Professional Documents
Culture Documents
September 2, 2009
Exercise 1: Tensors and Invariants
Tensor/Index Notation
Scalar (0th order tensor), usually we consider scalar elds function of space and time
p = p(x, y, z, t)
Vector (1st order tensor), dened by direction and magnitude
( u)
i
= u
i
If u =
u
v
w
then u
2
= v
Matrix (2nd order tensor)
(A)
ij
= A
ij
If A =
a
11
a
12
a
13
a
21
a
22
a
23
a
31
a
32
a
33
then A
23
= a
23
Kronecker delta (2nd order tensor)
ij
= (I)
ij
=
1 if i = j
0 if i = j
To indicate operation among tensor we will use Einstein summation convention (summation over repeated
indices)
u
i
u
i
=
3
i=1
u
i
u
i
i is called dummy index (as opposed to free index) and can be renamed
Example: Kinetic energy per unit volume
1
2
| u|
2
=
1
2
(u
2
+ v
2
+ w
2
) =
1
2
u
i
u
i
Matrix/Tensor operations
( a
b) = a
1
b
1
+ a
2
b
2
+ a
3
b
3
= a
i
b
i
=
ij
a
i
b
j
= a
j
b
j
(scalar, inner product)
( a
b)
ij
= ( a
b)
ij
= a
i
b
j
(diadic, tensor product)
(A
b)
i
= A
ij
b
j
(matrix-vector multiplication, inner product)
(AB)
ij
= (A B)
ij
= A
ik
B
kj
(matrix multiplication, inner product)
(AB)
ijkl
= (AB)
ijkl
= A
ij
B
kl
(diadic, tensor product)
(A : B) = A
ij
B
ij
(double contraction)
tr(A) = A
11
+ A
22
+ A
33
= A
ii
(trace)
(A)
ij
= A
ij
(A
T
)
ij
= A
ji
(transpose)
1
Permutation symbol
ijk
=
b =
e
1
e
2
e
3
a
1
a
2
a
3
b
1
b
2
b
3
= e
1
(a
2
b
3
a
3
b
2
) e
2
(a
1
b
3
a
3
b
1
) + e
3
(a
1
b
2
a
2
b
1
) =
ijk
e
i
a
j
b
k
( a
b)
i
=
ijk
a
j
b
k
ijk
ilm
=
jl
km
jm
kl
ijk
ijm
= {l = j} = 3
km
jm
kj
= 3
km
mk
= 2
km
ijk
ijk
= {m = k} = 2 3 = 6
Example: Rewrite without the cross product:
( a
b) ( c
d) = ( a
b)
i
( c
d)
i
=
ijk
a
j
b
k
ilm
c
l
d
m
=
ijk
ilm
a
j
b
k
c
l
d
m
=
(
jl
km
jm
kl
) a
j
b
k
c
l
d
m
= a
j
c
j
b
k
d
k
a
j
d
j
b
k
c
k
= ( a c)(
b
d) ( a
d)(
b c)
Tensor Invariants
b
ij
, b
2
ij
= b
ik
b
kj
= (b
ij
)
2
, b
3
ij
= b
ik
b
kl
b
lj
= (b
ij
)
3
Any scalar obtained from a tensor (e.g. b
ii
, b
2
ii
, . . . ) is invariant, i.e. independent of the coordinate system.
The principal invariants are dened by (
i
are the eigenvalues of B)
I
1
= b
ii
= tr(B) =
1
+
2
+
3
I
2
=
1
2
(b
ii
)
2
b
2
ii
=
1
2
[tr(B)]
2
tr(B
2
)
=
1
2
+
2
3
+
1
3
I
3
=
1
6
(b
ii
)
3
1
2
b
ii
b
2
jj
+
1
3
b
3
ii
= det(B) =
1
3
Decomposition of Tensors
T
ij
= T
S
ij
+ T
A
ij
symmetric and anti-symmetric parts
T
S
ij
=
1
2
T
ij
+ T
ji
= T
S
ji
symmetric
T
A
ij
=
1
2
T
ij
T
ji
= T
A
ji
anti-symmetric
The symmetric part of the tensor can be divided further into a trace-less and an isotropic part:
T
S
ij
=
T
ij
+
T
ij
T
ij
= T
S
ij
1
3
T
kk
ij
trace-less
T
ij
=
1
3
T
kk
ij
isotropic
This gives:
2
T
ij
= T
S
ij
+ T
A
ij
=
T
ij
+
T
ij
+ T
A
ij
In the Navier-Stokes equations we have the tensor
u
i
x
j
(deformation-rate tensor). The anti-symmetric part
describes rotation, the isotropic part describes the volume change and the trace-less part describes the defor-
mation of a uid element.
Operators
(p)
i
=
x
i
p (gradient, increase of tensor order)
p = p =
2
p =
2
x
i
x
i
p (Laplace operator)
u =
x
i
u
i
(divergence, decrease of tensor order)
( A)
j
=
x
i
A
ij
(divergence of a tensor)
( u)
ij
= ( u)
ij
=
x
i
u
j
(gradient of a vector)
( u)
i
=
ijk
x
j
u
k
(curl)
Gauss theorem (general)
Gauss theorem (divergence theorem):
F ndS =
F dV
or with index notation,
S
F
i
n
i
dS =
V
F
i
x
i
dV
In general we can write,
S
T
ijk
n
l
dS =
x
l
T
ijk
dV
Example: Put T
ijk
n
l
= T
ij
u
l
n
l
S
T
ij
u
l
n
l
dS =
x
l
(u
l
T
ij
) dV
or,
S
T( u n) dS =
( u)T+ ( u )T
dV
Identities
Derive the identity
(
F
G) = (
G )
F + (
G)
F (
F)
G(
F )
G
(
F
G)
i
=
ijk
x
j
klm
F
l
G
m
=
ijk
klm
x
j
F
l
G
m
=
kij
klm
x
j
F
l
G
m
=
3
(
il
jm
im
jl
)
x
j
F
l
G
m
=
x
j
F
i
G
j
x
j
F
j
G
i
=
F
i
x
j
G
j
+
G
j
x
j
F
i
F
j
x
j
G
i
+
G
i
x
j
F
j
=
G
j
F
i
x
j
+
G
j
x
j
F
i
F
j
x
j
G
i
F
j
G
i
x
j
= [(
G )
F + (
G)
F (
F)
G(
F )
G]
i
Show that
(
F) = 0
(
F) =
x
i
ijk
x
j
F
k
=
ijk
x
i
x
j
F
k
Remember that
ijk
=
jik
and that
x
i
x
j
F
k
=
x
j
x
i
F
k
and thus all terms will cancel.
ijk
x
i
x
j
F
k
= 0
Similarly, show that
(f) = 0
(f) =
ijk
2
x
j
x
k
f = 0 according to the same argument as above.
4