Stress

The following are the three important equations for the stress:

1. (definition)
ti = σij nj .

2. (balance of force)
σij,j + bi = 0.

3. (balance of moment)
σij = σji .

To prove the equilibrium equation and the balance of moment equation, you are reminded of the Gauss
theorem, i.e.
I ∫ ∫
ni vdl = v,i dS, (1)
∂D D
where the integral in the left hand side is a contour integral defined along the boundary of the domain D
and the integral in the right hand side is a 2-D integral over D.

Balance of the force on a continua is expressed as

I ∫ ∫
ti dl + bi dS = 0. (2)
∂D D
Using the definition of the stress, the first term in the integral above is rewritten as

I I
ti dl = σij nj dl (3)
∂D
∫ ∫
∂D

= σij,j dS, (4)

D

so that the above becomes

∫ ∫
(σij,j + bi )dS = 0, (5)
D
which is equivalent to

σij,j + bi = 0. (6)
The balance of moment is expressed as
I ∫ ∫
r × tdl + r × bdS = 0, (7)
∂D D
or
I ∫ ∫
ϵijk xj tk dl + ϵijk xj bk dS = 0. (8)
∂D D
Noting that the first integral above is
I I
ϵijk xj tk dl = ϵijk xj σkl nl dl, (9)
∂D ∂D

the above is written as

1
 ∫ ∫ ∫ ∫
(ϵijk xj σkl ),l dS + ϵijk xj bk dS = 0, (10)
D D
or

(ϵijk xj σkl ),l + ϵijk xj bk = 0 (11)

ϵijk xj,l σkl + ϵijk xj σkl,l + ϵijk xj bk = 0, (12)

ϵijk δjl σkl + ϵijk xj (σkl,l + bk ) = 0, (13)

ϵijk σkj = 0, (14)

which leads to

σij = σji . (15)

(Stress) Boundary conditions

When two materials are in contact each other, the traction force (physics) and the displacement (geometry)
must be continuous across the material boundary.
Example 1:
[

[
Along the boundary, n = (0, 1) so that

t1 = σ11 n1 + σ12 n2 = σ12 (16)

t2 = σ21 n1 + σ22 n2 = σ22 , (17)

so both σ12 and σ22 must be continuous across the boundary but σ11 does not have to be continuous.
Example 2:

2
 [

[
√ √
This time, the normal, n, is n = (1/ 2, 1/ 2) so that

√
t1 = σ11 n1 + σ12 n2 = (σ11 + σ12 )/ 2 (18)
√
t2 = σ21 n1 + σ22 n2 = (σ21 + σ22 )/ 2, (19)

so σ11 + σ12 and σ21 + σ22 must be continuous but individual components of the stress tensor do not have
to be continuous.

Principal stresses

W
Q

When the surface traction is perpendicular to the surface, we say that the state of stress is principal stresses.
When this happens,

t = αn. (20)
Combining eq.(20) with t = Sn (ti = σij nj ), it follows

Sn = αn (21)
= αIn, (22)

where I is the identity matrix.

Equation (22) is equivalent to

(S − αI)n = 0, (23)
which has a non-trivial solution if the matrix S − αI is singular, i.e.,

|S − αI| = 0. (24)

3
By solving eq.(24), the principal stresses, α, can be found. The principal direction, n, can be found by
back-substituting α into eq.(22).
In-3D, eq.(24) is written as
 
 σ11 − α, σ12 , σ13 

 σ21 , σ22 − α, σ23  = 0, (25)

 σ31 , σ32 , σ33 − α 
whose expansion yields

α3 − I1 α2 + I2 α − I3 = 0, (26)
where

I1 = σ11 + σ22 + σ33 , (27)

I2 = −σ12
2
− σ23
2
− σ312
+ σ11 σ22 + σ22 σ33 + σ33 σ11 (28)
     
 σ22 σ23   σ33 σ31   σ11 σ12 

=   +   +  , (29)
σ32 σ33   σ13 σ11   σ21 σ22 
I3 = −σ132
σ22 − σ23
2
σ11 − σ12
2
σ33 + 2σ12 σ13 σ23 + σ11 σ22 σ33 (30)
 
 σ11 σ12 σ13 
 
=  σ21 σ22 σ23  . (31)
 σ31 σ32 σ33 

It is clear that the quantities, I1 , I2 and I3 are all scalars as the cubic equation they define is invariant.

Therefore, I1 , I2 and I3 must be 0-th rank tensors and hence it follows that they can be expressed in tensorial
form of σij .
It is easy to see that
I1 = σii .
For I2 and I3 , we can choose the coordinate system in which the principal stresses are realized. In this
coordinate system, the stress invariants above become

I1 = σ1 + σ2 + σ3 , (32)
I2 = σ1 σ2 + σ2 σ3 + σ3 σ1 , (33)
I3 = σ1 σ2 σ3 , (34)

where σ1 , σ2 and σ3 are the principal stresses. If we can show that the above quantities can be written in
tensor invariants in this specific coordinate system, the tensor expressions must be valid in general coordinate
systems.

I1 = σii (35)
σii σjj − σij σij
I2 = (36)
( 2 )
I3 = σii σjj σkk − σij σjk σki − 3 (σii σkl σkl − σij σjk σki ) /6. (37)

Example
 
−1 2 2
σij =  2 −2 −1  . (38)
2 −1 −2
The eigenvalues can be found by solving

4
  
 −1 − λ, 2 2 
 
 2, −2 − λ −1  = −λ3 − 5λ2 + λ + 5 = −(λ − 1)(λ + 1)(λ + 5) = 0 (39)
 
 2 −1 −2 − λ 

so (λ1 , λ2 , λ3 ) = (−1, 1, 5) and the eigenvectors can be found by solving

    
−1 − λ, 2 2 n1 0
 2, −2 − λ −1   n2  =  0  . (40)
2 −1 −2 − λ n3 0
For λ1 = −1, the above equations are

0n1 + 2n2 + 2n3 = 0 (41)

2n1 − n2 − n3 = 0 (42)
2n1 − n2 − n3 = 0 (43)

from which
n1 = 0, n2 + n3 = 0,
so one can choose (after normalizing)  
0√
n =  −1/√ 2  . (44)
1/ 2
Similarly, for λ2 = 1,
 √ 
2/√6
n =  1/√6  , (45)
1/ 6
and for λ3 = −5,  √ 
−1/√ 3
n =  1/√3  . (46)
1/ 3
More examples on eigenvalues/eigenvectors (from ME5332)

Stress deviator
Any symmetric second rank tensor, vij , can be uniquely decomposed into the shear part (deviatoric part)
and the volume expansion part (hydrostatic part) as1

′ 1
vij = vij + δij vkk . (47)
3
Example:  
1 2 3
vij =  2 4 5 .
3 5 6
Noting vii = 1 + 4 + 6 = 11,
1 For 2-D, the decomposition is
′ 1
vij = vij + δij vkk .
2

5
 ′ 1
vij = vij − δij vkk (48)
 3   
1 2 3 11 0 0
1
=  2 4 5  −  0 11 0  (49)
3
3 5 6 0 0 11
 
−8/3 2 3
=  2 1/3 5 . (50)
3 5 7/3
2
It should be noted that the trace of the deviator tensor is always zero as
′
vii = 0. (51)
If we set i = j in eq.(47),

′ 1
vii = vii + δii vkk (52)
3
′
= vii + vkk . (53)

Since vii = vkk , it follows

′
vii = 0, (54)
i.e. the deviatoric part represents part of deformation that does not accompany volume expansion 3 .
Example: Strain energy is defined as
1
U ≡ σij ϵij . (59)
2
Using the decomposition of σij and ϵij as

′ 1
σij = σij + δij σkk , (60)
3
′ 1
ϵij = ϵij + δij ϵkk . (61)
3

( )( )
1 ′ 1 ′ 1
U = σij + δij σkk ϵij + δij ϵkk (62)
2 3 3
( )
1 ′ ′ 1 ′ 1 ′ 1
= σij ϵij + δij σij ϵll + δij ϵij σkk + δij δij σkk ϵll (63)
2 3 3 9
( )
1 ′ ′ 1 ′ 1 ′ 1
= σij ϵij + σjj ϵll + σkk ϵjj + σkk ϵll (64)
2 3 3 3
( )
1 ′ ′ 1
= σij ϵij + σkk ϵll . (65)
2 3
2 The sum of diagonal elements.
3

∆V (a + ∆a)(b + ∆b)(c + ∆c) − abc

= (55)
V abc
∆a ∆b ∆c
∼ + + (56)
a b c
= ϵx + ϵy + ϵz (57)
= ϵii . (58)

6
i.e. the deviatoric part and the hydrostatic part of the strain energy do not couple each other.

Shear deformation can be defined as the deviatoric part of either the stress or strain tensor (more later).
Similar to the stress invariants, I1 , I2 and I3 , of σij , the stress invariants, J1 , J2 and J3 of the deviatoric part
of the stress can be defined as

′
J1 = σii (66)
= 0, (67)
1 ′ ′ ′ ′
J2 = (σ σ − σij σij ) (68)
2 ii jj
1 ′ ′
= − σij σij , (69)
( 2′ ′ ′ ′ ′ ′ ′ ′ ′ ′ ′ ′
)
J3 = σii σjj σkk − σij σjk σki − 3(σii σkl σkl − σij σjk σki ) /6 (70)
′ ′ ′
= −σij σjk σki /6. (71)

Recall that metal failure is defined as the yielding point, which is caused by dislocation of atoms and the
shear force is required to cause dislocation. Hence the deviatoric part of σij is responsible for the yielding
condition and yielding must be a function of the stress deviatoric parts. Its simplest form is

J2 = c, (72)
or
′ ′
σij σij = c. (73)
If this equation is written in 2-D principal stresses, σ1 and σ2 , the von Mises failure criterion is derived,
i.e.  
σ1 0 0
σij =  0 σ2 0  , (74)
0 0 0
so

′ 1
σij = σij − δij σkk (75)
 3   σ1 +σ2 
σ1 0 0 3 0 0
=  0 σ2 0  −  0 σ1 +σ2
3 0  (76)
σ1 +σ2
0 0 0 0 0 3
 2σ1 −σ2 
3 0 0
=  0 2σ2 −σ1
3 0 , (77)
−σ1 −σ2
0 0 3

hence

( )2 ( )2 ( )2
′ ′ 2σ1 − σ2 2σ2 − σ1 −σ1 − σ2
σij σij = + + (78)
3 3 3
2 2
= (σ − σ1 σ2 + σ22 ) (79)
3 1
= c, (80)

i.e.
σ12 − σ1 σ2 + σ22 = c. (81)
′ ′
On the other hand, for the uniaxial yield strength, σY , σij σij can be computed as

7
  
σY 0 0
σij =  0 0 0 , (82)
0 0 0

   
σY 0 0 1 0 0
σ
′
=  0 0 0 −  0 
Y
σij 0 1 (83)
3
0 0 0 0 0 1
 2 
3 σY 0 0
=  0 − 13 σY 0 , (84)
0 0 − 13 σY
so
′ ′2 2
σij σ .
σij = (85)
3 Y
Hence the yield condition under biaxis stresses is written as

σ12 − σ1 σ2 + σ22 = σY2 . (86)

Mohr’s Circle
By drawing Mohr’s circle, the stress transformation in 2-D can be done graphically.

σx + σy σx − σy
σ = + cos 2θ + τxy sin 2θ, (87)
2 2
σx − σy
τ = − sin 2θ + τxy cos 2θ. (88)
2
Note that the angle of rotation, θ, is defined by comparing the two normals on the surfaces.

σ\

τ
τ[\ σ

σ[ θ

How to draw a Mohr circle

Each point on a Mohr circle represents the orientation of surface. Therefore, one can start by mapping each
side of the element to a corresponding point on the Mohr circle, i.e.

1. Normal Stress ⇒ Horizontal axis

(a) Tensile stress ⇒ Right side
(b) Compressive Stress ⇒ Left side

8
 2. Shear Stress ⇒ Vertical axis

(a) Clockwise rotation ⇒ Upper side

(b) Counterclockwise rotation ⇒ Lower side

3. Draw a circle by connecting the four corners.

4. Rotation of θ in the physical elements is translated into 2θ on the Mohr circle space.

Note that we avoided using terms such as “positive rotation” etc.. on purpose.

Example: Draw Mohr’s circle for (σx , σy , τxy ) = (3, −4, −2).

 FZURWDWLRQ

 FRPSUHVVLRQ  $ WHQVLRQ
%
$  
 
%

FFZURWDWLRQ
Example 1: Using Mohr’s circle, one can explain why the angle of the failure surface of a metal bar is 45
degrees. The maximum shear stress theory in failure of metals postulates that Failure (yielding) of metals is
predicted when the maximum shear stress reaches a critical value. When a metal bar is subject to uniaxial
tension, the corresponding Mohr’s circle is the one centered at (σ/2, 0) with a radius of σ/2. So the maximum
shear stress is at the top of the circle (σ/2, σ/2) which is rotation from the surface subject to the uniaxial
tension by 90 degrees which is translated into rotation of the physical element by 45 degrees.

Example 2: Drawing Mohr’s circle can also explain why hydrostatic stresses never cause failure in metals.
If an element is subject to hydrostatic stresses (σx = σy = σ), the corresponding Mohr’s circle is degenerated
to a single point whose maximum shear stress remains 0.

Strains
It should be noted that the stress is a physical quantity while the strain is a geometrical quantity.

O O

If one wants to measure the rate of deformation based on the original length, one can define
ℓ − ℓo
E≡ . (89)
ℓo
On the other hand, the rate of deformation can be also defined based on the present length as
ℓ − ℓo
e≡ . (90)
ℓ

9
Since the definitions are different, their values are different naturally. However, if one assumes that the
deformation is infinitesimal, i.e.

ℓ = ℓo (1 + ϵ), (91)
where ϵ is an infinitesimal quantity,

ℓo (1 + ϵ) − ℓo
E = (92)
ℓo
= ϵ. (93)
ℓo (1 + ϵ) − ℓo
e = (94)
ℓ
ϵ
= (95)
1+ϵ
= ϵ(1 − ϵ + ϵ2 − ϵ3 + . . .) (96)
∼ ϵ, (97)

the two definitions coincide.

When a point in a deformable continua moves from a to x as shown in the figure, the rate of deformation
can be measured by comparing the length of two infinitesimal vectors at the both points as

dx
da
u
x
a

ds2 = |dx|2 − |da|2 (98)

= dxi dxi − dai dai . (99)

If this quantity is to be expressed based on the original point, a, the position, x, can be regarded as a
function of a as

xi = xi (aj ), (100)
so that
∂xi
dxi = daj . (101)
∂aj
Substituting this to eq.(99), one obtains

10
 ds2 = dxi dxi − dai dai (102)
∂xi ∂xi
= daj dak − dai dai (103)
∂aj ∂ak
∂xi ∂xi
= daj dak − δjk daj dak (104)
∂aj ∂ak
( )
∂xi ∂xi
= − δjk daj dak (105)
∂aj ∂ak
≡ 2Ejk daj dak , (106)

where ( )
1 ∂xi ∂xi
Ejk ≡ − δjk . (107)
2 ∂aj ∂ak
This is called Green’s strain (Lagrangian strain in fluid mechanics).

On the other hand, if one wants to measure the rate of deformation based on the current position, the point
a can be regarded as a function of x as

ai = ai (xj ), (108)
so that
∂ai
dai = dxj . (109)
∂xj
Substituting this to eq.(99), one obtains

ds2 = dxi dxi − dai dai (110)

∂ai ∂ai
= dxi dxi − dxj dxk (111)
∂xj ∂xk
∂ai ∂ai
= δjk dxj dxk − dxj dxk (112)
∂xj ∂xk
( )
∂ai ∂ai
= δjk − dxj dxk (113)
∂xj ∂xk
≡ 2ejk dxj dxk , (114)

where ( )
1 ∂ai ∂ai
ejk ≡ δjk − . (115)
2 ∂xj ∂xk
This is called Cauchy’s strain (Euler’s strain in fluid mechanics).

In short, Eij and eij are different in values as the definitions are different. Nevertheless, their combination
with the infinitesimal elements are the same, i.e.

ds2 = 2Eij dai daj (116)

= 2eij dxi dxj . (117)

It is possible to express Green’s strains and Cauchy’s strains in terms of the displacement, i.e.

ui = xi − ai . (118)

11
For Green’s strains,
xi (aj ) = ui (aj ) + ai , (119)
so that
∂xi ∂ui
= + δij . (120)
∂aj ∂aj
( )
1 ∂xi ∂xi
Ejk = − δjk (121)
2 ∂aj ∂ak
( )
1 ∂ui ∂ui
= ( + δij )( + δik ) − δjk (122)
2 ∂aj ∂ak
( )
1 ∂ui ∂ui ∂uk ∂uj
= ( + + + δjk ) − δjk (123)
2 ∂aj ∂ak ∂aj ∂ak
( )
1 ∂uk ∂uj ∂ui ∂ui
= + + . (124)
2 ∂aj ∂ak ∂aj ∂ak
For Cauchy’s strains,

ak (xi ) = xk − uk (xi ), (125)

so that
∂ak ∂uk
= δki − . (126)
∂xi ∂xi
( )
1 ∂uk ∂uk
eij = δij − (δki − )(δkj − ) (127)
2 ∂xi ∂xj
( )
1 ∂ui ∂uj ∂uk ∂uk
= δij − (δij − − + ) (128)
2 ∂xj ∂xi ∂xi ∂xj
( )
1 ∂ui ∂uj ∂uk ∂uk
= + − . (129)
2 ∂xj ∂xi ∂xi ∂xj
If the deformation is small such that the second order terms in eqs.(124,129) are dropped, eq.(124) and
eq.(129) become the same and the both can be expressed as
( )
1 ∂ui ∂uj
ϵij ≡ + . (130)
2 ∂xj ∂xi
Note that this definition differs from the definition of engineering strain by 1/2 for the shear components.
Problem
A square plate is rotated from the first quarter to the third quarter as shown below. Compute the strains.

(a,b)

(x,y)

12
Answer
Since
x = −a
(131)
y = −b,

u = x − (−x)
= 2x,
(132)
v = y − (−y)
= 2y,
so
∂u
ϵ11 = ( ∂x = 2,
)
1 ∂u ∂v
ϵ12 = 2 ∂y + ∂x = 0, (133)
∂v
ϵ22 = ∂y = 2,

but the rotation above does not accompany any deformation (rigid body rotation). What went wrong ???

Solution
The displacement in the rotation above is not small so one has to use the formula for finite deformation
(either Eij (Green’s strains) or eij (Cauchy’s strain)). Let’s compute the Green strain, Eij .

u = −2a,
(134)
v = −2b,
so
( )
E11 = 1 ∂u1
+ + ∂u
∂u1 k ∂uk

(2 ∂a1
( ∂u )2 ( ∂v )2 )
∂a1
∂a1 ∂a1
1 ∂u ∂u
= + + + ∂a (135)
2 ∂a
1
( ∂a ∂a
2
)
= 2 (−2) + (−2) + (−2)
= 0.
Other Eij ’s can be shown to be 0 (verify!).

13