1

MATH45061: SOLUTION SHEET1 III

1.) Balance of angular momentum about a particular point Z requires that
Z
(R − Z) × ρV̇ dV = LZ .
Ω

The moment about the point Z can be decomposed into torque due to the body
force, F , and surface traction T :
Z Z
LZ = (R − Z) × ρF dV + (R − Z) × T dS;
Ω ∂Ω

and so
Z Z Z
(R − Z) × ρV̇ dV = (R − Z) × ρF dV + (R − Z) × T dS. (1)
Ω Ω ∂Ω

If linear momentum is in balance then

Z Z Z
ρV̇ dV = ρF dV + T dS;
Ω Ω ∂Ω

and taking the cross product with the constant vector Z − Y , which can be brought
inside the integral,
Z Z Z
(Z − Y ) × ρV̇ dV = (Z − Y ) × ρF dV + (Z − Y ) × T dS. (2)
Ω Ω ∂Ω

Adding equation (2) to equation (1) and simplifying

Z Z Z
(R − Y ) × ρV̇ dV = (R − Y ) × ρF dV + (R − Y ) × T dS.
Ω Ω ∂Ω

Thus, the angular momentum about any point Y is in balance.

If body couples, L, are present then they need to be included in the angular mo-
mentum balance as an additional term on the right-hand side
Z
L dV.
Ω

This term will be unaffected by the argument above, so we need to demonstrate that
the body couple distribution is independent of the axis about which the angular
momentum balance is taken. For a body couple to be independent of the body
force, the forces that constitute the couple must be in balance because otherwise
an additional contribution to the body force would result. We can write that at a
given point R, the body couple referred to a given axis is
X
LZ (R) = lim (R + δ (i) − Z) × F (i) ,
|δ (i) |→0 i

1
Any feedback to: [email protected]
 2

the limit of the sum of torques acting at slightly different points but the distance
between the points is tending to zero. We assume that such a limit exists and is
finite (which actually requires unbounded forces, because otherwise the forces will
simply cancel). The argument P is now essentially the same as above, because the
forces must be in balance then i F (i) = 0, and so
X X
(Z − Y ) × F (i) = (Z − Y ) × F (i) = 0,
i i

which means that

X 
LZ = lim (R + δ (i) − Z) × F (i) + (Z − Y ) × F (i)
|δ (i) |→0 i
X
= lim (R + δ (i) − Y ) × F (i) = LY ,
|δ (i) |→0 i

so the body couple distribution is a free vector distribution — it is independent of

the axis about which the angular momentum balance is taken. Thus, the result is
still true if body couples are included, provided that our body couple distribution
is defined in a manner consistent with the behaviour of couples that act over finite
distances.

2.) The difficulty of the argument depends on how rigorous you want to be. Here I shall
adopt a middle way. The starting point is to consider a short cylinder (pillbox) that
is formed by extending a disk D of fixed radius about the point R a distance h/2
in both the positive and negative normal directions to the disk.

N
D+
h/2
R Γ

h/2 D−

The surface of the cylinder consists of two disks D+ , D− and the curved surface Γh .
We assume that the total force on the cylinder tends to zero as the volume of the
cylinder tends to zero, i.e. the forces remain bounded, so
Z Z Z
lim T dS + T dS + T dS = 0.
h→0 Γh D+ D−

If the forces remain bounded (which follows from continuity of T ) then because the
area of the curved surface Γh → 0 as h → 0 the first integral vanishes and we have
that Z Z
T dS + T dS = 0.
D+ D−
 3

(If you want to make this really rigorous, you need to use the mean value theorem
for integration.) In this limit the two disks are the same and coincide with the
original disk D, but the normals equal an opposite, so
Z Z
T (N (R), R) dS + T (−N (R), R) dS = 0,
D D

which is true for any size of disk so

Z
T (N (R), R) + T (−N (R), R) = 0 ⇒ T (N (R), R) = −T (−N (R), R).
D

3.) This is a straightforward exercise in vector geometry. The key point is that in the
infinitesimal limit, the tetrahedron is straight-sided with three edges given by Gi .

G3

G3
− G1

1
G2 − G

G2

G1
We now use the standard result that the cross product of two vectors gives a vector
normal to both (with direction given by the right-hand rule) and with magnitude
equal to twice the area of the triangle defined by the two vectors. Thus the vector
area of the face whose edges are not aligned with a coordinate direction is
1
dS = N dS = (G2 − G1 ) × (G3 − G1 ),
2
where the direction of the normal is out of the tetrahedron. Hence,
1
N dS = (G2 × G3 − G2 × G1 − G1 × G3 + G1 × G1 ) ,
2
the final term is zero due to the properties of the cross product, so
1 1 1
N dS = G2 × G3 + G1 × G2 + G1 × G3 .
2 2 2
after using the antisymmetry property of the cross product.
The three terms on the right-hand side are simply the vector areas of the other
three faces of the tetrahedron; and the outer normals to these faces are given
√ by the
i i
contravariant base vectors G . Thus, outer unit normals are given by G / Gi i (no
summation) and we have the required result
X Gi dS(i)
N dS = √ .
G ii
i
 4

4.) Here we need only use the transformation properties of the two vectors. We know
that the vector components transform as
∂XIb
NIb = NJ = QIJ
b NJ .
∂XJ
and so multiplying by the inverse of the matrix gives

NIb Q−1
K Ib
= Q−1 Q b NJ = δKJ NJ = NK .
K Ib IJ

We use these vector transformation properties to determine that

TI = TJI NJ ⇒ TKb Q−1

b = TJI NL
IK
−1
b QJ L
b;

and then multiplying both sides by QM

cI

−1 −1
TM
c = QM
cI TJI NL
b QJ L
b ⇒ TM
c = QM b NL
cI TJI QJ L b,

which means that if in the transformed coordinates TM

c = TL
bMcNL
b , then

−1 −T T T
c = QM
TLb M cI TJI QJ L
b = QLJ
b TJI QI M
c = QLJ
b TJI QI M
c,

because Q is orthogonal. Thus the Cauchy stress tensor does satisfy Cartesian tensor
transformation properties.

5.) a.) The total surface force on the body is given by integrating the traction over the
surface Z Z Z
FS = T dS = −pN dS = ρf gX3 N dS.
∂Ω ∂Ω ∂Ω
We would like to apply the divergence theorem, but the integrand is not of the
correct form. In the examples class, I took the following approach: Consider
the three separate integrals,
Z Z
ρf gX3NI dS = ρf gX3 eI ·N dS, I = 1, 2, 3.
∂Ω ∂Ω

Then applying the divergence theorem to each integral, we obtain

Z Z Z
ρf gX3NI dS = ∇R · (ρf gX3 eI ) dV = ρf gδI3 dV,
∂Ω Ω Ω

which means that

F s = ρf gV e3 .
Alternatively, a more sophisticated method is to take the inner product with
a constant vector a so that
Z
F S ·a = ρf gX3 aI NI dS.
∂Ω

Now we can use the divergence theorem to write

Z Z
∂
F S ·a = (ρf gX3aI ) dV = ρf ga3 dV.
Ω ∂XI Ω
 5

Hence, because a is arbitrary it follows that

Z
F s = ρf g dVe3 = ρf gV e3 ,
Ω

where V is the volume of the body. Hence the resultant surface force is the
mass of water displaced ρf V multiplied by g, which gives the weight of water
displaced and acts in the positive X3 direction.
b.) If we let hRi be the centre of volume then the resultant torque about the centre
of volume is Z
L= (R − hRi) × ρf gX3 N dS.
∂Ω
In a Cartesian coordinate system, the component in the direction eI is
Z
LI = ρf g X3 eIJK (XJ − hXJ i)NK dS.
∂Ω

Thus, using the divergence theorem gives

Z
∂
LI = ρf g (X3 eIJK (XJ − hXJ i) dV
Ω ∂XK
Z
= ρf g δK3 eIJK (XJ − hXJ i) + X3 eIJK δJK dV.
Ω
Now δJK is symmetric on interchange of J and K, but eIJK is antisymmetric
on interchange of J and K so the last term must be zero. Thus,
Z
LI = ρf g eIJ3 (XJ − hXJ i) dV,
Ω

but hXJ i is a constant and so we have

Z 
LI = ρf geIJ3 XJ dV − V hXJ i ,
Ω

where V is the volume of the body. Now by definition of the centre of volume
Z
1
hXJ i = XJ dV,
V Ω
which means that

LI = ρf g eIJ3 [V hXJ i − V hXJ i] = 0 ⇒ L = 0,

as required.

6.) a.) A change from one set of Cartesian coordinates to another does not change the
angle between the coordinate axes. If the X3 direction and handedness remain
fixed, the only possible family of transformations are rotations about the X3
axis. These rotations can be described by a single angle θ, between the X1 and
Xb1 directions:
 6

X2
Xb2

Xb1
θ

θ
X1
b.) The transformation between the two coordinate systems is given by
    
Xb1 cos θ sin θ 0 X1
 Xb2  =  − sin θ cos θ 0   X2  .
Xb3 0 0 1 X3
Thus the transformation rule for the stress tensor is
∂XIb
TIbJb = QIN
b TN M QJM
b , where QIJ
b = ,
∂XJ
so
 
Tb1 b1 Tb1 b2 Tb1 b3
b =  Tb b Tb b Tb b
T 
12 22 23
Tb1 b3 Tb2 b3 Tb3 b3
   
cos θ sin θ 0 T11 T12 0 cos θ − sin θ 0
=  − sin θ cos θ 0  T12 T22 0   sin θ cos θ 0 
0 0 1 0 0 0 0 0 1
  
cos θ sin θ 0 T11 cos θ + T12 sin θ −T11 sin θ + T12 cos θ 0
=  − sin θ cos θ 0  T12 cos θ + T22 sin θ −T12 sin θ + T22 cos θ 0  ;
0 0 1 0 0 0
and so
Tb1 b1 = T11 cos2 θ + T22 sin2 θ + 2T12 cos θ sin θ,
Tb2 b2 = T11 sin2 θ + T22 cos2 θ − 2T12 cos θ sin θ,
Tb1 b2 = (T22 − T11 ) sin θ cos θ + T12 (cos2 θ − sin2 θ).
We can use the double angle formulæ to write
1 1
Tb1 b1 = T11 (1 + cos 2θ) + T22 (1 − cos 2θ) + T12 sin 2θ,
2 2
(T11 + T22 ) (T11 − T22 )
= + cos 2θ + T12 sin 2θ
2 2
1 1
Tb2 b2 = T11 (1 − cos 2θ) + T22 (1 + cos 2θ) − T12 sin 2θ,
2 2
(T11 + T22 ) (T22 − T11 )
= + cos 2θ − T12 sin 2θ,
2 2
(T22 − T11 )
Tb1 b2 = sin 2θ + T12 cos 2θ.
2
 7

c.) If Tb1 b2 = 0, then

(T22 − T11 ) 2T12

sin 2θ + T12 cos 2θ = 0 ⇒ tan 2θ = .
2 (T11 − T22 )

The maximum and minimum values of Tb1 b1 and Tb2 b2 occur when ∂Tb1 b1 /∂θ = 0
and ∂Tb2 b2 /∂θ = 0, respectively; and

∂Tb1 b1
= sin 2θ(T22 − T11 ) + 2T12 cos 2θ = 2Tb1 b2 ,
∂θ
∂Tb2 b2
= sin 2θ(T11 − T22 ) − 2T12 cos 2θ = −2Tb1 b2 .
∂θ
Hence, when Tb1 b2 = 0, the values of Tb1 b1 and Tb2 b2 are either at their maximum
or minimum. In other words, these are the principal stresses.
d.) The maximum value of Tb1 b2 occurs when ∂Tb1 b2 /∂θ = 0, i. e.

(T22 − T11 )
(T22 − T11 ) cos 2θ − 2T12 sin 2θ = 0 ⇒ tan 2θ = .
2T12
If we indicate the value of θ that extremises the normal stress by θn and that
corresponding to the maximum value of Tb1 b2 by θs , then

1 sin 2θn cos 2θs

tan 2θn = − ⇒ =− ,
tan 2θs cos 2θn sin 2θs
⇒ sin 2θn sin 2θs + cos 2θn cos 2θs = 0 ⇒ cos 2(θn − θs ) = 0;
and therefore
π π
2(θn − θs ) = ± ⇒ θn = θs ± .
2 4
Thus the angle at which Tb1 b2 gains its maximum value is ±45◦ from the angle
at which the Tb1 b1 is maximised or minimised.

7.) Note that we shall drop the overbars for convenience in the working here, but we are
working in the Eulerian coordinate system. The principal invariants of the Cauchy
stress tensor are given by

|Tji − µδji | = −µ3 + I1 µ2 − I2 µ + I3 ,

so
I1 = Tii,
1 i j

I2 = Ti Tj − Tji Tij ,
2
I3 = |Tji|.

8.) The principal invariants of the stress deviation tensor are given by
1
|Teji − µδji | = |Tji − Tkk δji − µδji | = −µ3 + J1 µ2 − J2 µ + J3 .
3
 8

Thus,
1 1
J1 = Tii − Tkk δii = Tii − Tkk × 3 = 0.
3 3
1 i




J2 = Ti − Tkk δii /3 Tjj − Tmm δjj /3 − Tji − Tkk δji /3 Tij − Tmm δij /3
2
1 
= J1 J1 − Tji Tij + Tkk Tjj /3 + Tii Tmm /3 − Tkk Tmm δji δij /9 .
2
We know that J1 = 0 and δji δij = δjj = 3, so
 
1 1 i j i j 1 i j  1
J2 = Ti Tj − Tj Ti = Ti Tj − Tji Tij − Tii Tjj = I2 − (I1 )2 /3,
2 3 2 3

1
J3 = Tji − Tkk δji
3

9.) a.) From the definition

T = TN N + T S ,
and taking the dot product with the unit normal gives

T · N = TN ;

and by the definition of the Cauchy stress tensor

TN = TJI NI NJ = TIJ NI NJ ,

after using the symmetry property of the Cauchy stress tensor as well.
b.) Taking the dot product of the stress vector with itself and using the fact that
N and T S are orthogonal gives

T · T = (TN N + T S )·(TN N + T S ) = TN2 + T S ·T S ,

so
|T S |2 = |T |2 − TN2 . (3)
If we work in the Cartesian coordinate system given by the principal axes of
stress then the Cauchy stress tensor is diagonal with entries corresponding
to the principal stresses, say σ(1) , σ(2) and σ(3) . Hence from the definition
TI = σ(I) NI (no summation), and so

|T |2 = (σ(1) N1 )2 + (σ(2) N2 )2 + (σ(3) N3 )2 .

In addition,
 2
T 2N = (TIJ NI NJ )2 = σ(1) N12 + σ(2) N22 + σ(3) N32 ,

and so by direct substitution into equation (3)

 2
|T S |2 = (TIJ NI NJ )2 = (σ(1) N1 )2 +(σ(2) N2 )2 +(σ(3) N3 )2 − σ(1) N12 + σ(2) N22 + σ(3) N32 .
2


= σ(I) (NI2 − NI4 ) − 2 σ(1) σ(2) N12 N22 + σ(1) σ(3) N12 N32 σ(2) σ(3) N22 N32 .
 9

This expression can be simplified by spotting that

N12 − N14 = N12 (1 − N12 ) = N12 (N22 + N32 ),

because N is a unit vector. Using similar expressions for N22 −N24 and N32 −N34 ,
we obtain

|T S |2 = N12 N22 (σ(1)

2 2
+ σ(2) − 2σ(1) σ(2) ) + N22 N32 (σ(2)
2 2
+ σ(3) − 2σ(2) σ(3) )

+N12 N32 (σ(1)

2 2
+ σ(3) − 2σ(1) σ(3) )
⇒ |T S |2 = [N1 N2 (σ(1) −σ(2) )]2 + [N2 N3 (σ(2) −σ(3) )]2 + [N1 N3 (σ(1) −σ(3) )]2 . (4)
c.) We continue to work in the coordinate system given by the principal axes of
stress. The cosine of the angle between a plane and the I-th coordinate axis is
given by cos θ(I) = NI = N ·eI , where N is the unit normal to the pane and
eI is a unit base vector in the coordinate direction. The result follows directly
from the formula A · B = |A||B| cos φ, where φ is the angle between the two
vectors A and B.
If a plane makes the same angle with each of the three principal axes of stress
then
cos θ(1) = cos θ(2) = cos θ(3) ⇒ N1 = N2 = N3 ,
but N is a unit vector, so
1
N12 + N22 + N32 = 1 ⇒ 3N12 = 1 ⇒ N12 = N22 = N32 = .
3
The square of the magnitude of the shear stress on this plane is found from
equation (4)
1 
|T S |2 = (σ(1) − σ(2) )2 + (σ(2) − σ(3) )2 + (σ(1) − σ(3) )2 .
9
In our chosen coordinate system the invariant J2 is found from question 8 to
be
1
J2 = I2 − (I1 )2 /3 = (TII TJJ − TIJ TIJ ) − (TII )2 /3
2
1  2 2 2

= (σ(1) + σ(2) + σ(3) )(σ(1) + σ(2) + σ(3) ) − σ(1) − σ(2) − σ(3)
2
−(σ(1) + σ(2) + σ(3) )2 /3.
 2 2 2

= σ(1) σ(2) +σ(1) σ(3) +σ(2) σ(3) − σ(1) + σ(2) + σ(3) + 2σ(1) σ(2) + 2σ(1) σ(3) + 2σ(2) σ(3) /3
1 2 2 2
 19
= σ(1) σ(2) + σ(1) σ(3) + σ(2) σ(3) − σ(1) − σ(2) − σ(3) =− |T S |2 ,
3 32
which means that the required magnitude is proportional to J2 .
 10

10.) a.) Using the hint, we consider

(XJ TIK ),K = XJ,K TIK + XJ TIK,K , (5)
by the product rule. The equations of equilibrium in Cartesian coordinates are
TJI,J + FI = TIJ,J + FI = 0, (6)
because F is a body force per unit volume (not per unit density) and the
Cauchy stress tensor is symmetric Using equation (6) in (5) gives
(XJ TIK ),K = δJK TIK + XJ (−FI ),
which gives the result in the hint
TIJ = (XJ TIK ),K + XJ FI .
Taking the volume average of the equation, we have
Z Z
1 1
hTIJ i = TIJ dV = (XJ TIK ),K + XJ FI dV,
V Ω V Ω
and using the divergence theorem on the first integral
Z Z 
1
hTIJ i = XJ TIK NK dS + XJ FI dV ,
V ∂Ω Ω

where NI are the Cartesian components of the outer unit normal to the body.
From the definition of the stress tensor TIK NK = τI on the body’s surface and
writing the equation in dyadic form gives the result (ahem, actually we also
use the symmetry property of the stress tensor)
Z Z 
1
hTi = R ⊗ τ dS + R ⊗ F dV .
V ∂Ω Ω

This is known as Signorini’s theorem and it demonstrates that the average

stress in the body can be determined directly from the surface traction τ and
the body force F .
b.) If F = 0 and the surface is loaded by a uniform pressure then τ = −pN , where
N is the outer unit normal to the surface. Hence from Signorini’s theorem
Z Z
1 p
hTi = R ⊗ (−pN ) dS = − R ⊗ N dS,
V ∂Ω V ∂Ω
because the pressure is uniform (and therefore p is constant). Using the diver-
gence theorem and working in component form
Z Z Z
RI NJ dS = ΩRI,J dV = δIJ Ω dV = δIJ V ;
∂Ω

so back in matrix form

hTi = −p I.
If we choose T = hTi = −p I, then the equilibrium equations, in the absence of
a body force, are satisfied because
TIJ,J = −p δIJ,J = 0.
 11

11.) This question is just to check that you understand the colon notation
D D DAIJ DAIJ
A : A = AIJ AIJ ⇒ (A : A) = (AIJ AIJ ) = AIJ + AIJ ,
Dt Dt Dt Dt
so
D DA
(A : A) = 2A : .
Dt Dt
12.) The second Piola–Kirchhoff stress tensor is given by

s = JF−1 σF−T .

Hence the material derivative, by the product rule is

˙ −1 σF−T + J Ḟ−1 σF−T + JF−1 σ̇F−T + JF−1 σ Ḟ−T .
ṡ = JF (7)

From the lecture notes (or working it out again),

J˙ = J divV = J tr(D), and Ḟ = LF.

Now because

FF−1 = I ⇒ ḞF−1 + FḞ−1 = 0 ⇒ Ḟ−1 = −F−1 ḞF−1 = −F−1 L,

and so
Ḟ−T = −LT F−T .
Using these results into the expression (7) we obtain the required stress rate
 
ṡ = JF−1 σ̇ + tr(D) σ − Lσ − σLT F−T .

The quickest way to establish the objectivity, or not of ṡ is to think about s. Under
a change in observer
s∗ = J ∗ (F∗ )−1 σ ∗ (F∗ )−T .
The Cauchy stress tensor is objective by construction (it is a Cartesian tensor), see
question 4, so σ ∗ = QσQT . We can look up the transformation properties of F in
the lecture notes to find that F∗ = QF. The Jacobian J is a scalar invariant because
F∗ = QF, which means that

J ∗ = |F∗ | = |QF| = |Q||F| = |F| = J,

because |Q| = 1 by the orthogonality of Q. Hence,

s∗ = J(QF)−1 QσQT (QF)−T = JF−1 QT QσQT QF−T = JF−1 σF−T = s,

which means that the second Piola–Kirchhoff stress tensor is invariant under Eu-
lerian observer transformation. This shouldn’t be a surprise, it’s a Lagrangian
measure of the stress. By taking the material derivative of the above equation we
have that
ṡ∗ = ṡ,
and so ṡ is also invariant under change of observer, so it is not objective.
 12

T
13.) If we invent the symbol A to represent the Truesdell rate, then
T DA∗
A∗= + tr(D∗ )A∗ + L∗T A∗ + A∗ L∗ ,
Dt

= Q̇AQT + QȦQT + QAQ̇T + tr(D)QAQT + (QLT QT + ΩT )QAQT + QAQT (QLQT + Ω)

h i
= Q Ȧ + tr(D)A + LT A + AL QT + ΩQAQT − QAQT Ω + ΩT QAQT + QAQT Ω,

and using the fact that Ω = −ΩT , we have

T T
A∗= QAQT ,

which proves that the Truesdell rate is objective.