EN0175 09 / 19 / 06

Today we introduce the application of index notations in vector and tensor algebra.

Index notation:

⎡1 0 0⎤
⎧1, i = j ⎢ ⎥
e.g. Kronecker delta: δ ij = ⎨ . In matrix form: 0 1 0
⎢ ⎥
⎩0, i ≠ j ⎢⎣0 0 1⎥⎦

v v v
( v
) (v v )
Dot product of vectors: a ⋅ b = (ai ei ) ⋅ b j e j = ai b j ei ⋅ e j = ai b jδ ij = ai bi

v v v
Cross product of vectors: c = a × b ,

v
The magnitude of cross product is: c = ab sin θ

The direction of cross product is:

v v
c b
vθ
a
v v v
For base vectors e1 , e2 , e3 :

v v v v v v v v
e1 × e1 = 0 , e1 × e2 = e3 , e1 × e3 = −e2
v v v v v v v v
e2 × e1 = −e3 , e2 × e2 = 0 , e2 × e3 = e1
v v v v v v v v
e3 × e1 = e2 , e3 × e2 = −e1 , e3 × e3 = 0

Or we can write the nine equations in a concise form as:

v v v
ei × e j = ε ijk ek

⎧1, ijk = 123, 231, 312

⎪
where ε ijk = ⎨− 1, ijk = 213, 132, 321 (called permutation symbol)
⎪0, otherwise
⎩

1
EN0175 09 / 19 / 06

2 3
Vector operations are useful in representing areas and volumes in a solid body.

Area of a parallelogram:

v v v v
n ( n ⊥ to the plane of a and b )
v v v
b A = ab sin θ = a × b
θ v r v v v
a A = An = a × b

v v v
v dA = dx ×d y
dy

v
dx

Volume of a parallelepiped:

v v v v v v v
(
V = A(c ⋅ n ) = c ⋅ A = c ⋅ a × b )
v v
c b
v
a

v v v v v v v v v
dV = d x3 ⋅ (d x1 × d x2 ) = d x1 ⋅ (d x2 × d x3 ) = d x2 ⋅ (d x3 × d x1 )

v
d z d yv In index notation, we can express volume of a parallelepiped
as
v
dx

2
EN0175 09 / 19 / 06

v v v
( v
)
V = c ⋅ a × b = ci ei ⋅ (a j e j × bk e k ) = c i a j bk ei ⋅ (e j × e k ) = ε ijk ci a j bk = ε ijk a i b j c k ,
v v v v v

We could also write this in matrix notation as

a1 a2 a3
V = b1 b2 b3 = a1 (b2 c3 − b3 c 2 ) − a 2 (b1c3 − b3 c1 ) + a3 (b1c 2 − b2 c1 ) = ε ijk ai b j c k
c1 c2 c3

⎡ A11 A12 A13 ⎤

⎢
For an arbitrary matrix A = A21
⎢ A22 A23 ⎥⎥ , its determinant is det ( A) = ε ijk A1i A2 j A3k .
⎢⎣ A31 A32 A33 ⎥⎦
More generally,

ε ijk Api Aqj Ark = det ( A)ε pqr

ε − δ relation:

⎡δ ip δ iq δ ir ⎤
⎢ ⎥
ε ijk ε pqr = ⎢δ jp δ jq δ jr ⎥
⎢δ kp δ kq δ kr ⎥⎦
⎣
Proof:

For any matrix,

A11 A12 A13
det ( A) = A21 A22 A23
A31 A32 A33

For any change in rows, the calculation rule of a determinant obeys

Ai1 Ai 2 Ai 3
ε ijk det ( A) = A j1 Aj 2 Aj 3
Ak1 Ak 2 Ak 3
Similarly, for any change in column,
Aip Aiq Air
ε ijk ε pqr det ( A) = A jp A jq A jr
Akp Akq Akr

Set Aij = δ ij ( A = δ ), det(I ) = 1 ,

3
EN0175 09 / 19 / 06

⎡δ ip δ iq δ ir ⎤
⎢ ⎥
ε ijk ε pqr = ⎢δ jp δ jq δ jr ⎥
⎢δ kp δ kq δ kr ⎥⎦
⎣
In particular,
⎡δ ii δ iq δ ir ⎤
⎢ ⎥
ε ijk ε iqr = ⎢δ ji δ jq δ jr ⎥ = δ jqδ kr − δ jrδ kq
⎢δ ki δ kq δ kr ⎥⎦
⎣
The above are called ε − δ relations.

By using the ε − δ relationship, we can calculate

v v v v v v v
ei × e j × ek = ε ijq eq × ek = ε ijqε qkr er = ε qij ε qkr er
= (δ ik δ jr − δ ir δ jk )er
v
v v
= δ ik e j − δ jk ei

v v v
a × b × c = ai b j ck ei × e j × ek = ai b j ck (δ ik e j − δ jk ei )
v v v v v
v v
= ai ci b j e j − b j c j ai ei
v v v v v v
= (a ⋅ c )b − b ⋅ c a ( )

Gradient of a scalar field T ( x, y, z )

v ∂ v ∂ v ∂
∇ = e1 + e2 + e3
∂x1 ∂x2 ∂x3

v ∂T v ∂T v ∂T v ∂T v
∇T = e1 + e2 + e3 = ei = eiT , i
∂x1 ∂x2 ∂x3 ∂xi

v
Divergence of a vector field u ( x, y, z ) :

v v ∂ v ∂u j ∂u
∇ ⋅ u = ei ⋅ u je j = δ ij = i
∂xi ∂xi ∂xi
∂u1 ∂u 2 ∂u3
= + + = u i, i
∂x1 ∂x2 ∂x3
e.g.

4
EN0175 09 / 19 / 06

v v
∂ 2u
∇ ⋅σ + f = ρ 2
∂t
∇ ⋅ σ = σ ij , j
σ ij , j + f i = ρ u&&i

v
Curl of a vector field u ( x, y, z ) :

v v ∂ v v
∇ × u = ei × u j e j = u j , iε ijk ek
∂xi

Tensors:

(Second rank) tensors are linear transformations that map one vector into another vector. Higher
order tensors can map one tensor into another tensor. One can also say that vectors are first rank
tensors and scalar are zeroth rank tensors.

For example,
v
n
v
t

v v
If n is the normal vector of a tiny area on the surface, t is surface traction, we will show that
v v
there exists a tensor which maps n to t ,

v v
σ n=t

where σ will turn out to be the stress tensor (discussed later).

An example of higher order tensor is the elastic modulus tensor. If σ is stress tensor and ε is
strain tensor, then we need a 4th order tensor to relate them to each other:

σ = Cε

where C is called the modulus tensor (to be discussed later).

5
EN0175 09 / 19 / 06

This relationship, called the generalized Hooke’s law, is the 3D generalization of the 1D Hooke’s
law,

σ = Eε .

Base tensors:

v
e3
⎡1⎤ ⎡0 ⎤ ⎡0 ⎤
v ⎢ ⎥ v ⎢ ⎥ v ⎢ ⎥
e1 = ⎢0⎥ , e2 = ⎢1⎥ , e3 = ⎢0⎥
v
e2 ⎣⎢0⎦⎥ ⎣⎢0⎦⎥ ⎣⎢1⎦⎥

v
e1

⎡1 0 0⎤ ⎡0 1 0 ⎤ ⎡0 0 1 ⎤
e1 ⊗ e1 = ⎢0 0 0⎥ , e1 ⊗ e2 = ⎢0 0 0⎥ , e1 ⊗ e3 = ⎢0 0 0⎥⎥ …
⎥ ⎥
v v ⎢ v v ⎢ v v ⎢

⎢⎣0 0 0⎥⎦ ⎢⎣0 0 0⎥⎦ ⎢⎣0 0 0⎥⎦

where “ ⊗ ” is the tensor product (or dyad product) symbol.

⎡0 ⎤ ith row
⎢ ⎥
v v ⎢ 1 ⎥ jth column
ei ⊗ e j =
⎢ ⎥
⎢ ⎥
⎣ 0⎦