VECTORS, TENSORS

AND INDEX NOTATION

Enrico Nobile

Dipartimento di Ingegneria e Architettura

Università degli Studi di Trieste, 34127 TRIESTE

March 12, 2021

Vectors & Tensors, E. Nobile March 12, 2021 1 / 36

OUTLINE

1 Vectors
The scalar product
The vector product
The scalar triple product
The vector triple product
The Nabla ∇ operator

2 Index notation
Summation convention and dummy index
Free index
The Kronecker delta
Rules of index notation
The Permutation Symbol

Vectors & Tensors, E. Nobile March 12, 2021 2 / 36

Review of vector algebra

The scalar product

The scalar product of two vectors a and b (dot
product) is usually written a · b, and is defined
as the scalar quantity:

a · b = |a| |b| cos θ

Some properties of the scalar product

a · (b + c) = a · b + a · c
a · a = |a|2
a · b = (ax i + ay j + az k) · (bx i + by j + bz k)
= ax bx + ay by + az bz

Vectors & Tensors, E. Nobile March 12, 2021 4 / 36

Review of vector algebra - cont.

The vector product

The vector product of two vectors a and b

(cross product) is usually written a × b (or
a ∧ b), and is defined as the vector quantity

a × b = |a| |b| sin θ n

Some properties of the vector product

b × a = −a × b
i j k
a × (b + c) = a × b + a × c
a×b = ax ay az
a × b = (ax i + ay j + az k) × (bx i + by j + bz k) bx by bz
= (ay bz − az by ) i + (az bx − ax bz ) j + (ax by − ay bx ) k

Vectors & Tensors, E. Nobile March 12, 2021 5 / 36

Review of vector algebra - cont.

Vector area A

A = a × b = |a| |b| sin θ n

Volume of a parallelepiped

V = c·(a × b) = a·(b × c) = b·(c × a)

Vectors & Tensors, E. Nobile March 12, 2021 6 / 36

Scalar triple product

We already saw that

a · (b × c) = b · (c × a) = c · (a × b)

A useful way of expressing the scalar triple product is in determinant form. Since

i j k
a×b= ax ay az
bx by bz

it follows that

Scalar triple product

ax ay az
a · (b × c) = bx by bz
cx cy cz

Vectors & Tensors, E. Nobile March 12, 2021 7 / 36

Vector triple product

Geometric interpretation

From the properties of the cross-product, it

follows that the vector a × (b × c) is
perpendicular both to a and to (b × c).
Since (b × c) is perpendicular to both b and
c, it follows that a × (b × c) must lie in the
plane containing both b and c (see figure).

Therefore we may write

a × (b × c) = αb + βc
where α and β are yet to be determined. This can be done evaluating the vector-product on the
left-hand side, and use the right-hand side to determine α and β.

Vectors & Tensors, E. Nobile March 12, 2021 8 / 36

Vector triple product - cont.

If a = ax i + ay j + az k, and analogously for b and c, then

b × c = i (by cz − bz cy ) − j (bx cz − cx bz ) + k (bx cy − cx by )

Using the determinant form of the vector product

i j k
a × (b × c) = ax ay az
by cz − bz cy cx bz − bx cz bx c y − c x by
= i [ay (bx cy − cx by ) − az (cx bz − bx cz )]
− j [ax (bx cy − cx by ) − az (by cz − bz cy )]
+ k [ax (cx bz − bx cz ) − ay (by cz − bz cy )]

Regrouping according to the form αb + βc

a × (b × c) = ibx (ay cy + az cz ) + jby (ax cx + az cz ) + kbz (ax cx + ay cy )

− icx (ay by + az bz ) − jcy (ax bx + az bz ) − kcz (ax bx + ay by )

Vectors & Tensors, E. Nobile March 12, 2021 9 / 36

Vector triple product - cont.

Adding and subtracting

i (bx ax cx ) + j (by ay cy ) + k (bz az cz )
to the previous expression finally gives

a × (b × c) = (a · c) b − (a · b) c (1)

and therefore

α=a·c
β =a·b

Vectors & Tensors, E. Nobile March 12, 2021 10 / 36

Gradient of a scalar

A vector operator, which is frequently used in fluid dynamics and heat transfer, is the nabla, (or
del) operator, defined as
∂ ∂ ∂
∇= i+ j+ k
∂x ∂y ∂z
When the nabla operator is applied on a scalar variable φ, it results in the gradient of φ, given
by:
∂φ ∂φ ∂φ
∇φ = i+ j+ k
∂x ∂y ∂z
Therefore, the gradient of a scalar field is a vector field and it indicates that the value of φ
changes in space in both magnitude and direction.

Vectors & Tensors, E. Nobile March 12, 2021 11 / 36

Gradient of a scalar - cont.

The projection of ∇φ in the direction of the unit vector el is given by

dφ
= ∇φ · el = |∇φ| cos (∇φ, el )
dl
and is called directional derivative of φ along the direction of the unit vector el .
The maximum value of the directional derivative is |∇φ| and is achieved when
cos(∇φ, el ) = 1, e.g. in the direction of ∇φ.
Therefore, the gradient of a scalar field φ indicates the direction and magnitude of the
largest variation of φ at every point in space.
Finally, ∇φ is normal to the φ = const. surface that passes through that point.

Vectors & Tensors, E. Nobile March 12, 2021 12 / 36

Operations on the ∇ operator

The scalar product of the ∇ operator with a vector u of components u, v and w in the x, y
and z direction, respectively, is the Divergence of the vector u, which is a scalar quantity
written as
∂u ∂u ∂u
∇·u= + +
∂x ∂y ∂z
The divergence of the gradient of a scalar variable φ is denoted by the Laplacian operator
and is a scalar given by

∂2φ
2 ∂2φ ∂2φ
∇ · (∇φ) = ∇ φ = + +
∂x2 ∂y 2 ∂z 2
The Laplacian of a vector follows from the above definition and is a vector given by

∇2 u = ∇2 u i + ∇2 v j + ∇2 w k

Vectors & Tensors, E. Nobile March 12, 2021 13 / 36

Additional vector operations

If φ is a scalar field, and v, v1 and v2 are vector fields, then the following relations apply:

∇ · (∇ × v) = 0
∇ × (∇φ) = 0
∇ · (φv) = φ∇ · v + v · ∇φ
∇ × (φv) = φ∇ × v + ∇φ × v
∇ (v1 · v2 ) = v1 × (∇ × v2 ) + v2 × (∇ × v1 ) + (v1 · ∇) v2 + (v2 · ∇) v1
∇ · (v1 × v2 ) = v2 · (∇ × v1 ) − v1 · (∇ × v2 )
∇ × (v1 × v2 ) = v1 (∇ · v2 ) − v2 (∇ · v1 ) + (v2 · ∇) v1 − (v1 · ∇) v2
∇ × (∇ × v) = ∇ (∇ · v) − ∇2 v
(∇ × v) × v = v · (∇v) − ∇ (v · v)

Vectors & Tensors, E. Nobile March 12, 2021 14 / 36

Additional vector operations - cont.

Let’s prove the second, e.g.

∇ × (∇φ) = 0
It can be expressed as
 
∂φ ∂φ ∂φ
∇ × (∇φ) = ∇ × i +j +k
∂x ∂y ∂z
i j k
∂ ∂ ∂
= ∂x ∂y ∂z
∂φ ∂φ ∂φ
∂x ∂y ∂z
     
∂ ∂φ
 ∂ ∂φ Z∂ ∂φ Z∂ ∂φ ∂ ∂φ ∂ ∂φ
 −  −j Z − ∂zZ  − 

=i +k
∂y ∂z  ∂z ∂y ∂x ∂z
Z ∂x
Z ∂x ∂y 
∂y ∂x
=0

Vectors & Tensors, E. Nobile March 12, 2021 15 / 36

Operations on the ∇ operator - cont.

Another quantity of interest is the curl of a vector field u, which is a vector obtained by the
cross product of the ∇ operator and vector u
 
∂ ∂ ∂
∇×u= i+ j+ k × (ui + vj + wk)
∂x ∂y ∂z
i j k
∂ ∂ ∂
=
∂x ∂y ∂z
u v z
     
∂w ∂v ∂u ∂w ∂v ∂u
= − i+ − j+ − k
∂y ∂z ∂z ∂x ∂x ∂y

Vectors & Tensors, E. Nobile March 12, 2021 16 / 36

The dyadic product

A dyad is a tensor of order two and rank one, and is the result of the dyadic product (also
called outer product or tensor product) of two vectors.
If a and b are two vectors, then the dyadic product is indicated as ab, a ⊗ b or abT ,
where T means transpose.
The dyadic product of two vectors a and b is a Tensor and is given, in matrix form, by
   
 a1  a1 b1 a1 b2 a1 b3
ab ≡ a ⊗ b ≡ abT = a2 [b1 b2 b3 ] =  a2 b1 a2 b2 a2 b3 
a3 a3 b1 a3 b2 a3 b3
 

Vectors & Tensors, E. Nobile March 12, 2021 17 / 36

The vector product of a vector and a tensor

The Vector Product (or Dot Product) of a vector a and a tensor T is a vector and is given by
  T
 T T11 T12 T13
a · T ≡ aT T = [a1 a2 a3 ]  T21 T22 T23 
T31 T32 T33
= [a1 T11 + a2 T21 + a3 T31 a1 T12 + a2 T22 + a3 T32 a1 T13 + a2 T23 + a3 T33 ]T
 
 a1 T11 + a2 T21 + a3 T31 
= a1 T12 + a2 T22 + a3 T32
a1 T13 + a2 T23 + a3 T33
 

Vectors & Tensors, E. Nobile March 12, 2021 18 / 36

A common mixed product

An important mixed product, which is frequently found in fluid mechanics, is the following
 
∂ ∂ ∂
[(u · ∇) u] = (ui + vj + wk) · i+ j+ k (ui + vj + wk)
∂x ∂y ∂z
 
∂ ∂ ∂
= u +v +w (ui + vj + wk)
∂x ∂y ∂z
     
∂u ∂u ∂u ∂v ∂v ∂v ∂w ∂w ∂w
= u +v +w i+ u +v +w j+ u +v +w k
∂x ∂y ∂z ∂x ∂y ∂z ∂x ∂y ∂z
 
∂u ∂u ∂u 

 u +v +w 


 ∂x ∂y ∂z 


 ∂v ∂v ∂v 
= u +v +w
 ∂x ∂y ∂z 
∂w ∂w ∂w 

 
 u +v +w

 


∂x ∂y ∂z

Vectors & Tensors, E. Nobile March 12, 2021 19 / 36

A common mixed product - cont.

It is easy to see that the same result can be obtained considering the vector (dot) product of the
vector u with the tensor obtained with the dyadic product of the divergence with the vector u:
  ∂u ∂v ∂w T
  ∂x ∂x ∂x 
  

T
T   ∂u ∂v ∂w 
u · (∇u) ≡ u (∇u) = [u v w] 
 ∂y ∂y ∂y


  
∂u ∂v ∂w
  
∂z ∂z ∂z
 T
∂u ∂u ∂u ∂v ∂v ∂v ∂w ∂w ∂w
= u +v +w u +v +w u +v +w
∂x ∂y ∂z ∂x ∂y ∂z ∂x ∂y ∂z
 
∂u ∂u ∂u 

 u +v +w 


 ∂x ∂y ∂z 


 ∂v ∂v ∂v 
= u +v +w
 ∂x ∂y ∂z 
∂w ∂w ∂w 

 
 u +v +w

 


∂x ∂y ∂z

Vectors & Tensors, E. Nobile March 12, 2021 20 / 36

Index notation: overview

The notation adopted to describe vectors is perhaps the most widely accepted.
There are a number of alternative notations, and one of the most useful is the index
notations, also called Cartesian notation.
The index notation is a powerful tool for manipulating multidimensional equations, and it
enjoys a number of advantages in comparison to the traditional vector notation:
I Conciseness: it allows standard results to be written in an immediately clear though compact
form.
I Quick proof of the results that we need.
There are other advantages - and a couple of disadvantages - but in addition it is important
to note that the index approach is the most natural notation when dealing with a
generalization of the vector: the tensor.
We will find convenient to relabel the coordinates (x, y, z) as (x1 , x2 , x3 ) or simply as xi ,
with the agreement that the index i can take any of the values 1, 2 and 3.
Using this index notation, a vector a may be denoted by ai . Therefore the components of
the vector ai are (a1 , a2 , a3 ).
There are times when the more conventional vector notation is more useful. It is therefore
important to be able to easily convert back and forth between the two.

Vectors & Tensors, E. Nobile March 12, 2021 22 / 36

Equality and summation convention

Equality
Two vectors ai and bi are equal if and only if

ai = bi (2)

If the index is unspecified then the equation in which it occurs is assumed to be valid for each of
the three possible values that the index can take, e.g. (2) is shorthand for

a1 = b1 , a2 = b2 , a3 = b3 (3)

The summation convention

If an index is repeated in a term, then it is assumed that the term is summed as the repeated
index takes the values 1, 2 and 3. Thus
3
X
ai bi means ai bi ≡ a1 b1 + a2 b2 + a3 b3 (4)
i=1

Using this convention, if ai and bi are two vectors (representing a and b), then the scalar
product a · b is represented by ai bi .

Vectors & Tensors, E. Nobile March 12, 2021 23 / 36

Dummy index

Dummy index
A repeated index is called a dummy index, since the precise letter used is irrelevant: ai bi has the
same meaning as aj bj , ap bp , as bs etc., that is a1 b1 + a2 b2 + a3 b3 .

A dummy index appears twice within an additive term of an expression. In the equation below,
j and k are both dummy indices

ai = ijk bj ck + Dij ej

The dummy index is local to an individual additive term. It may be renamed in one term - so
long as the renaming doesn’t conflict with other indices - and it does not need to be renamed in
other terms (and, in fact, may not necessarily even be present in other terms).

Vectors & Tensors, E. Nobile March 12, 2021 24 / 36

The summation convention: examples

3
X
λ = ai bi ≡ λ = ai bi ≡ λ = a1 b1 + a2 b2 + a3 b3
i=1

3
X  c1 = S11 x1 + S12 x2 + S13 x3
ci = Sik xk ≡ ci = Sik xk ≡ c2 = S21 x1 + S22 x2 + S23 x3
c3 = S31 x1 + S32 x2 + S33 x3

k=1

3 X
X 3
τ = Sij Sij ≡ τ = Sij Sij ≡ τ = S11 S11 + S12 S12 + . . . + S32 S32 + S33 S33
i=1 j=1
3
X
Cij = Aik Bkj ≡ Cij = Aik Bkj ≡ [C] = [A] [B]
k=1
3
X
Cij = Aki Bkj ≡ Cij = Aki Bkj ≡ [C] = [A]T [B]
k=1

Vectors & Tensors, E. Nobile March 12, 2021 25 / 36

Free index

Free index
An index that is not repeated (in a term) is called a free index.

In the equation below, i is a free index:

ai = ijk bj ck + Dij ej

The same letter must be used for the free index in every additive term. The free index may be
renamed if and only if it is renamed in every term.
Terms in an expression may have more than one free index so long as the indices are distinct.
For example the vector-notation expression

A = BT

is written
Aij = (Bij )T = Bji
in index notation.
This expression implies nine distinct equations, since i and j are both free indices.

Vectors & Tensors, E. Nobile March 12, 2021 26 / 36

Free index - cont.
The number of free indices in a term equals the rank of the term:

Notation Rank
scalar a 0
vector ai 1
tensor Aij 2
tensor Aijk 3

Technically, a scalar is a tensor with rank 0, and a vector is a tensor of rank 1. Tensors may
assume a rank of any integer greater than or equal to zero. You may only sum together terms
with equal rank.
The first free index in a term corresponds to the row, and the second corresponds to the column.
Thus, a vector (which has only one free index) is written as a column of three rows
 
 a1 
a = ai = a2
a3
 

and a rank-2 tensor is written as

 
A11 A12 A13
A = Aij =  A21 A22 A23 
A31 A32 A33

Vectors & Tensors, E. Nobile March 12, 2021 27 / 36

The Kronecker delta

We define a two-indexed quantity δij called the Kronecker delta by


1 if i = j
δij = (5)
0 otherwise

Thus δ11 = δ22 = δ33 = 1 and δ12 = δ21 = δ13 = δ31 = δ23 = δ32 = 0. Another name
for this quantity is the substitution operator, since it can be used to substitute one index for
another. This follows from the easiliy verified result

aj = δji ai (6)

The right-hand side is δj1 a1 + δj2 a2 + δj3 a3 , so if j = 1 in (6)

a1 = δ11 a1 + δ12 a2 + δ13 a3 = a1 , with similar results for j = 2 and j = 3.

Like in ordinary algebra, an equation will be formed from a collection of terms added
together and equated to zero.
Each term will include one or more indexed objects multiplied together.

Vectors & Tensors, E. Nobile March 12, 2021 28 / 36

Rules
1 The same index (subscript) may not appear more than twice in a product of two (or more)
vectors or tensors. Thus
Aik xk , Aik Bkj , Aij Bik Cnk
are valid, but
Akk xk , Aik Bkk , Aij Bik Cik
are meaningless.
2 The number and type of free indices in any term of a given equation must be the same.
Thus

xi = ui + ci ≡ x = u + c
ai = Aki Bkj xj + Cik uk ≡ a = AT Bx + Cu

are valid, but

xi = Aij
xj = Aik uk
xi = Aik uk + cj

are meaningless.

Vectors & Tensors, E. Nobile March 12, 2021 29 / 36

Rules - cont.

3 Free and dummy indices may be changed without altering the meaning of an expression,
provided that rules 1 and 2 are not violated. Thus

xi = Aik xk ⇔ xj = Ajk xk ⇔ xj = Aji xi

Vectors & Tensors, E. Nobile March 12, 2021 30 / 36

Examples

Example 1
If ai and bi correspond to vectors a and b, the index form of the vector equation

a + (a · b)b = 0

is the following
ai + aj bj bi = 0

Example 2
Verify that δij δij = 3
XX
δij δij = δij δij
i j

= δ11 δ11 + 
δ12
δ
12 + 
δ13
δ
13 + 
δ21
δ
21 + δ22 δ22 + 
δ32
δ
32 + 
δ31
δ
31 + 
δ32
δ
32 + δ33 δ33

δij δij = 3

Vectors & Tensors, E. Nobile March 12, 2021 31 / 36

Permutation Symbol

The Permutation Symbol or Levi-Civita symbol 1 , sometimes also called Alternating Unit
Tensor, is defined by

 1 if i, j, k is a cyclic permutation of 1, 2, 3
ijk = −1 if i, j, k is a non-cyclic permutation of 1, 2, 3 (7)
0 otherwise


The alternating unit tensor is positive when the indices assume any clockwise cyclical
progression, and it is negative when the indices assume any anticlockwise cyclic order, as
shown in the figure:

Since each subscript can only take on values 1, 2 or 3, it follows that ijk has 27 components.

1
Tullio Levi-Civita, 29 March 1873 Padua - 29 December 1941 Rome, was an Italian mathematician, most famous for
his work on absolute differential calculus (tensor calculus) and its applications to the theory of relativity.
Vectors & Tensors, E. Nobile March 12, 2021 32 / 36
Permutation Symbol - cont.

From the definition of ijk , it follows that

123 = 231 = 312 = +1 (8)

132 = 321 = 213 = −1 (9)

and all other ijk = 0, for example

223 = 233 = 311 = 0 (10)

These results imply the following identities

ijk = jki = kij (11)

That is, a cyclic permutation of the indices on the permutation symbol does not alter its value.
Also
ijk = −kji , ijk = −jik , ijk = −ikj (12)
that is, interchange of any two indices on the permutation symbol introduces a minus sign. This
corresponds to an anticyclic permutation of indices.

Vectors & Tensors, E. Nobile March 12, 2021 33 / 36

Permutation Symbol and vector product

With the introduction of the permutation symbol ijk it can be shown that the ith component of
the vector product of two vectors a and b (or ai and bi ) is

(a × b)i = ijk aj bk (13)

The right-hand side of (13) involves two repeated indices and therefore a double summation.
1 Firstly summation over the index j:

ijk aj bk = i1k a1 bk + i2k a2 bk + i3k a3 bk

2 Summing over the index k (and using the result in (10) that ijk is zero if it contains two
identical indices),

ijk aj bk = 
i11
a
1 b1 + i12 a1 b2 + i13 a1 b3

+ i21 a2 b1 + 
i22
a
2 b2 + i23 a2 b3

+ i31 a3 b1 + i32 a3 b2 + 
i33
a
3 b3


Vectors & Tensors, E. Nobile March 12, 2021 34 / 36

Permutation Symbol and vector product - cont.

Therefore if i takes the value 1:

1jk aj bk = 123 a2 b3 + 132 a3 b2 = a2 b3 − a3 b2

Similarly, if i = 2

2jk aj bk = 213 a1 b3 + 231 a3 b1 = a3 b1 − a1 b3

and if i = 3
3jk aj bk = 312 a1 b2 + 321 a2 b1 = a1 b2 − a2 b1
But these three terms - just replace 1, 2, 3 with x, y, z - are the components of a × b, and thus
(13) is verified.

Vectors & Tensors, E. Nobile March 12, 2021 35 / 36

The  − δ identity
There exists a basic identity (it is proved using the properties of determinants) connecting the 
symbol and the δ symbol. It is

ijk ipr = δjp δkr − δjr δkp (14)

It has numerous applications when developing vector identities. For example, it is possible to
re-prove, using the index approach, the identity (1), namely

a × (b × c) = (a · c) b − (a · b) c

Now the ith component of the left-hand side is, using (13) twice

ijk aj (b × c)k = ijk klm aj bl cm

After a cyclic permutation on the first  symbol on the right-hand side, it results

kij klm aj bl cm = (δil δjm − δim δjl ) aj bl cm

= aj bi cj − aj bj ci
= (a · c) bi − (a · b) ci
= {(a · c) b − (a · b) c}i

which proves the result.

Vectors & Tensors, E. Nobile March 12, 2021 36 / 36

Total differential of a function and gradient

Consider the total differential of a function of three variables, φ(x1 , x2 , x3 ). We have

∂φ ∂φ ∂φ
dφ = dx1 + dx2 + dx3 (15)
∂x1 ∂x2 ∂x3
In tensor notation, this is replaced by
∂φ
dφ = dxi (16)
∂xi
Equation (16) can be thought of as the dot product of the gradient of φ, namely ∇φ, and the
differential vector dx = ( dx1 , dx2 , dx3 ). Thus, the ith component of ∇φ, which we denote
as (∇φ)i , is given by

Gradient
∂φ
(∇φ)i = = φ,i (17)
∂xi

where a comma followed by an index is tensor notation for differentiation with respect to xi .

Vectors & Tensors, E. Nobile March 12, 2021 37 / 36

Divergence and curl

The divergence of a vector u is given by

Divergence
∂ui
∇·u= = ui,i (18)
∂xi

where again differentiation with respect to xi is denoted by , i.

The curl of a vector u is

Curl
∂uk
(∇ × u) = ijk = ijk uk,j (19)
∂xj

Vectors & Tensors, E. Nobile March 12, 2021 38 / 36