Appendix A

Index Notation

sec:A.0
In this book we use the standard notation that a bold variable denotes a vector. Some examples are the fluid
velocity u(x, y) or the particle velocity V(t). The fluid velocity is a vector field, i.e., it is dependent on the
space variable x, whereas the Lagrangian particle velocity is a just a time-dependent velocity vector. In index
notation they will be denoted as ui (x, t) and Vi , where the index goes i = 1, 2, and 3.
In this appendix, we will very briefly go over some of the index notation definitions and identities that will be
used in Chapter 4 and in the rest of the book. We caution that the treatment is neither extensive nor rigorous.
We are assuming that you already know all much of this material from your graduate level fluid mechanics
course. This appendix is only to serve as a quick refresher. If you are not familiar, you must take an applied
mathematics course or read a book on mathematics for physicists and engineers.
There are some simple standard rules of index notation that we will follow: (i) any index that appears only
once is called a free index, (ii) if an index repeats twice in a term then it implies sum over 1, 2 and 3 of that
repeated index, and (iii) if an index repeats more than twice in a term then you must have made an error.
According to the rule, (i) if there is no free index in a term, then that term is a scalar. An example will be
pressure, which will be denoted by the variable p, where it can be noted that it comes with no index of any
sort. (ii) If there is one free index, then the term is a vector. Examples are position x and velocities u and V,
which in index notation become xi , ui , and Vi . Here ui indicates the vector (u1 , u2 , u3 )T , which are the three
components of velocity. (iii) If there are two free indices, then the quantity is a second-rank tensor. An example
is stress tensor σ, which in index notation is written as σij , where i and j are both free indices. As i and j
take values 1, 2, and 3, we can see that the second rank tensor has 9 components (just as a vector has three
components), which can be written in a 3 × 3 matrix form as
 
σ11 σ12 σ13
σ = σij =  σ21 σ22 σ23  . (A.1)
σ31 σ32 σ33
For physical quantities, the indices have a definite meaning. For example, in the case of σij the first index
gives the direction of force per area (i.e., the direction of stress component) acting on a surface whose normal
is given by the second index. (iv) If there are three free indices, then the quantity is a third-rank tensor and
has 33 = 27 components. This list goes on. Note that scalar is a zeroth rank tensor and vector is a first rank
tensor.
We now introduce two important constant tensors. The first is Kronecker delta written as
(
1 if i = j
δij = (A.2)
0 if i 6= j .
The Kronecker delta is a constant second rank tensor and when written as a 3 × 3 matrix, it is an identity
matrix. The second useful tensor is Levy-Chevita Symbol which can be expressed as

 1
 if (i, j, k) = (1, 2, 3), (2, 3, 1), or(3, 1, 2)
ij = −1 if (i, j, k) = (1, 3, 2), (2, 1, 3), or(3, 2, 1) (A.3)

0 otherwise .


Levi-Chevita symbol is a constant third rank tensor, since it has three free indices i, j, and k. Both Kronerker
delta and Levi-Chevita symbol are handy tools in index notation, which we shall see soon.
The scalar, vectors, second and higher-order tensors obey standard mathematical operations such as addi-
tions, subtraction, multiplication, etc., in specific ways, which we shall go over quickly. You can on add or

161
162 Index Notation
subtract a scalar with a scalar, vector with a vector, tensor with a tensor, etc, and the sum or difference will
in turn result in a scalar, vector, tensor, etc. In physical terms, you can write velocity as a sum of two other
velocities - all three are vectors and all are of dimension length/time. Certainly, it does not make sense to add
pressure with velocity
As far as multiplication, we define three kinds of multiplication: (i) outer product, (ii) inner product, and
(iii) cross product. Let us apply these multiplications between two vectors a and b. In index notation, their
definitions are given below
Outer product : ab = ai bj ,
Inner product : a · b = ai bi , (A.4)
Cross product : a × b = ijk aj bk ,
From the above definitions, we see that the outer product of two vectors results in a second-rank tensor,
since there are two free indices at the end. The inner product of two vectors results in a scalar, since the
repeated index i indicates a sum over i = 1, 2, 3. The inner product of two vectors can thus be expanded as
a1 b1 + a2 b2 + a3 b3 . The cross product of two vectors results in another vector. This can be seen in the index
notation, where i is the only free (unmatched) index, while j and k are repeated. In an expanded form, the ith
component of the cross product can be written as
(a × b)i = i11 a1 b1 + i12 a1 b2 + i13 a1 b3
i21 a2 b1 + i22 a2 b2 + i23 a2 b3 (A.5)
i31 a3 b1 + i32 a3 b2 + i33 a3 b3 for i = 1, 2, 3 .
The above definitions of outer, inner and cross product can be generalized when a is a nth rank tensor and b
is a mth rank tensor (i.e, a has n free indices and b has m free indices). Their outer product will be a more
complex tensor of rank n + m. Whereas, the rank of their inner product will be smaller. If there is only one
free index that is common between a and b then the rank of a · b is n + m − 2. In the case of cross product,
one free index of a combines with one free index of b to yield a free index in the cross product and thus the
rank of a × b is n + m − 1.

prob:A.1.1
Problem A.1 With the above definitions, prove the following important properties of index notation.
• ai for i = 1, 2, 3 is equal to aj for j = 1, 2, 3.
• ai bj = ak bl for i, j, k, l = 1, 2, 3.
• ai bi = aj bj .
The first important property is that the indices are dummy variables. You can replace one dummy index with
another, since they all run 1, 2, 3. In case of a repeated index, if you decide to replace one of them, you must
do so for both. After replacement there should still be a repeated index.

prob:A.1.2
Problem A.2 With the above definitions, prove the following identities of Kronecker delta and Levi-Chevita
symbol. All these are very useful identities that will be repeatedly needed over the rest of the book. It is not
enough for you to prove these identities. You must remember them and start using them with ease as and when
needed .
• δii = 3.
• δij = δji for all i, j. I.e., the order of the two indices does not matter in Kronecker delta. This also means
Kronecker delta is symmetric.
• δij δil = δjl .
• Ai δij = Aj . This shows that Kronecker delta has the property of switching the free index of a tensor, through
the use of a repeated index.
• ijk = kij = jki for all i, j, k. This important property shows that the indices of Levi-Chevita symbol can
be cyclically varied.
• ijk = −ikj = −jik = −kji for all i, j, k. This important property shows that the indices of Levi-Chevita
symbol can be acyclically varied, but with a negative sign.
 Index Notation 163

• ijk δjk = 0 for all i.

Before leaving this appendix we point out that we will deal with tensor quantities that are products of
other tensor quantities. For example, the outer product of fluid velocity vector with itself is u u, which is also
sometimes written as u ⊗ u. It is a second rank tensor and in index notation will be denoted as ui uj . Similarly
xi xj xk will be a third rank tensor. On the other hand, ui uj xj is a vector, since a sum over j is implied and i
is the only remaining free index.
 Appendix B
Vector Calculus

sec:A.01
In this appendix we will very briefly go over some of the vector calculus identities that will use in Chapter
4 and in the rest of the book. We again caution that the treatment is neither extensive nor rigorous. This
appendix should only be used as a refresher.
In vector calculus, the important quantity is the gradient operator ∇, which is a vector and its three com-
ponents are the derivatives along three orthogonal directions. In Cartesian coordinates, the three components
are (∂/∂x, ∂/∂y, ∂/∂z)T . Thus, in index notation we have
∂
∇= for i = 1, 2, 3 , (B.1)
∂xi
where x1 , x2 , and x3 are also referred to as the x, y, and z coordinates. The gradient operator behaves as any
other vector and can be used in outer, inner and cross products. However, since it is a derivative operator, it
needs to operate on another quantity. I.e., it needs to be followed by another fluid mechanical quantity to the
right of it and this quantity must be a field variable (i.e., it must be a function of x) in order for the gradient
to be non-zero. For example, ∇ operating on a constant scalar, vector, or tensor (that is not a function of x)
is identically zero.
Let us now consider the three different products of ∇ and the velocity vector field u(x, t). Since the field u
is a function of space and time its gradient is in general non-zero. The three products are
∂uj
Outer product : ∇u = ,
∂xi
∂u1 ∂u2 ∂u3
Inner product : ∇·u = + + (B.2)
∂x1 ∂x2 ∂x3
∂uk
Cross product : ∇ × u = ijk ,
∂xj
where ∇u is known as gradient of u, ∇ · u is known as divergence of u, and ∇ × u is known as curl of u, which
yields the vorticity field. All these definitions involve x1 , x2 , and x3 derivatives of the three components of
velocity. All the rules of index notation discussed in the previous appendix apply to the gradient operator as
well. Note that ∇ can operate repeatedly. For example, ∇∇u corresponds to gradient operator outer product
with gradient operator outer product with velocity. Due to the two outer products, the result will be a third
rank tensor, which in index notation can be written as ∂ 2 uk /(∂xi ∂xj ). As i, j, and k run over 1, 2, 3, ∇∇u
can be seen to have 27 components. Of particular importance is the Laplace operator, which is defined as
∂2 ∂2 ∂2 ∂2
∇2 = ∇ · ∇ = = + + , (B.3)
∂xi ∂xi ∂x21 ∂x22 ∂x23
where the last relation comes from summing over the repeated index i.

prob:A.2.1
Problem B.1 Prove the following identities where p is a scalar field, A, B, and u are vector fields.
• ∇x is equal to identity matrix or in index notation ∂xj /∂xi = δij .
• ∇ × (∇p) ≡ 0.
• ∇ · (∇p) = ∇2 p .
• ∇ · (∇ × A) = 0.
• ∇ × (A × B) = A(∇ · B) + (B · ∇)A − B(∇ · A) − (A · ∇)B.

164
 Vector Calculus 165

• u · ∇u = (∇ × u) × u + 12 ∇(u · u).
• ∇2 u = ∇(∇ · u) − ∇ × (∇ × u)

We now consider some useful results that will be of great help in following the derivations presented in
Chapter 4. We start with the definition that r is the length of the position vector x. I.e., r = (x · x)1/2 . As an
example, we derive the following relations that are used in obtaining the growing and decaying solutions given
in Section 4.1.1:
∂r ∂(xj xj )1/2 1 ∂xj 1 xi
= = 1/2
2xj = xj δij =
∂xi ∂xi 2(xj xj ) ∂xi r r eq:A.B.4
k
(B.4)
∂r k−1 ∂r k−2
= kr = kr xi .
∂xi ∂xi

eaxm:A.2.1
Example B.2 We present here as an example some manipulations that you will need to obtain the relations
presented in Chapter 4. In this example, we start with the following results for pressure and particular velocity
given in (4.11) and (4.14)
xj a1 xi xj
p(x) = a1 ucj 3 and u = ucj 3 , (B.5)
r 2µf r
where we note a1 and µf are scalar constants and ucj is a vector constant. Our goal is to shown that the above
pressure and velocity satisfies the Stokes equation 0 = −∇p + µf ∇2 u. To do that we first obtain pressure
gradient as
 
∂  xj  δij 3xi xj
−∇p = − a1 ucj 3 = −a1 ucj − , (B.6)
∂xi r r3 r5
where we have the used the result obtained in (B.4). Towards obtaining the Laplacian of velocity, let us first
calculate the gradient of velocity as
   
∂ui ∂ a1 xi xj a1 xj xi 3xi xj xk
= ucj 3 = ucj δik 3 + δjk 3 − (B.7)
∂xk ∂xk 2µf r 2µf r r r5
Follow this procedure and take another derivative of the above expression to obtain
 
2 δij 3xi xj
µf ∇ u = a1 ucj − . (B.8)
r3 r5
This combined with the expression for the pressure gradient shows that that satisfy the Stokes equation.

The final section of the appendix is to recollect the Gauss theorem that you have learnt in vector calculus.
The Gauss theorem is written as Z Z
∇ · u dΩ = u · n dS , (B.9)
Ω S
where on the left hand side is the volume integral over the volume Ω and on the right hand side is the integral
over the surface S of that volume. Here u is any vector field and n is the outward pointing surface normal
vector. It is very important to refresh your understanding of this theorem.

prob:A.2.2
Problem B.3 In this problem you will derive two other variants of the Gauss theorem. Let u = aφ(x), where
a is a constant vector independent of x, whereas the scalar field φ is a function of space. Derive the following
generalization of the Gauss theorem
Z Z
∇φ dΩ = φn dS . (B.10)
Ω S
Instead let u = a × w(x), where a is a constant vector and w is a vector field. Derive the following
generalization of the Gauss theorem
Z Z
∇ × w dΩ = n × w dS . (B.11)
Ω S
166 Vector Calculus