Appl. Math. Mech. -Engl. Ed.

, 33(2), 223232 (2012)

DOI 10.1007/s10483-012-1545-9
c Shanghai University and Springer-Verlag
Berlin Heidelberg 2012
Applied Mathematics
and Mechanics
(English Edition)
Eigenfunction expansion method of upper triangular operator matrix
and application to two-dimensional elasticity problems based on
stress formulation

EBURILITU (), ALATANCANG ()

(School of Mathematical Sciences, Inner Mongolia University, Hohhot 010021, P. R. China)
Abstract This paper studies the eigenfunction expansion method to solve the two-
dimensional (2D) elasticity problems based on the stress formulation. The fundamental
system of partial dierential equations of the 2D problems is rewritten as an upper tri-
angular dierential system based on the known results, and then the associated upper
triangular operator matrix matrix is obtained. By further research, the two simpler com-
plete orthogonal systems of eigenfunctions in some space are obtained, which belong to
the two block operators arising in the operator matrix. Then, a more simple and conve-
nient general solution to the 2D problem is given by the eigenfunction expansion method.
Furthermore, the boundary conditions for the 2D problem, which can be solved by this
method, are indicated. Finally, the validity of the obtained results is veried by a specic
example.
Key words eigenfunction expansion method, two-dimensional (2D) elasticity problem,
upper triangular operator matrix, general solution
Chinese Library Classication O175.3
2010 Mathematics Subject Classication 47A70, 47A75, 47B99
1 Introduction
In the 1990s, based on the analogy between the theory of structural mechanics and opti-
mal control, the innite-dimensional Hamiltonian system was applied to the structural anal-
ysis. Then, a new systematic methodology for the theory of elasticity was established by
Zhong
[12]
. Subsequently, the method of separation of variables based on the Hamiltonian sys-
tem is widely developed, and a large number of non-self-adjoint problems in applied mechanics
are considered
[38]
. This method extends the traditional separation of variables. However,
the mathematical foundation, i.e., the symplectic eigenfunction expansion theorem, is partly
solved
[911]
, and there are still many unsolved problems. For example, the completeness of
the eigenfunction system of the innite-dimensional Hamiltonian operator appearing in the
two-dimensional (2D) elasticity problems based on the stress formulation has not been proved.
Therefore, the theoretical basis has not been established to solve the 2D problems with the
Received Feb. 25, 2011 / Revised Dec. 1, 2011
Project supported by the Specialized Research Fund for the Doctoral Program of Higher Education
of China (No. 20070126002) and the National Natural Science Foundation of China (No. 10962004)
Corresponding author ALATANCANG, Professor, Ph. D., E-mail: alatanca@imu.edu.cn
224 EBURILITU and ALATANCANG
method of separation of variables based on the Hamiltonian system. Moreover, given a mechan-
ics equation, in general, we can look for the various dierential systems
[12]
. The Hamiltonian
system is only one class of them. In all dierential systems, we believe that the upper triangular
dierential systems have certain advantages in solving which problems. Now, many researchers
are studying the upper triangular operator matrices, which has become a research hot point of
the operator theory.
In Ref. [13], the fundamental system of partial dierential equations of the 2D problems is
rewritten as an upper triangular dierential system. For the associated operator matrix, the
existence of the two the complete orthogonal systems of eigenfunctions in some space are ob-
tained, which belong to the two block operators arising in the operator matrix. The general
solution to the 2D problem is given by the eigenfunction expansion method. However, the
analytic expression formula of the general solution is very tedious and complicate. Therefore,
it is inconvenient to be applied.
Through an in-depth study of the 2D problem, in a new coordinate system, this paper
presents a more simple and practical general solution. Furthermore, the boundary conditions
for the 2D problem, which can be solved by the method, are indicated.
2 Properties of operator matrix
Let = {(x, y)|0 x h, 0 y l}, for some constants h and l, be a rectangular region in
the plane. We consider the following system of fundamental partial dierential equations based
on the stress formulation:

x
x
+

xy
y
+f
x
= 0,

y
y
+

yx
x
+f
y
= 0,
_

2
x
2
+

2
y
2
_
(
x
+
y
) + (1 +)
_
f
x
x
+
f
y
y
_
= 0,
(1)
where f
x
and f
y
are the body forces in the region . The boundary conditions for x = 0 and
x = h are

x
= 0, v = 0 for x = 0, h. (2)
Introduce the state functions
[13]

p =

x
y
+

y
y
,
q =

x
x
+

y
x
.
(3)
Then, we have

xy

y
p
q

0

x
0 1

x
0 0 0
0 0 0

x
0 0

x
0

xy

y
p
q

f
x
f
y
(1 +)
_
f
x
x
+
f
y
y
_
0

, (4)
Eigenfunction expansion method of upper triangular operator matrix and application 225
which is an upper triangular dierential system derived from Eq. (1), and the corresponding
upper triangular operator matrix
H =

0
d
dx
0 1

d
dx
0 0 0
0 0 0
d
dx
0 0
d
dx
0

. (5)
By the stress-displacement relationship
v
y
=
1
E
(
y

x
),
from the boundary conditions in Eq. (2), we have

y
= 0, p = 0 for x = 0, h. (6)
Here, we introduce the Hilbert space X = L
2
[0, h] and set W = XXXX
[10]
. By Eq. (6),
the domain of the operator H can be described as
D(H) =

2
(0) =
2
(h) = 0,
3
(0) =
3
(h) = 0,

i
is absolutely continuous,

i
X, i = 1, 2, 3, 4

. (7)
Write the operator H as the following block operator representation:
H =
_
A B
C A

_
, (8)
where
A =

0
d
dx

d
dx
0

, B =
_
0 1
0 0
_
, C =
_
0 0
0 0
_
, A

0
d
dx
d
dx
0

.
By the domain (7) of the operator H, we set
D(A) =
__

2
_
X X

2
(0) =
2
(h) = 0,
i
is absolutely continuous,

i
X, i = 1, 2.
_
, (9)
D(A

) =
__

4
_
X X

3
(0) =
3
(h) = 0,
i
is absolutely continuous,

i
X, i = 3, 4.
_
. (10)
Here, B and C are the bounded linear operators on the whole space X X. By the block
operators A and A

, we attempt to solve Eq. (1) with the boundary conditions in Eq. (2).
Therefore, we need the following completeness theorems.
Theorem 1 The block operator A has an orthogonal eigenfunction system in the Hilbert
Space X X. In particular, the eigenvalues and the associated eigenfunctions (away from a
226 EBURILITU and ALATANCANG
multiple of a constant ) of A are given by

U
k
=

cos
_
k
h
x
_
sin
_
k
h
x
_

k
=
k
h
,
(11)
where k = 0, 1, 2, .
Proof The eigenvalues and the associated eigenfunctions of A shown in Eq. (11) are
demonstrated by straightforward calculations. By calculation, we have
(U
k
, U
j
) =
_
h, k = j,
0, k = j,
where k, j = 0, 1, 2, . This shows that the eigenfunction system {U
k
|k = 0, 1, 2, }
possesses the orthogonality. Therefore, Theorem 1 is proved.
Theorem 2 The eigenfunction system {U
k
|k = 0, 1, 2, } dened in Eq. (11) of the
block operator A is complete in the sense of the Cauchy principal value in XX, i.e., for each
_
f
g
_
X X, there exist constant sequences {c
k
}
+
k=0
and {c
k
}
+
k=1
such that
_
f
g
_
= c
0
U
0
+
+

k=1
(c
k
U
k
+c
k
U
k
). (12)
Proof For k = 0, 1, 2, , set
c
k
=
1
h
_
h
0
_
f(t) cos
_
k
h
t
_
+g(t) sin
_
k
h
t
__
dt.
Then, substituting them into the expansion (12) yields
c
0
U
0
+
+

k=1
(c
k
U
k
+c
k
U
k
) =

1
h
_
h
0
f(t)dt +
+

k=1
2
h
_
h
0
f(t) cos
_
k
h
t
_
dt cos
_
k
h
x
_
+

k=1
2
h
_
h
0
g(t) sin
_
k
h
t
_
dt sin
_
k
h
x
_

. (13)
We observe that the sequences of the functions
1, cos
_

h
x
_
, cos
_
2
h
x
_
, , cos
_
k
h
x
_
, , (14)
sin
_

h
x
_
, sin
_
2
h
x
_
, sin
_
3
h
x
_
, , sin
_
k
h
x
_
, (15)
are the eigenfunction systems of the Sturm-Liouville problems

d
2
z
dx
2
+z = 0,
z

(0) = z

(h) = 0,

d
2
z
dx
2
+z = 0,
z(0) = z(h) = 0,
Eigenfunction expansion method of upper triangular operator matrix and application 227
respectively. Thus, the two components of the vector-valued function on the right-hand side
of Eq. (13) are the Fourier series expansions in terms of the corresponding orthogonal systems
associated with f and g in X = L
2
[0, h], respectively. Therefore, the expansion (12) holds.
Theorem 3 The block operator A

has an orthogonal eigenfunction system in the Hilbert

space X X. In particular, the eigenvalues and the associated eigenfunctions (away from a
multiple of a constant ) of A

are given by

U
k
=

sin
_
k
h
x
_
cos
_
k
h
x
_

k
=
k
h
,
(16)
where k = 0, 1, 2, .
Proof The eigenvalues and the associated eigenfunctions of A

shown in Eq. (16) are

demonstrated by straightforward calculations. By calculation, we have
(

U
k
,

U
j
) =
_
h, k = j,
0, k = j,
where k, j = 0, 1, 2, . This shows that the eigenfunction system {

U
k
|k = 0, 1, 2, }
possesses the orthogonality. Therefore, Theorem 3 is proved.
Theorem 4 The eigenfunction system {

U
k
|k = 0, 1, 2, } dened in Eq. (16) of the
block operator A

is complete in the sense of the Cauchy principal value in X X, i.e., for

each
_
f
g
_
X X, there exist constant sequences {c
k
}
+
k=0
and {c
k
}
+
k=1
such that
_
f
g
_
= c
0

U
0
+
+

k=1
(c
k

U
k
+c
k

U
k
). (17)
Proof For k = 0, 1, 2, , set
c
k
=
1
h
_
h
0
_
f(t) sin
_
k
h
t
_
+g(t) cos
_
k
h
t
__
dt.
Then, substituting them into the expansion (17) yields
c
0

U
0
+
+

k=1
(c
k

U
k
+c
k

U
k
) =

k=1
2
h
_
h
0
f(t) sin
_
k
h
t
_
dt sin
_
k
h
x
_
1
h
_
h
0
g(t)dt +
+

k=1
2
h
_
h
0
g(t) cos
_
k
h
t
_
dt cos
_
k
h
x
_

. (18)
According to the proof of Theorem 2, the two components of the vector-valued function on
the right-hand side of Eq. (18) are the Fourier series expansions in terms of the corresponding
orthogonal systems associated with f and g in X = L
2
[0, h], respectively. Therefore, the
expansion (17) holds.
Remark 1 Based on the domains (9) and (10), the above four theorems are obtained.
Similarly, based on the domain (9) and the domain
D(A

) =
_
_

4
_
X X

4
(0) =
4
(h) = 0,
i
is absolutely continuous,

i
X, i = 3, 4.
_
,
228 EBURILITU and ALATANCANG
we can also obtain similar results. Therefore, the boundary conditions in Eq. (2) can be replaced
by some other boundary conditions.
3 General solution
This section is devoted to the general solution to the 2D elasticity problems (1) and (2). Let
U =
_

xy

y
_
,

U =
_
p
q
_
, F =
_
q f
x
f
y
_
,

F =

(1 +)
_
f
x
x
+
f
y
y
_
0

.
Then, Eq. (4) can be written as
U
y
= AU +F, (19)

U
y
= A

U +

F. (20)
To obtain the general solution to the problems (1) and (2), we use the eigenfunction expansion
method.
By Theorems 3 and 4, we apply the separation of variables to the problem (20). First, we
consider the corresponding homogeneous case, i.e.,

U
y
= A

U. (21)
We look for the solutions of the form

U(x, y) =

T(y)

V (x). Then, substituting this expression

into Eq. (21) yields

(y)

T(y)
=
A

V (x)

V (x)
.
It is seen that both sides of the above equation are constant and equal to . Therefore, the
homogeneous equation (21) reduces to the following form:

(y) =

T(y), (22)
A

V (x) =

V (x). (23)
We notice that the eigenvalue problem of the operator A

can be obtained by separating

variables for the system (21). Theorems 3 and 4 show that =
k
=
k
h
(k = 0, 1, 2, ),
and

V (x) is a multiple of the eigenfunction

U
k
(x). Hence,

T(y) is a multiple of e

k
y
. By
superposing these solutions, the homogeneous equation (21) has the solution of the following
form:

U(x, y) =

T
0
(y)

U
0
(x) +
+

k=1
(

T
k
(y)

U
k
(x) +

T
k
(y)

U
k
(x)). (24)
Now, we apply the separation of variables to the inhomogeneous problem (20). From Theorems
3 and 4, we see that the inhomogeneous equation (20) still has the solution of the form (24),
and the inhomogeneous item

F in Eq. (20) has the following Fourier expansion:

F(x, y) =

F
0
(y)

U
0
(x) +
+

k=1
(

F
k
(y)

U
k
(x) +

F
k
(y)

U
k
(x)), (25)
Eigenfunction expansion method of upper triangular operator matrix and application 229
where

F
k
(y) =
1
(

U
k
(x),

U
k
(x))
(

F(x, y),

U
k
(x))
=
1 +
h
_
h
0
_
f
x
x
+
f
y
y
_
sin
_
k
h
x
_
dx (k = 0, 1, 2, ). (26)
Substituting Eqs. (24) and (25) into Eq. (20), we have

k
(y) =
k

T
k
(y) +

F
k
(y) (k = 0, 1, 2, ),
and hence

T
k
(y) =

d
k
e

k
y
+
_
y
0

F
k
(t)e

k
(yt)
dt (k = 0, 1, 2, ) (27)
with

d
k
being unknown constants to be determined. Therefore, the general solution to Eq. (20)
is as follows:

U = (x, y) =
_
p
q
_
, (28)
where
p =
+

k=1
_

d
k
e

k
y

d
k
e

k
y
+
_
y
0
(

F
k
(t)e

k
(yt)

F
k
(t)e

k
(yt)
)dt
_
sin
_
k
h
x
_
,
q =

d
0
+
+

k=1
_

d
k
e

k
y
+

d
k
e

k
y
+
_
y
0
(

F
k
(t)e

k
(yt)
+

F
k
(t)e

k
(yt)
)dt
_
cos
_
k
h
x
_
,

F
k
(t)e

k
(yt)

F
k
(t)e

k
(yt)
=
1 +
h
_
h
0
_
f
x
(x, t)
x
+
f
y
(x, t)
y
_
sin
_
k
h
x
_
(e

k
(yt)
e

k
(yt)
)dx.
Next, we substitute the component q of

U given in Eq. (28) into Eq. (19). Then, we obtain
an inhomogeneous problem. From Theorems 1 and 2, we see that Eq. (19) has the solution as
follows:
U(x, y) = T
0
(y)U
0
(x) +
+

k=1
(T
k
(y)U
k
(x) +T
k
(y)U
k
(x)), (29)
and the inhomogeneous item F has the following Fourier expansion:
F(x, y) = F
0
(y)U
0
(x) +
+

k=1
(F
k
(y)U
k
(x) +F
k
(y)U
k
(x)), (30)
where
F
k
(y) =
1
h
_
h
0
_
f
x
cos
_
k
h
x
_
+f
y
sin
_
k
h
x
__
dx
1
2
_

d
k
e

k
y
+

d
k
e

k
y
+
_
y
0
(

F
k
(t)e

k
(yt)
+

F
k
(t)e

k
(yt)
)dt
_
(k = 0, 1, 2, ). (31)
With the same method as used to solve Eq. (20), by direct calculations, we have
T
k
(y) = d
k
e

k
y
+
_
y
0
F
k
()e

k
(y)
d (k = 0, 1, 2, ), (32)
230 EBURILITU and ALATANCANG
where d
k
are unknown constants to be determined. Therefore, the general solution to Eq. (19)
is given by
U(x, y) =

d
0


d
0
y
1
h
_
y
0
_
h
0
f
x
(x, )dxd
0

+
+

k=1
_
E
k
E
k
_
, (33)
where
E
k
=
_
d
k
e

k
y
+d
k
e

k
y
+
_
y
0
(F
k
()e

k
(y)
+F
k
()e

k
(y)
)d
_
cos
_
k
h
x
_
, (34)
E
k
=
_
d
k
e

k
y
d
k
e

k
y
+
_
y
0
(F
k
()e

k
(y)
F
k
()e

k
(y)
)d
_
sin
_
k
h
x
_
. (35)
Remark 2 From the general solution (33), on the rectangular region , we have

xy
(x, 0) = d
0
+
+

k=1
(d
k
+d
k
) cos
_
k
h
x
_
, (36)

y
(x, 0) =
+

k=1
(d
k
d
k
) sin
_
k
h
x
_
. (37)
Moreover, we observe that the sequences of the functions (14) and (15) are the complete system
of orthogonal functions in L
2
[0, h], respectively. Therefore, under any boundary conditions

xy
(x, 0) =
0
(x) L
2
[0, h] and
y
(x, 0) =
0
(x) L
2
[0, h], the constants d
k
can be determined.
Furthermore, if f
x
(x, y) and f
y
(x, y) are given, the constants

d
k
can be determined under any
boundary conditions
xy
(x, l) =
l
(x) L
2
[0, h] and
y
(x, l) =
l
(x) L
2
[0, h].
Remark 3 The two components
xy
(x, y) and
y
(x, y) of the general solution U(x, y) are
the corresponding shear stress and the normal stress in Eq. (1). Furthermore, the normal stress

x
(x, y) can be determined by Eqs. (1) and (2).
4 Example
Let us consider a rectangular beam supported at the ends and subjected along the upper and
lower edges to the action of continuously distributed vertical forces of the intensity sin
_
4
h
x
_
,
respectively
[14]
. The Cartesian coordinate system is established with the origin at the corner
of the beam such that 0 x h and 0 y l, as depicted in Fig. 1.
Fig. 1 Geometry and coordinate system of beam
The boundary conditions are

y
= sin
_
4
h
x
_
,
xy
= 0 for y = 0, l.
Eigenfunction expansion method of upper triangular operator matrix and application 231
If f
x
(x, y) and f
y
(x, y) are given, based on the above conditions, the constants d
k
and

d
k
can be determined. Here, we take f
x
(x, y) = 0 and f
y
(x, y) = 0. Then, the outline of the
computing procedure is as follows. We expand the given boundary value functions
xy
(x, 0) = 0
and
y
(x, 0) = sin
_
4
h
x
_
into the Fourier series based on the orthogonal function systems (14)
and (15), respectively. Then, we have

xy
(x, 0) = 0 +
+

k=1
0 cos
_
k
h
x
_
, (38)

y
(x, 0) = sin
_
4
h
x
_
+
+

k=1,
k=4
0 sin
_
k
h
x
_
. (39)
Comparing the corresponding coecients of Eqs. (36)(39), the constants d
k
can be determined.
Similarly, we expand the given boundary value functions
xy
(x, l) = 0 and
y
(x, l) = sin
_
4
h
x
_
into the Fourier series using the orthogonal function systems (14) and (15), respectively. Com-
paring the coecients of the obtained two Fourier series and the coecients of the solution
U(x, l) (i.e., set y = l in the general solution (33)), the constants

d
k
can also be determined.
Then, we obtain

xy
(x, y) =E
4
=
2
4l +hsinh
4l
h
exp
4(l +y)
h
_
(l y) exp
4l
h
+y exp
8l
h
y exp
8y
h
+ (l +y) exp
4(l + 2y)
h
_
cos
_
4
h
x
_
, (40)

y
(x, y) =E
4
=
1
2(4l +hsinh
4l
h
)
exp
4(l +y)
h
_
(h + 4(l y))
exp
4(l + 2y)
h
+ (h 4y) exp
8y
h
(h + 4y) exp
8l
h
+ (h + 4(l +y)) exp
4l
h
_
sin
_
4
h
x
_
. (41)
In the present example, we take h = 10 and l = 1. In the example of [14], we take A = B = 1,
c =
1
2
, and l = 10. Then, some computed results for
xy
(x, y) and
y
(x, y) are presented in
Table 1.
Table 1 Some computed results
(x, y) (5.00, 0.25) (5.00, 0.55) (5.00, 0.75) (5.00, 0.81) (5.00, 0.94)
xy (present) 0.014 255 1 0.003 727 64 0.014 255 1 0.014 586 2 0.007 704 85
(x, y) (5.00, 0.25) (5.00, 0.05) (5.00, 0.25) (5.00, 0.31) (5.00, 0.44)
xy Ref. [14] 0.014 255 1 0.003 727 64 0.014 255 1 0.014 586 2 0.007 704 85
(x, y) (1.25, 0.25) (1.25, 0.55) (1.25, 0.75) (1.25, 0.81) (1.25, 0.94)
y (present) 0.996 608 0.994 127 0.996 608 0.997 707 0.999 689
(x, y) (1.25, 0.25) (1.25, 0.05) (1.25, 0.25) (1.25, 0.31) (1.25, 0.44)
y Ref. [14] 0.996 608 0.994 127 0.996 608 0.997 707 0.999 689
5 Conclusions
In the present paper, for the 2D elasticity problems based on the stress formulation (1), we
obtain the following results:
232 EBURILITU and ALATANCANG
(i) Compared with the results in [13], the two simpler complete orthogonal systems of eigen-
functions are obtained, which belong to the two block operators arising in the operator matrix
H. Furthermore, a more simple and convenient general solution to the 2D problem is given.
(ii) Under any boundary conditions
xy
(x, 0),
y
(x, 0),
xy
(x, l), and
y
(x, l) L
2
[0, h], the
coecients of the general solution (33) can be determined by the given specic f
x
(x, y) and
f
y
(x, y).
(iii) According to Remark 1, the boundary conditions in Eq. (2) can be replaced by other
boundary conditions.
References
[1] Zhong, W. X. Method of separation of variables and Hamiltonian system (in Chinese). Computa-
tional Structural Mechanics and Applications, 8(3), 229240 (1991)
[2] Zhong, W. X. A New Systematic Methodology for Theory of Elasticity (in Chinese), Dalian Uni-
versity of Technology Press, Dalian (1995)
[3] Yao, W. A. and Sui, Y. F. Symplectic solution system for Reissner plate bending. Applied Math-
ematics and Mechanics (English Edition), 25(2), 178185 (2004) DOI 10.1007/BF02437319
[4] Yao, Z., Zhang, H. W., Wang, J. B., and Zhong, W. X. Symplectic analysis for phonon dispersion
of carbon nanotubes based on inter-belt model (in Chinese). Chinese Journal of Solid Mechanics,
29(1), 1322 (2008)
[5] Xu, X. S., Wang, G. P., and Sun, F. M. Analytical and numerical methods of symplectic system for
Stokes ow in two-dimensional rectangular domain. Applied Mathematics and Mechanics (English
Edition), 29(6), 705714 (2008) DOI 10.1007/s10483-008-0602-3
[6] Zhou, J. F., Zhuo, J. S., and Li, Q. D. Variable separation method based on Hamilton system (in
Chinese). Journal of Hohai University, 28(6), 2731 (2000)
[7] Luo, J. H., Li, Q. S., and Liu, G. D. A biorthogonality relationship for three-dimensional couple
stress problem. Science in China, Series G: Physics, Mechanics & Astronomy, 52(2), 270276
(2009)
[8] Lim, C. W. and Xu, X. S. Symplectic elasticity: theory and applications. Applied Mechanics
Reviews, 63(5), 050802 (2010)
[9] Zhang, H. Q., Alatancang, and Zhong, W. X. The Hamiltonian system and completeness of
symplectic orthogonal system. Applied Mathematics and Mechanics (English Edition), 18(3), 237
242 (1997) DOI 10.1007/BF02453366
[10] Eburilitu and Alatancang. On feasibility of variable separation method based on Hamiltonian
system for a class of plate bending equations. Communications in Theoretical Physics, 53(3),
569574 (2010)
[11] Alatancang and Wu, D. Y. Completeness in the sense of Cauchy principal value of the eigenfunc-
tion systems of innite dimensional Hamiltonian operator. Science in China, Series A: Mathemat-
ics, 52(1), 173180 (2009)
[12] Alatancang, Zhang, H. Q., and Zhong, W. X. Pseudo-division algorithm for matrix multi-variable
polynomial and its application. Applied Mathematics and Mechanics (English Edition), 21(7),
733740 (2000) DOI 10.1007/BF02428369
[13] Huang, J. J., Alatancang, and Wang, H. Eigenfunction expansion method and its application
to two-dimensional elasticity problems based on stress formulation. Applied Mathematics and
Mechanics (English Edition), 31(8), 10391048 (2010) DOI 10.1007/s10483-010-1340-x
[14] Timoshenko, S. P. and Goodier, J. N. Theory of Elasticity, 3rd ed., McGraw-Hill, New York,
5356 (1970)