acted upon by the loads q<i(x) and t 2 (x) (Fig. 60), the corre-
sponding boundary conditions on the basis of (6.102) will be
written as
I
/mJt\2 -r- .
sin =
mnx
the conditions on the upper face y =h will be (6.103)
cos
Let us expand the prescribed loads in trigonometric series:
(6.104)
Sec. 45] ONE MODIFICATION OF FILON'S METHOD 193
in which the coefficients are found in accordance with Euler's

formulas:
(6.105)
sin
Substituting the expressions of the loads (6.104) in (6.103) and

comparing the coefficients in the right-hand and left-hand mem-
bers, we obtain
(6.106)
Hence, the coefficients of functions out to be pro-

(6.99) turn
portional to the coefficients loads (6.105)
of the expansion of
and their determination does not require the solution of a sys-
tem of equations [cf. equations (6.90b)]. The stress function
(6.100) takes the final form satisfying the conditions on the
upper and lower faces:
X (z) + B m ^m (z (z) (6. 107)
Z ~
_mny
I
The stresses are calculated by the usual formulas:
= - = .
sin
cos
Introducing here the expression of function (6.99), we obtain

the final formulas of stresses for a strip under given loads on
its long sides (in differentiation, as mentioned before, we take
13-1013
into account that
(6.108)
2 1 4,*;*
In the given strip under various loads, one has

design of a
only to substitute the
corresponding coefficients of expansion
(6.105) in formulas (6.108); the functions i|^ m (2), an d their deriv-
atives, however, are independent of loads. In most practical
problems not all the loads allowed for in formulas (6.108) are
involved, and the latter are simplified. For instance, in the
case of a strip supported at the ends and loaded over the upper
face (Fig. 61) one must set A m = B m = D m = 0.
The foregoing method of solving the problem possesses certain
advantages in cases where it is necessary to determine the
displacements of the points of a strip u and v, which is required,
for instance, in the solution of contact problems. The displace-
ments are found in the same way as in the case of the cantilever
considered in Sec. 41. Without presenting the appropriate course
of reasoning (see the work cited in the footnote at the beginning
of this section), we give some final results. The general formula
of the vertical component of displacement has the form
BT 14,
(6.109)
x
where Ffe m (2) (k=\, 2, 3, 4) denote the following antideriva-
tives of the functions t|) ftm (z) [see (6.97)]:
lp
*m (?) = -
I(**m c km ) cosh z + (akm -f- d t j sinn z +
Y is the angle of rotation and e is the vertical component of

rigid body displacement of the strip. Let us confine ourselves
to the case of vertical loads q\(x) and (J2(x) and, therefore,
assume subsequently that B m D
m ~0. The displacements of the
upper and lower faces of the strip are of prime interest in prac-
tical application; the following table gives the boundary values
of the functions yfhm(z) entering in formula (6.109):
By using them, we can easily find the displacement (deflection)

of the lower face (2 = 0):
v = 21 VI 1
(6.110)
If the load acts only on the upper face, we have
___ 2/_ V_L, . mnx

"
sin -.
(6.110a)
En 4*4m
With formula (6.109) it is possible to solve the problem of a

strip resting on an absolutely rigid foundation and subjected to
the load qz(x) over the upper face [formulas (6.104)]. In this
case 'we have the mixed problem (Sec. 37). If there is no friction
between the strip and the foundation, the conditions on the low-
er face will be: at 2
makes
= 0, = 0, ^= 0;
contact with the foundation. The solution of the prob-
full
it is assumed that the strip
lem leads to the following expressions for stresses:
= c- sin
(6.111)
cos
13*
196 _ PLANE PROBLEM IN CARTESIAN CO-ORDINATES
Setting 2=0 in the second formula,

the foundation exerts on the strip:
we
_
find the reaction
[Ch. VI
which
(6.112)
The displacement (settlement) of the upper face under the action

of the load is given by
The simplicity of the last formulas shows that the foregoing

method is practicable.
46. STRIP OF INFINITE LENGTH
The preceding sections have been concerned with a strip of

finite length /; at the same time it may be said that we deal
with an infinite strip, but one subjected to loads repeated peri-
odically in the portions of length / since the whole solution is
expressed in terms of periodic functions cos- and si
it possible to modify the solution obtained and

develops that it is
to extend problem for an infinite strip under an arbit-
it to the
rary load. With that end in view, let us return to the initial
particular solution (6.86) for the stress function (p(jc, y). By as-
signing discrete values X

= -^~-(m==l, 2, ..., oo) to the parame-
ter X, we obtain the problem solution
trigonometric of the in
series. Now we vary X continuously from
shall oo to +00; this
will lead to the solution of the problem in integrals. The coeffi-
cients A, B, C, D will also be considered as functions of A,; by
setting the coefficients Ki and Kz equal to unity, without loss in
generality, we write the solution (6.86) as
cp (x, y)
= sin lxY l (Xy) -f cos X*K 2 (Xy), (6.1 14)
where
Y l (Xy)
=A (X) +B
cosh Xy (X) sinh Xy +
+C (X) Xy cosh Xy +D (X) Xy sinh Xy,
(Xy)
== ^ cosh Xy +
(X) , (X) sinh Xy + (6.115)
+C (X) Xy cosh Xy -f- D (X) Xy sinh Xy.

Sec. 46] STRIP OF INFINITE LENGTH 197
The sum of solutions (6.114) will now be represented in the form

of the integral
+ 00 TW
<(.*:, y)= J cp<A= J /sir

(6.116)
obviously, it will also be a solution of equation (6.77). With the

stress function (6.116) we obtain the stresses
+ 00
X, = jp =
'
2 2
/ X cos \xYl (Xy) rfX +f X sin XjcK 2 (Xy) dX,
+ 00 + 00
yy =1^.
= Joo X2cosXjcK,(Xy)dX- J
X 2 sinXjcK2 (Xy)dX, (6.117)
oo
+ 00 +00
AT =
2 2
X sinXjcr,'(Xy)rfX-- X cosXjcK2(Xy)rfX.
y jj|y= jf f I
Here single and double primes denote the derivatives of the func-
tions Y\(hy) and Y^y) with respect to the argument "ky.
If loads <]i(x), ti(x), </ 2 (*), t 2 (x) are given on the upper and
lower faces (Fig. 60), the boundary conditions will be of the
form similar to (6.90a) (the limits of integration are omitted
for simplicity in writing):
for y = 0:
Y- y = <? 2 (x) = f cos IxY,

*2 (0) rfX +J X2 sin XjcK
2 (0) </X,
X_ y = < 2 (jc) = J X sin XjcKJ 2

(0) rfX +JX
2
cos Xjc^ (0) rfX;
for y = h:
[(6.118)
Y^= 9l (x)
= Jf X cos lxY (XA) 2
l
rfX
Jf
X 2 sin Xxr 2 (XA) rfX,
X =
y <, (jc)
= J X sin XjcKj (XA) dX
2
f
2
X cos XxK2 (XA) dX.
These conditions will be used as a basis for determining the

coefficients of functions (6.115):
C(X), D(X):
(6.119)
I, C(X),
198 PLANE PROBLEM IN CARTESIAN CO-ORDINATES [Ch VF
The boundary values of these functions enter in the integrands

of (6.118); they are to be found in the first place. In solving the
problem by means of series such an operation amounts to deter-
mination of coefficients according to Euler's formulas; in the
method under consideration, this requires the use of Fourier's
formula expressing an arbitrary function f(x) (satisfying the so-
called Dirichlet conditions) in the form of a double integral with
the integrand containing the same function f(x). The derivation
of this formula may be found in the courses on integral calcu-
lus; it has the form
rfX. (6.120)
00
Here x plays the role of a parameter; the inner integration is

carried out with respect to the variable a, and the outer with
respect to A,. Writing out the expression of the cosine, we put
(6.120) in the following form (the limits of integration are omit-
ted as before) :
/ X
( )
27 J COS Xjc J / (a) COS
[
Xa
flfaj
d\ +
sin X * / (*) sin Xa dL (6- 120a>
Its structure analogous to that of formulas (6.118), and

is from
their comparison we obtain the desired boundary values of func-
tions YI and V 2 and their derivatives. Indeed, assuming that
f(x)=Qi(x) in (6.120a) and comparing the result with the third
of formulas (6.118), we find
J <?, (a) sin
Setting now f(x) = ti(x) and comparing with the fourth formula
of (6.118), we obtain
4-co
\ r . ../
If
27lX 2
4-00
* ^
ZTtA^ J J
CO 00
In the same way, by the use of the first two formulas of (6.118)
we find four boundary values of the functions
Sec. 46] STRIP OF INFINITE LENGTH 199
The reader is recommended to derive these formulas as an exer-

cise. These eight boundary values of functions (6.115) and their
derivatives must be substituted in the left-hand members of these
equalities setting y
= Q or y = h, respectively, in the right-hand
members. This will give eight equations for the determination
of coefficients (6.119), and the boundary conditions for the strip
will be satisfied; it remains to determine the stresses according
to formulas (6.117).
The presence in the solution of integrals with infinite limits
requires certain restrictions on loads if they are not applied in
a portion of finite length but continue infinitely: in this case
their resultants must be finite; otherwise the integrals may loose
sense (become infinite). The integrals entering in the solution
are taken in finite form only in rare cases; one has to evaluate
them by the numerical method.
VII
Plane Problem in Polar

Co-ordinates
47. GENERAL EQUATIONS OF THE PLANE PROBLEM

IN POLAR CO-ORDINATES
Up to this point, in solving the problems of the theory of

elasticity we have used the Cartesian co-ordinates in which a
point M (XQ, j/o, 2o) is located by the intersection of three planes:
Fig. 63 Fig. 64
X=XQ, y=yti Z=Z Q Applying these co-ordinates we divide the

.
body under investigation into infinitesimal elements by three

systems of planes:
The surface on which one of the co-ordinates retains the con-

stant value is called a co-ordinate surface; in our case these sur-
faces are planes parallel to the co-ordinate planes, the resulting
elements being of the form of parallelepipeds (Fig. 63).
Sec 47] GENERAL. EQUATIONS OF PLANE PROB. IN POLAR CO-ORDINATES 201
In many problems, however, it is found more convenient to

select other systems of co-ordinates. In particular, consider cy-
lindrical co-ordinates in which a point (Fig. 64) is located by
three numbers: r, 9, z. In other words, a point is located, by the
intersection of the following co-ordinate surfaces: a circular
cylinder r r
=
a plane 9
,
=
0o passing through the axis Oz and a
=
plane Z ZQ parallel to Oxy. Accordingly, in this case we divide
the body under investigation into infinitesimal elements by the
following three systems of surfaces: a system of concentric cylin-
= = =
ders r r it r r2 ..., r r t-; a set of planes 6 61,
,
=
2
=
..., ,
Fig. 65 Fig. 66
6 = 0i containing the axis Oz, and a system of ânes z Zi, =

z=z 2 ..., z
,
=
Zi parallel to Oxy. The form of the resulting in-
finitesimal element is shown in Fig. 65.
Let us apply the cylindrical co-ordinates to the plane problem,
the axis of the prismatical body under investigation being paral-
lel to the axis Oz. The external load, according to our assump-
tions, will be parallel to the plane Oxy. In this case, as we have
seen, both for plane strain and for plane stress the axis Oz may
be removed formally with the result that the whole problem will
be solved as if in the plane, in polar co-ordinates r, 0.
Let us now derive the basic equations of the theory of elastic-
ity for the plane problem in polar co-ordinates. First we take
the differential equations of equilibrium (I). We
isolate from the
body an element abed with central angle d0 and the smallest
radius r. Its sides will be (Fig. 66)
ab =cd =
(7.1)
be =
202 PLANE PROBLEM IN POLAR CO-ORDINATES [Ch VII
We select two axes, mr and m0, and by using them designate

the stresses acting upon the faces of the element according to
the rule accepted at the beginning; these notations are shown in
the drawing. We set up the conditions of equilibrium of the
element abed by projecting the forces applied to it on the axes r
and 6. The body forces are neglected. The thickness of the ele-
ment along the axis Oz is taken to be unity. In projecting the
forces, in view of the infinite smallness of the angle dQ, we shall
assume
sin
.
= cos-2-=l.
1
Multiplying the stresses, indicated in the drawing, by the cor-

responding areas (7. 1) and projecting the resulting forces on the
axes r and 0, we find
, 4- + dr) - rfO /?, dr -
+^ dr - e., dr + (R, + ^ de) dr -f- +

(e, </e)
r + (e, + ^ dr) +
-f- (r dr) d6
- 9,rd9 = 0.
Cancelling out the obtained equations and disregarding infini-
tesimal quantities of the third order, we 'get
Further, dividing these equations by dr dQ and bearing in mind

that r
= /?fj, we
shall have the final differential equations of
equilibrium in polar co-ordinates:
r .
" Q
dr r d0
2/? (Ipp)
T "W "7"
For the equations of equilibrium in Cartesian co-ordinates (6.14)

corresponding to (Ipp), it has proved possible to find the gener-
al solution (6.16) expressed through the stress function cp(jc, y).
An analogous solution can be obtained for equations (Ipp) by
the use of the stress function <p(r, 9); there are several methods
of deriving this solution; however, we shall not present them
Sec 47] GENERAL EQUATIONS OF PLANE PROS. IN POLAR CO-ORDINATES 203
here; we write only the result:

1
r2
(VHIp)
'^
dr\r c
By substituting these expressions in equations (Ipp), we see that

they are satisfied identically. It is necessary, of course, to add
to them the condition of compatibility of strain components; in
Cartesian co-ordinates it has the form (IVp, Sec. 39)
V 2 (A^ +r =
y)
0. (IVp)
It is well to recall now that
represents an invariant under a transformation of co-ordinates

and, therefore, its magnitude will not change if we replace the
Cartesian co-ordinates by the polar co-ordinates. In the case of
plane strain (oy
= 0)
In the case of generalised plane stress (Z Z =Q)
Thus, X x +Yy will be an invariant in the plane problem and,

therefore,
On this basis we obtain directly from (IVp) the compatibility

condition in polar co-ordinates:
V2 (/? r +e =
o) 0. (7.2)
However, the invariant X X +Y V represents the Laplacian opera-

tor on the stress function in Cartesian co-ordinates:
= = yM*, y);
consequently, on the basis of (VIIIp)
is the Laplacian operator on the same function in

polar co-or-
dinates; hence the general conclusion: the symbol of the Lapla-
204 PLANE PROBLEM IN POLAR CO-ORDINATES [Ch. VII
cian operator in polar co-ordinates implies the following differ-

ential operation:
Consequently, the compatibility equation (7.2) in polar co-ordi-

nates has the form
Carrying out the operations indicated in the first parentheses,

we obtain the partial "differential equation of the fourth order
Fig. 67 Fig. 68
which must be satisfied by the stress function cp(r, 0); by the use
it we calculate the stresses
of according to formulas (VIIIp).
We now fix attention on deformations and displacements
(Fig. 67). Let us designate the displacements of points along
the r axis by u, and along the 9 axis by v, the elongation along
the r axis by e rr along the 6 axis by e^
\ and shear, i.e., the
distortion of the right angle bad, by e r^
By the same reasoning as before (Sec. 10), we obtain e rr = -p.

We proceed to the elongation e^. This elongation may be due to
two causes. First, if only radial displacements u occur, the unit
elongation of the linear element ad=rd0 is given by
(r + u)dti rdti _
~ u
'
rdb r
Due to displacements v along the axis there also results the

unit elongation
-r^f 5 the total elongation is
Sec 47] GENERAL EQUATIONS OF PLANE PROB. IN POLAR CO-ORDINATES 205
Let us now examine the shear. One sees from the drawing
(Fig. 68) that the shear is expressed as
*r , = (p-a) + 7 ;
(7.4)
next we note that
dv da du
Substituting these values in formula (7.4), we obtain

dv v . 1 du
Thus, in place of equations (III'p) for the plane problem we now

obtain the following system of equations:
du 1 dv
dv v 1 du (Hipp)
r
Hooke's law (V'p) for the case of generalised plane stress has
the previous form, only the notations of stresses and strains vary:
(Vpp)
"
2(1 + )
E
In the case of plane strain E and a should be replaced by
and GI according to formulas (6.5).
48. PROBLEMS IN WHICH STRESSES ARE INDEPENDENT

OF THE POLAR ANGLE
Let us apply the equations derived to the solution of some

problems. The problem is simplest to solve when stresses (VI lip)
are independent of the angle 9, i.e., when /? r e and /?o are the ,
same at all points of any circle with centre at the pole O.

The stress function (p can obviously be assumed to be independ*
ent of 6 in this case; equation (IXp) is then simplified:
Carrying out differentiation, we obtain

^!l_u^-^!l
*~
~dr<~ T Jr*
~ 7L^!l_i__L
2
"dr 7* 2 ~T~ (IX'p)
the expressions of stresses (VIIIp) become
~'' ~
(viirp)
Equation (IX'p) can easily be integrated by the use of the sub-

stitutionop
= r n which leads to the characteristic equation with
two multiple roots. As a result, the general solution of equation
(IXp) has the form
cp (r)
=A 1 n r -f- Br 2 1 n r + Cr + D. 2
(7.5)
Hence we find stresses according to (VIIIp)
(7.6)
By way of example, we shall use equations (7.6) for the solu-

tion of Lame's problem of a circular tube submitted to uniform
external and internal compression. Let the external pressure be
p 2 and internal pi.
,
Then, we obviously obtain the following conditions on the sur-

face (Fig. 69):
for r =a _r = -}-p l
or
for r= b (7.7)
To obtain Lame's solution in these conditions the arbitrary

constant B is set equal to zero; then equations (7.6) give
The constants A and C can easily be found from conditions (7.7);

by satisfying them we obtain
2
Pi
~
(7.8a)
a2
2L
+ !
b2 a2
Sec 48] PROBLEMS IN WHICH STRESSES ARE INDEPEN. OF POLAR ANGLE 207
In the solution of this problem we have made an assumption

by setting in advance B 0;
= with three arbitrary constants
A, B and C and twoconditions (7.7) we could solve the problem
under other assumptions as well; we can prove, however, that
Lame's solution (7.8a) is the one corresponding to the actual
stress distribution. A feature of this problem is the fact that we
encounter here a doubly connected contour
since the cross section of the tube is bounded
by two closed curves which do not intersect
each other; in the presence of a doubly connect-
ed or a multiply connected contour the problem
becomes more involved, general, and the in
solution may not be unique. This difficulty can
be avoided by applying either of the following
two methods.
First method. The solution of the
problem in Fig. 69
terms of displacements. Since stressesand
strains are independent of the polar angle 0, there remains only
the first of the equations of equilibrium (Ipp) in its simplified
form
T
~^ (7.9)
relations (Hipp) are also simplified:

u
dr
'
r (7.10)
Hooke's law (6.4) in the case of plane strain is
(7.11)
Hence
I u
- U 3)
i
substituting here from (7.10), we have

or,
n
*V -=E\ -
1
I du u
<s\ \ dr
+-
1
r dr
By introducing these expressions of stresses in the equation of

equilibrium (7.9), we arrive at the differential equation of the
second order with respect to u\ its general solution will involve
only two arbitrary constants, which are determined from the
conditions on the outer and inner contours. Hence we obtain the
same formulas for stresses (7.8) from (7.12). This method is

given in the courses on strength of materials.
Second method. The problem is solved, as before, in terms of
stresses. The equation of equilibrium (7.9) is supplemented by
the compatibility condition which is obtained by eliminating
the only displacement a from relations (7.10). The second of
these relations gives u = re b ^ hence ~ = ^ + r ~)r- Substituting

this in the first relation, we obtain the required compatibility
condition in the form of a simple differential equation of the first
order:
dr
Substituting here the values of strains from (7.11), we obtain

the compatibility condition in terms of stresses:
-=< 1
+<>) j5l î -
<
7 - 13a >
Unlike (7.2), turns out to be an equation of the first order

it
as the axisymmetrical character of the problem has imposed

certain restrictions on strains and stresses;
s^~ T^N. ^
e c ass of possible solutions of equation
l
(7.13a) is considerably narrower than that

i XX"* 7*Xx y
/y~ \tf /^X^-vy of equation (7.2); it does not involve the
\^^ solutions which do not correspond to the
O axisymmetrical problem and among these
Fig. 70 the solution corresponding to the value
B = in (7.5). Let us integrate the equilib-
rium equation (7.9) and the compatibility equation (7.13a); mul-
tiplying the former by (1+ai) and adding it to the latter, we
obtain
d8t dR r ^' ,
dr dr
whence
This enables us to eliminate o from (7.9), and it becomes
_
dR r
__l ,
2(/? r _ _
-C.)_
Integrating it, we find

n
Kr
_ ^r 1
_i
"1
which coincides with (7.8).

Sec 48] PROBLEMS IN WHICH STRESSES ARE 1NDEPEN. OF POLAR ANGLE 209
Consider a part of a circular ring bent by forces applied to

the end sections and reducing to couples (Fig. 70). The surface
conditions will be expressed as
for r =a R r 0;
for r == b R r
= 0;
b b
J
e (J
dr = 0;
fe>rdr
= M.
a a
Writing out these conditions, we get

= 0;
We have obtained four equations for determining three con-

stants A B and C; we note, however, that the third equation
9
is a consequence of the first two equations. Solving these equa-

tions, we find
. _
C= -JSJ-
(b
2
a2 +2 (b
2
In b a2 In a)],
where
N = (b* a 2 )2 4a 262
(in ^ .
Introducing these values in formulas (7.6), we obtain the

following stresses:
4M ( a*b* t
b
(7.14)
Comparing the stress 89 as computed by formulas (7.14) with

that given by the elementary theory of rods of large curvature
(the hypothesis of plane sections), one can see that the difference
obtained is not great; it must be attributed to the fact that the
elementary theory does not take into account the stresses R r
caused by the mutual pressure of individual curvilinear longi-
14-1013
tudinal fibres; these stresses produce additional deformation of

a rod l
.
Next consider in this problem the change of the central angle

dQ of an infinitesimal element abed (Fig. 71). Since shear de-
formations are absent (6,- = 0) in this case, the change of the
angle 6(d0) depends on two elongations: e tr and e^. Figs 716'
i
___
----
a P '
v
h '
'
Effect of Effect of
elongation err elongation egg
Fig- 71
and c show clealrly the effect of both of these factors; we obtain

from them the effect of the radial elongation e rr (Fig.
a' a"
ca'
= - -
e rr dr rfO
dr-\- e rr dr
fi
716):
the effect of the axial elongation e^ (Fig. 7lc)

bb' dd'
but
W _ = e^ + dd' -
dr (r + dr) d6 e^
therefore,
Thus, the total change of the angle d0 is
8(rfO)
= 51 (rf
and the unit change is
The reader will notice that in a similar case of pure bending of a

1
straight rod the stresses X,/, corresponding to the stresses R r of the above case,
were absent [formula (6.28) of Sec. 40]; consequently, there is no mutual
pressure of individual fibres in a straight rod under pure bending
Sec. 49] EFFECT OF CONCENTRATED FORCE 211
If we substitute here in strains their expressions in

place of
terms of stresses from (Vp) 47 and make use of equations
of Sec.
(7.14), we can easily see that en is independent of the radius r.
This will evidently show that plane cross sections remain plane
in the case of pure bending, thus supporting the hypothesis of
plane sections which is usually adopted in the elementary theory
of curved rods.
49. EFFECT OF A CONCENTRATED FORCE

(FLAMANT-BOUSSINESQ PROBLEM)
Consider (Fig. 72) a homogeneous elastic medium bounded by
a plane AB and extending indefinitely down from this plane.
Let a force P be applied at a point C of the medium; with regard
to the extension of the medium in the direction normal to the
plane of the drawing there may be tw o cases as always in the
r
plane problem: either this extension is very small (generalised

plane stress) or it is unlimited (plane strain); in the latter case
the force P must be assumed to be not concentrated but uniformly
distributed along a straight line normal to the plane of the draw-
ing and passing through the point C.
It is evident that the stresses will be very great and will exceed
the elastic limit near the point C; we shall, therefore, consider
only points lying outside a small region bounded by a circle
mn of radius p; in view of this, by using Saint-Venant's prin-
ciple we replace the load P by a load distributed along a semi-
circumference mn and equivalent to the load P\ this will
have little effect on the stresses at points remote from the
circle mn.
Letus attempt to satisfy the conditions of the problem by
the following simple assumptions:
(1) there are no shearing stresses at any point M
on the area
normal to the radius = r;
CM
(2) compressive normal stress R r on this area is inversely pro-
portional to the radius r and directly proportional to cos 6;
(3) normal stresses 60 on areas in the radial direction are
absent.
These assumptions lead to the following system of stresses:
(7.15)
14*
These stresses satisfy the conditions of the problem on the
surface AB\ indeed, the bounding plane AB (6= ~, cos Q = Q\

is free of stresses except for the point C where R r assumes the
indefinite
form-g-;
but the point C and even the small semicircle
mn have been excluded from our analysis. Furthermore, the coef-

ficientk in equation (7.15) can easily be selected so that the
normal stresses R r along the semicircumference mn will be equiv-
alent to the given load P, i.e., so that the sum of the projections
of the forces
along the semicircumference r =p on the axis Ox will be equal

to P:
70
+
7
f /?_,pcosflde =A (7.16)
Substituting here the value Rr from (7.15) and carrying out

integration, we easily obtain
op
k = lf. (7.17)
The stress distribution as defined by formulas (7.15) is some-

times called a simple radial distribution of stresses.
In order to finally accept the above-constructed distribution
of stresses (7.15) it is necessary to check whether it satisfies the
differential equations of equilibrium (Ipp) and the equation of
compatibility (IX'p). Since 6j =/?g =0, the equations of equi-
librium reduce to a single equation
Substituting in it the value of Rr from (7.15), we see that it is

satisfied identically. The equation of compatibility (IX'p) as-
sumes a simple form
d/? r-
Substitution of the value of R r in it shows that it is also satis-

fied identically.
Sec. 49] EFFECT OF CONCENTRATED FORCE 21$
*
It may be shown that stresses (7.15) are obtainable according
to formulas (VIIIp) by the use of the stress function
<f
= j-r^s\n^. (7.19)
In fact, these formulas, after the substitution of values (7.15).

are transformed into a system of differential equations:
1 d<p
lr W~T~~r*~dr*
|^
1 d 2 <p
== ~ k ~r~
, cos 9
t
'
dr*
"-"' dr
The last two of these equations will be satisfied if we set 9 =

= r/(6). Then the first equation will be transformed into an ordi-
nary differential equation
cp(6)
= -
and then
cp
= rf (6)
==
-|
/-0 sin + Ar cos + Br sin 0.
The last two terms can be disregarded since in Cartesian co-or-

dinates they represent a linear function
Ar cos + Br sin = Ax + By,

which has no influence on the stresses.
We have proved that stresses (7.15) satisfy all equations of
the theory of elasticity: on the basis of (7.17) they are ex-
pressed as '
n 2P cosO^
(7.20>
Their investigation leads us to the following conclusions:

(1) the area at any point M
(Fig. 72) which is normal to the
radius r is principal since there are no shearing stresses on it;
(2) let us construct a circle CMx (Fig. 72) passing through the
given point M
and let its diameter be C#=d; then
but
1214 PLANE PROBLEM IN POLAR CO-ORDINATES [Ch. VI [
therefore
cos
substituting this in (7.20), we find
(7.21)
Thus the stress R r is the same at all points of the circumference

concerned. Hence each circle with centre under the load P and
A 7
8 ^
m C /7
Fig, 72 Fig. 73
passing through its point of application C is a trajectory of equal

stresses Rr l
.
Let us now find the stresses in the plate on a horizontal area

at a distance x from the surface. From the conditions of equi-
librium of elementary prisms abc and aibî (Fig. 73) we obtain 2
Yx =R r
sin0 cos 0, (7.22)
j
Yy =R r s\n*Q. )
Introducing here the value of R r from (7.20) and taking into

account that
X
(7.23)
+y 2
we have
x
^ _. y2)2
2P
y
22
2
)
'
(7.24)
1
We may recall that the points inside the semicircle run are excluded.
2
Cf. formulas (1.15), of Sec. 4.
Sec. 49] EFFECT OF CONCENTRATED FORCE 215-
Fig. 74 shows
the diagrams of these stresses. In practical cal-
culations, itthe custom to distribute the action of the load
is P
at a given depth x under a certain angle; in the conditions of the
problem considered we have
Thus, for calculation one may assume conditionally that the load
P is distributed uniformly over an area about 1.6* long which
corresponds to an angle of 0~38 (Fig. 75).
diagram
diagram
Fig. 74 Fig. 75
If there are several loads applied at different points of the

straight line AB (Fig. 73), we shall readily find stresses (7.22)
Fig. 76
at any point by summing up the effects of individual loads. Fur-

ther, the problem can easily be extended to cover the case of
any continuous load (Fig. 76). If the intensity of the load at a
given point is p, the load acting on an infinitesimal element dy
of the straight line ab will be pdy, but we see from the drawing.
'216 PLANE PROBLEM IN POLAR CO-ORDINATES [Ch. VII
that
rdft
and the element of load is

prdti
cos (7.25)
We substitute this quantity for the load in the first of equations

(7.20) and get
Introducing this value of R r in equations (7.22), we obtain the

caused by one element of load (7.25):
.stresses
= cos 2
and similarly
The meaning of the angle 9 is clear from the drawing (Fig. 76).
If load extends from point a(9 = 9i) to point &(9 = 9 2 ), by
the
.summing up the stresses due to its individual elements, we obtain
the final expressions of stresses:
(7.26)
One sees from these formulas that it is necessary first of all to

represent thf load p as a function of the angle 9. If the load is
uniform, p is constant; putting it before the integral signs in
equations (7.26), we readily evaluate the latter and obtain
(7.27)
Sec. 50] WEDGE LOADED AT THE VERTEX 217
These formulas can be transformed to the Cartesian co-ordi-

nates x and y, taking into account equations (7.23) and adding to-
them
) = -^- or j
= arc tan .
The problems concerning the stresses in a medium bounded by

a plane are gaining in importance nowadays in the theory of
bases and foundations for the kinds of soil which maintain, under
moderate pressures, the properties close to those of a homogene-
ous elastic body.
50. WEDGE LOADED AT THE VERTEX

Solution (7.15) can be applied to a more general problem, in-
troducing in it an additional arbitrary parameter by replacing 9
by (9 9o) where 9 is so far an arbitrary angle. We obtain
then
The first term in the right-hand member, as we already know,

satisfies all basic equations of the plane problem. The second
Fig. 77
term satisfies them too, since it is obtainable from the first term
by rotating the polar axis through 90, i.e., by replacing 9 by
Consider (Fig. 77) a wedge to the vertex of which an arbi-

trarily directed force P is applied. Let the polar? axis Ox be di-
rected along the axis of the wedge. The constants k and & (7.28)
are chosen, just as in Sec. 49, from the condition that the normal
forces
along the arc mn of any radius p are equivalent to the given

force P. Projecting all forces on the axes Ox and Oy, we get
a
-a
-fa
/
Substituting here the value R. r = R r from (7.28), we obtain

ihe following equations for the determination of k and :
-fa -f-a
k cos 6
J cos 2 dQ +k sin 6
J sin cos 6 rfO = P cos p,
k cos 6
Ja cos sin dQ +k sin
J sin 2 rf6 == P sin p.
of
Evaluating the integrals entering here, we find

-I- a
Ja sin cos 6d0^0,

fa
A= f cos 2 8 rfO = 1 (2a 4- sin 2a), (7.29)

a
-fa
B= J sin 2 Gaf6 = l(2a sin2a).
Introducing this in the preceding equations, we obtain

Ak cos == P cos p; Bk sin 9 = P sin p.
Hence we readily find that
*-// (7.30)
Examining formula (7.28), we observe that it has the same

form as formula (7.15), but the polar angle in it (6 6 ) is
Sec 50] WEDGE LOADED AT THE VERTEX 219>
measured from a certain direction 6 (Fig. 77), defined by the

second formula of (7.30); since in general
B
then =f= P, and this direction is not coincident in the general
case with the direction of the load P. If we put 0. = -^, we shall
return to the "half-plane" considered in Sec. 49, but loaded by
Fig 78
an inclined force (at the angle to the normal to the plane);

then from (7.29) and (7.30)
-n>
Hence
2P cos(0
R r
R
Here (6(3) is (Fig. 77) the polar angle measured from the
direction the load P. Thus, solution (7.20) can immediately
of
be applied to the case when the load P (Fig. 72) acts not nor-
mally to the plane AB
but at any angle; one chould only re-
member that the polar angle 6 is always measured from the di-
rection of the load P. Consequently, this solution can be em-
ployed in the case of a load applied in the direction AB\ the pre-
vious solution (Sec. 49) also applies to the problem of a con-
tinuous tangential load over the plane AB. Combining the cases of
normal and tangential loadings, we can solve the problem in-
volving any inclined continuous load.
As illustrative examples of the analysis of a wedge consider
two cases: compression of a wedge (Fig. 78a) and bending of
a wedge by a force applied at the end (Fig. 786).
220 _
In the first
PLANE PROBLEM IN POLAR CO-ORDINATES
case
J3
=
_ [Ch. VII
and from equations (7.29) and (7.30) we have

A f\ 1
op
**
These values of k and 6 must be introduced in formula (7.28).

Similarly to (7.22), one can find stresses over the section of
the wedge mn (Fig. 78a) at a depth A: O the normal stress is ;
If CD (Fig. 78a) denotes the angle of inclination of the radius-

vector of a point (X Q y) in the section mn, the preceding for-
,
mula can be written down as

^ = ~~^ k 1
where
In the case of an acute wedge (angle a is small) the stress

Xx deviates slightly from the value
X* X
= (2<x + sin2a)
2/>
jt
'
i.e.,the stress distribution over the section mn is nearly uni-

form. If, however, the wedge is obtuse (angle a is not small),
the stresses are distributed highly nonuniforrnly; for instance,
at a = 45 X x varies
t
k i k
from to -.
In the case of bending (Fig. 786) we have
.and from equations (7.29) and (7.30)

2P
ft
o
-
2
A
K ~ 2a-sin2a
Prom (7.28) we have
* li i-=-*
But (y is the polar angle measured from the direction of
0j
Ihe load; consequently, solution (7.15), where k ha* the above
Sec. 50] WEDGE LOADED AT THE VERTEX 221
value, holds true in this case as well. The stresses over the sec-
tion mn, as previously, will be found similarly to (7.22) by the
use of equations (7.15) but bearing in mind that the angle 61,
as we have just said, should be measured from the direction of
the load as shown in Fig. 786. The stresses over the cross sec-
tion mn in this case are expressed by
(2a sin2a)
(7.32)
where M = Py is the magnitude of the bending moment at the

section.
If the angle of the wedge 2a is small, then expanding sin 2a
in a series, we obtain approximately
After some transformations the formulas of stresses can be rep-

resented as
v
Y, = --MX /tan ot\3
( )
.
sm<,
.
<7 ' 33>
where h is the height of the section mn, J is its moment of iner-

a" a
tia. For small a the quantity is close to unity; further-
more, since
then for small a the angle is close

toy everywhere and
sin0~l. It is easy to see, therefore, that the magnitude of the
normal stress Y y in equation (7.33) is close to that given by the
elementary solution. The distribution of shearing stresses differs
essentially from the elementary result according to the formula
one sees from equations that they vanish on the axis of

(7.33)
the wedge (at #=0) their maximum value at the
and reach
edges of the section. If, however, in place of the plane section
mn we make a section of the rod by a cylindrical surface shown
in the drawing by the arc m'n' with centre at C, then there will
apparently be no shearing stresses in this section at all.
51. GENERAL SOLUTION OF THE PLANE PROBLEM

IN POLAR CO-ORDINATES
Let us seek particular solutions for the basic differential equa-

tion (VIIIp) of the plane problem by using the method of sepa-
ration of variables applied to equation (IX) in Sec. 44:
= 0, (7.34)
where for the polar co-ordinates
its solutions will be sought in the form
cp
= r/?9, (7.36)
where R is one variable r, and

a function of is a function of
one variable 0. The factor r in (7.36) has been introduced for
convenience in carrying out further computations.
For substitution in (7.35) and (7.34) we calculate
Adding up these equalities by members, we obtain
V2<? = --0"+(r#"+3/?' + 4) e -
<7
- 37>
or
V2 2,
where
<h=>>e", (7.38)
,fc
= r(/?" + 3-. + (7.39)
)e.
Hence
V 2 ( V 2 cp) =V +V 2
<h
2
<j; 2
.
(7.40)
But the expressions of tpi and i|? 2

have the same structure as q>
in (7.36); therefore, we shall calculate
V2 and
Sec. 51] GENERAL SOLUTION OF PLANE PROBLEM IN POLAR CO-ORDINATES 223
according to the former formula (7.37), when calculating

if
n
we replace /? by -^ and by 0", and when calculating V
2
\|? 2
R r
R
we replace R by /?" +3 -f--^- leaving 6 unaltered.
In this way we obtain
It remains to differentiate the expressions

_*. and
.asindicated by primes in brackets and substitute the results in

(7.40). Having performed these computations, we write down
equation (7.34) in the following form (multiplying both of its
members by r3 ) :
/?0
IV
+ 2AQ" + BQ = 0, (7.41)
where the following notations are introduced:
(7 ' 42)
In equation (7.41) the variables can easily be separated; divid-

ing both of its members by /?, we find
e+.e=o; (7.43)
by partial differentiation with respect to r we obtain
2
(i)
V +(!)' e = -
(7-44)
First consider the case when
f
\
6. we
(A -n-}
get
2 -
224 PLANE PROBLEM IN POLAR CO-ORDINATES (Ch. VII
here m is a constant number. The variables are separated:
e"-f/tt
2
e = 0; (7.46)
(4)'-
2m *
(4)'
=0 -
( 7 46a >
-
The general solution of (7.46) is known:

= Dml cos m8 + Dm2 sin mO. (7.47)
Further we have
Substituting this in (7.41), we have
or, writing out this by the use of (7.42), we obtain
/?' V + 6/?'V + (5 2m 2
) /? r
2 2
(
1 + 2/n 2
) /? V+ ( 1 m 2 2
) /? = 0. (7.48)
The solutions of equation (7.48) are sought in the form
/? =r n
.
(7.49)
Introducing this in (7.48) and dividing it through by we rn ,
obtain after simplification the following characteristic equation:
Its four roots are
/i 1 = mH-l; n2 =m 1; #3 = /rc + 1; /z 4 = m 1. (7.50)
Let A/I take on positive integral values; if and =

1, m =
m
then there are no multiple roots among (7.50), and we obtain
from (7.49) four linearly independent particular solutions of equa-
tion (7.48); then for = 2, 3, 4, ... we obtain the following gen-
m
eral solution of this equation:
where C mft are arbitrary constants. Introducing this and (7.47) in

(7.36) and summing up with respect to m, we get
sin /6). (7.51)

Now consider the cases when (1) m = 0, (2) m=l. In the first
case equation (7.46) is replaced by 6" = 0, whence
e =D +D 01
6 02
.
The roots of (7.50) are multiple: n 3 =l; n2 1

=n = /i 4 = 1; they
give four independent solutions of equation (7.48):
r\ rlnr;
i lit
;
ylnr.
/? = C r+C01 02 r In r +C 03 y+C 04
~ In r.
According to (7.36) the function cp is
To
= (<V + C 2
02 r
2
In r +C +C 03 04 In r) (D01 8 +D 02 ). (7.52)
In the second case (w=l) the general solution (7.47) of equa-

tion (7.46) has the form
and roots (7.50) become: AZi

= 2; n4 = 2; n 2 ==Az 3 ==0. There are
four independent solutions of equation (7.48) corresponding to
them:
r2 ; 1; Inr.
-^-;
The general solution of equation (7.48) is now

R = Cnr 2
+ C^ -f C + C 13 14 In r.
The function <pi, according to (7.36), is
Ti
= (C u r + C y + C r + C
3
12 13 14
r In r) (D n cos fl
+D 12 sin 0). (7.53)
One may put here, however, Ci 3 = since rcos6 = A;, rslr\Q = y\

consequently, there results a function of the first degree in Car-
tesian co-ordinates.
Thus if, according to (7.45),
this leads to three types of solution: (7.51), (7.52), (7.53).

Let us now turn to the case
(4)
/=0 or A=c & (
7 -54)
where Ci is an arbitrary constant. Equation (7.44) takes the form
15-1013
V
(B =0,
I
i.e.,
B = c2 & (7.55)
or = 0; but this trivial case is of no interest and we omit it
because, according to (7.36), it leads to the solution
cp
= 0.
On the basis of (7.54) and (7.55) the basic equation (7.41)

reduces to
+c e=o. 2 (7.56)
We add here (7.54) and (7.55):
A ?!/?
= 0; B CiR = 0,
and, writing them out with the use of (7.42), we obtain
"V 3 + 5/?"r 2
R'
Here Ci and c 2 must be chosen so that equations (7.57) will have
solutions in common, and later use can be made, of course, only
of these common solutions; let us seek them in the form
R = rm .
(7.58)
Substituting this in (7.57), we obtain two characteristic equa-

tions:
/&2+ !=<?!; (//i
2
1)
2 = <?
2
.
Introducing these values of GI and c 2 in (7.56) and (7.57), we

reduce them to the form
-le = 0, (7.59)
^V + tf'r-^/Ô,
V /?V + = 2 2
1) ]/? 0.
\
( W >
[1 (/rc j
The solutions
r
m and r~
m
(7.61)
will satisfy both equations (7.60) (since the latter are not altered
when m is replaced by m). The solution of equation (7.59) is
sought in the form =e ft8

. The characteristic equation is
whence
2m = (m l)*
= (m I)
2
/
2
,
If
k=(m 1)1. (7.62)
'
//iÔ; m+\, (7.63)

then (7.62) gives four different roots and (7.59) has four partic-
ular solutions of the form
cos(/7i+l)0; sin(/7t+l)0; cos (m 1)0; sin (tn 1)0. (7.64)
Combining them with (7.61), we observe that we obtain, accord-

ing to (7.36), the values of the function cp coinciding with those
of which the previously obtained solution (7.51) is composed;
consequently, in the case (7.63) our investigation will give no
new solutions.
remains to investigate the cases: (1)
It w= 0, (2) m=l.
In
examining these cases, we shall seek again, for greater clearness,
the solutions of equations (7.60) in the form
JR =r n
.
We obtain the following characteristic equations:

n2 = m*, \
(/I
2
1)2
= (/H* 1)2; J
(7 ' 65 >
they have a common solution: n^ = m\ n 2 = m. For m = 0, fti

=
= ft 2 = we have two particular solutions:
r=l; rlnr = lnr.
Consequently,
/? =C +C 01 02 lnr. (7.66)
Equation (7.59) will be written down as
equation & + 2& +l=0 has two double roots:

4 2
Its characteristic
ki = i\ &2 = i which give four particular solutions:
cos 6; sin 6; Ocos0; sin 6. (7.67)
By using (7.66) and (7.67), we construct the stress function
<p
= (D 01 cos 6 + Z)02 sin +D 03 cos +D 04 sin 0) (C01 r +C 02
r In r).
In comparison with (7.53) there appear new solutions here:
To
= (Q)ir + C 02 r In r) (D 3 cos +D 04 sin 0). (7.68)
We proceed to the second case m=l. Equations (7.65) have the
roots rti=l, n 2 = 1 which
give two solutions:
r,
7. (7.69)
15*
Equation (7.59) will be written down as
e= D n cos 20 + D 12 sin 20 -f Dj + DH .
(7.70)
Expressions (7.69) and (7.70) give the following stress function:
<P
= (^V + C 2
13 ) ( Ai cos 29 + D 12
sin 20 +D +A
13
6 4 )-
Comparing this with (7.51) and (7.52), we notice that we obtain

no new solutions here.
Summarising the results of the integration of the basic differ-
ential equation of the plane problem (7.34), we obtain the final
value of the stress function as the sum of solutions (7.51),
(7.52), (7.53), (7.68). It is obvious that one can set all coefficients
D equal to unity in formula (7.51) and others listed above
without loss in generality; each term of the sum (7,51) can be
replaced by two terms collecting the coefficients of cos m6 and
sin w0; the constants C mh (k=l, 2, 3, 4) may be taken different
in the two cases; the same applies to formulas (7.52), (7.53)
and (7.68).
<P
= 1<PO + + + + +
2fo l<Pl 2<Pl 1?0 2
where
cp
=C r +C r01
2
02
2
In r +C +C 03 04 In r,
!CPO
= (Coir + Co r 2
2
2
In r + Co + Co 3 4 In r) 0,
Ti
=
To
= (
coi r + C^r In r) cos 0, (7.72)
ffo =(
rlere one can set C 3=C 1 3=C / 13 = 0, since the expressions of the
unction <p corresponding to these constants yield no stresses. The
expression of function (7.71) was found by J. H. Michell except
for the terms

2
Co2 r lnr-0; In r- 6; C^r Inr 6 cos 0; C02 rlnr sin 8, (7.73)
which did not enter in the expression given by him l

.
The components of the state of stress corresponding to these

terms are in polar co-ordinates:
For cp
= Co4 In r0
= Co4 Inr 1
For cp
=C 02 r In r sin
* =
/? e = C 02
(y
sin +7 cos
0)
.
For cp
=C 02 r In r0 cos
->
?r
= C^/ln
(7 COS
02
6
r\ 2 Inr .
Sin
n\
0J
,
/?o
^ Q2 (7 cos - sin
0)
.
If the pole of a system of polar co-ordinates lies within the

body under investigation, all these functions are multiple-valued
and cannot be used; in a simply connected region, however, these
components are single-valued, if the pole of the co-ordinate
system is placed outside an elastic body or at its boundary.
From solution (7.71) in the present chapter only some of the
first terms have been used. In fact, in both problems of Sec. 48
!
The most complete list of particular solutions of equation (7.34) contain-
ing both these solutions and two more solutions of the form cos (m In r)
cosh m0, r 2 cos (m In r) cosh m6 may be found in the book by C t B. Biezeno
and R. Grammel, Technische Dynamik, Berlin, Springer, 1939.
the solution is based on the function [cf. formulas (7.5)]
i<Po
=C ol r 2 4- C02 r 2 In r + C04 In r.
In problems of Sections 48-50 the functions icpo and 2 <po are used
partly [cf. formula (7.19)]. The remaining terms of function
(7.71) enable us to solve many important problems related to
the analysis of a wedge and closed ring The stresses correspond- l
.
ing to solution (7.72) will be found according to formulas (VIIp).
1
See S. P. Timoshenko, Theory of Elasticity, New York, McGraw-Hill
Book Co., 1934 and P. F. Papkovitch, Theory of Elasticity, Moscow, Oborongiz,
1939.
VIII
Torsion of Prismatical
Bars and Bending
52. TORSION OF PRISMATICAL BARS

Consider the problem of torsion of a prismatical or cylindri-
cal bar(with the cross section of arbitrary shape) by couples
lying in the planes of its extreme sections. The influence of the
own weight of the bar is neglected, i.e., we assume =Q X=Y=Z
in equations (VII) (Sec. 25).
The problem will be solved in terms of displacements (Sec. 25)
by the use of Saint-Venant's semi-inverse method, i.e., a
of displacements will be prescribed and
the remaining ones will be found from
Lame's equations (VI) and surface
conditions (II) or (Via).
We take the axes of co-ordinates ac-
cording to Fig. 79 and, following Saint-
Venant, assume that displacements u
and v have the same values as in the
case of a bar of circular cross section
[formulas (5.19) of Sec. 32]:
u = ry
1} = TJC2.
(8.1)
The assumption w = made for the cir-

cular section has to be rejected as we
introduce there by the hypothesis that Fig. 79
the plane cross sections of a bar remain
plane in torsion; in the meantime such a hypothesis in the general
case of an arbitrary cross section cannot be true l
accordingly,;
we assume that
, y), (8.2)
1
This will be proved at the end of the present section.
232 TORSION OF PRISMATICAL BARS AND BENDING [Ch. VIII
where the function (p defines the form of the warped surface of

the cross section; it is called the torsion function. We take it to
be independent of 2, i.e., we assume that the warping is the
same for all cross sections.
It is necessary first to check whether assumptions (8.1) and
(8.2) satisfy Lame's equations (VI) and to determine the function
cp so as to satisfy the conditions on the surface of the bar cor-
responding to the above-stated problem of its torsion. For substi-

tution in equations (VI) we calculate, by using (8.1) and (8.2),
du ,
dv
"57+17
(8.3)
The first two of equations (VI) are satisfied identically while the
last equation becomes
0. (8.4)
This is Laplace's equation in the plane; consequently, (p(*, y)

should be a harmonic function. For its determination it is neces-
sary to add surface conditions (II) to (8.4):
Xv = Xx cos (vx) -f- Xy
cos (vy) -f- X z cos (vz),
'
Yv = Yx cos (vx) + Y y
cos (vy) + Y, cos (vz), (8.5)
Zv = Zx cos (vx) + Z y
cos (vy) -f- Zz cos (vz). <
By using (8.1), (8.2), (8.3), Cauchy's equations (III) and Hooke's

7
law (V ), we calculate the components of the stress tensor:
]
!L
i *-
o
^-M
1
i
(8.6)
*.-<(*
(8.7)
Thus on our assumptions (8.1) and (8.2) only shearing stresses

act at cross sections, their distribution being the same for all
sections.
Consider the boundary conditions (8.5), applying them first to
the lateral cylindrical surface of the bar which is free of stresses
Sec. 52] TORSION OF PRISMATICAL BARS 233
according to the conditions of the problem, i.e., X V =Y V = Z V = Q.

On this surface
=cos-|-=0. Taking
(Fig. 79) cos(vz) into account
now (8.6), we observe that the first two of conditions (8.5) are
satisfied identically and the third assumes the form
X z cos (vx) +Y z cos (vy) = 0.
The meaning of this condition is very simple; if T denotes the

total shearing stress at a given point, then
X =T cos (tx)\
Z Y Z =T cos (ty),
and we obtain from the preceding equality
T [cos (tx) cos (vx) + cos (tY) cos (vy)]
or simpler
7* cos (fa;)
= 0, i.e., cos (fa;)
= 0.
Hence the shearing stress at a point on the contour of the

section is parallel to the tangent to the contour of the section at
Fig. 80
this point; this condition follows directly from the law of reciproc-
ity of shearing stresses since the lateral surface of the bar is
free of forces.
Let us now express the condition on the contour of the section
in terms of the torsion function <p by the use of (8.7). Multiplying
by the differential element of arc of the contour ds we obtain l
,
-y cos <**) ds + +x cos (*y) ds = -
(3*
But from Fig. 80 we have
ds cos (vx) = dy\ ds cos (vy) = dx,
The arc is assumed to grow,
1
i.e., ds>Q, when moving to the left as
indicated by an arrow in Fig. 80.
and therefore from (8.8)
or
*Ldy-$dX = X dX + ydy. (8.9)
This is the boundary condition (condition on .the contour of the
section) which should be added to (8.4) for the determination of

the harmonic function <p (x, y).
The condition on the contour of the cross section (8.8) may be put in the
alternate form; we write it as
-
cos (vx) + -- cos (vy) = y cos (vx) x cos (vy).
The left-hand member represents the derivative of the function (p along the
normal to the contour [cf. derivation of condition (Via) in Sec. 25]. To trans-
form the right-hand member we observe that from Fig. 80.
cos (vx) = cos (ty)\ cos (vy) = cos (tx),
where / is the tangent to the contour of the section, furthermore,

x r cos (rx\
y = r cos (ry),
where r is the radius-vector of a point M of the contour.
Therefore,
y cos (vx) x cos (vy) =

= r [cos (rx) cos (tx) + cos (ry) cos (ty)] = r cos (rt) = MN,
where MN (Fig. 80) is the projection of the radius-vector r on the tangent t.
On this basis the condition on the section contour (8.8) becomes
*L = rcos(rO. (8.10)
This condition defines the value of the derivative along the normal
at each point of the contour; thus, the problem of finding the torsion function
cp(#, y)} is reduced to the classical Neumann problem of determining a harmon-
ic function according to the values of its normal derivative prescribed on
the contour.
The solution of the torsion problem will assume an elegant

form and open new possibilities if, in addition to the harmonic
function cp(jc, y), we introduce another function, the conjugate
harmonic function \j?(#, y) so that these two functions satisfy
the Cauchy-Riemann conditions:
'
dy dx
Introducing this in the condition on the section contour (8.9),

we obtain
or
By integrating, we get
*
y
<|>(jc, y)= **~^ +C, (8.12)
where C is an arbitrary constant.

(8.12) is the equation of that contour of the cross
Obviously,
section for which displacements (8.1) and (8.2) will occur, i.e.,
u = tyz\ v = txz\ w= icp (x, y).
It may be said that equation (8.12) prescribes the values of the

function \|) (x, y)
on the contour of the section
<|>
= -
+ C, (8.13)
where r is the radius-vector of a point of the contour as before.

Hence, another classical problem is stated for the function (x, y), \|)
namely the Dirichlet problem: to find a harmonic function

according to the preassigned values of this function on the con-
tour of the section.
It now necessary to show that the obtained solution does
is
in correspond to the pure torsion of a bar. It has already
fact
been proved that the lateral cylindrical surface of the bar is free
of stresses. The forces applied at cross sections produce only
shearing stresses (8.7); it remains to show that these forces
reduce to a couple, i.e., they do not give a resultant or, in other
words, their resultant force vector is equal to zero.
If the function q> in (8.7) is replaced by its conjugate function
q according to the Cauchy-Riemann conditions (8.11), we have
*.-,*($-,).
We introduce a new function
<p(x, j;) |AT(^ ^"^ | (8.15)

\ I
and note that
^_ i i __, v| .
do
then we obtain from (8.14) the following simple formulas for

shearing stresses at a cross section
The function <D was proposed by L. Prandtl and, by analogy

with Airy's function in the plane problem, it is called the stress
function. It is important to note that, according to (8.12), this
function has a constant, though arbitrary, value at all points of
the contour:
(8.17)
Let us calculate the projection of the resultant force vector of

the shearing forces at the cross section on the axis Ox:
(F) x, y"
since (*/i) and G)(J/O) the values of the stress function at two
points of the contour are identical according to (8.17). In
exactly the same way it can be shown that the other projection
Y T of the resultant force vector vanishes. This proves that the
above solution of the problem corresponds to the pure torsion
of the bar.
In order to complete the solution of the problem it is neces-
sary to express the twisting moment in terms of the tangen- M t
tial tractions at the section. We obtain
M=t
f(Y z x-Xz y)dF=- fx^-dF- fy^dF. (8.18)
(F) (F) (F)
Let us calculate the first of the integrals in the right-hand

member
x ^ dF =
.~ f
x
JdO J
Applying to the inner integral an integration by parts, we have

*
- A dx =
(8-19)
lf y) and *(* ,
y)
are the values of the function (P on the contour of the section,

defined by formula (8.17); the value of the constant C will not
affect the magnitude of the twisting moment (8.18), since this
magnitude will not be changed if CD is replaced by CD + Ci where
Ci is also an arbitrary constant. If the cross section is simply
connected, i.e., if it is bounded by one closed contour, we may
set C = in (8.17) for the sake of simplicity; then
and (8.19) yields
By similar reasoning with regard to the second integral in the

right-hand member of (8.18) we find the same value for it. Thus,
finally,
Af,
= + 2 / ftodF, (8.20)
(/=)
or
2 2
(8.21)
If the twisting moment is prescribed, one can determine from

this formula the angle of twist per unit lengtji:
* = W> <
8 22 )
'
where
K= f (2<|>
jc
2
y
2
) dF (8.23)
(F)
depends on the dimensions and shape of the section contour and

represents a geometrical characteristic of the section. The quan-
tity [iK is the torsional rigidity of the bar.
238 _ TORSION OF PRISMATICAL BARS AND BENDING
From the solution here obtained one can draw an important

conclusion concerning the limits of applicability of the torsion
theory developed by C. A. Coulomb and based on the assumption
_ [Ch. VIII
that the plane cross sections of a bar remain plane. Indeed, in

this case
<w =
const and CP(A:, y) const. =
Then from (8.11) we find
hence
H= ;
f=*
y)
= const = C.
Substituting this in the equation of the section contour (8.12),
we get
x 2 -f-y 2 const. =
Consequently, the cross section is a circle. Coulomb's theory is
not applicable to any other cross sections and the cross sections
will warp in torsion.
53. SAINT-VENANT'S METHOD. SPECIAL CASES
The relation, established by equations (8.11) between the tor-

sion function op(x, y) and the function \|?(,v, y) which defines,
according to (8.12), the contour of the corresponding cross sec-
tion, permitted Saint-Venant to suggest the following elegant
procedure of solving a number of specific problems.
Let us take an analytic function f(z) of a complex variable
and separate the real and imaginary parts in it:
the functions U and V will be harmonic, and they also satisfy

the Cauchy-Riemann conditions:
dU^^dV^.
dx '
dU _ dV
dx
dy dy
Let us, therefore, set qp

= /; \|?=V. Then, if the equation
represents a closed curve, it will give the contour of the cross

section of the bar while the function U
will be the corresponding
Sec. 53] SAINT-VENANT'S METHOD, SPECIAL CASES 239
torsion function so that the equation
W=tU(x> y)
will determine the displacement in the direction of the z axis,

i.e., will give the warping of the plane cross section. The
it
stresses will be found according to formulas (8.14). Or, vice versa,

one can set
<f=V(x, y); (8.24)
then the equation of the contour (if it is closed) will be
_u (x, y) = * + +C
2
y2 1
(8.25)
and the displacement will be expressed as
w = tV(x, y). (8.26)

Consider some special cases.
A. A Bar of Elliptic Cross Section
Let us take an analytic function
D2 =D2
2
2 (x + /y)
2 =D 2 (x
2 - y + i2D xy.
2
) 2
Here
Assume that
Then, according to (8.25), we obtain the equation of the contour

by replacing C by Dii
or
or, ultimately,
1
It is taken into account here that if the function U+iV is analytic,
V iU is also an analytic function, since the Cauchy-Riemann conditions are
identical for them.
Setting here
-%- = a 2
,
^= b\ (8.28)
i-D,
we obtain
4 +"F
-&=l,
L
(8-29)
i.e., we have a bar of elliptic cross section.

From (8.28) we find the constants DI and D 2, expressing them
in terms of the semiaxes a and b of the ellipse of the cross
section:
Hence
w = 2D 2 y = il=L T*y. (8.30)
Next, according to (8.7), we find the shearing stresses ov$r the

section by replacing the symbol fx by G: \
(8.31)
Finally, we correlate the stresses over the section with the twist-
ing moment M t :
(8.32)
But
f f x dF= J^
. 2
f fy2 dF=Jjc
where J x and J y are the* moments of inertia of the ellipse about
its axes:
*
Jx _ nab 4
3
.
.
Jy
Therefore, we obtain from (8.32)
Hence, the angle of twist t can be expressed in terms of the

twisting moment:
Sec. 53] SAINT-VENANT'S METHOD. SPECIAL CASES 241
In the case of a circular section the elementary Coulomb's

theory yields
_ _ M t
GJp*
where J p is the polar moment of inertia of the section. In order
to compare these expressions let us correlate the quantity
with the polar moment of inertia of the ellipse:

.
Jp
_ /'x
-- I
\
/
Jy
_ "
3 3
F* a2 +b 2
a 6
i.e.,
T74
WJp 39AJp
and, consequently,
^ = -^r39.4y p
.
(8.33)
Replacing ay by z in equality (8.30), we obtain the equation

of the warped surface of the cross
section z = mxy, where
this surface is a hyperbolic para-

boloid; a set of horizontal planes
z = zn , where n= 1, 2, 3, . . .
,
cuts
it in hyperbolas
xy
* = m ,
referred to the asymptotes (Fig. 81).

By the use of equations (8.31)
we can find the maximum shear-
ing stresses. Indeed, the magni-
tude of the total shearing stress
Fig. 81
is given by
T~^2 2GT
^r + ir. (8.34)
Hence one can see that the stress T, as a function of x and */,
remains the same

- O I t
(8.35)
16-1013
at all points of the curve
^ b*
~ '
(8.36)
where C is a constant number. For different values of C equa-

tion (8.36) gives a family of curves [level lines of the function
n* i/)h eac h of these curves represents an ellipse with the semi-
axes
ai =a?C\ b^&C. (8.37)
If we assume that b<a, then
'
a, \a t a
and the ellipse (8.36) will be more elongated than the ellipse
(8.29) of the section contour. Equation (8.36) gives a family of
h /|2
similar and similarly located ellipses (the ratio -rr is con-
stant), the stress (8.35), which is constant for all points of the
Fig. 82
ellipse, becoming greater as the dimensions of the ellipse in-

crease (Fig. 82). It follows from this that the maximum value
of the stress (corresponding to ellipse 5) occurs at the extreme
points C and D
of the minor axis of the section. For these points
we have from (8.37) 61 =6=6 C 2

or C = -and, therefore, accord-
ing to equation (8.35)
At the points A and B the stress (being the same at all points
of the corresponding ellipse /) will be smaller; we have for them
a,
= a == a?C\ C = -L ;
2
T r*/" fl^ ^. T
B. A Bar of Triangular Cross Section

Let us take an analytic function
= +
C3 z3 C3 (x *y) 3 C3 (x* = 3xy
2
) + /C 3 (3x y
2
y
3
)
and assume
cp
= V= C 3 (3* y
2
y
3
). (8.38)
Then, according to (8.25), we obtain the equation of the contour
C3 (x* - 3xy 2 ) +1 +y = - C (A:

2 2
)
or, setting
C3 = |/3 3 ; C = y>i, (8.39)
we have
D 3 (x
3
3*y
2
) + x* + y = D v 2
(8.40)
This equation represents a complete hyperbola of the third
order with three asymptotes whose directions will be obtained
in accordance with the general rule
by equating to zero a set
of the terms of the third degree x 3 3xy 2 x (x -f- y 3) (x = Y
y 1^3) =
0; hence the equations of three straight lines parallel
to the asymptotes are
;c 0; = r
* 4- y 1/3 0; x y |/ j=0. =
Let us take three straight lines parallel to these:
b. (8.41)
By a suitable choice of the constants D t and D 3 curve (8.40)

can be shown to fall into these three straight lines, i.e., its equa-
tion will assume the form
D 3 (x a)(x + yY3)(x-b~ y 1/3) =
b 0,
or
D (x - a) [(x - - 3y =
3 ft)
2 2
]
0. (8.42)
In fact, writing out equations (8.40) and (8.42), we have
D^ - 3D,xy* + x + y - D, = 2 2
0, (8.43)
D^x* 3D^xy D (a + 26) x 3D ay +
2
3
2
-j- 3
2
+ D b (2a + b)x D ab = z 3
2
0. (8.44)
By comparing the coefficients of these equations, we obtain

the conditions for determining DI, D3 and b:
+ 2fr)=l; D 6 (2a + 6) = 0;
3
3D 3a=l; 0^ = 0,
16*
Hence we find
4_
3 (8.45)
and equation (8.44) becomes
(x a) [(x + 2af 3y
2
]
=
or
(x 0)C* + 2aH- yl/3)(jcH-2a y |/1j)
= 0,
i.e., curve (8.44) falls into three straight lines
which form (Fig. 83) an equilateral triangle with centroid at the

origin of co-ordinates. Thus we come to the problem of torsion
of a bar of triangular cross sec-
tion.
Equation (8.38), on the ba-
sis of (8.39), and (8.45), gives
w = -^- (3x 2
y y
3
), (8.46)
Fig. 83 Fig. 84
i.e., the cross section warps into the surface
or
It follows that the straight lines
which are the three altitudes of the triangle of the section, re-
main in the original plane of the section, since w=z=Q for all
points of these straight lines. Fig. 84 shows the form of the

warped cross section. With function (8.46) we find the stress
Sec. 53] SAINT-VENANT'S METHOD. SPECIAL CASES 24S
components X z and Y z according to equations (8.2) and (8.7).

Further, similarly to equation (8.32), we obtain the relation be-
tween the twisting moment t M
and the angle of twist t.
C. A Bar of Rectangular Cross Section

This problem was also solved by Saint-Venant; let us denote
the lengths of the sides of the section by b and c (Fig. 85) and
begin the solution of the problem with the determination of the
torsion function (p(*, y) in the form of a series. Particular solu-
tions of equation (8.4) will be found by the method of separation
of variables, which has already been applied previously, assuming
that
y = XY. (8.47)
The variables are immediately sepa-
rated:
X" Y" ,o I ^
and hence
X" = ,
\
I
(8 48)
'
As we shall see later, it is conven-

ient to set .1
^ ^T~' where /ra = 0, 1, 2, . . . .
2
When w = 0, equations (8.48) yield 85
Fig.
Substituting this in (8.47) and retaining only the term of the-

second degree, we get (setting also a = y=\)
cp
= jcy. (8.49>
When m = the solutions of equations (8.48) are
'
b
(8.50>
.
By/ summing up solutions (8.49) and (8.50), we obtain (setting

C m = for all values of m)
= *y + sin cosh .
(8.51)
"246 TORSION OF PRISMATICAL BARS AND BENDING [Ch. VIII
We adjust the arbitrary constants Cm m and D' m so as to ,

D
satisfy the conditions on the contour; they require, as we have
seen above, the shearing stresses to be directed along the tan-
gent to the contour; consequently, on the lateral sides x=-^

(Fig. 85) we must have according to (8.7)
. X, = or - = y; (8.52)
on the upper and lower sides y= -j
Y2 = or
1*-
= -*. (8.53)
Turning to (8.51), we find
It follows from this that condition (8.52) will be satisfied if we

assign m
only odd values: w=l, 3, 5, ...; in fact, then for
jc = we have
mnx = COS ( mrc \ ~
=0.
,
COS I
-y-j
To consider condition (8.53) we calculate from (8.51)
Introducing here the value y=-^, we write condition (8.53)'

in the form
w-D m smh-jr
m u mnc mnx
cm -b r^
-nine nk' \ -
s-> ( i _i_
-(Dm cosh
sin
) -^.
m=l, 3, 5, ...
In view of the fact that the second term in parentheses may

have two values, necessary it =0; in addition, we
is to set Dm
can, of course, take D m =l to obtain our condition in the final
form
-2*= S P.sinnr, (8-54)

m=l, 3, 5, ...
where
(8 .55)
It is evident from (8.54) that p m are coefficients of the Four-
ier series for the function 2x in the interval f

-j
. + "2 ) they
are determined in the usual way; multiplying both members of

mnx
(8.54) by sin
j- and integrating, we have
b
+T m-H
mT =-2
n bn T
p *s
J
m +""
l
or m = -o-o-( ir"2
On this basis we find from (8.55)

m+l
u
,
3
w ,
3
cosh
Introducing this in (8.51) and taking into account that D m =l

D m =0, we obtain the final expression of the torsion function:
m=l,3,5,...
(8.56>
Further, by using the Cauchy-Riemann conditions (8.11), we seek
its conjugate function ty(x, y)\ we have
,e
'"
sin cosh =_
m 2o cosh i
-J
By integrating the first of these equations we get

m+ l
(1) 2
=r cos -^- cosh
m 3
cosh
the second yields

m-f 1
m 3 cosh
1
Here the substitution is made
m-i
^~- = (-l) 2
for m-1, 3.5, ...
"248 TORSION OF PRISMATICAL BARS AND BENDING [Ch. VIII
where o>i and 02 are arbitrary functions. To have both of these

expressions of the function i|) identical, we must set
<2 (y)
=i
C is an arbitrary constant.
We obtain finally
-* ,
W vi, ,,
/w +l
L
cos cosh
_
m ,
3
COSh
z^
(8.57)
We calculate now the stress function according to formula
C = -j-to make
b2
(8.15) (we take here this function vanish on
the contour) :
.
1X 9 1 mnx
- mny /0 rov
(-D 2
,
cos cosh-- i
.
(8.58)
/
m=l,3, 5, ...
To determine the angle of twist t according to a preassigned

twisting moment we use relation (8.20); carrying out integration
in it we obtain l
^^^1
-2T
m = l,3, 5, ... J
= |itWc T ___ ^
Aq V f 1 64 6 *
* u ^^^ 1
.
tanh-jy-
m=l L f 3, 5, ... J
Hence
*= < 8 59 ) -
lifer-
where
1_^1A
m=l, 3, 5, ...
In further calculations we take into account that ( l)

m= 1 for
l. 3. 5, ...
We see that a depends only on the ratio of the sides of the
rectangle ; for a square, i.e., for--==

c
1, we get
c
=
For a very narrow rectangle we assume approximately that
= 0, and then
a = ~o 0.333 ... .
To complete the solution of the problem we calculate the

shearing stresses by using (8.16) and (8.58):
^-g-SHD" m ^'os^slnh^, 2
cosh -jr-
J
( 8 61 >
'
26u
m+ i
__^^
OU rwix
tr
Kz at
r
\
1
r\
2;c + Ur ^J
i
i>, (1)
2
X. 1
/
V
-t \
/
O
2
1
mTCC
sin
. til
-r cosh
i
"'. (8.62)
If b<c, the maximum value of the shearing stress occurs at
the point A (Fig. 85), i.e., forA: =

y, y
= 0; then we have
from (8.62)
9 . mite
[
M2cosh
_J
or by using (8.59)
Tm^ = (8.63)
-^>
where
u
m o
2
cosh -j
The stress at the point B (Fig. 85) will be obtained from (8.61),
setting A; = O, y = 4-:
'
2b
or
(8.64)
where
m-f 1
The values of the coefficients a, p and PI introduced here are

given in the table below:
54. SOLUTION OF THE TORSION PROBLEM

IN TERMS OF STRESSES. PRANDTL'S ANALOGY
The introduction of Prandtl's stress function makes it possible
to change the whole method of solving the torsion problem and
to find first the shearing stresses on the cross section (cf. Sections
17 and 36).
The torsion function y(x, y) and its conjugate \|?(A:, y) are
harmonic, i.e., they satisfy Laplace's equation
0. (8.65)
Let us see what differential equation the stress function <P(;t, y)

satisfies. We have found above (8.14) and (8.16) that
(8.66)
We eliminate tf by using the second of equations (8.65); differ-

entiating the first of equations (8.66) with respect to x and the
second with respect to y and adding up the results, we obtain
V 2 <D = 2pt. (8.67)
Thus the stress function satisfies Poisson's differential equation
(8.67). It may be seen that this equation represents the compat-
Sec. 54] SOLUTION OF THE TORSION PROBLEM IN TERMS OF STRESSES 251
ibility equation which should be added to the equations of

equilibrium, since the introduction of Prandtl's stress function
allows the problem to be solved in terms of stresses. Indeed,
Saint-Venant's compatibility equations are derived from Cauchy's
equations (III) through the elimination of the displacements
u, v, w from them. However, at the very beginning of this chapter
certain restrictions were imposed on the displacements by for-
mulas (8.1):
u = iyz\ v =
and, therefore, Cauchy's equations are considerably simplified:
Cyy
-C - zz U, Cx y -
dw dw
Differentiating the last two of them and subtracting the results,

we eliminate the only remaining displacement w:
e yz
dx
dexz
_
dy
Multiplying this equality by Lame's constant \i and using

Hooke's law
we obtain equation (8.67). In the torsion problem it is equiva-

lent to six equations of Beltrami (VII) of Sec. 36; just as in the
case of the axisymmetrical plane problem [Sec. 48, equation
(7.13)] such a simplification is a consequence of the restrictions
imposed on the displacements.
It is necessary to add to equation
(8.67) the previously ob-
tained boundary condition on the section contour (8.17)
$ = |itC. (8.68)
In the case of a simply connected cross section, i.e., a section

bounded by one closed contour and, consequently, without inner
cavities, the arbitrary constant C may be taken equal to zero.
The determination of the stress function will then be reduced to
the classical problem of integration of Poisson's equation (8.67)
with the condition that the function <D vanishes on the contour of
the section:
$ = (on the contour). (8.69)
If this problem is solved, the stresses on the cross section will
immediately be found according to formulas (8.66) and the angle

of twist t, according to equation (8.20).
252 TORSION OF PRISMATICAL BARS AND BENDING [Ch VIII
The cases considered in the preceding section can easily be

solved by this method. For instance, for an elliptic section
the solution of equation (8.67) can be taken in the form
(8.70)
which the boundary condition (8.69). Substituting this

satisfies
solution in (8.67), we find immediately the constant /(.
This method of solution of the problem led L. Prandtl, who
had proposed it, to the following analogy, which lends much
clarity to the corresponding computations and at the same time
makes it possible to elaborate a purely experimental method of
solving the problem for any contour of the cross section of
a twisted bar. Imagine a flexible inextensible membrane clamped
on an elastic contour identical with the periphery of the given
cross section; the tension is constant in all directions. If we
apply a uniform pressure p to the membrane, it may buckle to
some extent on account of small deformations of the elastic
contour itself the equation of equilibrium of the membrane was
!
derived by P. S. Laplace; it coincides with the equation given

in all courses on strength of materials for the analysis of thin
vessels having the form of bodies of revolution:
lL 4. fl_A
h
'
Pi Pa
where 0i and 02 are the principal stresses in the wall of the

vessel, pi and p2 are the principal radii of curvature of the wall,
and h is the wall thickness.
Taking into account that the tension of the membrane is con-
stant, i.e., o\=a 2 and that h<y\ ho<i is =
the magnitude of =H
this tension per unit length of the section of the membrane, we
obtain
JL-4--L
"^ JL.
P, P2 H '
if the buckling of the membrane is small, we may assume

= dx*
JL "~ .!!- '
1
^
~ <>**
dy*
'
p, p2
where z is the ordinate of the surface of the membrane, and we

obtain the following differential equation of the surface of the
1
One can see, however, that these deformations will be small as compared
to the sag of the membrane ^ and they can be neglected.
Sec. 54] SOLUTION OF THE TORSION PROBLEM IN TERMS OF STRESSES 253
membrane:
Jx*
= -&. r
_ 77'
P
(8.71)
dy
We add to it the obvious boundary condition

2 = (on the contour). (8.72)
Comparing (8.71) and (8.72) with (8.67) and (8.69) respectively,

we come to Prandtl's analogy: if the tension of the membrane H
and the load p are chosen so that
(8.73)
then
z = <$>\
(8.74)
hence, the ordinate of the surface of the buckled membrane gives

the value of the stress function at a given point of the cross sec-
tion, and the twisting moment is equal, according to (8.20), to
Fig. 86
double the volume bounded by the surface of the membrane anH

itsoriginal plane.
The distribution of the shearing stresses over the section can
also be represented visually with the aid of Prandtl's analogy.
Let any closed contour BC be passed through a chosen arbitrary
point A of the section (Fig. 86) and a tangent AT and an in-
ward normal Av to the contour be drawn at A. The projection of
the total stress t at this point on the tangent AT will be
t
(T}
=X z cos (xT) +- Y z cos (yT)\
254 TORSION OF PR1SMATICAL BARS AND BENDING [Ch. VIII
but
cos (xT) = 4- cos (yv),
cos (yT) = cos (xv).
Using also formulas (8.16), we get
t
( T)
= d* CQS ( ^ ) _j_ CQS (yiy)i
or
(8.75)
this is a generalisation of the property of the stress function

as expressed by formulas (8.16); the projection of the total stress
111
Fig. 87 Fig. 88
on any direction T is equal to the derivative of the function (D

along the normal v to this direction.
Now let us consider the surface of the buckled membrane and
imagine it has been cut by a number of equidistant planes par-
allel to the contour (Fig. 87). We obtain a series of closed con-
tours horizontals of the surface. At any point A of the horizon-
tal the derivative dQ)/dT along the tangent to it is equal to zero
since <D = const in this direction. Hence we conclude, according to
Sec 54] SOLUTION OF THE TORSION PROBLEM IN TERMS OF STRESSES 255
(8.75), that the projection of the total shearing stress t on the

normal Av to the horizontal is zero; consequently, the stress t
is directed along the tangent to the horizontal at this point. The
same formula (8.75) shows that the magnitude of the total shear-
ing stress is equal to the derivative of the function d> along the
inward normal to the horizontal or, which is the same, to the
gradient of the function (D at a given point:
Taking into account that the gradient (the tangent of the max-
imum angle of inclination of the surface of the membrane at
a given point to its initial plane) is proportional to the close-
ness of horizontals, we see that the system of horizontals gives
a visual picture of stress distribution over the section of a twist-
ed rod. Hence, one can draw, for instance, the following rather
general conclusions.
(1) If the section is bounded by a simple contour which is
everywhere convex, the closeness of horizontals increases as the
contour is approached, and it is there that one should expect the
greatest stresses.
(2) If the contour indicated in (1) above has two axes of sym-
metry and is somewhat elongated one of these
in the direction of
axes, the closeness of horizontals will be maximal at the ends of
the minor axis as, for instance, in the case of an ellipse or a
rectangle. (At the ends of the major axis the possibility of the
membrane buckling is restrained by its smaller width, measured
perpendicular to the major axis.)
(3)the cross section has the form of a very narrow rectangle
If
where side b is considerably smaller than side c (Fig. 88), then
at points remote from the short sides the surface of the corre-
sponding membrane can be assumed to be approximately cylindri-
cal; in view of the uniformity of the load p on the membrane,
the section of this surface by a plane perpendicular to the y axis
will be a parabola
}
But, according to PrandtFs analogy [see formulas (8.73) and

(8.74)], we have
Which is a funicular curve for a uniform load.

256 TORSION OF PRISMATICAL BARS AND BENDING [Ch VIII
Further, according to (8.20), we get
C I b*
j \-^
which has already been obtained previously. Consider in the

cross section a straight line */ = const, parallel to the x axis, which
is drawn far from the ends of the section. The shearing stresses
v
Y ~ ~~ M
d ^ t
x
are distributed according to a linear law along this straight line

(Fig. 88); the shearing stresses X z will apparently be negligible
far from the short sides of the rectangle.
Obviously, these results obtained for a narrow rectangle may
be applied to other sections having the form of a bent strip
(channels, I-beams, angles, unclosed rings, etc.) excepting, of
course, those points where the curvature of the contour of the
section changes abruptly; the configuration of the horizontals of
the membrane becomes more complicated here (for instance, near
the fillets and re-entrant corners of the section). These applica-
tions of Saint-Venant's theory and Prandtl's analogy are treated
in detail in the courses on strength of materials and we shall not
pursue the matter here.
In the case of a closed ring section one hits upon a feature
characteristic of any doubly connected or, in general, multiply
connected section bounded by several closed contours which do
not intersect each other. The differential equation (8.67) and the
boundary condition (8.68) for the stress function <D hold good, but
the constants C in (8.68) on each of the contours bounding the
cection will be different; for one of the contours, e.g., for the
outer contour, this constant can be chosen arbitrarily; the method
of determining the remaining constants was indicated by R. Bredt
l
.
Bredt's corresponding theorem, which follows from the general

Stokes's formula, can readily be obtained for our purposes from
Prandtl's analogy. In the case of a multiply connected section this
See A. Foppl and Drang und Zwatig, Bd. Berlin,

1
L. Foppl, 2. Aufl., 2,
1928.
Sec 54] SOLUTION OF THE TORSiv^xI PROBLEM IN TERMS OF STRESSES 257
analogy has to be constructed as follows (Fig. 89). The regions

FI and F 2 of the membrane, corresponding to the cavities in the
section of a twisted bar, are covered by absolutely rigid plates
glued to the membrane; after that a uniform pressure p is applied
on the whole of the section F.
T-ryrrrl
WM/2\
Fig. 89
Condition (8.68) requires that plates FI and F 2 move down-

ward parallel to their initial position (for that, generally speak-
ing, it is necessary to apply, in addition to the pressure, certain
Fig. 90
couples to the plates; these, however, will not be involved in sub-

sequent considerations). Let us write down the condition of equi-
librium of one such plate F m equating to zero the sum of the
projections of the forces applied to it on the axis normal to the
initial plane of the membrane. This condition will be of the form
given by the following equation (Fig. 90):
(8.78)
17-1013
258 TORSION OF PRISttATICAL BARS AND BENDING [Ch. VIH
where the integral is extended over the entire contour of the

plate concerned, and F m designates its area. We observe that
[cf. (8.76)]
sincp tancp = = grad O = /.
Furthermore, we put //, as a constant, before the integral sign

and bear in mind that according to (8.73) /? = 2jnt//. The equi-
librium equation (8.78) becomes
or (8.79)
this is the required particular case of Bredt's theorem of "stress

circulation" expressed by the integral in the left-hand member
of the second of equations (8.79). We
can write as many equa-
tions of this kind as there are inner cavities in a given section.
It may be shown that the left-hand members of equations (8.79)
depend linearly on the values of function d> on the inner contours

of the section; these values can be determined from (8.79). It is
to be noted that in the case of a multiply connected contour for-
mula (8.18), upon integration of the right-hand member by parts,
yields the following result in place of (8.20):
M = 2C F - 2
t Q Cm Fm - 2 f f 3> ds> (8.80)
where F is the area bounded by the outer contour; C is the value

of function F on this contour; F m and C m denote the same quan-
tities for the inner contours. The double integral is extended over
the area of the section minus the inner cavities. There remains in
(8.80) to substitute the values of the constants C and C m ob-
tained as indicated above.
55. CASE OF TRANSVERSE BENDING

Consider (Fig. 91) the case of the bending of a straight rod
with an arbitrary cross section, of length /, built-in at the left-
hand end and loaded at the right-hand end by a force Q which
we assume to be applied in the form of shearing stresses distrib-
uted over the end cross section.
This problem was solved by Saint-Venant by the use of the
semi-inverse method; we shall carry out its solution in terms of
stresses by applying the results of Sec. 36. In solving a similar
Sec. 55] CASE OF TRANSVERSE BENDING 259
problem in the case of plane strain or plane stress (Sec. 41), we

have made the following assumptions:
7 = M X = Q(lZ) Xi y
I z yy
I yx
y\ y n
U.
/L/2 j- j y\y
Following Saint-Venant, we shall also use the semi-inverse

method solve the present problem, though modifying slightly
to
the assumptions by considering that the stress Yz is different from
zero as well. Consequently, we make the following assumptions:
(8.81)
XX = Yy = Xy ~ 0.
The remaining stresses X Z X and Y z = Z y will be found
=Z from
the surface conditions so as to satisfy equations (I) and (VII) as
'
Fig- 91
well. Let us see what form these equations assume under our
assumptions (8.81). The equilibrium equations (I) will be
written as
^= 0;
dYz
=0; (8.82)
dZ K
Qx_
J (8.83)
Equations (8.82) show that the shearing stresses X z and Yz are

independent of z, i.e., they mu'st be distributed identically on all
cross sections and, of course, they must be distributed on the end
cross section B according to the same law. This restricts the
manner in which the force Q may be applied at the section B.
The force Q, when applied in a different manner, will produce
other stresses; according to Saint-Venant's principle, however, the
17*
260 _
difference
loaded end B.
We now
in
TORSION OF PRISMATICAL BARS AND BENDING
stress
proceed
distribution
to equations
will
(VII),
be essential only near
observing first
[Ch
that
VIII
the
in
-- -
our case
O 7z _ <?(*-*) X.
y.
j
It can easily be seen that the first, second, third and sixth of
equations (VII) are satisfied identically by virtue of assump-

tions (8.81). The fourth and fifth equations will be written as
or
V*J% = 0; ?'** = (T^yj- (8.84)
Let us also write down conditions (II) on the lateral surface

of the rod. The first two are satisfied identically since cos (vz)=0
on the lateral surface. There remains only the last equation
Z x cos (ox) + Z y cos (vy) = 0. (8.85)
Thus, the problem is reduced to the determination of two

stresses X z and Y 2 which are functions of x and y, from the condi-
,
tion that they satisfy equations (8.83) and (8.84) everywhere in

the rod and (8.85) on its lateral surface.
Subsequently it will also be necessary to require that these
shearing stresses on the end section reduce to the given load Q
applied at the centroid of the section. As previously, we can sim-
plify the solution of the problem by introducing the stress func-
tion; however, it is found convenient here to introduce two new
functions; let the stresses be prescribed as
(8.86)
Here cp(#, y) and f(y) are as yet arbitrary functions

l
. Substitut-
ing these expressions in equation (8.83), we see that it will be
satisfied identically. It remains to satisfy equations (8.84) and
the condition on the lateral surface (8.85).
Substituting stresses (8.86) in equations (8.84), we write them
down as
__
1
The introduction of the function f(y$ was proposed by S. P. Timoshenko.
Sec. 55] CASE OF TRANSVERSE BENDING 261
or
JL vV 2
co
V 0-
V>
-
v 2 cp
V ?
q(? _f J
dy /(I
From these equations we easily find
Weproceed to the condition on the lateral surface (8.85). As

in the torsion problem, we obtain
cos (vx) = -g- ; cos (vy) = ~.

Substituting this and (8.86) in condition (8.85), we have
- 2J >
ds dx ds
-
dy
or, bearing in mind that
d<pdx .
dy dy _ dy
'
dx ds '
dy ds ds
we get
dy
The question is reduced to the integration of equation (8.87)

subject to condition (8.88) on the lateral surface. Let us try to
adjust the arbitrary function f(y), introduced here, to make the
bracketed expression in equation (8.88) varnish on the contour of
the cross section, i.e., the following condition must be fulfilled
on the contour:
=0; (8.89)
equation (8.88) will then be reduced to the condition

~^f~
requiring that the stress function <p remain the same on the con-
tour of the section. It remains to consider the condition on the
end cross section. By projecting the forces applied to this section
on the axis Oy we have
262 TORSION OF PRISMATICAL BARS AND BENDING ICh. VIH
i.e.,the sum of the tangential tractions Q acts in a vertical

direction (Fig. 91). Furthermore,
__ 2
4.
r
~
2
Thus, the shearing stresses on the end section do in fact reduce

to the given load Q directed downward.
We have still one arbitrary constant C left in equation (8.87).
It is easy to check that for a section which is symmetrical with
respect to the plane xOz one should take C=0.

To solve the problem it is finally necessary to choose f(y) so
as to satisfy equation (8.89) on the contour. We introduce this
expression of f(y) in equation (8.87) and integrate it. Having
thus obtained (p(*, y) we introduce it in equation (8.86) and find
the shearing stresses.
If the cross section of a rod is not symmetrical with respect
to the principal centroidal axis Ox, the phenomenon of bending
becomes far more involved. The solution of the problem obtained
here holds true if the resultant of the load Q lies not in the prin-
cipal plane Oxz of the rod but in another plane parallel to it
and intersecting the cross section at a point called flexural centre.
In the case of thin-walled rods the flexural centre can be found
approximately by the elementary method given in the courses on
strength of materials.
Example. Bending of a circular rod. Let the contour of the sec-
tion be a circle:
jc
2
+y 2
r2 = 0.
(8.90)
To satisfy equation (8.89) on the contour one can obviously pre-

scribe the function f(y) as
After that equation (8.87) becomes
+C. (8.92)
'
Sec 551 CASE OF TRANSVERSE BENDING 263
Let us attempt to satisfy this equation by prescribing the function

<pin the following form:
y = m(r 2
x* y
2
) y.
For substitution in equation (8.92) we have
and equation (8.92) yields
Hence, obviously, C=0l

m= 1
(l+2a) Q
The stress function is finally found to be
_ 1 l+2a Q
Substituting this expression of the function cp and also f(y)

from equation (8.91) in equations (8.86), we find the shearing
stresses
x __ Q 3 + 2a /
2
*2
l-2a
"** 27 4(l + a) V
We observe that the shearing stresses depend on Poisson's ratio
o; assuming, for instance, that a = -^ we get
'** *
32 j y 11 '
(8<94)
Y = --*-.* }
Since, in general, Yz + 0, the total shearing stress
is acting in a direction other than vertical. At the points of the

section which are located on the co-ordinate axes (Jt=0, f/ = 0)
the component Yz is absent and t is vertical here. On the neutral
axis (#=0) we have
/ o l
r
) (
264
It
TORSION OF PRISMATICAL BARS AND BENDING
follows from this that the stresses X z

_
are distributed nonuni-
maximum value (at the
[Ch. VIII
formly along the neutral axis. Their

centre of the section for y = 0) is
~~
z 27 4(1 + a)
for o = -g-
we find from equations (8.94)
v
max X,-^ T r2 =
o
w -^r = T 7
- - Q7R
-== 11.375
T ,
/QQ^\
(8.95)
where F = nr 2 is the cross-sectional area of the rod. The elemen-

os
tary calculation according to the formula X = t= -~-
z gives
max*,, =
v 4Q
355-
The error, as compared to the exact value (8.95), is about

3.4 per cent. The stresses Yz are usually neglected in the ele-
mentary analysis,
IX
More General Methods

of Solving Elasticity Problems
56. GENERAL SOLUTION OF DIFFERENTIAL EQUATIONS

OF EQUILIBRIUM IN TERMS OF STRESSES. STRESS FUNCTIONS
The solution of problems of the theory of elasticity in terms

of stresses requires the simultaneous solution of two systems of
differential equations: the equations of equilibrium (I) and the
Beltrami-Michell compatibility equations (VII). We shall limit
ourselves to the case where body forces are absent; then all these
equations will be homogeneous. In this section it will be shown
that the system of equilibrium equations
=
0,
(9.1)
can be solved in the general form, all the components of the

stress tensor being expressed in terms of three arbitrary functions
of x, y, 2, called stress functions-, we have already encountered
similar solutions (though for narrower classes of problems) in
the plane problem (Airy's function) and in Saint-Venant's theory
of torsion (Prandtl's function).
If the general solution of the equations of equilibrium (9.1)
has been obtained, it will no longer be necessary to deal with
them in future and it will remain to integrate only Beltrami's
-equations; this will facilitate to some extent the solution of the
problem.
One of the forms of the general solution was proposed by
J. C. Maxwell in 1862; another form was given by G. Morera in
1892. These solutions can be obtained without great difficulty.
266 MORE GENERAL METHODS OF SOLVING ELASTICITY PROBLEMS [Ch. IX
1. Maxwell's solution. Let us take three arbitrary functions
<PI(* y> *); %(*. y> *); %(-* y. *) (

9 2>
-
and express the shearing stresses through them as follows:
y z
~~ **
_-dz dx '
dxdy* dydz'
These formulas are obtained from one another by cyclic changes
between
(x, y, z); (cp,, cp 2 , cp 3 ).
We shall see later that there are other possible ways of prescrib-
ing the shearing stresses, but in any case their formulas should
not be altered in form in interchanging the letters in the nota-
tion of stresses, as required by the law of reciprocity of shearing
stresses, for instance
Y ~~
dxdy
Substituting the expressions of Xy and X z from (9.3) in the first
of equations (9.1), we obtain
dXx __~ d*<? 3 d2?2
">" '
dx dx dy* dz 2 dx
Hence we find the component Xx by integrating with respect

to x and omitting an arbitrary constant of integration (i.e., an
arbitrary function of y and z) which is of no importance since
the functions cp 2 and cpa are arbitrary:
In the same way we obtain the remaining normal components

Yy and Zz from the second and third equations of (9.1); their
7
expressions, however, can be obtained from (9.4 ) by the above-
indicated cyclic changes. As a result, we have the following
solution of equations (9.1):
< 9 4>
-
M~
Z =-2H"
Functions (9.2) in this solution are called Maxwell's stress func-
tions.
Sec 56] SOLUT. OF DIFFEREN. EQUAT. OF EQUILIB. IN TERMS OF STRESSES 267
That solution (9.4) in general will be confirmed, if we prove

that for any tensor
(Xt Yx Zx \
(9.5)
which satisfies the equations of equilibrium (9.1), it is possible

to construct three functions (9.2). This can easily be done by
integrating twice the equalities in the right-hand column of (9.4):
<h
= ffXydxdy; <ti
= f fy z dydz;
cp 2
= J J Xg dx dz.
The constants of integration are omitted, as above. It remains to

ascertain that when these values of functions are substituted in
the equalities of the left-hand column of (9.4) the latter become
identities; for substitution in the first equality we find
dX y d^ 2 dXz
then it will be written down as
After differentiating it with respect to x, we find
dx dy dz
On the basis of the first of equations (9.1) this equality gives an

identity; in the same way we can prove that the other two equal-
ities (9.4) become identities.
2. Morera's solution. Let us take three arbitrary functions
y. *); <M* y. *) ( 9 6)
-
and express the normal stresses through them:

These formulas are also obtained one from another by cyclic

changes between
(x, y, z); (<h, <|) 2f <j> 3 )-
Introducing the values of stresses (9.7) in the equations of equi-

librium (9.1), we write them down as
'
'
dz dx dy dz
dYz d3 <\> 2
== '
(9.8)
~T~~dz~ dxdydz
dYz
+ dy dx dy dz
After differentiating them according to the scheme indicated orb

the left and adding them up by members, we obtain
dx dy dx dy dz dx
t
"+"
d^ 2
~ d^\
dz
'
\ dy )
whence we have
~ i
2
a
dz
/ah
( dx "^ dy
.
*h ah\
dz )'
The other two shearing components will be obtained by the

above-mentioned cyclic changes or by a change of signs in differ-
entiating equations (9.8). We thus arrive at the following form
of the solution of equations (9.1):
d 1 / dfy[
' A^ry == ~'2"~dz'\Jx'^ ^4^2 ^4^3
dF
dydz dy'
dzdx '
z 2 dx ( dy
~l~
dz **
dx (9.9)
dxdy
'
f
'
JC
= ~'2'Jy\~dz"^
d ^^3 1
r ~dx'
I C^l
Functions (9.6) in this solution are called Morera's stress func-

tions.The generality of Morera's solution can be proved by the
method described above for Maxwell's solution; functions (9.6)
are found by the double integration of the equalities in the left-
hand column of (9.9); their substitution in the equalities of the
right-hand column shows that they become identities on the basis
of equilibrium equations (9.1).
Sec. 56] SOLUT. OF DIFFEREN. EQUAT OF EQUILIB IN TERMS OF STRESSES 269
Maxwell's and Morera's solutions are not the only possible

ones; V. I. Blokh and Ju. A. Krutkov indicated a number of
other forms of the general solution of the equilibrium equations '.
These solutions are expressed in terms of the second derivatives
of stress functions; it is possible to seek solutions expressed in
terms of derivatives of higher orders. Let us take, for instance,
one arbitrary function F(x, y, z) and three constant parameters
a, b, c, also arbitrary, and write down the components of the stress
tensor in the following form:
' '
2
dy dz
2
dx dy dz 9
d'F
dz 2 dx 2
' ^-a
Yz ~~
"dydzdx 2
'
(9.10)
Zz = (a + b) dx 2 dy f',
Zx = b
dzdxdy 2
Substituting these values of the components in the equations of
equilibrium (9.1), we see that they are identically satisfied. Let
us set up a sum of solutions of the form (9.10):
:=2X
m-l
dy
2 Xy
m =1
m dxdydz 2 '
y
1 v
-y
7i '
2 ^ '
(9.11)
dz 2 dx 2
r ^H^ dydzdx 2
m =1
^=^(am dy
2 '
__^ m =1
m dzdxdy 2 '
it will also be a solution of equations (9.1).

If we now put
oo oo
V * m / .. .. _.\.
m =1 =1
(9.12)
and substitute these values in formulas (9.11), we shall obtain

Maxwell's solution (9.4). Since the parameters a m b m , ,
cm are
1
Ju. A. Krutkov, The Stress-function Tensor and General Solutions in Stat-
ics of the Theory of Elasticity. Moscow, Izd. AN SSSR, 1949; V. I. Blokh>
Prikl. matem. mekhan., i960, Vol 14, No 4, pp. 415-422
i
270 MORE GENERAL METHODS OF SOLVING ELASTICITY PROBLEMS [Ch IX
arbitrary, functions (9.12) are independent of each other, and since

the functions F m (x, y, z) are arbitrary, they are also arbitrary.
Let us now replace the parameters a m 6 m c m by others accord- , ,
ing to the formulas .
+ Cm = am am = J (Pm + Tm + J;
(9.13)
and set
m= 1 m=
S 1
(9.H)
m=l
Substituting these values in formulas (9.11), we obtain Morera's

solution (9.9). Formulas (9.12) and (9.14) present Maxwell's and
Morera's functions in the form of series; consequently, the param-
eters a m 6 m c m and a m p m y
, , entering in them should be
, , .
chosen so that the series converge and the formulas make

sense.
It will be observed that if we set (pi
= ^0 in
q) 2 Maxwell's for-
mulas (9.4) and if the function (p3
= cp(A;, y) (i.e., it does
not depend on z), it reprecents Airy's function in the plane prob-
lem and formulas (9.4) coincide with formulas (6.16) of Chap-
ter VI.
57. EQUATIONS OF EQUILIBRIUM IN CYLINDRICAL

CO-ORDINATES. THEIR GENERAL SOLUTION
The cylindrical co-ordinates have been employed in Sec. 47 for

the solution of the plane problem; by using them, we have obtained
the differential equations of equilibrium (Ipp) which do not in-
clude, however, certain components of the stress tensor corre-
sponding to the general three-dimensional problem of the theory
of elasticity. Fig. 92 shows these components:
Zz \ R Z = ZT \
= er (9.15)
Sec. 57J EQUATIONS OF EQUILIBRIUM IN CYLINDRICAL CO-ORDINATES 271
To derive the differential equations of the three-dimensional prob-

lem us isolate from a body the element shown in Fig. 65 (it
let
is repeated in Fig. 92). Let us apply to it the stresses involved
in the plane problem (Fig. 66) and those which we have intro-
duced anew; it is necessary, of course, to take into account the
increments of components (9.15), which they receive on account
of the increments of co-ordinates dr, dQ, dz in passing from one
face of the element to another, as shown in Fig. 66 in respect to
the components of the plane problem. By projecting all forces
applied to the element on the three
axes we obtain three differential
equations of equilibrium which repre-
sent a generalisation of equations
(Ipp) of Sec. 47 of the plane problem:
dr
(9.16)
The first terms of each of
three
these equations can be formally ob-
tained from equations (9.1) by re-
placing the notations:
x by r, dx by dr\
y by 6; dy by rdft.
The containing the com-

last terms, Fig. 92
ponents themselves, are present be-

cause of the curvilinearity of co-ordinate surfaces r = const (cf.
Fig. 64); because of this the area add' a? (Fig. 92) is not equal to
the area bcc'b' and the areas cdd'c' and abb'a' are not parallel
',
to each other; these circumstances and their effect can be traced

in the derivation of equations (Ipp) in Sec. 47.
As for equations (9.1), one can obtain the general solution for
equations (9.16), its forms being still more various than in the
case of Cartesian co-ordinates (see the afore-mentioned works of
V. I. Blokh and Ju. A. Krutkov). The general solution also con-
tains three arbitrary stress functions. We present one of the pos-
sible forms of the solution expressed in terms of the stress func-
l
See the author's paper, Some Generalisations

1
of Lame's Problem for

an Elastic Parallelepiped, Prikl. matem. mekhan., i 1953, Vol. 17, No. 4, pp.
465-469.
tions:
(9.17)
-2 2 r\* "I /^-22 ~T~

r* dr '
r
*-
<?r
(9.18)
v*
r <te r c>r dz
Z fl
= 4r c)0^ r2
1 (
A? r dr dz
The generality of this solution is proved, just as in Sec. 56, by

determining the functions f\, / 2 /a from the equations of the first ,
line of (9.18) and by substituting them in the remaining equa-

tions. As in the case of the plane problem in polar co-ordinates,
formulas (9.18) are considerably simplified if the state of stress
is independent of the co-ordinate 6; then all the derivatives with
respect to this co-ordinate vanish; we obtain
r ^ 2 J
1 r)
_L _^_ ( f _ f) /i \ .
'
(9.19)
dr r f)^ v2 c)^ /
7 =
^ 1
^(r/ 3)
72--
However, a thorough examination of these formulas shows a pos-

sibility of further simplification. In fact, let us introduce three
new stress functions in place of f it / 2 , fs :
dz*
whence we find
'
dr
and formulas (9.19) assume a very simple form:

:
~5T+ (te 2
'
^T ""^"'
2
(9.20)
'
r dr r2 dr
These formulas enable us to distinguish two categories of prob-

lems which are important in the sense of practical applications.
1. Assume x = 0; then /? =Z =0; thus (Figs 66 and 92) there J 1
remain only R 2 =Zr among the shearing stresses and we have a

Sec. 58] HARMONIC AND BIHARMONIC FUNCTIONS 273
state of axisymmetrical three-dimensional stress-, but according to

Hooke's law
and, therefore, we have an axisymmetrical three-dimensional

strain; the nonvanishing components of the stress tensor /? r 0) ,
Z z and R z = Z r are expressed in terms of two stress functions CD

and ty.
2. Assume o) = = 0.
\t>
In this case only two components of the
stress tensor are different from zero:
*
- 72"
~~~
dz '
--.
~~
r* dr
'
they are expressed in terms of a single stress function. This solu-

tion is employed in the problem of torsion of a rod having the
form of a body of revolution.
The general solutions of the equations of equilibrium, present-
ed in the last two sections, do not by themselves give the solu-
tion of the elasticity problem, since the stress functions involved
in them must be determined from the compatibility conditions
(for instance, from Beltrami's equations
Cartesian co-ordinates)
in
and from the conditions on the surface a body; these solu- of
tions, however, are of essential help in solving problems by the
variational method elaborated by A. Castigliano; this method is
discussed in Chapter XI with an appropriate use of the solutions
in question.
58. HARMONIC AND BIHARMONIC FUNCTIONS

In the preceding chapters we encountered harmonic and bihar-
monic functions. It will be seen from what follows that the
solution of all problems of the theory of elasticity is very closely
bound up with these functions; in this section and the next we
shall, therefore, acquire greater acquaintance of some properties
of harmonic and biharmonic functions and the general methods
of constructing the latter.
By a harmonic function
<?(*, y, z) (9.21)
is meant, as is known, a function which satisfies Laplace's equa-

tion
V 29 = 0. (9.22)
18-1013
Harmonic functions are sometimescalled the Laplacian poten-

tials or Laplace's potentials. Let us take a system of Cartesian
co-ordinates xyz in space; then the field of a function <P(A;, #, ^)
is defined as that part of space at all points of which the func-
tion cp(#, y, z) takes on particular finite values and remains
single-valued. the function cp is a potential, an additional re-
If
quirement is made to it: it is necessary that its first partial deriv-

atives exist within the field; one has often to require that deriv-
atives of higher order exist as well; for instance, it can be seen
from equation (9.22) that the derivatives of at least the second
order must exist for Laplace's potential.
If all we require from function (9.21) is that it satisfy Lap-
lace's equation, we
notice that there exists an infinite number of
such functions. For instance, if a certain harmonic function is
known, then proceeding from it, we can obtain any number of
new harmonic functions by mating use of the property of Lap-
lace's operator indicated in formula (5.54) of Sec. 36. Indeed, if
<p(#, y, 2, a, &,c,.. .) is a harmonic function and a, 6, c, ... are
any parameters of coefficients entering in its expression, then one
can easily show that the partial derivative
where t is any of the quantities x, y, z, a, fe, c, . . . is also a

harmonic function. Let us verify this:
but, since <p is a harmonic function, V 2

(p
= and, consequently,
which was to be proved. It is clear that if there exist consecu-

tive derivatives of the function
2 3
<?
<F
<?
y
dt*
'
dt*
' eK"'
they are all harmonic functions, too.

By way of example, let us consider the function
or, more compactly,

T = 7- (9.23)
Sec 58] HARMONIC AND BIHARMONIC FUNCTIONS 275
where r= |/"jc
2
-f-
2
y -\-z
2
is the radius-vector from the origin to a
point (x, y, z).
Let us first show that function (9.23) is harmonic. Note first
that from the relation
r2 = +y jc
2 2 - 2
(9.24)
we obtain
dx
or
dr
~dx
~x
in the same way
9 26 '
77=7 -7- <
'
On this basis we obtain from (9.23)

d M _ _d_ /M
\
_dr_ ~~~ _ __ i_ _ __ __ _
f
dx\r) dr ( r } dx r* r r3
1 . r3 _ Q-2 f^ r
r
-r + .
(9.27,
Similarly, differentiating with respect to y and z, we have

&__ n\ 1
2
dy
(9.28)
Adding these equalities by members, we find
= o
3 +3
-- _, o - =
which was to be proved. Hence function (9.23) is Laplace's po-
tential. The field of this function is the entire unbounded space
except for a single point the origin, where function (9.23) be-
comes infinite.
The gradient of function (9.23) has the following projections
on the co-ordinate axes [cf. formulas (9.27)]:
dx \ r )
~ JL
r3
'
dy
(9.29)
-L '
r3
18*
However, we observe that

-. = cos (r, *);
= cos (r, y);
~ = cos (r, 2),
and then formulas (9.29) can be rewritten as
v
,
y), (9.30)
-^rj^-TrCosCr,
Hence we conclude that
(1) the magnitude of the gradient

is-^-;
(2) directed along the radius-vector r towards the origin.
it is
This kind of field is obtained if we take the origin to be a
centre which attracts a material particle according to the law of
universal gravitation (inversely proportional to the square of the
distance).
As the function harmonic function, by differentiating
is a
it we obtain a series of new harmonic functions. For instance,

equations (9.29) give us three harmonic functions:
x _
r3 (9.31)
Functions (9.28) are obviously harmonic as well.

Let us prove now a lemma which will be very useful for what
follows: if tp(#, y, z) is a harmonic function, we have the identity
V 2 (^) = 2||; (9.32)
indeed, by successive differentiation of the function xty we obtain
- =X T-
_X (9.33)
dy dy
Adding up equalities (9.33) by members and observing that

V2 = according
\j)
to condition, we see the validity of (9.32).
This lemma can obviously be extended as follows: if \|)i, \J? 2 tys ,
Sec. 58] HARMONIC AND BIHARMONIC FUNCTIONS 277
are any harmonic functions, then
,)
= 2 -&,
(9.34)
On the basis of these identities we can easily prove the following

very important theorem. If <p(x, y, z) is a harmonic function,
then
is likewise a harmonic function.

Since the derivatives of the function cp will also be harmonic
functions, we have the right to set in (9.34)
Next, adding up identities (9.34) by members, we find
which was to be proved.

The function O can be given a simple vector representation.
In fact, if r is the radius-vector of a point (x, y, z), then
", y), (9.36)

: = r cos (r, z).
Further, if V cp is the gradient of the function cp or its derivative
along the normal |Vcp|=-^- f then
'
=== *"^
"7T~ ~H \~7T~ '
y] *
(9.37)
Substituting from (9.36) and (9.37) in (9.35), we get
,
z)].
"278 MORE GENERAL METHODS OF SOLVING ELASTICITY PROBLEMS [Ch IX
or
= -/ cos (-,r). (9.38)
But
d
WO
"35~ VW ')~~dr
'
therefore, we obtain from (9.38) the following vector expression

of the function:
0> = A:-^4-v
dx ^^ y dy
-4-z
^^ dz
^r-^-.
dr
(9.39)
'
59. BIHARMONIC EQUATION

A differential equation of the form
2 2
V V /(;t, y, z)
= (9.40)
is sometimes called biharmonic.

Afunction / which satisfies this equation is called biharmonic.
Equations of the type (9.40) play an important role in the theory
of elasticity. We have already encountered in the plane problem
the case where the function / depends only on two variables
(x,y)-, it has been shown there that the stress function must sa-
tisfy the biharmonic equation (see Sec. 39)
Later we shall show that Lame's equations (VI) of Sec. 25 in the

absence of body forces (X=Y=Z=Q) and Beltrami's equations
(VII) of Sec. 36 can also be reduced to biharmonic equations.
Bearing this in mind, we shall now indicate some types of so-
lutions of the biharmonic equation, which are necessary for appli-
cations. These solutions were obtained by J. Boussinesq; they are
closely connected with harmonic functions, as may quite naturally
be expected if we take into consideration a very close relation
between harmonic equation (9.22) and biharmonic equation (9.40).
We first note that each harmonic function cp(x, */, z) is a so-
lution of the biharmonic equation as it satisfies the equation
V 2 cp=0 and, therefore, it certainly satisfies equation (9.40):
V 2 V 2 cp^V 2 = 0. We thus establish that the first type of solutions
of equation (9.40) is represented by any harmonic function.
1
This follows from the definition of the derivative of a function of a point
-along a given direction.
Sec. 59] BIHARMONIC EQUATION 279*
Let us now show that if \|) 4 , \(?2, ^3 are some harmonic functions,
the functions
will be solutions of the biharmonic equation. This can easily be

done on the basis of the deduction summarised in formulas (9.34).
Indeed, by applying the Laplacian operator to both members of
the first of equations (9.34), we obtain
But since a harmonic function, will likewise be a har-

\|H is
-^
monic function; consequently, the right-hand member of the last
equality is equal to zero, and hence
V 2 V 2 (;c^) = 0, (9.42>
which was to be proved.
By using in the same way the second and third of equalities
(9.34), we can show that the remainder of functions (9.41) are
also solutions of biharmonic equation (9.40). thus arrive at We
the second type of solutions of the biharmonic equation in the
form of functions (9.41).
If we have found two solutions fi and / 2 of equation (9.40),
their sum likewise is a solution of this equation. Indeed, it can-
easily be verified that
By applying once again Laplace's operator to both members of

we obtain
this equality,
but the right-hand member of this equation is zero, since /i and

/2are solutions of equation (9.40). It follows that
which was to be proved. By the use of this conclusion we obtain

the third type of solutions of the biharmonic equation by sum-
ming up the solutions of the first and second type:
(9.43).
where cpi, q> 2 , cps, t|?i, t|?2, % are any harmonic functions. Instead
of these functions, one can, of course, take any of their partial
derivatives. Later we shall need, for instance, three solutions

of the biharmonic equation of the following form [of the type of
the third of equations (9.43)]:
(9.44)
where cpi, cp 2 , cp 3
and \|)
are harmonic functions.
Let us now show that a function of the form
r^ (9.45)
is a solution of the biharmonic equation if \|) is a harmonic

function. For this purpose let us set up the Laplacian operator
-of function (9.45) ;
we get
dx
" dx Y-T'
but from (9.25) we have
2r = 2*, (9.46)
consequently,
-. T
2
r 2 -T- .
dx '
dx
Further,
or, by using (9.46),
Similarly, we get
By adding up the last three equalities by members, we find

Sec. 59] BIHARMONIC EQUATION 281
as tp is assumed to be a harmonic function, the last term van-

ishes, and we obtain
(9.47>
But the right-hand member of this equation is the sum of two

solutions of the harmonic equation [cf. formula (9.35)]; therefore,
it too is a harmonic function. Consequently, the left-hand mem-
ber V 2 (r 2 \|)) also represents a harmonic function. Therefore
= ;
(9.48)
hence, function (9.45) is in fact a solution of the biharmonic

equation. This is a new, the fourth, type of solutions. As
dtp
"dx dy dz
are harmonic functions, the functions
' '
dx dy dz
will likewise be solutions of the biharmonic equation. By sum-

ming up these solutions with the solutions of the first type cpt,
q>2, <Pa
we obtain the fifth type of solutions:
(9.49)
dz
If we add to the right-hand members solutions of the first type

in the form of functions
d<\>
a* _ '
dy dz
where a is a constant number, we obtain solutions of the follow-

ing, the sixth, type:
(9.50)
60.REDUCTION OF LAME'S AND BELTRAMI'S

EQUATIONS TO BIHARMONIC EQUATIONS
The foregoing particular solutions of biharmonic equations of
the types (9.44) and (9.49) are of great importance in the solu-
tion of the elasticity problem in terms of displacements since
Lame's equations, the basic equations involved in this method,
(VI)
in the absence of body forces, i.e., when = Q can easily

X=Y=Z J
be reduced to three independent biharmonic equations. This cir-

cumstance considerably simplifies the solution of the problem.
Omitting body forces, we write equations (VI) in the form
(VIO of Sec. 25:
(9.51)
V2 w + (k+} )*L = 0t
where, according to (3.17),
Accordingly, the equations of Hooke's law are written in the

form (3.18)
Z, =
(9.52)
Xy =
Thus, equations (9.51) and (9.52) are expressed in terms of two
elastic constants jn and k.
Turning now to Lame's equations (9.51), we recall that [Sec. 36,
equation (5.55)] in our case, when there are no body forces, the
dilatational strain satisfies Laplace's equation, i.e., is a har-
Sec. 60] REDUCTION OF BASIC EQUATIONS TO BIHARMONIC EQUATIONS 283-
monic function
V2 = 0.
The derivatives
<36 d6 ()Q
' '
dx dy dz
will likewise be harmonic functions; therefore
On this basis, applying Laplace's operator to equations (9.51 ),.

we observe that they reduce to the form
(9.53>
= 0.
Of particular importance that these equations are

is the fact
independent, i.e., only one of the unknown
functions u, v, centers
in each of them, and since these equations are biharmonic, the
conclusions of Sec. 59 will enable us to find without difficulty
any number of their particular solutions
l
.
Let us now turn to Beltrami's equations [(VII) of Sec. 36]. We

had [see formulas (5.55) of Sec. 36 and (3.14) of Sec. 18),
V 2 6 = 0,
e=(3A, + 2|j,)e. Hence we obtain V = 0, i.e., in the
2
absence of body forces the function Q = X x+ y+Z

Y is harmonic.
z
We also take into account that, for instance,
then, by applying Laplace's operator to each of Beltrarni's equa-

tions (VII), we obtain
= 0;
(9.54>
Hence, in the absence of body forces, Lame's and Beltrami's

equations reduce to biharmonic equations. A very important con-
clusion follows therefrom: all the basic unknown functions in the
general problem of the theory of elasticity
U, Vj W\
y
yv * Y
M
y
7
^2'
v2
M 7
t'xv
y
**
y
are biharmonic functions.
Though, as we shall see not

1
later, all of these solutions will satisfy (he
initial equations (9.51).
61.BOUSSINESQ'S METHOD; APPLICATION OF HARMONIC

FUNCTIONS TO SEEKING OF PARTICULAR SOLUTIONS
OF LAME'S EQUATIONS
Considerations of Sections 59 and 60 show that solutions of

Lame's equations can be sought as biharmonic functions by
assigning their expressions in the form of (9.44), (9.49) and
(9.50); the harmonic functions (pi, 92, $3, ty involved in them
should be determined so as to satisfy Lame's equations and pre-
scribed boundary conditions.
We shall investigate solutions of the type (9.44):
(9.55)
solutions of the type (9.49):
(9.56)
We shall not use solutions of the type (9.50) here and shall only
indicate that they are needed for the problem of equilibrium of
an elastic sphere. We should note, however, the following: in-
stead of Lame's equations (9.51), we have obtained, by differen-
tiation, equations (9.53) of higher order but simpler form and
found the solutions of these equations (9.55) and (9.56). One
cannot say, however, that (9.55) and (9.56) will necessarily be
solutions of Lame's equations (9.51): the solutions of Lame's
equations will always satisfy equations (9.53), since the latter
are a consequence of Lame's equations; but because of the higher
order of equations (9.53), they will have a wider class of solu-
tions than equations (9.51). We
shall thus have to ascertain in
\vhat conditions (9.55) and (9.56) will satisfy equations (9.51); to
this end it is obviously necessary to substitute (9.55) and (9.56)
in (9.51) >.
It will be noted that this procedure has been

applied more than once in
1
Sections 32, 33, and 34.

Sec 61] BOUSSINESQ'S METHOD
Let us examine solution (9,55). For substitution in (9.51) we

calculate first the functions
0; W, VV,
We have
du _
~~
a ?l
dx dx dx 2 '
dv dy> 2
~dy~
dw ~" 3
~" Z ~* '
"
'
Adding up these equations by members and taking into account

that the function ip is harmonic, we find
Further we obtain
as the functions cpi and \|)

are harmonic, we have
V 2
Tl
=
, according to (9.34),
in exactly the same manner we find
(9.58)
The results (9.57) and (9.58) are introduced in equations (9.51):
or
By making reduction in the square brackets, we get

The same computations for the second and third equations yield
These three equations are satisfied only if
where C is an arbitrary constant.

Equation (9.59) represents just the required condition in which
functions (9.55) will be a solution of the system of Lame's equa-
tions (9.51). The arbitrary constant C in (9.59) is of no impor-
tance and we set it equal to zero Hence }
.
<9 60 >
-
We come to the following conclusion: a particular solution of

the system of Lame's equations
'
dy
can be obtained in the form
(9.61)
where q>i, <p2, q>3 are arbitrary harmonic functions while the har-
monic function is determined from equation
\|> (9.60). In this
way one
\|>
can obviously obtain a great number of particular
solutions, there being available the arbitrariness of the choice of
harmonic functions cpj, (p 2 , 93.
1
Indeed, without violating the generality of equation (9.59), we can set
C = D(/z-f3), where D is a new arbitrary constant; we denote \|?
Dz=*tyi will
obviously be a harmonic function, just as ty; then (9.59) becomes
Sec. <51J BOUSSINESQ'S METHOD 287
Let us now consider the form (9.56) of the solution of bihar-

monic equations (9.53) and establish
what conditions this in
solution satisfies Lame's equations (9.51). We set up, on the
basis of (9.56), the expression of the volume expansion 0. We
obtain
*!L
dx
*2L
dx
+
^ JL
dx \
r (
* J*h
dx)'
(9 62)
(J '^>
but
d
dx
/>
\
^
dx}~
\
d < r2 )
dx
d*
dx~r
I
r2 ^
dx*
'
or, on the basis of (9.25),
We introduce this in (9.62); in a similar manner we calculate

the remaining derivatives required for 0:
du dy
^. = ^L + 2y^ + r*^. (9.63)

J19.I.
2
r -
By adding up these equalities by members and remembering that
we find
(9.64)
where the following abbreviated notations are introduced !

:
.-* + +T&- <M5>
We now proceed to set up Laplace's operators:

V 2 tf, V 2 i;,
The first of equations (9.56) yields
(9.67)
since
1
See formula (9.39).
To write out the right-hand member of (9.67) we make use of
formula (9.47), replacing in it ^by--:
(9.68)
^ '
We transform the bracketed expression in the right-hand member

on the basis of the following considerations:
d*ty
dx 2
= dx
d I
\
a|/\
dx } dx '
T - \t
y
y y
dxdy dx \ dy /
'
JLfrJW
(*)
.
dxdz dx
Adding up these identities by members and using notation (9.66),

we find
~ ~~ '
dxdy dxdz dx dx
Introducing this in (9.68), we get
V =2 +4 = 2<* + 2*). (9.69)
In exactly the same way we obtain
(9.69)
We substitute (9.64) and (9.69) in the first of equations (9.51):
or
The other two of equations (9.51) yield:
Hence we conclude
(k + 1) 6 -f 2<|. H- 2 (* + 3) $=
Sec. 61] BOUSSINESQ'S METHOD 289
Without loss in generality of this equation we can set its con-

stant right-hand member equal to zero dividing the equation
1
;
through by 2(* + 3), we find
-
(9.70)
Recalling now that according to formula (9.39) of Sec. 58 the

function d> has the vector expression
equation (9.70) is rewritten as
r
"W "^ k +3 ^ =~ 2(6 + 3) '
or
r-^- + = at)> p0 , (9.71)

where
If we pass from the Cartesian co-ordinates x, y, z to another

system in which one of the co-ordinates is the radius-vector r
(for instance, to polar co-ordinates in space), the derivative p

of the function \|) along the direction r may be treated as the
partial derivative of \|) with respect to the variable r\ obviously,
the other two co-ordinates determining the direction of the radius-
vector r are assumed to remain constant. If the same reasoning
is applied to the function equation (9.71) may be looked upon
,
as an ordinary linear differential equation with respect to the

function ^. By integrating it, we find \|) as a function of the ra-
dius-vector r\ however, the arbitrary constant, which will be in-
volved here, will in general depend on the other two co-ordinates,
i.e., it will depend on the direction of the radius-vector r.
Equation (9.71) is integrated by the usual method of substitu-

tion:
ty
= UV, (9.72)
where U and V are two new functions of r. Substituting this in

'(9.71), we find
or
rUV + (rU + *IS) l/= r
p0 .
(9.73)
1
This constant can be incorporated in either of functions 0o or t|)
in the
left-hand member of the equation.
19-1013
We select the function U from the condition
r/'-)-a/ = or r-r- = a(/. (9.74)

Hence
dU _ <*
dr
U
Integration yields
lnf/ = alnr or / = r-<". (9.75)
On the basis of (9.74) and (9.75) equation (9.73) takes the form
dV
or
Multiplying both members by dr and integrating, we find
Substituting from this and from (9.75) in (9.72), we find the

function \|r.
.
(9.76)
Here, as pointed out above, the constant C depends in general

on the direction of the radius-vector r, i.e., when C is equal
to a constant number, (9.76) gives a law of variation of the
function with a motion in the chosen direction drawn from
t|>
the origin.
62. EFFECT OF A LOAD ON A MEDIUM BOUNDED BY A PLANE

(BOUSSINESQ'S PROBLEM)
A was treated in Sec. 49 for the case of plane
similar problem
strain; there that the load was distributed uniformly
we assumed
along an infinite straight line in the plane bounding the medium.
Here we shall indicate a method of solving the problem in the
general case when the medium under consideration is acted upon
by an arbitrary load; the medium itself will be called the half-
space for brevity, and the plane which bounds it the boundary.
The origin of co-ordinates is chosen somewhere on the bounda-
ry; the axes Ox and Oy are directed in the boundary plane and
the axis Oz normally to it along the outward normal. Then the
Sec. 62] EFFECT OF LOAD ON MEDIUM BOUNDED BY A PLANE 291
problem in question can be formulated as follows: it is required to

find the stresses and displacements at any point of the half-space
(i. e., for any x, y and z less than zero), if the stresses due to the
load X z y z Z 2 are prescribed on the boundary z
, , as functions =
of x and y. If the displacements u, v, w, rather than the stresses,
were prescribed on the boundary, the solution of the problem
could be found directly in the form (9.61). Indeed, if it is speci-
fied that for 2 =
on the boundary
tt=/i(* y); v=f*(x* y); w=fz(x, y),
then, putting z = in (9.61), we find that

<?,(*, y, 0) =/!(*, y); 2 cpa are be found from , to
the condition that they turn into preassigned functions ft, / 2 fs ,
on the boundary 2 = 0; this is the basic boundary value problem

of the theory of potential, which admits quite a definite solution;
after determining the functions q> lf q> 2 and <p 3 we find the func- ,
tion \j? by (9.60) and, introducing all this in (9.61), we complete

the solution of the problem.
However, it is the stresses that we consider to be given on the
boundary and, therefore, the procedure of solving the problem will
be somewhat complicated. Let the stresses on the boundary be
given in the form of functions of x and y as follows:
Z, = IMx, y); Z je
=^ 2 (x, y); Zy = O3 (.*, y). (9.77)
Since we are
solving problem the in terms of displacements
(Chapter IV, Sec. 24), we express the left-hand members of these
equalities in terms of displacements u, v, w by using Hooke's
law (9.52) and Cauchy's equations (III) of Sec. 24:
> --
(2
(9.78)
These equations are valid only on the boundary, i.e., for

2=0. Let us now assume that we have found three harmonic
functions:
<o
>,(.*, y, 2); z (x, y, 2); <
3 (x, y, z)
such that on the boundary 2=0 they turn into prescribed func-
tions (9.77), apart from a constant factor n which we introduce
19*
to simplify subsequent computations:

V*i(x> y, 0) *!(*, = y),
V, 0)
=* 2 (x f
y), (9.79)
y, Q)
= 3 (x, y).
We substitute these values of d>i, O 2 and <D 3 in the right-hand

members of equations (9.78) and replace a, vand w in the left-
hand members by their expressions from (9.61) and (9.57); we
obtain
*
3
In these equations we omit the terms involving the factor 2 (as

2= on the boundary) and note that on the basis of (9.57) and
(9.60)
.2 .
3
""
.
_ '
"
.
_ 2 d^ *
dy
'
dz dz 1 dz 1 dz
(9.80)
Then the foregoing equations become

d? 3 , 2 d<\>
2
"dz" '
k + l ~dz~~
dz (9.81)
dz "^
dy
= 3
.
dy
These equations, as already stated, hold on the boundary of the
half-space. Both their left-hand and right-hand members, how-
ever, are harmonic functions; it is known that if two harmonic
functions coincide on the boundary of a region, they are identical
in the whole region It follows from this that equations
l
.
(9.81)
are valid in the whole half-space 2<0. This conclusion enables
us, by eliminating the three functions cpi, cp 2 and cp 3 from (9.81)
and (9.60), to express ty in terms of coi, <o 2 co 3 which are assumed ,
to have been determined previously for any point of the half-

space.
1
This theorem is valid if the above functions, besides satisfying Laplace's
equation, have no singularities in the region of existence and particularly if
they do not become infinite.
Sec. 62] EFFECT OF LOAD ON MEDIUM BOUNDED BY A PLANE 293
We differentiate the first of equations (9.81) with respect to z,

the second with respect to x and the third with respect to y and
add up the results by members; in the equation thus obtained a
number of terms vanish since the functions cpi, q> 2 <pa and \|? are ,
harmonic and therefore

=V =V =V =
2
cp 2
2
cp 3
2
<}>
0.
We find the following result:
_ "
k ~~ 1 ^ ^~ I / *?! I
dx "^ dy "^ dz
<*?2 I <??3 \ _
~~ d*l "^
<toi
do) 3
dx ~^ dy
|
+
'
k 1 dz* dz \ ) dz
In order to eliminate the functions <pi, 92, cpa from this we make
use of equation (9.60) and obtain finally
After this has been done, the procedure of solving the problem
is as follows:
1. According to the prescribed functions cp lt d> 2 <J> 3 [see equa- ,
tions (9.77)], we find harmonic functions i,

co 2 co 3 satisfying ,
conditions (9.79) on the boundary.

2. By double integration we find the function \|) from (9.82).
3. From (9.81) we seek harmonic functions <pi, and <p in (9.61), we find dis-
placements u, v, w.
5. From Hooke's law (9.52) we find the stresses at an arbi-
trary point of the half-space.
Subsequently we shall confine ourselves to such cases of loading
on the boundary where the resultant is a finite quantity. Then,
as z increases beyond all bounds (i.e., for an infinitely removed
horizontal section of the half-space), all the displacements and
stresses must tend to zero; hence, on the basis of equations (9.61),
we note that as z-> oo the function \j? and its first derivatives
must likewise approach zero. We shall make use of this remark
in determining the function t]> from equation (9.82). By integrat-
ing it twice with respect to z from oo to an arbitrary value of 2,
we obtain
CO OO
But for z oo the left-hand members of both of these equal-

ities must vanish according to the condition; since the integrals
in the right-hand members also vanish, we conclude that
/1=/2 = 0,
and we finally obtain
* T / oo
/(* +
oo
+
The first stage of the foregoing procedure of solving our prob-
lem makes it necessary to solve the Dirichlet boundary value prob-
lem for harmonic functions coi, o) 2 003 [see formulas (9.79)]. As ,
shown by Boussinesq, this problem, however, can easily be solved

in the case where the load consists of a single concentrated
force applied to any point of the boundary; hence one can pass
to the case of an arbitrary load by approximately the same pro-
cedure as was employed in Sec. 49 in the corresponding plane
problem.
63. EFFECT OF A CONCENTRATED FORCE
NORMAL TO THE BOUNDARY AND APPLIED AT THE ORIGIN

The conditions on the boundary of the half-space [see equations
(9.77)] in this case are as follows:
1. Functions 2
= 3=0<I ) throughout the boundary.
2. Function (Di is zero throughout the boundary except at the
origin (*
= = 0),
*/ where it is infinite.
From conditions it is necessary to find [cf. conditions

these
(9.79)] harmonic functions coi, co 2 co 3 But it can be proved that ,
.
if a harmonic function is equal to zero on the boundary of a

region, it is equal to zero everywhere in the region; therefore, from
the conditions d>2 = 03 = on the bounrday of the half-space we
conclude that (02 = 0)3 = in the whole half-space.
It remains to find function G)i(A;, y, z) from the condition that
it is zero throughout the boundary (z
= 0) except at the origin
(*
= */ = z = 0), where it has a singularity, for it becomes infinite;
the nature of this singularity must be examined more closely to
determine better the behaviour of the function on the boundary
and in the vicinity of the singular point.
Let us cut out a hemisphere of radius r with centre at the
origin from the half-space; from the condition of equilibrium
of this hemisphere it follows (Fig. 93a) that
Sec 63] EFFECT OF CONCENTRATED FORCE NORMAL TO BOUNDARv 295
where Z v as usual, denotes the projection of the stress over the

,
area dS hemisphere on the axis Oz\ the in-

of the surface of the
tegral is extended over the surface of the hemisphere. According
to the mean value theorem for integrals we obtain
J
Zv dS = Zvm f dS = Z v
where Z vm is a certain average stress on the surface of the
hemisphere. Hence
and, finally,
(9.84)
This conclusion refers to a certain average stress on the surface

of the hemisphere; from this one can draw, however, an impor-
tant conclusion that in general the stresses in the half-space due
Fig. 93
to the force P, as the point of its application is approached, in-

crease at a rate of the same order as (9.84) where r is the dis-
tance from a given point to the point of application of the force.
But, as is seen from (9.77) and (9.79), the function coi is a
quantity of the same order as the stress; consequently, it should
likewise
(1) vanish everywhere on the boundary except at the origin;
(2) become infinite in the same order as ^ when r->0;

(3)be a harmonic function, i.e., satisfy the equation V 2 coi =
everywhere except at the origin
l
.
1
In addition, as pointed out above with respect to equations (9.81), it
must remain finite everywhere in the half-space, i.e., for 2<0, including the
case of x, y, 2 tending to infinity.
296 .
MORE GENERAL METHODS OF SOLVING ELASTICITY PROBLEMS [Ch. IX
All these conditions are satisfied by the last of functions (9.31)

indicated in Sec. 58; we take it in the form
(J)
i
= -7r.
Cz
where C is an arbitrary constant of integration; later we shall

relate it to the magnitude of the force P. Thus, we assume
CD
I
= ;
co
2
= = o)
3 0. (9.85)
We proceed to seeking the function \|>;

from equation (9.83) we
have
* r /
oo
/***
oo
T
But in Sec. 58 we saw that [cf. (9.27)]
JL-J-
r
~~
3
dz
Therefore,
We calculate the functions <pi, cp 2 , qpa from equations (9.81). The

first of them gives
2 9^1 c
2 dty _
~ Cz . 2 C z _
~"
k+2 Cz
dz 1
k + 1 dz r3
"t"
* + 1 2 r2 AJ
+ 1 r3
'
but
Therefore, we have
By using the second of equations (9.81), we find cp
foi _
~ ^3 ^_
~ C(l2a) A:
dz dx dx 2 r3
In taking this integral, we recall that
r = yV-K 2
, where 2 =
Sec. 63] EFFECT OF CONCENTRATED FORCE NORMAL TO BOUNDARY 297
and apply Euler's substitution

r = +f *,
where t is a new variable. We have
Introducing this in the integral, we obtain
( *L ~ A t tdt
+
~~~
*
~~
J r* ^J (*
2 2 8
5 ) *
2
+ 6
a
r(r *)
'
and, therefore,
In the same manner, by using the third of equations (9.81), we

l
find
Now we have only to introduce functions (9.86), (9.87), (9.88)

and (9.89) in equations (9.61) and obtain the following expres-
sions of displacements:
(9.90)
In addition we shall
the expression of the volume
derive
expansion required seeking the
in stresses. For this purpose it
is easiest to make use of equation (9.80) by introducing therein
the value of the function \|> from (9.86):
= -r = ^r = ^ /t
c (l-2^r-
N /rk rnv
(9-91)
With expressions (9.90) and (9.91) we can calculate the stress

components by the formulas of Hooke's law (9.52) and Cauchy's
equations (III). Omitting the corresponding computations, we
Here, as in (9.87) and (9.88), in evaluating the indefinite integrals we

1
omit arbitrary constants in view of the fact that functions cpi, <p2 cpa must ,
vanish as z~>o.
present the final result:
K,
= Cp{-2gL-(l-2
(9.92)
Zr =
+(i-: r (/-)*
It now remains to C involved in the
determine the constant
expressions of stresses so as to make the load on the boundary (at
the origin) reduce to the given force P. For that it is sufficient
to require that the resultant of the normal forces Z z dS (Fig. 936)
over any horizontal section of the half-space at a constant depth
z= h be equal to P (P is the magnitude of the force which
is assumed to be compressive). Hence we obtain the condition
where the integral is extended over the whole plane 2 = h

parallel to the boundary. Substituting for Z2 its expression (9,92),
we find
-3CpA 3 /4? = -A (9.93)
To evaluate the integral involved here we make use of the

polar co-ordinates p and cp (Fig. 936). Let us transfer the inte-
gral to these variables. We obtain
The element of area in polar co-ordinates, as is known, is dS =

= pdpdcp. Introducing all this in the integral of
equality (9.93),
we have
or
v2
Sec 63] EFFECT OF CONCENTRATED FORCE NORMAL TO BOUNDARY 299
we carry out integration with respect to cp and then with respect

to p:
p<*p _
~~
1
f <*(P
2
+ /*
2
) _
~~
1 1
3
'
1 2 J 5. 2 3/z
( p 2_|_/2 2)2 ( p2 _j_/z 2)2
The preceding equation thus yields

-
r .
and hence
C|x-2ic = />; C=
-^p (9.94)
It remains to introduce this value of C in equations (9.92) and

(9.90) to obtain the final expressions of stresses and displace-
ments.
As formulas (9.90) and (9.92), it is necessary to make the
for
reservation we did at the beginning of Sec. 49 for a similar plane
problem: expressions (9.90) and (9.92) are valid throughout the
half-space except for a small region near the origin (the point
of application of the concentrated force) where the stresses are
beyond the elastic limit of a given material. In this region Hooke's
law, on which the entire derivation is based, does not hold; the
applicability of our derivation in the remaining part of the half-
space, just as in the problem of Sec. 49, is determined by Saint-
Venant's principle.
The foregoing solution of the problem of a concentrated force
acting on the boundary of the half-space was based on the expres-
sions of displacements in the form (9.55) subject to condition
(9.60).
The problem of a concentrated force acting at a point of an
unbounded space is solved in a similar way; for that it is neces-
sary to accept as a starting point the expressions of displacements
(9.56) subject to condition (9.76) >.
A. E. H. Love hasshown that Boussinesq's solution (9.90) and

(9.92) concentrated force can be extended to the case of a
for a
distributed load of intensity </(, r]) acting on the boundary of
the half-space, where g, t] are the co-ordinates of a point in the
loaded part of the boundary; we can then assume that the ele-
ment of area dS at this point is acted upon by the elementary con-
centrated force
(9.95)
1
See E. Trefitz, Handbuch der Physik, Band VI. "Mechanik der elastischen
Korper", Berlin, 1928.
Summing up the effects of such forces, we obtain the displace-

ments and stresses in the elastic half-space due to the given
distributed load. Love's formulas for displacements in this case
are given in his treatise on the theory of elasticity, already
cited, and also in L. S. Leibenzon, Theory of Elasticity, 2nd ed.,
Moscow, Gostekhizdat, 1947.
It is very easy to obtain the most important formula in appli-
cations for the displacement w of a point on the boundary z =
perpendicular to it; for the elementary load (9.95) we have,
according to the third of formulas (9.90) and formula (9.94),
for 2 =
where
is the distance from the load (9.95) to the point of the bound-
ary (x, y) at which the displacement is being determined. The
expression of dw must be summed up with respect to the coordi-
nates and rj for all points of the loaded area; we obtain the
expression of the total vertical displacement of the point (x, y) of
the boundary:
(9.96)
The case of a load distributed on the boundary of the half-

space served H. Hertz as a starting point for solving the prob-
lem of compression of two bodies bounded by curvilinear sur-
faces. In Hertz's theory formula (9.96) is fundamental. This same
formula underlies many works concerning the theory of analysis
of soils as elastic foundations for various kinds of buildings and
structures.
We have considered in this section the case of a concentrated
force normal to the boundary of the half-space. If the force P
is directed in the plane of the base, e.g., along the axis Ox,
the whole procedure of solution will be changed only in the
sense that in place of (9.85) one must take ^= u>

3
= 0, <i)
2
= -^
and carry out all subsequent computations accordingly.
With the solutions for the cases of vertical and horizontal forces
one can, of course, obtain displacements and stresses for a force
inclined at an arbitrary angle to the boundary.
The foregoing Boussinesq's solutions of the form (9.55) and
(9.56) expressed in terms of three arbitrary harmonic functions
<pi, (pa* <ps are not general solutions of the equations of the theory
Sec. 64] SOLUTION OF PLANE PROBLEM 301
of even though they make possible the solution of a

elasticity
great number af problems. The general solution which is |>res-
sed, in the absence of body forces, in terms of three arbitrary bi-
harmonic functions was found by Academician B. <3. Galerkin l
.
64.SOLUTION OF THE PLANE PROBLEM OF ELASTICITY

BY MEANS OF FUNCTIONS OF A COMPLEX VARIABLE
In the preceding sections (58-63) it was ascertained that har-

monic functions of three variables x, y, z were of great impor-
tance in constructing the solutions of basic equations of the theory
of elasticity. Turning to the plane problem we note that we shall
work here with harmonic functions in the plane which are very
closely related to analytic functions of a complex variable
We shall now carry out sufficiently simple computations of a

formal character which will confirm this relation and at the same
time prove useful for what follows. We replace the independent
variables x and y by other variables z and z according to the
formulas
-- =
(9.97)
Obviously, z and z are two conjugate complex numbers. By

differentiating (9.97) with respect to x and y, we find
^L ~ i
*'
dz
dx dy
(9.98)
Let us consider a function f(x, y) and assume that its inde-

pendent variables x and y are replaced by the variables z and z
according to (&97). Then, in differentiating this function with
respect to x and y it is necessary to apply the rule of differen-
tiation of a composite function. For instance,
df _ df dz _ df dz
dx dz dx dz dx
See, for instance, L. S. Leibenzon, Theory of Elasticity, 2nd Moscow*

1
ed.,
Gostekhizdat, 1947, Chap. IV, See. 1ft
or, by using (9.98), we obtain
dx <)z dz
and similarly (9.99)
dy dz
From (9.99) we immediately find the equalities which will be

required later;
dx dy
(9.100)
dx
^ l ~~~
dz
'
dy
Differentiating (9.99) with respect to x and y respectively and

using again (9.98), we get
2
d 2f ()
f o d 2f d 2f
'
~dx 2 ~l)z 2
dzdz d~z 2
d2f d2f
dz dz dz 2
Adding up these equalities by members, we obtain the formula

of transformation of Laplace's operator to the variables z and z:
t"79 J!
v / = T7
O J
dx
O
+ TT
2
I J
2
:===4
A
:Hr
dz dz
r
" J
-
/C\ 1
(9.101)
^ 1 \
dy
Thus Laplace's equation V 2/=0 in the new co-ordinates becomes
dzdz
and, therefore, we immediately find its general solution in the
form
, (9.102)
where Fi and F2 are arbitrary analytic functions of the variables
z and z (respectively). The correctness of solution (9.102) is
verified by differentiation.
Any harmonic function can thus be represented as the sum of
two functions of complex variables z and z, according to (9.102).
This fact underlies several methods of application of a complex
variable to the plane problem of the theory of elasticity. The most
significant contributions in this respect have been made by
G. V. Kolosoff and especially by N. I. Muskhelishvili who devel-
l
G. V. Kolosoff, On One Application of Functions of a

1
Complex Variable,
to the Plane Problem of Elasticity. Jurjev, 1909.
Sec. 65] FILON'S METHOD 303
oped Kolosoff's method and worked out a complete theory of this

question
l
. Muskhelishvili's investigations and his method have
been widely used and inspired a great number of works constitut-
ing a special direction in the development of the theory of elas-
ticity during the last decades.
Among earlier investigations we should point out A. E. H. Love's
work set forth in his course on the theory of elasticity and also-
the method proposed by L. N. G. Filon as far back as 1903,
Filon showed that by changing the variables considered in the
preceding section it is possible to reduce Lame's equations in the
plane problem to the form that lends itself to integration in quad-
ratures, and to find their general solution. He gave, however, no<
essential applications of the important result obtained and the
method was forgotten. We present here Filon's solution because it
is obtained by a simple and natural course of reasoning and per-
mits the reader to approach that stage of analysis which opens
ways to the effective method of solving the specific problems elab-
orated by N. I. Muskhelishvili and his school.
65. L. FILON'S METHOD

Consider the case of plane strain in the absence of body forces;
Lame's equations for case will be obtained from (9.51) of
this
Sec. 60 if we set w = Q and regard u and v as functions of x and y
only:
(9.103)
where
0=^-4-4^-;
dx*dy
V2 ==-^
dx2+ -^-.
'
dy
2 (9.104)
^.IVT/
To transfer to new independent variables z and z we multiply

the second of equations (9.103) by i and add it to the first;
we obtain the following two equations:
(9-105)
N. Muskhelishvili, Some Basic Problems of the Mathematical Theory of

1
I.
Elasticity, Groningen, P. Noordhoff, 1953.

304 MORE GENERAL METHODS OF SOLVING ELASTICITY PROBLEMS (Ch. IX
We now replace the unknown functions u(x, y) and v(x, y) by the

new ones U and C/, according to the formulas
U=u ii).
(9.106)
By using (9.99), we have

du du ,
du
dx dz dz
dv . dv . dv
dy dz dz
Adding up these equalities by members, we have

a du . dv d / . .
x
. d .
or, on the basis of (9.106),
}
_dU . dU
'
(9.107)
dz d7
Thus is expressed in terms of new functions (9.106) and new

independent variables z and z. Furthermore, to transfer (9.105)
we have from (9.101) and (9.100)
dzdz
(9.108)
d2 U
dzdz
j!L
+ / Jli = 2 4i = 2 - (
dU
I
d[
L\
dx dy dz dz \ dz dz /
(9.109)
dx
/ =2 dz
=2 / 4-
dy dz \ dz dz ,
Introducing (9.108) and (9.109) in equations (9.105), we represent

them in the following form:
dz
Sec 65] FILON'S METHOD 305
By performing integration, we obtain (after some reduction with-

in the brackets)
where f{(z) and f'^(z) are arbitrary analytic functions of the

variables indicated in the parentheses. Primes denote their de-
rivatives. We solve the last equations with respect to dU/dz
and dU/dz:
Tg- 1* -i- o)
1
j l (Z) (ft -t- u/ 2 i*jj,
(9.110)
By integrating these equations, we obtain
(9.111)
where and 92(2) are two new arbitrary analytic functions

cpi(z)
of the variables z and z (respectively). Expressions (9.111) give
the general solution of Lame's equations (9.103); if U and are
known, we can find the displacement components u and v accord-
ing to (9.106):
=i
2i
(*+!) [*/ (z) + z/; (z)] +^ + (^) cp 2
(9.112)
In addition to these formulas let us write the expression of

the volume expansion by using (9.107) and (9.110):
A * r jn
20-1013
To determine the components of stress we take into consider-

ation Hooke's law in the form of equations (9.52):
%-(*+*) (9.114)
dv ,
da
By introducing here the values of u, v and from (9.112) and

(9.113) we obtain
2j- {2 (A +1) [/;(*) + /; (z)]

(*+!) [z/; (z) + zf'( (z)} + cpj (z) + cp 2
(z)}
.
= 7(Ff2) I
2 (k
_ _ _ (9-115)
1 )
[z/2
*
(z) 4- zf\ (z)}
- cp,'
(z)
- ?2 "
(z)}
,
1
I
The right-hand members of formulas (9.112), (9.113) and

(9.1 15) are, in general, functions of the complex variables z and z,
but according to the meaning of the problem they must be real.
This will be achieved if we adjust the four so far arbitrary analytic
functions _
/i (). /a (2), ?i() and cp 2 (J) (9.116)
so as to have _ __
/2 (J) = /,(z), 1
_ _ _
<P2 (2)
= 9l (2),
}
|
(9.117)
where /i(2) and 91(2) are the conjugate expressions to f\(z) and 1
q>i^) respectively, i.e., if
(9.118')
then
/2 (2) = /i (2)
=x (*, y) /t (je, y).
(9.118")
<p 2 (2)
=f , (2)
= xi (Jt, y) *<h (x, y).
Sec. 65] FILON'S METHOD 307
Indeed, it can easily be shown that if conditions (9.117) are

satisfied, the displacements u and v defined by equalities (9.112)
and, consequently, stresses (9.115) will be real quantities.
In fact,
(9.119)
and, using (9.97), we readily find that
Substituting all this in (9.112) and (9.113), we obtain after can-

cellation and simplification
(9.120)
k +2 dx '
where functions x Xi î are connected by the Cauchy-Riemann

equations:
' '
dx dy dx dy
^_ ' '
(9.121)
dy dx dy dx"
Thus formulas (9.120), subject to condition (9.117), give real

values for a, v and 6, as we set out to prove. It is evident that
formulas (9.115) will give real values for the components of stress.
Because of conditions (9.121) all the functions x ^ Xi. î. in-
volved here are harmonic.
Formulas (9.120), subject to conditions (9.121), give the general
solution of the plane problem for the case of plane strain in terms
of functions of real variables x, y.
One can, however, proceed with the solution in terms of func-
tions of the complex variables z and z taking as a basis the fore-
going formulas (9.111)-(9.115) and (9.117). Indeed, one can thus
easily obtain the fundamental formulas of the methods of A. E. H.
Love and N. I. Muskhelishvili mentioned at the end of Sec. 64.
20*
In addition to the last of formulas (9.120) we derive the expres-

sion of the component of rotation
2lO,
z
---du
= dx
dv
TT .
dy
By using formulas (9.112), (9.117) and (9.118), we get
2.=4tffc-/;<*>]=
Thus,
or, on the basis of (9.119), we find
(* + 2) 8 + 2/o>
2 =/;(*). (9.122)
This formula was obtained by A.E.H. Love from Lame's equa-

tions (9.103) and is fundamental in his method.
In Muskhelishvili's method the starting points are:
(1) the first of formulas (9.111) which, on the basis of (9.117),
takes the form
i &+ ?i &l '

< 9 123 >
'
(2) the following two formulas which can easily be obtained

from (9.115):
** + Y*=*f ((*) + /[(*).
(9.124)
The first of these formulas allows us to find the complex expres-

sion of Airy's stress function employed in Chapter VI. Indeed, the
left-hand member of this formula represents the Laplacian opera-
tion upon the stress function:
Its complex expression is given in formula (9.101); on this basis

we have
Double integration yields
where o>i (^) and 0)2(2) are arbitrary functions; it .is easy to verify
that the bracketed expression is a real quantity; in order that the
Sec. 65] FILON'S METHOD 309*
sum of the last two terms be real we must have

<i(z) =u 2 (z).
Thus the expression of the stress function is
Let us clarify the meaning of the functions 0)2(2) and co 2 (z)

involved; to this end we first find the stress Y v from this formula:
and then we obtain the same quantity according to the second of

7
formulas (9.115), taking into account (9.1 18 ) and 9.118"):
Comparing the expressions obtained, we find that
Formulas (9.123), (9.124) and (9.124') will coincide in this

condition with those given by N. I. Muskhelishvili if we replace
the functions /i and q) 4 introduced here by other functions cp and
t|) according
to the formulas
(9.125>
and if we set
_ _.
k+l
All our reasoning pertains to the case of plane strain; for the-
case of generalised plane stress Hooke's law takes the form
Hence
2
where
V = e xx + e yr
Therefore we have
Xx =
yy
i
where
, 2* 2a
*
1 a
Thus in the case of generalised plane stress all formulas Hold good
if the elastic constant k is replaced in them by k* '.
66 WAVE EQUATIONS
In Sections 26-28 of Chapter IV we established that Lame's
dynamic equations
(9.127)
permitted the solution of problems involving small motions of

an elastic body and considered the simplest cases of propagation
of small uniform vibrations of an infinite elastic medium, de-
scribed by the differential equation of the form
where / is a t. This is the simplest form of a

function of x and
"wave equation" l
.the present chapter we derive wave
To conclude
equations for the general problem of propagation of elastic vibra-
tions assuming, as before, that body forces are absent; therefore,
they are omitted in equations (9.127).
Let us apply to equations (9.127) the procedure that was used
in Sec. 36 to obtain equation (5.55), i.e., we differentiate them
M'th respect to x, y and 2, respectively, and add; we take into
account that (for constant density p)
Ox \~di*~
-.
dt* \dx>
Cf. equations, (4.8), (4.11), (4.26).

Sec. 66] WAVE EQUATIONS 311
the other two equalities of the same form are obtained replacing
x by y and by z; we get
(9.129)
or
2
V 2 0, (9.1290
-^Lr=a
where, as in (4.9), we designate
We now differentiate the third equation (9.127) with respect to

y and the second with respect to z and subtract one result from
the other:
&w _
~
dv 2
dw dv
^
2 /
\~~dy ~dz)
\
p W r) /
\~dy ~dz)
\ m
'
but, according to (2.9),

dw -- dv
-.
*~^
dz
_
-~ ^CO xK
dy
and, therefore, the preceding equation reduces to the previous form

(9.129)
, (9. 131 a).
where [cf. formula (4.12)]
&= -.
(9.132>
A cyclic change in (9.131a) gives two equations of the same form:
Equations having the structure of (9.129) and (9.131a,b) repre-

sent "wave equations" of the general form. The coefficients a and
6, as we saw in Sec, 26, represent the velocities of propagation of
plane waves of two kinds: waves of dilatation and waves of distor-

tion. Ifwe succeed in integrating equations (9.129) and (9.131a,b),
there will be determined everywhere: the volume expansion 0, i.e.,
the divergence of the displacement vector u(u, v, w), and the-
solenoidal vector (curl) cêo*, (o y co z ). ,
In vector analysis it is proved that, if 6 and co are known, one

can find the vector u at every point of the field, i.e., one can
determine the displacements of the points of an elastic space }
.
Consider two limiting cases separately:

1. = throughout the space;
2. 0)*
= 0)^ = 0)2 = throughout the
space.
In the first case the wave process takes place without change
-of volume; this, as is seen directly from Lame's equations (9.127),
results in wave equations of the general form for the components
of the displacement vector:
(9-133)
where notation (9.132) is introduced. Waves of this type are called
equivoluminal (variously called distortional, rotational, shear,

transverse, and, in seismology, S waves); they travel with the
velocity
-*/?
The second case x
o).
= o) = o) =
1/ z occurs when the displacements
u, v, w have a potential (see the end of Sec. 10 of Chapter II)
Proceeding from these relations, we transform the first of Lame's

equations (9.127):
fl
_ du ,
~T~
dv dw _
d**
"^ dz ~~ dx*
.
"*" "^
__ - 2<1>
~ v .
'
dx dy dy*
2
therefore,
It remains to substitute this in the first of equations (9.127);

making an analogous transformation in the other two equations
1
See N. E. Kochin, Vector Analysis, Moscow, Objed. natichno-tekhn. izd.,
.1938, p. 223.
Sec. 67] SOME PARTICULAR SOLUTIONS OF WAVE EQUATION
and introducing notation (9.130), we obtain

2
d' u
dt 2
(9.134)
w
d 2
j-
= a V w.9T79
2 2
We have again wave equations of the general form for the

displacement, just as in the case of equivoluminal waves (9.133);
waves, corresponding to the case under consideration, are called
irrotational waves of dilatation (variously called dilatational, bulk,
longitudinal, and, in seismology, P waves). Their velocity of pro-
pagation a, as we already know from Sec. 26, is considerably
greater than that of equivoluminal waves b. It is clear from the
foregoing that the fundamental problem of the theory of elastic
waves is the problem of integrating the general wave equation
where
dx + 4!( + ^/;
V 2/==
J |V 2 2 2 (9
v .136>
dy
'
dz '
c is the velocity ofwave propagation. The simplest wave equations

of uniform vibrations (plane waves), treated in Sections 26 and 28,
result from (9.135) if / depends on one co-ordinate only, for <
example, on jc, since then, according to (9.136),
and in place of (9.135) we get
dt 2
L_
-~ c 2
d *f
I)* 2
'
,
(
For this equation we have D'Alembert's general solution [see for-

mula (4.18)]
f = v(x-ct) + *t(x + ct)> (9.138)
where q> and \|?
are symbols of arbitrary functions.
67. SOME PARTICULAR SOLUTIONS OF THE WAVE EQUATION

We cannot dwell here on the general problem of integrating the
wave equation (9.135) but we may point out one simple property
of its particular solutions which is quite analogous to the property
of particular solutions of Laplace's equations indicated at the
-314 MORE GENERAL METHODS OF SOLVING ELASTICITY PROBLEMS [Ch. IX
beginning of Sec. 58; if any particular solution of equation (9.135)

is known
where a, p, y are some parameters or coefficients, then the

partial derivative of / with respect to any of the arguments in-
dicated in the parentheses of (9.139) will also be a particular solu-
tion of equation (9.135). The validity of this conclusion follows
directly from the fact that if denotes any of the above arguments,
then
_ d2 ((
~'
dt 2 ) dt 2 \ ,, ,
(9.140)
Indeed, since / is a solution of equation (9.135), then
o-
Differentiating both members with respect to ,

we have
and on the basis of (9.140) we get
hence
is
"where r=
_
a solution of equation
I/"* -f- y
2 2 2
(9.135), as
Consider an extremely important particular solution of the form
+
/ = /o(r,
z represents the radius-vector of an arbi-
trary point with respect to the origin. It is obvious that solution
(9.141) corresponds to a symmetrical distribution of factor /
with respect to the origin, so that the quantity / has the same
we
').
set out to prove.
'
(9-141)
value at all points of a sphere of radius r with centre at the

origin; this is a case of spherical waves having a source at a point
taken as the origin. Such waves are considered here only as a
mathematical image [i.e., as a particular solution of equation
(9.135)]; however, they can really be produced as a result of the
action of a momentary or, in general, of a varying force applied
at the origin.
By fo is meant any of the quantities

a, v, w, (Djf, a>2 , 0.
coy,
Sec 67] SOME PARTICULAR SOLUTIONS OF WAVE EQUATION 315-
Let us check the possibility of solution (9.141) by substituting

2
it in (9.135), and to that end calculate Laplace's operator V /a
[cf. formulas (9.25) -(9.27)]:
f)fo
Ox
_ dfp
dr
Or
dx
_ df Q x
dr r
'
x
~~
**
,
f)/
~~ ~~ df, r*-3 *
dx 2 dr 2 r2 '
dr r2 dr 2 r2 dr r
Calculating similarly the second derivatives with respect to if/-
and z, we get
2
d /o
~~ _ 2
f x 2
.
df r 2
-x 2
dx 2 dr 2 r2 "^ dr r3
'
^ J
Ar22
dr
J
r^22
~ dr
I
I
'
<>f r*-y*
r3
~ dfdr r 2
-z 2
dz 2 dr 2 r2 r3
Adding up these equalities by members and taking into account

that
* 2 + y 2 -r-z 2 = /-
2
>
we obtain
/0 ~~
V 2f _^ /o 2
"^
.
2 d/ _
~ 1 ^
^r 2 r dr r dr 2 V^/o;-
Substituting this in (9.135), we have

^L__L_
d*
~"
r2
c)r
2
or
r
~w '
but it is evident that
r ~~
d< 2 a* 2
and the last equation takes the form of the simplest wave equa-
tion [cf. (9.137)]:
therefore, we immediately write its general solution of the type

of (9.138):
where cp and ^ are arbitrary functions. Hence we obtain the re-

quired particular solution of equation (9.135), i.e.,the equation
,316 MORE GENERAL METHODS OF SOLVING ELASTICITY PROBLEMS [Ch. IX
of a spherical wave:
(9-142)
The first term on the right-hand side corresponds to a wave

propagated from the centre (the origin), the second term to a wave
travelling in the opposite direction (the reflected wave). Solution
(9.142), like solution (9.138) for the case of uniform vibrations,
represents a moving wave.
Very important particular solutions of the "standing wave" type
may be obtained for the general equation (9.135) by the method of
separation of variables, which was applied to the simplest wave
equation (4.28) in Sec. 28. For that it is necessary to seek solutions
in the form of a product of two functions
f=F(x,y,z)-T(t). (9.143)
Introducing this in (9.135), we obtain

F(JC, y, z)T"(t) = c^F(x, y, z)T(t)
or, dividing both members by F T, we separate the variables
r/ (0 .2 W(*.y.*) ~~ >2 _ '
r<0 F(*,y,z)
Avhere W is an arbitrary constant. This leads to two differential
equations:
0, (9.144)
y, z)
= 0. (9.145)
The particular solutions of equation (9.144)

T l
sin X/; T2 = cos \t
give harmonic vibrations; therefore, solution (9.143) takes one of
the forms:
/ = /?(jc,
1 y, *)sinM,
f2 = F(x, y, z)cosX/,
where F(x, y, z) must satisfy equation (9.145).
The dilatational and equivoluminal waves discussed here take
place in the interior of an elastic body; in the neighbourhood of
its surface waves of a different type are possible (Rayleigh waves),
while near the surface of contact of two elastic bodies special
waves occur (Love waves) !
.
1
See, for instance, L. S. Leibenzon, Theory of Elasticity, 2nd ed., Moscow.
Costekhizdat, 1947.
X
Bending of a Plate
68. GENERAL
A plate is a body of prismatical or cylindrical form whose
height h is small compared with the dimensions of the faces
(Fig. 94).
mark off the middle plane of the plate which divides
Let us
its height in half. This plane plays the same role in the theory
of plates as the axis and the neutral plane in the theory of bend-
ing of rods. Let us specify the system of co-ordinates, placing the
Fig. 94
axes Ox and Oy in the middle plane; the axis Oz is directed down-

ward.
In Chapters VI and VII we discussed the problems related to
stresses and deformations of such a plate in the case when the
external load and the corresponding reactions act in the middle
plane (generalised plane stress). If the load does not exceed a cer-
tain limit, the middle plane does 'not deflect. We shall now turn
to another case where the load and the supporting reactions are
normal to the middle plane; in this case, the middle plane &l the
318 BENDING OF A PLATE [Ch X
plate deflects, and we deal with the bending of a flat plate. It

should be noted that, in general, we encounter here not only bend-
ing (as it is understood in the elementary theory of bending
of a beam), since in a great majority of cases bending is accompa-
nied by torsion of a plate.
Below we consider the approximate theory of plate bending,
which is valid for plates whose thickness is small compared with
their other dimensions but whose deflections are small compared
with the thickness. Such plates are called "moderately thick plates'"
or, in Galerkin's terminology, "thin slabs". In the study of stresses
and deformations of such plates one may employ the same methods
as in the approximate theory of bending of rods, but in this case
the theory of bending is considerably complicated both by the
above-indicated phenomenon of twisting and by the fact that one
has to relate the stress and strain distribution to a whole plane
(the middle plane) rather than to a single line (the axis of a
rod). This leads to the replacement of ordinary differential equa-
tions of bending by partial differential equations whose integration
is much more involved.
Henceforth, to simplify the theory, we shall make the following
assumptions:
1. In the analysis of strains we disregard (just as in the ele-
mentary theory of bencling of rods) the stresses Z ? which occur on
,
account of the mutual pressure of horizontal layers of the plate,

and, consequently, the strains e zz of the plate in the direction of
its thickness; because of this we take the generalised Hooke's law
(V) for a plate in the following form:
(10.1)
2. We introduce a hypothesis analogous to the hypothesis of
plane sections in a rod. If a plate is bent into a cylindrical sur-

face, the above hypothesis may be accepted in the form as it is
formulated for a rod: plane cross sections of a plate subjected to
bending remain plane and normal to the deflected middle plane.
If, however, the plate is not bent into a cylindrical surface, our
hypothesis will be formulated as: a linear element mn (Fig. 94)

within a plate perpendicular to the middle plane remains straight
and normal to this plane after its deflection during bending. This
assumption was first proposed by G. Kirchhoff; it is sometimes
called the hypothesis of linear elements.
3. Since the deflections of a plate are small, we assume that
the points of the middle plane displace only in the direction of
Sec. 69} BASIC EQUATIONS OF BENDING AND TORSION OF PLATE 319
the z axis, i.e., for them

# = 0; i> = 0; wf(x, y).
This obviously implies that deformations, i.e., extensions and

shears, are absent in the middle plane.
69. BASIC EQUATIONS OF BENDING AND TORSION OF A PLATE

The foregoing assumptions make it possible to derive the basic
equations of the approximate theory of the bending of moderately
thick plates. According to the third of these assumptions, the dis-
placements u and v vanish in the middle plane; they are different
Fig. 95
from zero at points outside this plane, but Kirchhoff's hypothesis

enables us to express them in terms of deflections w(x, y) of the
points of the middle plane. Indeed, Fig. 95 shows a section of the
plate by a plane parallel to Oxz. Let OK be the trace of the middle
plane after deflection. Let us take any point d at a distance cd = z
from the middle plane. Taking into account the second and third
points of the assumptions made above, we have from the drawing
u = 2 sin a
,
2 tan a = dw
2-^-,
where a is the angle of inclination of the tangent to the trace of
the middle plane; we take the minus sign in the equality because,
for instance, in the case represented in Fig. 95 z>(\ tan a>0,
u<0. Thus
= -*-" (10-2)
Making a section of the plate by a plane parallel to Oyz, we find

by the same reasoning
*" 00-3)
320 BENDING OF A PLATE [Ch. X
With relations (10.2) and (10.3) we can obtain the expressions of

the following three components of the strain tensor from Cauchy's
formulas (III):
xx
_ du _
dx 2 '
dx
'
~ dy
~ Z
dy
2 (10.4)
~ ~ d 2
w
*" '
dx dy dx dy
We now solve the equations of Hooke's law (10.1) for stresses
E ,
l Otfyy),
Yy =
j
E
4. a) 2(1
Substituting here the values of e xx e yy e xy from (10.4), we obtain, ,
the expressions of three of the components of the stress tensor in

terms of deflection w(x, y):
___ Ez / d2w Pw\

1 _a 2 I
dX 2 dy*)'
r,& Ez
dx* )
(10.5)
dxdy'
The remaining three components X z Yz Z
, , z can now be obtained
from the equations of equilibrium (I) which have not yet been
utilised; the first of them gives
dXz dX dX
Ez d
[ a2 Ox
where
In the same way we obtain from the second equation of (I)
dz
~ 1 s'
d
dy
2
w '
Sec. 69] BASIC EQUATIONS OF BENDING AND TORSION OF PLATE 321
It remains to integrate these two equations with respect to z,
taking into account that
and -
are independent of z\ we have
(10.6)
The loads and the supporting reactions of the plate are assumed
to be normal to its upper and lower planes and, therefore, on these
planes, i.e., for z = y there should be
We chose accordingly the value of the arbitrary constant in for-

mula (10.6):
and obtain the final expressions of the shearing stresses:
Turning, at last, to the third equation of (I), we find
dYz
dz dx dy
or, substituting here the expressions of the shearing; stresses from

(10.7)',
Hence by integrating with respect to z, we can obtain the expres-

sion of Z z which
,
characterises the mutual pressure of horizontal
layers of the plate; however, the stress Z 2 in itself is of no interest
because in absolute' value it does not exceed the intensity of the
external load on the plate; the latter, however, is negligibly small
in comparison with the stresses (10.5) set up by the bending of
the plate" and, therefore, Zz was disregarded in the formulas of
Hocikfc's law- (10.l)v Equation (10.8) will be used for another, more
important purpose; by integrating it throughout the thickness of
O1
the plate, i.e., from y to +-o* we eliminate z from it and

thus obtain a differential equation for determining the function
w(x, y) through which it was found possible to express the remain-
ing displacements [formulas (10.2) and (10.3)] and the stresses
[formulas (10.5) and (10.7)].
Taking into account that
4
T-
we obtain from (10.8)
It is assumed that the load q(x, y) is applied on the upper sur-

face of the plate (z = then for the chosen system of co-
-g-J;
ordinates (Fig. 94) we have
g (x, y) = Z., (- 4) = - Z, (- *-)
.
Substituting these values in (10.9), we obtain the required differ-

ential equation w ith respect to deflection w(x, y):
r
~.
(10.10)
Here [see formula (6.5)]
Eh*
a'
_E^~' 110.11)
This quantity is called the flexural rigidity of a plate, by analogy

with the rigidity of a rod or a bar of rectangular cross section
with dimensions b and h in the elementary theory of bending:
or for b=\ B=
In the expanded form equation (10.10) is written down as
It is sometimes called the equation of Sophie Germain after the

young French investigator who first obtained it in the final form
in 1815.
ANALYSIS OF RESULTS OBTAINED
'
Sec. 70] 323
In solving each particular problem, one has to add to equation

(10.10) the boundary conditions, i.e., the fixing conditions of a
plate; we shall discuss them later. Now we shall proceed to clarify
the physical meaning of the formulas obtained.
70. ANALYSIS OF THE RESULTS OBTAINED
Turning to formulas (10.5), we note that the second deriva-

tives
d2w o A
- and
give approximately the curvatures of the curves in which the

middle surface of a plate is cut by planes */ = const and Ar=const
passed through a given point. These curvatures characterise the
phenomenon of bending of a plate and show that the stresses
Xx and Yy are set up by bending; they vary in proportion to z r
i.e., according to a linear law, across the thickness of the plate.

The same law governs the variation of normal stresses at the cross
section of a rod in the elementary theory of bending and these
stresses determine the bending moment at the cross section of the
rod. When a plate is subjected to bending there occur two bending
moments at each point which act at the sections normal to the K
and y axes; to calculate them we isolate an element with dimen-
sions dx, dy, h from a plate (Fig. 96) by four vertical planes; the
elementary strips of width dz on the faces of the element are acted
upon by the forces
Xx dy dz\ Y y dx dz.
The bending moments acting on the faces of theelement will be
calculated as the moments of these forces about the y and x
axes:
+4
M dy=
l
f Xx zdydz = dy j
Xx zdz\ M^dx = dx f Y y dzz
_A A a
2 "2 f
~2
Dividing these equalities by dy and dx respectively, we obtain the

bending moments in the plate per unit length of the cross section:
,h h
+T +T
M= v
f Xx zdz\ M = f 2 YyZdz. (10.11)
h h
T
s*
21
If we substitute here the expressions of stresses (10.5) and carry

out integration, then, taking notation (10.11) into account, we
finally get
d*w (10.12)
M,=-D( dy*
In addition to the bending moments, the faces of the isolated ele-
ment are acted upon by shearing forces due to the presence of
Fig. 96 Fig. 97
the shearing stresses Z X =XZ and Z y =Y z which

,
are determined
by formulas (10.7); induced on the faces of the element are the
forces
Zx dydz\ Z v dxdz. <e
The required shearing forces will be calculated as the resultants of

these forces:
t. ti
^ f Zx dydz\ N 2 dx = f Z y dxdz.
It is necessary to substitute here the expressions of 7stresses (10.7)

and carry put integration; taking Jnto account (10.8 ) and (10.11),
we obtain the expressions of the shearing forces per unit length
of -the section
N l
= D^^w\ N = ~-D~V*w. 2 (10.13)
V
The bending moments (10.12) in conjunction with the shearing

forces (10.13) determine the phenomenon of. bending of a plate.
Sec. 70] ANALYSIS OF RESULTS OBTAINED 325
It remains to consider the effect of the shearing stresses Xy [the

third of formulas (10.5)] on the deformation of the plate.
We have seen above that the second derivatievs -5-?-
ox
and -^
oy
give the curvatures of the sections of the middle surface of the
plate; the mixed derivative . . can be written down in one of the
following forms:
d I dw \ d
dy \ Ox
where tan a* and tan a y are the angular coefficients of the tan-
gents to the aforementioned sections; thus the mixed derivative
is the rate of change of the slope of the tangent to the section when
the latter moves in the direction normal to its plane; in this sense
the mixed derivative characterises the "torsion" of the middle
surface of the plate. The shearing stresses Yx according to (10.5),
,
vary directly as z, i.e., linearly across the thickness of the plate;

the resultant of the system of elementary forces X y dx dz (Fig. 97)
vanishes and they will reduce to a couple with the moment l
= J Y x zdydz,
which is called the twisting moment. Substituting here the expres-

sion of the shearing stress Y x from (10.5) and carrying out inte-
gration, we obtain the twisting moment per unit length of the sec-
tion:
H^-Dd)^. (10.14)
The shearing stresses X y acting on the face normal to the y

axis will also give the twisting moment, // 2 but it is equal ,
to HI\ this follows both from the law of reciprocity of shearing

stresses Y X = X V and from the fact that formula (10.14) is not
altered by interchanging x and y.
1
The moment considered positive if, when viewing from the side of the
is
positive x axis, we
see the rotation in the clockwise direction; the mixed deriv-
ative, which represents the "torsion" of a surface, is positive when the tan-
gent rotates counterclockwise [see above the second form of formula (a)].
This is taken into account in the sign of the right-hand member of formula
(10.14).
22-1013
326
to
Summarising
BENDING OF A PLATE _
the arguments set forth in this section, we come
the conclusion that the deformation of the plate under the
action of the load q(x, y) applied to its upper plane is deter-
[Ch. X
mined by the differential equation (10. 10'). This deformation

results from: (1) bending produced by the bending moments MI
and M2 [formulas (10.12)] and the shearing forces and 2 N {
N
[formulas (10.13)]; (2) torsion produced by the twisting moment
//=//i [formula (10.14)]. Both of these phenomena are generally
inseparable in a plate. Indeed, let us revert to the considerations
of Sec. 38 in regard to the case represented in Fig. 47. If the plate
is replaced by a flooring composed of separate rods, each of them
will bend under the action of the load acting on it irrespective of
the neighbouring rods; let them now be tied together in a solid
slab (plate). If we load only one rod, whose cross section mnpq
is shown in Fig. 47, then, deflecting, it will carry along the
adjacent rods applying to their faces those shearing forces which

we have designated here by N\ and A/ 2 these forces will cause
;
rotation of the cross section, i.e., the twisting of the rod.
M =M =N =N =
l 2 l 2 0; H= const. (10.15)
Then formula (10.14) is reduced to a differential equation in the

function w(x, y)
-^ H
00-16)
It is easy to verify by substitution that its particular solution

satisfying conditions (10.15) can be taken in the form
LJ
This equation indicates that the middle surface of the plate under
pure torsion is a hyperbolic paraboloid (or the so-called oblique
plane). It has been ascertained above that the mixed derivative of
the deflection w represents the torsion of a surface which is equal
to the angle of twist per unit length designated in Chapter VII by
T. If we also take into account that
let- formula (3.12)],
then formula (10.17) will be written down as
H
Sec. 70J
We
//=
, ANALYSIS OF RESULTS OBTAINED
consider the case where the plate is twisted by two couples

//' per unit length of the section; if this length is denoted
by c, then
_ 327
where M t is the twisting moment applied to the whole section, and

we obtain
Suppose that a rectangular plate is twisted by torques applied to

its four edges. Formula (10.18) gives the unit angle of twist. If
we eliminate, for instance, the couple //', we shall have the case
of simple torsion by torques considered in Chapter VIII. It should
be expected that in this case the angle of twist will be one half
as large as that given by (10.18), and we obtain
T _ M t
'
ych*
This result is coincident with Saint-Venant's formula (8.59), if
we set a ==-T in it for the case of a very narrow rectangular sec-

tion (see Table at the end of Sec. 53).
The case of pure bending of a plate occurs if we set
M, = N = //=
2 0; Af ! = M = M *= const.
2 (10. 19)
In this case formulas (10.12) give two differential equations:
a ^ ~~ ~D
~Jx* ~+~ ~Jy 2
~
'
d w
^ ddx*w ~ MD
2 2
,
- -
(3 '
2
dy
by solving them for the derivatives, we obtain

M ,inon\
-
00-20)
Their solution, satisfying conditions (10.19), is
It follows from this that the middle plane of the plate under pure
bending is converted into a paraboloid of revolution. If the approx-
imate values of the curvatures were replaced by the exact ones in
22*
equations (10.20),
d2w d2w 1
we would find directly from these equations that the middle plane
is bent into a spherical surface of radius
~~
M
The pure bending of a plate occurs only in exceptional cases, for
instance, in a circular plate loaded along its contour by uniformly
distributed bending moments of intensity M
per unit length of the
contour.
71. BOUNDARY CONDITIONS FOR A PLATE

As pointed out earlier, the boundary conditions are the con-
ditions on the surface of plate which fnust be prescribed in
a
advance in order to obtain the solution of equation (10.10') cor-
responding to the specific problem under consideration. Such con-
ditions include the load q(x, y) on the upper and lower planes
of the plate but it is taken into account in the formulation of
the general problem of bending of plates and it enters in the
free term of equation (10.10'). It remains to clarify the conditions
on the lateral surface, i.e., at the edges of the plate, depend-
ing on the fastening or supporting conditions. For simplicity's
sake, let us begin with the case of a rectangular plate (Fig. 98)
whose edges are parallel to the axes Ox and Oy. Ignoring the
fixing conditions indicated in the drawing, we note that at the
left-hand or the right-hand edge (* = or x = a), for instance, there
act at each point
bending moment M^ \
twisting moment H, I
(10.22)
shearing force N 19 j
which we can specify in advance and, moreover, we can impose

the condition on the deflection w
requiring that it be equal to zero
or to a given value. The question of
boundary conditions in the
problem of bending of a plate proved to be exceedingly difficult
and it was treated in detail by a number of the greatest scientists
of the 19th century (S. D. Poisson, G. Kirchhoff, Kelvin and
Tait,
H. Lamb and others). We can elucidate it here
only briefly, pro-
ceeding from the following considerations.
Let the load q(x, y) be absent; in this case, the
bending of
the plate can nevertheless take place under the action of the forces
_
Sec. 71] BOUNDARY CONDITIONS FOR PLATE
applied at the edges (for instance, the forces M, //, N indicated

above) or as a consequence of the deflections w prescribed here.
In this case equation (10.10') becomes homogeneous
dx *
~
r o^
|
_ 329
and coincides with equation (IX) of Chapter VI for the stress

function in the plane problem of the theory of elasticity. By using
it in the solution of the problem for a region bounded by a given
contour, we obtain a unique, quite definite solution by specifying

two conditions on this contour: in the first problem, the values of
the normal and shearing stresses are assigned on the contour; in
Simply supported edge
Fig. 98 Fig. 99
the second, the displacements u and v of the points of the contour

are prescribed; in the mixed problem, a combination of one of the
displacements and one of the stresses can be prescribed (if, for
instance, a strip rests on rigid foundation, the displacement norm-
al to their surface of contact will be zero, and so will the shearing
stress in the absence of friction). The kindred character of equa-
tions (10.100 and (IX) indicates that in the problem of bending
of a plate one may not prescribe more than two conditions on its
contour; any third condition may bring about a contradiction in
the statement of the problem, which is somewhat restricted by the
adopted hypothesis of linear elements. By confining ourselves to
two conditions, we have to meet the most important requirements
imposed by the manner in which the plate is fixed or supported.
Consider the fundamental cases encountered in design practice
(Fig. 98).
1. The left-hand edge is clamped. Accordingly, we impose the
conditions: for jc = 0, w= 9
-^
= 0, requiring that the deflection
and the angle of inclination of the tangent to the middle plane
at the support be absent.
2. The back edge is simply supported. Consequently, the deflec-
tion and the bending moment M
2 [formulas (10,12)] must be zero
at the support; thus
for y = 0, w= 2 (10.23)
dy
In these two cases no conditions have been imposed on shearing
forces and twisting moments; but, being applied to the edge
resting on the support, they will have some effect only on the mag-
nitude of the supporting reactions of the plate and on the stress
distribution near this edge.
3. The right-hand edgeis free. This case presents the greatest
difficulties. No
restrictions are imposed here on the deflection;
however, all three forces MI, //, NI must be equal to zero and there
\0\
Fig. 100
is no reasonto disregard any of them by treating them as second-

ary. Kirchhoff suggested the following way out of this difficulty.
Let us suppose that the face x a is acted upon by distributed
torsional couples. Their distribution is assumed to be nonuniform,
in general, the magnitude of the moment of these couples per
unit length of the face at a given point being denoted by HI The
magnitude of the moment per length dy of the face is obviously
equal to H^dy.
It will be noted that from the viewpoint of statics the distributed
torsional couples are equivalent to a certain shearing force. Indeed,
the couple with the moment Hidy may be applied (Fig. 100) by
means of two equal and opposite forces HI acting at the edges of
an area of length dy.
The couple (
//i + -^-L dy\ dy on the adjacent area of length dy
may also be applied as two opposite forces H { -j-

-j-f- dy with the
arm dy. Having done this for all areas of the face concerned we
see that the forces applied at points and n on the boundary m
of the two areas reduce to a single force -jr-dy per length dy.
Sec. 72] ELLIPTIC PLATE CLAMPED AT EDGE 331
We observe, however, that there will remain two nonvanishing

finite concentrated forces H\ and H\ at theedges A and B oi
the face.
Hence we conclude that the distributed torsional couples oi
intensity HI are statically equivalent to the distributed shearing
force which has the intensity
4^-
dy
(10.24
v
and two concentrated shearing forces H\ and //fat the ends, i.e.
at the corners of the plate; the contour is smooth and withoui
if
corners, these concentrated forces are absent. Proceeding fron

this, Kirchhoff proposed that the three boundary conditions at th(
free edge be combined into two by equating to zero the bending
moment MI and the shearing force A/\ and by adding to the lattei
the term (10.24), which reflects the influence of the twisting mo
ment //. Now we arrive at the following two conditions at the
free edge:
d*w ,
d*w r,
dx*
or (10.25
d3w
This form of the boundary conditions for a free edge is conven-

tional.
There have now appeared some publications whose object i<
to eliminate the restrictions on the foregoing approximate theorj

of bending of plates imposed by Kirchhoff's hypothesis of lineal
elements; this hypothesis, as we see, makes it difficult to satisfy
all the necessary boundary conditions. On the other hand, it is
somewhat contradictory, since, taking into account the shearing

stresses Z X = X Z and Z y = Yz this theory eliminates the possibility
,
of the pertinent shearing strains e xz and e yz their consideratior \
permits the refinement of the theory and of the respective boundary

conditions Consider some special cases of bending of plates.
l
.
72. ELLIPTIC PLATE CLAMPED AT THE EDGE

Let us take the expression of deflection in the form
;
(io.26;
E. Reissner, J. Math, and Phys., 1944, Vol. 23, pp. 184-191; L. Bolle
1
Bull, techn. Suisse Romande, 1947, Nos. 11 et 25 oct.; B. F. Vlasov, Vestn

Mosk. Univ., 1957, No. 2, pp. 25-34; Izv. AN
SSSR, Otd. tekhn. nauk, 1958
No 12, pp. 124-127; Dokl. AN Azerb. SSR, 1957, Vol. 13, No. 9, pp. 955-959,
23*
designating for the sake of brevity

_L2 ~r 2112 i
i
A (
(10.27)
a
we have
(10.28)
We observe that the deflection is zero on the elliptic contour:
^= -
+ {1-
1=0. (10.29)
Consequently, we shall obtain the solution of the problem of bend-

ing of an elliptic plate if function (10.26) satisfies the basic equa-
tion (10.10').
Taking the first derivatives of deflection (10.28)
dw c> T r dU dw o / /
^^
' '
dx dx d oy
we see that they vanish everywhere on the contour (10.29); hence

we readily conclude that the middle plane remains horizontal on
the contour; consequently, we have a plate clamped at the edge.
Equation (10.10') involves the fourth derivatives of deflection; one
sees from expression (10.26) that they will all be constant num-
bers; it is evident from this that equation (10.10') can be satisfied
if <7 = const, i.e., if the load is continuous and uniform.
The calculation of the successive derivatives gives
dw A t
dx*
= 8-rX
O C
+ 4-5-
9
2 I A C
dy*
= 84- i-^r*; (10.30)
dy*
c
dx* dx 1 dy* dy* IF
Substituting the values of the derivatives in equation (10.10'), we
find the constant c, which is not yet determined:
=:=
q
c
16
Introducing this value in (10.26), we get the equation of the

deflected middle plane.
The reader is recommended to do the following exercises:
Sec. 73] RECTANGULAR PLATE. NAVIER'S SOLUTION 333
1. Find the maximum deflection of the plate at the centre.

Making use of formulas (10.12), (10.13), (10.14) and expres-
2.
sions of derivatives (10.30), find the general expressions of Af 4 ,
M ,Hi=H2,Ni
2 and N 2.
Calculate the maximum bending moments Mi and
3. 2 at the M
ends of the axes of the contour (moments of clamping) and at the
centre of the plate,
4. Setting b = a (circular plate), find the quantities indicated in
the problems above.
73. RECTANGULAR PLATE. NAVIER'S SOLUTION

Consider a plate simply supported at the edge and subjected to
an arbitrary load. In this case, the solution of the basic equation
(10. 10'), i.e., the expression of the deflection w, has to be sought
in the form of an infinite trigonometric series. Let us take it in
the following form which, as may easily be seen, satisfies the
supporting conditions on the four sides of the contour
l
:
r sm
.
** sin
m-\ n=l
If we denote for the sake of brevity
A mn sln^- S \ n
^- = Umn , (10.32)
then series (10.31) will be represented as
m=l /i=l
(10.33)
each line of this expression is a series with respect to the sub-

script ft, while each column is a series with respect to the sub-
script m.
1
Indeed, on the contour, i.e., at *~0, x=a, #=0, yb we have w = Q,
__..-0, -3-5-
= 0, and the conditions for the simply supported edge will be
satisfied.
Differentiating expression (10.32) and substituting its deriva-

tives for w in the left-hand member of equation (10.10'), we get
sin sin
therefore, equation (10.10') will be written as

OO 00
2 . .
nny , v /-m o^\
m-^-^x, y), (10.34)
or more compactly:
= q(x, ,
y),
.
/1A
(I0.34a)
, v
n= 1 /i = 1
where
(10.35)
It remains to choose the coefficients of the series so as to satisfy
equation (10.34a) identically over the whole area of the plate; in

other words, according to (10.34a), our problem consists in ex-
panding the function q(x, y) in a trigonometric series in two vari-
ables x and */; this is carried out by the usual Euler's procedure
for calculating the coefficients of a trigonometric series; since we
have a double series, the procedure should be applied twice. First
we multiply both members of equality (10.34a) by
where k is an integer, and integrate between the limits and b:

uo oo b b
\n V~^ ^ . nmx r .
mty .
kny , r / v .
kny ,
/n=l n=l
(10.36)
The composition of this double sum is, of course, the same as in

expression (10.33), but it is known that if n --
k then
y
sin -
sin
/sir
if, however, AZ = &, then
(10.37)
Sec. 731 RECTANGULAR PLATE. NAVIER'S SOLUTION 325
Hence easy to conceive that in the sum of the left-hand

it is
member all terms vanish except those corresponding
of (10.36)
to the kth column of (10.33) in which m has all values 1, 2, 3,
. oo, and on the basis of (10.37) equality (10.36) becomes
. .
,
oo b
sl " *" y
j S Cmk
m =l
sin
^T~ =/?(* y) d y- (io.38)
Both sides of this equality are functions of x only. We now apply

the above procedure to equality (10.38): we multiply both sides
by
sin
^- dx
and integrate from to a; all terms except the one having the
subscript
= i vanish in the left-hand member; by analogy
m with
(10.37) we have
. o
/.sm
s
then from equality (10.38) we obtain

a b
b a n =
C C C fa*
sin
-2"2 '*
J J
Hence we find the coefficient C ik ,
r* 4
and further, on the basis of (10.35), we write
( fq(x,
This is the general expression of any coefficient of series (10.31);
replacing the symbols i and k again by and n, we obtain m
2 +-\\'
b 2
J
b
f f q(X '
y} Sin ^ Sin
^ dX dy -
(1 '
39)
Given the distribution of load q(x, y), it is possible to evaluate

the integrals involved here and, substituting the values of coef-
ficients Amn in equation (10.31), to find the equation of the de-
flected middle plane. Consider two fundamental cases which are
essential in design practice.
1. Let the load q be continuous and uniform over the whole

area of the plate. Then, for calculating the coefficient A mn given
by equation (10.39), we have
(10.40)
Consequently, the double integral has fallen into two simple

integrals; we evaluate them:
For m and n odd these integrals are equal to -^ and respec- ,
tively; for m and n even they vanish. On this basis, integral (10.40)
is different from zero only for m and n odd and then it is equal to
4qab
Substituting this value in (10.39), we get
where m= 1, 3, 5, . . . , oo; /z== 1, 3, 5, . . . ,

oo.
We observe that with increasing m and n the magnitude of the
coefficient A mn drops off sharply and series (10.31) converges
rapidly; we obtain ultimately
,.,*-., , tiny
sin sin ,
where m=
l, 3, 5, . . , oo; n=l, 3, 5, .
,
oo. . .
In order to find the maximum deflection at the centre of the plate

we set #=0.5a; j/0.56; then
16
/f- m, ^
+
Sec. 73] RECTANGULAR PLATE. NAVIER'S SOLUTION 337
Here D should be replaced by its expression from (10.11). For

calculation it is convenient to set~ = p, and to represent the last
formula as
The coefficient a is expressed by the double series [cf. formula

(10.42)] and depends on the ratio of the sides \JL (see Table).
With the expression of the deflection (10.41) we can find, by

equations (10.12), (10.13) and (10.14), the forces at any point of
the plate, for instance:
1
Inasmuch as the edges are simply supported, the maximum

moment occurs at the centre of the plate for jc = 0.5a, y = 0.5b:
The shearing forces and supporting reactions of the plate on the

contour are calculated in the same manner. The results are pre-
sented in the table compiled by Academician B. G. Galerkin.
The quantities
max
represent the maximum unit supporting reactions at the middle

of the sides of the contour. R = mabq are the concentrated reac-
tions at the corners of the plate; in calculation a is taken to
be 0.3.
2. A plate is bent by a concentrated force P applied at any
=
point x = c, y d. We replace the force P by a continuous load
q distributed over an infinitesimal area dx dy:
or q=. (10.43)
Then, the function q(x, y) in formula (10.39) is zero everywhere

= =
except at the point x c, y d, where it is equal to (10.43). In
the integral of the right-hand member of equation (10.39) there
remains only one nonvanishing element which corresponds to the
point (c, d):
/
x . mnx
q(x, y)sm a o
P . mnc . rind , , ^ . mnc . rind

^n sin-prf^^Psin sin-^--
then, according to equation (10.39),
4/>sin^^-sin
jT\mn
m2
Sec 74] RECTANGULAR PLATE. LEVY'S SOLUTION 339
From (10.31) we obtain the equation of the deflected middle plane
mite ,
rind
4P
n2 \
2 (10.44)
2
b )
74. RECTANGULAR PLATE. LEVY'S SOLUTION

The method given by M. Levy is more general than Navier's
method; at the same time it is closely related to the solutions of
L. N. G. Filon and C. Ribiere of the plane problem for a rectangle
Fig. 101
presented in Sec. 44, which is due to the above-mentioned close

relationship between the basic equations:
for the deflection of a plate
V2V2 ^ _ ? (* y)
(10.45)
and for the stress function in the plane problem

'
2
V V 2
cp
= 0.
Levy's method is valid for all cases where (Fig. 101) two op-
posite edges of a rectangular plate, for instance and AC, are OB
simply supported, while the other two, OA and BC, may be sup-
ported in any manner or be free. Bearing in mind these condi-
tions, we seek the solution of equation (10.45) in the form
(10.46)
m=l
340
Since
we have
_
d2w
gpr
=
BENDING OF A PLATE
rc
2
F i "* /m sm
= (M d z*_ n
=a\
V^
W~ a
o
2 ^
2
/
(y)
\
_ mnx
,
[Ch. X
i.e., the conditions the simply supported edges OB and

at AC
are Substituting (10.46) in the left-hand member
satisfied. of
equation (10.45), we obtain
(10.47)
or more compactly
mnx =q
V D*m (y) sin
,
'
,
(10.48)
ra=l
where
2
D^==/UV)(y)_ 2 (
J
^) /^(y) + (-~L)Vm (y)- (10.49)
The left-hand member represents a Fourier series in terms of sines;

therefore,we must think of the right-hand member as being ex-
panded in the same series
where F m (y) are the coefficients in the series which are depend-
ent on y since the prescribed load q (x, y) is generally dependent
on it; substituting this in the right-hand member of (10.48), we
equate the coefficients of the respective terms on the left-hand
and right-hand sides:
and according to (10.49), leads to an ordinary differential

this,
equation of the fourth order for finding the function fm(y), which
is as yet unknown.
In practice, this process is reduced to the following formal
manipulation: we multiply both members of (10.47) by sin
-^^
(where Ai=l, 2, . .
.) and integrate from zero to a; taking into
Sec. 74] RECTANGULAR PLATE. LEVY'S SOLUTION 341
account that
J sin - -
mnx .
sin
nnx
dx
,
is equal to zero for m^n and to

y for m = n, we obtain the
required differential equation

a
A
j-n\r\
IV) / \
(y)-2(
n / H \
)
2
fn(y) +
* / \ i /
(
nn \
)
4
*
fn(y)
=
/ \
D-a
2
>
(10.50)
Upon integration in the right-hand member of this equation,
it will obviously become a function of y alone.
The general solution (10.50) will be composed of
of equation
the general solution of the homogeneous equation which coincides
with equation (6.82) of Sec. 44, if we set in the latter
and of any particular solution of the complete equation (10.50);

we denote this particular solution by
the general solution of the homogeneous equation can be writ-

ten at once, on the basis of the arguments of Sec. 44, in the
form of formula (6.88) by replacing m by n and / by a:
(10.51)
Therefore, the general solution of equation (10.50) is
(10.52)
Introducing this value of f n (y) in (10.46), we find the solution

of equation (10.45), which corresponds to the prescribed load
q(x, y) and satisfies the boundary conditions at the simply sup-
ported edges OB and AC. It remains to adjust the arbitrary
constants
A , C rt ,
Dn (10.53)
so that every individual term of the sum (10.46) will satisfy

the boundary conditions prescribed at the other two edges OA
and BC\ the number of these conditions is always four, and

determine constants (10.53). Consider, for instance,
they suffice to
the case where the edges OA and BC are also simply supported;
then the boundary conditions are
Jt
dy
s-
= 0. (10.54)
or, according to (10.46), we have
S*/(0)sm-^- =
//-v\ mix ^
0;
/IA nnx r\
or on account of the arbitrariness of A; we get
(10.55)
Writing out these conditions by the use of (10.52), we obtain

four equations:
( 10 .56)
After substituting in the right-hand members of these equations

the appropriate values of the particular solution
7(y). (10.57)
we must solve them with respect to constants (10.53). The solu-

tion is greatly simplified if, taking advantage of the free choice
of the particular solution (10.57), we construct it so that
f tt (Q)
= Ql /*(0) = 0. (10.58)
Sec. 74] RECTANGULAR PLATE. LEVY'S SOLUTION 343
Then the first two equations of (10.56) give
A n = D n = Q. (10.59)
Solving the other two equations, we get
{t
<
CO sh
fflib
/-
~ftt / , \
(10.60)
2I::lsinh 2
a a
The particular solution (10.57) which satisfies conditions (10.58)

will be derived by Cauchy's method. For that, we first find from
(10.51), by a suitable choice of constants (10.53), the particular
solution of the homogeneous equation
which satisfies the initial conditions
/(0) = /;(0) = /;(0) = 0; /;?(0)

= 1; (10.61)
we denote it by Y n (y)', it is given by
n 2 (rw) (10.62)
v '
By differentiating, we find that this solution satisfies conditions

(10.61). If, for brevity, we introduce in the right-hand member of
(10.50) the notation
-q (x, y) sin dx = <? (y), (10.63)
then, following Cauchy's method, we find the required particular

solution of equation (10.50) in the following form:
(10.64)
e
or in the expanded form
a . , tin (y 1\) r , . . mix , , /1A aA .
mi
smh a
\
I a (x, 7])sm d
dx^dri. (I0.64a)
x
JJ I
1
Substituting the values (10.59), (10.60) and (10.64) in (10.52),

we finally find the function f n (y), and, introducing it in (10.46),
we obtain the solution for a rectangular plate simply supported
along its contour and subjected to any prescribed load q( x, y).
The forces will be determined in the usual way according to for-
mulas (10.12), (10.13) and (10.14).
This solution is more convenient
for application than Navier's
solution given in Sec. 73, because of the good convergence of
the correspoding series. In a similar manner one can derive
solutions for other boundary conditions at the edges OA and
BC\ it is only necessary to modify equations (10.56) accordingly
for finding constants (10.53); the further procedure of solution
holds.
Many problems concerned with the bending of plates were
solved by Academician B. G. Galerkin. All these results are
collected in his fundamental work Thin Elastic Plates, Lenin-
grad-Moscow, Gosstrojizdat, 1933.
75. CIRCULAR PLATE

the investigation of a circular plate it is most convenient
In
to use the polar co-ordinates r and 0; therefore, all the basic
equations of the bending of plates will be transformed to the
polar co-ordinates according to the formulas
Inasmuch as the left-hand member of the basic equation (10.10')

is completely analogous structure to the left-hand member of
in
equation (IX) of Sec. 39 for the plane problem, we can use the
computations of Sec. 47 for the transformation in question by
replacing the function cp by w. Thus we immediately obtain
y2 W_
~~
d2w
dx*
"" d w
2
.
_
~ d*w
dr 2
.
'7"
1 dw 1
. d2w
dr ~T~"7*' ~dP"
'
dy*
and the basic equation (10.10) becomes [cf. (IXp) in Sec. 47]
2 , 1 d 1 d2 \ / d2w 1 dw , 1 d*w
(10.65)
Here we shall deal only 'with the case where the load is in-
dependent of 6, i.e., when it is distributed similarly in all direc-
tions from the centre of the plate. In this case the deflected middle
surface must obviously be a surface of revolution, the deflection
w does not depend on and equation (10.65) is simplified [cf.
Sec
al
75]
__
equation (IX'p) in Sec.
The general solution

solution w of the
d*w 2
47]:
2 d*w
CIRCULAR PLATE
I d*w .
of this equation is the sum of the gener-

equation without the last term
d3w 1 d2w 1
I dw\
do;
nn
345
< ia66
fifi
.
>
and any particular solution of the complete equation (10.66),

however, we know
the general solution of equation (10.67)
already
[equation (7.5) of Sec. 48]
Cj In r w= +
C 2 r 2 In r C 3 C 4 r 2 (10.68) + + ;
therefore, the general solution of (10.66) is written as

w=C l
In r + C 2r2 In r + C3 + C 4 r 2
-\-w. (10.69)
The particular solution w

can readily be found if the load q
is (continuous uniform load over the whole area of the
constant
plate); the form of equation (10.66) itself reveals that one may
assume w =cr 4 where c is a number to be determined; substi-
,
tuting this value in equation (10.66), we easily find that c

=
= q/64D and, therefore,
w9TL
~ 64D '
Substituting this in equation (10.69), we obtain the general

solution for a circular plate under a continuous uniform load:
(10.70)
To calculate momentslet us use equations (10.12) and (10.14),

when transformed the polar co-ordinates. If the axis Ox is
to
directed along the radius-vector r of a given point, we have
d 2 w~ d w dw
dx 2 ~~~~dr 2 t
2
"5y2"~7~5T
1
dx dy
~ *
Introducing this in equations (10.12) and (10.14), we find
(10.71)
Here M { is the bending moment over the area normal to the

radius, M 2 is the moment over the radial' section; the twisting
24-1013
346 _ _
moments over
its centre, we have
solution (10.70);
BENDING OF A PLATE
the basic sections under consideration are absent,

as expected, because of the symmetry of the plate and loading.
Consider several particular modes of fixing a plate on its con-
tour. If we take a solid plate without a circular cutout around
to drop the first two terms in the general
otherwise, w, -^- and -^ become

[Ch.
infinite
X
at
the centre of the plate, i.e., the deflection and curvature of the
plate will be equal to infinity. Hence we obtain
(10.72)
The remaining two arbitrary constants can be determined from

the fixing conditions.
(a) If the plate of radius a is clamped along its contour, we
obtain the following conditions on the contour: for r = a w = Q,
=
dw/dr Q. Determining C3 and C* from these conditions and sub-
stituting their values in equation (10.72), we get
Further, according to equations (10.71), we find
M = -~- [a
}
i
16 l
2
(1
v + a)' ;
(10.74)
The reader is recommended to verify these formulas and to

find max Mon the contour of the plate and at its centre, as well
as to find the deflection at the centre. All these results are to be
compared with the results of Sec. 72 in the case b a.
(b) If the plate is simply supported along its contour, the con-
ditions on the contour for r=a are
determining C3 and C4 from these conditions, we obtain from

equation (10.72)
(10.7S)
Further, we find, according to equations (10.71),
(10.76)
Sec. 76] MEMBRANE ANALOGY. MARCUS'S METHOD 347
The reader is recommended to calculate max and the deflectionM

at the centre of the plate.
(c) Consider also the case of a plate with a simply supported
=
edge in the absence of load (<7 0) but subjected to a uniformly
distributed moment Mi =M
along the contour. In this case, from
equation (10.72), we obtain
The constants have to be determined from the conditions for
We readily find that
This solution appears to be necessary in the design of the end

plates of circular cylindrical reservoirs. Here we encounter a
case of pure bending of a plate.
In conclusion we present the results which are obtained in the
case of a plate carrying a centrally placed concentrated force P.
If the plate is clamped along its contour, we have
If the plate is simply supported, we have
Calculation of moments according to equations (10.71) offers

no difficulty.
Combining these solutions with those given above and equating
to zero the deflection at the centre, one can solve the problem
of a plate supported along the contour and at the centre and
subjected to a continuous uniform load.
76. MEMBRANE ANALOGY. MARCUS'S METHOD

Let us indicate a new form assumed by the basic equations
of a plate in introducing the Laplacian operator on the deflection
d2w d2w .
and the so-called reduced sum of the moments [see formulas

(10.12)]
a = jH|+^k = _)v^. (10.80)
24*
This leads to the differential equation
DV 2w = 3K. (10.81)
The expressions of shearing forces (10.13) can be written down
^r
dx \dx* ~dy*')
d d*w d*w (10.82)
i \
--I-
J
These relationships are analogous to the known relationship for

a straight rod
Next, equation (10.10) on the basis of (10.81) is rewritten as
-r =? or V2aw== <? 00-83)
For a straight rod there is a similar relation

d*M _____
dx 2
~ 4'
We shall now consider, parallel with the plate, a flexible inex-

tensible membrane (film) clamped on the same contour as the
plate in question and subjected to the same load q(x y)\ the t
differential equation of the surface of the membrane was obtained

in Sec. 54 [see equation (8.71)]:
d*z .d*z q
or more compactly
V 2 * =--:. (10.84a)
For a funicular curve we have a similar equation:

d*z ________ g
dx*~ H '
where H is the pole distance.

Let us now compare equations (10.81), (10.83) and (1084a):
= -|-, (10.83)
= jjr. (10.848)
Sec. 76] MEMBRANE ANALOGY. MARCUS'S METHOD
Hence we draw the following conclusions similar to those

based on the analogy between the elastic line and the funicular
curve.
1. If we set //=!, then z = 2R; consequently, when the mem-
brane is subjected to tension equal to unity, its sagged surface
under the load q gives the diagram of the reduced sum of the
moments 2Ji due to this load. (Compare with the diagram as M
a funicular curve or funicular polygon for the given load.)
2. If we set // = D, <7=3K, then z = o>; consequently, when the
membrane is subjected to tension equal to the flexural rigidity
D, the sagged surface of the membrane under the load in the
form of the 2)1 diagram gives the deflected surface of the plate.
(Compare with the elastic line as a funicular curve due to the
loading by the diagram.) M
By using this analogy between plate and membrane, H. Marcus
devised his method of analysis of a plate; he replaces a mem-
brane by a network (cf. the approximate replacement of a funi-
cular curve by a funicular polygon) and thereby transforms its
differential equation (10.84a) into a finite-difference equation; pro-
ceeding in this way, he replaces the integration of the differential
equation (10.84a) by the solution of a system of equations of the
l
first degree .
1
See H. Marcus, Die Theorie elastischer Gewebe und ihre Anwendung auf
die Berechnung biegsamer Flatten, Berlin, Springer, 1924.
XI
Variational Methods
of the Theory of Elasticity
77. VARIATIONAL PRINCIPLES OF THE THEORY OF ELASTICITY.

FUNDAMENTAL INTEGRAL IDENTITY
By variational principles of the theory of elasticity are meant

certain basic theorems expressed in the form of integral equal-
ities connecting stresses, strains and displacements throughout
the volume of a body and based on the properties of the work
of elastic forces. The variational principles are of practical in-
terest in the sense that they form the basis of the methods which
permit the effective solution of problems in many cases where the
classical approach to the integration of the basic equations of the
theory of elasticity presents difficulties which have as yet not
been surmounted. In this section we shall fix our attention on
an integral transformation which enables us to simplify further
computations.
Consider a certain state of equilibrium or of small elastic vi-
brations of a body, which is characterised by the stresses
V V 7 V 7 V
jf yi ^*2 2* .* y
and the displacements

u, v, w
.and which satisfies the basic equations
~5T+~^ ]r
~dT + X = ?~WV
dYx dY v dYz d*v
-^+i7+-aF+ rP = Pir- Oi-i)

dZx dZ dZ
~5jM 57" "~dT + ^P =
z
+ y
with external forces acting on the surface of the body:

Xv = X x cos (vx) + X y
cos (vy) +X z cos (vz),
Y v = Y x <;os(vx) + r y cos(vy) + Y z cos(vz), \ (11,2)
Zv Zx cos (vx) -f- Zy cos (vy) -f Z2 cos (vz).
Sec 77] VARIATIONAL PRINCIPLES OF THEORY OF ELASTICITY 351
At the same time consider three functions

it' (x, y, z, t)\ v' (x, y, 2, t)\ w' (x, y, z, t), (11.3)
which are continuous throughout the volume of the body and

have continuous partial derivatives of the first and second order
with respect to x, y, z and t in the same region. Let us regard
them as displacements produced in the body under the action of
some forces which are of no interest as yet. Because of dis-
placements (11.3) there are induced strains
du' , dw' , dv'
e xx~ dx '
"y*~ dy
"T"
dz
,
^yy~
dv'
dy
'
"**
,
:
===
-^r
dz
du'
+ dw'
dx
.
r
-
'
(11.4)
, dw' . , dv' . du'

' e '
zz dz *y dx '
dy
Let us form the following integral extended over the surface

of the body
(S)
Substituting here the values of Xv Yv Z

, , v from (11.2) and col-
lecting the coefficients of cosines, we get
/= [Pcos(tFj:) + Qcos(t>y)H-/?cos(TC)]d5, m.6>

(S)
where
Next, as in Sec. 37, we transform this surface integral into a

volume integral
where
'
dx
dX y dYy dZ y ' '
dw
_
'
dQ du dv
f ,
y Y 7
dw'
(11.8)
352 VARIATIONAL METHODS OF THEORY OF ELASTICITY [Ch XI
For substitution in (11.7), we add up these equalities by mem-

bers, making use of equations (11.1) in adding up the first three
columns on the right-hand side, and of equalities (11.4) in adding
up the remaining columns. Hence we find.
'
+X e'
x xx +Y y
e'
yy +Z z
e'
zz +Y
(11.9)
We introduce the value (11.9) in (11.7) and replace / in the
left-hand member of the latter by its expression (11.5); we then
obtain
f 9 [(x
4S) (t)
(1UO)
This results from formal computations; nevertheless,
equality
attaching one or another meaning, on the one hand, to the state
of stress
X Y r Zz Y , Zx X y
,
(11. 11 a)
and to the respective displacements

u, v, w (11. lib)
and, on the other hand, to the displacements

a', v', w' (11.12)
we obtain all the general theorems of the theory of elasticity.

As one ofthe examples, we assume that the state of stress
(11. 11 a) corresponds to the equilibrium of a body (i.e.,
,p_^.
r=:
p^^.
=ip ~- = Oj and that the actual displacements
a, 0, w are taken as a', v', w'\ then equality (11.10) is transformed
into (5.62) and yields Clapeyron's theorem.
78. LAGRANGE'S VARIATIONAL EQUATION

Let the state of stress (11. 11 a) and displacements (11. lib) cor-
respond to certain small elastic vibrations of a body (and, in
7
particular, to the equilibrium state of a body). As a a' and w' ,
Sec. 78]
we
do
_
displacements, and we denote
LAGRANGE'S VARIATIONAL EQUATION
take arbitrary variations of the displacements u, v, w which

not violate the geometrical constraints
u!
imposed on these
= 8# ; i)
9
= 8-0; w f
l
Iw.
_ 35$
Equation (11.10) assumes the following form:
(S) CO
(11.13)
\-
In the right-hand member it is taken into account that

d
dv' . du'
i.e., 6e 6^, represent the variations of the strain com-

i1coc , . . .
ponents corresponding to the state of stress (11. 11 a). The inte-

grand in the right-hand member of (11.13) will be transformed
by using formula (3.25):
The examples of geometrical constraints are:

1
1. a body rests on immovable supports at some

If
points, the following
conditions must be imposed on the variations:
Ba =s tv = %w = 0.
2. If an element dx at a given point is restrained from rotating in the

planes Oxy and Oxz, then [cf. formulas (2.3), (2.4), (2.8)]
and hence the conditions follow which are imposed on the variations at this
point:
354 VARIATIONAL METHODS OF THEORY OF ELASTICITY [Ch. XI
Here 8W is the variation of the elastic energy W

resulting
from the variation of the displacements 6u, f*v, 6w. Then
(T) (t) (t)
represents the variation of the elastic energy in the whole body.

Equality (11.13) yields the variational equation
(S) (t)
= *fV*, (11-14)
(-0
which formulates D'Alembert's general principle in relation to

an elastic body.
We very frequently with problems of equilibrium of an
deal
elastic body; in this case accelerations are equal to zero and
equation (11.14) formulates Lagrange's principle, i.e., the prin-
ciple of virtual displacements:
(S) (T)
t.
(11.15)
It should be emphasised that in forming equations (11.14) and

(11.15) we vary the displacements u, v, w and the corresponding
strains, while the external forces
A* v \f
I v^
"7 \? V ^7
y v ,
Ap, /p, ^p,
the stresses
yf v 7 Y 7 x
and the inertia forces
d 2u d2v
are not varied. Bearing this in mind, we put the variation sign
before the integral signs in the left-hand member of equation
(11.15) and write it down as
(S) (T)
where
(S) (T)
Sec. 78] LAGRANGE'S VARIATIONAL EQUATION 355-
U is the work done by all external forces applied to the body

on virtual displacements. But this is not the work A actually
done by the external forces: the equilibrium will be attained only
when the external forces increase infinitely slowly from zero to
their final values (producing no accelerations); then, according
to Clapeyron's theorem, U = 2A and equation (11.15) can be
written in either of the following two forms:
\U- fWd* = 0, (ll.lGa)

I (T)
Wdi = 0. (11.1Gb).
["-/
The quantity
is the force function and = F is the potential energy of the

H
system of external and internal forces. Equation (ll.lGa) gives
the necessary condition for a maximum or minimum of the po-
tential energy of the system if an elastic body is in equilibrium.
On the basis of Lejeune-Dirichlet's criterion the potential energy l
has a minimum value if the equilibrium is stable. Instability of

elastic equilibrium may occur, as is known, in the case of thin
bodies: bars, plates and shells.
If displacements are prescribed on the surface of a body (i.e.,
if we have the second problem of the theory of elasticity), their
variations should be equal to zero; if, at the same time, the body
forces are absent, the variation of the work done by external
forces is zero:
and equation (ll.lGa) becomes

5
fWd* = 0. (11.17)
It formulates the theorem of minimum elastic energy: the actual

state of stress and strain is that for which the total elastic energy
of the body has a minimum value. We
shall call this theorem the
firsttheorem of minimum and point out once again that the dis-
placements and strains corresponding to the state of equilibrium
are varied while the stresses are not; in fact, according to (11.13a)
It is given in the courses on theoretical mechanics.

in forming the force function

F=U f Wd^ (11.18)
whose minimum is to be sought according to (11.16a), the elas-

tic energy W is calculated as the work of forces
L(Xx e xx + Y y e yy + Zz ezz + Yz e yz + Zx e zx + Xy e xy \ (11.19)
and then stresses are expressed in terms of strains by the gener-

alised Hooke's law; thus, W
is represented as a function of strain
components.
However, in investigating the bending of bars, plates and
shells of small thickness the hypotheses of plane sections and
linear elements introduced there permit the elastic energy to be
calculated as the work of the bending and twisting moments
and shearing forces. For instance, in the bending of a straight
bar we isolate its element by two adjacent sections; then, neg-
lecting the work done by the shearing force, we have the energy
of bending of this element
1 JI/T dX
where l
's the curvature of the axis of the bar.
The relation between moment and deformation yields
~~~T"'
The energy of bending per unit length of the bar (dx=l) is
2
w=v(l-} .
But because of the simple approximate relation between defor-

mation and deflection of the bar v
~~ '
P dx*
we are in a position to express the elastic energy as a function
of the displacement v:
For example, for a beam under a distributedload of intensity

q(x) the work of external forces in equation (ll.lGa) is
Sec. 78]
_
and the equation becomes
LAGRANGE'S VARIATIONAL EQUATION
_ 357
Thus, the problem reduces to the determination of such a func-

tion v(x) that minimises the integral
Jt. (11.20)
In the case of the bending of a moderately thick plate we can

also calculate the elastic energy as a function of the deflection
w by using the results of Chapter X; on the basis of Clapeyron's
theorem we obtain per unit area of the middle surface
here
Pi
t
,
2
* r
ar e the curvatures of the middle surface,
1
= W
t 2 is the torsion.
d d
Substituting these values and the values of moments (10.12) and

(10.14) in the preceding expression, we obtain the elastic energy
of the isolated element:
+20 )
it over the whole of the middle surface we find the

integrating
total energy of the plate. It does not involve the work of the
shearing forces A/\ and N2 since the corresponding shearing ,
strains e xz and e yz according to the hypothesis of linear elements,

,
are zero. By using expression (11.2T), we form equation (ll.lGa)

for a plate bent by a load q(x, y). It requires the value of the
integral
358 VARIATIONAL METHODS OF THEORY OF ELASTICITY [Ch XT
extended over the whole middle surface of the plate to be a

minimum.
The problems of finding functions which minimise a definite
integral are treated in the calculus of variations. Here we shall
dwell only on the so-called direct method of the calculus of
variations which was first introduced by W. Ritz in solving the
problem of bending and vibrations of a plate, in relation to in-
tegral (11.21).
79. RITZ-TIMOSHENKO METHOD

The essence of this method is thai the form of unknown func-
tions u, v, w
prescribed in such a manner that the boundary
is
conditions for them (the conditions on the surface of a body) are
satisfied. At the same time these functions should contain a suffi-
cient number of arbitrary parameters by varying which we can
vary the functions themselves without violating the geometrical
constraints imposed on the body (i.e., constraints preventing
displacements).
In the case of the bending of a bar [formula (11.20)], the de-
flection is prescribed in the form
v (x) = Cl v, (x) + c 2 v 2 (x) + . . .

-f- c n vn (x), (1 1
.22}
where c\, ..., c n are arbitrary parameters, and Vi(x), v 2 (x)>

c2 ,
..., v n (x) are functions which are chosen arbitrarily, in gener-

al, but satisfy the conditions imposed on displacements. For
instance, in the case of a cantilever fixed at the left-hand end
(x Q) all functions must satisfy the conditions:
forjc = ^(jc) = 0,
-^j-
= 0. (11.23)
In the case of a beam simply supported at the ends (x = and

# = /), the conditions will be:
iorx = andx = l vk (x) = 0. (11.24)
These geometrical conditions are alone obligatory; the satisfac-
tion of statical conditions is but desirable. For instance, in the
case of a cantilever subjected to a distributed load only there
must be:
for*-/
(11.25)
for*-/
Sec. 79] RITZ-TIMOSHENKO METHOD 359
The of these conditions is not necessary, generally

satisfaction
speaking, but they determine a more or less successful choice
of functions Vf.(x), and the success of the application of the
method itself in a given problem rests entirely on this choice.
For example, if we assume for a simply supported beam
m m
...+cm x (l x, .
(11.26)
the geometrical conditions (11.24) will be satisfied, while the

statical conditions are not satisfiedbecause the bending moments
at the ends do not vanish; this circumstance, however, is alle-,
viated if a sufficient number of terms is taken in formula (11.26).
Assuming
mnx ,., . ~_v
,
(11.27)
we satisfy both the geometrical and statical conditions; there-

fore, we obtain a good result even for a small number of terms
in formula (11.27).
To illustrate the procedure of solution we shall carry it through
in the case of a uniform load. The work done by the load is
The terms of this sum vanish for m even; therefore,
(x)dx = ^-^cm ^; m = l, 3, 5,
The elastic energy is
^-2; c> -H
:
4
The force function is
m
it is a polynomial of the second degree in the coefficients c m

of function (11*27). In order to obtain a minimum of this func-
tion we equate to zero its partial derivatives with respect to c m :
m
Hence
The maximum deflection (at x=//2) is
If we
retain only the first terms in formula (11.27), then, taking
into account that the error in an alternating series is less than
the first omitted term, we see that this error is less than
-^
= - = 0.0041 = 0.4%
and
J ~ __
ÊJ
~ 382.5/
instead of the exact value
304% j-
F=
In the case of the bending of a rectangular plate having the
sides a and b long and simply supported along the contour, its
leflection can be expressed as
.
nny
1
,
., om
Substituting this value in the expression of the elastic energy

(11.21') and carrying out integration, we obtain
->2
'mn-
If the plate is subjected to a single concentrated load P at a

loint *=!, f/=ri, the work done by the external forces is
The force function will be expressed as
Here we satisfy not only geometrical but also statical conditions on the
1
onlour, since the supporting bending moments MI and M* vanish.

Sec. 79] RITZ-TIMOSHENKO METHOD 361
By setting its derivatives with respect to the coefficients Cmn

equal to zero, we obtain a system of equations of the form
a
-
b
--+- 4 \ a2 '
b2
\C mn
)
whence we find the coefficients C mn :
Sln
mr.l .
Sin
n^
-f n 2 !2 ) 2 ~~a~ ~T '
Substituting this in (11.30), we obtain the equation of the de-

flectedmiddle plane of the plate:
mrc; .
/ZTC7]
~"fl~
Sin
T" sin
. mnx
-
.
sin
nny
r
If we assumehere that m=
1, 2, 3, .,
oo and n= 1, 2, 3, oo, . . . . . ,
we obtain the exact solution of Navier presented in Sec. 73 of

Chapter X; retaining the first term alone in (11.30), we find the
approximate equation:
71$ 7CYJ
sin sin r1
a b . nx . TCV
sin sin :-
+
.
Dn* (a
2
62) 2
If the load is applied at the centre u=4- >

^ == T) > the maxi-
mum deflection will be at the same point:
W max f = D^ ^ 2
_|_ 2)2
This first approximation differs from the exact value in the

case of a square plate by less than 10 per cent.
The foregoing examples refer to bars and plates. In the three-
dimensional problem of the theory of elasticity in the case of
equilibrium of a body of arbitrary shape the general variational
equation (11.15) is applied as follows.
Let us suppose that a body is acted upon by the surface trac-
tions
Xv * Yv , Zv
and the body forces
Xp, Kp, Zp
and, in addition, geometrical constraints (for instance, fixing)
are imposed on it, because of which the displacements of some
of its points are predetermined; for simplicity, we assume that
these constraints are imposed only on the surface of the body;
hence on the whole surface or part of it where the constraints
are imposed the displacements must have prescribed values
u = u(x, y, 2); v = v(x, y, 2); w=w(x,y,z), (11.31)
for instance, they must vanish.

Let
u (x, y, z)\ v (x, y, z): W Q (X, y, z) (11.32)
now be arbitrary functions; if constraints (11.31) are imposed,

these functions must conform on the surface (or on the corre-
sponding part of it) to the constraints, i.e., there must be
UQ = U\ v^^v; WQ=W\ (11.33)

furthermore, let
fm(x* y. *); ?,(*. y, *); <M*, y, *) (m= i, 2, 3, . .

.)
also be arbitrary functions but vanishing at the points of the

surface of the body, where the constraints are imposed. Let us
specify the displacements of points of the body by the follow-
ing formulas:
u = u + Iam/m ,
(11.34)
dm, b mj c m are arbitrary constant coefficients. Varying them, one

can obtain from (11.34) a set of the states of strain in the body
satisfying the conditions imposed by the constraints, since at any
values of the constants a m 6 m c m conditions (11.31) are satis-
, ,
fied.
Substituting the values (11.34) in equations (III) of Chapter

IV and hence in formula (3.33) of the elastic energy
^=TP + ^(^ + ^ + ^ + l^(^ + ^+^)]- dl-35)
we observe that the integral in the right-hand member of equa-

tion (11.15), which represents the elastic energy of the whole body,
appears to be a homogeneous function V of the second degree in
the constants a m fr m cm, ,
:
Further, for substitution in (11.15) we calculate the variations

of displacements (11.34), taking into account that only the con-
Sec. 79] RITZ-T1MOSCHENKO METOD 363
slants a m ,
fc m ,
c m are varied:
8w = 5ty m U, y, z)8c m .
Introducing these values of the variations in equation (11.15) and

collecting the coefficients of the variations- of the constants 6a m ,
66 m 6Cm, we obtain the system of equations of the form

,
(S) (T)
:
=0 f (11.36)
(S) CO
where m=l, 2, 3, .... Inasmuch as all variations 6a m 6ft m 6c m , ,
are arbitrary and independent of each other, equations (11.36) are

possible only if the coefficients of all variations are zero; hence
we arrive at the following system of equations:
da m
(S) (t)
dV
(11.37)
(S) (T)
dV
Oc m
(S)
The left-hand members of these equations are linear functions

of the constants a m b m c m while the right-hand members are
, ,
independent of these constants; the number of equations is equal

to the number of unknowns a m b m c m , , .
Since* V is a homogeneous function of the second degree in

am> b m Cm, the coefficients of these unknowns in the left-hand
,
members of (11.37) form a symmetric matrix, and the system of

equations (11.37) is, in the terminology of structural mechanics,
canonical.
Introducing the. values of the coefficients a mt b mt c m obtained ,
from (11.37), in (11.34), we find the displacements; further, ac-

cording to formulas (III), we determine the strains e xx e yyt ..., ,
364 VARIAT1ONAL METHODS OF THEORY OF ELASTICITY [Ch XI
e xy and, at last, from the equations
Ky X6 (11.38
we find the stresses. It is to be noted that at the surface of the

body these stresses in general, correspond to the pre-
will not,
scribed loads X v y,, Z v since in the process of solution it is
, ,
not required to satisfy the statical conditions on the surface
Xv = Xx cos (vx) -f- X y

cos (vy) 4- X 2 cos (vz),
Yv = Yx cos (vx) 4 ^ cos y (z;y) 4- Y z cos (TJZ), (11.39)
Zv = Zx cos (UK) 4~ Z> cos (i;y) 4 Z^ cos (vz).
These conditions will be satisfied approximately, and the greater
the number of terms taken in expressions (11.34) and, conse-
quently, the greater the number of constants a m b m c m introduced , ,
in the solution, the more exactly

these conditions be satis- will
fied. We met with this circumstance above in considering the
problem of the bending of a bar and specifying the displacements

in the form (11.26) where the statical conditions = at x = Q M
and x = l were not satisfied.
80, CASTIGLIANO'S VAR1ATIONAL EQUATION
In Chapter IV it was pointed out that three differential equa-

tions of equilibrium
OX Z
dx Oy ~0z~ ^" p
= '
dYx dK,
dx dy ^-+r p =o, (I)
dZ x
dx
4 '
<?y
^
-f
rfZ,
^ =
are inadequate for determining the stresses
*,, ry ,
z,, A V, K., z,. (11.40)
and that the problem of the theory of elasticity is therefore

statically indeterminate; hence it is possible to obtain any num-
ber of systems of stresses (11.40) satisfying equations (I) and the
Sec. 80] CASTIGLIANO'S VARIATIONAL EQUATION 365
prescribed conditions on the surface of a body

v X =X
x cos (vx) y
cos (vy) +X +X
z cos (vz),
Yv = Yx cos (u*) + K cos (vy) + K, cos y (TO), (II)
Zv = Z^ cos (vx) + Z cos (vy) Z cos y -+- 2 (TO).
Let us introduce, in addition to (11.40), another system of stresses

hich differs from it by infinitesimal quantities
(11-41)
and also satisfies the equations of equilibrium
The quantities &X X t>Y y 6Z Z 6^ y 6YZl 6Z X will be called the varia-

, , , ,
tions of stresses. For generality, we assume that stresses (11.41)

are due to the load which differs from the former by an infinitesi-
mal quantity so that on the surface
X v (Xx cos (vx) + (X y
cos
= (Yx cos (vx) + (Y y

cos (vy) +
(11.43)
cos cos
where 6X V 6y v 6Z V are the variations

, , of surface tractions. Sub-
tracting by members equations (I) from (11.42), and equations
(II) from (11.43) we obtain:
= IXX cos (vx) + ZX y

cos (vy) + *X cos Z (TO),
j 45)
Consequently, the variations of stresses themselves must satis-

the equations of equilibrium (11.44) without body forces and
e surface conditions (11.45).
Let the system of stresses (11.40) actually exist in an elastic
body under the action of given surface and body forces
xm r f) ,
zv ; XP, rp zp
, .
25-1013
Let us see what makes it different from the whole variety of

systems of stresses (11.41) close to it and how it can be deter-
mined from this variety. To this end we revert to the fundamental
equality (11.10), leaving out the forces of inertia
d 2u d2v
since we
are dealing with the equilibrium of a body.
We
take for the displacements ', v w' and the corresponding
f
',
strains e' xx e' yy ..., e' zx involved in (11.10) the displacements

, ,
and strains actually existing in an elastic body and denote them,

as usual, by
w\
ppppf>p
^xx^ ^yy
u, v,
^zz^ ^xy> ^yz* ^zx*

%
}
\
>
(11.46)
we connect the strains with the actually existing stresses by

Hooke's law:
(V)
c, y tii
Further, we take for the stresses denoted in (11.10) by X Xl Y y ,
Z : X y Y z Z x the variations
, , ,
of the actualappearing in the right-hand members of

stresses
(V); we can do
because variations (11.47), according to the
this
condition, satisfy the equations of equilibrium (11.44). Accord-
ingly, X D Y v Z v have to be replaced in (11.45) by the variations
, ,
^ V^ 5* I/" 5* *7
Oy\ v , O/ , 02-r,y
of the loads specified on the surface of the body. Finally, the

body forces Xp, 7p, Zp must also be replaced by their variations
/ \ & V^ \ / 7 \. 5 V/" / SJ
(Ap), (i p), (Z,p;,
however, we have assumed above

that the body forces are iden-
tical in the neighbouring states of stress [cf. equations (I) and
(11.42)]; therefore, these variations should be set equal to zero;
it follows that the second integral on the left-hand side of equal-
ity (11.10) vanishes and may be rewritten as
f (?>Xv u +W v + *Z w) = f (8*,^ + W
v v rfS y
e yy +
(S) <*)
,J</T. (11.48)
Sec
We may
80)
_
(Xx e xx
CASTIGLIANO'S VARIATIONAL EQUATION
note that the integral in the right-hand member repre-

sents the variation of the doubled elastic energy of the body
e + Zzzz
+ Y yyy
y yy e xy + Y zyz
e + Xyxyz zz e +Z e y z yz
_ x zx ) d*
367
(T)
[cf. formula (3.32) of Chapter III] on the assumption that the

strains (and displacements) are not subject to variation, and the
stresses alone are varied, their variations satisfying the equations
l
of equilibrium (11.44) .
We
now introduce the expressions of strains in terms of
stresses (V) in the integrand on the right-hand side of (11.10);
this integral will then be expressed as
+ Z W + XJ
z y
YJY, + ZglZx )\d*. (11.49)

But it is obvious that
= 8 (Xx Yy ), etc.
etc.
Substituting these values in integral (11.49) and putting the

variation sign 6 before the brackets and then before the integral
sign, we represent it as
(t)
t r o n\. -i 1
(11.50)
Comparing this with formula (3.34) of Chapter III, we see that

the integrand represents the elastic energy per unit volume of
the body, and integral (11.50) can be written as
(t)
This means that in the process of variation we

1
compare the actual state

of stress with others which are close to it and also satisfy the equations of
equilibrium.
25*
Substituting this in the right-hand member of (11.48), we obtain

Castigliano's variational equation
f (8A> + 8K + 8Z ,w)</S =
t ,t> t
8
/ Wdt. (11.51)
(S) (T)
The left-hand member of this equality represents the work done

by the variations of the surface tractions on the actual displace-
ments; the right-hand member involves the variation of the elastic
energy of the body resulting from the variation of the actual
state of stress.
The right-hand member of (11.51) may conveniently be put in
the form of (11.49) where the variations of the stresses X x Y y , ,
. . . Zx must satisfy the equations of equilibrium (11.44) while the

,
variations of the surface tractions

^ V">
*
V> ^V
must be calculated by formulas (11.45) in terms of the variations
of stresses.
The variational equation (11.51) gives the desired relation be-
tween the actual stressed state of equilibrium of an elastic body
and the neighbouring states which also satisfy the equilibrium
conditions.
In the discussion above we did not touch upon the compatibil-
ity conditions [equations (IV) of Chapter IV], which take the
form of Beltrami's equations for an isotropic body in the absence
of body forces (Chapter V, Sec. 36). These equations should also
be satisfied; it is important to note, however, that the compatibil-
ity equations (IV) appear to be a consequence of the variational
equation (11.51) and can be deduced therefrom
1
Consequently, .
when applying the variational equation (11.51) there is no need

to be concerned with satisfying the conditions of compatibility
(continuity of strain components).
81. APPLICATION OF CASTIGLIANO'S VARIATIONAL

EQUATION TO THE PROBLEM OF TORSION
OF A PRISMATICAL ROD
To apply Castigliano's
principle in this problem it is necessary,
assign a system of stresses satisfying the equations
first of all, to
of equilibrium (I) and the surface conditions; this state of stress
can easily be obtained by using Prandtl's stress function U(x, y)
1
See L. S. Leibenzon, Theory of Elasticity, 2nd ed., Moscow, Gostekhizdat,
1947, Chap. XI, Sec. 122.
Sec. 81] APPLICATION OF CASTIGLIANO'S VARIAT1ONAL EQUATION 369
{formula (8.16)]:
Xx = Y = Z> = Xy == Oj
y z
X = du Y = (11.52)
We confine ourselves to the case of a simply connected cross

section; the equilibrium equations are then satisfied; should we
also require the function U to vanish on the contour, then the
lateral surface of the rod will be free of loads and, consequently,
the conditions on it will be satisfied. The conditions at the ends
of the rod, as we saw in Chapter VIII, are prescribed, according
to Saint-Venant's principle, in integral form only: the resultant
force vector at either of the bases is equal to zero, and the re-
sultant moment vector is expressed in terms of the stress function:
f f Udxdy. (11.53)
The displacements of the points of the rod, according to (8.1)

and (8.2), are given by the formulas
# = T yz; v = txz\ w = ty(x. y), (11.54)
The variation of stresses (11.52) will be obtained by giving the

variation 6(/ to the stress function; then
IXX = ?>Yy = t>Zz = IXy = 0;
and the variation of the elastic energy (11.50) will be expressed

as
The stresses X V9 Y v> Z v on the lateral surface are equal to zero

and, if the variation of the stress function vanishes here, we have
and, consequently, the surface integral in the left-hand member

of (11.48), taken over the lateral surface, becomes zero. LeJ: us
now turn to the bases of the rod; the stress distribution is iden-
tical in all cross' sections, the bases of the rod included; hence,
if we vary stresses (11,52), the conditions on the bases will also
be varied; the integral in the left-hand member of (11.48) will
be different from zero, and it should be evaluated. On the bases
we have, from (II),
Xv = X z\ Y v = Yz \ Zv = 0.
Let us assume that the lower base 2 = is fixed in its plane; then
u=v= and the corresponding integral becomes zero. It remains
to evaluate the integral over the upper base z = l where / is the
=
length of the rod; therefore, putting z l in (11.54), we write
down equation (11.51) as
(860 * dS.
(S) (S)
(11.55)
Consider the first of the integrals on the right-hand side:
(11.56)
We apply integration by parts to the inner integral:
y
fy(W)dy = ylU\- y.
flUdy.
y.
Since, according to the condition, the variation f*U is zero on

the contour, the first term in this formula vanishes (Fig. 102),
and we have
=- flUdy.
fy-jy(W)dy
Vi
Substituting this in (11.56), we find that
If we apply an analogous transformation to the second term in

the right-hand member of (11.55), we obtain the same result;
therefore, (11.55) yields
J (*Xv ii + W v + 8Z w) dS = - 2*1 f f W dx dy.

v v
(S)
The variational equation (11.51) takes the form
It is taken into account here that
2(1 + ) _ l
Sec. 81] APPLICATION OF CASTIGLIANO'S VARIATIONAL EQUATION 371
and that when integrating with respect to z the stress distribu-

tion is independent of z. Dividing the preceding equation by ,
we transpose the terms to the left-hand side

all and put the varia-
tion sign before the braces:
This equality requires that the function U(x, y), corresponding

to the actual stress distribution in torsion, should minimise the
Fig. 102
value of the integral appearing in the braces. The condition

{11.57) is equivalent to the differential equation (8.67)
with the condition on the contour U=0; this can be proved. In

fact, carrying out variation in (11.57), we obtain
-a (H.58,
Applying integration by parts and interchanging the order of

variation and differentiation, we have
But in the substitution outside the sign of the inner integral the
values of Xi and x2 refer to the points of the contour at which,
372 _ VARIAT10NAL METHODS OF THEORY OF ELASTICITY
according to the condition, 66/=0 and, therefore,

r r d*U
_ [Ch X*
J J
Similarly, we obtain
"*"
Substituting this in (11.58) and taking f*U out of the brackets,
we have
But since the variation 6U is arbitrary, the integral on the left-

hand side of equality may vanish only if the bracketed
this
expression is equal to zero at all points of the region of inte-
gration, i. e., at all points of the cross section. Hence, for all
points of the section
d*U
this coincides with equation (8.67), which is the compatibility

condition in the torsion problem; this is quite natural, for, as
stated at the end of Sec. 80, Saint-Venant's compatibility con-
ditions (IV) appear to be a consequence of Castigliano's varia-
tional equation.
The variational equation (11.57) provides a very important
means solving problems of the torsion of a rod, having a
for
complex cross section; the solution of such problems by the usual
method of integrating equation (8.67) with the boundary condition
(7=0 is obtainable in extremely rare cases; in the meantime the
variational equation (11.57) makes it possible to find an approxi-
mate solution in the case of a simply connected section with suf-
ficient accuracy.
In fact, let the contour of the cross section be given by the
equation
F(x,y) = Q*. (11.59)
Let us take the stress function in the following form:
,(x,y)] 9
(11.60]
where Ci, C2 , ,
C n are arbitrary constants, fi(x, y), f2 (x, y),
. . . . . ,
..., fn(x> y) are arbitrary functions. The function satisfies the U

1
Since the section is assumed to be simply connected, the function F(x\u]
' '
: ^
must not vanish inside the section.
Sec. 82] FIRST PROBLEM OF THEORY OF ELASTICITY 373
boundary condition of the problem because it vanishes on the

contour of the section. Its variation will be obtained by varying
the coefficients C m The variation 6C/ also vanishes on the contour.
.
Substituting this value of U in equation (11.57) and carrying out

integration, we find that the integral in its left-hand member will
be expressed as a quadratic function, i.e., a polynomial of the
.second degree in the constants Cm :
. . .
-f- /IQIÎ ~T~ ^02^2 I
and to obtain an approximate solution

of the problem it is necess-
ary to find a minimum
of this function, i.e., to equate to zero its
partial derivatives with respect to Ci, C2 C n Then, as men- , . . .
,
.
tioned in Sec. 79, we obtain a canonical system of linear equations

to find the coefficients C m and the function (/, according to
,
(11.60), will be determined. Using this method, L. S. Leibenzon

solved a number of torsional problems for bars of various cross
sections. (See his Variational Methods for Solving Problems of the
Theory of Elasticity, Moscow, Gostekhizdat, 1943.)
82. FIRST PROBLEM OF THE THEORY OF ELASTICITY;

SECOND THEOREM OF MINIMUM ENERGY
The first problem of the theory of elasticity was defined in
Sec. 37 of Chapter V
as the case where loads are prescribed on the
surface of a body; these loads, according to the meaning of the
problem, must be the same both in the actual state of the body
and in all neighbouring states which are obtained by variation;
hence, we must equate to zero the variations of loads v Yv Zv X , ,
on the whole surface of the body in equation (11.51); then we

get the following variational equation:
0. (11.61)
This is a necessary condition for the elastic energy of a body

in the actual state of equilibrium to have a or a mini- maximum
mum value in comparison with all possible neighbouring states
of equilibrium; this equation will be obtained in expanded form
by setting integral (11.49) equal to zero. To decide whether the
elastic energy is a maximum or a minimum, we form the total
variation of the elastic energy. The variation given by formula
XI 1.49) is the first variation; we see that it is formed as the first
374
_
differential
formal sense;
VARIAT10NAL METHODS OF THEORY OF ELASTICITY
of a
for
function of several independent variables in a
example, the first differential of a function
_ [Ch. XI
*(*, y) is
The second differential will be obtained by differentiating dzwith

respect to x and y, but considering dx and dy constant:
By analogy with this we get the second variation of the energy
by varying the stresses Xx ,

Yy ..., Zx in integral (11.49)
,
but
considering 6^, 6Ky ,
. . .
,
>ZX constant; consequently,
= ^J
I- *)[(*,
It will be noted that the expression in the braces represents
the elastic energy per unit volume produced by the stresses
At the end of Sec. 21 of Chapter III it was pointed out that

the elastic energy is always positive; hence
82 / ( Wdt\>0. (11.62)
\(t) /
Reverting to the analogy with differentiation of functions, we

recall that the condition
is a necessary condition for a maximum or a minimum of the

function z\ if it turns out that
d*z > 0,
then the function z has a minimum. Thus conditions (11.61) and
(11.62) show that the actual stressed state of equilibrium of an
Sec 83] APPROXIMATE METHOD BASED ON VARIATIONAL EQUATION 375
elastic body differs from all neighbouring states

of equilibrium
in that it minimises the This is the principle of
elastic energy.
least work known in structural mechanics. To apply it, the elastic
energy, according to (11.50), (11.51) and (11.61), should be ex-
pressed as a function of stresses:
+ 2(1 + <,)(X + Y + Z. (11.63)
The problems of structural mechanics deal with rods or bars

to which, when
subjected to tension, compression and bending,
it is permissible to apply the hypothesis of plane sections; this
makes it possible to express the elastic energy as a function of
bending and twisting moments and also longitudinal and shearing
forces rather than as a function of stresses:
~~ -
J EF GFoi ~T~ GC
here C
the torsional rigidity of the bar; the energy is related
is
to unitlength of the axis of the bar and not to unit volume;
therefore, the variational equation (11.61) requires in this case a
minimum of the integral
N 2
Q
2
M
taken throughout the length 5 of the axis of the bar; this condition,
as is known, permits the solution of statically indeterminate prob-
lems. The general problem of the theory of elasticity for a body on
the surface of which there are given loads is also statically inde-
terminate; to solve it one may apply an approximate method
based on the variational equation (11.61) and analogous to the
Ritz-Timoshenko method presented in Sec. 79.
83. APPROXIMATE METHOD BASED ON VARIATIONAL

EQUATION (11.61)
This method has been proposed by P. F. Papkovitch and con-
sists in the following. Let us attempt to find a state of stress in
a body satisfying the equations of equilibrium (I) and the con-
ditions specified on the surface of the body; we denote the com-
*
ponents of the corresponding stress tensor by
Z?\ Kf, Z<>, Xf\ (11.64)

376 VARIATIONAL METHODS OF THEORY OF ELASTICITY [Ch Xt
Let us now find several states of stress with the components

v(^) v("0 r7("l) iX'W) *7(tn^ v(^) /1 1 rC*\
Aje i T y , .2 ,
/ z , ^jf , Ay , (II. DO)
where m=l, 2, 3, . . . .
These states of stress must also satisfy the equations of equilib-

rium (I), but at the same time the whole surface of the body
must remain free of stresses so that on the surface, according
to (II),
Xxm} cos (ox) + X m} cos (oy) + X m} cos (vz) = 0,
(
y
(
Y^ cos (ox) 4- Yym) cos (oy) + Y m) cos (vz) = 0, (

z .
(II .66)
cos (we) + Zy cos (vz) =

m) ( m} m)
Zi cos (i>y) 4- 2i 0,
The states of stress (11.64) and (11.65) do not satisfy, in

general, the compatibility conditions, i.e., Beltrami's equations
(VII) or Mitchell's equations (Chapter V, Sec. 36) because this
requirement would make the solution of the problem
]
extremely
difficult, but it is not obligatory, as we have seen, in Castigliano's
method; at the same time the state of stress (11.65) subject to
conditions (11.66), as such, is impossible, since it would give
initial stresses ("self-stressed state"), whose absence is assumed
in advance (see Sec. 21) according to the hypothesis of the nat-
ural state. We
shall not, however, consider this state of stress by
itself but, using (11.64) and (11.65), we shall construct a new
state of stress
Yy Yy 4" C m Ky '>
ZX ZX (11.67)
:===
i"r($S) i
v/'"* *7( fft) y y(
Z
'
*Z "I J-^m^'Z >

Ay .Ay
where Cm are arbitrary constants.

It is evident that stresses .(11.67) :
satisfy the equations of equilibrium;

(1)
(2) satisfy the conditions on the surface of the body;
(3) contain linearly several arbitrary constants Cm permitting ,
the state of stress of the body to be varied so as to satisfy the

conditions of equilibrium and the surface conditions, which is
required for Castigliano's variational equation (11.61) apply, to

i.
e., for finding
a minimum of the elastic energy of the body.
The greater the number of the states of stress (11.65) and of the
corresponding constants Cm we introduce, the wider the possibil-
ities of varying the energy, the closer the solution obtained to
1
If stresses (11.64) satisfied the compatibility conditions, they would give
a complete solution of the problem.
Sec. 83] APPROXIMATE METHOD BASED ON VARIATIONAL EQUATION 377
the exact solution and the more closely we come to the satis-
faction of the compatibility conditions.
If we substitute the values of stresses (11.67) in the expression
of the elastic energy
4-2(l (1 1
.68)
and carry out integration, it will be represented as a function

(polynomial) of the second degree in the constants Cm :
i = l, 2, ..., M; y=l, 2, ..., TV; * = 1, 2, . .

., N\
let us write down this expression in extended form:
.
-\-A mn C m C n -\-...
(11.69)
To obtain a minimum of the elastic energy we must equate to

zero the derivatives of expression (11.69) with respect to all
constants C m this will give a system of N linear equations for
:
the determination of these constants, such as
Aml C 1 -\-A m2 C 2 ~{-
dW (11.70)
~dC^
:
+ A nmC
We note that from the first equation of (11.70) we have
and from the second
therefore,
"mn "n (11.71)
Hence, the system of equations (11.70) is canonical, according to

the terminology of structural mechanics; this will provide an
important means of checking rather complicated calculations of
the coefficients AH of system (11.70).
It should be noted that equations (11.70) may be regarded as generalisa-

tion of canonical equations of structural mechanics; this generalisation is nat-
urally brought about as one passes from a bar system to a body of arbitrary
shape; the constants C m play the role of redundant unknowns, but their num-
ber is infiinitely large in general, since an elastic body is a system with an
infiinite number of redundant unknowns. By assigning a finite number of
unknowns C m in sums (11.67), we come to an approximate solution of the
problem by reducing the number of redundant unknowns. Formulas (11.67)
1
themselves are a generalisation of the formulas for moments of transverse

and longitudinal forces in bar systems; if, for instance, the redundant un-
knowns of such a system are denoted by Ci, C 2 C 3 C N the moment at
, ,
. . . , ,
a given section of the system will be expressed by a formula similar to (11.67):
M=M +C M +C M +
((
1 1 2 2
... +CN M N = M V + 2Cm M m (
, (11.72)
where M() is the moment of a given load; MI, Af 2 ..., M N are the moments
,
due to unit loads Ci = l, C 2 = 1, C N ~ 1.

. .
.,
The coefficients Ai$ in (11.69) and (11.70), which are integrals of squares
and pairwise products of functions involved in expressions (11.67), represent
a generalisation of known Mohr's integrals
MM' .
The analogy between (11.67) and (11.72) shows that the state of stress
(11.64), which satisfies the equations of equilibrium and the boundary condi-
tions but does not satisfy the compatibility conditions, corresponds, in the
case of the bar system, to the state of stress in the basic statically determinate
system of the force method (when the redundant constraints are removed in
the given system). The state of stress (11.65) is analogous to the "unit" states
of stress of the basic system which are due to the action of unit loads C m l. =
Consequently, these quantities may be interpreted as generalised forces in
formulas (11.67).
All the arguments of this section rest upon the condition that
the stresses on the surface of a body are not varied since it is
assumed that they are given. However, in the case where the
semi-inverse method is applied the stress distribution sometimes
is not given on certain parts of the surface, only the resultant
force vector (or the resultant) and the resultant moment vector
being given on these parts of the surface. For instance, in Chap-
ter VIII, where we considered the problems of torsion and bend-
ing of a prismatical bar, we prescribed on its bases: in bending,
the load Q subject to the condition that the moment of the
tangential tractions producing it be equal to zero; in torsion, the
twisting moment z M
subject to the condition that the resultant
force vector of the tangential tractions producing it be equal to
Sec. 84] LAME'S PROBLEM FOR ELASTIC RECTANGULAR PRISM 379
zero. The
stress distribution in all cross sections of the bar ap-
pears be
to identical; thus, varying the stresses throughout the
bar, we must admit their variation on its bases as well. In such
cases it becomes necessary to resort to the variational equation
of the general form (11.51) instead of (11.61). The next section
is concerned with the application of Castigliano's method to the
general problem of a bar of rectangular cross section.
84. LAME'S PROBLEM FOR AN ELASTIC

RECTANGULAR PRISM
Imagine an elastic rectangular prism whose surface is acted

upon by loads either normal or tangent to its faces; it is required
to determine the stresses at each point of the prism. Lame for-
mulated this problem in 1852 and emphasised its great impor-
tance in the sense of practical application, since its solution would
provide a means of checking the approximate elementary methods
of investigating the bending and torsion of beams of rectangular
cross section and also the compression of columns by a nonuni-
form load. At the same time, Lame pointed out the extreme
difficulties this problem involved in the solution of the differen-
tial equations (VI) with the conditions specified on the surface
of the prism; indeed, the problem thus formulated has yet to be
solved. In the meantime Castigliano's method permits the solu-
tion of Lame's problem to be carried through in some particular
cases of loadings on the faces of the prism. Such a solution, as
ascertained above, will be approximate because in practice we
have to work with a finite number of terms in the sums of for-
mulas (11.67) and, consequently, with a finite number of coeffi-
cients Cm that are varied. The larger this number, the better the
compatibility conditions will be satisfied; at the same time, there
are great possibilities here for satisfying the boundary condi-
tions on the faces of the prism, and this is important, if all three
basic dimensions of the prism are of the same order and the
alleviation of the boundary conditions by the use of Saint-Ve-
nant's principle is not justified. Let us consider the procedure of
solving the problem in greater detail.
According to (11.67), the stress tensor should be represented
as the sum of two tensors:
The first, which satisfies the equations of equilibrium (I) and

the given boundary conditions, will briefly be termed the basic
tensor. Its construction offers no difficulty in principle but it is
found to be more or less involved depending on the complexity
of given loads. The second tensor on the right-hand side of
(11.73) must also satisfy the equations of equilibrium (I); it does
not depend on the given load, since the surface of the prism
must be free of stresses; hence it can be constructed once and
for all for a given prism: we shall call it the correcting tensor.
Consider one of the methods of constructing it l
.
Fig. 103
Let a prism be given with the edges of length a, b, c (Fig. 103).

Let us set up the following three systems of functions:
)
= cos -cos = 0, I. 2, ...),
=0, 1, 2, ...),
(11.74)
= 0, 1, 2, ...)
2
They possess the following properties .
1. Each of these functions and its first derivative vanish at

both ends of the corresponding interval a, b, c\ this is also seen
in the graphs of several of the first functions of systems (11.74),
represented in Fig. 104.
1
See the author's paper, Prikl. matem. i mekhan., 1951, Vol. 15, No. 2,
pp. 137-148.
2
See the author's paper, Prikl. matem. i mekhan., 1946, Vol. 10, No. 1,
pp. 193-208.
2. Any function, which can be expanded in a trigonometric

series (Fourier series), can be represented in ,the corresponding
interval with the same accuracy by a polynomial Composed of func-
tions (11.74); if, for example, a function f(x) is defined in the
interval jc = 0, x a it is possible to choose the coefficients c m in
t
the equality
/ (x) = CoP (X) + CjPj (X) + C^P 2 (X)+ . . .

+ Cm Pm (X)+ ...
so as to make the right-hand member represent the function f(x)

as exactly as one pleases. Such a property of a system of functions
*^:
Fig. 104
is called its completeness or closeness (we shall not dwell on a

certain distinction between these concepts). The system of func-
tions (11.74) will be used to construct the correcting tensor as
follows. Since it must satisfy the differential equations of equil-
ibrium (I), we form its components with the aid of stress func-
tions; let us make use, for instance, of Maxwell's functions cpi, cp 2 ,
<p 3 and write down the expressions of the components in the form
of (9.4) of Sec. 56. The stress functions themselves will be repre-
sented as triple sums composed of functions (1 1.74):
(x) P n (y) Pp
m n p
<P 2
= 2 S 2 Bmnp Pm (x) P n (y) Pp (z), (11.75)
Cmn,Pm (x) P n (y) Pp (z),

m n p
26-1013
where A mnp B mnp Cmnp are arbitrary constant coefficients admit-

, ,
ting the variation of functions (11.75). Hence, the components of

the correcting tensor will be expressed as
, f P m (x)P"n (y)Pp (z) +

B mnpPm (x)P n (y)P"P (z)l
/Yy mnpPm (X) P n (y) Pp (z) +
rn n p
+ C mnfP^(x)P,,(y)P p (z)},
(11.76)
A mnpPm (x)P"n (y)Pp (z)],

x =-y n nP P'm (x)P'n(y)P P (Z),
=-sss
m n p
,
-
= 2SS (*).
It can easily be proved that in the state of stress described by

formulas (11.76) all the faces of the prism remain free of loads.
Consider, for instance, the faces x = =
and x a normal to the
axis Ox\ they are acted upon by the stresses x Yx = y ZX X Z
X ,
X ,
which vanish here; this follows from the properties of functions

(11.76):
Pm( 0) = 0; P m( a) = 0;
PL(0) = 0; P'm (a)
= 0.
J
I
(1 }
In the same way we prove that the remaining two pairs of

faces
y = 0; y = b\
2 = 0; z c
are also free of stresses at arbitrary values of the coefficients

Amnp, B mnp , Cmnp. The above-mentioned completeness of the sys-
tems of functions permits the construction, through the
(11.74)
use of Maxwell's functions (11.75), of a very wide class of the
states of stress (11.76) which leave the whole surface of the prism
free of loads; thus, formulas (11.76) represent the correcting tensor
in a very general form. By adding to it the basic tensor correspond-
ing to the loads prescribed on the surface of the prism, we obtain
the general tensor which satisfies the differential equations of
equilibrium of the prism, the boundary conditions of the prob-

lem formulated here and contains an unlimited number of the
parameters A mnp B mnp Cmnp to be varied; these parameters can
, ,
be determined by using Castigliano's variational method con-

sidered in Sec. 83, so that the problem is eventually reduced to
the solution of a system of linear equations of the form (11.70).
It may be noted that the correcting tensor (11.76) is slightly
different in form from that which was first proposed by P. F.
Papkovitch in formulas (11.67); by assigning a number of numer-
ical values 0, 1, 2, 3, A/" .1 to the index
. .
,
in these formu- m
las, we introduce a simple series of the coefficients
C ,
Cj, C2 , ..., CN _ l (11.78)
and for their determination we obtain a system of N linear

equations in N unknowns. If, however, we assign the same nu-
merical values to the indices m, n, p in formulas (11.76), the
coefficients form a triple series and their number will be 3W 3 ;
the number of equations for their determination will be the same.

Let us assign, for instance, to a first approximation, a single
value, m = n=p = Q, to all the indices; we then introduce three
coefficients:
A B C (11 79)
To a second approximation we introduce two values of the indices:

w = 0, 1; n = 0, 1; p = 0, 1; then we obtain eight coefficients:
A)00 AoO> A)10> A)01 \ /ii on\
A A A A I
(ll.oUj
Altogether we shall introduce 24 coefficients A mnpl B mnpj Cmnp .
Thus, a tendency to improve the accuracy of the solution entails

a considerable increase in the number of unknowns introduced
in the problem and creates difficulties of a technical character
associated with the solution of systems with a large number of
equations; this is due, of course, to a complicated three-dimen-
sional nature of the problem itself. The form of solution (11.67)
proposed by P. F. Papkovitch is simpler with regard to construc-
tion, but to obtain the same degree of accuracy it will, ob-
viously, require the introduction of a considerably greater num-
ber of terms in series (11.78). The solution in the form (11.76)
can be somewhat simplified by using incomplete approximations-,
for instance, in introducing a new value of the indices n, p we m t
may retain only those coefficients in the triple indices of which

this new value is met just once; thus, in introducing the second
approximation (11.80), we retain only the coefficients of the first
linewhere the index "1" is met just once. Then the total niim-
26*
384 _ VARIAT1ONAL METHODS OF THEORY OF ELASTICITY
her of the coefficients will be 3X4=12, i.e., it will be reduced

to one half as compared to the complete second approximation.
_ [Ch. XI
The
application of Castigliano's method to Lame's problem
will be illustrated by the simplest example when the prism is
compressed or stretched by equal loads distributed arbitrarily
over two opposite faces, say, over the faces 2 = and z=c\ then
the boundary conditions on these faces will be
X = Y = 0; Z = F(x,
Z 2 z y) lorz =c
X_ 2 = r^ 2 = 0' J Z_ 2 = F(x, y) forz = 0;
the remaining faces must be free of loads. All these conditions

will be fulfilled if the basic tensor is taken in the form
Substitution in the equations of equilibrium (I) shows that they

are also satisfied identically. (Such a state of stress can exist
if the prism is replaced by a block of the same form made up
of separate bars parallel to the axis Oz on the assumption that
there is neither friction nor mutual pressure between them; each
bar will then carry that part of the load which acts on its ends
and the stress Z z developed in it will be constant along the length
and will be independent of z\ the compatibility conditions will
not, of course, be satisfied then.) It remains to construct the gen-
eral stress tensor as the sum of the basic (11.73) and correcting
(11.76) tensors; it differs from the tensor (11.76) only in that the
term F(x, y) will be added in the formula of the component Z z .
Fig. 105 shows a special case of a prism with a square base

a = fe=l, c = 2 subjected to loads which are distributed over the
upper and lower bases according to the law
The surface characterising this distribution is shown on the upper

base of the prism. The load (as the prism itself) has two axes of
symmetry; it is, therefore, necessary to retain in formulas (11.75)
only even-numbered functions P m Pn* Pp having the graphs
,
(Fig. 104) symmetrical about the middle of the segments a, b, c.

The solution of this problem to a second incomplete approximation
has led to a system of 12 equations
l
with respect to the co-
See the
1
author's paper, Prikl. matem. i
niekhan., 1951, Vol. 15, No. 2,
pp. 137-148.
Fig. 105 shows the surface of distribution of stresses Z z at the
section passing through the middle of the altitude of the prism.
In spite of a sharp nonuniformity of
the load acting on the bases, the dis-
tribution of stresses at the mid-
section is sufficiently close to uni-
form distribution, in full accord with
Saint-Venant's principle.
Castigliano's variational method
made it possible to obtain the solu-
tion of Lame's problem for a prism
in othermore complicated cases of
loading. V. P. Netrebko treated thel
problems of the torsion of a rec-

tangular prism under a given distri-
bution of shearing stresses over its
bases, and also the cases of the
so-called torsion with warping res-
traint when one or both bases cannot
warp (as follows from Saint-
Venant's theory) but must remain
2
plane E. S. Kononenko found the
solution of the problem of a prism
compressed between two absolutely
rigid plates in the presence of com-
plete cohesion on the surfaces of con-
tact; the problem was solved to a
second complete approximation (with
24 coefficients) and the stress dis-
tribution within the prism was first
clarified in detail in these complex
loading conditions.
By using the cylindrical co-ordi- Fig. 105
nates the method presented in this
section can be extended to cover the cases of a circular cylinder,
a tube or a tube sector subjected to given loads on the sur-
1
Vestnik Mosk. Univ., 1954, No. 12, pp. 15-26; 1956, No. 6, pp. 11-25.
2
In Investigations on the Theory of Structures, Moscow, GosstroyizdaU
1954, No. 6, pp. 455-468; 1957, No. 7, pp. 437-466.
,386 VARIATIONAL METHODS OF THEORY OF ELASTICITY [Ch. XI
face The construction of the correcting tensor for such bod-

l
.
dies can be carried out in a very general form through the use,
for instance, of formulas (9.18) which contain three stress func-
tions. Several problems of this kind were solved in the papers
of V. N. lonov, V. M. Ljubimov and E. R. Miroshnichenko 2 .
1
The author's paper. Prikl. matem. i
mekhan., 1953, Vol. 17, No. 4,
pp. 465-469.
2
V. N. Investigations on the Theory of Structures, Moscow,
lonov, in
Gosstrojizdat, pp. 413-436; E. R. Miroshnichenko, The Problem
1957,
7, No.
of a Cylinder Compressed Between Rigid Plates Without Sliding, Moscow,
Jviosk. lesotekhn. inst., 1957; V. M., Ljubimov. Prikl. matem. i mekhan., 1957,
Vol. 21, No. 4.
REFERENCES
1. A. E. H. Love,A Treatise on the Mathematical Theory of Elasticity, 4th ed.,.

Cambridge, Univ. Press, 1927.
2. L. S. Leibenzon, Theory of Elasticity, 2nd ed., Moscow, Gosteknizdat, 1947.
3. S. P. Timoshenko, Theory of Elasticity, New York, McGraw-Hill Book.

Co., 1934.
4. P. F. Papkovitch, Theory of Elasticity, Moscov. Oborongiz, 1939.
5. A. Foppl and L. Foppl, Drang und Zwang, 2 Aufl, Berlin, Bd. 1, 1924,
Bd. 2, 1928
6. N. I. Muskhelishvili, Some Basic Problems of the Mathematical Theory of

Elasticity, Groningen, P. Noordhoff, 1953.
7. E. Trefftz, Handbuch der Physik. Band VI. Mechanik der elastischen Kor-
per, Berlin, 1928.
8. P. Pfeiffer, Handbuch der Physik. Band VI. Mechanik der elastischen Kor~
per, Berlin, 1928.
9. G. V. Kolosoff, Application of a Complex Variable to the Theory of Elastic-

ity, Moscow, Objed. nauchno-tekhn. izd., 1935.
10. S. V. Serensen, Fundamentals of Applied Elasticity, Kiev, Gos. nauchno-
tekhn. izd., 1934.
11. A. N. Krylov, On Some Differential Equations of Mathematical Physics^

Leningrad, Izd. AN SSSR, 1932.
12. C. L. M. N. Navier, Resume des leqons sur I

'application de la mecanique,
3-me ed. avec des notes et des appendices par Barre de Saint-Venant^
Paris, Dunod, 1864.
13. E. Mathieu, Theorie de I'&lasticit6 des corps solides, Paris, Gauthier-Villars,.

1890.
NAME INDEX
Foppl, A., 256, 387

Foppl, L., 256, 387
Airy, G. B., 156 Fourier, J. B. J., Ill
B G
Beltrami, E., 139
Galerkin, B. G., 301, 338, 344
Bernoulli, J., 9, 126
Galilei, G., 9
Beizeno, C. B., 229
Germain, Sophie, 9, 322
Blokh, V. L, 269 229
Grammel, R.,
Bolle, L, 331
Green, G., 10
Bredt, R., 256
Boussinesq, J., 211, 278, 284, 290
H
Hertz, H., 300
Castigliano, A., 273, 364 Hooke, R., 9, 73
Cauchy, A. L., 10, 29, 48, 235
Clapeyron, B. P. E., 10, 143
Coker, E. G., 188
Coulomb, C. A., 9, 116, 238
lonov, V. N., 386
D
D'Alembert, J. le R., Ill, 313, 354
Dirichlet, P. G. L., 198, 235 10
Jasinsky, F. S.,
Jastrzbski, N. F., 10
Euler, L., 9, 87, 126, 193, 297
Kelvin, Lord, 10, 328

Kirchhoff, G., 143, 318, 328
Kochin, N. E., 312
Filonenko-Boroditch, M. M., 189, 271, Kolosoff, G. V., 10, 303, 387
380, 385, 386 Kononenko, E. S., 385
Filon, L. N. G., 181, 188-189, 303 Krutkov, Ju. A., 269
Flamant, 211 Krylov, A. N., 387
NAME INDEX 389
Lagrange, J. L., 9, 352 Papkovitch, P. R, 187-188,375, 387

Lamb, H., 328 Pfeiffer, P., 387
Lame, G.,10, 29, 79, 99, 209, 379 Poisson, S. D., 10, 74, 328
Laplace, P. S., 100, 250 Ponomarev, S D., 42
Leibenzon, L. S., 300, 316, 368, 373, Prandtl, L., 236, 250-252
387
Levy, M., 155, 199, 339
Ljubimov, V. M., 386
Love, A. E. H., 299, 303, 316, 387 Rayleigh, Lord, 10, 316
Reissner, E., 331
RibieYe, C., 181, 188
M Ritz, W., 358
Marcus, H., 347

Marietta, E., 9
Maslov, G. N., 187 Saint-Venant, B,, de, 10, 12, 55, 102,
Mathieu, E , 10, 387
118, 238, 259
Maxwell, J. C.,265 10,
Serensen, S. B., 387
Mitchell, J. H., 10, 139, 229
Miroshnichenko, E. R., 386
Mohr, O., 378
Morera, G., 265
N. 303, Tait, P. G., 328
Muskhelishvili, L, 10, 29, 56,
309, 387 Thomson, Sir'W. See Kelvin, Lord, 10,
328
Timoshenko, S. P., 10, 140, 186, 230,
N 260, 358, 387
Trefftz, E., 299, 386
Navier, C. L. M. H., 10, 333, 387
Netrebko, V. P., 385
Neumann, K. G., 234
Novozhilov, V. V,, 342
Vlasov, B. F., 331
Ostrogradsky, M, V., 10 Young, T., 74

SUBJECT INDEX
Biharmonic function, 162, 273, 282

Body forces, 17
Airy's function, 156, 162, 164, 270, 308 Bodies, anisotropic, 73
Amplitude of vibration, 105 homogeneous, 73
Analogy, Prandtl's, 252 isotropic, 73
Anisotropy, 73, see Bodies nonhomogeneous, 73
Areas, octahedral, 39 Boundary conditions, 24, 96
principal, 31 Boussinesq's problem, 290
Axisymmetrical three-dimensional Bredt's theorem, 256
stress, 273 Buckling of bar, 138
Bulk modulus, 81
Bars, bending of, 44

pure, 211, 163 Cantilever, bending of, 163
buckling of, 138 Castigliano's variational equation, 364
torsion of, 71, 116, 231 Cauchy's equations, 48, 97
Beams, bending of, built-in at ends, 176 Cauchy's quadric, 30, 39
of rectangular cross section, 131, Cauchy-Riemann conditions, 235, 238
152 Clapeyron's theorem, 143, 352
supported at ends, 171 Closeness of system of functions, 381
Beltrami-Mitchell equations, 138, 251, Coefficients of Lam6, 78, 94
278, 282 Compatibility, conditions of, 55, 97,
Bending of 151, 203
beam, built-in at ends, 176 Compendium of basic equations, 96
of rectangular cross section, 131, Completeness of system of functions,
152 381
supported at ends, 171 Components, of displacement, 19, 45
cantilever, 163 of rotation, 51
plate, 316 of strain, 49
pure, 327 of stress, 28, 84
prismatical bar, 44 Compression of wedge, 219
pure, 126, 163 Condition, Levy's, 155
rod, 44 Conditions, Cauchy-Riemann, 235, 238,
pure, 71 307
transverse, 258 Dirichlet's, 198
of circular cross section, 262 of compatibility (continuity) of
strip, 181, 189 strain components, 55, 97, 151, 203
wedge, 219 on surface, 24, 96
Bending moment, 323 Constants, elastic, 88
SUBJECT INDEX 391
Contact problem, 143, 189 Equilibrium, general equations of, 18,

Continuity, 12 265
Co-ordinates, cylindrical, 200 of elastic sphere, 284
polar, 200 Euler's substitution, 297
Coulomb's theory of torsion, 116, 238 Euler's theorem on homogeneous func-
Criterion, Lejeune-Dirichlet's, 355 tions, 87
Curved bar, pure bending of, 210 Expansion, volume, 312
Cylindrical co-ordinates, 201 Extension, 48
Field of function, 274

D'Alembert's principle, 354 Flexural rigidity of plate, 322
Density, 17 Forces, body, 10-12, 16
Diagrams, stress-strain, 73 external, 11
Dilatation, cubical, 64 internal, 10
waves of, 311 shearing, 324
irrotational, 313 Formula, Fourier's, 198
Dirichlet's conditions, 198 Green-Ostogradsky's, 25
Displacements, 19, 45 Formulas, Love's for displacements,.
virtual, principle of, 354 300
Distortion, waves of, 311 Fourier's series, 1 1 1
Divergence, of displacement vector, Function, analytic, 238

311 biharmonic, 162, 273, 278
force, 355
harmonic, 232, 273
stress, of Airy, 156, 162, 164, 270,
308
Effect of load on medium bounded by of Prandtl, 236, 250
plane, 290 torsion, 231, 250
Ellipsoid, stress, 36, 37 Functions, stress of Maxwell, 266
Elongations, 45 of Morera, 268
principal, 63
End plate of cylindrical reservoir, 347
Equation, biharmonic, 278
Laplace's, 232, 273 Germain's equation, 322
Levy's, 151, 155 Gradient of Prandtl's stress function,.
of transverse vibration of string, 111
254
Sophie Germain's, 322
Green-Ostrogradsky formula, 25
variational, of Castigliano, 364
Lagrange, 352
of
wave, 310 H
Equations, Beltrami-Mitchell, 139, 251,
278, 282 Harmonic function, 232, 273
canonical, 363 Hertz's problem of two bodies in
Cauchy's 48, 97 compression, 300
differential, of Navier, 21, 96, 110, Homogeneity, 73, 74
202, 270 Homogeneous deformation, 52
for plane problem in polar co-ordi- Homogeneous stress, 162
nates, 200 Hooke's law, 74, 85
geometrical, 97 generalised, 76, 94, 97, 149, 205
Lame's, 99, 141, 278, 282 in shear, 74, 117
of bending for plates, 319 Hypothesis, of continuous structure
of compatibility, 55, 97, 151, 203 of solids, 12
of equilibrium in statics, 18 of linear elements, Kirchhoff's, 31&
of torsion for plates, 316, 319, 326 of natural state, 87, 137
392 SUBJECT INDEX
I N
Identities, Saint-Venant's, 55 Navier's equations, 21, 96, 110, 202,
Initial stress, 86 270
Integrals, Mohr's 378 Neumann's problem, 234
Invariants of Neutral axis, 127
strain deviator, 65 Nonhomogeneity, 73
strain tensor, 65 Number, Poisson's 74
stress deviator, 40
stress tensor, 34
Isotropy, 73, 89 O
Oblique planes, 326
Octahedral areas, 39
Operator, Laplacian, 74, 203-204
Kirchhoff's hypothesis of linear ele-
ments, 318
Paraboloid, 241, 326, 327

Lagrange, principle of, 354 Period of vibration, 105
variational equation of 352 Photoelastic method, 156
Lame, coefficients of, 78, 94 Plane problem, equations for, in po-
ellipsoid of, 35-38 lar co-ordinates, 200
equations of, 99, 141, 277, 282 solution of, in terms of stresses, 158
Laplacian operator, 100, 203 Plane strain, 147, 151
potential, 274 Plane stress, 27
Law, Hooke's, 74, 85 Plane stress, generalised, 151
of reciprocity (pairing, conjunga- Planes, oblique, 326
tion) principal, 128
of shearing stresses, 21, 42 Plates, bent by couples, 327
Least work, principle of, 375 boundary conditions for, 328-331
Lejeune-Dirichlet's criterion, 355 flexural rigidity of, 322
Level lines of function, 242 of special forms:
Levy's condition (equation), 155 circular, 344
Limit of proportionality, 74 331
elliptic,
Love waves, 316 rectangular, 328, 333
theory of moderately thick, 318
M Poisson's number, 74
Poisson's ratio, 74
Maxwell's stress functions, 266 Polar co-ordinates, 200
Mean hydrostatic stress, 42 Potential, displacement, 53
Membrane, 252 elastic forces having, 83
Method of sections, 11 Laplacian, 274
photoelastic, 156 Potential elastic energy, 145, 356
Ritz-Timoshenko, 358 minimum of, 355, 373
Minimum elastic energy, theorems of, Prandtl's analogy, 252
355, 373 Prandtl's function, 236, 250
Modulus, of rigidity, 75 Principal areas, 31
in shear, 74 Principal elongations, 63
in tension, 74 Principal planes, 128
of volume expansion, 81 Principal strain, 66
Young's, 74 Principal stress, 31
Mohr's integrals, 378 Principal of D'Alembert, 354
Molecular theory of structure of bo of Lagrange, 354
dies, 10 . of least work, 375
Morera's stress functions, 268 of Saint- Venant, J19
SUBJECT INDEX 393
of superposition, 72, 137 Shearing force, 324

of virtual displacements, 354 Shearing strain, 44
Prism, rectangular, elastic, 379 Shearing stress, 14, 235
stretched of its own weight, 132 Simple radial distribution of stresses,
Problem, Boussinesq's, 290 212
contact, 143, 189 Slabs, thin, 318
Dirichlet's, of finding a harmonic Solenoidal vector (curl), 311
function, 235, 294 Solution, uniqueness of, 136, 143
Flamant-Boussinesq, 211 Sphere, equilibrium of, 284
Hertz's, of two bodies in compres- Stability of bar, see Buckling of bar,
sion, 300 138
Lame's, of circular tube in uniform State of stress, 13
compression, 206 Statics, equations of equilibrium, 18
mixed, of elasticity, 143 Stokes's theorem, 256
Neumann's, 234 Strain, 9, 45
Propagation of waves, 106-108, 312 compatibility of, 53, 55, 151, 203
Proportional limit, 74 components of, 45, 49
Pure bending, of curved rods, 210 dilatational, 64
of plates, 327 extensional, 48
of prismatical bars, 126, 163 finite, 67, 70
homogeneous, 52
identical relations between compo-
nents of (see Compatibility of,
above), 45
Quadric, Cauchy's, 30, 39 in multiply connected body, 58
invariants of, 64, 65
plane, 147
principal, 67
pure, 53
Radial stress, distribution of, 212 shearing, 48
Ratio, Poisson's, 74 Strain deviator, 65
Rayleigh waves, 316 invariants of, 65
Retaining wall, 177 Strain surface, 62
Rigid body displacement, 52 Strain tensor, 62
Rigidity, flexural, of plate, 322 invariants of, 64, 65
modulus of, 74 Stress, 11, 12
torsional, 237 11
average,
Ring, closed, 230 elements of, 28
Ritz-Timoshenko method, 358
homogeneous, 162
Rods, bending of, 44 86
initial,
pure, 70 invariants 29
of,
transverse, 258
mean hydrostatic, 42
cylindrical, 16, 262
of large curvature, 209 normal, 15
Rotation, components 51 octahedral, 39, 79
of,
plane (see Plane stress), 27, 157
principal, 31
shearing, 15, 235
total, 12
Saint-Venant, equations of, 55, 97, 151, Stress circulation, 258
203 Stress deviator, 39
identities of, 55 invariants of, 40
principle of, 119 Stress ellipsoid, 35-38
Sections, method of, 11 Stress functions, 151, 156, 161, 164
Self-stressed state, 376 236, 250, 266, 268, 271, 308
Series, Fourier's, 111 Stress-strain relations, 84
Shear, 44 Stress surface, 30, 38
394 SUBJECT INDEX
Stress, tensor, 28 U
invariants of, 34
Uniqueness of solution of
String, equation of transverse vibra- elasticity
tion equations, 136, 143
Unit elongation, 47
of, 111
Strip, in bending, 189
rectangular, 181
on absolutely rigid foundation, 189,
195 Variation of stress, 365
Superposition principle of, 72, 137 Variational equation, Castigliano's, 364
Symmetric matrix, 363 Lagrange's, 352
Velocity, of wave propagation, 106-
108
Vibrations, elastic, 19
harmonic, 104
Tensile test diagram, 73 longitudinal, of bar, 109
Tensor, basic, 380 uniform, 104
transverse of string, 111
correcting, 380
finite-strain, 67, 70 uniform, 104
of relative displacements, 56 Virtual displacements, principle of, 354
spherical, 37
strain, 62 W
stress, 29
Theorem, Bredt's, 256 Wave equation, 311
Clapeyron's, 143, 352 Wave length, 105
Euler's on homogeneous functions, Waves, equivoluminal, 313
87 Love, 316
Stokes's, 256 of dilatation, 312
Theorems of minimum elastic energy, irrotational, 313
355, 373 of distortion, 311
Torque, see Twisting moment, 116, Rayleigh, 316
236, 325 reflected, 316
Torsion, Coulomb's theory of, 116, 238 spherical, 314
of plates, 318, 326 standing, 316
of prismatical bars, 231, 368 velocity of, 106-108
of particular forms of section: Wedge, bending of, 219
circular, 71, 116 compression of, 219
elliptic, 239 loaded at vertex, 217
rectangular, 245 Work, done by elastic forces, 82
triangular, 243 by external loads, 143
of rod in form of body of revolu- by surface tractions, 145
tion, 273 least, principle of, 375
Torsion function, 232, 250
Torsional rigidity, 238
Trajectory of equal stresses, 214
Twisting moment, 116, 237, 325 Young's modulus, 74
To the Reader
Peace Publishers would be glad to have your opinion

of the translation and the design of this book.
Please send your suggestions to 2, Pervy Rizhsky
Pereulok, Moscow, U. S. S. R.
Printed in the Union of Soviet Socialist Republics

Filonenko-Borodich - Theory of Elasticty

Uploaded by

Copyright:

Available Formats

Filonenko-Borodich - Theory of Elasticty

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Filonenko-Borodich - Theory of Elasticty

Uploaded by

Copyright:

Available Formats

M.

Translated from the Rust

PEACE PUBLISHERS MOSCOW

Chapter I. Theory of Stress 13

1. State of Stress in a Body 13

Invariants of the Stress Tensor. Lame's Ellipsoid 29

Chapter II. Geometrical Theory of Strain 45

10. Displacement Components and Strain Components, and Rela-

Chapter III. Generalised Hopke's Law 72

*21. Stress-strain Relations; Hypothesis of the Natural State of

Chapter IV. Solution of the Elasticity Problem in Terms of Displacements 95

24. Compendium of Basic Equations of the Theory of Elasticity .

Chapter V. Solution of the Elasticity Problem in Terms of Stresses . . . 115

29. The Simplest Problems 115

Chapter VI. Plane Problem in Cartesian Co-ordinates 147

38. Plane Strain 147

Chapter VII. Plane Problem in Polar Co-ordinates 200

47. General Equations of the Plane Problem in Polar Co-ordinates 200

49. Effect of a Concentrated Force (Flamant-Boussinesq Problem) 211

Chapter VIII. Torsion of Prismatical Bars and Bending 231

52. Torsion of Prismatical Bars 231

56. General Solution of Differential Equations of Equilibrium in

Terms of Stresses. Stress Functions 265

Chapter X. Bending of a Piate 317

68. General 317

Name Index 388

Subject Index 390

The theoryof elasticity developed into an important branch

theories indicated above. In Russia an analogous treatise by

Clapeyron, B. de Saint-Venant, G. Green, J. C. Maxwell, Lord

the lines of forces of interaction between

moved, for example B, the system of forces of interaction applied

s called the average stress of internal forces in the body across

mathematical analysis, we shall contract the contour of the ele-

it has a vector character; it is represented, therefore, as a vector p

applied at the point M

with later, in Chapter I.

hypothesis of the continuous structure of solids. It must be borne

1. STATE OF STRESS IN A BODY

Beginning the study of the state of stress in a body, let us

Imagine (Fig. 2) a solid cut by a plane into two parts A and B.

tion on it); we shall denote the projection of the stress p on

since stresses X v Y v Z v express the action of the part B on the

part A, while stresses X~ v Y- v Z- v express the action of the part A

on the part B\ these actions are equal in magnitude, but opposite

The projection Y y is normal to the area and is called the normal

Fig. 5a, b shows the notation of stresses on the areas normal to

following system of stresses on the areas parallel to the co-

One can easily see that the stresses

1. Fig. 6 represents the cross sec-

Some authors now use the following

on the areas normal to the axes OX, OY and OZ are denoted,

the shearing stresses on the same areas are denoted by

2. pIFFERENTIAL EQUATIONS OF EQUILIBRIUM

Imagine that we isolate in a solid an infinitesimal element in

components. Altogether there will be 6 X 3

unit of volume and having the dimension 2 4

If .the area abed of the element (Fig. 9) is acted upon

stress Xx + ~jjj-dx since, in passing to the area a'b'c'd', we