You are on page 1of 131

Applied Elasticity

Marijan Dravinski

September 2000
Applied Elasticity (ME 509)
by
M. Dravinski

1
Chapter 1

Tensors and Dyadics

1.1 Cartesian Tensors


If we have two sets of right-handed Cartesian coordinate systems {xi } with the
basis {ei } and {x0i } with the basis {ei }, i = 1, ..., 3, with the same origin, the
coordinates of a point P can be expressed in both coordinate systems (Fig.1.1
). Then,

x = x1 e1 + x2 e2 + x3 e3 = xi ei (1.1)
x = x01 e01 + x02 e02 + x03 e03 = x0i e0i (1.2)

where summation over repeated indices is understood. Thus, we have

u±v = (ui ± vi )ei (1.3)


u·v = ui vi (1.4)
 
e1 e2 e3
u×v =  u1 u2 u3  (1.5)
v1 v2 v3
= (u2 v3 − u3 v2 )e1 + (u3 v1 − u1 v3 )e2 + (u1 v2 − u2 v1 )e3 (1.6)

where

ei · ej = δ ij (1.7)
ei × ei = 0, i = 1, 2, 3 (1.8)
e2 × e3 = −e3 × e2 = e1 (1.9)
e3 × e1 = −e1 × e3 = e2 (1.10)
e1 × e2 = −e2 × e1 = e1 (1.11)

Then from Eqns. 1.1 and 1.2 it follows that

x01 = e01 · e1 x1 + e01 · e2 x2 + e01 · e3 x3 (1.12)

2
X3
X’3

X’2

X2

X1
X’1

Figure 1.1: Two Cartesian coordinate syatems.

or
x01 = 11 x1 + 12 x2 + 13 x3 (1.13)
with

11 = e01 · e1 = cos ]{x01 , x1 } (1.14)


12 = e01 · e2 = cos ]{x01 , x2 } (1.15)
13 = e01 · e3 = cos ]{x01 , x3 } (1.16)

Similarly, we have that

x02 = 2k xk (1.17)
x03 = 3k xk (1.18)

where
ij = e0i · ej = cos ]{x0i , xj } (1.19)
Or in matrix notation

x0 = Lx; L=[ ij ]; i, j = 1, 2, 3 (1.20)

The last result can be written in index notation as

x0i = ij xj ; i, j = 1, 2, 3 (1.21)

Therefore, Eq. 1.20 represents transformation of vector x form coordinate sys-


tem {xi } to coordinate system {x0i } with matrix L denoting direction cosine
matrix or linear transformation matrix.
It is evident from Eq. 1.20 that

e0i = Lei ; i = 1, 2, 3 (1.22a)

3
But since the basis {e0i } are orthonormal, i.e.,

e0i · e0j = e0T 0


i ej = δ ij (1.23)

it follows that  
e0T
1 £ ¤
LT L =  e0T
2
 e01 e02 e03 =I (1.24)
e0T
3

or that the transformation matrix L is an orthogonal matrix. Therefore, it


follows from Eq. 1.20 that
x = LT x0 (1.25)

1.1.1 Index Notation


Throughout we shall employ the following index notation conventions:

• Range convention: Every index takes on values 1,2, and 3 unless stated
differently,
• Summation convention: Repeated (dummy) indices are summed up over
their range, and
• Free index convention: Indices that are not dummy one are called free
indices.

1.1.2 Alternating Symbol


This symbol is defined as following

 1 for even permutations of i,j,k
ijk = 0 otherwise (1.26)

−1 for odd permutations of i,j,k

Consequently, we can write

ei × ej = ijk ek (1.27)
(ei × ej ) · ek = ijk (1.28)
u×v = ijk uj vk ei (1.29)
(u × v)i = ijk uj vk (1.30)

Namely,

u×v = 1jk uj vk e1 + 2jk uj vk e2 + 3jk uj vk e3 (1.31)


= ( 123 u2 v3 + 132 )e1 (1.32)
+( 231 u3 v1 + 213 )e2 (1.33)
+( 312 u1 v2 + 321 )e3 (1.34)
= (u2 v3 − u3 v2 )e1 + (u3 v1 − u1 v3 )e2 + (u1 v2 − u2 v1 )e3 (1.35)

4
Example 1 Show that ijk ijk = 6.

Solution 1

ijk ijk = + ij2 ij2 + ij3 ij3


ij1 ij1
= i21 i21 + i31 i31
+ i12 i12 + i32 i32
+ i13 i13 + i23 i23
= 321 321 + 231 231
+ 312 312 + 132 132
+ 213 213 + i23 i23 = 6

Example 2 Use alternating symbol to evaluate ∇ × u, where u is a vector.

Solution 2 It can be easily verified that


∂uk
(curlu)i = (∇ × u)i = ijk uk,j = ijk (1.36)
∂xj

or that
∇×u= ijk uk,j ei (1.37)

1.1.3 General Cartesian Tensors


Let a physical quantity T be defined in a three dimensional Euclidean space,
such that for a given choice of the basis {ei } it is possible to specify T in terms
of an ordered set of 3n numbers denoted by tijk... containing a total of n indices.

Definition 1 If the components of T in the new vector basis {e0i } become t0ijk... ,
where

t0ijk... = ip jq kr ...tpqr... (1.38)


then T is called a Cartesian tensor of order n.

Note: Multiplication of a tensor T by a scalar α produces another tensor


αT of the same order

αt0ijk... = ip jq kr ...αtpqr... (1.39)


Therefore, n=0,1, 2 corresponds to a scalar, vector, and a matrix, respec-
tively.

5
Tensor Properties
1. Two tensors of the same ordered can be added: A + B = tensor of the
same order as A.
2. Operation of addition of tensors is commutative and associative, i.e.,
A + B = B + A; (A + B) + C = A + (B + C).
3. Two tensors are equal if their difference is the null tensor, i.e., A − B = 0
or aijk... − bijk... = 0.

Tensor Products
Let u and v be two vectors (tensors of order 1) and S and T be two tensors of
order 2. Then we have

u0i = ik uk ; vi0 = ik vk (1.40)


s0ij = ik jm skm ; t0ij = ik jm tkm (1.41)

Consider a set of nine numbers aij defined by the product

aij = ui vj (1.42)

It follows that
a0ij = u0i vj0 = ik jm uk vm = ik jm akm (1.43)
which implies that A is a tensor of order 2.
Similarly, the numbers defined by

bijk = tij uk (1.44)

obey the following transformation law

b0ijk = t0ij u0k = ip jq tpq kr ur = ip jq kr bpqr (1.45)

which states that B is a tensor of order 3 with 27 components (33 ). Also, the
set of 34 = 81 numbers defined by

cijkl = sij tkl (1.46)

represents the components of a tensor of order four.

Definition 2 The products in Eqns. 1.42, 1.45 and 1.46 are called outer prod-
ucts of the tensors.

They can be extended to higher order tensors. An outer product between


two tensors generates a tensor whose order is the sum of the orders of the
corresponding tensors.

Definition 3 An inner product between two tensors is characterized by the pres-


ence of one or more of dummy indices.

6
The following are examples of inner products:

α = ui vi (1.47)
pi = tij uj (1.48)
qik = sij tkj (1.49)

It is easy to show that α, p, and q are tensors of order zero, one , and two,
respectively.
Note: The order of the resulting tensor is always less in the inner products
than in the corresponding outer products.
Apparently

α = aii (1.50)
pi = bijj (1.51)
qik = cijkj (1.52)

where the quantities on the RHS of the last three equations are obtained by
putting two indices equal and then summing over this index. This operation is
called contraction. So, a scalar α is obtained by contracting the second order
tensor A, the vector p is obtained by contracting the third order tensor B, etc.
In general, the contraction of a tensor of order n (>2) results in a tensor of
order n-2.

Symmetry of Tensors
Definition 4 Tensor W with components wij is symmetric if

wij = wji (1.53)

and it is antisymmetric if
wij = −wji (1.54)

Note: For any antisymmetric tensor w11 = w22 = w33 = 0.


This definition can be extended to tensors of higher order, the symmetry
and antisymmetry being defined with respect to a particular pair of indices

Definition 5 Tensor W with components wijkl... is symmetric in the pair of


indices jk if
wijkl... = wjikl... (1.55)

Symmetry and antisymmetry of a tensor are intrinsic properties, indepen-


dent of coordinate system in which it is represented. For example, for a sym-
metric tensor W of order 2 we have
0 0
wij = ir js wrs = ir js wsr = is jr wrs = jr is wrs = wji (1.56)

Or in another words, if W is symmetric in coordinate system {xi }; i = 1, 2, 3,


it is also symmetric in the coordinate system {x0i }; i = 1, 2, 3.

7
The Quotient Rule
Theorem 1 Let W be an entity in {xi }; i = 1, 2, 3 coordinate system repre-
sented by an ordered nine quantities wij . Suppose for all vectors v the scalars

ui = wij vj (1.57)

are components of a vector u. Then W is a second order tensor.

Proof. From the assumption we have that

u0i = wij
0 0 0
vj = wij js vs = ir ur = ir wrs vs (1.58)

from which it follows that


0
(wij js − ir wrs )vs =0 (1.59)

for all vs . Therefore,


0
wij js − ir wrs =0 (1.60)
By multiplying the last equation by ts we obtain
0
wij ts js = ir ts wrs (1.61)

Due to orthogonality relationship ts js = δ tj it follows that


0
wit = ir ts wrs (1.62)

which completes the proof.


The proof extends to tensors of higher order. In general we have

Theorem 2 Let the components of a tensor of order m be uijk... and the com-
ponents of a tensor of order n be vrst... . Then if relationship

uijk... = wijk... vrst... (1.63)

holds for all uijk... and vrst... , then wijk... re components of a tensor of order
m+n.

Example 3 The set of scalars uijkm is such that for all second order tensors
with components ekm , the scalars σ ij = uijkm ekm are the components of a second
order tensor σ. Then, uijkm are the components of a fourth order tensor.

Solution 3

σ 0ij = u0ijkm e0km = u0ijkm kr ms ers = il jn σ ln = il jn uln rs ers (1.64)

implies that
(u0ijkm kr ms − il jn uln rs )ers =0 (1.65)
for all ers . Therefore,
u0ijkm kr ms = il jn uln rs (1.66)

8
By multiplying the last result with tr os it follows that

u0ijkm kr tr ms os = il jn tr os uln rs (1.67)

Using the orthogonality relationships kr tr = δ kt and ms os = δ mo we obtain

u0ijto = il jn tr os uln rs (1.68)

which completes the proof.

1.1.4 Isotropic Tensors


Definition 6 A Cartesian tensor is said to be isotropic if its components are
identical regardless of how the axes of the coordinate system are rotated.

The following properties can be deduced for isotropic tensors:

1. There are no nontrivial isotropic tensors of order 1.


2. Any isotropic tensor of order 2 can be written as αI.
3. The alternating symbol ijk is a component of a 3rd order isotropic tensor
and any 3rd order isotropic tensor can be written as α².
4. If C is an isotropic tensor of order 4, then

cpqrs = αδ pq δ rs + βδ pr δ qs + γδ ps δ qr (1.69)

The first property is self evident. The second property follows from

δ 0ij = ir js δ rs = δ ij (1.70)

The third property can be shown as following:


0
ijk = ir js kt rst
= i1 j2 k3 + i2 j3 k1 + i3 j1 k2
− i1 j3 k2 − i2 j1 k3 − i3 j2 k1
 
i1 j1 k1
= det  i2 j2 k2 = ijk
i3 j3 k3

which completes the proof.

1.1.5 Tensor Fields


Definition 7 If the components wijk... of the tensor W are functions of the
coordinates {xi }, i = 1, 2, 3 of the points in given region, then W is called a
tensor filed or tensor function of position.

9
The main feature of a tensor field is ease of multiple partial derivatives of
the components wijk... . Usually, they are denoted by

∂wijk...
= wijk...,p (1.71)
∂xp
∂ 2 wijk...
= wijk...,pq (1.72)
∂xp ∂xq

etc. For example,


ϕ,i (1.73)
represents components of a gradient of a scalar field.

1.1.6 Tensor Gradient


Definition 8 In an open region R, let the components wijk... of tensor W of
order n be continuously differentiable functions of {xi }, i = 1, 2, 3. Then,

wijk...,p (1.74)

are the components of a tensor field of order n+1 called the tensor gradient of
W.

Proof. (of consistency). From


0
wrst... = ri sj ... ri sj ...wij... (1.75)

it follows that 0
∂wrst... ∂wij... ∂xp
= ri sj ... (1.76)
∂x0q ∂xp ∂x0q
Since
0
xp = qp xq (1.77)
we obtain that 0
∂wrst... ∂wij...
= ri sj ... qp (1.78)
∂x0q ∂xp

Special Cases
• For ϕ a scalar field ϕ,i denotes the components of a 1st order tensor.
Thus gradϕ = ∇ϕ = ϕ,i ei . Similarly ϕ,ij are components of a second
order tensor, and ϕ,ii = ∇2 ϕ.
• For a vector filed f , fi,j denotes components of a 2nd order tensor and
fi,i = ∇ · f =divf . Similarly, (Curlf )i = ijk fk,j or ∇ × f = ijk fk,j ei .

Using tensor notation we can verify many vector identities. For example:

10
1. div(u × v) = v·Curlu − u·Curlv
2. Curl(Ωu) =ΩCurlu − u×gradΩ
3. CurlCurlu =grad(divu)−∇2 u

To prove identity 1 we proceed as following. Let

w = u × v = ²ijk ui vj ek (1.79)

Then

divw = (1.80)
= vj jki ui,k − ui ikj vj,k (1.81)
= vj (Curlu)j − ui (Curlv)i ¤ (1.82)

Proof of identity 3 can be constructed in similar fashion. Namely, let

w =Curlu = ∇ × u = ²ijk uk,j ei (1.83)

Then

Curlw = ijk wk,j ei = ijk ( kmn un,m ),j ei (1.84)


= kij kmn (un,mj )ei (1.85)

Using the so called − δ equality

kij kmn = δ im δ jn − δ in δ jm (1.86)

it follows that

Curl(Curlu) = (δ im δ jn − δ in δ jm )(un,mj )ei (1.87)


2
= (uj,ij vj − ui,jj )ei = ∇(∇ · u) − ∇ u¤ (1.88)

1.2 The Divergence Theorem


Theorem 3 Let V denote a volume of a region space bounded by a piecewise
smooth surface S and its interior and let ui (x) be the components of a vector
filed in a fixed vector basis {ei } that span the space. Assume that the first partial
derivatives ui,j exist and are continuous in V . Then,
Z Z Z Z
divudV = u · ndS ⇐⇒ ui,i dV = ui ni dS (1.89)
V S V S

where n is the outward unit normal to S at the surface element dS.


Proof. See Kellog, O. D. (1953). Foundations of Potential Theory, Dover,
New York, pp. 37-39.

11
1.2.1 Generalization to Tensors of Any Order
Theorem 4 Let us consider a tensor filed Ajkl... . Let the region V with bound-
ary surface S be within the region of definition of Ajkl... . Assume that every
component of Ajkl... is continuously differentiable. Then,
Z Z

Ajkl... dV = ni Ajkl... dS (1.90)
V ∂xi S

Proof. For u =A(x)ei , i = 1, 2, 3, Eq. 1.89 implies that


Z Z
∂A
dV = Ani dS
V ∂xi S

for every Ajkl... . Therefore,


Z Z

Ajkl... dV = ni Ajkl... dS
V ∂xi S

Exercise 1 Show that


Z Z Z Z
ϕ,i dV = ϕni dS ⇐⇒ ∇ϕdV = ϕndS (1.91)
V S V S

Exercise 2
Z Z Z Z
ui,i = ui ni dS ⇐⇒ ∇·u= u · ndS (1.92)
V S V S

Exercise 3
Z Z Z Z
ijk uk,j dV = ijk uk nj dS ⇐⇒ ∇×u= n × udS (1.93)
V S V S

1.3 Dyadics
Consider a vector v expressed in terms of the basis {ei } or in terms of the basis
{e0i }
v =vi ei = vi0 e0i (1.94)
This direct representation can be extended to the higher order tensors.

Definition 9 Consider now a second-order tensor A with components aij , i, j =


1, 2, 3. We form the following expression

a = aij ei ej (1.95)

where the base vectors are juxtaposed in a specific order and the summation
convention applies. Then the last equation defines a dyadic.

12
Since
e0i = ij ej & ei = 0
ji ej (1.96)
Eq. 1.95 implies that
0 0 0 0
a = aij ki ek mj em = (aij ki mj )ek em (1.97)

However, aij ki mj = a0km and thus

a = aij ei ej = a0km e0k e0m (1.98)

These expressions are similar to those for v in Eq. 1.94, except that the base
vectors ei and e0i have been replaced by the juxtaposed pairs ei ej and e0i e0j .
Consequently Eq. 1.98 can be interpreted as a direct representation of the
second-order tensor A.
Algebraic operations can be carried out directly with a provided certain rules
of operations on the pair ei ej are followed.

• Since v =vi ei = vi0 e0i it follows that

(ek · a) · em = (ek · aij ei ej ) · em = aij δ ki δ jm = akm (1.99)

Note that
(em · a) · ek = amk (1.100)
Therefore, the ordering of ei ej in Eq. 1.98 is critical in carrying out
algebraic operations on a and in general it cannot be changed arbitrarily.
• All the notions associated with the second-order tensor A (symmetry,
transpose, orthogonality, eigenvalues, and principal axes) may be used for
dyadics.
• Special dyadics

— For vectors u and v:


ui vj ei ej ≡ dyad (1.101)
— The dyadic corresponding to the unit tensor I is

i = δ ij ei ej = e1 e1 + e2 e2 + e3 e3 (1.102)

is called unit dyadic.

The following results can be proved:

u·i = i·u=u (1.103)


(a × u) · v = a · (u × v) (1.104)
v·(a × u) = (v·a) × u (1.105)
(u×a) × v = u × (a × v) (1.106)

13
x3

(r,φ,z)

z
O x2
r
φ

x1

Figure 1.2: Cylinrical coordinate system.

For example, the proof of the identity u·i = i · u = u can be constructed as


following

u·i = (uk ek ) · δ ij ei ej = uk δ ki δ ij ej = uj ej = u (1.107)


i · u = δ ij ei ej · uk ek = δ ij ei δ jk uk = ui ei = u¤ (1.108)

Similar proofs can be constructed for the other identities.

1.4 Vector Identities in Cylindrical Coordinates


Let’s consider cylindrical coordinate system {r, ϕ, z} with unit bases {er , eϕ , ez }
(Fig. 1.2)
The corresponding Cartesian components are given by

x1 = r ∗ cos ϕ; x2 = r ∗ sin ϕ; x3 = z (1.109)

and an element of an arc of a curve C can be written as

ds2 = dr2 + (rdϕ)2 + dz 2 (1.110)

Consequently it follows that


∂ ∂ ∂
∇f (r, ϕ, z) = (er + eϕ + ez )f (1.111)
∂r r∂ϕ ∂z
Recall that unit tangent vector along a space curve is given by
∂x
t= (1.112)
∂s

14
where s denotes a parameter along the curve. The position vector x can be
written as
x =x1 e1 + x2 e2 + x3 e3 = r cos ϕe1 + r sin ϕe2 + ze3 (1.113)
then from Eqns. 1.112 and 1.113 it follows that
∂x
er = = cos ϕe1 + sin ϕe2
∂r
∂x
eϕ = = − sin ϕe1 + cos ϕe2 (1.114)
r∂ϕ
∂x
ez = = e3
∂z
Since ei , i = 1, 2, 3 are constant vectors we obtain the following
∂er ∂er ∂er
= 0; = eϕ ; =0
∂r ∂ϕ ∂z
∂eϕ ∂eϕ ∂eϕ
= 0; = −er ; =0 (cy8)
∂r ∂ϕ ∂z
∂ez ∂ez ∂ez
= 0; = eϕ ; =0
∂r ∂ϕ ∂z
Let u be a vector with components (ur , uϕ , uz ) along (er , eϕ , ez ), i.e.
u =ur er + uϕ eϕ + uz ez (1.115)
Then
∂ ∂ ∂
∇u = (er + eϕ + ez )(ur er + uϕ eϕ + uz ez )
∂r r∂ϕ ∂z
∂ur ∂uϕ ∂uz
= er ( er + eϕ + ez )
∂r ∂r ∂r
1 ∂ur ∂uϕ ∂uz
+ eϕ ( er + ur eϕ + eϕ − uϕ er + ez )
r ∂ϕ ∂ϕ ∂ϕ
∂ur ∂uϕ ∂uz
+ez ( er + eϕ + ez ) (1.116)
∂z ∂z ∂z
from which it follows that
∂ur 1 ∂uϕ ∂uz
∇u = er er + (ur + )eϕ eϕ + ez ez
∂r r ∂ϕ ∂z
1 ∂uz ∂uϕ ∂ur ∂uz
+ eϕ ez + ez eϕ + ez er + er ez (1.117)
r ∂ϕ ∂z ∂z ∂r
∂uϕ 1 ∂ur
+ er eϕ + ( − uϕ )eϕ er
∂r r ∂ϕ
Consequently, we have that
∂ur 1 ∂uϕ ∂uz
∇·u= + (ur + )+ (cy10)
∂r r ∂ϕ ∂z

15
x3

(R,φ,θ)

R
θ

O x2

x1

Figure 1.3: Spherical coordinate system.

Note in order to calculate ∇ · u simply perform the dot products between


each pair of unit vectors in Eq. 1.117.
Similarly, by noting that eϕ × ez = er , etc. ... it follows that
∂ ∂ ∂
∇ × u =(er + eϕ + ez ) × (ur er + uϕ eϕ + uz ez ) (1.118)
∂r r∂ϕ ∂z
or that
1 ∂uz ∂uϕ ∂ur ∂uz ∂uϕ uϕ 1 ∂ur
∇ × u = er ( − ) + eϕ ( − ) + ez ( + − ) (1.119)
r ∂ϕ ∂z ∂z ∂r ∂r r r ∂ϕ

Example 4 ∇2 f (r, ϕ, z) =?.

Solution 4 From
∇2 f = ∇ · ∇f (1.120)
and
u =∇f (1.121)
using Eq. 1.111 we obtain
∂2f 1 ∂f 1 ∂2f ∂2f
∇2 f = + + + (1.122)
∂r2 r ∂r r2 ∂ϕ2 ∂z 2

1.5 Vector Identities in Spherical Coordinates


Let’s consider spherical coordinate system {R, θ, ϕ, } with unit bases {eR , eθ , eϕ }
(Fig. 1.3)
Then corresponding Cartesian coordinates are given by

x1 = R sin θ cos ϕ; x2 = R sin θ sin ϕ; x3 = R cos ϕ (1.123)

16
Then the element of an arc of a curve is defined by
ds2 = dR2 + (Rdθ)2 + (R sin θdϕ)2 (1.124)
and
∂ ∂ 1 ∂
∇f (R, θ, ϕ) = (eR + eθ + eϕ )f (1.125)
∂R r∂θ R sin θ ∂ϕ
Therefore we have that
x = sin θ cos ϕe1 + R sin θ sin ϕe2 + R cos θe3 (1.126)
and that
∂x
eR = = sin θ cos ϕe1 + sin θ sin ϕe2 + cos θe3
∂R
1 ∂x
eθ = = cos θ cos ϕe1 + cos θ sin ϕe2 − sin θe3 (1.127)
R ∂θ
1 ∂x
eϕ = = − sin ϕe1 + cos ϕe2
R sin θ ∂ϕ
Therefore,
∂eR ∂eR ∂eR
= 0; = eθ ; = − sin θeϕ
∂R ∂θ ∂ϕ
∂eθ ∂eθ ∂eθ
= 0; = −eR ; = cosθeϕ (1.128)
∂R ∂θ ∂ϕ
∂eϕ ∂eϕ ∂eϕ
= 0; = 0; = − sin θeR − cos θeθ
∂R ∂θ ∂ϕ
Based on these results we have that
∂ ∂ 1 ∂
∇u =(eR + eθ + eϕ )(uR eR + uθ eθ + uϕ eϕ ) (1.129)
∂R r∂θ R sin θ ∂ϕ
which leads to
∂uR ∂uθ ∂uϕ
∇u = eR ( eR + eθ + eϕ )
∂R ∂R ∂R
eθ ∂uR ∂uθ ∂uϕ
( eR + uR eθ + eθ − uθ eR + eϕ )
R ∂θ ∂θ ∂θ
eϕ ∂uR ∂uθ
+ ( eR + uR sin θeϕ + eθ + uθ cos θeϕ
R sin θ ∂ϕ ∂ϕ
∂uϕ
+ eϕ − sin θuϕ eR − cos θuϕ eθ )
∂ϕ
which after some algebra becomes
∂uR 1 ∂uθ 1 ∂uϕ
∇u = eR eR + (uR + )eθ eθ + (sin θuR + cos θuθ + )eϕ eϕ
∂R R ∂θ R sin θ ∂ϕ
1 ∂uϕ 1 ∂uθ 1 ∂uR
+ eθ eϕ + ( − cos θuϕ )eϕ eθ + ( − sin θuϕ )eϕ eR
R ∂θ R sin θ ∂ϕ R sin θ ∂ϕ
∂uϕ ∂uθ 1 ∂uR
+ eR eϕ + eR eθ + ( − uθ )eθ eR (1.130)
∂R ∂R R ∂θ

17
From the last result we obtain that
∂uR 2 1 ∂uθ 1 ∂uϕ
∇·u= + uR + ( + cot θuθ ) + (1.131)
∂R R R ∂θ R sin θ ∂ϕ
Similarly, it follows that
1 ∂uϕ 1 ∂uθ
∇ × u = eR ( + cot θuϕ − )
R ∂θ sin θ ∂ϕ
1 ∂uR ∂uϕ uϕ
+eθ ( − − )
R sin θ ∂ϕ ∂R R
∂uθ uθ 1 ∂uR
+eϕ ( + − ) (1.132)
∂R R R ∂θ
or  
e Reθ R sin θeϕ
1  ∂R ∂ ∂ 
∇×u= ∂R ∂θ ∂ϕ (1.133)
R sin θ
uR Ruθ R sin θuϕ

Example 5 Evaluate ∇2 f in spherical coordinates.

Solution 5 Using the fact that ∇2 f = ∇ · ∇f and Eqns. 1.125 and 1.131 it
follows that
∂2f 2 ∂f 1 ∂2f ∂f 1 ∂2f
∇2 f = + + ( + cot θ + (1.134)
∂R2 R ∂R R2 ∂θ2 ∂θ sin2 θ ∂ϕ2

1.6 Eigenvalue Problem for Real Symmetric Ten-


sors of Order Two
Definition 10 Let A be a real symmetric tensor of order 2 with components
aij in the vector basis {ei }. Let v be a vector with components vi in {ei }. Then
the eigenvalue problem for A is defined by

Av =λv ⇐⇒aij vj = λvi (1.135)

for λ being a scalar.

Equation 1.135 can be written as

(A−λI)v = 0 ⇐⇒ (aij − λδ ij )vj = 0 (1.136)

The system 1.136 has nontrivial solution if and only if

det(A−λI) =0 (1.137)

or  
a11 − λ a12 a13
det  a21 a22 − λ a23  = 0 (1.138)
a31 a23 a33 − λ

18
Equation 1.138, called the characteristic equation, is cubic in λ and it may be
written as
D(λ) = λ3 − IA λ2 + IIA λ − IIIA = 0 (1.139)
where the invariants IA , IIA , and IIIA are defined by

IA = aii
1
IIA = (aii ajj − aij aji ) (1.140)
2
IIIA = det A = ijk a1i a2j a3k

It is easy to show that IA = trace(A) and that


· ¸ · ¸ · ¸
a11 a12 a11 a13 a22 a23
IIIA = det + det + det (1.141)
a21 a22 a31 a33 a32 a33

The roots of the characteristic equation are called the eigenvalues of A and
the corresponding nontrivial solutions v of Av =λv are called the eigenvec-
tors of A. There are three eigenvalues λk , k = 1, 2, 3 of A and corresponding
eigenvectors are denoted by v(k) , i.e.,

Av(k) =λk v(k) ; k = 1, 2, 3. (1.142)

Clearly, if v is solution of Eq. 1.136, so is αv, where α denotes a scalar.


Consequently, the length of the eigenvectors is arbitrary. We shall normalize
them to unit vectors.

Theorem 5 If A is a real symmetric second-order tensor then all eigenvalues


of A are real.

Proof. From

aij vj = λvi (1.143)


aij vj = λvi (1.144)

where an overbar denotes the complex conjugate it follows that

aij vi vj = λvi vi = aij vj vi (1.145)


aij vj vi = λvi vi (1.146)

By subtracting the last two equations we obtain

(λ − λ)vi vi = 0 (1.147)

Since vi vi 6= 0, it follows that λ = λ.

Theorem 6 If A is a real symmetric second-order tensor then the eigenvectors


corresponding to two distinct eigenvalues of A are orthogonal.

19
Proof. Suppose that the two distinct eigenvalues are denoted by λ1 and λ2 .
Then from
(1) (1)
aij vj = λ1 vi (1.148)
(2) (2)
aij vj = λ2 vi (1.149)

it follows that
(1) (2) (1) (2)
λ1 vi vi = aij vi vj (1.150)
(2) (1) (1) (2)
λ2 vi vi = aij vi vj (1.151)

and thus
(1) (2)
(λ1 − λ2 )vi vi =0 (1.152)
(1) (2)
Since λ1 6= λ2 it follows that vi vi = 0.

20
Chapter 2

Strain Tensor

Let us assume that a point of an elastic body before deformation is specified by


P (a1 , a2 , a3 ). At a later instant of time the body is deformed and the point P
moved to Q(x1 , x2 , x3 ). We are assuming the change of body to be continuous
and the transformation of P to Q to be one-to-one and that it can be specified
as
xi = xi (a1 , a2 , a3 ); i = 1, 2, 3 (2.1)
i.e., coordinates of a point in deformed state can be expressed in terms of coor-
dinate of the point in undeformed state. We also assume that this has a unique
inverse
ai = ai (x1 , x2 , x3 ); i = 1, 2, 3 (2.2)
or that coordinates of a point in undeformed state can be expressed in terms of
coordinates of the same point in deformed configuration.
We are interested in describing the strain of the body fibers, i.e., with stretch-
ing and distortion of the body. If. P P 0 P 00 are three closely spaced points forming
a triangle in the original configuration, and if they change to QQ0 Q00 in deformed
configuration, the change of the area and the angles of the triangle is completely
defined if we know the change in the length of the sides.
Consider now an infinitesimal element connecting P (a1 , a2 , a3 ) with P 0 (a1 +
da1 , a2 + da2 , a3 + da3 ). The square of length ds0 of P P 0 is given in the original
configuration by
ds20 = dai dai (2.3)
Since P moves to Q(x1 , x2 , x3 ) and P 0 to Q0 (x1 + dx1 , x2 + dx2 , x3 + dx3 ) it
follows then that the square of length ds of QQ0 in deformed configuration is
given by
ds2 = dxi dxi (2.4)
Using Eq.2.1 we obtain that
∂xi ∂xi ∂xi ∂xi
ds2 = daj dak = daj dak (2.5)
∂aj ∂ak ∂aj ∂ak

21
Q’’ Q’

P’ u Q(x1,x2,x3)
P(a1,a2,a3)
X3 P’’

O X2
X1

Figure 2.1: Deformed and undeformed states of an elastic body.

and consequently
∂xi ∂xi ∂xα ∂xβ
ds2 − ds20 = ( − δ jk )daj dak = (δ αβ − δ jk )daj dak (2.6)
∂aj ∂ak ∂aj ∂ak

Similarly, by using Eq.2.2 and


∂aα ∂aα
ds20 = dxi dxj (2.7)
∂xi ∂xj

we obtain
∂aα ∂aβ
ds2 − ds20 = (δ ij − δ αβ )dxi dxj (2.8)
∂xi ∂xj
Thus we define Green strain tensor
1 ∂xα ∂xβ
Eij = (δ αβ − δ jk ) (2.9)
2 ∂aj ∂ak

and Almansi strain tensor


1 ∂aα ∂aβ
eij = (δ ij − δ αβ ) (2.10)
2 ∂xi ∂xj

Therefore, we can write

ds2 − ds20 = 2Eij dai daj (2.11)

ds2 − ds20 = 2eij dxi dxj (2.12)


Using quotient rule it follows that Eij and eij are components of second-
order tensors.

22
Remark 1 It should be noted that Green tensor is defined in terms of unde-
formed coordinates ai while Almansi strain tensor is defined in terms of de-
formed coordinates xi .

Since components of displacement vector are given by

ui = xi − ai (2.13)

we can obtain different forms of the Green and Alamansi strain tensors which
are more commonly used
1 ∂uα ∂uβ
Eij = [δ αβ ( + δ αi )( + δ βj ) − δ ij ]
2 ∂ai ∂aj
1 ∂uα ∂uβ ∂uα ∂uβ
= [δ αβ ( + δ βj + δ αi + δ αi δ βj ) − δ ij ]
2 ∂ai ∂aj ∂ai ∂aj
1 ∂ui ∂uj ∂uα ∂uα
= ( + + ) (2.14)
2 ∂aj ∂ai ∂ai ∂aj

Similarly,
1 ∂ui ∂uj ∂uα ∂uα
eij = ( + − ) (2.15)
2 ∂xj ∂xi ∂xi ∂xi
In unabridged notation (x, y, z) for (x1 , x2 , x3 ), (a, b, c) for (a1 , a2 , a3 ), and
(u, v, w) for (u1 , u2 , u3 ) we have that components of the Green strain tensor
(Lagrangian strain tensor) is given by

∂u 1 ∂u 2 ∂v ∂w 2
Eaa = + [( ) + ( )2 + ( ) ]
∂a 2 ∂a ∂a ∂a
1 ∂u ∂v ∂2u ∂2v ∂2w
Eab = [ + +( + + )]
2 ∂b ∂a ∂a∂b ∂a∂b ∂a∂b
..
. (2.16)

while the components of the Almansi strain tensor ( Eulerian strain tensor) are
given by
∂u 1 ∂u 2 ∂v ∂w 2
exx = − [( ) + ( )2 + ( ) ]
∂x 2 ∂x ∂x ∂x
1 ∂u ∂v ∂2u ∂2v ∂2w
exy = [ + −( + + )]
2 ∂y ∂x ∂x∂y ∂x∂y ∂x∂y
..
. (2.17)

Remark 2 u, v, and w are considered as functions of a, b, a, the position points


in the body in undeformed configuration when the Lagrangian strain tensor is
evaluated whereas they are considered as functions of x, y, z (the positions points
in deformed configuration) when Eulerian strain tensor is evaluated.

23
If the components of displacement field ui are such that their first derivatives
are so small that the squares and the product of their partial derivatives of ui
are negligible, then eij reduces to Cauchy’s infinitesimal strain tensor

1 ∂ui ∂uj 1
eij = ( + ) = (ui,j + uj,i) (2.18)
2 ∂xj ∂xi 2

which in unabridged notation gives


∂u ∂v ∂w
exx = ; eyy =
; ezz =
∂x ∂y ∂z
1 ∂u ∂v 1 ∂u ∂w
exy = eyx = ( + ); exz = ezx = ( + )
2 ∂y ∂x 2 ∂z ∂x
1 ∂v ∂w
eyz = ezy = ( + ) (2.19)
2 ∂z ∂y
Remark 3 In the infinitesimal displacement theory, the distinction between the
Lagrangian and Eulerian strain tensor disappears. Namely
1 ∂ui ∂uj 1 ∂ui ∂ak ∂uj ∂ak
eij = ( + )= ( + )
2 ∂xj ∂xi 2 ∂ak ∂xj ∂ak ∂xi
1 ∂ui ∂xk ∂uk ∂uj ∂xk ∂uk
= [ ( − )+ ( − )]
2 ∂ak ∂xj ∂xj ∂ak ∂xi ∂xj
1 ∂ui ∂uk ∂uj ∂uk
= [ (δ kj − )+ (δ ki − )]
2 ∂ak ∂xj ∂ak ∂xj
1 ∂ui ∂uj
= [ + ] = Eij
2 ∂aj ∂ai

2.1 Geometric Interpretation of Infinitesimal Strain


Components
2.1.1 Component exx
For that purpose we examine an elastic element of length dx and dy along the
x- and y-axis, respectively. We assume that ∂u∂x > 0. Then, if the element is
extended in the x-direction uniformly along the y-axes we have

ds0 = dx
∂u ∂u
ds = dx + u + dx − u = (1 + )dx
∂x ∂x
which implies that

ds2 − ds20 = 2eij dxi dxj = 2exx dx2

24
y

u u+uxdx

x
P dx P’

Figure 2.2: Elastic element streched in horizontal direction.

or
ds − ds0 2exx 2
= ds0 = exx
ds0 ds + ds0 2 + ∂u
∂x
Using the assumption of small strain yields then
ds − ds0
exx =
ds0
or that the component exx measures the change of length per unit length of a
fiber parallel to the x-axis. Consequently, the component exx describes extension
(compression) of a material fiber. Similar conclusions follow for the components
eyy and ezz .

2.1.2 Component exy


To see the geometrical interpretation of this component of the starin tensor,
let us consider again the small rectangle of the material which undergoes the
following deformations
Apparently, before deformation the angle ](OP , OP 0 ) = π/2. After defor-
mation we have that
∂u
tan α =
∂y
∂v
tan β =
∂x
Suppose that tan α > 0 for α increasing in clockwise (CW) direction, while
tan β > 0 for β increasing in CCW direction. For small strains we may take
that tan α ≈ α and tan β ≈ β. Therefore,
1 ∂u ∂v 1
exy = ( + ) = (α + β)
2 ∂y ∂x 2

25
y Q’’

α Q’(u+uydy,v+vydy)
P’ P’’
Q(u+uxdx,
dy β
v+vxdx)
O’(u,v)

x
O dx P

Figure 2.3: Shearing strain interpretation.

which implies that the strain component exy is equal to one half of the amount
(in radians) the right angle between two fibers is diminished due to deformation.
In engineering, the strain components eij ; i 6= j are called the shearing
strains.

2.1.3 Simple Shear


Consider a case ∂v/∂x = ∂u/∂x = 0, i.e. u and v remain unchanged in the
x-direction. Also, we assume that ∂u/∂y > 0. Then the deformed element looks
like
and we have that
1 ∂u
exy =
2 ∂y

2.1.4 Rotation Vector ω


Consider a small element dxdy with ∂u/∂y > 0 and ∂v/∂x > 0. Then,
we have that
∂v
tan α ≈ α = >0
∂x
∂u
tan β ≈ β= >0
∂y

where α denotes an angle for which OP rotates about the z-axis and β represents
an angle for which OP 0 rotates about the z-axes. Consequently,
1 ∂v ∂u
ωz = ( − ) ≡ measure of rotation
2 ∂x ∂y

26
y

P’’ Q’’ P’ Q’

dy

x
O dx P
u

Figure 2.4: Simple shear.

Q’(uydy, P’
vydy) dy

β
Q(uxdx,vxdx)
α
x
O P
dx

Figure 2.5: Definition of rotation vector.

27
Remark 4 If there is no shear strain then exy = 0 and thus

∂u ∂v
=−
∂y ∂x

Consequently, ω z = α, which implies rigid-body-rotation (i.e. ](OQ, OQ0 ) =


π/2).

In general we have that rotation vector is defined by


1
ω = ∇×u (2.20)
2

2.1.5 Rotation Tensor Ω


Definition 11 For infinitesimal displacement field ui (x1 , x2 , x3 ) the antisym-
metric second order tensor Ω, defined by
1
Ωij = (uj,i − ui,j ) (2.21)
2
is called the rotation tensor.

It is interesting to observe the following


1 1 1
ijk Ωjk = ijk uk,j = (∇ × u)i = ω i (2.22)
2 2 2
or that the rotation tensor is related to the rotation vector. In addition we have
1 1
lmi ω i = ijk lmi Ωjk = [δ jl δ km − δ jm δ kl ]Ωjk = Ωlm
2 2
Therefore,
Ωij = ijk ω k (2.23)

2.1.6 Physical Interpretation of Rotation Vector and Ro-


tation Tensor
Theorem 7 Vanishing of the symmetric strain tensor is necessary and suffi-
cient condition for a neighborhood of a particle to move like a rigid body.

Proof. Necessary condition. Whenever neighborhood of a particle moves


as a rigid body , then ds = ds0 .Thus

ds2 − ds20 = 2eij dxi dxj = 0

for all dxi dxj . Therefore, eij = 0.


Sufficient condition. When eij = 0, then 2eij dxi dxj = 0 and thus ds = ds0 .

28
Theorem 8 When the strain tensor vanishes at point P, the infinitesimal rota-
tion of the RB motion of a neighborhood of P is given in terms of the rotation
vector ω.

Proof. Let’s consider point P 0 in the neighborhood of P . Let P (xi ) and


P (xi + dxi ). Then, the relative displacement of P 0 with respect to P is given
0

by
∂ui
dui = dxj
∂xj
1 1
= (ui,j + uj,i )dxj + (ui,j − uj,i )dxj
2 2
= eij dxj − Ωij dxj = − ijk ω k dxj

or that
du = ω×dx
The last results indicates that the relative displacement of P 0 with respect to P
is equivalent to an infinitesimal rotation ω about an axis through P in direction
of ω.

2.2 Compatibility and Strain Components


The question arises of how to determine the displacements ui , i = 1, 2, 3 when
the components of strain tensor eij are known or how to integrate the differential
equations
1
eij = (ui,j + uj,i ) (2.24)
2
Apparently, there are three unknowns and six equations. Consequently,
Eqns.2.24 will not have a single solution in general if the functions eij are spec-
ified arbitrarily. Therefore, one may suspect that a solution may exist only if
the functions eij satisfy certain conditions.
Since strain components only involve relative positions of points in the body,
and since RB-motion correspond to zero strain we expect to have the solution
ui to within a RB-motion.
If the stearins are prescribed arbitrarily we may encounter the following
situations
For a single valued continuous solution to exist, the points C and D must
meet perfectly in the strained configuration. In order to achieve that the strain
components must satisfy certain conditions. They are to be satisfied by the
strain components eij and can be obtained by elimination of ui from Eq. 2.24.

29
D C

A B

Figure 2.6: Fibers in a body before deformation.

D C

A B

Figure 2.7: Fibers in a body which may result from arbitrarily prescribed strains.

30
Therefore,
1
eij,kl = (ui,jkl + uj,ikl )
2
1
ekl,ij = (uk,lij + ul,kij )
2
1
eik,jl = (ui,kjl + uk,ijl )
2
1
ejl,ik = (uj,lik + ul,jik )
2
and subsequently

eij,kl + ekl,ij − eik,jl − ejl,ik = 0; i, j, k, l = 1, 2, 3 (2.25)

These equations known as the compatibility equations were first derived by St.
Venant (1860).
In general, Eq. 2.25 represents 34 = 81 equations. However, due to symme-
try of the strain tensor there are only six independent equations. In unabridged
notation they are listed as
∂ 2 exx ∂ ∂eyz ∂ezx ∂exy
= (− + + )
∂y∂z ∂x ∂x ∂y ∂z
∂ 2 eyy ∂ ∂ezx ∂exy ∂eyz
= (− + + )
∂x∂z ∂y ∂y ∂z ∂x
∂ 2 ezz ∂ ∂exy ∂eyz ∂ezx
= (− + + )
∂x∂y ∂z ∂z ∂x ∂y
∂ 2 exy ∂ 2 exx ∂ 2 eyy
2 = + (2.26)
∂x∂y ∂y 2 ∂x2
∂ 2 eyz 2
∂ eyy ∂ 2 ezz
2 = +
∂y∂z ∂z 2 ∂y 2
∂ 2 exzx 2
∂ ezz ∂ 2 exx
2 = +
∂z∂x ∂x2 ∂z 2
In two-dimensional case we have that u = u(x, y); v = v(x, y); w ≡ 0. There-
fore, ezx = ezy = ezz = 0 and ∂()/∂z ≡ 0. In that case the compatibility
equations reduce to a single equation
∂ 2 exy ∂ 2 exx ∂ 2 eyy
2 = + (2.27)
∂x∂y ∂y 2 ∂x2
Example 6 2D Case. Suppose that the strain filed is specified by exx = 3x +
sin y, eyy = x3 , exy = 12 (x cos y + ax2 y). Determine a suitable value for ”a” and
then calculate displacement filed u and v.

Solution 6 Compatibility equation (2.27) implies that

− sin y + 2ax = − sin y + 6x

31
and consequently a = 3. Now
∂u
exx = = 3x + sin y
∂x
which implies that
3 2
u= x + x sin y + C1 (y)
2
Similarly, from
∂v
eyy = = x3
∂y
we get
v = x3 y + C2 (x)
Consequently, from
1 ∂u ∂v 1
exy = ( + ) = (x cos y + 3x2 y)
2 ∂y ∂x 2
we have that
1 1
(x cos y + 3x2 y + C10 (y) + C20 (x)) = (x cos y + 3x2 y (2.28)
2 2
from which it follows that
C10 (y) = −C20 (x)
for all x and y. Therefore,

C20 (x) = −C10 (y) = const = A

and

C1 = −Ay + C
C2 Ax + B

where A, B, and C are constants. Finally, we obtain


3 2
u(x, y) = x + x sin y − Ay + C
2
v(x, y) = x3 y + Ax + B

32
Chapter 3

Stress Tensor

Consider an elastic body B subjected to a system of external forces Pi , i =


1, ..., n (see Fig. 3.1)
Suppose we cut the body into two parts B1 and B2 by a plane so that
B = B1 ∪ B2 . In order for separate parts to be in equilibrium has to apply to
parts B1 and B2 the forces F and −F, respectively which are the resultant
of internal forces acting on the plane of the section S. Consider now an
element of area ∆S with corresponding internal force resultant ∆F. Then, the
stress vector or traction is defined by
∆F
Tn ≡ lim (3.1)
∆S→0 ∆S
Since Tn will depend upon the orientation of the surface ∆S (specified by the
unit normal vector n), superscript n is introduced with the traction vector.
Choice of surface S is arbitrary. It is of interest, however, to consider some
special cases in which the surface ∆S = ∆Sk is parallel to one of the co-
ordinate planes. Let the normal to ∆Sk be in the positive direction of the
xk -axis, and let the stress vector acting on ∆Sk be denoted by Tk with compo-
nents T1k , T2k , and T3k along the axes x1 , x2 , and x3 ,respectively. For this special
case the following convention is used

Tjk = σkj ; k, j = 1, 2, 3 (3.2)

Therefore, for the surface perpendicular to x1 -axis we have stress vector com-
ponents σ 1j , for the surface perpendicular to x2 -axis we have stress vector com-
ponents σ 2j , and for the surface perpendicular to x3 -axis we have stress vector
components σ 3j , j = 1, 2, 3. Graphically, the components of a stress vector can
be depicted as in Fig. 3.3
Thus, σ ii , i = 1, 2, 3 (no summation) are defined as normal stresses, while
σ ij , i 6= j are known as shear stresses.

Definition 12 Sign Convention: Positive normal stress vector points away


from the material. Positive shear stress σ ij , i 6= j means the stress vector is in

33
P1
P1 F

1 2
1
S n

x3 Pn

x2
O
x1

Figure 3.1: An elastic body subjected to a system of forces.

the positive direction of xj −axis when the positive direction of the xi −axis point
out of the body.

3.1 Symmetry of the Stress Tensor


Let us consider an infinitesimal cube of size dx1 dx2 dx3 subjected to stresses σ ij
and a body force b (see Fig. 3.4) P
Since the box is in equilibrium, then Mx1 = 0, where x1 is an axis parallel
with x1 and going through the centroid of the cube. Thus
dx2 ∂σ 23 dx2
σ23 dx1 dx3 + (σ 23 + dx2 )dx1 dx3 (3.3)
2 ∂x2 2
dx3 ∂σ 32 dx3
−σ 32 dx1 dx2 − (σ 32 + dx3 )dx1 dx2 (3.4)
2 ∂x3 2
must be zero. By taking the limit as ∆V → 0 and dxi → 0 we obtain
1 ∂σ 23 1 ∂σ 32
σ 23 + lim dx2 − σ 32 − lim dx2 = 0
dx2 →0 2 ∂x2 dx2 →0 2 ∂x2

which leads to
σ23 = σ 32

34
D F

D S
Figure 3.2: The resultant internal force on an infinitesimal area.

35
s 33

s 32

s s 23
31

s 13
s 22

s 21
s 12

s 11

x3

x2
x1

Figure 3.3: The components of stress vector.

36
x3
s Ds
+
33 3

s 32 Ds
+ 32

s 31 Ds
+ 31
s +Ds 23 23

s 23 s 21 r b3
s+D
s r b2 s Ds
2 2
2
r b1 21+ 21
dx3
s 31 x2
s 32
dx1
dx2
x1 s 33

Figure 3.4: An infinitesimal box of material in equilibrium.

P P
Similarly, Mx2 = 0 provides σ 13 = σ 31 and Mx3 = 0 results in σ 12 = σ 21 .
Therefore, provided there are no internal moments proportional to a volume we
have that
σ ij = σ ji ; ij = 1, 2, 3 (3.5)
or that the stress tensor is symmetric.

3.2 Equations of Equilibrium


For equilibrium
P of an infinitesimal
P box (see Fig. 3.4) we must have that sum of
the forces F = 0. Thus from Fx1 = 0 it follows that
∂σ 11
−σ 11 dx2 dx3 + (σ 11 + dx1 )dx2 dx3
∂x1
∂σ 21
−σ 21 dx1 dx3 + (σ 21 + dx2 )dx1 dx3
∂x2
∂σ 31
−σ 31 dx1 dx2 + (σ 31 + dx3 )dx1 dx2 + ρb1 dx1 dx2 dx3
∂x3
must be zero. By taking the limit lim∆V →0 the last equation implies that
∂σ11 ∂σ 21 ∂σ 31
+ + + ρb1 = 0
∂x1 ∂x2 ∂x3

37
x2 n
T

D S n

s 1 s 31

s 21
x1

x3
Figure 3.5: Traction boundary conditions.

P P
Similar equations can be obtained for Fx2 = 0 and Fx2 = 0 which can be
summarized as the equations of equilibrium

σij,j + ρbi = 0; i, j = 1, 2, 3 (3.6)

or equivalently
divσ + ρb = 0 (3.7)

3.3 Boundary Conditions


Consider a segment of an elastic body with surface ∆S subjected to a surface
traction Tn (Fig.3.5). Unit normal on ∆S is n.
Then projection of ∆S on x1 x2 -plane is given by

∆S3 = ∆S cos ](n,x3 ) = ∆Sn3

Similarly it follows that

∆S2 = ∆Sn2
∆S1 = ∆Sn1

38
p0

a S1 x

n S2
y
Figure 3.6: An elastic wedge problem.

P
For equilibrium we must have Fx1 = 0 or

T1n ∆S = σ 11 ∆S1 + σ 12 ∆S2 + σ 13 ∆S3 + ρb1 ∆V


= (σ 11 n1 + σ 12 n2 + σ 13 n3 )∆S + ρb1 ∆V

By taking the limit lim ∆S → 0 and noting that lim∆xi →0 ∆V /∆S = 0 we


obtain
T1n = σ 11 n1 + σ 12 n2 + σ 13 n3
Similarly, we have that

T2n = σ 21 n1 + σ 22 n2 + σ 23 n3
T3n = σ 31 n1 + σ 32 n2 + σ 33 n3

or
Tin = σ ij nj (3.8)
The last result represents the Cauchy stress formula which relates external trac-
tions to internal stresses. Since Tn and n are vectors, σ is a tensor of order
2.

Example 7 Boundary conditions for an elastic wedge (Fig. 3.6)

Solution 7

σ 22 (x, 0) = −p0 ; x ∈S1


−σ 11 sin α + σ12 cos α = 0; x ∈S2
−σ 12 sin α + σ22 cos α = 0; x ∈S2

39
3.4 Plane State of Stress
Definition 13 State of stress in which σ33 = σ 31 = σ 32 ≡ 0 is called plane
state of stress in the plane x1 x2 .

According to transformation law σ 0ij = ir js σrs , the transformation matrix


is easily satisfied by single angle θ
 
cos θ sin θ 0
L =  − sin θ cos θ 0 
0 0 1

Then it is easily shown that


σxx + σ yy σ xx − σ yy
σ0xx = + cos 2θ + σ xy sin 2θ
2 2
σxx + σ yy σ xx − σ yy
σ 0yy = − cos 2θ − σ xy sin 2θ (3.9)
2 2
σ xx − σ yy
σ 0xy = − sin 2θ + σ xy cos 2θ
2
Remark 5 Sum of normal stresses remains unchanged (an invariant) in dif-
ferent coordinate systems, i.e.,

σ xx + σ yy = σ 0xx + σ 0yy (3.10)

Case 1 σ 0xy = 0. Then we have that

2σ xy
tan 2θ0 = (3.11)
σ xx − σ yy

Since
∂σ 0xx
= 2σ 0xy (3.12)
∂θ
thus σ 0xy = 0 implies extremum in normal stresses. By using

cos 2θ0 = (1 + tan2 2θ0 )−1/2 (3.13)


sin 2θ0 = tan 2θ0 (1 + tan2 2θ0 )−1/2 (3.14)

we obtain from Eqns.3.9 that


r
σ xx + σ yy σ xx + σ yy 2
σ0xx (θ0 ) = σI = + ( ) + σ 2xy (3.15)
2 2
r
0 σ xx + σyy σ xx + σ yy 2
σ yy (θ0 ) = σ II = − ( ) + σ 2xy (3.16)
2 2
where σ I and σ II are maximum and minimum normal stresses, respectively.
These stresses are known as the principal stresses.

40
Case 2 max σ 0xy . Then
∂σ 0xy
=0
∂θ
which implies that
σ xx − σ yy
tan 2θ∗ = − (3.17)
2σ xy
from which it follows that
r
σ xx + σyy 2
σ 0xy (θ∗ ) = σ 0xy max = ( ) + σ2xy (3.18)
2
Since
1
tan 2θ0 = −
tan 2θ∗
we have that
θ∗ = θ0 ± 45o (3.19)

which states that the directions of the principal stresses are bisected by
directions of maximum shear stresses.

3.5 Linear Momentum and the Stress Tensor


Suppose we remove from a body a closed region V + ∂V . The surface ∂V is
subjected to a distribution of surface tractions Tn (x,t). Each mass element of the
body may be subjected to a body force per unit mass f (x,t). Then the principle
of balance of linear momentum states that: Instantaneous rate of change of the
linear momentum of a body is equal to resultant external force acting on the
body at that time, or
Z Z Z
∂ ∂u
ρ dV = ρf dV + Tn dS (3.20)
∂t V ∂t V ∂V

Note: ρdV = dm is the particle mass. Also, Eq. 3.20 is based on a La-
grangian description, and V and S move with the particle. Since the particle
mass is constant in time we obtain from 3.20 that
Z Z Z
∂2u
ρ 2 dV = ρf dV + Tn dS (3.21)
V ∂t V ∂V

Using Cauchy stress formula it follows than that


Z Z Z
ρüi dV = ρfi dV + σ ij nj d
V
ZV Z∂V
= ρfi dV + σ ij,j dV
V V
or Z
(σ ij,j + ρfi − ρüi )dV = 0 (3.22)
V

41
Since V may be an arbitrary part of the body, we get that whenever the inte-
grand in Eq. 3.22 is continuous that
σ ij,j + ρfi = ρüi (3.23)
which is Cauchy’s first law of motion.

3.6 Balance of Angular Momentum


The principle of balance of angular momentum states: Time rate of change of
angular momentum about O equals to the moment of the forces about O acting
in V. Thus
Z Z Z

x × u,t dV = x × ρf dV + x × Tn dS (3.24)
∂t V V ∂V

Since ∂(ρdV )/∂t = u,t × u,t = 0 we obtain that


Z Z Z
n
ρ klm xl üm dV = ρ klm xl fm dV + klm xl Tm dS (3.25)
V V ∂V

Using the Gauss theorem and equation of motion we obtain for the surface
integral
Z Z Z
n
klm xl Tm dS = klm xl σ mr nr dS = klm (xl σ mr ),r dV
∂V
Z∂V V

= klm [σ ml + xl (ρüm − ρfm )]dV (3.26)


V

Substituting Eq. 3.26 into Eq. 3.25 yields that


Z
klm σ ml dV = 0
V

or that
klm σ ml

which implies that


σ ml = σ lm
the stress tensor is symmetric.

3.7 Principal Stresses


Definition 14 In general state of stress, the stress vector Tn acting on a sur-
face with outer normal n depends on n. We are asking in what direction n
the stress vector becomes normal to the surface on which there are no shearing
stresses (Fig.3.7). Such surface will be called the principal plane, its normal
the principal axis, and its value of normal stress acting on the principal plane a
principal stress.

42
n
T
n
* n

S*
S

Figure 3.7: The principal stress.

Therefore, along the principal axes we have

Tn = λn

or that
σ ij nj = λni (3.27)
which is an eigenvalue problem. This can be written in the form

(σ ij − λδ ij )nj = 0 (3.28)

Condition for a nontrivial solution of Eq. 3.28 requires that

det(σ−λI) =0 (3.29)

which is the characteristic equation of the problem. The last result can be
written in the form
λ3 − Iσ λ2 + IIσ λ − IIIσ = 0 (3.30)
where

λ1 = σ 1 ; λ2 = σ 2 ; λ3 = σ 3
Iσ = σ ii
1
IIσ = (σ ii σjj − σ ij σ ij )
2
IIIσ = det σ (3.31)

43
Thus in the coordinate system {ei } the stress tensor σ is real and symmetric.
Then, in the coordinate system {e0i }, where the basis vectors are along the
principal axes (the eigenvectors of Eq. 3.28) {e0i }, the stress tensor becomes
diagonal, i.e.,  
σ1 0 0
σ0 =  0 σ2 0  (3.32)
0 0 σ3
In that case the stress tensor invariants become

Iσ = σ1 + σ2 + σ3
IIσ = σ1 σ2 + σ1 σ3 + σ2 σ3
IIIσ = σ1 σ2 σ3 (3.33)

Example 8 The stress tensor is defined by


 
3 5 8
σ = 5 1 0 
8 0 2

Determine the principal stresses and the principal directions. Write down the
value of stress invariants.

Solution 8 The stress invariants are

Iσ = σ ii = 3 + 1 + 2 = 6
1
IIσ = (σ ii σ jj − σ ij σ ij )
2
= σ 11 σ 22 + σ 11 σ 33 + σ 22 σ 33 − σ 212 − σ 213 − σ 223
= 3 · 1 + 3 · 2 + 1 · 2 − 52 − 82 − 02 = −78
IIIσ = det σ = −108

Corresponding characteristic equation is given by

λ3 − Iσ λ2 + IIσ λ − IIIσ = 0

or
λ3 − 6λ2 + −78λ + 108 = 0
which produces

σ1 = 11.8242
σ2 = 1.2848
σ3 = −7.1090

The principal directions: For λ = λ1 we have

(σ−λ1 I)n(1) = 0

44
Tn
s nt
n
s nn

P
t

Figure 3.8: Stresses on an oblique plane.

which results in  
0.7300
n(1) = ±  0.3372 
0.5945
Similarly, the last two eigenvectors are calculated to be
   
−0.480 −0.6818
n(2) = ±  −0.8424  ; n(3) = ±  0.4204 
0.5368 0.5988

Note that the eigenvectors are normalized so that | n(s) |= 1, s = 1, 2, 3.

3.8 Normal and Shearing Stresses on an Oblique


Plane
The normal stress σ nn on the plane P is the projection of the stress vector Tn
in the direction of the unit normal n of plane P (see Fig. 3.8). Thus

σ nn = Tn · n =σ ij nj ni (3.34)

45
200 psi

a 600
100 psi
100 psi
a
x2

200 psi

O x1

Figure 3.9: A skewed plate problem.

or

σ nn = σ 11 n21 + σ 22 n22 + σ 33 n23


+2σ 12 n1 n2 + 2σ 13 n1 n3 + 2σ 23 n2 n3 (3.35)

Similarly, the shearing stress


p
σnt = Tn · t = | Tn |2 −σ2nn (3.36)

Example 9 The skewed plate of unit thickness is loaded uniformly along the
sides of the plate as shown by Fig. 3.9. Determine the elements of the stress
tensor and normal stress on a plane making 45o with the x1 and x2 axes.

Solution 9 From balance of forces in x1 − and x2 −directions for section a-a


we have that

σ 11 (A sin 60) − 100A − 200(A cos 60) sin 30 = 0


σ 12 (A sin 60) − 200(A cos 60) cos 30 = 0

which provides

σ 11 = 100 3psi
σ 12 = 100psi

46
In addition we have that

σ 22 = 200 sin 60 = 100 3psi

The stress tensor is then specified by


 √ 
100 3 100
√ 0
σ =  100 100 3 0 
0 0 0

For the surface on a 45o plane the unit normal is defined by n =( √12 , √12 , 0).
Then,

σ nn = σ ij ni nj = σ 11 n21 + σ 22 n22 + 2σ 12 n1 n2 = 100(1 + 3)psi

3.9 Equations of Motion in Cylindrical and Spher-


ical Coordinates
Recall that equations of motion can be written as

divσ+ρb =ρu,tt (3.37)

Namely,

∇·σ = ei · σ kl ek el = σ kl,i δ ik el = σlk,k el
∂xi
Therefore, equations of motion can be written as

∇ · σ+ρb =ρu,tt (3.38)

Thus in curvilinear coordinates we need to evaluate ∇ · σ.

3.9.1 Cylindrical Coordinate System (r, θ, z)


In that case we have

b = (br , bθ , bz )
u = (ur , uθ , uz )
 
σ rr σ rθ σ rz
σ =  σ θr σ θθ σ θz 
σ zr σ zθ σ zz
We may write then

σ = σ ij ei ej
= er (σ rr er + σ rθ eθ + σ rz ez )
+eθ (σθr er + σ θθ eθ + σ θz ez )
+ez (σ zr er + σzθ eθ + σ zz ez )

47
or
σ = er tr +eθ tθ +ez tz
where
tr = σ rr er + σ rθ eθ + σ rz ez
tθ = σ θr er + σ θθ eθ + σ θz ez
tz = σ zr er + σ zθ eθ + σ zz ez
Recall that
∂ 1 ∂ ∂
∇ · u = (er + eθ + ez ) · (ur er + uθ eθ + uz ez )
∂r r ∂θ ∂z
∂ur 1 1 ∂uθ ∂uz
= + ur + +
∂r r r ∂θ ∂z
Therefore we can write that
∂ 1 ∂ ∂
∇·σ = (er + eθ + ez ) · (er tr +eθ tθ +ez tz )
∂r r ∂θ ∂z
∂tr 1 1 ∂tθ ∂tz
= + tr + +
∂r r r ∂θ ∂z
which provides
∂σ rr ∂σrθ ∂σrz
∇·σ = er + eθ + ez
∂r ∂r ∂r
1
+ (σ rr er + σrθ eθ + σ rz ez )
r
1 ∂σ rθ ∂σθθ ∂σ θz
+ ( er + σ rθ eθ + eθ − σ θθ er + ez )
r ∂θ ∂θ ∂θ
∂σ zr ∂σ zθ ∂σ zz
+ er + eθ + ez (3.39)
∂z ∂z ∂z
Based on Eqns. 3.38 and 3.39 we obtain the equations of motion in cylindrical
coordinates to be
∂σ rr 1 ∂σrθ ∂σ zr 1
+ + + (σ rr − σ θθ ) + ρbr = ρür
∂r r ∂θ ∂z r
∂σ rθ 1 ∂σ θθ ∂σ θz 2
+ + + σ rθ + ρbθ = ρüθ (3.40)
∂r r ∂θ ∂z r
∂σ rz 1 ∂σ θz ∂σ zz 1
+ + + σ rz + ρbz = ρüz
∂r r ∂θ ∂z r
Similar procedure can be used in the spherical coordinate system.
Example 10 Invariants of the stress tensor. Let σ be a stress tensor in {xi }, i =
1, 2, 3  
4 1 2
σ = 1 6 0 
2 0 8
Find the invariants Iσ , IIσ , IIIσ .

48
Solution 10 The invariants are given by

Iσ = σ ii = 4 + 6 + 8 = 12
1
IIσ = (σ kk σ ll − σkl σ lk )
2 · ¸ · ¸ · ¸
σ 11 σ 12 σ 11 σ 13 σ 22 σ23
= det + det + det
σ 21 σ 22 σ 31 σ 32 σ 31 σ33
= σ 11 σ 22 + σ 11 σ 33 + σ 22 σ 33 − σ 212 − σ 213 − σ223 = 99
IIIσ = det σ = 160 (3.41)

Now let us rotate x1 x2 axes about x3 − axis for 45o counterclockwise. Then the
transformation matrix L becomes
 √ √ 
2 2
0
 2√ √2 
L = − 2 2
0 
2 2
0 0 1

and thus in the new coordinate system the stress tensor assumes the values
 √ 
6 1 √2
σ 0 = LσLT =  √1 4
√ − 2 
2 − 2 8

The stress tensor invariants for σ 0 then are

Iσ0 = σ 0ii = 18
1 0 0
IIσ0 = (σ σ − σ 0kl σ 0lk ) = 99
2 kk ll
IIIσ0 = det σ 0 = 160

Thus the invariants remain the same in the two coordinate systems.

3.10 Mean and Deviator Stress Tensor


Experiments show that yielding and plastic deformation of many metals are
essentially independent of applied mean stress σ m defined by
1
σm = (σ 11 + σ22 + σ 33 )
3
1
= (σ 1 + σ 2 + σ3 ) (3.42)
3
Most plasticity theories postulate that the plastic behavior of materials is related
primarily to that part of the stress tensor which is independent of σ m . Therefore,
the stress tensor can be written as

σ = σm + σd (3.43)

49
where

σm = diag[σm , σ m , σ m ] = [δ αβ σm ] (3.44)
 1 
3 (2σ 11 − σ 22 − σ 33 ) σ 12 σ 13
σd =  σ 21
1
(2σ 22 − σ 11 − σ 33 ) σ 23 (3.45)
3
1
σ 31 σ 32 3 (2σ 33 − σ 11 − σ 22 )

where σ m is the mean stress tensor while σ d is the deviator stress tensor. The
latter ”measures” deviation from the state of stress from a spherically symmetric
state (i.e., from the state of stress that exist in an ideal, frictionless fluid).
If the coordinate axes {xi }, i = 1, 2, 3, are the principal one, then σ 11 =
σ 1 , σ22 = σ2 , σ 33 = σ3 ,and σ ij = 0, for i 6= j. The invariants of σ m and σ d are
given by

Iσm = 3σ m ; IIσm = 3σ 2m ; IIIσm = σ 3m


Iσd = 0
1
IIσd = − [(σ 1 − σ 2 )2 + (σ 2 − σ3 )2 + (σ 1 − σ 3 )2 ]
6
1
IIIσd = (2σ 1 − σ 2 − σ 3 )(2σ 2 − σ 3 − σ 1 )(2σ 3 − σ 1 − σ2 ) (3.46)
27
The eigenvalues of the deviator tensor σ d are (the principal stresses of σ d ) are
1
S1 = σ1 − σm = [(σ1 − σ 3 ) + (σ 1 − σ2 )]
3
1
S2 = σ2 − σm = [(σ2 − σ 3 ) + (σ 2 − σ1 )]
3
1
S3 = σ3 − σm = [(σ3 − σ 1 ) + (σ 3 − σ2 )] (3.47)
3
Since S1 + S2 + S3 = 0, only two principal stresses of σ d are independent.

3.11 Generalized Hooke’s Law


Cauchy generalized Hooke’s law σ xx = Eexx into a statement that the compo-
nents of stress are linearly related to the components of strain. Thus in tensor
notation
σ ij = Cijkl ekl (3.48)
where Cijkl is a fourth order tensor of elastic constants of the material. For
isotropic materials C should be an isotropic tensor of order 4 or

Cijkl = λδ ij δ kl + µδ ik δ jl + νδ il δ jk (3.49)

Therefore,
σ ij = λekk δ ij + µeij + νeji

50
Since stress tensor is symmetric we must have µ = ν and consequently
σ ij = λekk δ ij + 2µeij (3.50)
The inverse relation is easily follows to be
µ ¶
1 λ
eij = σ ij − σkk δ ij (3.51)
2µ 3λ + 2µ
Since E = 2µ(1 + ν) and µ = λ(1 − 2ν)/2ν, we can write
1+ν ν
eij = σ ij − σkk δ ij (3.52)
E E
In unabridged notation the last result becomes
1
exx = [σ xx − ν(σ yy + σ zz )]
E
1
eyy = [σ yy − ν(σ xx + σ zz )]
E
1
exx = [σ zz − ν(σ xx + σ yy )]
E
1 1 1
exy = σ xy ; exz = σ xz ; eyz = σ yz (3.53)
2µ 2µ 2µ
Frequently, Hooke’s law is expressed as
ν
σ ij = 2µ( ekk δ ij + eij ) (3.54)
1 − 2ν
Remark 6 Stresses are not unbounded as ν → 1/2 (an incompressible mater-
ial).

Generalized Hooke’s law σij = Cijkl ekl implies the following symmetry prop-
erties of tensor C due to symmetry of σ and e
Cijkl = Cjikl = Cijlk (3.55)
In general, tensor C has 81 components. However due to symmetry conditions
expressed by Eq. 3.55, only 36 of these components are independent. This is
also clear from the fact that σ ij = Cijkl ekl is a set of six linear homogeneous
equations relating the six independent components of σ to six independent
components of e.
We next assume that there exits a strain energy function W (C) from which
the stress tensor can be derived through
∂W
σ ij = (3.56)
∂eij
For small strains, W may be expressed in the power series
1
W = W0 + bij eij + Cijkl eij ekl + · · · (3.57)
2

51
where W0 , bij , ... are constants. Since the strain energy of the material has been
assumed to be zero in the undeformed state, W0 must be zero. Furthermore,
since e is symmetric we have that
1 1
W = bij eij + Cijkl eij ekl + ... = bji eij + Cijkl eij ekl + ...
2 2
1 1
= bij eij + Cjikl eij ekl + ... = bij eij + Cijlk eij ekl + ...
2 2
1
= bij eij + Cklij eij ekl + .... (3.58)
2
Therefore we conclude that

bij = bji
Cijkl = Cjikl = Cijlk = Cklij (3.59)

Noting that
∂(Cijkl eij ekl )
= 2Cmnkl ekl (3.60)
∂emn
we have that
∂W
σ ij = = bij + Cijkl ekl + ... (3.61)
∂eij
From Eq.(3.61) we conclude that the linear terms in the stress-strain relations
arise from the quadratic terms (in strains) of W. If the material is stress free in
the undeformed state, then bij = 0. It follows then that the linear constitutive
equation or the generalized Hooke’s law for an elastic solid is

σ ij = Cijkl ekl (3.62)


1
W = Cijkl eij ekl (3.63)
2
Remark 7 The existence of the strain energy function introduces an additional
symmetry condition
Cijkl = Cklij (3.64)
which reduces the number of independent constants of the tensor C from 36
to 21. Assuming the isotropic materials, the number of independent constants
reduces to two (λ, µ =Lamé constants).

Isotropic materials have the strongest possible symmetry (i.e., a complete


symmetry about a point). There are some weaker symmetries naturally oc-
curring (e.g., wood, crystals) or constructed (e.g., fiber-reinforced composites)
materials. For these materials the stress strain law must be form-invariant un-
der certain transformations. Solids that do not have point symmetry are called
anisotropic.

52
3.12 Anisotropic Materials
Let’s introduce the notation
σ 11 = σ 1 ; σ 22 = σ 2 ; σ 33 = σ 3 ; σ 23 = σ 4 ; σ 31 = σ 5 ; σ 12 = σ 6
e11 = e1 ; e22 = e2 ; e33 = σe; 2e23 = e4 ; 2e31 = e5 ; 2e12 = e6 (3.65)
Then the generalized Hooke’s law becomes
σ ij = cij ej ; i, j = 1, 2, 3, ..., 6 (3.66)
Obviously, the element cij are related to Cijkl and [cij ] is a 6-by-6 matrix. Since
Cijkl = Cklij, the so called stiffness matrix [cij ] is symmetric, i.e.,
cij = cji (3.67)
Therefore, there are 21 independent material constants.
With this new notation we have that
1 1
W = cij ei ej = σ i ei
2 2
∂W
σi = (3.68)
∂ei
In expanded from the strain energy function becomes
1
W = c11 e21 + c12 e1 e2 + c13 e1 e3 + c14 e1 e4 + c15 e1 e5 + c16 e1 e6
2
1
+ c22 e22 + c23 e2 e3 + c24 e2 e4 + c25 e2 e5 + c26 e2 e6
2
1
+ c33 e23 + c34 e3 e4 + c35 e3 e5 + c36 e3 e6
2
1
+ c44 e24 + c45 e4 e5 + c46 e4 e6
2
1
+ c55 e25 + c56 e5 e6 (3.69)
2
Most solids exhibit symmetry properties with respect to certain rotations
of the body or reflection about one or more planes. These properties further
reduce the number of elastic constants. Obviously, the elastic constants cij , in
general, depend upon the reference frame, since the stress components σ i and the
strain components ei vary with different coordinate systems. For certain solids
the elastic constants cij may remain invariant under a given transformation of
coordinates. It is this invariance property that determines the elastic symmetry
of the solid.

3.12.1 Material With One Plane of Symmetry


Say the plane of symmetry is the x1 x2 −plane. Then cij must be invariant under
the transformation
x01 = x1 ; x02 = x2 ; x03 = −x3 (3.70)

53
which defines the transformation matrix
 
1 0 0
L = 0 1 0  (3.71)
0 0 −1
and the strain tensor
 
e11 e12 −e13
e0 = LeLT =  • e22 −e23  (3.72)
• • e33
where • denote the symmetric part of the matrix. It follows than from Eq.
(3.72) that
e023 = −e23 ; e031 = −e31
with all the other components unchanged. Thus

e04 = −e4 ; e05 = −e5 ; e0i = ei ; i = 1, 2, 3, 6 (3.73)

The strain energy function W 0 is obtained from the right hand side of Eq.
3.69 by changing the signs of e4 and e5 . Since the material is symmetric about
the x1 x2 −plane, we must have that

W0 = W (3.74)

Consequently, we have that

c14 = c15 = c24 = c25 = c34 = c46 = c56 = 0 (3.75)

Therefore, we are left with 13 (=21-8) independent elastic constants. Conse-


quently, for an elastic material with x1 x2 −plane of symmetry, the generalized
Hooke’s law σ i = cij ej can be written as
    
σ1 c11 c12 c13 0 0 c16 e1
 σ 2   c12 c22 c23 0 0 c26   
     e2 
 σ 3   c13 c23 c33 0 0 c   e 
   36   3 
 σ4  =  0 0 0 c c 0   e  (3.76)
   44 45  4 
 σ5   0 0 0 c45 c55 0   e5 
σ6 c16 c26 c36 0 0 c66 e6

3.12.2 Orthotropic Materials


Definition 15 If a material has three mutually orthogonal planes of symmetry,
it is called orthotropic.

Wood is an example of such material. Say that the planes of symmetry are
the coordinate planes. Then, the symmetry about the x1 x2 −plane implies that
(see Eq. 3.75)

c14 = c24 = c34 = c56 = c15 = c25 = c35 = c46 = 0

54
Similarly, the symmetry about the x1 x3 −plane requires that

c16 = c26 = c36 = c45 = c15 = c25 = c35 = c46 = 0 (3.77)

Comparison of Eqns. 3.75 and 3.77 shows that Eq. 3.77 requires only four addi-
tional constants to vanish. Thus the number of independent constants reduces
to 9 (=13-4). Finally for material with symmetry about the x1 x3 −plane implies
that
e04 = −e4 ; e06 = −e6
and thus
c14 = c16 = c24 = c26 = c34 = c36 = c45 = c56 = 0 (3.78)
which bring no new conditions. Therefore, if x1 x2 − and x2 x3 −planes are planes
of symmetry, then x1 x3 −plane is also a plane of symmetry. The Hooke’s law
for orthotropic material is then
    
σ1 c11 c12 c13 0 0 0 e1
 σ 2   c12 c22 c23 0 0 0   
     e2 
 σ 3   c13 c23 c33 0 0 0   e3 
    
 σ4  =  0 0 0 c44 0 0    (3.79)
     e4 
 σ5   0 0 0 0 c55 0   e5 
σ6 0 0 0 0 0 c66 e6

It is obvious from Eq. 3.79 that e4 = e5 = e6 = 0 implies that σ 4 = σ 5 =


σ 6 = 0, which means that the principal directions of stress coincide with the
principal directions of strain. This is not so for monoclinic materials (see Eq.
3.76) where e4 = e5 = e6 = 0, implies that σ 6 = c16 e1 + c26 e2 + c36 e3 which
may not vanish.

3.12.3 Transversely Isotropic Materials


If an orthotropic solid exhibits symmetry with respect to arbitrary rotations
about one of the axes (such as the x3 -axis), then it is called transversely
isotropic. Say that the system {x0i } is obtained from the system {xi } by a
rotation about the x3 -axis for an angle θ. Then x0 = Lx, where
 
cos θ sin θ 0
L =  − sin θ cos θ 0 
0 0 1

Then cij must be invariant under the transformation L. Since

e0ij = lir ljs ers

55
we can calculate that

e011 = e11 cos2 θ + e22 sin2 θ + 2e12 sin θ cos θ


e022 = e11 sin2 θ + e22 cos2 θ − 2e12 sin θ cos θ
e033 = e33
e023 = e23 cos θ − e31 sin θ
e031 = e23 sin θ + e31 cos θ
e012 = −(e11 − e22 ) sin θ cos θ + e12 (cos2 θ − sin2 θ)

or

e01 = e1 cos2 θ + e2 sin2 θ + e6 sin θ cos θ


e02 = e1 sin2 θ + e2 cos2 θ − e6 sin θ cos θ
e03 = e3
e04 = e4 cos θ − e5 sin θ
e05 = e4 sin θ + e5 cos θ
e06 = −2(e1 − e2 ) sin θ cos θ + e6 (cos2 θ − sin2 θ) (3.80)

The strain energy function W 0 is obtained by replacing e0i with ei in Eq. 3.69. If
the solid is symmetric about an arbitrary rotation about the x3 −axis we must
have W 0 = W for an arbitrary θ. By equating the coefficient of e21 we get

c11 = c11 cos4 θ + c22 sin4 θ + 2(c12 + 2c66 ) sin2 θ cos2 θ

which can be rearranged to

c11 (1 + cos2 θ) = c22 (1 − cos2 θ) + 2(c12 + 2c66 ) cos2 θ (3.81)

for all θ. Therefore, we must have

c11 = c22 ; c11 = −c22 + 2(c12 + 2c66 ) (3.82)

which implies that


1
c66 =
(c11 − c12 ) (3.83)
2
By equating the coefficients of e1 e3 and e24 in W 0 = W it follows that

c13 e1 e3 = c13 e01 e03 + c23 e02 e03


= c13 (e1 cos2 θ + e2 sin2 θ + e6 sin θ cos θ)e3
+c23 (e1 sin2 θ + e2 cos2 θ − e6 sin θ cos θ)e3

and therefore

c13 = c13 cos2 θ + c23 sin2 θ


1 1 1
c44 = c44 cos2 θ + c55 sin2 θ
2 2 2

56
which implies that
c13 = c23 (3.84)
c44 = c55 (3.85)
Thus for transversely isotropic material we need only 9-4=5 independent
elastic constants. The Hooke’s law becomes
    
σ1 c11 c12 c13 0 0 0 e1
 σ2   c12 c11 c23 0 0 0  e2 
    
 σ3    
  =  c13 c23 c33 0 0 0  e3 
 σ4   0 0 0 c44 0 0  e4 
    
 σ5   0 0 0 0 c44 0  e5 
σ6 0 0 0 0 0 c66 e6
1
c66 = (c11 − c12 ) (3.86)
2

3.12.4 Isotropic Materials


Here, the elastic constants cij are independent of the orientation of the co-
ordinate axes. We start with the stress-strain relationship for an orthotropic
material. The symmetry with respect to rotation about the x3 −axis implies
that
1
c11 = c22 ; c66 = (c11 − c22 )
2
c13 = c23 ; c44 = c55 (3.87)
Similarly, the symmetry with respect to rotation about the x1 −axis requires
that
1
c22 = c33 ; c44 = (c22 − c33 )
2
c12 = c13 ; c55 = c66 (3.88)
By combining Eqns. 3.87 and 3.88 results in
c11 = c22 = c33
c12 = c13 = c23 = λ
1
c44 = c55 = c66 = (c11 − c12 ) = µ
2
which gives
c11 = c22 = c33 = λ + 2µ (3.89)
Therefore, Hooke’s law for an isotropic solid is given by
    
σ1 λ + 2µ λ λ 0 0 0 e1
 σ2   λ λ + 2µ λ 0 0 0  e2 
    
 σ3   λ λ λ + 2µ 0 0 0  e3 
    
 σ4  =  0 0 0 µ 0 0  e4  (3.90)
    
 σ5   0 0 0 0 µ 0  e5 
σ6 0 0 0 0 0 µ e6

57
Based on this result we have

σ1 = λ(e1 + e1 + e1 ) + 2µe1
σ2 = λ(e1 + e1 + e1 ) + 2µe2
σ3 = λ(e1 + e1 + e1 ) + 2µe3
σ4 = µe4
σ5 = µe5
σ6 = µe6

or
σ ij = λekk δ ij + 2µeij (3.91)
This result can be written as

σ ij = Cijkl ekl (3.92)


Cijkl = λδ ij δ kl + µ(δ ik δ jl + δ il δ jk ) (3.93)

The strain energy function then becomes


1 1
W = σij eij = λ(ekk )2 + µeij eij
2 2
1
= λ(ekk )2 + µ(e21 + e22 + e23 + 2e223 + 2e213 + 2e212 ) (3.94)
2

3.13 Infinitesimal Strains in Cylindrical and Spher-


ical Coordinates
Recall that the elements of strain tensor are defined by
1
eij = (ui,j + uj,i )
2
Then, the strain dyadic is given by
1
e = eij ei ej = (uj,i ei ej + ui,j ei ej ) (3.95)
2
Since

uj,i ei ej = ei (uj ej ) = ∇u
∂xi
∂ T
ui,j ei ej = (ui,j ej ei )T = [ej (ui ei )]T = (∇u) (3.96)
∂xj
we have that
1 T
e=
(∇u+(∇u) ) (3.97)
2
This equation can be used to obtain the strain-displacement relations in various
curvilinear coordinate systems.

58
3.13.1 Cylindrical Coordinates (r, θ, z)
The displacement vector is then written as

u =ur er + uϕ eϕ + uz ez

and the components of the infinitesimal strain dyadic e are given by


 
err erθ erz
 eθr eθθ eθz  (3.98)
ezr ezθ ezz

By recalling that
∂ ∂ ∂
∇u = (er + eϕ + ez )(ur er + uϕ eϕ + uz ez )
∂r r∂ϕ ∂z
we have
∂ur 1 ∂uϕ ∂uz
∇u = er er + (ur + )eϕ eϕ + ez ez
∂r r ∂ϕ ∂z
1 ∂uz ∂uϕ ∂ur ∂uz
+ eϕ ez + ez eϕ + ez er + er ez (3.99)
r ∂ϕ ∂z ∂z ∂r
∂uϕ 1 ∂ur
+ er eϕ + ( − uϕ )eϕ er
∂r r ∂ϕ
By noting, for example, that eθz = eθ · e · ez , Eqns. 3.97 and 3.99 lead to the
following results
∂ur 1 ∂uθ ∂uz
err = ; eθθ = ( + ur ); ezz =
∂r r ∂θ ∂z
1 ∂uθ 1 ∂uz
eθz = ezθ = ( + )
2 ∂z r ∂θ
1 ∂ur ∂uz
ezr = erz = ( + )
2 ∂z ∂r
1 ∂ur ∂uθ uθ
erθ = eθr = ( + − ) (3.100)
2 r∂θ ∂r r

3.13.2 Spherical Coordinates (R, θ, ϕ)


Now the displacement vector is given by

uR eR + uθ eθ + uϕ eϕ

and the components of the infinitesimal strain dyadic are given by


 
eRR eRθ eRϕ
 eθR eθθ eθϕ  (3.101)
eϕR eϕθ eϕϕ

59
Since
∂uR 1 ∂uθ 1 ∂uϕ
∇u = eR eR + (uR + )eθ eθ + (sin θuR + cos θuθ + )eϕ eϕ
∂R R ∂θ R sin θ ∂ϕ
1 ∂uϕ 1 ∂uθ 1 ∂uR
+ eθ eϕ + ( − cos θuϕ )eϕ eθ + ( − sin θuϕ )eϕ eR
R ∂θ R sin θ ∂ϕ R sin θ ∂ϕ
∂uϕ ∂uθ 1 ∂uR
+ eR eϕ + eR eθ + ( − uθ )eθ eR (3.102)
∂R ∂R R ∂θ
and by using Eq. 3.97 we get
∂uR 1 ∂uθ 1 ∂uϕ
eRR = ; eθθ = ( + uR ); eϕϕ = [ + sin θuR + cos θuθ ]
∂R R ∂θ R sin θ ∂ϕ
1 1 ∂uθ ∂uϕ
eθϕ = eϕθ = ( + − cot θuϕ )
2R sin θ ∂ϕ ∂θ
1 1 ∂uR ∂uϕ uϕ
eϕR = eRϕ = ( + − )
2 R sin θ ∂ϕ ∂R R
1 1 ∂uR ∂uθ uθ
eRθ = eθR = ( + − ) (3.103)
2 R ∂θ ∂R R
This procedure can be extended to other curvilinear coordinate systems.

3.14 Constitutive Equations in Cylindrical and


Spherical Coordinates
In dyadic notation, we can write generalized Hokke’s law for isotropic materials
as
σ = λdI + 2µe (3.104)
where σ, I, and e are stress, unit, and strain dyadics, respectively and

d = ekk = ∇ · u (3.105)

Using Eq. 3.97 then produces


T
σ = λdI + µ(∇u+(∇u) ) (3.106)

or
σ ij ei ej = λdδ ij ei ej + 2µeij ei ej (3.107)

60
Then by comparing the coefficients of the like terms ei ej of Eq. 3.107produces
in cylindrical coordinates (r, θ, z)
∂ur
σ rr = λd + 2µerr = λd + 2µ
∂r
2µ ∂uθ
σ θθ = λd + 2µezz = λd + ( + ur )
r ∂θ
∂uz
σ zz = λd + 2µezz = λd +
∂z
∂ur ∂uz
σ zr = σ rz = 2µezr = µ( + )
∂z ∂r
∂uθ 1 ∂uz
σ θz = σ zθ = 2µeθz = µ( + )
∂z r ∂θ
1 ∂ur ∂uθ uθ
σ rθ = σ θr = 2µerθ = µ( + − ) (3.108)
r ∂θ ∂r r
where
∂ur ur 1 ∂uθ ∂uz
d=∇·u= + + + (3.109)
∂r r r ∂θ ∂z
Similar approach in spherical coordinate system results in
∂uR
σRR = λd + 2µeRR = λd + 2µ
∂R
2µ ∂uθ
σ θθ = λd + 2µeθθ = λd + ( + uR )
R ∂θ
2µ ∂uϕ
σ ϕϕ = λd + 2µeϕϕ = λd + ( + sin θuR cos θuθ )
R sin θ ∂ϕ
µ 1 ∂uθ ∂uϕ
σ θϕ = 2µeθϕ = ( + − cot θuϕ )
R sin θ ∂ϕ ∂θ
1 ∂uR ∂uϕ uϕ
σ ϕR = 2µeϕR = µ( + − )
R sin θ ∂ϕ ∂R R
1 ∂uR ∂uθ uθ
σ Rθ = 2µeRθ = µ( + − ) (3.110)
R ∂θ ∂R R
where
∂uR 2uR 1 ∂uθ 1 ∂uϕ
d=∇·u= + ( + cot θuθ ) + (3.111)
∂R R R ∂θ R sin θ ∂ϕ

3.15 Dilatation d
Let us consider the physical interpretation of d = ekk . For that purpose let’s
consider a cube of size 1 × 2 × 3 before deformation. Let e11 , e22 , and e33 be
the principal strains, i.e., eij = 0 for i 6= j. Then, after the deformation we have
that
1 → 1+ 1 e11

2 → 2+ 2 e22

3 → 3+ 3 e33

61
Remark 8 The fibers of the cube will remain perpendicular to each other.

The initial and final volumes of the cube are given by

V0 = 1 2 3
V = (1 + e11 )(1 + e22 )(1 + e33 )V0

Then the change in volume is

V − V0 = V0 (e11 + e22 + e33 + ...)

and by neglecting the higher order terms (infinitesimal strain theory) we obtain
the relative change of volume to be
V − V0
= ekk = d (3.112)
V0
Thus the dilatation represents the relative change of volume of the material
during the deformation.

62
Chapter 4

Boundary Value Problems


of Linear Elasticity

Consider an elastic solid of volume V and boundary B = Bu ∪ Bσ (see Fig. 4.1)


where Bu and Bσ denote the portions of the boundary where displacement or
stress field is being prescribed, respectively.

Then the equilibrium equations are given by

σ ij,j + fi = 0; x ∈V (4.1)

where f denotes body force per unit volume. Constitutive equations are given
by
σ ij = Cijkl ekl (4.2)
where the tensor of elastic properties satisfies the following symmetry relations

Cijkl = Cjikl = Cijlk = Cklij (4.3)

Strain-displacement relations are given by


1
eij = (ui,j + uj,i ) (4.4)
2
while the compatibility conditions are prescribed by

eij,kl + ekl,ij − eik,jl − ejl,ik = 0 (4.5)

The boundary conditions are specified in the following form

ui = ui (B); x ∈Bu (4.6)


σ ij nj = Tin (B); x ∈Bσ (4.7)

Definition 16 If only displacements are prescribed on the boundary B, Eqns.


4.1 to 4.7 define the displacement boundary value problem.

63
T n
Bs

/V=B Bu

Figure 4.1: Boundary value problem of elasticity.

Definition 17 If only tractions are prescribed on the boundary B, Eqns. 4.1


to 4.7 define the stress boundary value problem,

Definition 18 If both displacements and tractions are prescribed on the bound-


ary B, Eqns. 4.1 to 4.7 define the mixed boundary value problem.

At an in interface C between two different materials, certain continuity con-


ditions must be enforced. For a perfect bond both displacement and traction
fields must be continuous across the interface, i.e.,
(1) (2)
ui = ui ; x ∈C (4.8)
n(1) n(2)
Ti = Ti ; x ∈C (4.9)

For isotropic materials we have that

Cijkl = λδ ij δ kl + µ(δ ik δ jl + δ il δ jk )

which results in the constitutive equations of the form

σ ij = λekk δ ij + 2µeij (4.10)

Remark 9 Note that


2
σ kk = (3λ + 2µ)ekk = 3(λ + µ)ekk
3

64
or
s = 3κd (4.11)
where
s = σ kk
d = ekk
2
κ = λ+ µ (4.12)
3
with κ being the bulk modulus.
Remark 10 The inverse form of the constitutive equation 4.10 is given by
1+ν ν
eij = σ ij − sδ ij (4.13)
E E
In cylindrical coordinates (r, θ, z) the equations of equilibrium are
∂σ rr 1 ∂σ rθ ∂σ zr 1
+ + + (σ rr − σ θθ ) + ρbr = 0
∂r r ∂θ ∂z r
∂σ rθ 1 ∂σ θθ ∂σ θz 2
+ + + σ rθ + ρbθ = 0 (4.14)
∂r r ∂θ ∂z r
∂σ rz 1 ∂σ θz ∂σ zz 1
+ + + σ rz + ρbz = 0
∂r r ∂θ ∂z r
The stress-strain relationships are
σrr = λd + 2µerr ; σθθ = λd + 2µezz ; σ zz = λd + 2µezz
σzr = σ rz = 2µezr ; σ θz = σ zθ = 2µeθz ; σ rθ = σθr = 2µerθ (4.15)
where d = ∇ · u = err + eθθ + ezz .
The strain displacement equations are given by Eq. 3.100.
In spherical coordinates (R, θ, ϕ) the equations of equilibrium are given by
∂σ RR 1 ∂σθR 1 ∂σ ϕR 1
+ + + (2σ RR − σ θθ − σ ϕϕ + σ θR cot θ) + fR = 0
∂R R ∂θ R sin θ ∂ϕ R
∂σ Rθ 1 ∂σ θθ 1 ∂σ ϕθ 1
+ + + [(σ θθ − σ ϕϕ ) cot θ + 3σ rθ ] + fθ (4.16)
= 0
∂R R ∂θ R sin θ ∂ϕ R
∂σ Rϕ 1 ∂σ θϕ 1 ∂σ ϕϕ 1
+ + + (2σ θϕ cot θ + 3σ Rϕ ) + fϕ = 0
∂R R ∂θ R sin θ ∂ϕ R
The stress-strain and strain-displacement relations are given by Eq. 3.110.

4.1 Navier’s Equations


Equations of equilibrium are given by Eq. 4.1. Then by using constitutive
equations for an isotropic solid (Eq. 4.10) together with definition of strain the
equilibrium equations become
(λ + µ)uj,ji + µui,jj + fi = 0 (4.17)

65
or
(λ + µ)∇(∇ · u)+µ∇2 u + f = 0 (4.18)
These equations are known as Navier’s equations. In cylindrical and spherical
coordinates we can utilize the expressions for ∇f, ∇ · u, and ∇2 f in order to
obtain Navier’s equations for these curvilinear coordinates.

4.2 Uniqueness of Solution


Theorem 9 Let the tractions be prescribed over a part Bσ of the boundary
B = ∂V and displacements be prescribed over the remaining part of Bu (note
Bσ ∪ Bu = B). Then the solution of elastostatic problem is unique.σ rθ = σθr =
2µerθ

Proof. Assume that two solutions u(1) and u(2) are possible. Then we have
that
(1) (2)
σ ij,j + fi = 0; σij,j + fi = 0
where
(1)(2) (1)(2)
σ ij = Cijkl uk,m
The boundary conditions are given by
(1) (2)
ui = ui = ui (B) ; x ∈Bu
(1) (2)
σ ij nj = σ ij nj = ti (B); x ∈Bσ

Let us define
(1) (2)
ui ≡ ui − ui
σ ij = Cijkm uk,m

Then it is easy to show that

σ ij,j = 0; x ∈V
ui = 0; x ∈Bu
σij nj = 0; x ∈Bσ

Let W be the strain energy function corresponding to the displacement ui . Then


Z Z Z
2 W dV = σ ij eij dV = σ ij ui,j dV
V
ZV V
Z
= (σ ij ui ),j dV − σ ij,j ui dV
V V
Z Z
= σ ij ui nj dB − σ ij,j ui dV = 0
B V

since σ ij,j = 0; x ∈V and σij nj = 0; x ∈Bσ while ui = 0; x ∈Bu .

66
R
Since W is nonnegative, the fact that V W dV = 0, implies that W = 0 in
V. Since W is positive definite quadratic form in strain components, it cannot
be zero unless all the strain components are zero. Thus eij = 0 implies that
(1) (2)
ui ≡ ui − ui represents a rigid body motion. Since ui = 0 on Bu we must
(1) (2)
have that ui = 0 everywhere in the body and thus we have that ui = ui .
Remark 11 This proof applies to the linear case only.
Theorem 10 (Clapeyron’s) If a solid is in equilibrium under the action of a
given system of body and surface forces, then the strain energy of deformation
is equal to one-half the static work that would be done by the external forces.
Proof. From the uniqueness proof we have that
Z Z Z
2 W dV = σ ij eij dV = σ ij ui,j dV
V V V
Z Z
= (σ ij ui ),j dV − σij,j ui dV
ZV ZV
= σ ij ui nj dB + fi ui dV
ZB Z V
= Tin ui dB + fi ui dV
B V
Z Z
= Tn · udB + f · udV
B V

which completes the proof.


Theorem 11 (Betti-Rayleigh Reciprocity (BRR) Theorem)If an elastic body is
subject to two systems of body and surface forces, then the work done by the first
system of forces acting through the displacements of the second system is equal
to work done by the second system of forces acting through the displacements of
the first system.
Proof. Let us consider two equilibrium states: u(1) due to body forces
f and tractions t(1) and u(2) due to f (2) and t(2) . Then the work Ω done by
(1)

the forces f (1) and tractions t(1) if they acted through the displacement u(2) is
Z Z
(1) (2) (1) (2)
Ω = fi ui dV + ti ui dB
V B
Z Z
(1) (2) (1) (2)
= fi ui dV + σij nj ui dB
ZV ZB
(1) (2) (1) (2)
= fi ui dV + (σij ui ),j dV
ZV V
Z
(1) (1) (2) (1) (2)
= (f + σ ij,j )ui dV + σ ij ui,j dV
V V
Z
(1) (2)
= σ ij ui,j dV (4.19)
V

67
Thus Z Z
1 (1) (2) (2) (1) (2)
Ω= σ ij (ui,j + uj,i )dV = Cijkl ekl eij dV (4.20)
2 V V
(1) (2) (1)
Since Cijkl = Cklij , then the integrand Cijkl ekl eij is symmetric in ekl and
(2)
eij . Thus we have that
Z Z
(2) (1) (2) (1)
Ω = Cijkl ekl eij dV = σ ij ui,j dV
V V
Z Z
(2) (1) (2) (1)
= fi ui dV + ti ui dB (4.21)
V B

By comparing Eqns. 4.19 and 4.21 we obtain that


Z Z Z Z
(1) (2) (1) (2) (2) (1) (2) (1)
fi ui dV + ti ui dB = fi ui dV + ti ui dB (4.22)
V B V B

or equivalently Z Z
(1) (2) (2) (1)
σ ij ui,j dV = σij ui,j dV (4.23)
V V
which completes the proof.

Example 11 Using Betty-Rayleigh resiprocity theorem derive the change of vol-


ume for an elastic body subjected to a body force and surface tractions.
(1)
Solution 11 Let ui = Axi be the solution of a problem of an isotropic solid
(1)
of arbitrary volume with no body forces subjected to the traction ti = 3κAni ,
where κ = λ + 2µ/3 ≡bulk modulus. Then it follows that
(1)
eij = Aδ ij ⇒ σ ij = 3κAδ ij

Let ui , eij , fi and ti denote a second possible state of the body. Then by putting
(1)
fi = 0 and using the Betty-Rayleigh reciprocity theorem we obtain
Z Z Z
(1) (2) (2) (1) (2) (1)
σij eij dV = fi ui dV + ti ui dB
V V B

(2) (2) (2)


By substituting eij = eij , fi = fi and ti
= ti we get
Z Z Z
3κeii dV = fi xi dV + ti xi dB
V V B
R
But V
eii dV = ∆V ≡ increase in volume due to deformation, thus
Z Z
1
∆V = { fi xi dV + ti xi dS}
3κ V S

which represents increase in volume of an elastic body due to the body force fi
and the surface traction ti .

68
4.3 Exact Solutions for Some Problems
For some problems in elastostatics, Navier’s equations can be integrated directly.
The integration constants can be found by using the boundary conditions.
Example 12 Consider a cylindrical tube of inner and outer radii a and b, re-
spectively subjected to internal pressure p and the external pressure q. Solve for
the unknown displacement field in the tube.
Solution 12 Assume that there is no displacement in the z − direction, i.e.
uz ≡ 0. Due to symmetry we have
ur = u(r); uθ = uz = 0
The boundary conditions are then given by
σ rr (r = a) = −p; σrr (r = b) = −q
Recall that the equations of equilibrium are given by (no body forces)
(λ + µ)∇(∇ · u) + µ∇2 u = 0
Furthermore, in cylindrical coordinates (r, θ, z) we have that
∂ ∂ ∂
∇f = (er + eθ + ez )f
∂r r∂θ ∂z
∂ur ur 1 ∂uθ ∂uz 1
∇·u = + + = u,r + u
∂r r r ∂θ ∂z r
u
∇(∇ · u) = er (u,r + ),r
r
We still need the term ∇2 u in cylindrical coordinates. In general we have that
∂ ∂ ∂
∇2 u = ∇ · ∇u = (er + eθ + ez ) · (∇u)
∂r r∂θ ∂z
However
∂ur 1 ∂uϕ ∂uz
∇u = er er + (ur + )eϕ eϕ + ez ez
∂r r ∂ϕ ∂z
1 ∂uz ∂uϕ ∂ur ∂uz
+ eϕ ez + ez eϕ + ez er + er ez
r ∂ϕ ∂z ∂z ∂r
∂uϕ 1 ∂ur
+ er eϕ + ( − uϕ )eϕ er
∂r r ∂ϕ
therefore,
∂ 1
∇u = ur,rr er er + [ (ur + uθ,θ )],r eθ eθ + uz,zr ez ez
∂r r
1
+( uz,θ ),r eθ ez + uθ,zr ez eθ + ur,zr ez er + uz,rr er ez
r
1
+uθ,rr er eθ + [ (ur,θ − uθ )],r eθ er
r

69
Consequently

er · (∇u) =ur,rr er + uz,rr ez + uθ eθ
∂r
∂ ∂ ∂
where we have used the fact that ∂r er = ∂r eθ = ∂r ez = 0. Similarly we get


ez · (∇u) = uz,zz ez + uθ,zz eθ + ur,zz er
∂z
1 ∂ 1 ur 2 1
eθ · (∇u) = ( ur,r − 2 − 2 uθθ + 2 ur,θθ )er
r ∂θ r r r r
2 1 1 1
+( 2 ur,θ + uθ,r − 2 uθ + 2 uθ,θθ )eθ
r r r r
1 1
+( 2 uz,θθ + uz,r )ez
r r
By collecting the last three results we obtain
1 ur 2 1
∇2 u = (ur,rr + ur,r − 2 − 2 uθ,θ + 2 ur,θθ + ur,zz )er
r r r r
1 1 1 2
( 2 uθ,θθ + uθ,r − 2 uθ + 2 ur,θ + uθ,rr + uθ,zz )eθ
r r r r
1 1
+(uz,zz + uz,r + 2 uz,θθ + uz,rr )ez
r r
For axisymmetric case u =ur er = u(r)er , and ∂/∂θ = ∂/∂z ≡ 0 and we get
1 ur 1
∇2 u = (ur,rr + ur,r − 2 )er = (u,r + u),r er
r r r
Therefore, the equation of equilibrium for the cylinder problem becomes
d du 1
(λ + 2µ) ( + u) = 0
dr dr r
which can be solved for displacement
B
u = Ar +
r
Corresponding stress field follows from the equation σ rr = λd + 2µerr , where
du u
d = ∇·u= + = 2A
dr r
du B
err = =A− 2
dr r
Thus
B
σ rr = 2(λ + µ)A − 2µ
r2

70
Similarly
B
σ θθ = λd + 2µeθθ = 2(λ + µ)A + 2µ
r2
u B
eθθ = =A+ 2
r r
σ zz = λd = ν(σrr + σ θθ )
σ rr + σ θθ
σ rr + σ θθ = 2(λ + µ)d ⇒ d =
2(λ + µ)
Then the boundary conditions imply
B
σ rr (a) = −p ⇒ 2(λ + µ)A − 2µ = −p
a2
B
σrr (b) = −q ⇒ 2(λ + µ)A − 2µ 2 = −q
b
which can be solved for
pa2 − qb2 a2 b2 (p − q)
A= 2 2
;B =
2(λ + µ)(b − a ) 2µ(b2 − a2 )
Thus the displacement and stress fields are given by
B
u = Ar +
r
pa2 − qb2 a2 b2 (p − q)
σ rr = −
b2 − a2 (b2 − a2 )r2
pa2 − qb2 a2 b2 (p − q)
σθθ = +
b2 − a2 (b2 − a2 )r2
pa2 − qb2
σ zz = 2ν 2
b − a2
If there is no external pressure (q = 0) we have
pa2 b2
σ rr = (1 − ) ⇒ compression
b2 − a2 r2
pa2 b2
σ θθ = (1 + ) ⇒ tension
b2 − a2 r2
The maximum (minimum) stresses are then
a2 + b2
σmax = σ θθ (r = a) = p >p
b2 − a2
2a2 p
σmin = σ θθ (r = b) = 2 < σ min
b − a2
For thin cylinders d = b − a << 1 and a, b and r are nearly equal. Thus we have
pa2 b2 pa2 b2 pa
σθθ = (1 + ) = (1 + )≈
b2 − a2 r2 d(b + a) r2 d

71
1
which implies that σ θθ ∝ d. For thick pipes on the other hand we have that
b >> a which implies that

pa2 − qb2 q
A = 2 2
≈−
2(λ + µ)(b − a ) 2(λ + µ)
a2 b2 (p − q) a2 (p − q)
B = ≈
2µ(b2 − a2 ) 2µ

and thus
q p − q a2
u = − r+
2(λ + µ) 2µ r
2
a
σ rr = −q − (p − q) 2
r
a2
σ θθ = −q + (p − q) 2
r
For p = 0 we get that σ θθ (r = a) = −2q, i.e. twice the stress without the hole.
Finally, for r >> a we have that
qr
u = −
2(λ + µ)
σ rr = σθθ = −q

Example 13 Determine the stress in a rotating shaft (ω = const. and no lon-


gitudinal deformation).

Solution 13 Due to symmetry, ur = u(r), and uθ = uz = 0. The Navier’s


equation becomes
d du 1
(λ + 2µ) ( + u) + ρω 2 r = 0
dr dr r
Let
ρω 2
C≡
λ + 2µ
then the equation of equilibrium becomes
d du 1
( + u) = −Cr
dr dr r
which can be solved for
B r3
u = Ar + −C
r 8

72
For a bounded solution at r = 0 we choose B = 0. Thus
r3
u = Ar − C
8
du u 1
d = ∇·u= + = 2A − Cr2
dr r 2
du C
σ rr = λd + 2µerr = λd + 2µ = 2(λ + µ)A − (2λ + 3µ)r2
dr 4
u C
σ θθ = λd + 2µeθθ = λd + 2µ = 2(λ + µ)A − (2λ + µ)r2
r 4
C 2
σ zz = λd = 2λA − λr
2
Now the boundary condition at the surface of the shaft is given by
C
σ rr (r = a) = 0 ⇒ 2(λ + µ)A − (2λ + 3µ)a2 = 0 ⇒
4
2λ + 3µ
A = Ca2
8(λ + µ)
Consequently, the displacement and the stress fields are given by
Cr 2λ + 3µ 2
u = ( a − r2 )
8 λ+µ
C
σrr = (2λ + 3µ)(a2 − r2 )
4
C 2λ + 3λ 2
σ θθ = (2λ + µ)( a − r2 )
4 2λ + µ
λC 2λ + 3λ 2
σ zz = ( a − r2 )
2 2λ + µ
The the displacement and stress fields on the surface of the shaft are

Ca3
u(r = a) = (λ + 2µ)
8(λ + µ)
Ca2 µ
σ θθ (r = a) =
2
2 λµ
σ zz (r = a) = a
4(λ + µ)
Maximum stress at r = 0 is then
1 2
σrr (0) = σ θθ (0) = Ca (2λ + 3µ)
4

Example 14 Consider a thin plate in x1 x2 −plate with an elastic circular in-


clusion loaded as shown in Fig. 4.3. Determine the unknown displacement and
stress fields in the plate and the inclusion.

73
s 0

r
q
s 0 s
l,m 2 2 a 0

l,m 1 1

s 0

Figure 4.2: An elastic plate with a circular inclusion loaded at infinity.

74
Solution 14 Due to axisymmetry u =u(r)er and the solution of the Navier’s
equation becomes
Bi
u(i) = Ai r + ; i = 1, 2
r
so that the stress field is given by

(i)
2µi Bi
σ rr = 2(λi + µi )Ai −
r2
(i) 2µi Bi
σ θθ = 2(λi + µi )Ai +
r2
(i) (i) (i)
σ zz = ν i (σ rr + σ θθ ) = 4ν i (λi + µi )Ai

where Ai , Bi ; i = 1, 2 are the unknowns. Requirement of finite displacement at


(1)
r = 0 requires that B = 0. In addition as r → ∞, we must have that σ rr → σ 0
or that
σ0
2(λ1 + µ1 )A1 = σ 0 ⇒ A1 =
2(λ1 + µ1 )
Continuity of displacement and traction fields at r = a then is expressed as

u(1) (a) = u(2) (a)


t(1) (2) (1) (2)
r (a) = tr (a) ⇔ σ rr (a) = σ rr (a)

which gives
B1
A1 a + = A2 a
a
2µ1
2(λ1 + µ1 )A1 − B1 = 2(λ2 + µ2 )A2
a2
The last two equations can be solved for the unknowns

σ0 (λ1 + 2µ1 )
A2 =
(λ1 + µ1 )(µ1 + λ2 + µ2 )
λ1 + µ1 − λ2 − µ2
B1 = σ 0 a2
(λ1 + µ1 )(µ1 + λ2 + µ2 )

Therefore, we can evaluate both displacement and stress fields throughout the
medium. In particular, the interface stresses between the plate and the inclusion
are calculated to be

σ (1) (2)
rr (a) = σ rr (a) = 2(λ2 + µ2 )A2
(1) 2µ B1
σ θθ (a) = 2(λ1 + µ1 )A1 + 12
a
(i)
σzz = 4ν i (λi + µi )Ai ; i = 1, 2

Therefore, σ θθ and σ zz have discontinuity at the interface r = a.

75
Example 15 Consider a spherical shell a ≤ R ≤ b subjected to internal pres-
sure p and external pressure q. Thus the boundary conditions are specified by

σ RR (R = a) = −p
σ RR (R = b) = −q

Due to point symmetry uR = u(R); uθ = uφ ≡ 0. For the Navier’s equations (no


body forces)
(λ + µ)∇(∇ • u) + µ∇2 u = 0
we have for spherical coordinates with point symmetry that
du 2u d du 2u
∇•u= + ⇒ ∇(∇ • u) = eR ( + )
dR R dR dR R
We still need the ∇2 u−term. Thus we consider
µ ¶
∂ ∂ 1 ∂
∇2 u =∇ • ∇u = eR + eθ + eφ • (∇u)
∂R R∂θ R sin θ ∂φ

But
u u
∇u =u,R eR eR + eθ eθ + eφ eφ
R R
and thus
∂ d du
eR • (∇u) = eR
∂R dR dR
∂ 1 du u
eθ • (∇u) = ( − )eR
R∂θ R dR R2
1 ∂ 1 du u
eφ • (∇u) = ( − )eR
R sin θ ∂φ R dR R2
Consequently we obtain

d2 u 1 du u
∇2 u = [ 2
+ 2( − 2 )]
dR R dR R
d du u
= ( +2 )
dR dR R
Therefore, the Navier’s equations reduce to
d du u
( +2 )=0
dR dR R
Consequently, it follows that the displacement field is given by
B
u = AR +
R2

76
and the corresponding stress filed is
du
σ RR = λd + 2µeRR = λ∇ • u + 2µ
dR
du u du
= λ( + 2 ) + 2µ
dR R dR
B
= (3λ + 2µ)A − 4µ 3
R
du u u
σθθ = λd + 2µeθθ = λ( + 2 ) + 2µ
dR R R
B
= (3λ + 2µ)A + 2µ 3 (4.24)
R
du u u
σ φφ = λd + 2µeφφ = λ( + 2 ) + 2µ (4.25)
dR R R
B
= (3λ + 2µ)A + 2µ 3 (4.26)
R
Then from the boundary conditions we have that
B
(3λ + 2µ)A − 4µ = −p
a3
B
(3λ + 2µ)A − 4µ 3 = −q
b
from which it follows that
pa3 − qb3
A =
(3λ + 2µ)(b3 − a3 )
(p − q)a3 b3
B =
4µ(b3 − a3 )
Corresponding displacement and stress fields are then
pa3 − qb3 (p − q)a3 b3 1
u = 3 3
R+
(3λ + 2µ)(b − a ) 4µ(b3 − a3 ) R2
pa3 − qb3 (p − q)a3 b3 1
σ RR = −
(b3 − a3 ) (b3 − a3 ) R3
pa3 − qb3 (p − q)a3 b3 1
σθθ = σφφ = 3 +
(b − a3 ) 2(b3 − a3 ) R3
If q = 0 we obtain
pa3 4µ b3
u = 3 3
( R + 2)
4µ(b − a ) 3λ + 2µ R
3 3
pa b
σ RR = − 3 ( − 1) ≤ 0
b − a3 R3
pa3 b3
σθθ = σφφ = 3 (1 + )>0
b − a3 2R3

77
The maximum stresses at R = a are
pa3 b3
σ RR−Max = σ φφ−Max = (1 + )
b3 − a3 2a3
For a thin shell, d = b − a << a it follows that
pa3 b3 pa
σ θθ = σ φφ = 2 2
(1 + 3
)≈
(a + ab + b )d 2R 2d
For a very thick sphere, b >> a, we have that
qR (p − q)a3
u = − +
3λ + 2µ 4µR2
3
a
σRR ≈ −q − (p − q) 3
R
a3
σ θθ = σ φφ ≈ −q + (p − q) 3
2R
Consequently at R = a we have that
1
σRR = −p; σ θθ = σ φφ = (p − 3q)
2
Thus for a thick sphere and no internal pressure (p = 0) we have that
3
σθθ = − q
2
demonstrating that there is a stress concentration taking place at the cavity.

4.4 Special Models


4.4.1 1-D Models
Definition 19 If the body force and the components of the stress tensor depend
upon one spatial variable, say x1 , we talk about 1-D models.
Equation of equilibrium then becomes
σ i1,1 + fi = 0 (4.27)
and we distinguish three cases:
1. Longitudinal Strain
u1 = u1 (x1 ); u2 = u3 = 0 (4.28)
Then the strain and stress tensors are given by
 
u1,1 0 0
e =  0 0 0 
0 0 0
 
(λ + 2µ)u1,1 0 0
σ =  0 λu1,1 0  (4.29)
0 0 λu1,1

78
Equation of equilibrium is given by
(λ + 2µ)u1,11 + fi = 0 (4.30)

2. Longitudinal Stress
σ 11 = σ 11 (x1 ); σ ij = 0; i, j 6= 1 (4.31)
Thus the stress tensor is defined by
 
σ 11 0 0
σ= 0 0 0  (4.32)
0 0 0
Since
σ 22 = λekk + 2µe22 = 0
σ 33 = λekk + 2µe33 = 0
it is easy to show that
e22 = e33 (4.33)
The from the condition σ 22 = 0 it follows that
λ
e22 = e33 = − e11 = νe11 (4.34)
2 (λ + µ)
where ν denotes the Poisson’s ratio. Finally, the stress filed is described
by
3λ + 2µ
σ 11 = µ e11 = Ee11 (4.35)
λ+µ
where E denotes Young’s modulus.
3. Shear, defined in terms of displacement field which is in the plane perpen-
dicular to the x1 −axis
u =u2 (x1 )e2 + u3 (x1 )e3 (4.36)
Then the strain and stress tensors are given by
 1 1

0 2 u2,1 2 u3,1
e =  2 u2,1
1
0 0 
1
u3,1 0 0
 2 
0 µu2,1 µu3,1
=  µu2,1 0 0  (4.37)
µu3,1 0 0
The Navier’s equation reduce to two uncoupled equations
λu2,11 + f2 = 0
λu3,11 + f3 = 0 (4.38)

79
4.4.2 Two-dimensional Problems
Definition 20 The body forces and stresses are independent of one spatial co-
ordinate, say x3 .

Therefore, ∂x3 ≡ 0 and the Navier’s equations reduces to

σ3β,β + f3 = 0 (4.39)
σ αβ,β + fα = 0; α, β = 1, 2 (4.40)

or

σ31,1 + σ 32,2 + f3 = 0 (4.41)


σ11,1 + σ 12,2 + f1 = 0 (4.42)
σ21,1 + σ 22,2 + f2 = 0 (4.43)

It can be seen that Eqn. 4.41 is decoupled from the Eqns. 4.42 and 4.43.

Antiplane Strain Model


Definition 21 In this model the displacement field is assumed to be of the form

u3 = u3 (x1 , x2 ); ui = 0; i = 1, 2 (4.44)

Physically, this corresponds to an infinitely long cylinder (along the x3 −axis


loaded uniformly with tractions in the x−direction. Then the corresponding
strain and stress tensors are
 1

0 0 2 u3,1
e =  0 0 1
2 u3,2
 (4.45)
1 1
u3,1 u
2 3,2 0
 2 
0 0 µu3,1
σ =  0 0 µu3,2  (4.46)
µu3,1 µu3,2 0

Navier’s equations then reduce to a single equation

σ 3α,α + f3 = 0 (4.47)

or
µ∇2 u3 (x1 , x2 ) + f3 = 0 (4.48)
Now the boundary conditions follow from the Cauchy’s law in the following
form
∂u3
T3n = µ(∇u3 ) • n =µ (4.49)
∂n
where n denotes an outward unit normal on the boundary surface S.

Example 16 State the boundary conditions for antiplane strain model depicted
by Fig. ??.

80
s 0

r
q
s 0 s
l,m 2 2 a 0

l,m 1 1

s 0

Figure 4.3: An antiplane strain BVP.

Solution 15 For different surfaces the boundary conditions are stated as fol-
lows:
∂w
| x |> a; 0 ≤ y ≤ h; ⇒ µ= T0
∂x
∂w
x = −a, − ≤ y ≤ h; ⇒ µ = −T0
∂x
∂w
y = h; | x |≤ a; ⇒ µ = T0
∂y

Plane State of Stress (PSS)


Definition 22 Plane state of stress parallel to xy−plane has σ zz = σ zx =
σ zy ≡ 0.
Physically that corresponds to a thin plate in the xy−plane loaded with
traction vectors in the same plane.
The strain field follows from
1
eij = [(1 + ν)σ ij − νσkk δ ij ]
E

81
to be
1
exx = (σ xx − νσ yy )
E
1
eyy = (σ yy − νσ xx )
E
ν
ezz = − (σxx + σ yy )
E
exz = eyz = 0 (4.50)

From
ν
σ ij = 2µ[ ekk δ ij + eij ] (4.51)
1 − 2ν
we obtain first that σzz = 0 implies that
ν
ezz = − (exx + eyy ) (4.52)
1−ν
Based on this result the corresponding stress field is calculated to be
2µ E
σ xx = (exx + νeyy ) = (exx + νeyy )
1−ν 1 − ν2
E
σ yy = (eyy + νexx )
1 − ν2
E
σ xy = 2µexy = exy (4.53)
1+ν
where we have used the fact that E = 2µ(1 + ν).
Note:
E
σ xx + σ yy = (exx + eyy )
1−ν
∂u ∂v
exx + eyy = + (4.54)
∂x ∂y
Equation of equilibrium becomes
∂σ xx ∂σ xy
+ + fx = 0
∂x ∂y
∂σ xy ∂σ yy
+ + fy = 0 (4.55)
∂x ∂y

Substitution for stresses in terms of strains and then using eij = (ui,j +uj,i )/2
leads to the equations of equilibrium for PSS to be

∂2u ∂2u 1 + ν ∂ ∂u ∂v
µ( + 2)+µ ( + ) + fx = 0
∂x2 ∂y 1 − ν ∂x ∂x ∂y
∂2v ∂2v 1 + ν ∂ ∂u ∂v
µ( 2 + 2 ) + µ ( + ) + fy = 0 (4.56)
∂x ∂y 1 − ν ∂y ∂x ∂y

82
or
1+ν
µ∇2 u + µ ∇∇ • u + f = 0 (4.57)
1−ν
· ¸
u
u =
v

Plane Strain (PS) Model


Definition 23 We say that the model is of the plane strain type if

w ≡ 0; u = u(x, y); v = v(x, y) (4.58)

Physically, this model can be envisioned in terms of a long cylinder along


the z-axis subjected to tractions which are uniform along the cylinder’s axis.
First, based on the prescribed from of the displacement field we obtain that
ezz = 0 implies that
σ zz = ν(σ xx + σ yy ) (4.59)
Using that result the stresses (4.51) are calculated to be

ν ∂u ∂v ∂u
σ xx = 2µ[ ( + )+ ]
1 − 2ν ∂x ∂y ∂x
ν ∂u ∂v ∂v
σ yy = 2µ[ ( + )+ ]
1 − 2ν ∂x ∂y ∂y
∂u ∂v
σxy = µ( + ) (4.60)
∂y ∂x
The Navier’s equations for plane strain model become
1
µ∇2 u + µ ∇∇ • u + f = 0 (4.61)
1 − 2ν
· ¸
u
u =
v

By comparing Eqns. 4.61 and 4.57 it is evident that the only difference
1+ν 1
between the two is in the factors 1−ν and 1−2ν . Therefore, if ν in Eq. 3.35 is
replaced by ν/ (1 + ν), then Eqns. 4.61 and 4.57 will be identical. This suggests
that any solution of plane strain equation of equilibrium may be solved as a
plane stress problem by replacing the true value of ν by the ”apparent value”
ν/(1 + ν). Conversely, any plane stress problem may be solved as a plane strain
problem by replacing true ν by an apparent ν/(1 − ν).

83
4.4.3 Solution of 2D Elastostatic Problems in Terms of
Stresses
Equations of equilibrium
∂σ xx ∂σ xy
+ + fx = 0 (4.62)
∂x ∂y
∂σ xy ∂σ yy
+ + fy = 0 (4.63)
∂x ∂y
consist of two equations with three unknowns. Thus we need to use the com-
patibility equation
∂ 2 exx ∂ 2 eyy ∂ 2 exy
2
+ 2
=2 (4.64)
∂y ∂x ∂x∂y
By replacing the components of strain tensor in terms of stresses we obtain for
a plane strain model
1
exx = [(1 − ν 2 )σ xx − ν(1 + ν)σ yy ]
E
1
eyy = [(1 − ν 2 )σ yy − ν(1 + ν)σ xx ]
E
1+ν
exy = σ xy (4.65)
E
and consequently the compatibility equation becomes
∂2 ∂2 ∂ 2 σ xy
[(1 − ν)σ xx − νσ yy ] + [(1 − ν)σ yy − νσ xx ] = 2 (4.66)
∂y 2 ∂x2 ∂x∂y
By adding Eqns. 4.62 and 4.63 we obtain
∂ 2 σxx ∂ 2 σ yy ∂fx ∂fy ∂ 2 σ xy
2
+ 2
+ + = −2 (4.67)
∂x ∂y ∂x ∂y ∂x∂y
Finally, by adding Eqns. 4.66 and 4.67 we obtain the compatibility equation for
a plane strain problem to be
1 ∂fx ∂fy
∇2 (σ xx + σ yy ) = − ( + ) (4.68)
1 − ν ∂x ∂y
∂2 ∂2
∇2 = 2
+ 2
∂x ∂y
Therefore a plane strain problem in terms of stresses is defined by partial
differential equations
∂σ xx ∂σ xy
+ + fx = 0
∂x ∂y
∂σ xy ∂σ yy
+ + fy = 0
∂x ∂y
1 ∂fx ∂fy
∇2 (σ xx + σ yy ) = − ( + ) (4.69)
1 − ν ∂x ∂y
x ∈ D

84
where D denote the interior of the body domain. The following boundary
conditions must be satisfied

Tx = σ xx nx + σ xy ny
Ty = σ yx nx + σ yy ny (4.70)
x ∈ ∂D

where ∂D denotes the boundary of the domain.


Note for plane stress problem the compatibility equation becomes
∂fx ∂fy
∇2 (σ xx + σ yy ) = −(1 + ν)( + ) (4.71)
∂x ∂y

85
Chapter 5

Plane Elasticity in
Cylindrical Coordinates

5.1 Plane Strain (PS) Model


Let displacement filed in polar coordinates (r, θ) for plane strain model be given
by
ur = ur (r, θ); uθ = uθ (r, θ); uz = 0 (5.1)
The corresponding strain and stress fields are given then (see Set #6)
1 1
err = ur,r ; eθθ = (uθ,θ + ur ); eeθ = (ur,θ + ruθ,r − uθ ); (5.2)
r 2r
erz = eθz = ezz = 0

σrr = λd + 2µur,r ; σθθ = λd + (uθ,θ + ur ); σ zz = λd;
r
ur 1
d = ur,r + + uθ,θ (5.3)
r r
1 uθ
σrz = σ θz = 0; σ rθ = µ( ur,θ + uθ,r − )
r r
Equations of equilibrium are then specified by (see Set#6)
1 1
σ rr,r + σ rθ,θ + (σrr − σ θθ¯)fr = 0
r r
1 2
σ rθ,r + σ θθ,θ + σ rθ + fθ = 0 (5.4)
r r

86
The strain-stress equations are given by
1 1+ν
err = (σ rr − λd) = [(1 − ν)σ rr − νσ θθ ]
2µ E
1 1+ν
eθθ = (σ θθ − λd) = [(1 − ν)σ θθ − νσrr ] (5.5)
2µ E
1+ν
erθ = σ rθ
E
Remark 12 Recall in Cartesian coordinates we have that

σ 11 = λ(e11 + e22 ) + 2µe11


σ 22 = λ(e11 + e22 ) + 2µe22
σ r + σ 22 = 2(λ + µ)(e11 + e22 ) = 2(λ + µ)d

while in polar coordinates it follows that

σrr = λd + 2µerr
σ θθ = λd + 2µeθθ
σ rr + σ θθ = 2(λ + µ)(err + eθθ ) = 2(λ + µ)d

Therefore
σ 11 + σ 22 = σ rr + σ θθ (5.6)

Recall that the compatibility equation for PS model is


1
∇2 (σ11 + σ 22 ) + ∇ • f =0
1−ν
which in polar coordinates becomes
∂2 1 ∂ 1 ∂2 1 ∂fr fr 1 ∂fθ
( 2
+ + 2 2 )(σ rr + σ θθ ) + ( + + )=0 (5.7)
∂r r ∂r r ∂θ 1 − ν ∂r r r ∂θ

5.2 Plane Stress (PSS) Model


In this case the model is defined by

σ rz = σθz = σ zz ≡ 0
∂σ rr ∂σ θθ ∂σ rθ
= = ≡0
∂z ∂z ∂z
d = err + eθθ + ezz

The stress-strain relations are defined by

σ rr = λd + 2µerr
σ θθ = λd + 2µeθθ (5.8)
σ zz = λd + 2µezz ≡ 0

87
Vanishing of σ zz results in
λ 2µ
ezz = − (err + eθθ ) ⇒ d = (err + eθθ ) (5.9)
λ + 2µ λ + 2µ
Also
σ rr + σ θθ
σ rr + σ θθ = 2λd + 2µ(err + eθθ ) ⇒ d = (5.10)
3λ + 2µ
Thus the strain-stress equations become
1 1
err = (σ rr − λd) = (σrr − νσ θθ )
2µ E
1
eθθ = (σ θθ − νσ rr ) (5.11)
E
1+ν
erθ = σ rθ
E
Recall in Cartesian coordinates the compatibility equation for plane stress model
is given by
∇2 (σ 11 + σ 22 ) + (1 + ν)∇ • f =0
and

σ 11 = λd + 2µe11 ; σ 22 = λd + 2µe22 ;
λ 2µ
e33 = − (e11 + e22 ); d = (e11 + e22 )
λ + 2µ λ + 2µ
so that
σ 11 + σ 22 = 2λd + 2µ(e11 + e22 ) = (3λ + 2µ)d
In polar coordinates σ zz = 0 implies that
λ 2µ
ezz = − (err + eθθ ) ⇒ d = (err + eθθ )
λ + 2µ λ + 2µ
d = err + eθθ + ezz

Consequently it follows that

σ rr + σθθ = 2λd + 2µ(err + eθθ ) = (3λ + 2µ)d

Therefore, we have that


σ 11 + σ 22 = σ rr + σ θθ (5.12)
and the compatibility equation becomes

∂2 1 ∂ 1 ∂2 ∂fr fr 1 ∂fθ
( 2
+ + 2 2 )(σ rr + σ θθ ) + (1 + ν)( + + )=0 (5.13)
∂r r ∂r r ∂θ ∂r r r ∂θ

88
5.3 Stress Function in Cartesian Coordinates
Assume that the body forces can be derived from a potential V so that

fx = −V,x ; fy = −V,y (5.14)

Then, equilibrium equations become

(σ xx − V ),x + σ xy,y = 0
σ xy,x + (σ yy − V ),y = 0 (5.15)

Let us introduce a function φ(x, y) such that

σxx − V = φ,yy
σ yy − V = φ,xx (5.16)
σ xy = −φ,xy

Then the equations of equilibrium

φ,xyy − φ,xyy = 0
−φ,xxy + φ,xxy = 0

are trivially satisfied. The function φ is known as the Airy’s stress function.
Since the equilibrium equations are automatically satisfied by the stress func-
tion, we are left to check the compatibility equation (PSS)

∇2 (σ xx + σ yy ) = −(1 + ν)(fx,x + fy,y )

By substituting into the last equation for the stress function, the compatibility
equation reduces to
∇4 φ(x, y) = −(1 − ν)∇2 V (5.17)
where
∇4 φ = φ,xxxx + 2φ,xxyy + φ,yyyy (5.18)
For plane strain (PS) model the compatibility equation in terms of the stress
function becomes
1 − 2ν 2
∇4 φ(x, y) = − ∇ V (5.19)
1−ν
In the case of zero body forces, the compatibility equations (for both PSS
and PS) reduce to a single biharmonic equation

∇4 φ(x, y) = 0 (5.20)

89
5.3.1 Solution of a Biharmonic Equation
Let’s consider a biharmonic equation

∇4 φ(x, y) = ∇2 (φ,xx + φ,yy ) = 0 (5.21)

By introducing the mapping

z = x + iy
z = x − iy (5.22)

it follows that
∂2 ∂2 ∂2
∇2 = + = 4 (5.23)
∂x2 ∂y 2 ∂z∂z
4

∇4 = ∇2 ∇2 = 16 2 2 (5.24)
∂z ∂z
Therefore, the biharmonic equation reduces to

∂ 4 φ(z, z)
=0 (5.25)
∂z 2 ∂z 2
Multiple integration of this equation results in

φ(z, z) = zF (z) + G(z) + zH(z) + J(z) (5.26)

where F, G, H and J are arbitrary analytic functions of complex variable.

Remark 13 We have used the following convention for defining a conjugate


function G(z) of G(z)

G(z) = u(x, y) + iv(x, y)


G(z) = u(x, −y) + i(x, −y)
G(z) = u(x, y) − iv(x, y) (5.27)

Since we want φ to be real we require that

φ − φ = 2 Im φ = 0

or

z([F (z) − H(z) + z[(H(z) − F (z)] + G(z) − J(z) + J(z) − G(z) = 0 (5.28)

Therefore, we must have that

F (z) = H(z) & G(z) = J(z) (5.29)

and the stress function becomes

φ(z, z) = zF (z) + zF (z) + G(z) + G(z) (5.30)

90
Let us express the analytic functions F and G as
1
F (z) = [ψ (x, y) + iψ 2 (x, y)]
2 1
1
G(z) = ψ (x, y) + iψ 4 (x, y) (5.31)
2 3
then the stress function reduces to

φ(x, y) = xψ 1 (x, y) + yψ 2 (x, y) + ψ 3 (x, y) (5.32)

where

ψ1 = 2 Re F (z)
ψ2 = 2 Im F (z) (5.33)
ψ3 = 2 Re G(z)

As we recall, F (z) and G(z) are assumed to be analytic functions.


Now for an analytic function f (z) = u(x, y) + iv(x, y) the real and imaginary
parts must satisfy Cauchy-Riemann equations

u,x = v,y & u,y − −v,x (5.34)

This implies that u and v must be harmonic functions in the region of analyticity
of f (z), i.e.,
∇2 u(x, y) = 0 & ∇2 v(x, y) = 0 (5.35)
Therefore we must have that

∇2 ψ i = 0, i = 1, 2, 3 (5.36)

and Eq. 5.32 represents the most general solution of the biharmonic equation.

Remark 14 Often it is possible to avoid the follow this general procedure of


solving the biharmonic equation which is illustrated in the following two exam-
ples.

Consider stress function being a polynomial of the third degree


a3 3 b3 2 c3 d3
φ= x + x y + xy 2 + y 3 (5.37)
6 2 2 6
This stress function automatically satisfies the biharmonic equation ∇4 φ = 0
and the stresses are calculated to be

σxx c3 x + d3 y; σyy = a3 x + b3 y; σ xy = −b3 x − c3 y (5.38)

The constants a3,..., b3 can be chosen arbitrarily. However, stress function of the
form a polynomial of order four
a4 4 b4 3 c4 d4 e4
φ= x + x y + x2 y 2 + xy 3 + y 4 (5.39)
12 6 2 6 12

91
in order to satisfy the biharmonic equation ∇4 φ = 0 the coefficients must satisfy
the following
a4 + 2c4 + e4 = 0 (5.40)
Since ∇4 φ = 0 is a linear partial differential equation the principle of super-
position can be used in order to solve certain problems exactly.

5.3.2 Stress Function and Displacement Field


Plane Strain Case. For this model the strains can be calculated (see Set #8)
according to
1
exx = [σ xx − ν(σ xx + σ yy )]

Thus we have that
2µu,x = φ,yy − ν∇2 φ (5.41)
Similarly
1
eyy = [σyy − ν (σ xx + σ yy )]

leads to
2µv,y = φ,xx − ∇2 φ (5.42)
Let us assume that we can find function ψ such that

∇2 φ ≡ ψ ,xy (5.43)

Since ¡ ¢
∇2 ∇2 φ = ∇2 ψ ,xy = 0 (5.44)
we obtain that ψ must be a harmonic function, i.e.,

∇2 ψ = 0 (5.45)

Equation 5.43 results in the following equations

φ,xx = −φ,yy + ψ ,xy (5.46)


φ,yy = −φ,xx ψ ,xy (5.47)

By substituting Eq. 5.46 into Eq. 5.41 and Eq. 5.47 into Eq. 5.42 we obtain
the following equations

2µu,x = −φ,xx + (1 − ν)ψ ,xy (5.48)


2µv,y = −φ,yy + (1 − ν)ψ ,xy (5.49)

Integration of the last two equation provides the following relationships between
the displacements and the stress function

2µu = −φ,x + (1 − ν)ψ ,y (5.50)


2µv = −φ,y + (1 − ν)ψ ,x (5.51)

92
L b

x O 2h

Figure 5.1: Bending of a cantiliver beam.

where
∇2 φ = ψ ,xy & ∇2 ψ = 0 (5.52)
For plane stress (PSS) model corresponding equations are derived to be

2µu = −φ,x + (1 + ν)−1 ψ ,y (5.53)


−1
2µv = −φ,y + (1 + ν) ψ ,x (5.54)

Example 17 Bending of a cantiliver beam loaded at its end section (Fig. ??.

Solution 16 Let’s try


d4 3
φ = −b2 xy + xy
6
Then the stresses are calculated to be

σxx = φ,yy = d4 xy
d4 2
σ xy = −φ,xy = b2 − y
2
σ yy = φ,xx = 0

Now the boundary conditions at y = ±h are


d4 2
σ yy = 0 & σ xy = 0 ⇒ b2 = h
2
Solution 17 The force balance at the loaded end x = 0 can be stated as
Z h Z h
d4
P = (−σ xy )bdy = −b (b2 − y 2 )dy
−h −h 2
d4
= −b(b2 2h − 2h3 )
2∗3

93
which leads to the result
3P
d4 −
2bh3
Consequently the stress filed can be determined to be
3P xy 3P
σ xx = − ; σyy = 0; σ xy = − (h2 − y 2 )
2bh3 4bh3
Using the moment of inertia I of the cross section I = 2bh3 /3 it follows then
that
P xy P
σ xx = − ; σ yy = 0; σ xy = − (h2 − y 2 ) (5.55)
I 2I
These results denote the exact solution provided the sharing forces at the end are
distributed according to the same parabolic law as the shearing stress σ xy in Eq.
5.55 and if the intensity of the normal forces at the built-in end is proportional
to y. If the forces at the ends are distributed in another way, the solution is not
exact. However, this solution is found to be acceptable for cross sections away
from the ends. As for the displacement field we have that
2µu = −φ,x + αψ ,y
2µv = −φ,y + αψ ,x
∇2 ψ= 0 & ∇2 φ = ψ ,xy
½
1 − ν; P S
α =
(1 + ν)−1 ; P SS
In order to find the function ψ we proceed as follows. Equation
d4 3
∇2 φ = ∇2 (−b2 xy + xy ) = d4 xy = ψ ,xy
6
implies that
d4 2 2
x y + f (x) + g(y)
ψ=
4
Since ∇2 ψ = 0 there follows that
d4 2
f 00 (x) + g 00 (y) + (x + y 2 ) = 0 ⇒
2
d4 d4
f 00 (x) + x2 = −g 00 (y) − y 2 = const = a0
2 2
So
d4 2 2 a0 2 d4
ψ= x y + (x − y 2 ) − (x4 + y 4 ) + a1 x + b1 y + c1
4 2 24
Based on this result we can evaluate
d4 d4
φ,x = −b2 y + y 3 ; φ,y = −b2 x + xy 2
6 2
d4 2 d4 3
ψ ,x = xy + a0 x − x + a1
2 6
d4 2 d4 3
ψ ,y = x y − a0 y − y + b1
2 6

94
so that the displacement filed can be evaluated to be

2µu = −φ,x + αψ ,y
d4 d4 d4
= b2 y + y 3 + α( x2 y − a0 y − y 3 + b1 )
6 2 6
2µv = −φ,y + αψ ,x
d4 d4 d4
= b2 x + xy 2 + α( xy 2 + a0 x − x3 + a1 ) (5.56)
2 2 6
or
αd4 2 (1 + α)d4 3
2µu = x y− y − (a0 α − b2 )y + b1 α
2 6
(1 − α)d4 2 αd4 3
2µv = − xy − x + (a0 α + b2 )x + a1 α
2 6
These solutions cannot satisfy the required boundary conditions at x =

u = v = 0; x = ; −h < y < h

However the unknown constants a0 , a1 , b1 can be determined by using an approx-


imate ”engineering” boundary conditions at x =

u = v = 0; v,x = 0; x = ; y = 0

Thus

u( , 0) = b1 α = 0 ⇒ b1 = 0
αd4 2 αd4 2
v,x ( , 0) = − + (a0 α + b2 ) = 0 ⇒ a0 α + b2 =
2 2
αd4 3 αd4 3 d4 3
v( , 0) = − + + a1 α = 0 ⇒ a1 = − (5.57)
6 2 3
Therefore the displacement filed is calculated to be (for PSS α = (1 + ν)−1 )

αd4 3
2µv(x, 0) = − x + (a0 α + b2 )x + a1 α
6
or
P
v(x, 0) = (x3 − 3 2 x + 2 l3 )
6EI
The strength of material solution is given by

P 3
v(0, 0) =
3EI
It should be emphasized that our solution is an approximate one since it does not
satisfy all the boundary conditions. The results become more and more accurate
as the distance from the ends increases.

95
5.4 Airy Stress Function in Polar Coordinates
Recall for plane strain model in polar coordinates we have the strain-displacement
equations of the form
err = ur,r
1
eθθ = (uθ,θ + ur )
r
1
erθ = (ur,θ + ruθ,r − uθ )
2r
erz = eθz = ezz = 0 (5.58)
Corresponding equilibrium equations are
1 1
σ rr,r + σ rθ,θ + (σ rr − σ θθ ) + fr = 0
r r
1 2
σ rθ,r + σ θθ,θ + σ rθ + fθ = 0 (5.59)
r r
Stress-strain equations are specified by
1+ν
err = [(1 − ν)σ rr − νσ θθ ]
E
1+ν
eθθ = [(1 − ν)σ θθ − νσ rr ] (5.60)
E
or
2µerr = (1 − ν)σ rr − νσ θθ = −σθθ + (1 − ν)(σrr + σ θθ )
2µeθθ = (1 − ν)σ θθ − νσrr = −σrr + (1 − ν)(σ rr + σ θθ ) (5.61)
Equilibrium equations for zero body force can be written as
(rσ rr ),r + σ rθ,θ − σ θθ = 0
(r2 σ rθ ),r + rσ θθ,θ = 0 (5.62)
In order to introduce the stress function in polar coordinates we define first the
transformation from Cartesian to polar coordinates
x = r cos θ; y = r sin θ
y
r2 = x2 + y 2 ; θ = tan−1
x
sin θ cos θ
r,x = cos θ; r,y = sin θ; θ,x = − ; θ,y = (5.63)
r r
Therefore, we have that
cos θ
φ,y = φ,r r,y + φ,θ θ,y = φ,r sin θ + φ,θ (5.64)
r
1 1 1
φ,yy = ( φ,r + φ ) cos2 θ + φ,rr sin2 θ + ( φ,θ ),r sin 2θ (5.65)
r r2 ,θθ r
1 1 2 2 1
φ,xx = ( φ,r + φ ) sin θ + φ,rr cos θ − ( φ,θ ),r sin 2θ (5.66)
r r2 ,θθ r

96
s s s
s
q yy
q r
xy
s
y s s rq
xy
xx
y
r s
s rr
s rq
rr s q r s q

q q
x
O x

Figure 5.2: Two elements in equilibrium.

Let’s consider now two elements of a solid as depicted by Fig.5.2. Equilibrium


conditions for the first element yield

σ xx = σ θθ sin2 θ + σ rr cos2 θ − σ rθ sin 2θ = φ,yy (5.67)

while the equilibrium of the second provides

σ yy = σθθ cos2 θ + σ rr sin2 θ + σ rθ sin 2θ = φ,xx (5.68)

Then from Eqns. 5.65 to 5.68 it follows that


1 1
σ rr = φ + φ
r ,r r2 ,θθ
σ θθ = φ,rr
1
σ rθ = −( φ,θ ),r (5.69)
r
which expresses stresses in terms of the stress function in polar coordinates.
It is easy to verify that stresses calculated from Eqns. 5.69 identically satisfy
equations of equilibrium Eq. 5.62.
Now the compatibility equation in cylindrical coordinates is given by

∇2 (σ xx + σ yy ) = ∇2 (σ rr + σ θθ ) = 0

or
∂2 1 ∂ 1 ∂2 ∂2 1 ∂ 1 ∂2
∇4 φ = ( 2
+ + 2 2 )( 2 + + 2 2) = 0 (5.70)
∂r r ∂r r ∂θ ∂r r ∂r r ∂θ
Also,
∇2 φ = φ,xx + φ,yy = σ xx + σ yy = σ rr + σ θθ (5.71)
From strain-stress equations 5.60 it follows that

2µur,r = −σθθ + (1 − ν)(σrr + σ θθ )

97
or
2µur,r = −φ,rr + (1 − ν)∇2 φ; (P S) (5.72)
Now let’s introduce an auxiliary function ψ such that

∇2 φ = (rψ ,θ ),r (5.73)

then Eqn. 5.72 becomes

2µur,r = −φ,rr + (1 − ν)(rψ ,θ ),r

which after integration results in

2µur = −φ,r + 1 − ν)rψ ,θ ; (P S) (5.74)

Similarly, Eqns.5.60,5.69 and 5.73 result in


1
2µ (ur + uθ,θ ) = −σ rr + (1 − ν)(σ rr + σ θθ )
r
1 1
= −( φ,r + 2 φ,θθ ) + 1 − ν)∇2 φ
r r
From the last result it follows then that
1
2µuθ,θ = −φ,r − φ,θθ + (1 − ν)r∇2 φ + φ,r − (1 − ν)rψ ,θ
r
1
= − φ,θθ + (1 − ν)r[∇2 φ − ψ ,θ ]
r
1
= − φ,θθ + (1 − ν)r2 ψ ,rθ
r
which after integration provides
1
2µuθ = − φ,θ + (1 − ν)r2 ψ ,r ; (P S) (5.75)
r
Therefore, Eqns. 5.74 and 5.75 relate stress function φ and auxiliary function
ψ to displacement field in polar coordinates. In order to examine further the
properties of the auxiliary function ψ we recall that
1 σ rθ 1 1
erθ = (ur,θ + ruθ,r − uθ ) = = − ( φ,θ ),r
2r 2µ 2µ r
Therefore
1
2µ(ur,θ + ruθ,r − uθ ) = −2r( φ,θ ),r
r
or
1 1
−φrθ + (1 − ν)rψ ,θθ + r( φ − φ + 2(1 − ν)rψ ,r + (1 − ν)r2 ψ ,rr )
r2 ,θ r ,rθ
1 2
+ φ,θ − (1 − ν)r2 ψ ,r = φ − 2φ,rθ
r r ,θ

98
or
1 1
(1 − ν)r3 (ψ ,rr + ψ ,r + 2 ψ ,θθ ) = 0
r r
Thus we have that the auxiliary function ψ must satisfy

∇2 ψ = 0 (5.76)

By replacing ν with ν/(1 + ν) we obtain the corresponding plane stress


equations relating the displacement field to stress function φ and the auxiliary
function ψ according to
1
2µur = −φ,r + rψ (5.77)
1 + ν ,θ
1 1 2
2µuθ = − φ,θ + r ψ ,r (5.78)
r 1+ν

5.4.1 Solutions for Stress Function in Polar Coordinates


Let’s consider a special case of stress function of the form

φ(r, θ) = gn (r) cos nθ (5.79)

Then we have that


∂2 1 ∂ 1 ∂2
∇2 φ = ( 2
+ + 2 2 )gn (r) cos nθ
∂r r ∂r r ∂θ
d2 gn 1 dgn n2
= ( 2 + − 2 gn ) cos nθ = Gn (r) cos nθ (5.80)
dr r dr r
and thus
d2 1 d n2 2
∇4 φ = ( + − ) gn (r) cos nθ
dr2 r dr r2
2 2
d 1 d n
= ( 2+ − 2 )Gn (r) cos nθ (5.81)
dr r dr r
Thus biharmonic equation ∇4 φ = 0 becomes

d2 1 d n2 d2 1 d n2
( + − )( + − )gn (r) = 0 (5.82)
dr2 r dr r2 dr2 r dr r2
Case n=0
In this case Eqn. 5.82 becomes

d2 1 d d2 1 d
( + )( + )g0 (r) = 0 (5.83)
dr2 r dr dr2 r dr
Let
d2 1 d
G0 (r) = ( + )g0 (5.84)
dr2 r dr

99
Then Eqn. 5.83 becomes
dG
r2 G000 + rG00 = 0; G0 = (5.85)
dr
Now introduction of new variable t via
dt 1
r = et ⇒ = (5.86)
dr r
results in
dG 1 dG
=
dr r dt
d2 G d dG 1 d 1 dG 1 dG 1 d2 G
= ( )= ( )=− 2 + 2 2
dr2 dr dr r dt r dt r dt r dt
Consequently Eqn. 5.85 becomes
d2 G0
=0 (5.87)
dt2
which can be solved for
G0 (t) = A + Bt (5.88)
Now from Eqn. 5.84 it follows that
d2 1 d 1 d2 d 1 d2 g0
( 2
+ )g0 = 2 (r2 2 + r )g0 = 2 2 = A + Bt (5.89)
dr r dr r dr dr r dt
Integrating the last result twice produces
1 B
g0 (t) = (A − B)e2t + te2t + Ct + D (5.90)
4 4
or
g0 (r) = a0 r2 + b0 r2 ln r + c0 + d0 ln r = φ(r, θ); n = 0 (5.91)
Case n=1
In this case Eqn. 5.82 becomes
d2 1 d 1
( + − )2 g1 = 0 (5.92)
dr2 r dr r2
Let
d2 1 d 1
( + − )g1 (r) = G1 (r) (5.93)
dr2 r dr r2
then Eqn. (5.92) implies

r2 G001 + rG01 − G1 = 0 (5.94)

Using the substitution defined by Eqn. 5.86, Eqn. 5.94 becomes


d2 G1
− G1 = 0 (5.95)
dt2

100
which solution is given by

G1 (t) = Aet + Be−t (h9e39)

Therefore Eqn. 5.93 can be written in the form

d2 1 d 1 1 2 d2 d
( 2
+ − 2 )g1 = 2
(r 2
+ r − 1)g1
dr r dr r r dr dr
1 d2 g1
= ( − g1 ) = Aet + Be−t
r2 dt2
or
d2 g1
− g1 = Ae3t + Bet (5.96)
dt2
The homogeneous and particular solutions of the last equation are given by

g1h = b1 et + d1 e−t (5.97)


g1p = a1 e3t + c1 tet (5.98)

and the solution is specified by

g1 (t) = a1 e3t + b1 et + c1 tet + d1 e−t

or
g1 (r) = a1 r3 + b1 r + c1 r ln r + d1 r−1 (5.99)
Case n>1
In this case Eqn. 5.82 becomes

d2 1 d n2
( 2
+ − 2 )2 gn (r) = 0 (5.100)
dr r dr r
As before, let
d2 1 d n2
Gn (r) = ( + − )gn (r) (5.101)
dr2 r dr r2
Then Eqn. 5.100 implies

1 n2
G00n + G0n − 2 Gn = 0 (5.102)
r r
Introducing a change of variable according to Eqn. 5.86 the last equation be-
comes
d2 Gn
− n2 Gn = 0 (5.103)
dt2
which can be solved to give

Gn = Aent + Be−nt (5.104)

101
Consequently
d2 1 d n2 1 2 d2 d
( + − )gn (r) = (r + r − n2 )gn
dr2 r dr r2 r 2 dr 2 dr
1 d2 2
= ( − n )gn = Gn
r2 dt2
which implies that
d2
( − n2 )gn (t) = Ae(n+2)t + Be(−n+2)t (5.105)
dt2
Homogeneous and particular solutions of the last equation are given by

gnh = bn ent + dn e−nt (5.106)


gnp = an e(n+2)t + cn e(−n+2)t (5.107)

So the solution for gn is given by

gn (t) = an e(n+2)t + bn ent + cn e(−n+2)t + dn e−nt

or
gn (r) = an rn+2 + bn rn + cn r−n+2 + dn r−n ; n > 1 (5.108)

5.4.2 Stress and Displacement Fields


Based on the stress function just derived we proceed with evaluation of corre-
sponding stress and displacement fields. For that purpose we utilize the follow-
ing equations

1 1
σ rr = φ + φ
r ,r r2 ,θθ
σ θθ = φ,rr
1
σ rθ = −( φ,θ ),r
r

2µur = −φ,r + αrψ ,θ


1
2µuθ = − φ,θ + αr2 ψ ,r
r
 (1 − ν); P S
α =

(1 + ν)−1 ; P SS
∇2 φ = (rψ ,θ ),r
∇2 ψ = 0

Case n=0

102
φ = a0 r2 + b0 r2 ln r + c0 + d0 ln r
1 1 d0
σ rr = φ + φ = 2a0 + b0 (1 + 2 ln r) + 2 (5.109)
r ,r r2 ,θθ r
d0
σθθ = φ,rr = 2a0 + b0 (3 + 2 ln r) − 2 (5.110)
r
σ rθ = 0 (5.111)
For displacement field we have
2µur = −φ,r + αrψ ,θ (5.112)
Now since
∇2 φ = (σrr + σ θθ ) = (rψ ,θ ),r ⇒
4a0 + 4b0 (1 + ln r) = (rψ ,θ ),r
we get that
c + 4a0 r + 4b0 r ln r = rψ ,θ (5.113)
Using the last results in Eqn. 5.112 results in
d0
2µur = −(2a0 r + 2b0 r ln r + b0 r + ) + α(4a0 r + 4b0 r ln r + c)
r
or
d0
2µur = 2(2α − 1)a0 r + b0 r + 2(2α − 1)b0 r ln r − + 2µu0r (5.114)
r
Now from Eqn. 5.113 we get that
c c
ψ ,θ = 4a0 + 4b0 ln r + ⇒ ψ = 4a0 θ + 4b0 θ ln r + θ + D
r r
and thus
r2 ψ ,r = 4b0 θr (5.115)
Based on this result we get the tangential component of displacement
1
2µuθ = − σ ,θ + αr2 ψ ,r + 2µu0θ
r
or
2µuθ = 4αb0 rθ + 2µu0θ (5.116)
Case n=1

φ = (a1 r3 + b1 r + c1 r ln r + d1 r−1 ) cos θ (5.117)


c1 2d1
σ rr = (2a1 r + − 3 ) cos θ (5.118)
r r
c1 2d1
σ θθ = (6a1 r + + 3 ) cos θ (5.119)
r r
c1 2d1
σrθ = (2a1 r + − 3 ) sin θ (5.120)
r r

103
Since
2c1
∇2 φ = σ rr + σ θθ = (8a1 r + ) cos θ = (rψ ,θ ),r ⇒
r
rψ ,θ = (4a1 r2 + 2c1 ln r) cos θ + c

the displacement field ur follows from Eqn. 5.74 to be

2µur = −φ,r + αrψ ,θ


d1
= −[(3 − 4α)a1 r2 + b1 + (1 − 2α)c1 ln r + c1 − ] cos θ
r2
+2µu0r (5.121)

Based on the result that


2c1 c
ψ ,θ = (4a1 r + ln r) cos θ + ⇒
r r
2c1 c
ψ = (4a1 r + ln r) sin θ + θ + D ⇒
r r
r2 ψ ,r = (4a1 r2 − 2c1 ln r + 2c1 ) sin θ − cθ

it follows from Eqn. 5.75 that


d1
2µuθ = [(1 + 4α)a1 r2 + b1 + 2αc1 + (1 − 2α)c1 ln r + ] sin θ
r2
+2µu0θ (5.122)

Case n>1
In this case the stress function is given by

φ = (an rn+2 + bn rn + cn r−n+2 + dn r−n ) cos nθ

and the corresponding stress fields is given by

σ rr = −[(n − 2)(n + 1)an rn + (n − 1)nbn rn−2 (5.123)


+(n − 1)(n − 2)cn r−n + n(n + 1)dn r−n−2 ] cos nθ
σ θθ = −[(n + 2)(n + 1)an rn + n(n − 1)nbn rn−2 (5.124)
+(n − 2)(n − 1)cn r−n + n(n + 1)dn r−n−2 ] cos nθ
σ rθ = n[an (n + 1)rn + bn (n − 1)rn−2 + cn (1 − n)r−n (5.125)
−(n + 1)dn r−n−2 ] sin θ (5.126)

Since
∇2 φ = σ rr + σ θθ = (rψ ,θ ),r
we have that
rψ ,θ = (4an rn+1 + 4cn r−n+1 ) cos nθ

104
Therefore, Eqn. 5.74 produces

2µur = [(n + 2 − 4Ga)rn+1 + nbn rn−1 − (n − 2 + 4α)cn r−n+1


−ndn r−n−1 ] cos nθ + 2µu0r (5.127)

Similarly, from the result

ψ ,θ = 4(an rn + cn r−n ) cos nθ ⇒


r2 ψ ,r = 4(an rn+1 − cn r−n−1 ) sin nθ

we obtain from Eqn. 5.75 the tangential component of displacement to be

2µuθ = [(n + 4α)an rn+1 + nbn rn−1 + (n − 4α)r−n+1 cn


+ndn r−n−1 ] sin nθ + 2µu0θ (5.128)

5.4.3 Axisymmetric Case


Plane Strain Model and n=0

α = 1−ν
2µur = 2(1 − 2ν)a0 r − b0 r + 2(1 − 2ν)b0 r ln r − d0 r−1
+2µu0r

or
1+ν
ur = [2(1 − 2ν)a0 r − b0 r + 2(1 − 2ν)b0 r ln r − d0 r−1 ] + u0r
E
4(1 − ν 2 )
uθ = b0 rθ + u0θ
E

Plane Stress Case

α = (1 + ν)−1
1
ur = [2(1 − ν)a0 r − (1 − ν)b0 r + 2(1 − ν)b0 r ln r
E
−(1 + ν)d0 r−1 ] + u0r
4b0 rθ
uθ = + u0θ
E
Example 18 Consider a semicircular arch of inner and outer radii being a and
b, respectively loaded as shown by Fig. 5.3

Determine the stress and displacement fields.

105
2P

r
P q P
Q d Q
M M
Figure 5.3: A semicircular arch problem.

Solution 18 The boundary conditions along the curved surfaces of the arch are
specified by
σ rr = σ rθ = 0; r = a, b; 0 < θ < π (5.129)
while along the end section θ = 0, π we must have

Z b
σ θθ dr = −P ; θ = 0, π
a
Z b
σrθ dr = −Q; θ = 0, π
a
Z b
σ θθ rdr = −M − P d (5.130)
a

Let’s try the stress function of the form

φ = f (r) + g (r)(A cos θ + B sin θ) (5.131)

Then from Eqns. 5.91 and 5.99 we get

f (r) = a0 r2 + b0 r2 ln r + c0 + d0 ln r
g(r) = a1 r3 + b1 r + c1 r ln r + d1 r−1 (5.132)

106
The stress filed follows from Eqns. 5.109 to 5.111 and 5.118 to 5.120 to be

σ rr = 2a0 + b0 (1 + 2 ln r) + d0 r−2 + (2a1 r + c1 r−1 − 2d1 r−3 )


(A cos θ + B sin θ)
σ θθ = 2a0 + b0 (3 + 2 ln r) − d0 r−2 + (6a1 r + c1 r−1 + 2d1 r−3 )
(A cos θ + B sin θ)
σrθ = (2a1 r + c1 r−1 − 2d1 r−3 )(A sin θ − B cos θ) (5.133)

Then the boundary conditions (5.129) become

2a0 + b0 (1 + 2 ln a) + d0 a−2 = 0
2a0 + b0 (1 + 2 ln b) + d0 b−2 = 0
2a1 a + c1 a−1 − 2d1 a−3 = 0
2a1 b + c1 b−1 − 2d1 b−3 = 0 (5.134)

The end conditions (5.130) then become

B[a1 r2 + c1 ln r + d1 r−2 ]ba = Q


A[3a1 r2 + c1 ln r − d1 r−2 ]ba = −P
[a0 r2 + b0 r2 (1 + ln r) − d0 ln r]ba = −(M + P d) (5.135)

The unknown coefficients are found to be


M0 2
a0 = − [b − a2 + 2(b2 ln b − a2 ln a)]
N2
2M0 2
b0 = (b − a2 )
N2
4M0 2 2 b
d0 = a b ln
N2 a
b
M0 = −(M + P d); N2 = (b2 − a2 )2 − 4a2 b2 (ln )2
a
A = P ; B = −Q
a2 b2 b
d1 = − ; N1 = a2 − b2 + (a2 + b2 ) ln
2N1 a
1 a2 + b2
a1 = ; c1 = −
2N1 N1

107
P
O
y
r
q

Figure 5.4: A normal line load on an elastic half-space.

Then the stress field is given by


4M0 a2 b2 b r r
σ rr = − (− 2 ln − b2 ln + a2 ln )
N2 r a b a
P cos θ − Q sin θ a2 b2 a2 + b2
+ (r + 3 − )
N1 r r
4M0 a2 b2 b r r
σθθ = − ( ln − b2 ln + a2 ln + a2 − b2 )
N2 r2 a b a
P cos θ − Q sin θ a2 b2 a2 + b2
+ (3r − 3 − )
N1 r r
P sin θ − Q cos θ a2 b2 a2 + b2
σ rθ = (r + 3 − ) (5.136)
N1 r r
If M = Q = 0 we get
P cos θ a2 + b2 a2 b2
σ rr = (r − + 3 )
N1 r r
2 2
P cos θ a +b a2 b2
σ θθ = (3r − − 3 )
N1 r r
P sin θ a2 + b2 a2 b2
σ rθ = (r − + 3 ) (5.137)
N1 r r
Problem 1 Normal Line Load on a Half-Space
Let’s consider the problem depicted by Fig.(5.4).
We try now to solve this problem by choosing the following stress function
P
φ=− rθ sin θ (5.138)
π

108
O e
y
C -s rr

x
Figure 5.5: Local stress field on a small semicircle at the origin.

Then the stress field becomes


1 1 2P
σ rr = φ + φ =− cos θ
r ,r r2 ,θθ πr
σ θθ = φ,rr = 0
1
σ rθ = −( φ,θ ),r = 0 (5.139)
r
It should be noted that the surface of the half-space is stress free (excluding
the origin). For that reason it is necessery to examine the situation under the
load more carefuly. Suppose we cut-out an infinitesimal semicircle of material
at the origin as shown by Fig(5.5).
Due to the original loading, the resultant force acting on the cylinder C must
be in the x-direction and of magnitude P. Then
Z π/2 Z π/2
2P
−σ rr r cos θdθ = cos2 θdθ = P (5.140)
−π/2 π −π/2

independently of ε as ε → 0. From the stress-strain relationship

(1 + ν)
err = ur,r = [(1 − ν)σ rr − νσθθ ]
E
1 (1 + ν)
eθθ = (uθ,θ + ur ) = [(1 − ν)σ θθ − νσrr ]
r E
1 1 uθ σ rθ
erθ = ( ur,θ + uθ,r − ) =
2 r r 2µ

109
we get

2(1 − ν 2 )P
ur,r = − cos θ (5.141)
πEr
1 2ν(1 + ν)P
(uθ,θ + ur ) = cos θ (5.142)
r πEr
1 1 uθ
( ur,θ + uθ,r − ) = 0 (5.143)
2 r r
The from Eqn.(5.141) we have that

2(1 − ν 2 )P
ur = − ln r cos θ + g 0 (θ) (5.144)
πE
and then from Eqn.(5.142) follows

2(1 + ν)P 2(1 − ν 2 )P


uθ = sin θ + ln r sin θ − g(θ) + f (r) (5.145)
πE πE
for an arbitrary functions f and g. By substituting (5.144) and (5.145) into
(5.143) we obtain

2(1 + ν)(1 − 2ν)P


g 00 (θ) + g (θ) + sin θ = −rf 0 (r) + f (r) = const = 0 (5.146)
πE
where primes denote differentiation. The last equation can be solved for
(1 + ν)(1 − 2ν)P
g = θ cos θ + a ∗ cos θ + b ∗ sin θ (5.147)
πE
f = d∗r (5.148)

for arrbitrary constants a, b, and c.


Since Eqn.(5.148) represents the rigid-body motion (RBM) we may assume
d = 0. Consequently, the displacement fields can be calculated from Eqns.(5.144)
and (5.145)

2(1 − ν 2 )P (1 + ν)(1 − 2ν)P


ur = − ln r cos θ − θ sin θ + u0r (5.149)
πE πE
2ν(1 + ν) 2(1 − ν 2 )P
uθ = P sin θ + ln r sin θ
πE πE
(1 + ν)(1 − 2ν)P
+ (sin θ − θ cos θ) + u0θ (5.150)
πE
where the superscript 0 denotes the RBM.
In order to determine the RBM’s we assume: i) all the points on the x-axis
do not have any lateral displacement, and ii) one point at distance d from the
origin along the x-axis is fixed. Thus

2(1 − ν 2 )P
ur |θ=0,r=d = 0 ⇒ u0r = ln d cos θ
πE

110
which imples
(1 + ν) d
ur = P [2(1 − ν) cos θ ln − (1 − 2ν)θ sin θ] (5.151)
πE r
(1 + ν) d
uθ = P [sin θ22(1 − ν) sin θ ln − (1 − 2ν)θ cos θ] (5.152)
πE r
where we have used for convenience u0θ = −2(1 − ν) ln d sin θ. We should note
that uθ (θ = 0) = 0.
At the half-space surface (θ = ±π/2) we have thus
(1 + ν)(1 − 2ν)P
ur = −
2E
(1 + ν)P d
uθ = ± [1 − 2(1 + ν) ln ] (5.153)
πE r
which imples that the material is displacemed toward the origin.
We should note that the solution is not valid at the point of the load appli-
cation. Also, there is a log-singularity at infinity.
For a PSS model, the stress and displacement fileds can be obtained by
replacing ν with ν/(1+ν), while keeping µ = E/{2(1+ν)} unchanged. Therefore
we have that the displacement field is given by
P d
ur =
[2 cos θ ln − (1 − ν)θ sin θ]
πE r
P d
uθ = [(1 + ν) sin θ − 2 sin θ ln − (1 − ν)θ cos θ] (5.154)
πE r
Problem 2 Stress in a wedge subjected to a line loade at the vertex.
Consider an infinite elastic wedge subjected to a load P per unit length at
the vertex along the axis of the wedge (Fig.(5.6)).
Lets’s try the stress function
φ = crθ sin θ (5.155)
which implies the following stress field (PS)
1 1 2c
σ rr =
φ + φ = cos θ
r ,r r2 ,θθ r
σ θθ = φ,rr = 0
1
σrθ = −( φ,θ ),r = 0
r
2νc
σ zz = ν(σ rr + σ θθ ) = cos θ (5.156)
r
Let’s consider the equilibrium of the OAB portion of the wedge.
Z α
σ rr r cos θdθ + P = 0 (5.157)
−α
Z α
σ rr r sin θdθ = 0 (5.158)
−α

111
y
B
s rr
r
P q x
O C
2a
A
Figure 5.6: The wedge problem.

By substittuting the value of σrr in the last two equations we get from
Eqn.(5.157)
P
c=−
2α + sin 2α
and thus
2P cos θ
σrr = − ; σ θθ = σ rθ = 0 (5.159)
r(2α + sin 2α)
Note that
2P
σ rr (θ = π/2) = − cos θ
πr
which is the same result as for the half-space problem.
Problem 3 Wedge with a line load perpendicular to the x-axis.
Let’s consider the second wedge problem depicted by Fig.(5.141).
In this case we try
φ = drθ cos θ (5.160)
which provides
2d
σ rr = − sin θ (5.161)
r
σθθ = σrθ = 0 (5.162)
Balance of forces on the portion of the wedge can be stated as
Z α
Q
σrr r sin θdθ + Q = 0 ⇒ d =
−α 2α − sin 2α

112
y

s rr
Q r
q x
O C
2a

Figure 5.7: Second wedge problem.

and thus
2Q
σ rr = − sin θ (5.163)
r(2α − sin 2α)
σ θθ = σ rθ = 0 (5.164)

5.4.4 Stresses Near a Corner


Consider an elastic wedge as shown by Fig.(5.8).
We recall that the stress and displacement fields can be evaluated in terms
of the Stress function φ and the auxiliary function ψ, where

∇4 φ = 0 & ∇2 ψ = 0 & ∇2 φ = (rψ ,θ ),r (5.165)

Let’s try
φ = rβ+1 F (θ) & ψ = rm G(θ) (5.166)
where β and m are unknown constants while F and G are unknown functions.
By substituting Eqn.(5.166) into (5.165) we get

∂2 1 ∂ 1 ∂2
( 2
+ + 2 2 )2 rβ+1 F (θ) = 0
∂r r ∂r r ∂θ
or
d2 2
[(β − 1)2 + ] F (θ) = 0
dθ2

113
y

r
O q x
2q o

Figure 5.8: An elastic wedge.

The last equation becomes

F (iv) + [(β + 1)2 + (β − 1)2 ]F 00 + (β 2 − 1)2 F = 0 (5.167)

Similarly for the function ψ we get

G00 + m2 G = 0 (5.168)

Characteristic equation for Eqn.(5.167) is given by

λ4 + [(β + 1)2 + (β − 1)2 ]λ2 + (β 2 − 1)2 = 0

with the corresponding roots

λ2I = −(β − 1)2 ; λ2II = −(β + 1)2 (5.169)

Therefore, the solution for the two functions can be written as

F (θ) = b1 sin(β + 1)θ + b2 cos β + 1)θ + b3 sin(β − 1)θ + b4 cos(β −(5.170)


1)θ
G(θ) = a1 cos mθ + a2 sin mθ (5.171)

Using the third equation of Eqn.(5.165) we must have

[(β + 1)2 F (θ) + F 00 (θ)]rβ−1 = (m + 1)rm G0 (θ) (5.172)

In order for Eqn.(5.172) to be vaild for 0 < r < ∞, we must have

β−1=m (5.173)

114
Then from Eqns.(5.170) ,(5.171) and (5.172) it follows that
4b3 4b4
a1 = − ; a2 = (5.174)
β−1 β−1
Based on these results functions φ and ψ become
φ = rβ+1 [b1 sin(β + 1)θ + b2 cos(β + 1)θ + b3 sin(β − 1)θ + b4 cos(β (5.175)
− 1)θ
4
ψ = [−b3 cos(β − 1)θ + b4 sin(β − 1)θ]rβ−1 (5.176)
β−1
Now displacement filed can be evaluates according to
2µur = −φ,r + αrψ ,θ
1
2µuθ = − φ,θ + αr2 ψ ,r
r
which provides
2µur = {−(β + 1)[b1 sin(β + 1)θ + b2 cos(β + 1)θ] (5.177)
+(4α − β − 1)[b3 sin(β − 1)θ + b4 cos(β − 1)θ]}rβ
2µuθ = {−(β + 1)[b1 cos(β + 1)θ − b2 sin(β + 1)θ] (5.178)
−(4α + β − 1)[b3 cos(β − 1)θ − b4 sin(β − 1)θ]}rβ
Now the stress field can be calculated using Eqns.(5.156) and (5.175)
σrr = −{β(β + 1)[b1 sin(β + 1)θ + b2 cos(β + 1)θ]
+β(β − 3)[b3 sin(β − 1)θ + b4 cos(β − 1)θ]}rβ−1 (5.179)
σ θθ = β(β + 1)[b1 sin(β + 1)θ + b2 cos(β + 1)θ]
+b3 sin(β − 1)θ + b4 cos(β − 1)θ]rβ−1 (5.180)
σ rθ = −β{(β + 1)[b1 cos(β + 1)θ − b2 sin(β + 1)θ]
+(β − 1)[b3 cos(β − 1)θ − b4 sin(β − 1)θ]}rβ−1 (5.181)
Since the both wedge faces are stress free we have that
σ θθ = σ rθ = 0; θ = ±θ0 (5.182)
Using Eqns.(5.180)-(5.182) we obtain
b1 sin(β + 1)θ0 + b2 cos(β + 1)θ0 + b3 sin(β − 1)θ0 + b4 cos(β − 1)θ0 = 0
−b1 sin(β + 1)θ0 + b2 cos(β + 1)θ0 − b3 sin(β − 1)θ0 + b4 cos(β − 1)θ0 = 0
(β + 1)[b1 cos(β + 1)θ0 − b2 sin(β + 1)θ0 ] + (β − 1)[b3 cos(β − 1)θ0 − b4 sin(β − 1)θ0 ] = 0
(β + 1)[b1 cos(β + 1)θ0 + b2 sin(β + 1)θ0 ] + (β − 1)[b3 cos(β − 1)θ0 + b4 sin(β − 1)θ0 ] = 0
or
b1 sin(β + 1)θ0 + b3 sin(β − 1)θ0 = 0 (5.183)
b1 (β + 1) cos(β + 1)θ0 + b3 (β − 1) cos(β − 1)θ0 = 0 (5.184)
b2 cos(β + 1)θ0 + b4 cos(β − 1)θ0 = 0 (5.185)
b2 (β + 1) sin(β + 1)θ0 + b4 (β − 1) sin(β − 1)θ0 = 0 (5.186)

115
uq uq
y ur y ur

O q x q x
O
-q -q

symmetric antisymmetric
ur ur
uq
uq

Figure 5.9: Symmetric and antisymmetric wedge displacements.

For a nontrivial solution of the systems (5.183)-(5.184) and (5.185)-(5.186)


we must have

(β − 1) sin(β + 1)θ0 cos(β − 1)θ0 − (β + 1) cos(β + 1)θ0 sin(β − 1)θ0 = 0


(β − 1) sin(β11)θ0 cos(β + 1)θ0 − (β + 1) cos(β − 1)θ0 sin(β + 1)θ0 = 0

or

sin 2θ0 β − β sin 2θ0 = 0 (5.187)


sin 2θ0 β + β sin 2θ0 = 0 (5.188)

Therefore, for a wedge with both faces free, β must satisfy Eqn.(5.187) and
it produces antisymmetric displacements (see the b1 − and b3 −components of
displacements in Eqns.(5.177) and (5.178)). In order to produce symmetric
displacements (b2 − and b4 −terms) β must satisfy Eqn.(5.188).The symmetric
and antisymmetric diplacements for the wedge are shown by Fig.(5.9).
The solutions for β are in general complex numbers. For the displacement
field to be finite everywhere we need to have (see Eqns.(5.177) and (5.178))

Re β > 0 (5.189)

For the faces of the wedge being clamped, the boundary conditions are

ur = uθ = 0; θ = ±θ0 (5.190)

116
Stress−free Wedge
2.5

symmetric θ =0.9 π
0
1.5

0.5

−0.5

−1

−1.5

antisymmetric
−2

−2.5
−2 −1.5 −1 −0.5 0 0.5 1 1.5 2
β

Figure 5.10: Equations for β for a stress-free wedge with θ0 = 0.9π.

Then from Eqns.(5.190),(5.177), and (5.178) we obtain

b1 (β + 1) sin(β + 1)θ0 − b3 (4α − β − 1) sin(β − 1)θ0 = 0 (5.191)


b1 (β + 1) cos(β + 1)θ0 + b3 (4α + β − 1) cos(β − 1)θ0 = 0 (5.192)
b2 (β + 1) cos(β + 1)θ0 − b4 (4α − β − 1) cos(β − 1)θ0 = 0 (5.193)
b2 (β + 1) sin(β + 1)θ0 + b4 (4α + β − 1) sin(β − 1)θ0 = 0 (5.194)

Thus for a nontrivial solution for the two systems (5.191),(5.192) and (5.193),(5.194)
we have

(4α − 1) sin 2θ0 β + β sin 2θ0 = 0 (5.195)


(4α − 1) sin 2θ0 β − β sin 2θ0 = 0 (5.196)

where Eqns.(5.195) and (5.196) refer to antisymmatric and symmetric solution,


respectively.
It is apparent from Eqns.(5.179)-(5.181) that the stresses are proportional
to rβ−1 . Thus, the stresses are finite at the corner for Re β ≥ 1 and unbounded
for Re β < 1.
Let’s evaluate the Eqns.(5.187) and (5.188). We choose the angle θ0 = 0.9π
and π/3. The results are displayed by Figs.(5.10) and (5.11).

117
Stress−free Wedge
3

θ =π /3
0
2

antisymmetric
1

−1

−2
symmetric

−3
−2 −1.5 −1 −0.5 0 0.5 1 1.5 2
β

Figure 5.11: Equations for β for a stress-free wedge and θ0 = π/3.

118
y

 
r


  x
O a

Figure 5.12:

It can be shown that for 0 < θ0 < π/2 the stresses are finite at the corner
(e.g., Fig.(5.11)). For π/2 < θ0 < π, the stresses are unbounded at the corner
(e.g., Fig.(5.10)). For the details see Williams[?].

Problem 4 Circular Hole in a Plate

Let’s consider a problem of a plate as shown by Fig.(5.12).


5.199Therefore, in the far field we must have

σxx = σ 0 ; σ yy = σ xy = 0 (5.197)

Thef from the equilibrium of the elements depicted by Fig.(??) we have that

σ rr = σ xx cos2 θ + σ yy sin2 θ + σ xy sin 2θ


σ θθ = σ xx sin2 θ + σ yy cos2 θ − σ xy sin 2θ (5.198)
σrθ = (σyy − σ xx ) sin θ cos θ + σ xy cos 2θ

Consequently, Eqns.(5.197) and (5.198) produce


σ0
σ rr = (1 + cos 2θ)
2
σ0
σ θθ = (1 − cos 2θ) (5.199)
2
σ0
σ rθ = − sin 2θ
2
Let’s try the stress function of the form

φ = f (r) + g(r) cos 2θ (5.200)

119
Then from the general results in polar coordinates we have that
f (r) = A ln r + Br2 ln r + Cr2 + D (5.201)
g(r) = αr2 + βr4 + γr−2 + δ (5.202)
Since the stresses must be finite when r → ∞, we get B = β = 0.
From general results for stresses in polar corrdinates (with n = 0 and n = 2)
by taking into account that the stresses should remain bounded in the far filed,
we have that
A 6γ 4δ
σ rr = + 2C − (2α + 4 + 2 ) cos 2θ
r2 r r
A 6γ
σ rr = − 2 + 2C + (2α + 4 ) cos 2θ (5.203)
r r
6γ 2δ
σ rθ = (2α − 4 − 2 ) sin 2θ
r r
Now the boundary conditions at the hole are given by
σ rr = 0; σrθ = 0; r=a (5.204)
which become
A 6γ 4δ
+ 2C − (2α + 4 + 2 ) cos 2θ = 0
a2 a a
6γ 2δ
(2α − 4 − 2 ) sin 2θ = 0
a a
Since at infinity the fields repesented by Eqns.(5.199) and (5.203) must be
the same we have that
σ0
2C − 2α cos 2θ = (1 + cos 2θ)
2
σ0
2C + 2α cos 2θ = (1 − cos 2θ)
2
σ0
2α sin 2θ = − sin 2θ
2
and thus
σ0 σ0
C= ; α=−
4 4
so that the boundary conditions at the hole imply
σ0 2
A = −2Ca2 = − a
2
6γ 4δ
2α + 4
+ 2 = 0
a a
6γ 2δ
2α − 4 − 2 = 0
a a
Consequently, we obtain
σ0 σ0 2 σ0 4
A = − a2 ; δ= a ; γ=− a
2 2 4

120
and the stress filed becomes
σ0 a2 3a4 4a2
σ rr = [(1 − 2 ) + (1 + 4 − 2 ) cos 2θ]
2 r r r
σ0 a2 3a4
σ θθ = [(1 + 2 ) − (1 + 4 ) cos 2θ] (5.205)
2 r r
σ0 3a4 2a2
σ rθ = − (1 − 4 + 2 ) sin 2θ]
2 r r
The maximum stress is then given by

σ θθ max |r=a = 3σ 0 ; r = a; θ = π/2, 3π/2

121
Chapter 6

Torsion

6.1 Saint-Venant Theory


The paper by Saint-Venant (1853) to the French Academy contains not only the
author’s theory of torsion but it also gives an account of all that was known at
that time in the theory of elasticity.
Let’s consider an elastic cylinder with an axis z, and with the ends at z = 0
and z = L (Fig.(6.1)).
The shaft is subjected at its ends to a ditributed shearing stresses whose
resultant moment is a torque T. Let’s define the rotation angle about the z−axis
as θ. Then recall that
dθ T
= (6.1)
dz Jµ
where J is the polar moment of inertia and µ is the shear modulus.
Let’s define the cross section od the shaft perpedicular to the z−axis as
shown by Fig.(6.2).
For components of displacement field ui , Saint-Venant assumed that as the
shaft twistas the plane ross-sections are worped but the projections on the
xy−plane rotate as a rigid body, i.e.,
ur = 0
uθ = αrz (6.2)
uz = αφ(x, y)
Since
ur = ux cos θ + uy sin θ
uθ = −ux sin θ + uy cos θ
we have that
ux = ur cos θ − uθ sin θ (6.3)
uy = ur sin θ + uθ cos θ (6.4)

122
y

x
T

Figure 6.1: An elastic cylinder.

 zy

r  zx


R x
O

/R=C
n
Figure 6.2: Cross-section of the cylinder subjected to torsional loading.

123
Then from Eqns.(6.2)-(6.4) we have

u = −αyz
v = αxz (6.5)
w = αφ(x, y)

where we have used notation u = ux , v = uy , and w = uz .


The equaions of equilibrium σ ij,j = 0 become

σ xx,x + σ xy,y + σ xz,z = 0


σ yx,x + σ yy,y + σ yz,z = 0 (6.6)
σ zx,x + σ zy,y + σ zz,z = 0

while the boundary condtions on the lateral surface C are

σ xx nx + σ xy ny = 0
σ yx nx + σyy ny = 0 (6.7)
σ zx nx + σ zy ny = 0

Boundary conditions at z = 0 and z = L can be written as (see Fig.(6.2))

σzz = 0 (6.8)
Z
σ zx dxdy = 0 (6.9)
ZR
σ zy dxdy = 0 (6.10)
R
Z
(xσzy − yσ zx )dxdy = T (6.11)
R

where T is the torque applied at the ends. Equations (6.9)-(6.11) state that the
stress field σ zx and σ zy are equipollent to a torque T.
Now Eqns.(6.2) imply that the following stresses are zero

σ xy = σ xx = σ yy = σ zz = 0 (6.12)

while the noziro components of the stress tensor are


∂φ
σ xz = αµ( − y) (6.13)
∂x
∂φ
σyz = αµ( + x) (6.14)
∂y

Therefore, the equations of equilibrium (6.6) reduce to

∂2φ ∂2φ
+ 2 = 0; (x, y) ∈ R (6.15)
∂x2 ∂y

124
Boundary conditions (6.7) become
∂φ
= ynx − xny ; on C (6.16)
∂n
The boundary conditions (6.9)-(6.11) now imply
Z Z
σ zx dxdy = αµ(φ,x − y)dxdy
R R
Z n
£ ¤ £ ¤ o
= αµ x(φ,x − y) ,x + x(φ,y + x) ,y dxdy
ZR
© ª
= αµ x(φ,x − y)nx + x(φ,y + x)ny dx
Zc · ¸
∂φ
= αµ x − ynx + xny ds = 0
C ∂n
due to Eqn.(6.16). Similarly, it follows that
Z
σ zy dxdy = 0
R
so it follows that the end conditions (6.9) and (6.10) are automatically staified.
The end condition (6.11) becomes
T = αµJ (6.17)
where Z
J= (x2 + y 2 + xφ,y − yφ,x )dxdy
R
the polar moment of inertia when the cross-section is circular. The same symbol
is retained for noncircular cross-section.
Therefore, the problem of torsion reduces to solving the following boundary
value problem
∇2 φ(x, y) = 0; (x, y) ∈ R (6.18)
∂φ
= ynx − xny ; (x, y) ∈ C (6.19)
∂n
The ends sections of the shaft are free to warp, and if stresses prescribed on
the end sections are exactly the same as those given by the solution, then the
exact solution is obtained, and the solution is unique. If the strerss distribution
acting on the end sections, while equipollent to the tprque T, does not agree
exactly with those given by Eqns.(6.13) and (6.14), then only an approximate
solution is obtained.
According to the principle proposed by Saint-Venant, the rror in the approx-
imation is signifficant only in the neighborhood of the end sections.
According to the Divergence Theorem, we have that
Z Z
∂φ
∇2 φdxdy = ds = 0 (6.20)
R C ∂n
Therefore, condition for existence of the a solution φ is that the integral of
∂φ/∂n calculated over the entire boundary C must vanish.

125
6.2 Prandtl Theory
In this approach we take the stress components as the principal unknowns. If
we assume that σ xz and σ yz are nonzero, then the only equilibroum equation
to be cosidered is
σ zx,x + σ zy,y = 0 (6.21)
Prandtl proposed to use a stress function ψ(x, y) so that

σ xz = ψ ,y ; σ yz = −ψ ,x (6.22)

Evidently, the equation of equiliberium (6.21) is automatically satisfied. The


boundary condtions on C and the end conditions are specified by Eqns.(6.7)-
(6.11). The compatibility conditons are given by

eij,lk + ekl,ij − eik,jl − ejl,ik = 0 (6.23)

Using the inverse of the Hooke’s law


1+ν ν
eij = σ ij − σkk δ ij
E E
the compatibility conditions become
ν
σ ij,lk + σ kl,ij − σ ik,jl − σ jl,ik = (Θ,kl δ ij + Θ,ij δ kl − Θ,jl δ ik − Θ(6.24)
,ik δ jl
1+ν
Θ = σ kk

Since only six of the 81 equations are linearly independent, by setting k = l


in the last equation and by taking into account that σ ij,j = 0 (no body forces)
the compatibilty equations reduce to
1 ν
σij,kk + Θ,ij − Θ,kk δ ij = 0
1+ν 1+ν
Since σ kk = 0, we get that compatibilty equations become

∇2 σ xz = 0
∇2 σ yz = 0

which i terms of the stress function ψ implies


∂ψ ∂ 2
∇2 ( ) = 0⇒ ∇ ψ=0
∂y ∂y
∂ψ ∂ 2
∇2 ( ) = 0⇒ ∇ ψ=0
∂x ∂x
or
∇2 ψ(x, y) = const in R (6.25)
On the boundary C we have that boundary conditions

σ zx nx + σ zy ny = 0

126
become
∂ψ dy ∂ψ dx dψ
+ = =0
∂y ds ∂x ds ds
or
ψ(x, y) = const on C
For a simply connected regions without loss of generality we can set

ψ(x, y) = 0 on C (6.26)

We still need to examine the end conditions. The condition σ zz = 0 is


satisfied by the starting assumptions. As for the other conditions we have
Z Z Z
∂ψ
σ zx dxdy = dxdy = ψny ds = 0
R R ∂y C

where we have used Eqn.(6.26). Similarly


Z Z Z
∂ψ
σ zy dxdy = − dxdy = − ψnx ds = 0
R R ∂x C

and
Z
T = (xσ zy − yσzx )dxdy
R
Z Z
= − (xψ ,x + yψ y )dxdy = − [(xψ),x + (yψ),y − 2ψ]dxdy
ZR Z R
= − (xψnx + yψny )ds + 2 ψdxdy
C R

which implies Z
T =2 ψdxdy
R
Therefore, in Prandtl’s formulation the torsion of a shaft can be posed as
following boundary-value problem

∇2 ψ(x, y) = const; f or (x, y) ∈ R (6.27)


ψ = 0; f or (x, y) ∈ C (6.28)
Z
2 ψdxdy = T (6.29)
R

We still need to determine the constant in Eqn.(5.183). This can be done


through boundary conditions on displacements. Namely, from

σ zx = µ(w,x + u,z ) = µ(w,x − αy)


σ zy = µ(w,y + v,z ) = µ(w,y + αx)

we have that
σ xz,y − σ yz,x = −2αµ

127
or
∇2 ψ(x, y) = −2αµ (6.30)
where we have used Eqns.(6.22). By comparing Eqns.(6.30) and (6.27) we see
that the unknwn constant is equal to -2αµ.
Therefore, Prandtl’s torsion problem becomes

∇2 ψ(x, y) = −2αµ; f or (x, y) ∈ R (6.31)


ψ = 0; f or (x, y) ∈ C (6.32)
Z
2 ψdxdy = T (6.33)
R
σ xz = ψ ,y ; σ yz = −ψ ,x (6.34)

Example 19 Torsion of an elliptic bar.

In that case the cross-section boundary C is defined by

x2 y 2
C: + 2 =1 (6.35)
a2 b
Let’s try the Prandtl’s stress function of the form

x2 y 2
ψ = A( + 2 − 1) (6.36)
a2 b
Then, using Eqn.(6.31) we obtain

a2 b2
A=− αµ
a2 + b2
and therefore
a2 b2 x2 y 2
ψ=− αµ( + 2 − 1) (6.37)
a2 + b2 a2 b
Now Eqn.(6.33) becomes
Z
2a2 b2 x2 y 2
T =− αµ ( + 2 − 1)dxdy (6.38)
a2 + b2 R a2 b

Consequently, we cosider the integral


Z
x2 y 2
I = ( 2 + 2 − 1)dxdy
R a b
Z a Z √ 2 2 b 1−x /a
x2 y2
= 2 dx ( + − 1)dy
−a 0 a2 b2
Z a
4b x2 3/2 abπ
= − (1 − 2
) dx = −
3 −a a 2

128
Based on the last result qn.(6.38) states that

πa3 b3
T = αµ (6.39)
a2 + b2
The stress field is calculated to be
2T y
σ zx = ψ ,y = − (6.40)
πab3
2T x
σ zy = −ψ ,x = (6.41)
πa3 b
The resultant stress can be calculated according to

τ resul = (σ 2xz + σ 2yz )1/2


2T 1 x2 1 1
= [ 2 − 2 ( 2 − 2 )]1/2
πab b a b a
The maximum resulting stress is then
2T
τ max = ; at x = 0; y = ±b
πab2
Saint-Venant Approach
Since the stresses σ zx = ψ ,y and σ zy = −ψ ,x are known, we have that

σxz = µ(αφ,x − αy) = ψ ,y (6.42)


σ yz = µ(αφ,y + αx) = −ψ ,x (6.43)

Equation (6.42) provides that

1 b2 − a2
φ,x = y + ψ ,y = 2 y
αµ a + b2
and thus
b2 − a2
φ= xy + f (y) (6.44)
a2 + b2
Using Eqns.(6.43) we have that

1 b2 − a2
φ,y = −x − ψ ,x = 2 x
αµ a + b2

and using Eqn.(6.44) we get

f 0 (y) = const = 0

which implies that the worping function is given by

a2 − b2
φ=− xy (6.45)
a2 + b2

129
y

+ -
x
- +

Figure 6.3: w=const lines for an elliptical shaft under torsional loading.

Consequently, the displacement component w is given by

a2 − b2
w = αφ = − αxy (6.46)
a2 + b2
Therefore, the curves w = const are represented by a family of hyperbolas as
shown by Fig.(6.3).
It is evident from Eqn.(6.46) that for a corcular shaft (a = b) there is no
worping of the cross-sections z = const, i.e., w = 0.

130

You might also like