Vector Analysis

PYL101: Electromagnetics & Quantum Mechanics

Semester-I, 2023-2024

Prof. Rohit Narula1

1 Department of Physics
The Indian Institute of Technology, Delhi

© 2016-2023 Rohit Narula. All rights reserved.

© 2016-2023 Rohit Narula. All rights reserved.
Copyright Notice

Ï My lectures and course materials, including presentations, tests, exams, outlines, and
similar materials, are protected by copyright and for the sole use of students
registered for PYL101: Electromagnetics & Quantum Mechanics in Sem-I 2023-2024
only. I am the exclusive owner of copyright in those materials I create. I encourage
you to take notes and make copies of course materials for your own educational
use. However, you may not, nor may you knowingly allow others to reproduce or
distribute lecture notes, audio, video, screenshots and other course materials
publicly or to internal university repositories such as the BSW without my express
written consent. This includes providing materials to commercial course material
suppliers such as CourseHero and other similar services. Students who publicly
distribute or display or help others publicly distribute or display copies or modified
copies of my course materials may be in violation of applicable law.

© 2016-2023 Rohit Narula. All rights reserved.

Color Codes

Ï Black: regular text.

Ï Red: important concept, emphasis,. . .
Ï Green: optional, but worth thinking about.
Ï Blue: required HW, test your understanding!
Ï Orange: in jest; take with a pinch of salt!

© 2016-2023 Rohit Narula. All rights reserved.

References

Ï Introduction to Electrodynamics, David J. Griffiths [IEDJ]

Ï Chapter I, 1., Vector Analysis

© 2016-2023 Rohit Narula. All rights reserved.

Applications of Electromagnetism

Ï Define electromagnetism.
© 2016-2023 Rohit Narula. All rights reserved.
What is Electromagnetism?

Ï Electromagnetism is a branch of physics which describes the interaction between

charged particles.
Ï Charges come in only two flavors: positive (+), and negative (−).
Ï By interaction we mean the (electromagnetic) forces which the charges exert on
each other.
Ï The EM force is ’carried/mediated’ by electromagnetic fields composed of the
electric fields (E ) and magnetic fields (B ).

© 2016-2023 Rohit Narula. All rights reserved.

Fields

Ï Fields?
Ï Simplified: A field is a physical quantity, represented by a number, or a vector that
has a value for each point in space-time, i.e., (x, y, z, t ).
1. e.g., pressure (scalar field)
2. e.g., wind velocity (as seen on a weather report) (vector field)
3. e.g., the E and B fields of electromagnetism. (vector fields)
Ï More precisely: A field is a physical quantity, represented by a tensor1 (e.g., a
number is a rank-0 tensor, a vector is a rank-1 tensor), that has a value for
each point in space-time, i.e., (x, y, z, t ).
Ï e.g., stress tensor (rank-2 tensor field)

1 def: a tensor is an algebraic object (e.g., vector or scalar or other tensors) that describes a linear
mapping from one set of algebraic objects to another.
© 2016-2023 Rohit Narula. All rights reserved.
How are Electric and Magnetic Fields Produced?

Ï How are the electric and magnetic fields produced?

Ï A stationary charged particle: produces a static electric field.
Ï Similarly, a steady current in a wire (also) produces a static magnetic field.
Ï An accelerating charged particle produces an electromagnetic (EM)
wave/radiation, i.e., a synchronized oscillation of electric and magnetic fields which
have the property that they travel through empty space/vacuum at the speed of
light c . (e.g., light, X-rays)

© 2016-2023 Rohit Narula. All rights reserved.

The Theoretical Basis of Classical Electromagnetism
Ï Only these 4 Maxwell’s equations in the SI unit convention,

∇·D = ρf , (Gauss’ law of electricity)

∇·B = 0, (Gauss’ law of magnetism)

∂B
−∇ × E = , (Faraday’s law of induction)
∂t
∂D
∇×H = Jf + , (Ampere’s law)
∂t
where ρ f is the free charge density and, J f is the free current density.
. . . combined with the Lorentz force law

F = qE + q(v × B )

, form the basis of classical electromagnetism, and optics.

© 2016-2023 Rohit Narula. All rights reserved.

The Theoretical Basis of Classical Electromagnetism

Ï Maxwell’s equations in the SI unit convention, constitute a set of 4 equations for

both [E (D )] and [B ( H )] which are. . .
Ï coupled,
Ï first-order,
Ï partial differential

© 2016-2023 Rohit Narula. All rights reserved.

Vectors vs. Scalars

Figure: [Image from kullabs.com]

Ï Vector (noun): a quantity having a direction as well as a magnitude.

Ï e.g., velocity, acceleration, force and momentum.
Ï Scalar (noun): a quantity that has a magnitude, but no direction.
Ï e.g., mass, charge, density, and temperature.
Ï Vectors have magnitude, and direction, but not location.
Ï Typographically, vectors shall be represented by a bold face, e.g., A .

© 2016-2023 Rohit Narula. All rights reserved.

Vector Operations

Ï We will encounter four kinds of vector operations:

Ï one addition (A + B ),
Ï and three kinds of ’multiplication’.
Ï Multiplication by a scalar (k A ),
Ï Dot product of two vectors ( A · B ),
Ï Cross product of two vectors ( A × B ).

© 2016-2023 Rohit Narula. All rights reserved.

Vector Addition

Ï For graphically representing vector addition, recall the triangle rule (from high
school).
Ï Vector addition is commutative, i.e.,

A +B = B + A
Ï Vector addition is also associative, i.e.,

(A + B ) +C = A + (B +C )

© 2016-2023 Rohit Narula. All rights reserved.

Multiplication by a scalar

Ï Multiplication of a vector by a positive scalar a multiplies the magnitude but leaves

the direction unchanged.
Ï Scalar multiplication is distributive, i.e.,

a(A + B ) = a A + aB

© 2016-2023 Rohit Narula. All rights reserved.

Dot Product of Two Vectors

Ï The dot product of two vectors is defined (≡) by,

A · B ≡| A || B | cos θ
where θ is the angle they form when placed tail-to-tail.
Ï A · B yields a scalar, hence the alternative name scalar product.
Ï The dot product is commutative, i.e.,

A ·B = B · A
Ï The dot product is distributive, i.e.,

A · (B +C ) = A · B + A · C

© 2016-2023 Rohit Narula. All rights reserved.

Dot product of two vectors
Ï Given vectors in component form, i.e.,
X
A = a 1 x̂ 1 + a 2 x̂ 2 + a 3 x̂ 3 + . . . = a i x̂ i
i
X
B = b 1 x̂ 1 + b 2 x̂ 2 + b 3 x̂ 3 + . . . = b i x̂ i
i

Ï Component-wise we can define the dot product as,

X
A · B ≡ a1 b1 + a2 b2 + a3 b3 + a4 b4 + . . . ≡ ai bi
i

Ï Also note that,

A · A =| A |2 = A 2
where A 2 or |A|2 are just shorthands for A · A .
Ï and,
0· A = 0

© 2016-2023 Rohit Narula. All rights reserved.

Orthogonality and Projections
Ï Two vectors A and B are orthogonal2 if and only if,

A ·B = 0

Ï The scalar projection3 of B onto A is defined as,

B·A
P B ,A =
A
Ï The vector projection of B onto A is defined as,
B·A
P B ,A = A
A2

2 The term perpendicular describes a property of two vectors, orthogonal is a related property of any
collection of vectors (i.e., a collection of vectors is orthogonal if and only if all of them are pairwise
perpendicular), and normal is a relation between a vector and an object such as the tangent plane at a point
of a smooth surface.
3 Is P
A,B = P B ,A ?
© 2016-2023 Rohit Narula. All rights reserved.
Cross Product of Two Vectors

Ï The cross product of two vectors is defined by,

A × B ≡| A || B | sin θ n̂

where θ is the angle they form when placed tail-to-tail, and n̂ is a unit vector4
pointing ⊥ to the plane of A and B .
Ï The correct orientation of n̂ is determined by the right-hand rule, e.g., A × B above
points into the page.
4 a hatted n̂ denotes a unit vector
© 2016-2023 Rohit Narula. All rights reserved.
Cross Product of Two Vectors

Ï A × B is itself a vector (hence the alternative name vector product).

Ï The cross product is distributive,

A × (B +C ) = (A × B ) + (A × C )

Ï But the cross product is not commutative,

B × A 6= A × B

and instead,
B × A = −(A × B )
Ï Geometrically, | A × B | is the area of the parallelogram generated by A and B .

© 2016-2023 Rohit Narula. All rights reserved.

Cross Product of Two Vectors
Ï Given component-wise 3d vectors we can conveniently calculate the cross product
via the determinant5 . ¯ ¯
¯ x̂ 1 x̂ 2 x̂ 3 ¯¯
¯
A × B = ¯¯ A 1 A2 A 3 ¯¯
¯B B2 B3 ¯
1

Ï Expand the above determinant and check that it matches with the expression 13,
Chapter I of [IEDJ].
Ï The vector A × B is orthogonal to both A and B .
Ï Two non-zero vectors A and B are parallel/anti-parallel if and only if

A×B =0

5 You must be able to calculate determinants for both 3 × 3, and 2 × 2 matrices.

© 2016-2023 Rohit Narula. All rights reserved.
Scalar Triple Product

Ï The scalar triple product between vectors A , B and C is given by

A · (B × C )
Ï Geometrically, A ·(B × C ) is the volume of the parellopiped generated by A , B and C .
Ï Cyclic order preserves sign, i.e.,

A · (B × C ) = B · (C × A) = C · (A × B )

© 2016-2023 Rohit Narula. All rights reserved.

Scalar Triple Product
Ï While non-cyclic permutations reverse sign, i.e.,

A · (C × B ) = B · (A × C ) = C · (B × A) = −A · (B × C )

Ï In component form, the scalar triple product evaluates to a determinant,

¯ ¯
¯ A1 A2 A 3 ¯¯
¯
A · (B × C ) = ¯¯ B 1 B2 B 3 ¯¯
¯C C2 C3 ¯
1

Ï The dot and cross can be interchanged (keeping the same cyclic order),

A · (B × C ) = (A × B ) · C

Ï Problem: What’s wrong with (A · B ) × C ?

© 2016-2023 Rohit Narula. All rights reserved.

Vector Triple Product
Ï The vector triple product between vectors A , B and C is given by

A × (B × C )

which geometrically amounts to?

Ï . . . "a vector in the plane spanned by B and C , which is also perpendicular to A ".
Ï It can be simplified via the BAC-CAB identity,

A × (B × C ) ≡ B (A · C ) −C (A · B )
Ï Let’s regroup the brackets. Since cross-products are not associative, i.e.,

(A × B ) × C 6= A × (B × C )

it turns out that this regrouping,

(A × B ) × C = −A(B · C ) + B (A · C )

is an entirely different vector6 !

6 Unlike the scalar triple product A · (B × C ) where at worst you’ll be off by a sign.
© 2016-2023 Rohit Narula. All rights reserved.
The Position Vector

Ï The location of a point in three dimensions can be described by listing its Cartesian
coordinates (x, y, z )
Ï The vector 7 to that point from the origin is called the position vector,

r ≡ x x̂ + y ŷ + z ẑ
Ï Its magnitude,
q
r= x2 + y 2 + z2
is simply the distance from the origin O .
7 The position vector is not truly a vector in the rigorous sense since we’d mentioned earlier that vectors
be freely transposed, while the position vector is tethered to the origin. To this extent it’s more aptly a
pseudo-vector.
© 2016-2023 Rohit Narula. All rights reserved.
The Position Vector

Ï The unit vector

r
r̂ =
r
points radially outward.
Ï The infinitesimal displacement vector8 from (x,y,z) to (x+dx,y+dy,z+dz) is,

d l = d x x̂ + d y ŷ + d z ẑ
Ï I misspoke in the lecture. In order to have d l point radially outward we must have
d x : d y : d z = x : y : z , i.e., d x, d y, d z must have the same relative proportions as
x, y, z .
8 d l does not point in any particular direction since d x 6= d y 6= d z , in general.
© 2016-2023 Rohit Narula. All rights reserved.
The Separation Vector s

Ï In electrodynamics, one frequently encounters problems involving two points,

typically, a source point, r 0 , where an electric charge is located, and a field point,
r , at which you are calculating the electric or magnetic field.
Ï The separation vector s from the source point to the field point is then,

s ≡ r − r 0 = (x − x 0 )x̂ + (y − y 0 ) ŷ + (z − z 0 )ẑ

© 2016-2023 Rohit Narula. All rights reserved.

Ordinary Derivative

Figure: www.themathpage.com

Ï Given a function f (x), the ordinary derivative d f /d x represents the rate of change
of f w.r.t x .
Ï Alternatively, it tells us how rapidly f varies when we change x by an infinitesimal
(tiny) fraction d x .
Ï Geometrically, it gives us the slope of the graph of f vs. x .
© 2016-2023 Rohit Narula. All rights reserved.
The Partial Derivative

Figure: www.khanacademy.com

Ï The partial derivative of a function of several variables is its derivative with respect
to only one of those variables, with the others held constant (as opposed to the total
derivative, in which all variables are allowed to vary).
∂f df
Ï Q: Given f (x, y) = x 2 y 3 , compute ∂y , and dy .
∂f df
Ï Ans: ∂y = 3x 2 y 2 and dy = 3x 2 y 2 + 2x y 3 dd xy
© 2016-2023 Rohit Narula. All rights reserved.
The Del ∇ Operator
Ï The del operator is defined as,

∂ ∂ ∂
∇ ≡ x̂ + ŷ + ẑ
∂x ∂y ∂z

Ï Note, ∇ is not a vector! Instead, it’s an operator, which in the grad(ient) case, maps a
differentiable scalar function f to a vector function ∇ f .
Ï Nor does it multiply what appears on the right of it. (∇T ??)
Ï Instead, it’s a vector operator, or an instruction to act on whatever appears to its
right.
Ï It’s really just (very) clever notation9 and acts in 3 ways:
Ï On a scalar function, say T : ∇T (gradient)
Ï On a vector function, say v , via the dot product: ∇ · v (divergence)
Ï On a vector function, say v , via the cross product: ∇ × v (curl)

9 So clever that it allows us to use ∇ "like a vector", but being an operator it alone doesn’t have a meaning!
© 2016-2023 Rohit Narula. All rights reserved.
The Gradient
Ï Given a scalar function, say T (x, y, z), the change in the quantity T when x, y, z are
varied infinitesimally i.e., by (d x, d y, d z)10 is given by the total differential,
d T = T (x + d x, y + d y, z + d z) − T (x, y, z)
∂T ∂T ∂T
=( )d x + ( )d y + ( )d z ≡ (∇T ) · (d l )
∂x ∂y ∂z
where
∂T ∂T ∂T
∇T ≡ x̂ + ŷ + ẑ
∂x ∂y ∂z
is, by definition, the gradient of T , and is evidently a vector function.
Ï Using then the definition of the dot product we can of course also write,
(∇T ) · (d l ) =| ∇T || d l | cos θ

10 What if the changes in x, y, z were larger (i.e., not infinitesimal)? As a hint consider f (x) and expand it
about the point x 0 via a Taylor series as
∂f 1 ∂2 f
f (x) = f (x 0 ) + ∆x + (∆x)2 + . . .
∂x 2! ∂x 2 © 2016-2023 Rohit Narula. All rights reserved.
The Gradient and the Directional Derivative
Ï We may think of the gradient as "a multi-variable generalization of the derivative".
Ï To make this concrete, consider the definition for the directional derivative of a
function f (x, y, z, . . .) along an arbitrary unit vector û as,
f (a + h û) − f (a)
D û f (a) ≡ lim = ∇ f (a) · û
h→0 h | {z }
the gradient of f at a

which is the rate of change of f at the point a in the direction û .

Ï Now, ask yourself ’Along which direction is D û f (a) maximised?’
Ï The directional derivative can be rewritten as,

D û f (a) = ∇ f (a) · û = |∇ f (a)| cos θ

so the largest the D û f (a) can be is when θ = 0, i.e., when û is in the direction of
the gradient ∇ f (a).
Ï Thus the gradient ∇ f (a) points in the direction of the greatest increase of f ,
i.e., the direction of steepest ascent.
© 2016-2023 Rohit Narula. All rights reserved.
Properties of the Gradient

Ï The gradient ∇T , or alternatively, ∇T

at an arbitrary location points in the direction
|∇T |
of maximum increase of the function T around that particular location/point.
Ï The magnitude | ∇T | gives the slope (rate of increase) along this maximal direction.
Ï When the gradient vanishes, i.e., ∇T = 0 at (x i , y i , z i ) then the entire set, {(x i , y i , z i )}
represent the stationary or critical points of the function T (x, y, z).
Ï These stationary/critical points are either:
Ï local maxima (extremum)
Ï local minima (extremum)
Ï saddle/minimax points
Ï If you want to locate the critical points(s) of a scalar function of two (or more)
variables, simply set its gradient to zero.

© 2016-2023 Rohit Narula. All rights reserved.

Exercise: The Gradient Of a Hill

Ï First, assume a very simple hill of elevation,

h(x, y) = −x 2 − y 2 + 100

having a height 100 m. What do the gradient and the contour lines look like?
Ï Second, draw a more arbitrary, realistic hill using contour lines (topological map).
Ï Q:What does the gradient ∇h(x, y) look like?

© 2016-2023 Rohit Narula. All rights reserved.

The Gradient Of a Hill

Ï On a contour/topological map, the gradient points in a direction which minimizes

the distance between adjacent contour lines. Thus the gradient is always ⊥ to a
contour line.
Ï Do all the gradient vectors point exactly toward the global peak?
Ï Ans: No, they do not! The gradient ∇h(x, y) is a local quantity and only gives the
magnitude and the direction of maximum increase around the given point (i.e.,
locally) in question, say (x i , y i ).
Ï Dwell on the mnemonic: "water always flows in the direction opposite the direction
of the gradient."
© 2016-2023 Rohit Narula. All rights reserved.
The Divergence

Ï The divergence of a vector function v in 3d-space is defined as:

∂v x ∂v y ∂v z
∇·v ≡ + +
∂x ∂y ∂z
Ï The divergence ∇ · v is itself a scalar function.
Ï Is ∇ · v defined for each point, or do we have one value per v ?
© 2016-2023 Rohit Narula. All rights reserved.
The Divergence

Ï The diagrams above plot a vector field11 with the (scaled) magnitude and direction
of the vector valued function v at selected points/grid in Cartesian space.
Ï Geometrically, the divergence measures how much the vector v spreads out
(diverges) from the point in question.
11 In vector calculus, a vector field is an assignment of a vector to each point in a subset (grid point)) of
space.
© 2016-2023 Rohit Narula. All rights reserved.
An Intuitive Picture of Divergence

Ï Consider a liquid-flow analogy. If you drop some sawdust12 at a particular location

on the surface and they seem to spread out(in), as opposed to staying stationary, or
simply translate, we have location of positive(negative) divergence.
Ï A point of positive divergence is a source, or faucet; a point of negative divergence
is a sink, or drain.
12 Why sawdust?
© 2016-2023 Rohit Narula. All rights reserved.
The Divergence: be careful visualizing!

Ï One might be lulled into thinking that Fig.(a) represents the E due to a single, static
charge. However, it cannot possibly represent v = k rr̂2 , since the radial vectors above
are getting longer!
Ï Instead, say if Fig.(a) represents v = r r̂ , then it has a constant divergence13
everywhere, i.e., ∇ · v = 3.
Ï You might think that the vector function given by Fig.(c) has a zero divergence
everywhere. But that would be incorrect, since the arrows/vectors, starting from the
bottom are getting longer.
13 In spherical coordinates ∇ · v = 1 ∂ (r 2 v ) + . . .
r 2 ∂r r
© 2016-2023 Rohit Narula. All rights reserved.
The Curl

Ï The curl of a vector function v is defined as,

¯ ¯
¯ x̂ ŷ ẑ ¯¯
¯
¯∂ ∂ ∂ ¯
∇ × v = ¯ ∂x ∂y ∂z ¯¯ (vector function)
¯
¯v x v y v z ¯

Ï Geometrically, it measures of how much v swirls around the point in question.

Ï Intuitively, imagine standing at the edge of a pond. Float a small and light
paddlewheel; if it starts to rotate at that location, then you’ve placed it at a point of
non-zero curl.
© 2016-2023 Rohit Narula. All rights reserved.
The Curl

Ï The three vector fields above all have zero curl.

© 2016-2023 Rohit Narula. All rights reserved.

The Curl

Ï Whereas the functions above have a substantial curl, pointing in the ẑ direction, as
the natural right-hand rule would suggest.

© 2016-2023 Rohit Narula. All rights reserved.

Addition and Scalar Multiplication Rules

Ï Just like with ordinary calculus, certain mathematical rules hold for vector
derivatives,

Ï The above rules are fortunately quite intuitive!

© 2016-2023 Rohit Narula. All rights reserved.

Product Rules
Ï While the product rules get a little more complicated. . . 14

14 Proving these rules once is advised.

© 2016-2023 Rohit Narula. All rights reserved.
Second Derivatives

Ï By applying ∇ twice, we can construct five species of second derivatives

© 2016-2023 Rohit Narula. All rights reserved.

(1) The Divergence of a Gradient aka The Laplacian

Ï When given a scalar function T , the Laplacian is defined as,

2 ∂2 T ∂2 T ∂2 T
∇ T
|{z} ≡ ∇ · (∇T ) = + +
∂x 2 ∂y 2 ∂z 2
just shorthand

Ï Note, the Laplacian of a scalar function T is a scalar function!

Ï On the other hand, given a vector function v , the Laplacian is then defined as,

∇2 v
|{z} ≡ (∇2 v x )x̂ + (∇2 v y ) ŷ + (∇2 v z )ẑ
just shorthand
6≡ ∇ · (∇v )

and is a vector function.

© 2016-2023 Rohit Narula. All rights reserved.

(2) The curl of a gradient

Ï . . . is always zero. i.e.,15

∇ × (∇T )= 0

15 Do the proof!
© 2016-2023 Rohit Narula. All rights reserved.
(3) Gradient of a Divergence

Ï
∇(∇ · v )
. . . doesn’t show up much in the study of electromagnetism so we won’t bother
about it.

© 2016-2023 Rohit Narula. All rights reserved.

(4) The divergence of a curl

Ï . . . is also always zero, i.e.,16

∇ · (∇ × v )= 0

16 Do the proof!
© 2016-2023 Rohit Narula. All rights reserved.
(5) The curl of a curl

Ï
2
∇ × (∇ × v ) = ∇(∇ · v ) − ∇ v
| {z } |{z}
(3) (1)

is just the gradient of a divergence minus the Laplacian of a vector, or (3) − (1).

© 2016-2023 Rohit Narula. All rights reserved.

Line Integrals

Ï A line integral is an expression of the form

Zb
v |{z}· dl
a
dot product

where v is a vector function, and d l is the infinitesimal displacement vector.

Ï The integral must be carried out along a prescribed path P from point a to point
b , and we should instead write Z b
v · dl
a P
explicitly mentioning the path P .
© 2016-2023 Rohit Narula. All rights reserved.
Line Integrals

Ï If the path P is a closed loop (i.e., a = b ) we put a circle around the integral sign
as, I
v · d l 6= 0
P

where it’s important to note that, in general, closed line integrals don’t have to be
zero, in general!

© 2016-2023 Rohit Narula. All rights reserved.

A Line Integral Example

Ï Problem: Calculate the line integral for the vector function

v = y 2 x̂ + 2x(y + 1) ŷ

from the point a = (1, 1, 0) to b = (2, 2, 0), along the paths (1) and (2) in the above
figure.
© 2016-2023 Rohit Narula. All rights reserved.
A Line Integral Example: Along path (1)

Ï Along the horizontal segment of (1) (only x varies) dy = dz = 0, so

Z Z2
2 2
d l = d x x̂, y = 1, v · d l = y d x, v · d l = (1) dx = 1
1

© 2016-2023 Rohit Narula. All rights reserved.

A Line Integral Example: Along path (1)

Ï For the vertical segment of (1) (only y varies) d x = d z = 0, so

Z Z2
d l = d y ŷ, x = 2, v · d l = 2x(y + 1)d y, v · d l = 4 (y + 1)d y = 10
1
Ï Thus, summing the horizontal and vertical parts of (1) we get,
Z
v · d l = 1 + 10 = 11
(1)
© 2016-2023 Rohit Narula. All rights reserved.
A Line Integral Example: Along path (2)

Ï For path (2), x = y ⇒ d x = d y 17 and, d z = 0

d l = d x x̂ + d y ŷ, v · d l = x 2 d x + 2x(x + 1)d x = (3x 2 + 2x)d x,

, and finally, Z
v · d l = 10
(2)

17 This step is crucial for a general path where variables are interdependent.
© 2016-2023 Rohit Narula. All rights reserved.
A Line Integral Example: a → (1) → (2) → a

Ï For the circuitous path a → (1) → (2) → a , we simply sum the contributions of paths
(1) and (2) and get, I
v · d l = 11 − 10 = 1

© 2016-2023 Rohit Narula. All rights reserved.

Surface/Double and Flux Integrals

Ï A surface/double integral is an expression of the form

Z
f da
S
where f is a scalar function, d a is an infinitesimal patch of area over a prescribed surface
S.
Ï However, the flux18 of a vector function v through S is defined as,
Z
·
v |{z} da
S
dot product

18 in analogy with liquid flow, i.e., if v describes the flow a mass of liquid per unit area per unit time.
© 2016-2023 Rohit Narula. All rights reserved.
Flux Integrals

Ï Since there are two opposite directions for the surface normal, we choose the
direction that points radially outward with the origin as a reference point.
Ï If the surface is closed19 we write,
I
v · d a 6= 0
S

19 A closed surface in 3 − d is exemplified by a balloon.

© 2016-2023 Rohit Narula. All rights reserved.
A Flux Example

Ï Problem: Calculate the flux

v = 2xz x̂ + (x + 2) ŷ + y(z 2 − 3)ẑ

over five sides (excluding the bottom).

© 2016-2023 Rohit Narula. All rights reserved.
A Flux Example: Surface 1

Ï Note, that for surface (i ) variables x and y are independent of each other20 , and the
integrals for each can be carried out independently.
20 Why?
© 2016-2023 Rohit Narula. All rights reserved.
A Flux Example: Surface 2

© 2016-2023 Rohit Narula. All rights reserved.

A Flux Example: Surface 3

© 2016-2023 Rohit Narula. All rights reserved.

A Flux Example: Surface 4

© 2016-2023 Rohit Narula. All rights reserved.

A Flux Example: Surface 5

© 2016-2023 Rohit Narula. All rights reserved.

A Flux Example: Total

Ï Note that the surface S over which the flux is calculated need not be closed.

© 2016-2023 Rohit Narula. All rights reserved.

Evaluating a Surface Integral over a triangular S

Ï Problem: Evaluate Ï
x y d xd y
S
over the triangular region with ⊥ sides of unity, as above.
Ï Strategy:
1. Calculate T (x, y = const .) × the area of each infinitesimally short horizontal strip of
Î R1 Rx=1−y
height d y [INNER SUM: x : 0 → (1 − y) ( S → y=0 x=0 )], and then,
Î R1 Rx=1−y
2. Sum these horizontal strips up one by one [OUTER SUM: y : 0 → 1 ( S → y=0 x=0 )].
© 2016-2023 Rohit Narula. All rights reserved.
Evaluating the Surface Integral over a triangular S

Ï Thus, keeping y constant, let’s focus now on the [INNER SUM: x : 0 → (1 − y)], i.e.,
Zx=1−y
y(1 − y)2
H (y) = xy dx =
x=0 |{z} 2
y=const.

where the (1 − y) limit accounts for the fact that the length of these strips varies.
Ï Finally, we perform the [OUTER SUM: y : 0 → 1] and
Z1 Z1
y(1 − y)2 1
H (y) d y = dy =
y=0 y=0 2 24
R
Ï HW: Find S x y d xd y by instead summing vertical strips.
© 2016-2023 Rohit Narula. All rights reserved.
Volume Integrals
Ï Given a scalar function T , volume integrals are expressed as,
Z
T dτ
T

and d τ is an infinitesimal volume element.

Ï For a vector function v , on the other hand, the associated volume integral is,
Z Z
v d τ = (v x x̂ + v y ŷ + v z ẑ)d τ
Z Z Z
= x̂ v x d τ + ŷ v y d τ + ẑ v z d τ

Ï In Cartesian coordinates,21
d τ = d xd yd z

21 Spherical and cylindrical coordinates are other candidates.

© 2016-2023 Rohit Narula. All rights reserved.
A Volume Integral Example

Ï Problem: Calculate the volume integral of T = x y z 2 .

Ï It’s easy to see from the prismatic volume that z does not depend on either x or y ,
and ranges from 0 to 3.
Ï This means that the integral over z can be factored out.
© 2016-2023 Rohit Narula. All rights reserved.
A Volume Integral Example

Ï
Z Z3 Ï
2 2
x yz dτ = z dz x y d xd y
0 | {z }
done earlier!
Z1 ·Z1−y ¸
=9 y xd x d y
y=0 x=0
Z
9 1 3
= y(1 − y)2 d y =
2 0 8
© 2016-2023 Rohit Narula. All rights reserved.
The Fundamental Theorem for Gradients

Ï Given a scalar function T (x, y, z), by changing x , y and z infinitesimally, i.e., by d x , d y and
d z , the variation in T is,
d T = (∇T ) · d l

Ï If we keep advancing from a to b along units of d l i we can accumulate the total change in
the scalar function T , Zb
(∇T ) · d l = T (b) − T (a)
a

also known as the fundamental theorem for gradients22 .

22 There is also a direct proof using the integral in the LHS above.
© 2016-2023 Rohit Narula. All rights reserved.
The Fundamental Theorem for Gradients

Ï Remarkably, the RHS, i.e., T (b) − T (a) makes no reference to the actual path taken.
R
Ï It implies that, ab (∇T ) · d l is independent of the path taken from a to b .23
Ï It also implies that,
I
(∇T ) · d l = 0

23 In practice, even though the integral is independent of the path, we must pick a specific (if convenient)
route in order to evaluate it explicitly.
© 2016-2023 Rohit Narula. All rights reserved.
The Fundamental Theorem for Divergences

Ï The fundamental theorem for divergences states that:

Z I
(∇ · v )d τ = v ·da
V S

Ï It is alternatively known as, Gauss’ theorem, Green’s theorem, or simply the

divergence theorem.
Ï Physical Analogy: Imagine we’re interested in measuring the flux (RHS) of an
incompressible fluid through a closed area S .
Ï The divergence theorem states that instead of measuring the flux directly, we
could’ve equivalently, sum up all the (liquid) sources inside the volume V enclosed
by the surface S .

© 2016-2023 Rohit Narula. All rights reserved.

The Fundamental Theorem for Divergences

Ï Q: Verify the divergence theorem for v = 2x x̂ + y 2 ŷ + z 2 ẑ for the sphere of radius

8π
unity. [Ans: LHS = RHS = 3 ]

© 2016-2023 Rohit Narula. All rights reserved.

The Fundamental Theorem for Curls

Ï Also known as Stoke’s theorem states that,

Z I
(∇ × v ) · d a = v · dl
S P

which says that the integral of a curl (of a vector function) over a surface S is equal
to the value of the function integrated over the boundary P ) enclosing that surface.
Ï Sticking with the liquid analogy, since the curl measures the twist of v , if we are
interested in the total swirl, we can equivalently just measure how much the flow v
follows the closed boundary P enclosing the surface S .
Ï As earlier, by convention we select the orientation of d a pointing outward, and the
sense of the line d l to be anti-clockwise.
© 2016-2023 Rohit Narula. All rights reserved.
The Fundamental Theorem for Curls

R
Ï As a consequence of the theorem, S (∇ × v ) · d a depends only on the boundary line
P , but not on the particular surface S used, as long as it’s circumscribed by P .
Think of an intact soap bubble across a fixed loop. It doesn’t matter whether the
bubble is convex, concave, or combinations thereof, as long as the loop
circumscribing it is fixed.
Ï As a consequence we may deform S for mathematical convenience as long as it
satisfies the boundary P .
H
Ï S (∇ × v ) · d a = 0 for any closed surface, since the boundary line, like the mouth of
a balloon, shrinks down to a point.
Ï HW: Do example 1.11 of [IEDJ].

© 2016-2023 Rohit Narula. All rights reserved.

Spherical Coordinates

Ï An alternative to using the usual Cartesian coordinates described by (x, y, z) is to

use spherical coordinates described by (r, θ, ϕ) where,
1. r is the distance from the origin (the magnitude of the position vector r )
2. θ (the angle down from the z axis) is called the polar angle
3. ϕ (the angle around from the x axis) is the azimuthal angle.
© 2016-2023 Rohit Narula. All rights reserved.
Spherical Coordinates

Ï From the above figure we observe that,

x = r sin θ cos ϕ
y = r sin θ sin ϕ
z = r cos θ
Ï A general vector A can be represented in spherical coordinates as,
A = A r r̂ + A θ θ̂ + A ϕ ϕ̂
where r̂ , θ̂ , and ϕ̂ form an orthogonal basis set.
© 2016-2023 Rohit Narula. All rights reserved.
Spherical Coordinates

Ï In terms of Cartesian unit vectors we can write,24 ,

r̂ = sin θ cos ϕx̂ + sin θ sin ϕ ŷ + cos θ ẑ
θ̂ = cos θ cos ϕx̂ + cos θ sin ϕ ŷ − sin θ ẑ
ϕ̂ = − sin ϕx̂ + cos ϕ ŷ
24 Do the proof.
© 2016-2023 Rohit Narula. All rights reserved.
The inifinitesimal elements in spherical coordinates

Ï Beware that in the spherical coordinate system,

d l 6=d r rˆ + d θ θ̂ + d ϕϕ̂
d τ6=d r d θd ϕ
Ï An infinitesimal displacement in the r̂ direction is simply d r , and thus,
d lr = d r
Ï An infinitesimal element of length in the θ̂ direction is,
d lθ = r d θ
Ï An infinitesimal element of length in the ϕ̂ direction is,
d l ϕ = r sin θd ϕ
© 2016-2023 Rohit Narula. All rights reserved.
The inifinitesimal elements in spherical coordinates

Ï Thus the general infinitesimal displacement d l is,

d l = d r r̂ + r d θ θ̂ + r sin θd ϕϕ̂

Ï The infinitesimal volume element d τ, in spherical coordinates, is the product of

the three infinitesimal displacements:

d τ = d l r d l θ d l ϕ = r 2 sin θd r d θd ϕ

© 2016-2023 Rohit Narula. All rights reserved.

The surface element in spherical coordinates

Ï Suppose you’re integrating over the surface of a sphere of radius r , here,

d a 1 = d l θ d l ϕ r̂ = r 2 sin θd θd ϕr̂

Ï On the other hand, if the surface lies in the x y plane, as above25

d a 2 = d l r d l ϕ θ̂ = r d r d ϕ θ̂

25 What happened to the sin θ term in the RHS of d a , i.e., d l = r sin θd ϕ?

2 ϕ
© 2016-2023 Rohit Narula. All rights reserved.
More Formulae for Spherical Coordinates

Ï Things get complicated26 . . .

26 The above formulae will be provided, if required, during the exam, and needn’t be memorized.
© 2016-2023 Rohit Narula. All rights reserved.
Cylindrical Coordinates

Ï Do as HW.

© 2016-2023 Rohit Narula. All rights reserved.

Need for the Dirac Delta Function
Ï Given a body of mass M located at r 0 , we may express it in terms of its mass
density ρ(r ) as, Z
M = ρ(r )d τ
V
Ï But what does the mass density of a point mass located at r 0 look like?
Ï It can be unequal to zero only at a single point, i.e.,

ρ(r ) = 0 ∀r 6= r 0

Ï The volume integral, however,

Z
ρ(r )d τ = M (finite)
V

is finite provided r 0 lies within the volume V .

© 2016-2023 Rohit Narula. All rights reserved.

Need for the Dirac Delta Function

Ï It was a long time coming.

Ï Since there was no function that encoded such a property, Paul A. M. Dirac
invented one, writing ρ(r ) as,

ρ(r ) = M × δ(r − r 0 )

, and requiring that,

Z ½
1 if r 0 ∈ V
δ(r − r 0 )d τ =
V 0 otherwise
δ(r − r 0 ) = 0 ∀r 6= r 0

© 2016-2023 Rohit Narula. All rights reserved.

The One-Dimensional Dirac Delta Function

Ï The one-dimensional Dirac delta function27 , δ(x − a), can be pictured as an infinitely
high, but infinitesimally narrow spike, with area 1, located at x = a , i.e.,
(
0 x 6= a
δ(x − a) ≡
∞ x=a
and, Z∞
δ(x − a)d x = 1
−∞
27 If x has units of length, what’s the unit of δ(x)?
© 2016-2023 Rohit Narula. All rights reserved.
Properties of the One-Dimensional Dirac Delta Function
Ï Technically, δ(x) is not a function at all, since its value is not finite at x = 0.
Ï It’s even, i.e., δ(x) = δ(−x).
Ï An important characteristic of the Dirac delta function is its sifting property28 ,
Z∞ Z∞
f (x)δ(x − x 0 )d x = f (x 0 ) δ(x − x 0 )d x = f (x 0 )
−∞ −∞

where, loosely speaking, the delta function picks out the value of f (x) at x = x 0 , i.e.,
f (x 0 ).
Ï Another curious property of the Dirac delta function is,

1
δ(kx) = δ(x)
|k |

where k is any (non-zero) constant

28 Prove the sifting property!
© 2016-2023 Rohit Narula. All rights reserved.
The One-Dimensional Dirac Delta Function

Ï If f (x) is differentiable,
Z∞
f (x)δ0 (x − x 0 )d x = − f 0 (x 0 )
−∞

Ï The δ(x) may also be seen as the derivative of the Heaviside step function29 ,

d
δ(x − a) = Θ(x − a)
dx

29 This property might be invoked while discussing square potenial barriers in the QM part of this course.
© 2016-2023 Rohit Narula. All rights reserved.
The Three-Dimensional Dirac Delta Function

Ï In three dimensions,

δ3 (r − r 0 ) ≡ δ(x − x 0 )δ(y − y 0 )δ(z − z 0 )

| {z }
shorthand

where r 0 = x 0 x̂ + y 0 ŷ + z 0 ẑ is the location where δ3 (r − r 0 ) is centered. and30 ,

Z
δ3 (r − r 0 )d τ = 1
V

Ï Also, Z
f (r )δ3 (r − r 0 )d τ = f (r 0 )
V

30 All you need to ensure is that r lies within the volume V .

0
© 2016-2023 Rohit Narula. All rights reserved.
Divergence of r̂ /r 2

Ï Consider the vector function31 ,

1
v= r̂
r2
Ï From the figure, evidently v has a LARGE positive divergence at the center,
Ï . . . and yet. . . (do the math)
Ï
1 ∂ 2 1 ∂
∇·v = (r v r ) = 2 (1) = 0 everywhere!?
r ∂r
2 r ∂r
which is certainly not what we were expecting!
31 Which is eerily reminiscent of E (r ) = 1
r̂ , the electric field due to a single, static point charge.
4πϵ0 r 2
© 2016-2023 Rohit Narula. All rights reserved.
Divergence of r̂ /r 2

Ï However, when we consider the divergence theorem, i.e.,

Z I
(∇ · v )d τ = v ·da
V S
the RHS is. . .
Ï ..., I Zµ ¶
1
v ·da = r̂ · (r 2 sin θd θd ϕr̂ ) = 4π
S r2
© 2016-2023 Rohit Narula. All rights reserved.
Divergence of r̂ /r 2

R
Ï But we’d just (albeit näively) found that the LHS, i.e., V (∇ · v )d τ = 0, which
contradicts the divergence theorem!
R
Ï However, V (∇ · v )d τ = 0 is incorrect –the source of the problem being the point
r = 0, where v = ∞, i.e., it BLOWS UP.
Ï Indeed, ∇ · v is actually zero everywhere except the origin.32
32 Beware! The intuition of observing the spreading of sawdust at any of the non-central points seems to
suggest a non-zero diverence, at least to my eye.
© 2016-2023 Rohit Narula. All rights reserved.
Divergence of r̂ /r 2

Ï We resolve this paradox by realizing that the volume integral of ∇ · (r̂ /r 2 ) must yield
a constant 4π. (Why because the RHS of the divergence theorem just said so.)
Ï Thus, we can write using the definition of the Dirac delta function,

∇ · (r̂ /r 2 ) = 4πδ3 (r )

Ï Alternatively, 1 2
4π ∇ · (r̂ /r ) is a concrete representation of the Dirac delta function
δ3 (r ).
© 2016-2023 Rohit Narula. All rights reserved.
Proof of the Sifting Property
Ï Q: Prove the sifting property of the Dirac delta function, i.e.,
Z∞
f (x)δ(x)d x = f (0)
−∞

Ï We need to consider a concrete representation of the Dirac delta function such as

the top-hat function T (x) defined by
(
1
|x| ≤ ϵ
T (x) ≡ 2ϵ
0 |x| > ϵ

which in the limϵ→0 goes into the Dirac delta function. Note that the area under
curve of the top-hat function is 1.
Ï In other words,
δ(x) = lim T (x)
ϵ→0

© 2016-2023 Rohit Narula. All rights reserved.

Proof of the Sifting Property

Ï Consider Z∞ Z
1 ϵ
f (x)T (x)d x = f (x)d x
−∞ 2ϵ −ϵ
F (ϵ) − F (−ϵ)
=
2ϵ
where F (x) is the anti-derivative of f (x).
Ï Taking the limϵ→0 we get
F (ϵ) − F (−ϵ)
lim = f (0)
ϵ→0 2ϵ

© 2016-2023 Rohit Narula. All rights reserved.

A Nice Proof of the Sifting Property

Ï Recognize that δ(x) is only appreciable over a tiny region x ≤ |ϵ|, and thus
Z∞ Zϵ
f (x)δ(x)d x = f (x)δ(x)d x
−∞ −ϵ

Ï Since the interval x ≤ |ϵ| is tiny, and the function f (x) is well-behaved, we might as well take
it nearly constant in the extremely narrow interval around x = 0, Ri.e., f (0). Thus, we may take
∞
it out of the integral and use the defining property of δ(x) that −∞ δ(x)d x = 1
Zϵ Zϵ
f (x)δ(x)d x = f (0) δ(x)d x = f (0)
−ϵ −ϵ

© 2016-2023 Rohit Narula. All rights reserved.

Alternative Representations of δ(x)
optional

Ï There are several alternative representations of the Dirac delta function.

Ï Think about whether si n(x)
x is a candidate.
Ï However, note that limx→0 si n(x)
x = 1 and thus falls short of at least one of the
requirements of the Dirac delta function δ(x), i.e., its value must be ∞ at x = 0.
si n( xϵ )
Ï However, your guess wasn’t far away since limϵ→0 π1 x is a valid representation of
δ(x).
Ï Another commonly suspected candidate is 1
.
x2
Ï Now while it does blow up at x = 0, and is clearly even, it falls short of δ(x) since
isn’t zero at all points apart from x = 0.
Ï Again you were close, and instead the function limϵ→0 2 ϵ 2 is a valid representation
x +ϵ
of δ(x)

© 2016-2023 Rohit Narula. All rights reserved.

The Scalar Potential V
Ï When ∇ × E = 0 everywhere33 , Stokes’ theorem tells us that over any & every
possible closed path, I
E ·dl = 0

Ï On the other hand, the converse of the fundamental theorem of gradients34

allows us to write, I
−∇V · d l = 0

Ï Thus, a static E can be written as35 the gradient of a scalar potential V ,

E = −∇(V +const.)
Ï The potential is not unique, i.e., any constant can be added to V without affecting
its gradient, i.e., the (negative) electric field −E .
33 ∇ × E = 0 is guaranteed only in the electrostatic regime.
34 Converse of the fundamental theorem of gradients: "If the integral of v over every closed loop in the
domain of v is zero, then v is the gradient of some scalar-valued function, say V ."
35 The negative sign is purely a matter of convention.
© 2016-2023 Rohit Narula. All rights reserved.
Curl-less (or irrotational) fields v i r r

Ï The condition ∇ × E = 0 everywhere is equivalent to,

Rb
Ï a E · d l is independent of path, for any given end points, a consequence of the
fundamental
H theorem of gradients.
Ï E · d l = 0.
Ï E is the gradient of some scalar function: E = −∇V
Ï Irrotational fields v i r r , i.e., ∇ × v i r r = 0, are also known as conservative fields.36
Ï We’ll soon prove in the chapter on Electrostatics that the work done in a moving a
test charge against a background of static charges is independent of the path taken.

36 In mechanics, a conservative force F = −∇U is a force with the property that the total work done
R
W = ab F · d l in moving a particle between point a to b is independent of the path taken.
© 2016-2023 Rohit Narula. All rights reserved.
The Vector Potential A

Ï Maxwell’s equations guarantee that

∇·B = 0

Ï This allows B to be written as the curl of a vector potential A ,

B = ∇ × (A+∇T )

since mathematically (4) the divergence of a curl is always zero, i.e., ∇ · (∇ × v ) = 0

Ï The vector potential A is not unique, i.e., the gradient of any scalar function +∇T
can be added to A withough affecting its curl, i.e., the magnetic field B , since
mathematically (2) the curl of a gradient is always zero, i.e., ∇ × (∇T ) = 0.

© 2016-2023 Rohit Narula. All rights reserved.

Divergence-less (or solenoidal) fields

Ï The condition ∇ · B = 0 everywhere is equivalent to,

H
Ï B · d a = 0 for any closed surface, a direct consequence of the divergence theorem.
Ï B is the curl of some vector function: B = ∇ × A
R
Ï S B · d a is independent of the details of the precise open surface S (P ), once its
periphery P is set.37

37 This means we can deform the surface as we please as long as the boundary is kept fixed.
© 2016-2023 Rohit Narula. All rights reserved.
The Helmholtz Theorem
Ï "A well-behaved (i.e., goes to zero at infinity) vector field, say F , is uniquely specified by its
divergence and curl (and, in the case of a finite region, additionally by its normal component
over the entire boundary.)"
Ï i.e., suppose we know over all space.38

∇ · F (r ) = D(r ) and, ∇ × F (r ) = C (r )

Ï Then the unique vector field F is given by

F (r ) = −∇U (r ) + ∇ × W (r ) Helmholtz Decomposition

where,
D(r 0 ) C (r 0 )
z }| { z }| {
Z Z
1 ∇ · F (r 0 )
0 1 ∇ × F (r 0 )
0
U (r ) = d3r 0 and, W (r ) = d3r 0
4π all space |r − r 0 | 4π all space |r − r 0 |

where the ∇0 denotes that the derivatives are to be taken w.r.t. source points r 0 .
38 Note that since (4): ∇ · (∇ × v ) = 0, we must have ∇ · C = 0 for consistency.
© 2016-2023 Rohit Narula. All rights reserved.
Why’s Helmholtz Theorem Useful?
Ï Since a vector field is completely specified once its divergence and curl are known (a
purely mathematical result), and we know that the study of electromagnetism
involves the vector fields E and B , we can already guess. . .
Ï . . . the laws of electromagnetism:

∇·E = something1
∇×E = something2
∇·B = something3
∇×B = something4
which look exactly like Maxwell’s equations39 , which are thus mathematically
sufficient to reconstruct E , and B .
Ï Even further, we can almost solve for the fields E and B via the Helmholtz
decomposition, even without explicitly knowing what the R H S = somethingi are!
39 If you’re wondering where D and H went, here we’re discussing the so-called microscopic
representation of Maxwell’s equations. D and H appear in the macroscopic formulation where the material
medium is built into the equations. Both formulations are equally general.
© 2016-2023 Rohit Narula. All rights reserved.
Why’s Helmholtz Theorem Useful?
Ï Considering the static40 version of Maxwell’s equations

ρ(r )
∇·E =
ϵ0
∇×E =0
∇·B =0
∇×B = µ0 j (r )

we’re gratified by noting that a given set of stationary charges ρ(r ), and steady currents
j (r ), may only generate one possible steady E , and one possible steady B field.
Ï Similarly, if all the sources ρ , and currents j are zero everywhere, then the only physical
solution is E = B = 0.
Ï This implies that static electric and magnetic fields cannot generate themselves, i.e., there
must be stationary charges and steady currents generating them!

40 i.e., all charges are stationary, and currents steady.

© 2016-2023 Rohit Narula. All rights reserved.
Limitations of our Treatment of the Helmholtz Theorem

Ï Our discussion for the Helmholtz theorem only works for time-independent sources, and
currents.
Ï For the time-dependent case, while the divergence and curl still uniquely identify the vector
field, the Helmholtz decomposition looks a bit different, though I will not be writing it
down explicitly.
Ï A second issue is that we haven’t grappled with the application of the boundary conditions
(i.e., the normal component of the field must be known at all points of the periphery) that
need to be imposed if we’re considering a finite region of space.

© 2016-2023 Rohit Narula. All rights reserved.