Classical Electrodynamics

PH 302
Instructor : Golam Dastegir Al-Quaderi
ii
Contents

1 Introduction and Mathematical Preliminaries 1

1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 The Gradient . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.3 The Divergence . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.4 The Curl . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.5 The Operator Nabla ~ . . . . . . . . . . . . . . . . . . . . . . . 12
1.6 The Tensorial Form of Operators . . . . . . . . . . . . . . . . . . 12
1.6.1 Tensorial Form of Operators . . . . . . . . . . . . . . . . 15
1.7 Identities of Vector Calculus . . . . . . . . . . . . . . . . . . . . . 15
1.8 Operators in Orthogonal Curvilinear Coordinate Systems . . . . 20
1.9 Dirac Delta Function . . . . . . . . . . . . . . . . . . . . . . . . . 27
1.9.1 The Fundamental Property of the Delta Function . . . . . 28
1.9.2 Representations of the Delta Function . . . . . . . . . . . 29
1.9.3 Properties of the Delta Function . . . . . . . . . . . . . . 34

iii
iv CONTENTS
Chapter 1

Introduction and
Mathematical Preliminaries

1.1 Introduction
Classical electrodynamics (CED) is one of the basic pillars of physics. The uni-
fication of electricity and magnetism by Maxwell in the nineteenth century can
be regarded as the first step toward the unification of all the fundamental forces
of nature, a goal which physicists are yet to achieve. The fundamental basis of
CED was completely established theoretically by Maxwell by 1864 and experi-
mentally by Hertz by 1888. CED is classical in the sense that it best describes
phenomena involving electric and magnetic fields and sources where quantum
effects are not important. Maxwells equations are one of the basic sets of equa-
tions in the classical field theory and can easily be taken into the quantum field
theory realm, without a change of the form/structure by the method of quanti-
zation, giving rise to Quantum Electrodynamics (QED), one of the best tested
theories of modern physics. In everyday life, CED has immense applications,
like use in electrical appliances, power distribution, radio communication, opti-
cal fibers, photocopiers, etc. A good understanding of CED is thus an integral
part of the training of a physicist, both theorist and experimentalist.
In this chapter, we basically review vector calculus, find out some of its
interesting results and introduce some mathematical tools necessary for the
study of CED.

1.2 The Gradient

Consider a displacement vector d~s = ds n in some particular direction described
by the unit vector n. Let the unit vectors along the x,y and the z-axis be denoted
by ex , ey and ez , respectively. Then

n = ex + ey + ez

1
2CHAPTER 1. INTRODUCTION AND MATHEMATICAL PRELIMINARIES

where , , are the direction cosines of the direction of n, i.e.

= ex .n, = ey .n, = ez .n
Thus,
p
d~s = dsn = ds(ex + ey + ez )/ 2 + 2 + 2
p
= ds (, , )/ 2 + 2 + 2
= ex dx + ey dy + ez dz
which implies
p p p
dx = ds / 2 + 2 + 2 , dy = ds / 2 + 2 + 2 , dz = ds / 2 + 2 + 2
Let, = (x, y, x) be a continuously differentiable function of (x, y, z). Then
the rate of change of in the direction of d~s is
d (x + x, y + y, z + z) (x, y, z)
= lim
ds s0 s
x y z
= + +
x s y s z s
We can also write

d = dx + dy + dz
x y z
p
= ( + + ) ds / 2 + 2 + 2
x y z
d 1
= ( + + )p
ds x y z + 2 + 2
2

We wish to find the direction (i.e. a particular n = (, , ) ) in which the

function varies most rapidly. For this, we want to determine those values of
, and which maximizes d/ds. This is found from the conditions
d d d
= 0, = 0, =0
ds ds ds
Now,
d 1 1 1
= 2( + + ) 2
z ( + + 2 )3/2
2
p
ds x + +
2 2 2 2 x y
h i 1
= ( 2 + 2 )
x y z ( + + 2 )3/2
2 2

Hence we get

( 2 + 2 ) = 0
x y z

( 2 + 2 ) = 0
y z x

(2 + 2 ) = 0
z x y
1.2. THE GRADIENT 3

Solving for z from the first equation of the above, we get

2 + 2
=
z x y
Replacing this into the last equation, we get

2 + 2 2


( + 2 ) (2 + 2 ) + = 0
x y

2 2 + 4 + 2 2 (2 + 2 + 2 )


= 0
x y
/x
=
/y
From the above and by similar calculations, we get:
/x /y /z
= , = , =
/y /z /x
Recall that
2 + 2 + 2 = 1
which gives
 (/x)2 (/z)2 2 
2 2
+ + =1
(/y)2 (/y)2
(/y)2
2 =
(/x)2 + (/y)2 + (/z)2
(/y)
= p = k(/y)
(/x) + (/y)2 + (/z)2
2

Hence we get
 
n(max change in ) = n = (, , ) = k , ,
x y z
where k is a proportionality constant, which is multiplied with all the compo-
nents, hence does not change the values of /x, /x and /x, relative
to each other. Hence, the direction of the vector (
x , y , z ) is the same as
that of (, , ) = k(
x , y , z ).

Hence we get that:

 
The vector
x , y , z , gives the direction of the maximum rate of change
of . We call the vector above as the gradient of a scalar field :

~ = , , = ex + ey + ez
 
grad =
x y z x y z
4CHAPTER 1. INTRODUCTION AND MATHEMATICAL PRELIMINARIES

We can arrive at the same conclusion by the following argument as well. We

have
d ~ ~ d~s
= ().n = ().
ds ds
~ ~
d = ().d~s = || |d~s| cos

where is the angle between ~ and d~s. If we fix the magnitude |d~s| and
search for the various directions (that is vary ), the maximum change in will
occur when = 0 (i.e. for cos = 1). Thus, for a fixed distance |d~s|, d is the
greatest, when we move in the same direction as . ~

~ points in the direction of the maximum rate of change of the

The gradient
function .

~ gives the rate of change of the function along the maximal

The magnitude ||
direction.

The Fundamental Theorem for Gradient

Recall that the fundamental theorem of calculus states that
Z b Z b
df
dx = F (x)dx = f (b) f (a)
a dx a

where F (x) = df
dx is the derivative of the function f . Thus
The integral of a derivative over an interval is given by the value of the function
at the end points (boundaries) of the interval.

In comparison to the above, for the gradient (which is a kind of derivative), we

have the fundamental theorem for gradient as:
The line integral of the gradient of a scalar field between two points a to b, is
equal to the difference of the scalar field at the boundaries
Z ~b
~
().d~l = (~b) (~a)
~
a
P

where the end points are in general points in the space greater than dimension
1, and hence denoted by vectors.

Two immediate corollaries from the above theorem are:

R ~b
~
().d ~l is independent of the path taken from ~a to ~b if is defined throughout
~
a
the path.

~ ~l = 0, since the beginning and the end points are identical.

H
().d
1.3. THE DIVERGENCE 5

1.3 The Divergence

We define the divergence of a vector field F~ (x, y, z) as

1
I
div F~ = lim F~ .nda (1.1)
V 0 V S

where V is a volume enclosed by the closed surface S, n is a unit normal vector

at any point on the surface, pointing outward, da is a differential element of
area, at any point on the surface.

Geometrical Interpretation of Divergence

The geometrical interpretation of the above definition of the divergence is as
follows:

We compute the net flux coming out of the closed surface S enclosing a point
~r and divide by the volume V , that contains ~r and is enclosed by S, in the
limit when the volume V has been shrunk to a point at ~r (i.e. in the limit of
infinitesimal volume). Then the divergence is the total outward flux per unit
volume at that point.

Consider a small parallelepiped ABCDEF GH, with sides of length x, y

and z parallel to the coordinate axes. Let the coordinates of the points be

A (x0 , y0 , z0 ), B (x0 + x, y0 , z0 )
C (x0 + x, y0 , z0 + z), D (x0 , y0 , z0 + z)
E (x0 , y0 + y, z0 ), F (x0 + x, y0 + y, z0 )
G (x0 + x, y0 + y, z0 + z), H (x0 , y0 + y, z0 + z)

with volume V = xyz.

Consider the two faces ABCD and EF GH lying parallel to the x-z plane. The
surface integral yields approximately (for any point in the surfaces)

xz F~ (x, y0 , z).(ey ) + F~ (x, y0 + y, z).(ey )

 
Ix-z =
 
= xz Fy (x, y0 + y, z) Fy (x, y0 , z)

Fy 
(xz)y
y 

x,y0 ,z

Similarly, for the surface integrals along the pair of surfaces lying parallel to the
6CHAPTER 1. INTRODUCTION AND MATHEMATICAL PRELIMINARIES

x-y and y-z planes, we get

Iy-z = yz F~ (x0 , y, z).(ex ) + F~ (x0 + x, y, z).(ex )

 
 
= yz Fx (x0 + x, y, z) Fx (x0 , y, z)

Fx 
(yz)x
x 

x0 ,y,z

xy F~ (x, y, z0 ).(ez ) + F~ (x, y, z0 + z).(ez )

 
Ix-y =
 
= xy Fz (x, y, z0 + z) Fz (x, y, z0 )

Fz 
(xy)z
z 

x,y,z0

Combining these results, in the limit V = xyz 0, and noticing that, the
volume shrinks to the point (x0 , y0 , z0 ), we get
1
div F~
 
= lim Ix-z + Iy-z + Ix-y
xyz0 xyz
" 
1 Fy 
= lim xzy
xyz0 xyz y 

x,y0 ,z
 
Fx  Fz  i
+yzx + xyz
x  z 
 
x0 ,y,z x,y,z0
h F Fy Fz i
x
= + +
x y z (x0 ,y0 ,z0 )
~ ex + ey + ez , we can write the divergence of


Using the operator x y z
a vector field F~ (x, y, z) as

~ F~ = Fx + Fy + Fz
div F~ .
x y z

Observe that, the divergence of a vector field F~ (x, y, z) is itself a scalar, since
the divergence of a scalar is meaningless. The divergence . ~ F~ is a measure of
~
how much the vector F spreads out or diverges from the point in question.

Thus, if a vector field has some source at some point, then its divergence will
be positive, vectors will point outward from the source point, while if a vector
field has some sink at some point, its divergence at that point will be negative;
vectors will converge toward that point. On the other hand, if a vector field
has neither a source nor a sink at some point, it will be divergence-less at that
point.

Gauss Divergence Theorem

1.4. THE CURL 7

Gauss divergence theorem states that:

~ F~ , over some
The volume integral of the divergence of a vector field i.e. of .
volume V is equal to the surface integral of the outward normal component of
the vector field over the closed surface. The latter is just the outward flux of
the vector field from the closed surface.
Z I
~ ~
.F dV = F~ .n da
V S

Gauss divergence theorem is also called the fundamental theorem for diver-
gence, in comparison with the fundamental theorem of calculus. It relates the
volume integral of the divergence of a vector function with the integral of the
function at the boundary of the region (the surface enclosing the volume).

1.4 The Curl

The curl of a vector field F~ (x, y, z) is defined equivalently by the followings:

1
I
1. curl F~ = lim n F~ da
V 0 V S1
1
I
2. n.curl F~ = lim F~ .d~l
S0 S2 C

where V is the volume enclosed by the closed surface S1 while C is the closed
curve, traversed in the positive sense around an open surface S2 .

The First Definition of Curl

To define curl from the first definition, we again take the parallelepiped ABCDEF GH,
with sides of length x, y and z parallel to the coordinate axes, considered
in the section for the divergence.

Consider the two faces ABCD and EF GH lying parallel to the x-z plane. The
surface integral yields approximately (for any point in the surfaces)

xz (ey ) F~ (x, y0 , z) + (ey ) F~ (x, y0 + y, z)

 
Jx-z =
 
= xz ex Fz (x, y0 + y, z) ez Fx (x, y0 + y, z) ex Fz (x, y0 , z) + ez Fx (x, y0 , z)



= xz ex Fz (x, y0 + y, z) Fz (x, y0 , z) ez Fx (x, y0 + y, z) Fx (x, y0 , z)
" #
Fz Fx
(xz)y ex ez
y y
x,y0 ,z

Similarly, for the surface integrals along the pair of surfaces lying parallel to the
8CHAPTER 1. INTRODUCTION AND MATHEMATICAL PRELIMINARIES

x-y and y-z planes we get

Jy-z = yz (ex ) F~ (x0 , y, z) + (ex ) F~ (x0 + x, y, z)
 
 
= yz ey Fz (x0 + x, y, z) ez Fy (x0 + x, y, z) ey Fz (x0 , y, z) + ez Fy (x0 , y, z)



= yz ey Fz (x0 + x, y, z) Fz (x0 , y, z) ez Fy (x0 + x, y, z) Fy (x0 , y, z)
" #
Fz Fy
(yz)x ey ez
x x
x0 ,y,z

xy (ez ) F~ (x, y, z0 ) + (ez ) F~ (x, y, z0 + z)

 
Jx-y =
 
= xy ey Fx (x, y, z0 + z) ex Fy (x, y, z0 + z) ey Fx (x, y, z0 ) + ex Fy (x, y, z0 )



= xy ey Fx (x, y, z0 + z) Fx (x, y, z0 ) ex Fy (x, y, z0 + z) Fy (x, y, z0 )
" #
Fx Fy
(xy)z ey ex
z z
x,y,z0

Again, combining these results, in the limit V = xyz 0, and noticing

that, the volume shrinks to the point (x0 , y0 , z0 ), we get
1
curl F~
 
= lim Jx-z + Jy-z + Jx-y
xyz0 xyz
" !
1 Fz Fx
= lim xzy ex ez
xyz0 xyz y y
x,y0 ,z
! !
Fz Fy Fx Fy
+yzx ey ez + xyz ey ex
x x z z
x0 ,y,z x,y,z0
" #
Fz Fy
Fx Fz
Fy Fx

curl F~ = ex + ey + ez
y z z x x y
(x0 ,y0 ,z0 )

~ we can write the curl of a vector field as a deter-

Using the vector operator ,
minant:  
 ex ey ez 
curlF~ =  x y
 

z 
 Fx Fy Fz 

The Second Definition of Curl

Consider a rectangular loop ABCD around a point P (x , y , z ), which the
mid-point of the rectangle (the mid-point is used for ease of calculation, any
other point-within the loop would also do), taken parallel to the x-z plane. We
will compute the line integral of the vector field F~ around the loop ABCD and
equate it with the surface integral over the face with the unit normal vector
pointing in the ey direction.

The choice of the unit vector is determined by the sense of traversal along the
 1.4. THE CURL 9

boundary of the loop ABCD. If we curl up the fingers of the right hand along
the path of the line integral, the thumb will point in the direction of the unit
vector normal to the surface.

1
The coordinates of the corner points of the loop are:

1 1 1 1
A (x x, y , z z), B (x + x, y , z z)
2 2 2 2
1 1 1 1
C (x + x, y , z + z), D (x x, y , z + z)
2 2 2 2

Denoting the boundary of the surface ABCD by C, we get

I Z Z Z Z
F~ .d~l = F~ .d~l


+ + +
C AB BC CD DA

But
Z Z
F~ .d~l = Fx dx
AB AB
F x |(x ,y ,z 1 z) x
2

1 1 1 1 1 
= Fx (x + x, y , z z) + Fx (x x, y , z z) x
2 2 2 2 2

"
1 1 1 Fx 
Z

F~ .d~l Fx (x , y , z z) + x
2 2 2 x (x ,y ,z 12 z)

AB
#
1 1 Fx 
+ Fx (x , y , z z) x x
2 2 x (x ,y ,z 12 z)


1
= Fx (x , y , z z)x
2
h 1 Fx  i
Fx (x , y , z ) z  x
2 z (x ,y ,z )

where F z |(x ,y ,z 1 z) is the average value of the integrand at the mid-point

2
along AB.
1 It is interesting to note that, the curl is defined for a point. We can shrink the loop to

any point within the loop, or even on the loop and get the curl at that point. It is not
necessary that the point P must be the mid-point of the loop ABCD. The reason
for this to work, is that if the vector field is smooth enough (i.e. Taylor series expansion
can be done), there is no abrupt change in the vector field and hence we should not except
any abrupt change in the curl (which is a derivative). Then in the infinitesimal limit, the
difference between contributions from different loops going through or around the point P ,
having different shapes, are vanishingly small.
 10CHAPTER 1. INTRODUCTION AND MATHEMATICAL PRELIMINARIES

Similarly, we get

Z Z
F~ .d~l = Fz dz
BC BC
F z |(x + 1 x,y ,z ) z
2

1 1 1 1 1 
= Fz (x + x, y , z z) + Fz (x + x, y , z + z) z
2" 2 2 2 2
1 1 1 Fz 
Z

F~ .d~l Fz (x + x, y , z ) z
2 2 2 z (x + 12 x,y ,z )

BC
#
1 1 Fz 
+ Fz (x + x, y , z ) + z z
2 2 z (x + 12 x,y ,z )


1
= Fz (x + x, y , z )z
2
h 1 Fz  i
Fz (x , y , z ) + x  z
2 x (x ,y ,z )

and
Z Z
F~ .d~l = Fx dx
CD CD

F x |(x ,y ,z + 1 z) x
2

1 1 1 1 1 
= Fx (x x, y , z + z) + Fx (x + x, y , z + z) x
2" 2 2 2 2
1 1 1 F
Z 
x
F~ .d~l Fx (x , y , z + z) x
2 2 2 x (x ,y ,z + 21 z)

CD
#
1 1 Fx 
+ Fx (x , y , z + z) + x x
2 2 x (x ,y ,z + 12 z)


1
= Fx (x , y , z + z)x
2
h 1 Fx  i
Fx (x , y , z ) + z  x
2 z (x ,y ,z )

and
Z Z
F~ .d~l = Fz dz
DA DA
F z |(x 1 x,y ,z ) z
2

1 1 1 1 1 
= Fz (x x, y , z + z) + Fz (x x, y , z z) z
2 2 2 2 2
 1.4. THE CURL 11
"
1 1 1 Fz 
Z

F~ .d~l Fz (x x, y , z ) + z
2 2 2 z (x 21 x,y ,z )

DA
#
1 1 Fz 
+ Fz (x x, y , z ) z z
2 2 z (x 21 x,y ,z )


1
= Fz (x x, y , z )z
2
h 1 Fz  i
Fz (x , y , z ) x  z
2 x (x ,y ,z )
Combining the contributions, we get
I  F Fx 
z
F~ .d~l xz
x z (x ,y ,z )

ABCD

The area element is S = xz. Thus in the limit S 0, we get

1
I
ey . curl F~ = ( curl F~ )y = lim F~ .d~l
S0 xz ABCD
 F Fx 
z
=
x z (x ,y ,z )


Similarly, considering loops oriented along the x-y and y-z planes, we can cal-
culate the z- and the x-components of the curl. The final result is as before
 F Fy   F Fz   F Fx 
z x y
curl F~ = ex + ey + ez
y z z x x y

Geometrical Interpretation of Curl

The curl measures the twist of the vectors in the vector field in a region; a
region of high curl is a whirlpool - if we put a tiny paddle, which responds to
the vector F , then the paddle will rotate. The curl of a vector field (say) F~
~ F~ , is a measure of how much the vector F~ curls around the point
i.e.
in question.

Now, the surface integral of the curl of a vector field F~ over a surface is the flux
of the curl through the surface. This represents the total amount of swirl; we
can determine the total swirl just as well by going around the edge and find
out how much the flow is following the boundary.

Stokes Theorem
Stokes theorem states that:

The surface integral of the curl of a vector field over an open surface S2 is
equal to the line integral of the vector field along the boundary C of the surface,
traversed in the positive sense.
Z   I
curlF~ .nda = F~ .d~l
S2 C
12CHAPTER 1. INTRODUCTION AND MATHEMATICAL PRELIMINARIES

Stokes theorem is also called the fundamental theorem of curl. This is because,
the integral of a derivative (here the curl), over a region (here a patch of a
surface), is equal to the value of the function at the boundary of the region
(here the perimeter of the patch). As in the case of the divergence theorem, the
boundary term itself is an integral - specifically, a closed line integral.

1.5 ~
The Operator Nabla
~ as
We introduce the vector differential operator del or nabla

~ ex + ey + ez

x y z
Here, del is not a vector in the usual sense. It is without any specific meaning,
until we provide it with a function to act upon. It is a differential operator as
well - an instruction to differentiate what follows.

We can express the gradient, divergence and the curl using the nabla operator
as follows:

grad ~
=
div F~ ~ F~
= .
curl F~ = F~
~

1.6 The Tensorial Form of Operators

Einsteins Summation Convention
Introduce Einsteins summation convention that every repeated index is summed
over all the possible values of it. With this, we can express sums very compactly
like
3
~B ~ = A1 B1 + A2 B2 + A3 B3 =
X
A. Ai Bi = Ai Bi
i=1

where we adopt the notation of denoting the x- component by the subscript (or
superscript, as appropriate) of 1, and so on.

Kronecker Delta
The Kronecker delta is defined as

1 if =
=
0 otherwise

The contraction with the Kronecker delta is according to the rule

Aij is = Asj , F = F , etc.

1.6. THE TENSORIAL FORM OF OPERATORS 13

Orthogonality of Basis Unit Vectors

For convenience, we will use orthogonal coordinate system, e.g. the Cartesian
coordinate system, spherical polar coordinate system, cylindrical coordinate sys-
tem, etc. All these have the special property that the basis unit vectors are
perpendicular to each other, at any point. In Cartesian coordinate system, the
basis unit vectors are ex = e1 , ey = e2 , ez = e3 , are their dot and cross products
give

ex .ex = ey .ey = ez .ez = 1

ex .ey = ey .ez = ez .ex = 0
ex ex = ey ey = ez ez = 0
ex ey = ez , ey ez = ex , ez ex = ey

The first two relations can be conveniently written as

ei .ej = ij

Notice the dot product symbol, ei . ej is very different from ei ej , the first is a
scalar, while the second is a unit dyad.
The proof of the third and the fourth relations are left as exercises (which require
the use of Levi-Civita symbol introduced below).

Levi-Civita Symbol
The Levi-Civita symbol in three dimension is a completely anti-symmetric tensor
of rank three which is normalized to 1:

1 if (ijk) is an even permutation of (123)
ijk = 1 if (ijk) is an odd permutation of (123)
0 otherwise

With this, the cross product of two vectors can be expressed as

~B
A ~ = e1 A2 B3 e1 A3 B2 + e2 A3 B1 e2 A1 B3 + e3 A1 B2 e3 A2 B1
~ B)
= ei (A ~ i
= ei (ijk Aj Bk )

It may be noticed that the scalar-triple product of the unit basis vectors is
nothing but the Levi-Civita symbol as seen below:




e1 . e2 e3 = e1 .e1 = 1, e1 . e3 e2 = e1 . e1 = 1




e2 . e3 e1 = e2 .e2 = 1, e2 . e1 e3 = e2 . e2 = 1




e3 . e1 e2 = e3 .e3 = 1, e3 . e2 e1 = e3 . e3 = 1

The anti-cyclic orders give 1, because of the anti-commuting nature of the

cross product. If any two of the basis vectors involved in the triple product is
14CHAPTER 1. INTRODUCTION AND MATHEMATICAL PRELIMINARIES

the same, it is obviously zero. Hence, we can write

ei . ej ek = ijk

The Levi-Civita tensor has a special contraction property that will prove to be
very helpful
ijk klm = il jm im jl

Compact Notation for Partial Derivatives

Next, we introduce a compact notation for the partial derivative

x = 1 , y = 2 , z = 3
x y z

Higher derivatives are expressed in the obvious manner

2 3 4
= 1 2 , = 12 2 , = 3 12 2 , etc.
xy x2 y zx2 y

Symmetry Properties of Tensors

Recall that, any general tensor of rank greater than or equal to two can be
expressed as a sum of symmetrized and anti-symmetrized parts in a pair of
indices :
1
1

Aij = A(ij) + A[ij] = Aij + Aij + Aij Aij
2 2
where by symmetry property,

A(ij) = A(ji) and A[ij] = A[ji]

Contraction between a symmetric tensor and and anti-symmetric tensor leads

to a null or zero result:

S T = S T since T is anti-symmetric
= S T since S is symmetric
= Sji Tji since and are dummy indices, replacing i and j
= S T replacing j and i
= 0 since if something is equal to its opposite, it is necessarily zero

Application of the above result can be immediately seen in vector algebra, e.g.
~ A
A.( ~ B)
~ = Ai (A
~ B)
~ i = Ai ijk Aj Bk = ijk (Ai Aj )Bk = 0

where ijk is anti-symmetric in the first two indices while, Ai Aj is symmetric

in those indices.
1.7. IDENTITIES OF VECTOR CALCULUS 15

1.6.1 Tensorial Form of Operators

With the above definitions and tools been developed, we can now express the
vector operators like gradient, divergence and curl in tensor notations:
3
~
X
= ei = ei i
i=1
xi
3
~ F~
X Fi
. = = i Fi
i=1
xi

~ F~ F3 F2
F1 F3
F2 F1

= e1 + e2 + e3
x2 x3 x3 x1 x1 x2
3
X
= ei ijk j Fk = ei ijk j Fk
i=1

1.7 Identities of Vector Calculus

Product Rules
Using the tensorial notation, the product rules for vector calculus can be derived
very easily:
~ + g) = ei i (f + g) = ei i f + ei i g = f
(f ~ + g ~
~ A
.( ~ + B)
~ = i (Ai + Bi ) = i Ai + i Bi = (. ~ A)
~ + (.~ B)
~
~ (A
~ + B)
~ = ei ijk j (Ak + Bk ) = ei ijk j Ak + ei ijk j Bk = (~ A)
~ + (
~ B)
~

For a constant k, we simply have

~
(k) ~
= k , ~
.(k ~ = k(.
A) ~ A),
~ ~ (k A)
~ = k(
~ A)
~

There are two ways to make scalar as the product of two function:

fg (product of two scalar functions)

~B
A. ~ (dot product of two vector functions)

and two ways to make a vector:

fA~ (scalar times a vector)

~B
A ~ (cross product of two vector functions)

Accordingly, there are six product rules:

Two for the gradient:

~ g)
(f ~ + (f
= f (g) ~ )g
~ A.
( ~ B)
~ ~ (
= A ~ B)
~ +B ~ (
~ A)
~ + (A.
~ )
~ B
~ + (B.
~ )
~ A
~
16CHAPTER 1. INTRODUCTION AND MATHEMATICAL PRELIMINARIES

Two for the divergence:

~
.(f ~ = f (.
A) ~ A)
~ + A.(
~ f
~ )
~ A
.( ~ B)
~ ~
= B.( ~ A)
~ A.(
~ ~ B)
~

Two for the curl:

~ (f A)
~ ~ A)
= f ( ~ A ~ (f
~ )
~ (A
~ B)
~ ~ )
= (B. ~ A~ (A.
~ )
~ B
~ + A(
~ .
~ B)
~ B(
~ .
~ A)
~

Proof of Product Rules

~ g) = ei i (f g) = ei (i f )g + f (i g)


(f


= (ei i f )g + f (ei i g)
~ )g + f (g)
= (f ~
To prove
~ A.
( ~ B)
~ =A
~ (
~ B)
~ +B
~ (
~ A)
~ + (A.
~ )
~ B
~ + (B.
~ )
~ A
~
consider,
~ A. ~ B)
~ = ei i Aj Bj


LHS = (



= ei i Aj Bj + ei Aj i Bj



= ei i Aj Bj + Aj i Bj
RHS = A~ ( ~ B)
~ +B ~ (~ A) ~ + (A.
~ )
~ B~ + (B.
~ )
~ A~
~ B ~ ~ A ~



= ei ijk Aj k
+ ei ijk Bj k



+ Aj j ei Bi + Bj j ei Ai



= ei ijk Aj klm l Bm + ei ijk Bj klm l Am



+ ei Aj j Bi + Bj j Ai
h

i
= ei Aj il jm im jl l Bm + Bj il jm im jl l Am



+ ei Aj j Bi + Bj j Ai
h





i
= ei Aj i Bj Aj j Bi + Bj i Aj Bj j Ai + Aj j Bi + Bj j Ai
h

i
= ei Aj i Bj + Bj i Aj
= LHS

~ ~ =



.(f A) ei i . f ej Aj



= ei .ej i f Aj = ij f (i Aj ) + Aj (i f )



= f i Ai + Ai i f
= f (.~ A) ~ + A.(
~ f~ )
1.7. IDENTITIES OF VECTOR CALCULUS 17

To prove
~ A
.( ~ B)
~ = B.(
~ ~ A)
~ A.(
~ ~ B)
~

consider,
~ A ~ B)
~ = i A~B ~


LHS = .(

 i 
= i ijk Aj Bk = ijk (i Aj )Bk + Aj (i Bk )

RHS = ~
B.( ~ A)
~ A.(
~ ~ B)
~
~ A~ Ai ~ B ~



= Bi i i
= Bi ijk (j Ak ) Ai ijk (j Bk )
= B ( A ) (with i , j , k )
A k ( Bk ) (with i , j )
= ( A )B (from the cyclic property of )
+ A k ( Bk ) (from anti-symmetric property of k )
= ijk (i Aj )Bk (with i, j, k)
+ Aj ijk (i Bk ) (with i, j)
 
= ijk (i Aj )Bk + Aj (i Bk )
= LHS

LHS ~ (f A)
= ~ = ei ijk j (f Ak ) = ei ijk (j f )Ak + ei ijk f (j Ak )
RHS ~ A)
= f ( ~ A~ (f ~ )
= f ei ijk (j Ak ) ei ijk Aj (k f )
= ei ijk (j Ak ) ei ikj Ak (j f ) ( with k j, j k)
= ei ijk (j Ak ) + ei ijk Ak (j f ) (from anti-symmetric property of ijk )
= LHS

~ (A ~ B)
~ = ei ijk j A~B ~ = ei ijk klm j Al Bm



LHS = k



= ei il jm im jl Al (j Bm ) + (j Al )Bm
h i
= ei Ai (j Bj ) Aj (j Bi ) + Bj (j Ai ) Bi (j Aj )

RHS ~ )
= (B. ~ A~ (A. ~ )
~ B
~ + A( ~ .
~ B)~ B( ~ .~ A)~
= Bj (j )(ei Ai ) Aj (j )(ei Bi ) + (ei Ai )(j Bj ) (ei Bi )(j Aj )
h i
= ei Bj (j Ai ) Aj (j Bi ) + Ai (j Bj ) Bi (j Aj )
= LHS
18CHAPTER 1. INTRODUCTION AND MATHEMATICAL PRELIMINARIES

Second Derivatives
~ twice on a scalar/vector field, in
By applying the vector differential operator
appropriate manners, we can construct five species of second derivatives which
are either scalars or vector.

For the gradient, we have two further derivatives, curl and divergence of the
~
vector field :

(1) Divergence of a gradient: ~ )

.( ~
(2) Curl of a gradient: ~ ()
~

For the divergence, since it is a scalar - we can take only the gradient of it:
~ .(
(3) Gradient of Divergence: ( ~ F~ ))

For the curl, we have two further derivatives, curl and divergence of the vector
~ F~ :
field
~
(4) Divergence of a curl: .( ~ F~ )
(5) Curl of a curl: ~ (
~ F~ )

The Laplacian or div (grad)

For frequent use in vector calculus and in the formulation of physical theories,
the divergence of a gradient is given a special name, the Laplacian operator,
~ )
denoted by .( ~ 2 . It is a scalar operator, and hence, the vector symbol
is not put on the operator , appearing in the Laplacian.
   
~ )
.( ~ = ex + ey + ez . ex + ey + ez
x y z x y z



= ei . ej
xi xj

= ei .ej = ij =
xi xj xi xj xi xi
2
= = i i
x2i

We can however, form the Laplacian of a vector, which is convenient way of

combining, the Laplacian of the components of a vector, into another vector:
~ 2 ex Ax + ey Ay + ez Az
2 A

= (2 Ax )ex + (2 Ay )ey + (2 Az )ez

= i i ej Aj
= ej i i Aj
1.7. IDENTITIES OF VECTOR CALCULUS 19

curl (grad)
The curl of a gradient for a sufficiently smooth scalar field is always zero:

~
~ =0

To prove this we use the following result:

If a scalar field = (x, y, z), is smooth enough, i.e. its partial derivatives
up to second order (i , i j , i, j = 1, 2, 3), exist and the second order partial
cross-derivatives are continuous, then the order of the differentiation is imma-
terial (i.e. the cross-derivatives are equal ):

2 2
= , for i, j = 1, 2, 3
i j j i

Then we have

~
~ ~ k
= ei ijk j ()
= ei ijk j k
= ei ijk k j
1

= ei ijk j k + k j = ei ijk (j k)
2
= 0

where we have
used the fact that contraction of a symmetric tensor (here
1
2 j k + k j = (j k) ) with an anti-symmetric tensor ( here ijk ) gives
zero.

grad (div)
The gradient of the divergence of a vector field F~ , can be written conveniently
as
~ F~ = ei i j Fj
~ .

Note that, it is different from the Laplacian of the vector field F~ , since the
former is i i F~ = ej i i Fj ; the summation is done between different objects.

div (curl)
The divergence of the curl of a sufficiently smooth vector field is always zero:

~
.( ~ F~ ) = 0

Proof:
Since, we assume that the vector field F~ is sufficiently smooth, the cross-partial
derivatives of any particular component of F~ are equal; e.g. i j Fl = j i Fl .
20CHAPTER 1. INTRODUCTION AND MATHEMATICAL PRELIMINARIES

Then
~ ~ F~ ) = i ei . ej jkl k Fl



.(
= (ei .ej )i jkl k Fl
= ij i jkl k Fl using orthogonality of basis vectors
= jkl j k Fl contraction with Kronecker delta
= 0 contraction of anti-symmetric jkl with symmetric (j k)

Notice that, we have made use of the fact that cross-partial derivatives of the
components of F~ are equal, in constructing the symmetric tensor (j k) Fl =
1
2 (j k + k j )Fl .

curl (curl)
This second derivative is also widely used in physics, occurring in many appli-
cations. It comes out to be
~ ~ F~ = ~ .
~ F~ 2 F~

That is
curl (curl) F~ grad (div F~ ) - Laplacian F~

Proof:
~
~ F~ ~ F~



= ei ijk j k
 
= ei j ijk klm l Fm


= ei j il jm im jl l Fm
= ei j i Fj ei j j Fi



= ei i j Fj j j ei Fi
~ . ~ F~ 2 F~


=

Notice that, we again used the assumption of sufficient smoothness of the vector
field F~ in changing the order of the partial derivatives i.e. in j i Fl = i j Fl .
Considering all the second derivatives above, we really, then have only two
kinds of derivatives: the Laplacian and the gradient-of-divergence. Of these,
the Laplacian is of fundamental importance and widely used.

1.8 Operators in Orthogonal Curvilinear Coor-

dinate Systems
Orthogonal Curvilinear Coordinate Systems
Any three numbers u1 , u2 , u3 uniquely specifying the position of a point in space
are called the coordinates of the point in space. We consider the special kind of
coordinate systems which are:
1.8. OPERATORS IN ORTHOGONAL CURVILINEAR COORDINATE SYSTEMS21

1. Orthogonal System
The unit vectors in the increasing directions of the coordinates are or-
thogonal i.e. perpendicular to each other.
2. Curvilinear System
The surfaces of constant values of the coordinates (u1 , u2 , u3 ), called the
coordinate surfaces, intersect each other, in general, along curves, called
the coordinate curves or lines.

In an orthogonal curvilinear coordinate system:

1. The coordinate surfaces intersect at right angles with each other.
2. The tangent vectors to the coordinate lines at any point are perpendicular
to each other.

The u1 -, u2 - and u3 -coordinate curves of a curvilinear system are analogous to

the x-, y- and z-coordinate axes of a rectangular system. Surfaces of constant
values of u1 , u2 and u3 in rectangular system (where u1 = x, u2 = y, u3 = z)
are the y-z, x-z and x-y planes respectively.

As another example, for the cylindrical coordinate system:

= 0 cylinder of radius 0
= 0 half-infinite plane containing the z- axis
z = z0 plane parallel to the x-y plane
The coordinate curves for cylindrical coordinates are:
= 0 , = 0 straight line parallel to the z axis
= 0 , z = z0 straight line pointing radially outward in the z = z0 plane
z = z0 , = 0 circle of radius 0 in the z = z0 plane

Unit Basis Vectors in Curvilinear Coordinate Systems

By a basis of a generalized coordinate system u1 ,u2 ,u3 we mean any set of
three vectors ~e1 , ~e2 , ~e3 of (locally) fixed length, pointing tangent to the positive
direction of the of the coordinate curves. Thus, ~e1 is tangent to the coordinate
curve of u1 (u2 and u3 fixed). The basis is said to be local, since it varies from
point to point.

Consider a general point P (u1 , u2 , u3 ) (x, y, z). Since, both the sets of
coordinates refer to the single point P and there is a 1-to-1 mapping between
the triplets (x, y, z) and (u1 , u2 , u3 ). Thus, the rectangular coordinates can be
expressed as functions of the generalized coordinates (u1 , u2 , u3 ) as:
x = x(u1 , u2 , u3 ), y = y(u1 , u2 , u3 ), z = z(u1 , u2 , u3 )
22CHAPTER 1. INTRODUCTION AND MATHEMATICAL PRELIMINARIES

Unit Tangent Vectors and Scale Factors

Let ~r = xe1 + ye2 + ze3 be the position vector of the point P . Then we have
~r = x(u1 , u2 , u3 )e1 + y(u1 , u2 , u3 )e2 + z(u1 , u2 , u3 )e3 = ~r(u1 , u2 , u3 )
The tangent vector to the direction of u1 , at the point P (for which u2 and u3
~
r
are fixed) is given by u 1
. Then, a unit tangent vector in this direction is
,
~r  ~r 
e1 =
 
u1 u1
 

so that
~r  ~r 
= h1 e1 , where, h1 = 
 
u1 u1

Similarly, if e2 and e3 are unit tangent vectors to the u2 - and u3 -curves at P ,
respectively, then we have
~r  ~r 
= h1 e2 , where, h2 = 
 
u2 u2

~r  ~r 
 
= h1 e3 , where, h3 = 
u3 u3


Unit Vectors Normal to the Coordinate Surfaces

At each point, we can form another set of unit vectors which are normal to
the coordinate surfaces. At a point P , u ~ 1 is a vector, normal to the surface
u1 = c1 , where c1 is a constant. Similarly, we have u~ 2 and u
~ 3 vectors normal
to the surfaces u2 = c2 and u3 = c3 where c2 and c3 are constants. Thus a unit
vectors in the directions normal to the coordinate surfaces are given by
.  .  . 
E1 = u ~ 1 ,
~ 1 u E2 = u ~ 2 ,
~ 2 u E3 = u ~ 3 
~ 3 u

Thus, at each point P of a coordinate system, there exist, in general, two sets of
unit vectors e1 , e2 , e3 tangent to the coordinate curves and E1 , E2 , E3 , normal
to the coordinate surfaces. The two sets become identical if and only if
the curvilinear coordinate system is orthogonal.

The two sets of vectors u ~

r ~
r
, u ~
, ur ~ 1 , u
and u ~ 2 , u
~ 3 form a reciprocal
1 2 3
systems of vectors. Note that, in general, the basis vectors can be neither of
unit length nor mutually perpendicular. Coordinate systems whose basis vectors
intersect at right angles are said to be orthogonal.

Arc Length, Surface Element and Volume Element

We have ~r = ~r(u1 , u2 , u3 ). This gives,
~r ~r ~r
d~r = du1 + du2 + du3 = h1 du1 e1 + h2 du2 e2 + h3 du3 e3
u1 u2 u3
1.8. OPERATORS IN ORTHOGONAL CURVILINEAR COORDINATE SYSTEMS23

is the differential change of length of the vector. The differential arc length ds
is defined as the length of the vector d~r i.e.
  
ds = d~r.d~r = (hdu)i ei . (hdu)j ej
= (hdu)i (hdu)j ei .ej

For orthogonal system, we have ei .ej = ij . This gives,

ds2 (ds)2 = h21 du21 + h22 du22 + h23 du23

Along a u1 -curve, u2 and u3 are constants, so that the differential displace-

ment vector along the u1 -direction is d~r = h1 du1 e1 . Similarly, we can consider
differential displacements along the increasing directions of u2 and u3 .

Thus, the differential arc lengths along the coordinate curves at the point P
are:
ds1 = h1 du1 , ds2 = h2 du2 , ds3 = h3 du3

The surface elements tangential to the u2 -u3 curves at the point P is:
   
dA1 = (h2 du2 e2 ) (h3 du3 e3 ) = h2 h3 e2 e3 du2 du3 = h2 h3 du2 du3
   

   
where we have used the fact that e2 e3  = e1  = 1.

Similarly, we have
 
dA2 = (h3 du3 e3 ) (h1 du1 e1 ) = h3 h1 du3 du1
 
 
dA3 = (h1 du1 e1 ) (h2 du2 e2 ) = h1 h2 du1 du2
 

The volume element for an orthogonal curvilinear coordinate system is given by

 
dV = (h1 du1 e1 ).(h2 du2 e2 ) (h3 du3 e3 ) = |h1 h2 h3 |du1 du2 du3
 
 
since e1 .e2 e3  = 1 and the modulus has been taken since the volume element
can not be negative.

Jacobian of Transformation
The Jacobian of the transformation (x, y, z) (u1 , u2 , u3 ) is given by
 
 x y z 
 u1 u1 u1 
 
 x, y, z  (x, y, z)  
 x y z
J = =  u  = h1 h2 h3

u 1 , u2 , u3 (u1 , u2 , u3 )  2 u2 u2 
 
 x
 u y z 
u u

3 3 3
24CHAPTER 1. INTRODUCTION AND MATHEMATICAL PRELIMINARIES

This is because, the determinant is equal to

x y z
x y z
x y z

e1 + e2 + e3 . e1 + e2 + e3 e1 + e2 + e3
u1 u1 u1 u2 u2 u2 u3 u3 u3
~r ~r ~r
=( ).( )( ) = (h1 e1 ).(h2 e2 ) (h3 e3 )
u1 u2 u3
= h1 h2 h3 (e1 .e2 e3 ) = h1 h2 h3

~
r ~
r ~
r
If the Jacobian is identically zero, then u 1
, u 2
, u 3
are coplanar vectors and
the curvilinear coordinate transformation breaks down. This implies that there
is a relation between the coordinates x, y and z having the form F (x, y, z) = 0.

Gradient, Divergence, Curl and Laplacian

The differential vector operators gradient, divergence, curl and the Laplacian
can compactly be expressed in terms of the general orthogonal curvilinear co-
ordinates (u1 , u2 , u3 ).

~ be a vector function of position of the orthog-

Let, be a scalar field and let A
onal curvilinear coordinates u1 , u2 , u3 , i.e.
= (u1 , u2 , u3 )
~ = A(u
A ~ 1 , u2 , u3 ) = A1 (u1 , u2 , u3 )e1 + A2 (u1 , u2 , u3 )e2 + A3 (u1 , u2 , u3 )e3

Then we have the following expressions for the vector differential operators:

~ = grad = 1 1 1
e1 + e2 + e3
h1 u1 h2 u2 h3 u3
~ F~ = div F~ 1 h i
. = (h2 h3 A1 ) + (h3 h1 A2 ) + (h1 h2 A3 )
h1 h2 h3 u1 u2 u3
 
 h1 e1 h2 e2 h3 e3 
 
 
~ F~ = curl F~ 1 


= u1 u2 u3 

h1 h2 h3 

 
 h 1 F1 h 2 F2 h 3 F3 

1 h  h2 h3   h3 h1   h3 h1 i
2 = Laplacian = + +
h1 h2 h3 u1 h1 u1 u2 h2 u2 u3 h2 u3

Special Orthogonal Coordinate Systems

1. Cylindrical Coordinates (, , z):
x = cos , y = sin , z=z
where, 0, 0 < 2, < z <
h = 1, h = , hz = 1
1.8. OPERATORS IN ORTHOGONAL CURVILINEAR COORDINATE SYSTEMS25

2. Spherical Polar Coordinates (r, , ):

x = r sin cos , y = r sin sin , z = r cos
where, r 0, 0 , 0 < 2
hr = 1, h = r, h = r sin

3. Elliptic Cylindrical Coordinates (u, v, z):

x = a cosh u cos v, y = a sinh u sin v, z=z
where, u 0, 0 v < 2, <z <
p p
hu = a sinh2 u + sin2 v, hv = a sinh2 u + sin2 v, hz = 1
The coordinate surfaces of this coordinate system are sheets of confocal
ellipses and hyperbolas parallel to the z-axis.

Gradient, Divergence, Curl and Laplacian in Special Orthogonal Co-

ordinate Systems

We consider a smooth scalar field V = V (, , z) and a smooth vector field

F~ = F~ (, , z) = F e + F e + Fz ez , with components F = F1 , F = F2 , Fz =
F3 . The gradient, divergence, curl and the Laplacian for these fields in special
orthogonal coordinate systems are given by:
1. Cylindrical Coordinate System
For cylindrical coordinate system (, , z), we have
u1 = , u2 = , u3 = z
e1 = e , e2 = e , e3 = ez
h1 = h = 1, h2 = h = , h3 = h z = 1
Hence, we have,

~ 1 V 1 V 1 V
V = e1 + e2 + e3
h1 u1 h2 u2 h3 u3
V 1 V V
= e + e + ez
z

~ F~ 1 h


i
. = h 2 h 3 F1 + h 3 h 1 F2 + h 1 h 2 F3
h1 h2 h3 u1 u2 u3
1h

i
= F + F + Fz
z

   
 h1 e1 h2 e2 h3 e3   e e ez 
 
   
~ F~ 1 
 1  

= 
u3  =

h1 h2 h3  u1 u2
  z 

   
 h 1 F1 h 2 F2 h 3 F3   F F Fz 
26CHAPTER 1. INTRODUCTION AND MATHEMATICAL PRELIMINARIES
" #
2 1 h2 h3 V
h3 h1 V
h1 h2 V

V = + +
h1 h2 h3 u1 h1 u1 u2 h2 u2 u3 h3 u3
1
h V
1 V
V
i
= + +
z z
1  V  1 2V 2V
= + 2 2
+
z 2

2. Spherical Polar Coordinate System

For spherical polar coordinate system (r, , ), we have

u1 = r, u2 = , u3 = ;
e1 = er , e2 = e , e3 = e
h1 = hr = 1, h2 = h = r, h3 = h = r sin

Hence, we have,

~ 1 V 1 V 1 V
V = e1 + e2 + e3
h1 u1 h2 u2 h3 u3
V 1 V 1 V
= er + e + e
r r sin

~ F~ 1 h


i
. = h 2 h 3 F1 + h 3 h 1 F2 + h 1 h 2 F3
h1 h2 h3 u1 u2 u3
1 h 2


i
= r sin F r + r sin F + rF
r2 sin r
1 2
1
1
= 2
r Fr + sin F + F
r r r sin r sin

   
 h1 e1 h2 e2 h3 e3   er re r sin e 
 
   
~ F~ 1 
 1 

= 
u3  =
  
h1 h2 h3  u1 u2
 r sin  r
2 

   
 h 1 F1 h 2 F2 h 3 F3   Fr rF r sin F 

" #
2 1 h2 h3 V
h3 h1 V
h1 h2 V

V = + +
h1 h2 h3 u1 h1 u1 u2 h2 u2 u3 h3 u3
1 h 2 V
r sin V
r V
i
= 2
r sin + +
r sin r r r r sin
1 2 V
  1  V  1 2V
= 2
r + 2 sin + 2 2
r r r r sin r sin 2
1.9. DIRAC DELTA FUNCTION 27

3. Elliptic Cylindrical Coordinate System

For elliptic cylindrical coordinate system (u, v, z), we have
u1 = u, u2 = v, u3 = z;
e1 = eu , e2 = ev , e3 = ez
p p
h1 = hu = a sinh2 u + sin2 v, h2 = hv = a sinh2 u + sin2 v, h3 = hz = 1
Hence, we have,

~ 1 V 1 V 1 V
V = e1 + e2 + e3
h1 u1 h2 u2 h3 u3
1 h V V i V
= p eu + ev + ez
a sinh2 u + sin2 v u v z

~ F~ 1 h


i
. = h 2 h 3 F1 + h 3 h 1 F2 + h 1 h 2 F3
h1 h2 h3 u1 u2 u3
1 h p
2 2

p 2 2

i
= sinh u + sin v F u + sinh u + sin v Fv
a (sinh2 u + sin2 v) u v

+ Fz
z
 
 h1 e1 h2 e2 h3 e3 

 
~ F~ 1 

=  
h1 h2 h3  u1 u2 u3 

 
 h 1 F1
h 2 F2 h 3 F3 
 p p 
2 2
 a sinh u + sin v eu a sinh2 u + sin2 v ev ez 
 
 
1 


=

2 2 2  u v z 
a (sinh u + sin v)  
 p p
 a sinh2 u + sin2 v Fu a sinh2 u + sin2 v Fv

Fz 
" #
2 1 h2 h3 V
h3 h1 V
h1 h2 V

V = + +
h1 h2 h3 u1 h1 u1 u2 h2 u2 u3 h3 u3
1 h V
  V
   2 2 2 V i
= + + a (sinh u + sin v)
a2 (sinh2 u + sin2 v) u u v v z z
2 2 i 2
1 h V V V
= 2 2 2
+ 2
+
a (sinh u + sin v) u
2 v z 2

1.9 Dirac Delta Function

The Dirac delta function is a distribution or a kind of improper function. It
is useful in describing quantities that exist only at a point, a line, or a surface,
28CHAPTER 1. INTRODUCTION AND MATHEMATICAL PRELIMINARIES

i.e. quantities that are constrained to one or more dimensions (less than the
number of dimensions than that of the space in which it resides). A -function
in one dimension may be visualized as a function that is zero everywhere, except
at the point where its argument vanishes. However, as it turns out, that at this
point, the -function has a value of infinity, but in such a way that the integral
of the delta function around this point (since at all other points the integral
gives zero contribution) is unity.

1.9.1 The Fundamental Property of the Delta Function

Consider the function () (x) given by
1
for 2 < x <
2
() (x) =

0 for |x| >

2

where is some positive number. We note the following properties for the
function () (x):
/2
1
Z
() (x)dx = [ ( )] = 1
/2 2 2
Z
() (x)dx = 1

for x=0
lim () (x) =
0+
0 for x 6= 0

Note that, as 0+ , the height of the hat becomes larger and larger but, its
width becomes vanishingly small, in such a way that the area under the spike
is always unity.

Now consider the integral:

Z
dx () (x)f (x)

where f (x) is some arbitrary function, well-defined for x = 0 and does not blow
up at any other point.

If is sufficiently small, the effective variation of f (x) over the integration

interval [/2, /2] is negligible. Thus, f (x) remains practically equal to
f (0) and we can replace f (x) within the integration sign by the constant value
of f (0) and factor it out:
Z Z
dx () (x)f (x) f (0) dx () (x) = f (0)

1.9. DIRAC DELTA FUNCTION 29

The smaller , the better the approximation. We therefore examine the limit
0+ , and get the fundamental property of the delta function as
Z
dx (x)f (x) = f (0)

which is valid for any function f (x) defined at the origin and not blowing up
at any other point. More generally, we may define the delta function centered
around a point x0 6= 0, i.e. (x x0 ) by:
Z
dx (xx0 )f (x) = f (x0 )

For arbitrary integration limits a and b, we have:

Z b
dx (x)f (x) = f (0) if 0 [a, b]
a
= 0 if 0
/ [a, b]

The Sifting Property

The word sift has the meaning of to put flour or some other fine substance
through a sieve or sifter or to sift something out of some other thing means to
separate something from a group of things1 . In this sense, the property of the
Dirac delta function to pick out one particular value of the function, at which
the argument of the delta function is zero, is called the sifting property2 .
Sometimes, it is also called the shifting property, in the sense that it shifts
all the values of the function to one particular value 3 .

1.9.2 Representations of the Delta Function

Besides the hat function above, we can define various other parametric repre-
sentations of the delta function, all of which approach the true delta function at
some appropriate limit of the parameter. These may arise naturally in physical
applications, and may well be used to derive physically meaningful results out of
the mathematical calculations using the properties of the Dirac delta function.

The following functions approach (x) when the parameter approaches zero
from the positive side:

1 |x|/
1. e
2
1 Oxford
Advanced Learners Dictionary, 7th ed.
2 Principles
of Optics, M. Born and E. Wolf, 7th ed. Cambridge University Press; Optics,
Eugene Hecht, 2nd ed., Addison-Wesley Publishing Company.
3 Fourier Transform Booklet by Will Hossack, Modern Optics, School of Physics, The

University of Edinburgh.
30CHAPTER 1. INTRODUCTION AND MATHEMATICAL PRELIMINARIES

Proof:
Z
1 1 0 1
Z Z Z
() |x|/ x/
dx (x) = dx e = dx e + dx ex/
2 2 2 0
1 h x/ i0 1 h i
= e + () ex/
2 2 0
1 1
= [1 0] [0 1] = 1
2 2

1 |x|  |x|/
2. 1+ e
4
Proof:
Z
1 |x|  |x|/
Z
()

dx (x) = dx 1 + e
4
Z 0
1 x  x/ 1 x
Z
dx 1 + ex/
 
= dx 1 e +
4 4 0
1 h x/ x x/ i0 1 h x i
= e e + ex/ + () ex/ + ex/ + ex/
4 4 0
1 1
= [1 0 + 1 0 0 0] [0 + 0 + 0 1 0 1] = 1
4 4
Here, we have taken the limit 0+ at the last step and used the fact that
the exponential function goes to zero, for its argument of x , faster than
any other algebraic function.

1 2 2
3. ex / Gaussian Function

Proof:
Z Z
1 1
Z
() x2 /2 2
dx (x) = dx e = dz ez

Z Z 1
1 2 1 t 2 t
= dz ez = 2 e dt
0 2
1 1
= ( ) = 1
2

1
4. Lorentzian Function
x + 2
2

Proof:

1 2 1
Z Z Z
()
dx (x) = dx 2 2
= dx 2
x + 0 x + 2
Z /2
2 sec2 d 2 1
= = =1
0 (1 + tan2 )
2 2
1.9. DIRAC DELTA FUNCTION 31

2 3  1 2
5.
x2 + 2
Proof:
Z
2 3 1 2 2 3  1 2
Z  Z
dx () (x) = dx
2 2
= dx 2
x + x + 2
44 /2 2 1 2 4 /2
Z  Z
= sec d 2 = cos2 d
0 (1 + tan2 ) 0
4 1 /2
Z
= d (1 + cos (2))
2 0
2  1 /2 
= + sin (2) 0 =1
2 2
Also, in the limit 0+ , the denominator has one power of more than the
numerator and hence blows up.

1  sin (x/) 
6. Sinc Function
x
Proof:
Z
1  sin (x/)  2  sin (x/) 
Z Z
dx () (x) = dx = d(x/)
x 0 x/
2  sin y  2
Z
= dy = =1
0 y 2
R
where we have used the result 0 dz (sin z)/z = /2, found from contour inte-
gration technique.

Now for x 6= 0
x
1 y
Z
sin (x/) = dy cos ( )
0
But cos (y/) is a highly oscillating function in the limit 0+ assuming peak
values 1 and 1, equal number of times and having a net area under the curve
zero, for any non-zero upper limit (x 6= 0) of the integration. As 0+ ,
each part of the graph is magnified in height, but we still have equal
numbers of positive and negative peaks. Hence, the limit
lim sin (x/) = 0
0+

For x = 0, the limit 0+ has to be taken in the last step after taking the
limit x 0. Since, limy0 [(sin y)/y] = 1, we have

1 sin (x/)  1h sin (x/) i 1
lim+ = lim+ lim = lim+
0 x 0 x0 x/ 0


x=0
32CHAPTER 1. INTRODUCTION AND MATHEMATICAL PRELIMINARIES

Thus,

1 sin (x/) 0 for x 6= 0
lim =
0+ x
for x=0

= (x)

Alternative Proof :
We can arrive at the same conclusions by looking at the plot of the function
sin (x)/(x), where = 1/ and as 0+ , . This function is well
behaved for any finite value of . The width of the central peak is of the order
= 1/, since the first zero of the function occurs at x = x = .

Hence, as 0+ , the central peak grows narrower and taller. However, the
area under the curve, has a value of unity. As we move from the origin, the
consecutive peaks have heights less than the previous peaks. Hence, in the limit
0+ , the area under the curve is essentially equal to the area under the
central peak, which has a height of infinity. Hence,

1 sin (x/)
lim = (x)
0+ x

 sin (x/) 2
7.
x
Proof:
Z
 sin (x/) 2 2  sin (x/) 2
Z Z
()
dx (x) = dx = d(x/)
x 0 x/
Z
2  sin y  2 2
= dy = =1
0 y 2
R
2
where we have used the result 0
dz sin(z)/z = /2, found from contour
integration technique.

The plot of the function (/) sin2 (x/)/x2 , is a symmetric plot and having
 

a central peak with width since the first zero of the function occurs at
x/ = x = . As 0+ , the central peak grows narrower and taller.
However, the area under the curve, has a value of unity. As we move from the
origin, the consecutive peaks have heights less than the previous peaks. Hence,
in the limit 0+ , the area under the curve is essentially equal to the area
under the central peak, which has a height of infinity. Hence,

 sin (x/) 2
lim = (x)
0+ x
1.9. DIRAC DELTA FUNCTION 33

42  sin (x/) 3
8.
3 x
Proof:
Z
42  sin (x/) 3 Z  sin (x/) 3
8
Z
dx () (x) = dx = d(x/)
3 x 3 0 x/
Z
8  sin y  3 8 3
= dy = =1
3 0 y 3 8
R
3
where we have used the result 0 dz sin(z)/z = 3/8, found from contour
integration technique.

The plot of the function (42 /3) sin3 (x/)/x3 , is similar to the plot of sin (x/)/(x) ,
   

and we can arrive at the conclusion that in the limit 0+ , (42 /3) sin3 (x/)/x3
 

approaches to the Dirac delta-function:

42  sin (x/) 3
lim+ = (x)
0 3 x

33  sin (x/) 4
9.
2 x
Proof:
Z
33  sin (x/) 4 3  sin (x/) 4
Z Z
dx () (x) = dx = d(x/)
2 x 0 x/
3  sin y 4 3
Z
= dy = =1
0 y 3
R
4
where we have used the result 0 dz sin(z)/z = /3, found from contour
integration technique.

The plot of the function (33 /2) sin4 (x/)/x4 , is similar to the plot of (/) sin2 (x/)/x2 ,
   

and we can arrive at the conclusion that in the limit 0+ , (33 /2) sin4 (x/)/x4
 

approaches to the Dirac delta-function:

33  sin (x/) 4
lim+ = (x)
0 2 x

1
Z
10. dk eikx = (x) Fourier Transform of Unity
2

Proof:
Z
1 1 h eikx ik= 1 h sin (kx) i 1 h sin (x/) i
dk eikx = = lim = lim = (x)
2 2 ix k= k x 0+ x
where = 1/k.
34CHAPTER 1. INTRODUCTION AND MATHEMATICAL PRELIMINARIES

1.9.3 Properties of the Delta Function

1. Delta Function is an Even Function

(x) = (x)

Proof:
Z Z Z
(x)f (x)dx = (y)f (y)dy = (y)f (y)dy

Z
= f (0) = (x)f (x)dx

(x) = (x)

2. Delta Function of the constant multiple of a variable

1
(ax) = (x), a 6= 0
|a|

Proof:
Take a > 0. Then,
Z
1 y 
Z
(ax)f (x)dx = (y)f dy
a a
Z
1 1
= f (0) = (x)f (x)dx
|a| |a|
1
(ax) = (x) for a > 0
|a|

For a < 0, take a = c, c > 0. Then,

Z Z
1  y 
Z
(ax)f (x)dx = (cx)f (x)dx = (y)f dy
c c
1  y 
Z
= (y)f dy
c c
Z
1 1
= f (0) = (x)f (x)dx
a |a|
1
(ax) = (x) for a < 0
|a|

3.
1 
(x2 a2 ) =

(x a) + (x + a)
2|a|
Proof: If a 6= 0, the delta function peaks out for two values of x i.e. for
x = a. Thus we have,
1.9. DIRAC DELTA FUNCTION 35

Z Z
(x2 a2 )f (x)dx
 
= (x a)(x + a) f (x)dx

Z a+ Z a Z   
= + + (x a)(x + a) f (x)dx
a+ a
Z a+ Z   
= + (x a)(x + a) f (x)dx
a

Near x = a, (x a)(x + a) 2a(x + a) and similarly, near x = a,

(x a)(x + a) 2a(x a). Integral over the range a + to a gives
no contribution. Hence, using the second property above, we get

Z Z a+ Z  
(x2 a2 )f (x)dx

= + (x a)(x + a) f (x)dx
a
Z a+  
Z  
= 2a(x + a) f (x)dx + (x a)(2a) f (x)dx
a

1 
Z

= (x a) + (x + a) f (x)dx
2|a|
1 
(x2 a2 )

= (x a) + (x + a)
2|a|

  X 1
g(x) = (x xi )
4. xi |g (xi )|
g(xi )=0,

g (xi )6=0

Proof: Let the simple zeros of function g(x) be situated x = xi , i = 1, ...

Only at these points, the delta function has non-zero values. Hence we
can decompose the integral into a sum of integrals over small intervals
containing only one zero in each:

Z Z xi +
  X  
f (x) g(x) dx = f (x) (x xi )g (xi ) dx
xi xi

where we have approximated the function g(x) within the small intervals
36CHAPTER 1. INTRODUCTION AND MATHEMATICAL PRELIMINARIES

as g(x) g(xi ) + (x xi )g (xi ) = (x xi )g (xi ). Thus we have
Z XZ xi +
   
f (x) g(x) dx = f (x) (x xi )g (xi ) dx
xi xi
xi +
XZ 1
= f (x) (x xi )dx
xi xi |g (xi )|
X1
= f (xi )
xi
|g (xi )|
XZ 1
= f (x)(x xi )dx
xi |g (xi )|
Z h X
1 i
= (x xi ) f (x)dx
xi |g (xi )|
g(xi )=0,

g (xi )6=0

Hence
  X 1
g(x) = (x xi )
xi |g (xi )|
g(xi )=0,

g (xi )6=0

Note that, if any of the points xi is a zero of order

 higher
 than one of the
function g(x), then the above expression for g(x) makes no sense.

It may also be noted that the results (1) to (3) above are special cases
of the general result (4) above with g(x) equal to x, ax and x2 a2 ,
respectively.

5. Z
(y x)(x z)dx = (y z)

Proof: Consider the function

Z
F () (y, z) = () (y x) () (x z)dx

with the rectangular function representation of the delta function

1
for y 2 < x < y + 2
()
(y x) = (x y) =
0 for |x y| > 2

and similarly for () (x z).

The function F () (y, z) is zero as long as |y z| , that is the two

rectangles, each of width , centered around x = y and x = z, respectively
1.9. DIRAC DELTA FUNCTION 37

are separated by distance greater than or equal to (the two rectangles

do not overlap). If there is any overlap, the function F () (y, z) will be
non-zero.

The maximum value of the function F () (y, z) occurs when there is maxi-
mum overlap, i.e. for y = z. Then the value of the function is
Z
()
F (y, z) = () (y x) () (x z)dx

/2
1
Z
= 2
dx
/2
1
=

Between this maximum, and the minimum value of 0, the variation of
F () (y, z) with respect to y z, is linear. Hence, we see immediately that
F () (y, z) approaches (y z) as 0.