PHYS504: Mathematical Physics I

Course Notes
Duncan John Mowbray, BMath, BSc, MSc, PhD

Abstract
Vector and multivariate calculus of lines and surfaces in R3 , scalar and
vector fields, the “∇” operator, physical interpretation of grad, div, and
curl, surface and line integrals, curvilinear coordinates, Green’s, Stokes’,
and Gauss’ theorems. Optimization and Lagrange multipliers. Applications
include conservation laws, electromagnetic fields, heat flow, and the wave
equation. Fourier series, the Fourier transform, Parseval’s formula, separa-
tion of variables, Legendre, Bessel, and Green’s functions are also discussed.

1
Contents
1 Curves in R3 6

2 Surfaces in R3 11

3 Scalar & Vector Fields and the “∇” Operator 15

4 Physical Interpretation of Grad, Div, and Curl 24

5 Line Integrals 27

6 Green’s Theorem 33

7 Surface Integrals 39

8 Orientable Surfaces 42

9 Maxwell’s Equations 52

10 The Heat Equation 60

11 Fourier Transform 72

12 Sturm-Liouville Theory 74

13 Green’s Functions 78

14 Solutions to the Exercises 80

1 Curves in R3 80

2 Surfaces in R3 86

3 Scalar & Vector Fields and the “∇” Operator 90

4 Physical Interpretation of Grad, Div, and Curl 95

5 Line Integrals 99

6 Green’s Theorem 101

2
7 Surface Integrals 103

8 Orientable Surfaces 105

3
Methodologies for Learning Mathematics
• There are no stupid questions, only stupid answers.

• We learn by making mistakes, identifying them, and correcting them.

Deductive Learning Inductive Learning

Mathematics is a series of defini- Mathematics is a set of tools and for-
tions, theorems, and corollaries from mulas which we use to solve physics
which physically meaningful results problems.
can be derived.
I prefer to learn by deducing results I prefer to learn by solving exam-
based on theorems we have proven. ples, and plugging expressions into
formulas we are provided.
I prefer to test the hypotheses to I prefer to use simple examples to
check a result. determine if a result makes sense.
I would prove a theorem by deducing I would prove a theorem by testing it
the result based on the hypotheses. for many different examples.
The mathematical elegance and gen- The usefulness of a theorem to solve
erality of theorems is what is impor- relevant problems is what is impor-
tant to me. tant to me.

4
Discovery of the Vector Integral Theorems
Cauchy’s Integral Theorem 1825 Green’s Theorem 1828

Augustin-Louis Cauchy (1789–1857) George Green (1793–1841)

Gauss’ Divergence Theorem 1831 to 1839 Stokes’ Theorem 1854

Carl Friedrich Gauss (1777–1855) Sir George Gabriel Stokes (1819–1903)

5
1 Curves in R3
Definition 1.1 (Parametric Representation). A curve C in R3 can be written para-
metrically with parameter t as C : x = x(t), y = y(t), z = z(t), t0 ≤ t ≤ t1 .
Definition 1.2 (Non-parametric Representation). A curve C in R3 can be written
non-parametrically with x as parameter as C : y = y(x), z = z(x), x0 ≤ x ≤ x1 .
Definition 1.3 (Vector Form). A curve C in R3 can be written in vector form as
r = r(t), t0 ≤ t ≤ t1 where r = x(t)î + y(t)ĵ + z(t)k̂ is the position vector of
point P (x, y, z) of the parameter t.
Exercise 1.1. Plot r = (α cos t, α sin t, βt), α ≥ 0, β constants for 0 ≤ t ≤ 4π.
Exercise 1.2. Plot r = (αt cos t, αt sin t, βt), α ≥ 0, β constants for 0 ≤ t ≤ 4π.
Exercise 1.3. Plot r = (t, t2 , t3 ) for 0 ≤ t ≤ 2.
dr
Question. Can we define the derivative dt
of a vector r(t) with respect to the
scalar parameter t?
Definition 1.4 (Vectorial Derivative).
dr r(t + ∆t) − r(t) dx dy dz
≡ ṙ(t) ≡ lim = î + ĵ + k̂.
dt ∆t→0 ∆t dt dt dt
Definition 1.5 (Velocity). If t is time, v(t) ≡ ṙ is the velocity vector along the
curve C.
Definition 1.6 (Speed). If t is time, v ≡ |ṙ| is the speed along the curve C.
Definition 1.7 (Arclength). The
´ arclength
´ element ds ≡ |ṙ|dt on C in terms of t,
and the arclength s = s(t) = ds = |ṙ|dt as a function of t along C.
Note. In general, t is not arclength. But if |ṙ| = 1 on C, then s ≡ t + k, so s and
t are equal up to a constant of integration.
Definition 1.8 (Tangent Vector). The unit tangent vector along C is defined by
ṙ
T̂(t) ≡
|ṙ|
Lemma 1.1. The unit tangent vector along C satisfies
dr
T̂ =
ds
6
Proof.  
ṙ dr dr dr ds ds dr
T̂ ≡ = = =
|ṙ| dt dt ds dt dt ds
where we have used the Chain Rule, and the definition of arclength.

Lemma 1.2. Any unit vector T̂ is perpendicular to its vectorial derivative

dT̂
⊥ T̂.
ds

Proof. Since T̂ is a unit vector, T̂ · T̂ ≡ 1. Taking the derivative with respect to

arclength s:
d  
T̂ · T̂ = 0
ds
dT̂ dT̂
· T̂ + T̂ · =0
ds ds
dT̂
2 · T̂ = 0
ds

dT̂
⇒ ⊥ T̂.
ds

Definition 1.9 (Curvature).

dT̂
κ(s) ≡
ds
is the curvature scalar and
dT̂
κ(s) ≡
ds
is the curvature vector.

Definition 1.10 (Radius of Curvature).

1
≡ ρ(s)
κ(s)

is the radius of curvature of C.

7
Lemma 1.3.
d2 r
κ(s) = .
ds2
Proof.
dT̂
κ(s) ≡ by definition of κ
ds
dr
T̂ ≡ by definition of T̂
ds
d dr d2 r
⇒ κ(s) = = 2.
ds ds ds

Definition 1.11 (Principal Normal).

κ(s)
N̂ ≡
κ(s)
is the principal normal vector to C in the direction of curvature κ(s).
Lemma 1.4.
dT̂
≡ κ(s)N̂
ds
and N̂ ⊥ T̂.
Definition 1.12 (Moving Triad or Trihedral). If B̂ ≡ T̂ × N̂, then {T̂, N̂, B̂}
forms a right hand coordinate system which moves along C called the moving
triad or moving trihedral along C.
Summary.
r = r(t)
ds
≡ |ṙ|
dt
dr
T̂ ≡
ds
dT̂
κN̂ ≡
ds
dT̂
κ ≡ ≥0
ds
B̂ ≡ T̂ × N̂

8
Lemma 1.5. The derivative of B̂ with respect to arclength s is parallel to N̂

dB̂
∝ N̂.
ds
Proof. Since B̂ ⊥ T̂ by definition, B̂ · T̂ = 0. Taking the derivative with respect
to arclength s:
d  
B̂ · T̂ ≡ 0
ds
dB̂ dT̂
· T̂ + B̂ · ≡ 0
ds ds
dB̂
· T̂ + B̂ · (κN̂) ≡ 0.
ds
Since B̂ ⊥ N̂ by definition, B̂ · N̂ = 0.

dB̂ dB̂
⇒ · T̂ = 0 ⇒ ⊥ T̂
ds ds
Since B̂ is a unit vector, by Lemma 1.2,

dB̂
⊥ B̂.
ds
dB̂ dB̂
Hence ds
is in the direction of N̂ only and ds
∝ N̂.
Definition 1.13 (Torsion). The torsion τ (s) of C is defined by

dB̂
≡ −τ (s)N̂.
ds
Note. The “–” sign results in τ > 0 ⇔ the curve C twists out of the oscillating
plane, that is, the T̂ N̂ plane, in the direction B̂.
Note. Torsion measures the “rate at which C twists out of a plane with respect to
arclength”.
Lemma 1.6. τ = 0 along C ⇔ C is a plane curve.
Note. Up to position and orientation in space, any space curve is uniquely speci-
fied by the two scalar functions κ(s) and τ (s).

9
Theorem 1.1.
dN̂
= τ (s)B̂ − κ(s)T̂.
ds
Proof. N̂ = B̂ × T̂ by definition of B̂. Taking the derivative with respect to
arclength s:
dN̂ d  
= B̂ × T̂
ds ds
dB̂ dT̂
= × T̂ + B̂ ×
ds ds
= −τ (s)N̂ × T̂ + B̂ × κ(s)N̂
= τ (s)B̂ − κ(s)T̂,
since B̂ = −N̂ × T̂ and T̂ = −B̂ × N̂ by definition of cross product.
Definition 1.14 (Frenet-Serret formulae).
dT̂
= κN̂
ds
dB̂
= −τ N̂
ds
dN̂
= τ B̂ − κT̂
ds
is called the Frenet-Serret formulae, which describes the intrinsic differential ge-
ometry of space curves.
Definition 1.15 (Acceleration). If t is time, a(t) ≡ r̈ is the acceleration vector
along the curve C.
Lemma 1.7. If t is time,
dv
a= T̂ + κv 2 N̂.
dt
Proof.
dv d   dv dT̂
a = r̈ = = v T̂ = T̂ + v
dt dt dt dt
dv dT̂ ds
= T̂ + v
dt ds dt
dv
= T̂ + κv 2 N̂.
dt
10
Definition 1.16 (Tangential Acceleration). If t is time, at ≡ dv
dt
T̂ is the tangential
dv
acceleration, and dt is the time rate of change of speed along C.

Definition 1.17 (Normal Accelartion). If t is time, an ≡ κv 2 N̂ is the normal

acceleration, and κv 2 = v 2 /ρ is the centripetal acceleration.

Exercise 1.4. For motion on a circle of radius α, and angular velocity ω, provide
the centripetal acceleration and centripetal force.
Solution 1.1. ρ = α ⇒ κ = α1 , v = ds dt
= αω, κv 2 = α1 (αω)2 = αω 2 . So
2
centripetal acceleration is αω 2 = vα = κv 2 . So centripetal force is mass times
2
centripetal acceleration = maω 2 = mva = mκv 2 .
Exercise 1.5. For a circular helix r = (α cos θ, α sin θ, βθ), α ≥ 0, and θ is the
polar angle, find T̂, N̂, B̂, κ, and τ as functions of θ.
Note. The helix is the only space curve for which both κ(s) and τ (s) are constant.
Exercise 1.6. For motion on a curve r = (t, t2 , 32 t3 ) find v, s, T̂, N̂, B̂, κ, and τ
as functions of t.
Question. How far is travelled along the curve for 0 ≤ t ≤ 1?
Exercise 1.7. Using the chain rule, prove that the curvature κ(s) and torsion τ (s)
along the curve C : r(t) may be expressed in terms of derivatives of r with respect
...
to time t, that is, ṙ, r̈, and r as
...
|ṙ × r̈| (ṙ × r̈) · r
κ= and τ = .
|ṙ|3 |ṙ × r̈|2

2 Surfaces in R3
Definition 2.1 (Non-Parametric Representation). A surface S in R3 can be written
non-parametrically using one equation between x, y, z of f (x, y, z) = 0. Choos-
ing x, y as parameters on S, we can use the special form z = f (x, y).

Definition 2.2 (General Vector Form). A surface S in R3 can be written in vector

form as r = r(u, v), (u, v) ∈ Ruv where r = x(u, v)î + y(u, v)ĵ + z(u, v)k̂ =
(x(u, v), y(u, v), z(u, v)).

11
Definition 2.3 (Level Curves). The level curves of a surface S defined non-parametrically
in the form f (x, y) = z, with x, y as parameters, are the curves f (x, y) = C for
some constant C.

Definition 2.4 (Coordinate Curves). The coordinate curves on a surface S are

obtained by fixing one of the parameters so that r ≡ r(u, v0 ), with v = v0 fixed,
or r ≡ r(u0 , v), with u = u0 fixed.

Exercise 2.1. Sketch the level and coordinate curves of f , and hence sketch the
surface z = f (x, y) for (a) f (x, y) = x2 + y 2 , (b) f (x, y) = x2 − y 2 , and (c)
f (x, y) = x2 .

Definition 2.5 (Boundary). The boundary of a region R is denoted by ∂R.

Definition 2.6 (Partial Derivative). The partial derivative of f (x, y) with respect
to x at a fixed point (x0 , y0 ) is

∂f (x, y) df (x, y0 ) df
≡ fx (x0 , y0 ) ≡ = (x0 , y0 ).
∂x (x0 ,y0 ) dx x=x0 dx

Definition 2.7 (Tangent Vector). A tangent vector to a coordinate curve r =

r(u, v0 ) through (u) , v0 ) is

dr dr ∂
(u, v0 ) = (u0 , v0 ) = r(u, v) ≡ ru (u0 , v0 ).
du u=u0 du ∂u (u0 ,v0 )

Definition 2.8 (Normal Vector). A normal vector to S at (u, v) is ±ru × rv .

Definition 2.9 (Unit Normal). A unit normal n̂ to a surface S : r = r(u, v) is

ru × rv
n̂ = ±
|ru × rv |
.

Definition 2.10 (Tangent Plane). The tangent plane at (u0 , v0 ) to a surface S :

r = r(u, v) is defined by the tangent vectors ru (u0 , v0 ) and rv (u0 , v0 ).

Lemma 2.1.
|A × B| ≡ |A||B| sin γ
which is the area of the parallelogram defined by A and B.

12
Definition 2.11 (Surface Area Element). The surface area element dS of a surface
S : r = r(u, v) is
dS ≡ |ru × rv |dudv.

Note. Effectively, dS is the area of the parallelogram formed by ru and rv times

the differential elements of the two parameters u and v. Rewriting for finite dif-
ferences:
r(u0 + ∆u, v0 ) − r(u0 , v0 ) r(u0 , v0 + ∆v) − r(u0 , v0 )
∆S(u0 , v0 ) = sin γ∆u∆v
∆u ∆v
= |r(u0 + ∆u, v0 ) − r(u0 , v0 )| |r(u0 , v0 + ∆v) − r(u0 , v0 )| sin γ

where 0 ≤ γ ≤ π is defined to be the angle between ru and rv at (u0 , v0 ).

Definition 2.12 (Surface Area). The surface area S of a surface corresponding to

the region Ruv in the uv-plane is
¨ ¨ ¨
∂r ∂r
S= dS = |ru × rv |dudv = sin γdudv,
Ruv Ruv Ruv ∂u ∂v

where 0 ≤ γ ≤ π is defined to be the angle between ru and rv .

Definition 2.13 (Vector Arclength). The vector arclength element of a curve C is

ds ≡ T̂ds which is tangent to C and of magnitude ds: |ds| = ds, so that

ds = ṙdt.

Definition 2.14 (Vector Surface Element). A vector surface element to a surface

S is dS ≡ n̂dS whose magnitude is |dS| = dS and whose direction is in n̂, so
that
ru × rv
dS ≡ n̂dS = |ru × rv |dudv = (ru × rv )dudv.
|ru × rv |
Lemma 2.2. For a surface S defined non-parametrically in the form f (x, y) = z,
with x, y as parameters, dS = 1 + fx2 + fy2 dxdy and n̂ = (−f
√ x ,−f2 y ,1)2 .
p
1+fx +fy

13
Proof.
r = r(x, y) = (x, y, f (x, y))
⇒ rx = (1, 0, fx )
ry = (0, 1, fy )
⇒ rx × ry = (−fx , −fy , 1)
⇒ dS = |rx × ry |dxdy, by definition of dS,
q
= 1 + fx2 + fy2 dxdy.
ru × rv
n̂ = ± , by definition of n̂,
|ru × rv |
(−fx , −fy , 1)
= ±p .
1 + fx2 + fy2
We choose the “+” solution for n̂ since it is the upward pointing unit normal with
respect to the xy-plane.
Lemma 2.3. For a surface S defined non-parametrically in the form f (x, y) = z,
with x, y as parameters, dS = sec γdxdy, where γ is the angle between the unit
normal of the surface n̂ and the unit normal vector in the z-direction k̂.
Proof.
cos γ = n̂ · k̂, since γ is the angle between n̂ and k̂, which are both unit vectors,
(−fx , −fy , 1) · (0, 0, 1)
= p , by Lemma 2.2
1 + fx2 + fy2
1
= p
1 + fx2 + fy2
p p
so that sec γ = 1 + fx2 + fy2 , and by Lemma 2.2 ds = 1 + fx2 + fx2 dxdy =
sec γdxdy
Exercise 2.2. Calculate the surface area ofp a sphere x2 + y 2 + z 2 = α2 (a) using
x, y as parameters on one hemisphere z = a2 − x2 − y 2 (the hard way), and (b)
using spherical polar coordinates (r, θ, ϕ) with r = α (the easy way).
Theorem 2.1 (Pappus’). For a toroid with a radius r from its centroid to the center
of area A, its volume V = 2πrA, and its surface area S = 2πr`.
Exercise 2.3. Find the surface area of the paraboloid r(u, v) = (u cos v, u sin v, ku2 )
for 0 ≤ z ≤ 1, using cylindrical coordinates (r, θ, z).

14
3 Scalar & Vector Fields and the “∇” Operator
Definition 3.1 (Scalar Field). A scalar field is a scalar function of position V (x, y, z).
Example 3.1. The temperature in a room T (x, y, z)
Example 3.2. The pressure in a region P (x, y, z).
Example 3.3. The electrostatic potential V (x, y, z), whose gradient ∇V = grad V
is related to the electric field vector E(x, y, z).
Example 3.4. The charge density ρ(x, y, z).
Definition 3.2 (Vector Field). A vector field is a vector function of position V(x, y, z).
Example 3.5. Electric field vector E(x, y, z). On a charge q the electric force is
F = qE.
Example 3.6. Magnetic flux density B(x, y, z). The force on a moving charge q
is F = qv × B.
Example 3.7. Velocity field of a steady fluid flow in a region v(x, y, z).
Definition 3.3 (∇ Operator). The “∇” operator, also called “del” and “nabla”, is
the differential operator
 
∂ ∂ ∂ ∂ ∂ ∂
∇≡ , , = î + ĵ + k̂ ,
∂x ∂y ∂z ∂x ∂y ∂z
in Cartesian coordinates (x, y, z),
 
∂ 1 ∂ ∂ ∂ 1 ∂ ∂
∇≡ , , = êr + êθ + êz ,
∂r r ∂θ ∂z ∂r r ∂θ ∂z
in cylindrical coordinates (r, θ, z), and
 
∂ 1 ∂ 1 ∂ ∂ 1 ∂ 1 ∂
∇≡ , , = êr + êθ + êϕ ,
∂r r sin ϕ ∂θ r ∂ϕ ∂r r sin ϕ ∂θ r ∂ϕ
in spherical coordinates (r, θ, ϕ),
Definition 3.4 (Gradient). The gradient of a scalar field V (x, y, z) is the vector
field
   
∂V ∂V ∂V ∂ ∂ ∂
grad V = ∇V ≡ , , = , , V (x, y, z)
∂x ∂y ∂z ∂x ∂y ∂z
.

15
Note. The gradient of V measures the rate of change of V with distance.
Example 3.8. Pressure gradient ∇P .
Example 3.9. Temperature gradient ∇T .
Definition 3.5 (Divergence). The divergence of a vector field V(r) = (V1 , V2 , V3 )
is the scalar field
 
∂ ∂ ∂ ∂V1 ∂V2 ∂V3
div V = ∇ · V ≡ , , · (V1 , V2 , V3 ) = + + ,
∂x ∂y ∂z ∂x ∂y ∂z
in Cartesian coordinates r = (x, y, z),
 
1 ∂ 1 ∂ ∂ 1 ∂(rV1 ) 1 ∂V2 ∂V3
div V = ∇ · V ≡ r, , · (V1 , V2 , V3 ) = + + ,
r ∂r r ∂θ ∂z r ∂r r ∂θ ∂z
in cylindrical coordinates r = (r, θ, z),
 
1 ∂ 2 1 ∂ 1 ∂
div V = ∇ · V ≡ r , , sin ϕ · (V1 , V2 , V3 )
r2 ∂r r sin ϕ ∂θ r sin ϕ ∂ϕ
1 ∂(r2 V1 ) 1 ∂V2 1 ∂(sin ϕV3 )
= 2 + +
r ∂r r sin ϕ ∂θ r sin ϕ ∂ϕ
∂V1 2 1 ∂V2 1 ∂V3 cot ϕ
= + V1 + + + V3
∂r r r sin ϕ ∂θ r ∂ϕ r
in spherical coordinates r = (r, θ, ϕ).
Note. The divergence measures the local “expansion rate” of the vector field V.
It is related to conservation laws in physics and the presence of sources or sinks
in the field.
Definition 3.6 (Divergence Free or Solenoidal). A vector field V is divergence
free or solenoidal if div V = ∇ · V = 0 in a region.
Note. The magnetic field is divergence free or solenoidal since div B = ∇ · B ≡
0. This means there are no sources or sinks of the magnetic field, and hence
isolated magnetic “charges”, called magnetic monopoles, do not exist, that is,
there is no evident for isolated North or South magnetic poles.
Note. For fields with sources, we obtain laws of the form div(εE) = ∇ · E = ρ,
where E is the electric field strength,  is the permittivity of space, and ρ is the
charge density. This means the E field is divergence free only in regions with no
electric charges. This holds only for static charge distribution fields.

16
Definition 3.7 (Curl). The curl of a vector field V(r) = (V1 , V2 , V3 ) is the vector
field
 
∂ ∂ ∂
curl V = ∇ × V ≡ , , × (V1 , V2 , V3 )
∂x ∂y ∂z
î ĵ k̂
∂ ∂ ∂
= ∂x ∂y ∂z
V1 V2 V3
 
∂V3 ∂V2 ∂V1 ∂V3 ∂V2 ∂V1
= − , − , − .
∂y ∂z ∂z ∂x ∂x ∂y

in Cartesian coordinates r = (x, y, z),

 
1 ∂V3 ∂V2 ∂V1 ∂V3 1 ∂rV2 1 ∂V1
∇×V = − , − , − ,
r ∂θ ∂z ∂z ∂r r ∂r r ∂θ

in cylindrical coordinates r = (r, θ, z),

   
1 1 ∂ sin ϕV3 ∂V2 1 ∂V1 ∂rV3 ∂rV2 ∂V1
∇×V = − , − , − ,
r sin ϕ ∂θ ∂ϕ sin ϕ ∂θ ∂r ∂r ∂ϕ

in spherical coordinates r = (r, θ, ϕ).

Note. The curl measures the local rotation or “swirling” property of the vector
field V .

Definition 3.8 (Irrotational). A vector field V is irrotational if curl V = ∇×V =

0 in a region.

Definition 3.9 (Simply Connected Region). A simply connected region i one in

which every simple closed curve can be shrunk to a point without leaving the
region.

Definition 3.10 (Scalar Potential). A scalar potential φ of a vector field V satisfies

V = ± grad φ = ±∇φ.

Lemma 3.1 (Scalar Potential). If a vector field V is irrotational on a simply-

connected region R, that is, curl V = ∇ × V ≡ 0, then there exists a scalar
potential φ of V such that V = ± grad φ = ±∇φ.

17
Proof. Assuming ∇ × V = 0, so
 
∂V3 ∂V2 ∂V1 ∂V3 ∂V2 ∂V1
− , − , − =0
∂y ∂z ∂z ∂x ∂x ∂y

yields three partial differential equations for the coefficients of V:

∂V3 ∂V2 ∂V1 ∂V3 ∂V2 ∂V1
= , = , =
∂y ∂z ∂z ∂x ∂x ∂y
Integrating with respect to x, y, and z, and reversing the order of integration and
differentiation, yields
ˆ ˆ ˆ
V1 dx = V2 dy = V3 dz

Since the integrals of the three coordinates are equal, this implies there exists a
scalar function φ whose derivatives are equal to three components of V, that is,
∂φ ∂φ ∂φ
∃φ : V1 = ± , V2 = ± , V3 = ±
∂x ∂y ∂z
so that V = ±∇φ.

Lemma 3.2. ∇ × ∇V = curl grad V ≡ 0 if V (x, y, z) is a twice continuous

differentiable scalar field.

Proof.
 
∂V ∂V ∂V
∇ × ∇V = ∇× , ,
∂x ∂y ∂z
 
∂ ∂V ∂ ∂V ∂ ∂V ∂ ∂V ∂ ∂V ∂ ∂V
= − , − , −
∂y ∂z ∂z ∂y ∂z ∂x ∂x ∂z ∂x ∂y ∂y ∂x
 2 2 2 2 2 2

∂ V ∂V ∂ V ∂V  ∂ V
 ∂ V
=  − ,  − ,  −
∂y∂z ∂z∂y ∂z∂x ∂x∂z ∂x∂y ∂y∂x

so curl grad V = ∇ × ∇V ≡ 0.

Theorem 3.1. Let F be a continuously differentiable vector field in is a simply-

connected region R of R3 . Then the following five conditions are equivalent.

18
 1. ˆ b
F · dr
a
is path independent in R.

2. ˛
F · dr = 0
C
for all closed curves C ⊂ R.

3. curl F = ∇ × F ≡ 0 in R.

4. F = ±∇φ for some continuously differentiable field φ in R, the scalar

potential of F.

5. F·dr = F1 dx+F2 dy+F3 dz is an exact differential in R of some “potential”

function φ which is continuously differentiable in R.

Summary.
 
∂ ∂ ∂
∇ ≡ , ,
∂x ∂y ∂z
 
∂V ∂V
∇V = grad V ≡ ,
∂x ∂y
∂V1 ∂V2
∇ · V = div V ≡ +
∂x ∂y
 
∂V2 ∂V1
∇ × V = curl V ≡ 0, 0, − in k̂ direction.
∂x ∂y

Lemma 3.3 (Combinations of the ∇ Operator). The only valid combinations of

the ∇ operator are

∇ · (∇V ) = div grad V

∇ × (∇V ) = curl grad V
∇ × (∇ × V) = curl curl V
∇ · (∇ × V ) = div curl V
∇(∇ · V) = grad div V

19
Definition 3.11 (Laplacian ∇2 V ). The Laplacian of a scalar field V (r) is the
scalar field
   
2 ∂ ∂ ∂ ∂V ∂V ∂V
∇ V = div grad V ≡ , , · , ,
∂x ∂y ∂z ∂x ∂y ∂z
∂ 2V ∂ 2V ∂ 2V
= + +
∂x2 ∂y 2 ∂z 2
in Cartesian coordinates r = (x, y, z),
   
2 1 ∂ 1 ∂ ∂ ∂V 1 ∂V ∂V
∇V ≡ r, , · , ,
r ∂r r ∂θ ∂z ∂r r ∂θ ∂z
∂ 2V 1 ∂V 1 ∂ 2V ∂ 2V
= + + +
∂r2 r ∂r r2 ∂θ2 ∂z 2
in cylindrical coordinates r = (r, θ, z), and
   
2 1 ∂ 2 1 ∂ 1 ∂ ∂V 1 ∂V 1 ∂V
∇V ≡ r , , sin ϕ · , ,
r2 ∂r r sin ϕ ∂θ r sin ϕ ∂ϕ ∂r r sin ϕ ∂θ r ∂ϕ
∂ 2V 2 ∂V 1 ∂ 2V 1 ∂ 2V cot ϕ ∂V
= 2
+ + 2 2 2
+ 2 2
+ 2
∂r r ∂r r sin ϕ ∂θ r ∂ϕ r ∂ϕ
in spherical coordinates r = (r, θ, φ).
Definition 3.12 (Laplacian ∇2 V). The Laplacian of a vector field V(r) = (V1 , V2 , V3 )
is the vector field

∇2 V ≡ ∇2 V1 , ∇2 V2 , ∇2 V3

Definition 3.13 (Laplacian Operator). The Laplacian operator is the differential

operator
∂2 ∂2 ∂2
∇2 ≡ + + = div grad = ∇ · ∇
∂x2 ∂y 2 ∂z 2
in Cartesian coordinates (x, y, z).
Definition 3.14 (Inverse-Square Law Potentials). Any scalar field V which satis-
fies ∇2 V = 0, Laplace’s Equation, is called a harmonic function.
Example 3.10. Inverse square law of force.
• gravitation forces ⇒ gravitational potential.

20
 • electrostatic forces ⇒ electrostatic potential.

Exercise 3.1. In R3 , what is the gravitational potential function φ, where the grav-
itational force is
GmM
F(x, y, z) = − 3 r.
r
Find a φ such that
GmM
∇φ = r
r3
and show that φ ∝ 1r is the gravitational potential.
Note. In physics, we choose φ such that F = − grad φ so that energy conserva-
tion takes the form “kinetic energy + potential energy = total energy = constant”.
Otherwise, with ‘+’ in the definition of φ we would obtain “kinetic energy - po-
tential energy = constant”.
Note. Given an irrotational force F = (F1 , F2 , F3 ), then one may integrate F =
−∇φ to find the scalar potential φ(x, y, z). This involves solving three coupled
partial differential equations:
∂φ
F1 = −
∂x
∂φ
F2 = −
∂y
∂φ
F3 = − .
∂z
Integrate one of the three equations, pick up an arbitrary function of integration
of the other two variables, and use the remaining two equations to determine this
function.
Example 3.11. Electrostatic field E = −∇V , where V is the electrostatic scalar
potential.

Lemma 3.4. ∇ · ∇ × V = div curl V ≡ 0 if V = (V1 , V2 , V3 ) is a twice

continuous differentiable vector field.

21
Proof.
 
∂V3 ∂V2 ∂V1 ∂V3 ∂V2 ∂V1
∇·∇×V = div − , − , −
∂y ∂z ∂z ∂x ∂x ∂y
2 2 2 2 2
∂ V3 ∂ V2 ∂ V1 ∂ V3 ∂ V2 ∂ 2 V1
= − + − + −
∂x∂y ∂x∂z ∂y∂z ∂y∂x ∂z∂x ∂z∂y
∂ 2 V3 ∂ 2 V2 ∂ 2 V1 ∂ 2 V3 ∂ 2 V2 ∂ 2 V1
= − + − + −
∂x∂y ∂x∂z ∂y∂z ∂x∂y ∂x∂z ∂y∂z
∂ 2 V3 ∂2 V3 ∂ 2 V2 ∂2 V2 ∂ 2 V1 ∂ 2
V
1
=  −
 +  − +  −

∂x∂y ∂x∂y ∂x∂z ∂x∂z  ∂y∂z ∂y∂z
    
= 0

Definition 3.15 (Vector Potential). A vector potential A of a vector field V satis-

fies V = curl A = ∇ × A.
Lemma 3.5 (Vector Potential). If a vector field V is solenoidal, div V = ∇·V ≡
0, then there exists a vector potential A such that V = curl A = ∇ × A.
Proof. Assuming ∇ · V = 0, so
∂V1 ∂V2 ∂V3
+ + =0 (1)
∂x ∂y ∂z
and integrating (1) with respect to x, y, and z
¨ ¨ ¨
V1 dydz + V2 dxdz + V3 dxdy = 0
ˆ ˆ  ¨
V2 dz + V3 dy dx = − V1 dydz. (2)

We may assume without loss of generality that

∂A1
V2 = − C2 (3)
∂z
∂A1
V3 = − + C3 . (4)
∂y
Substituting (3) and (4) into (2)
ˆ ˆ  ¨
∂A1 ∂A1
− C2 dz + + C3 dy dx = − V1 dydz
∂z ∂y
¨ ¨ ¨
⇒ C3 dxdy − C2 dxdz = − V1 dydz (5)

22
We may further assume, again without loss of generality, that C3 and C2 may be
expressed as partial derivatives with respect to x, so that
∂A2
C3 = (6)
∂x
∂A3
C2 = . (7)
∂x
Substituting (6) and (7) into (5) yields
¨ ¨ ¨
∂A2 ∂A3
dxdy − dxdz = − V1 dydz
∂x ∂x
ˆ ˆ ¨
− A2 dy + A3 dz = V1 dydz (8)

Differentiating (8) with respect to y and z yields

∂A2 ∂A3
V1 = − + ,
∂z ∂y
so that we may express V = (V1 , V2 , V3 ) in terms of A1 , A2 , and A3 as
∂A3 ∂A2
V1 = −
∂y ∂z
∂A1 ∂A2
V2 = −
∂z ∂x
∂A3 ∂A1
V3 = − .
∂x ∂y
Thus for the vector field V there exists a vector potential A = (A1 , A2 , A3 ) such
that V = ∇ × A.
Example 3.12. The magnetic field B. Because div B ≡ 0 from Maxwell’s Equa-
tions, there exists a vector potential A such that B = curl A. A is called the
magnetic vector potential.
Lemma 3.6.

curl curl V = grad div V − div grad V

∇ × (∇ × V) = ∇(∇ · V) − ∇2 V

where div grad V ≡ ∇2 V ≡ (∇2 V1 , ∇2 V2 , ∇2 V3 ).

23
Exercise 3.2. Prove Lemma 3.6.
Lemma 3.7 (Linearity of ∇). Let φ, ψ be scalar fields and V1 , V2 be vector fields

∇(φ + ψ) = ∇φ + ∇ψ
∇ · (V1 + V2 ) = ∇ · V1 + ∇ · V2
∇ × (V1 + V2 ) = ∇ × V1 + ∇ × V2

Lemma 3.8. Let φ be a scalar field and A, B be vector fields

∇ · (φA) = ∇φ · A + φ∇ · A
∇ · (A × B) = (∇ × A) · B − (∇ × B) · A
∇ × (A × B) = (B · ∇)A − B(∇ · A) − (A · ∇)B + A(∇ · B)
∇(A · B) = (B · ∇)A + (A · ∇)B + B × (∇ × A) + A × (∇ × B)

Exercise 3.3. Consider the vortex field

(−y, x, 0)
V= , (x, y) 6= (0, 0).
x2 + y 2
Show that V is irrotational, that is, curl V = ∇ × V ≡ 0 for all (x, y) 6= (0, 0),
so that in simply connected regions there exists φ such that V = ∇φ, and find
φ(x, y, z). Show that V is solenoidal, that is div V = ∇ · V = 0 for all (x, y) 6=
(0, 0), so that in simply connected regions there exists A such that V = curl A =
∇ × A, and find A by assuming (1) Az ≡ 0 and (2) A1 (x, y) = 0 = A2 (x, y).

4 Physical Interpretation of Grad, Div, and Curl

Definition 4.1 (Flow Lines). For a vector field V(V1 , V2 , V3 ), the flow lines are
the paths traced through the vector field satisfying
dr
v = V or = V or ṙ = V
dt
where t is time.
Note. To find flow lines, integrate
dr
=V
dt
24
 dx dy dz
⇒
= V1 (x, y, z), = V2 (x, y, z), = V3 (x, y, z).
dt dt dt
These equations may be integrable depending on V
dx dy dz
⇒ dt = = =
V1 V2 V3
in non-parametric form (two equations and three unknowns).
Definition 4.2 (Integral Curves). For a vector field V(V1 , V2 , V3 ) the integral
curves are the paths traced through the vector field satisfying
ˆ
r = r(t) = Vdt.

Definition 4.3 (Isotimic Surfaces). For a scalar potential φ(x, y, z) the surfaces
φ(x, y, z) = C where C is a constant are called the isotimic surfaces of φ.
Lemma 4.1 (Direction of ∇φ). If φ(x, y, z) is a scalar field with φ = C (C is
a constant) its isotimic surfaces, then grad φ = ∇φ is a vector field everywhere
perpendicular to the isotimic surfaces φ = C and grad φ = ∇φ points in the
direction in which φ increases most rapidly with respect to displacements ds while
− grad φ = −∇φ is in the direction in which φ decreases most rapidly.
Definition 4.4 (Directional Derivative). The directional derivative of φ in the di-
rection of a unit vector û is
dφ ∂φ
= = Dû φ = ∂û φ = grad φ · û = | grad φ| cos γ = |∇φ| cos γ.
ds û ∂ û
varies sinusoidally with γ, the angle between û and ∇φ.
Lemma 4.2 (Magnitude of ∇φ). If φ(x, y, z) is a scalar field the maximum rate
of increase of φ is given by | grad φ| = |∇φ|.
Note. Given a surface in R3 of the form f (x, y, z) = C (C is a constant), a normal
is given by grad φ = ∇φ.
Note. Given a vector field V = (V1 , V2 , V3 ) in R3 , div V is a scalar field in
R3 , div V is the outward flux per unit volume at that point, or the flux outward
through a surface S divided by volume V enclosed by S.
Exercise 4.1. Find the flow lines of V = (1, 2x) in R2 in parametric and non-
parametric form.

25
Exercise 4.2. Find the flow lines of V = (−ωy, ωx, 0) in parametric and non-
parametric form for rigid body motion where ω is the angular velocity about the
z-axis.
Definition 4.5 (Equipotentials). For a scalar potential φ(x, y) the curves φ(x, y) =
C where C is a constant are called the equipotentials.
Example 4.1. Isotherms T = constant.
Example 4.2. Isobars P = constant.
Note. The definition of divergence in terms of flux allows us to formulate a differ-
ential form of conservation laws.
Definition 4.6 (Flux). If φ is a scalar field which flows with velocity field v, then
the flux of φ is J ≡ φv in units of φ per unit volume per unit time.
Example 4.3. If φ = ρ is mass density in kg/m3 , then J = ρv is the mass flux in
kg/m3 s.
Example 4.4. If φ = σ is charge density in C/m3 , then J = σv is the charge flux
in C/m3 s.
Note. Divergence of a vector field V is the outflow per unit volume in the limit as
volume → 0. So outflow per unit time through the border of ∆V , i.e., ∂∆V , is
(div V)∆V .
Theorem 4.1 (Continuity Equation). If the physical quantity corresponding to
a vector field V is conserved, then its density φ must change correspondingly,
yielding
∂φ
∇·J+ = 0,
∂t
where J ≡ φV.
Proof. For a volume ∆V , (∇ · J)∆V is the outflow per unit time from ∆V , and
∂φ
∂t
∆V is the change in the amount of φ inside ∆V per unit time. If quantities are
conserved, then the outflow from ∆V per unit time must balance the change in
the density within ∆V per unit time, so that
∂φ
(∇ · J)∆V + ∆V = 0.
∂t
Taking the limit ∆V → 0+ yields the continuity equation
∂φ
∇·J+ = 0.
∂t

26
Example 4.5. φ = ρ = mass density ⇒ J = ρv = momentum density and mass
conservation ⇔
∂ρ
∇ · (ρv) + = 0.
∂t
Example 4.6. φ = σ = charge density ⇒ J = σv = current density and charge
conservation ⇔
∂σ
∇ · (σv) + = 0.
∂t
Exercise 4.3. Try finding expressions for (1) div V and (2) ∇2 V in spherical
coordinates in terms of r, θ, ϕ derivatives.

x = r sin θ cos ϕ
y = r sin θ sin ϕ
z = r cos ϕ

using chain rule.

5 Line Integrals
Definition 5.1 (Line Integral of a Scalar Field). The line integral from P0 (t = t0 )
to P1 (t = t1 ) on the oriented curve C in R3 , r = r(t), t0 ≤ t ≤ t1 of a scalar
quantity f (x, y, z) is
ˆ P1
C f (r)ds.
P0

where ds is the arclength element along C from P0 to P1 .

Note. In general, the line integral depends on the curve C and the end points P0
and P1 . Along some other curve C 0 between the same end points we will not
always obtain the same answer.
mass
Note. If f is density of material on a wire, length
, then
ˆ P1
C f ds
P0

is the total mass of the wire.

27
Definition 5.2 (Line Integral of a Vector Field). The line integral from P0 (t = t0 )
to P1 (t = t1 ) on the oriented curve C in R3 , r = r(t), t0 ≤ t ≤ t1 of a vector
quantity F(F1 , F1 , F3 ) may be written as
ˆ P1 ˆ P1 ˆ P1 ˆ P1
C F · dr = C F · T̂ds = C F · ds = C F · ṙdt
P0 P0 P0 P0

since
dr dr
T̂ ≡ and ṙ ≡ .
ds dt
Note. Definition 5.1 is just a special case of Definition 5.2 with f (x, y, z) ≡ F· T̂.

Definition 5.3 (Work). If F is the force vector field, then the line integral
ˆ P1 ˆ P1
W = C F · dr = C dW
P0 P0

represents the work done by the force field F on a particle when moved along the
curve C from P0 to P1 , since F · dr = force × displacement = dW .

Note. The work integral in general depends on the path C from P0 to P1 .

Lemma 5.1. If F = ±∇φ then F · dr = ±dφ.

Proof.
 
∂φ ∂φ ∂φ
F · dr = ±∇φ · dr = ± , , · (dx, dy, dz)
∂x ∂y ∂z
 
∂φ ∂φ ∂φ
= ± dx + dy + dz
∂x ∂y ∂z
 
∂φ dx ∂φ dy ∂φ dz
= ± + + dt
∂x dt ∂y dt ∂z dt
dφ
= ± dt
dt
so F · dr = ±dφ.

Lemma 5.2. The following statements are equivalent.

´P ´P
1. The integral P01 F · dr = P01 dW is path independent.

28
 2. F is a conservative force field.

3. Total energy is conserved.

4. There exists a scalar potential φ of F such that F = −∇φ.

5. curl F = ∇ × F ≡ 0 on simply connected regions.

6. F · dr is an exact differential of −dφ and F · dr is an exact ordinary differ-

ential equation.

Lemma 5.3. The work done by a force field F on a particle when moved along a
path C from P0 to P1 is equal to the difference in kinetic energy at P1 and P0 , that
is, WP0 →P1 = K.E.(P1 ) − K.E.(P0 ).

Proof.
ˆ P1 ˆ P1 ˆ P1
WP0 →P1 ≡ dW = F · dr = mr̈ · dr by Newton’s 3rd Law
P0 P0 P0
ˆ P1 ˆ P1
dṙ 1d
= m · dr = m [ṙ · ṙ] dt
P0 dt P0 2 dt
ˆ P1 ˆ
m m P1
= d(ṙ · ṙ) = d(v 2 )
2 P0 2 P0
   
1 2 1 2
= mv − mv = K.E.(P1 ) − K.E.(P0 )
2 P1 2 P0

´
Exercise 5.1. Find C f (x, y, z)dx where f (x, y, z) = x + y + z and C : x = cos t,
y = sin t, z = t for 0 ≤ t ≤ π2 .
´B
Exercise 5.2. Evaluate I = C A (x + y)dx along each of

1. circle x = cos θ, y = sin θ from (1,0) to (0,1).

   
y=0 x=0
2. broken line ∪ from (1,0) to (0,1).
0≤x≤1 0≤y≤1
Note. Different answers imply this line integral is not path independent. Here
F = (x + y, 0, 0) ⇒ curl F = (0, 0, −1) 6= 0. Since curl F ≡ 6 0 we do not have
path independence.

29
Exercise 5.3. Find the work done by the force field F = (2xy, x2 , 0) in the xy-
plane when moving from (0,0,0) to (1,1,0) along
   
y=0 x=1
1. ∪ ,
0≤x≤1 0≤y≤1
2. y = x, 0 ≤ x ≤ 1,
   
x=0 y=1
3. ∪ .
0≤y≤1 0≤x≤1
Note. All three paths gave the same answer. This suggests that we might have
a path independent line integral. In fact, we do F · dr = 2xydx + x2 dy =
−d(−x2 y) = −dφ where φ = −x2 y is a potential function. We have an exact
differential in the integrand, that is, F = − grad φ, so that
ˆ B ˆ B
C F · dr = C −dφ = φ[A] − φ[B].
A A

Lemma 5.4 (Conservation of Energy). Conservation of energy holds for conser-

vative force fields F = −∇φ, where the scalar field φ is the potential energy of
the force field.

Proof. If F = −∇φ for some scalar field φ, then employing Lemma 5.3

WP0 →P1 = K.E.(P1 ) − K.E.(P0 ) (9)

Since F = −∇φ, by Lemma 5.1 F · dr = −dφ and

ˆ P1 ˆ P1
WP0 →P1 = F · dr = − dφ = φ(P0 ) − φ(P1 ) (10)
P0 P0

Since φ is the potential energy P.E. of the force field F, equating (9) and (10), we
find

K.E.(P1 ) − K.E.(P0 ) = P.E.(P0 ) − P.E.(P1 )

⇒ K.E.(P0 ) + P.E.(P0 ) = K.E.(P1 ) + P.E.(P1 ).

⇒ total energy E = K.E. + P.E. is conserved, since E(P0 ) = E(P1 ).

30
Lemma 5.5. If ˆ P1
F · dr
P0

is path independent in a region R, then the line integral around any closed curve
C ⊂ R is zero ˛
F · dr = 0.
C

Proof. Let P0 and P1 be any two points on a closed curve C, so that C = CP0 →P1 ∪
CP1 →P0 . Then
˛ ˆ ˆ
F · dr = F · dr + F · dr.
C CP0 →P1 CP1 →P0

Since we assumed the work integral is path independent, we have

˛ ˆ P1 ˆ P0 ˆ P1 ˆ P1
F · dr = F · dr + F · dr = F · dr − F · dr = 0,
C P0 P1 P0 P0

so that ˛
F · dr = 0
C
for all closed curves C in R.

Lemma 5.6. If the line integral around any closed curve C ⊂ R is zero
˛
F · dr = 0.
C

then ˆ P1
F · dr
P0

is path independent on R.

Proof. Let P0 and P1 be two points on a closed curve C, so that C = CP0 →P1 ∪
CP0 1 →P0 . Then
˛ ˆ ˆ
0= F · dr = F · dr + F · dr.
C CP0 →P1 0
CP
1 →P0

31
 ˆ ˆ ˆ ˆ P1
⇒ F · dr = − F · dr = F · dr = F · dr
CP0 →P1 0
CP 0
CP P0
1 →P0 0 →P1

where we have reversed the direction of the curve CP0 1 →P0 to go from P0 → P1 .
Since the two integrals are equal, and CP0 →P1 and CP0 0 →P1 are arbitrarily chosen
curves going from point P0 to P1 , this implies
ˆ P1
F · dr
P0

is path independent.
Note. The potential φ(x, y, z) can be defined by a line integral from some arbi-
trary reference point (x0 , y0 , z0 ) to the point (x, y, z) in question. In physics, this
reference point (x0 , y0 , z0 )is typically taken to be the potential at infinity, or far
from the origin, x0 , y0 , z0 → ∞. In this way
ˆ (x,y,z)
φ(x, y, z) ≡ − C F · dr.
(x0 ,y0 ,z0 )

Question. Does ∇ × F ≡ 0 in R ⇒ path independence in R?

Answer. No, in general.
Exercise 5.4. Show that for the vortex field
(−y, x, 0)
V= , (x, y) 6= (0, 0),
x2 + y 2
while ∇ × V ≡ 0 everywhere V is defined,
˛
F · dr 6= 0
C

for every closed curve C in R2 /(0, 0).

Note. The potential of the vortex field φ = arctan( xy ) is not continuous in R.
In order to get the converse, that is, ∇ × F ≡ 0 in R ⇒ path independence in
R, we must restrict ourselves to simply-connected regions R in which curl F =
∇ × F ≡ 0.

32
Definition 5.4 (Line Integral as a Limit of a Sum). Partition the curve C into n
segments of lengths ∆si (i = 1, . . . , n) by points of subdivision ri (i = 0, . . . , n).
Denote ∆ri ≡ ri − ri=1 . Then |∆ri | ≈ ∆si if ∆si is sufficiently small. In each
segment, choose some point Ri on C (i = 1, . . . n). Form the sum
n
X
f (Ri )∆si .
i=1

Let n → ∞ and max |∆si | → 0. Then define

ˆ P1 n
X
C f (x, y, z)ds ≡ lim f (Ri )∆si .
n→∞
P0 max |∆si | → 0 i=1

Lemma 5.7 (Uniqueness of Line Integral). If C is piecewise smooth and f is

continuous on C then the limit
n
X
lim f (Ri )∆si .
n→∞
max |∆si | → 0 i=1

is independent of partition, how the Ri are chosen, and how we let n → ∞.

6 Green’s Theorem
Definition 6.1 (Continously Differentiable). A function P is continuously differ-
entiable on a region R if P and its partial derivatives are all continuous on R.

George Green (1793–1841)

33
Theorem 6.1 (Green’s Theorem). Let P (x, y) and Q(x, y) be functions which are
continuously differentiable in R. Orient ∂R, the whole boundary of R, so that R
is always on the left when one advances along the curve in the positive direction,
that is, if T̂ is the unit tangent to ∂R and if N̂ is the unit vector perpendicular to
T̂ lying in the xy-plane and pointing into R, then T̂ × N̂ = k̂, so {T̂, N̂, k̂} is a
right handed coordinate system, where N̂ is ± the principal normal to ∂R. Then
˛ ¨  
∂Q ∂P
(P dx + Qdy) = − dxdy
∂R R ∂x ∂y
.

George Stokes (1819–1903)

Theorem 6.2 (Stokes’ Theorem in R2 ). If F = (P, Q, 0), then F · dr = P dx +
Qdy,  
∂Q ∂P
curl F = ∇ × F = 0, 0, −
∂x ∂y
and we can rewrite Green’s Theorem in vector form as
˛ ¨
F · dr = ∇ × F · k̂dA
∂R R
.
Definition 6.2 (Circulation).
˛ ˛
F · dr = F · T̂ds
C C

is the circulation of the vector field F around the closed curve C.

34
Definition 6.3. ¨
∇ × F · B̂dA
R
is the area sum of the local “swirl”, or curl F throughout the interior of R.

Carl Friedrich Gauss (1777–1855)

Theorem 6.3 (Gauss’ Divergence Theorem in R2 ). If F = (Q, −P, 0), then F ·
dr × k̂ = F · (dy, −dx, 0) = Qdy + P dx,
∂Q ∂P
div F = ∇ · F = −
∂x ∂y
and we can rewrite Green’s Theorem as
˛ ¨
F · dr × k̂ = ∇ · FdA.
∂R R

Note.
1. Only Definition 6.3 requires R to be a simply connected region. Theorems
6.2 and 6.3 and Definition 6.2 hold when R is not simply connected.
2. A vector field satisfying any of these conditions is called conservative.
3. Stokes’ Theorem is an integral form of a relationship from which we can
deduce path independence.
Note. Green’s Theorem is a 2D version of both Gauss’ Divergence Theorem and
Stokes’ Theorem in R3 . It relates the integral around the boundary ∂R of a region
R ⊂ R2 to a double integral over R, where we orient ∂R positively with the region
“on the left”.

35
Theorem 6.4 (Green’s Theorem). Let R be a closed bounded region of the xy-
plane whose boundary ∂R consists of a finite number of non-intersecting simple
closed curves each of which is piecewise smooth. Consider the integral of the
normal component of F outwards around ∂R. Let n̂ be the unit outward pointing
normal to R on ∂R. Then
˛
F · n̂ds = flux of F outward through ∂R.
∂R
F · n̂ds = (Q, −P, 0) · (T̂ × k̂)ds
= (Q, −P, 0) · (dr × k̂)
= (Q, −P, 0) · ((dx, dy, 0) × (0, 0, 1))
= (Q, −P, 0) · (dy, −dx, 0)
= Qdy + P dx

so Green’s Theorem can be written

˛ ¨
F · n̂ds = ∇ · FdA.
∂R R

Definition 6.4 (Flux). ˛

F · n̂ds
∂R
is the flux of F outward through ∂R.

Definition 6.5. ¨
∇ · FdA
R
is the sum of the local divergence over the interior of R.

Theorem 6.5 (Green’s Theorem). Let P and Q be continuously differentiable in

R and on ∂R, then
˛ ¨  
∂Q ∂P
P dx + Qdy = − dxdy.
∂R R ∂x ∂y

Proof. (For x-simple and y-simple regions where every horizontal and vertical
line intersects ∂R in at most two points.)

36
 P and Q are independent arbitrary functions. So we can prove Green’s Theo-
rem, and similarly Stokes’ and Gauss’ Theorems, by proving the separate results
˛ ¨
∂P
P dx = − dxdy (11)
∂y
˛∂R ¨ R
∂Q
and Qdy = dxdy (12)
∂R R ∂x

since by linearity we obtain Green’s Theorem by adding (11) and (12). Once
proven for such regions, we can extend to non-simple regions by subdivision. We
also can extend to non-simply connected regions by subdivision.
Let C1 : y = y1 (x) be the bottom half of ∂R and C2 : y = y2 (x) be the top
half of ∂R so that ∂R = C1 ∪ C2 . Let a ≤ x ≤ b on C1 and C2 . Consider
˛ ˆ b ˆ a
P dx = C1 P (x, y1 (x))dx + C2 P (x, y2 (x))dx
∂R x=a x=b
ˆ b
= − (P (x, y2 (x)) − P (x, y1 (x))) dx
a
ˆ b ˆ y2 (x) !
∂P
= − (x, y)dy dx
a y=y1 (x) ∂y
ˆ b ˆ y2 (x)
∂P
= − dydx, an iterated integral
a y=y1 (x) ∂y
¨
∂P
= − dxdy is a double integral
R ∂y

Similarly for Q.

Lemma 6.1 (Green’s Theorem in Polar Coordinates). In polar coordinates in the

xy-plane Green’s Theorem is
˛ ¨  
∂Q ∂P
(P dr + Qdθ) = − drdθ
∂R R ∂r ∂θ
.

Note. Green’s theorem has the same form in Cartesian and polar coordinates. In-
deed, Green’s Theorem is coordinate invariant.

37
Lemma 6.2. If u = u(x, y) and v = v(x, y) is an invertible mapping in the xy-
plane, then ˛ ¨  
∂Q ∂P
(P du + Qdv) = − dudv
∂R R ∂u ∂v
is a coordinate invariant result.
Exercise 6.1. Prove Lemma 6.2.
Exercise 6.2. Verify Green’s Theorem for
˛
(xy + y 2 )dx + x2 dy
C

where C is the curve formed by y = x and y = x2 for 0 ≤ x ≤ 1.

Exercise 6.3. Evaluate
˛
I= (y − sin x)dx + cos xdy
C

counter-clockwise around the triangle with vertices (0, 0), ( π2 , 0), and ( π2 , 1).
Lemma 6.3 (Areas Using Green’s Theorem).
˛ ¨  
∂Q ∂P
(P dx + Qdy) = − dxdy
∂x ∂y

The double integral is the area A of R if ∂Q

∂x
− ∂P
∂y
≡ 1:
¨ ˛
A= dxdy = (P dx + Qdy).
R ∂R

Note. There are many possible choices for P and Q which allow the calculation
of areas:
1. P = 0, Q = x ⇔ F = (0, x, 0)
˛
∂Q ∂P
⇒ − ≡1⇒A= xdy
∂x ∂y ∂R

2. P = y, Q = 0 ⇔ F = (y, 0, 0)
˛
∂Q ∂P
⇒ − ≡1⇒A=− ydx
∂x ∂y ∂R

38
 3. P = 12 y, Q = 21 x ⇔ F = 21 (−y, x, 0)
˛
∂Q ∂P 1
⇒ − ≡1⇒A= xdy − ydx
∂x ∂y 2 ∂R

Lemma 6.4 (Work Integral Using Green’s Theorem). The work integral can be
converted into a double integral
˛ ¨
W = F · dr = ∇ × F · k̂dA
∂R R

for F = (P, Q, 0).

7 Surface Integrals
Definition 7.1 (Flux Integral). The flux integral of a vector field V through a
surface S is the integral of the normal component of V with respect to area over
S ¨
flux = Φ = V · n̂dS.
S

Example 7.1. If V = v is the velocity field of a fluid flow, then the flux is the
volume per second flowing across S.
Example 7.2. If S is a closed surface with no boundary edge and if n̂ is the out-
ward pointing unit normal on S, then the flux is the net loss of fluid from within
S per second.
Example 7.3. If V = E is the electric field vector, then
¨
ΦE = flux of E through S = E · n̂dS.
S

Example 7.4. If V = B is the magnetic flux density (flux per unit area) then
¨
ΦB = flux of B through S = B · n̂dS.
S

Example 7.5. The following is also a surface integral:

¨
f (x, y, z)dS
S

as a special case of the more general form for scalar fields.

39
Example 7.6. If ρ is the mass density of a material per unit area, then
¨
total mass = ρdS.
S

Example 7.7. If σ is the charge density of charge on S in Coulombs per m2 , then

¨
total charge = σdS
S

Note. To evaluate a surface integral, express S in parameter form in terms of two

parameters (u, v) and rewrite the surface integral as a double integral in u and v.
So if S : r = r(u, v), (u, v) ∈ Ruv , then

dS = |ru × rv |dudv
ru × rv
n̂ =
|r × rv |
¨u ¨
S = F · n̂dS = F · dS
¨ S S

= F(r(u, v))(ru × rv )dudv

Ruv

Definition 7.2 (Surface Integral as a Limit of a Sum). Partition the surface S into
n patches of area ∆Si (i = 1, . . . , n). Let ri be any point on the ith patch. Let n̂i
be the normal to the surface at ri . Form the sum
n
X
F(ri ) · n̂i ∆Si .
i=1

Let n → ∞ while all ∆Si → 0. If this limit exists independent of how we let
n → ∞, how max(∆Si ) → 0, of how ri are chosen, and of how S is partitioned,
then the surface integral is
¨ n
X
F · n̂dS ≡ lim F(ri ) · n̂i ∆Si .
n→∞
S max ∆Si → 0 i=1

Lemma 7.1 (Existence and Uniqueness of the Surface Integral). A sufficient con-
dition for the existence and uniqueness of the surface integral is that S be piece-
wise continuously differentiable, that is, piecewise smooth.

40
Definition 7.3 (Jacobian). The Jacobian of the coordinate transformation (u, v) →
(ξ, η) is
 ∂u ∂u 
∂(u, v) ∂ξ ∂η
= det ∂v ∂v
∂(ξ, η) ∂ξ ∂η

Lemma 7.2 (Change of Variables). The change of variables (u, v) → (ξ, η) in a

double integral has the form
¨ ¨
∂(u, v)
f (x, y, z)dudv = f (x(ξ, η), y(ξ, η), z(ξ, η)) dξdη.
Ruv Rξη ∂(ξ, η)

∂(u,v)
where ∂(ξ,η)
is the Jacobian of the transformation.

Note. Recall n̂ · k̂dS = sec γdxdy using (x, y) as parameters. So

¨ ¨
f ds = f sec γdxdy
S Rxy

if (x, y) are suitable parameters on S.

˜
Exercise 7.1. Evaluate I = S x2 zdS where S is the curved cylindrical surface
x2 + y 2 = 1, 0 ≤ z ≤ 1 using cylindrical coordinates (r, θ, z) with r = 1.
Exercise 7.2. Find the centroid z, that is, the position of the center of mass, of a
hemispherical shell of constant density using spherical polar coordinates on S :
r = α.

Definition 7.4 (Solid Angle). The solid angle Ω subtended by a surface S at a

point 0 defined as ¨ ¨
r · n̂dS dS
Ω≡ 3
≡ 2
.
S r sphere r

Definition 7.5 (Extension of Radius). The extension of radius is the ratio of ar-
clength to radius, so that
ˆ ˆ
arclength dS r · n̂ds
extension of radius = = = .
radius circle r C r2
˜
Exercise 7.3. Evaluate the flux integral I = S A· n̂dS for A = 18z î−12ĵ+3y k̂,
where n̂ points upward and S is the plane 2x + 3y + 6z = 12 in the first octant
using (x, y) as surface parameters.

41
8 Orientable Surfaces
Definition 8.1 (Orientable Surface). A surface S is orientable if there exists a
uniquely defined normal n̂ that varies continuously over all S, so that n̂ orients S.
Lemma 8.1. If a surface is orientable it is two-sided. Otherwise S is non-orientable,
and may be one-sided.
Example 8.1 (Plane). An orientable open surface S with boundary ∂S 6= 0.
Example 8.2 (Sphere). An orientable closed surface S with boundary ∂S 6= 0.
Example 8.3 (Mobius Strip). A non-orientable open surface with one edge and
one side.
Example 8.4 (Klein Bottle). A non-orientable closed surface with no edge, one
side, and no inside or outside. One cannot make a Klein bottle in R3 , but can in
R4 , in which case there is no self intersection.
Note. Stokes’ Theorem only applies on orientable surfaces, and does not hold on
a Mobius strip.
Definition 8.2 (Coordinate Free curl). If F is a vector field defined over an open
region of R3 containing a surface ∆S with boundary ∂∆S and outward pointing
unit normal n̂ then the coordinate free definition of curl F is
˛
1
curl F · n̂ = ∇ × F · n̂ ≡ lim F · T̂ds,
∆S→0 ∆S ∂∆S

so for ∆S  1 ˛
curl F · n̂∆S ≈ F · T̂ds.
∂∆S

George Stokes (1819–1903)

42
Theorem 8.1 (Stokes’ Theorem). If S is an orientable surface in R3 which is
piecewise smooth with edge ∂S and unit normal n̂ and if F is a continuously
differentiable vector field in an open region of R3 containing S, then
˛ ¨
F · T̂ds = ∇ × F · n̂dS
∂S S

provided ∂S is oriented with respect to n̂ according to the right hand rule if S is

simply connected, or the total ∂S of S is oriented such that if one traverses the
boundary in the positive sense, with head pointing in the direction n̂, then S is on
one’s left, if S is multiply connected.

Proof. From the coordinate free definition of curl,

˛
1
curl F · n̂ = ∇ × F · n̂ = lim F · T̂ds,
∆S→0 ∆S ∂∆S

which is the circulation per unit area at the point in question in the direction of the
required component of curl F. Of course, to find the whole of curl F we need
its components in three linearly independent directions in R3 , for example, î, ĵ, k̂
directions.
˛
1
(curl F)x = curl F · î = lim F · T̂ds
∆S→0 ∆S ∂S
˛
1
(curl F)y = curl F · ĵ = lim F · T̂ds
∆S→0 ∆S ∂S
˛
1
(curl F)z = curl F · k̂ = lim F · T̂ds
∆S→0 ∆S ∂S

So for ∆S  1
˛
1
curl F · n̂ ≈ F · T̂ds
∆S
˛ ∂S

⇒ F · T̂ds ≈ curl F · n̂∆S

∂∆S

Partition S into n patches of area ∆Si with normals n̂i at some point ri in each
patch i = 1, . . . , n. For each patch
˛
F · T̂ds ≈ curl F|ri · n̂i δSi .
∂∆Si

43
Summing over i = 1, . . . , n:
Xn ˛ n
X
F · T̂ds ≈ curl F(ri ) · n̂i ∆Si .
i=1 ∂∆Si i=1

Now the left hand side is just the circulation around the boundary ∂S because all
interior parts of the line integrals cancel piecewise, so
˛ Xn
F · T̂ds ≈ curl F(ri ) · n̂i ∆Si .
∂S i=1

Now let n → ∞ such that ∆Si → 0. Then the right hand side goes over into a
surface integral by definition, and
˛ ¨
F · T̂ds = curl F · n̂dS.
∂S S

The result is thus true in all coordinate systems, so its form is coordinate invariant.
If S is not simply connected, split S = S1 ∪ S2 and apply Stokes’ Theorem on
each sub-area S1 and S2 , repeating until all Si are simply connected.
Definition 8.3 (Circulation). ˛
F · T̂ds
∂S
is the circulation of the vector field F around ∂S.
Corollary 8.1. The flux of the curl of a vector field over any closed surface S is
zero, that is, ‹
Φ= curl F · n̂dS = 0.
S

Proof. Let F be a vector field and S be a closed surface. Then by definition of S,

∂S = ∅. Applying Stokes’ Theorem, the flux of the curl of F is
‹ ˛
curl F · n̂dS = F · T̂ds = 0.
S ∂S=∅

Lemma 8.2. ¨ ¨
curl ·F · n̂dS = ∇ × F · n̂dS
S S
is a surface-independent surface integral.

44
Proof. Let S1 and S2 be two surfaces bound by the same curve C, and orient S1
and S2 consistently with the orientation of C. Applying Stokes’ Theorem on the
closed surface S1 ∪ S2 , by Lemma 8.1
‹
curl F · n̂dS = 0
¨ ¨ S1 ∪S2

⇒ curl F · n̂1 dS + curl F · (−n̂1 )dS = 0

S1 S2
¨ ¨
⇒ curl F · n̂1 dS = curl F · n̂1 dS.
S1 S2

Since S1 and S2 were arbitrarily chosen surfaces oriented by the curve C, this is a
surface-independent surface integral.

Lemma 8.3. If F is irrotational, curl F ≡ 0, and

¨
curl F · n̂dS = 0.
S

Lemma 8.4. If curl F 6= 0 then

¨ ˛
curl F · n̂dS 6= 0 and F · dr 6= 0.
S ∂S

Definition 8.4 (Piecewise Smooth Boundary). Let V be a region of R3 , then ∂V

is a piecewise smooth boundary of V if ∂V is one or more closed surfaces.

Example 8.5. A closed outer surface and “holes” inside V .

Definition 8.5 (Coordinate Free div). Let a volume ∆V be bounded by the closed
surface ∂∆V with outward pointing unit normal n̂ on ∂∆V , then
‹
1
div F = ∇ · F ≡ lim F · n̂dS,
∆V →0 ∆V ∂∆V

the outward flux per unit volume at a point, so for ∆V  1

‹
div F∆V ≈ F · n̂dS.
∂∆V

45
Carl Friedrich Gauss (1777–1855)

Theorem 8.2 (Gauss’ Divergence Theorem). Let V be a region of R3 with a

piecewise smooth boundary ∂V . Orient ∂V using the outward unit normal n̂
on all parts of ∂V . Assume F is continuously differentiable in an open region of
R3 containing V and ∂V , then
‹ ˚ ˚
F · n̂dS = ∇ · FdV = div FdV.
∂V V V

Proof. Partition V into n sub-regions of volumes ∆Vi , i = 1, . . . n. On each

sub-region, by the coordinate-free definition of div with ∆Vi  1
‹
div F(ri )∆Vi ≈ F · n̂i dS,
∂∆Vi

where ri is any point inside ∆Vi , for all i ∈ {1, . . . , n}. Summing from 1 to n:
n
X n ‹
X
div F(ri )∆Vi ≈ F · n̂i dS.
i=1 i=1 ∂∆Vi

The right hand side is just the flux outward through the boundary of S itself be-
cause on all interior surfaces the flux out of one patch equals the flux into the
adjacent patch, so that the cancel piecewise, yielding
n
X ‹
div F(ri )∆Vi ≈ F · n̂dS
i=1 ∂V

46
where n̂ is the outward normal on the whole of ∂V . Letting n → ∞ such that all
∆Vi → 0, by definition of the triple integral we get
˚ ˚ ‹
div FdV = ∇ · FdV = F · n̂dS.
V V ∂V

Note. The proof of Gauss’ Divergence Theorem is coordinate independent, so

Gauss’ Divergence Theorem is true in any coordinate system.
Note. Gauss’ Divergence Theorem does not apply to a Klein bottle since it is a
one-sided surface.
Note. The left hand side of Gauss’ Divergence Theorem is just the flux of F
through ∂V , where “outward” means “away from V ”.
Corollary 8.2. If F is solenoidal then
‹
F · n̂dS = 0,
∂V
Corollary 8.3. The flux through any closed surface vanishes.
Proof. Replace F with curl G = ∇ × G in Gauss’ Divergence Theorem. Then
because div curl G = ∇ · ∇ × G ≡ 0,
‹ ˚
∇ × G · n̂dS = ∇ · (∇ × G)dV = 0
∂V V
as found from Stokes’ Theorem applied to a closed surface.
Note. Correspondence between path independent line integrals and surface Inde-
pendence of surface integrals:
Theorem 8.3 (Path Independence). On simply-connected regions
ˆ b
C F · dr is path independent
˛ a
⇔ F · dr = 0
C
⇔ curl F = ∇ × F ≡ 0 (irrotational)
⇔ F = ± grad φ = ±∇φ (scalar potential)
⇔ F · dr is an exact differential ± dφ
⇔ F · dr = 0 is an exact differential equation
⇔ F is a conservative vector field

47
Reason. curl grad φ = ∇ × ∇φ ≡ 0
Proof. By Stokes’ Theorem if S is any piecewise smooth surface with boundary
C then ˛ ¨
F · dr = curl F · n̂dS = 0
C S
if curl F ≡ 0.

Theorem 8.4 (Surface Independence).

¨
F · n̂dS is surface independent
‹ S

⇔ F · n̂dS = 0
⇔ div F = ∇ · F ≡ 0 (solenoidal)
⇔ F = curl A = ∇ × A (vector potential)

Reason. div curl A = ∇ · ∇ × A ≡ 0

Proof.

1. By Gauss’ Divergence Theorem

‹ ˚
F · n̂dS = div FdV ≡ 0
∂V V

if div F = ∇ · F ≡ 0.

2. By Stokes’ Theorem applied to A on a closed surface S

˛ ‹ ‹
0= A · dr = curl A · n̂dS = F · n̂dS.
∂S=∅ S S

48
Hermann von Helmholtz (1821-1894)
Theorem 8.5 (Helmholtz’s Decomposition Theorem). Any vector field F which
is finite, uniform, continuous, and vanishes at infinity can be expressed as the sum
of the gradient of a scalar field, i.e., the scalar potential, ±φ and the curl of a
vector field, i.e., the vector potential A of zero divergence

F = ± grad φ + curl A = ±∇φ + ∇ × A,

with div A = ∇ · A ≡ 0. If the integral of F over all of R3 is finite then φ and A

are unique.
Note. F is the sum of an irrotational part (±∇φ) which contributes only diver-
gence, and a solenoidal part (∇ × A) which contributes only curl.
Note. The choice of restriction div A = ∇ · A ≡ 0 is one of many possible
restrictions. It can always be made without loss of generality.
Lemma 8.5. Let F = ±∇φ+∇×A, then div F = ±∇2 φ and curl F = −∇2 A.
Proof.

div F = ∇ · (±∇φ + ∇ × A) = ±∇ · ∇φ + ∇ · ∇ × A = ±∇2 φ

since div curl A ≡ 0 by definition.

curl F = ∇ × (±∇φ + ∇ × A) = ±∇ × ∇φ + ∇ × ∇ × A = −∇2 A

since curl grad φ ≡ 0 and curl curl A = grad div A − ∇2 A = −∇2 A since
div A ≡ 0.

49
Note. These equations suggest that a 4D formulation, that is, a space-time formu-
lation, will allow a combination into one operator.

Theorem 8.6 (Gauss’ Divergence Theorem).

˚ ‹
div FdV = F · n̂dS
V ∂V

Proof. (Gauss’ Divergence Theorem for regions V which are xy-simple)

Let z = z1 (x, y) be the lower boundary surface of V and z = z2 (x, y) be the
upper boundary of V . Let Rxy be the projection of V onto the xy-plane. Then
x, y are suitable surface parameters.
Now F = (F1 , F2 , F3 ) ⇒

∂F1 ∂F2 ∂F3

div F = ∇ · F = + +
∂x ∂y ∂z
and F · n̂dS = ±F1 dydz ± F2 dxdz ± F3 dxdy so that
‹ ¨ ¨ ¨
F · n̂dS = ± F1 dydz ± F2 dxdz ± F3 dxdy.
∂V ∂V ∂V ∂V

So Gauss’ Divergence Theorem can be reformulated as

˚   ¨ ¨ ¨
∂F1 ∂F2 ∂F3
+ + dxdydz = ± F1 dydz± F2 dxdz± F3 dxdy
V ∂x ∂y ∂z ∂V ∂V ∂V

Now F1 , F2 , F3 are arbitrary independent functions. So we can prove Gauss’

Divergence Theorem by separately proving the three results
¨ ˆ  ¨
∂F1
dx dydz = ± F1 dydz (13)
∂x
¨ ˆ
V
 ¨ ∂V
∂F2
dy dydz = ± F2 dxdz (14)
∂y
¨ ˆ
V
 ¨ ∂V
∂F3
dz dydz = ± F3 dxdy (15)
V ∂z ∂V

50
We will prove (15), as the proofs of (13) and (14) follow similarly.
˚ ¨ ˆ z=z2 (x,y)
∂F3 ∂F3
dzdxdy = dxdy dz
V ∂z Rxy| z=z1 (x,y) ∂z
¨
= [F3 (x, y, z2 (x, y)) − F3 (x, y, z1 (x, y))] dxdy
Rxy
¨ ¨
= F3 dxdy − F3 dxdy
z=z2 (x,y) z=z1 (x,y)
‹
= F3 dxdy
∂V

since the angle between n̂ and k̂ is acute on z = z2 (x, y) and obtuse on z =

z1 (x, y).
Exercise 8.1. Verify Gauss’ Divergence Theorem for F = 4xz î − y 2 ĵ + yz k̂ and
V the cube bounded by x = 0, x = 1, y = 0, y = 1, z = 0, z = 1.
Note. An arbitrary vector field F contains three arbitrary functions F1 , F2 , and
F3 . The set {φ, A} contains four functions φ, A1 , A2 , and A3 . Hence we can
impose one functional relation involving φ and A without loss of generality. In
Helmholtz’s Theorem we restrict A by div A = ∇ · A ≡ 0
∂A1 ∂A2 ∂A3
⇔ + + ≡ 0.
∂x ∂y ∂z
Even more so when dealing with electromagnetism, we have two vector fields
E and B, which consist of six independent functions. Maxwell’s Equations are
a set of two scalar and two vector equations ⇒ eight scalar equations for six
unknown functions E1 , E2 , E3 , B1 , B2 , B3 . Hence Maxwell’s Equations are over-
determined.
If we reformulate Maxwell’s Equations in terms of an electrostatic scalar po-
tential φ and a magnetic vector potential A, we have only four unknowns. In fact,
the E and B fields are not independent, that is, there are actually only four inde-
pendent functions in an electromagnetic field. The form of the equations involving
the potentials is the simplest and of exactly right number.
A is determined only up to an arbitrary gradient:

F = curl(A + grad φ) = curl A

This freedom is called gauge freedom.

51
Exercise 8.2. Verify Stokes’ Theorem for F = (2x − y)î − yz 2 ĵ − y 2 z k̂ where S
is the upper 21 hemisphere x2 + y 2 + z 2 = 1.
Exercise 8.3. Use Stokes’ Theorem to evaluate
¨
I= curl F · n̂dS,
S

where n̂ is the upward pointing normal, F = (x2 + y − 4)î + 3xy ĵ + (2xz + z 2 )k̂
and S is the curved surface of the paraboloid z = 4−(x2 +y 2 ) above the xy-plane,
so z = 4 − r2 .
Exercise 8.4. Evaluate ¨
I= r · n̂dS
S

where S is the sphere of radius α with center (0, 0, 0).

Exercise 8.5. Evaluate ˚
I= div FdV
V

where F = (x2 + y 2 + z 2 )(x, y, z) and V : x2 + y 2 + z 2 ≤ α2 .

9 Maxwell’s Equations
Theorem 9.1 (Electrostatic Force). The electrostatic force F between two charges
q1 and q2 separated by a distance r is the central force
q1 q 2 r
F= .
4πεr2 r
Definition 9.1 (Electric Field). The electric field vector E due to a charge q1 where
q2 ≈ 0 is the test charge is
F q1
E= = r.
q2 4πεr3
Definition 9.2 (Permittivity). The permittivity of the medium ε = ε0 εr , where
ε0 ≈ 8.854 × 1012 F/m is the permittivity of free space and εr is the relative
permittivity of the medium.
Note. One cannot apply Gauss’ Divergence Theorem on a surface S if S encloses
a charge q. Similarly, E has a singularity at r = 0.

52
Physical Law 1 (Gauss’ Law of Electrostatics). The flux of the E field ΦE out-
ward through any closed surface S equals the ratio of the total charge q contained
within S to the permittivity ε
‹
q
ΦE = E · n̂dS = .
S ε

Proof. Let S 0 be a spherical surface enclosing q, of radius r, lying inside S, and

centered on q. Apply Gauss’ Divergence Theorem in a region V between S and
S0
˚ ¨ ‹ ‹
div EdV = E · n̂dS = E · n̂dS + E · n̂dS.
V ∂V S0 S

But for inverse square laws, which E obeys, div E = ∇ · E ≡ 0 everywhere that
r 6= 0. So the left hand side is 0.
‹ ‹ ‹
qr  r 
⇒ E · n̂dS = − E · n̂dS = − 3
− dS
S S0 ‹ S 0 4πεr r
q
= dS, since r · r = r2
4πεr2 S 0
q q
= 2
4πr2 =
4πεr ε
So we obtain Gauss’ Law of Electrostatics:
‹
q
E · n̂dS = .
S ε

By linearity, each charge inside a surface S contributes its flux over S, so simply
add to obtain the total flux.
Recall. The definition of solid angle, this says if q is inside S then S subtends a
solid angle 4π at q, and if q is outside S, the flux through S is zero.

Maxwell’s 1st Equation (Gauss’ Law of Electrostatics). Let ρ be the charge den-
sity within a medium with permittivity ε, then the electric field E satisfies
ρ
div E = ∇ · E = . (M1)
ε

53
Proof. Let ρ be the charge density of a distribution of charges inside S, where the
charge in any small volume q = ρ∆V → 0 as the volume ∆V → 0. This means
there are no field singularities in S. Then by Gauss’ Law of Electrostatics
‹ ‹ ˚
1
E · n̂dS = E · n̂dS = ρdV.
S ∂V ε V

Since there are assumed to be no singularities, we may apply Gauss’ Divergence

Theorem to V
˚ ˚
1
⇒ div EdV = ρdV
ε
˚ nV V
ρo
⇒ div E − dV ≡ 0.
V ε
ρ
But V is arbitrary, so the integrand itself must be zero: ⇒ ∇ · E =
ε
Note. From the definition of the E field, ∇ × E ≡ 0 for all r. This implies
there exists a scalar electric potential φ such that E = −∇φ . So Maxwell’s
1st Equation (M1) can be rewritten in terms of the electrostatic potential φ as
Poisson’s Equation in regions with charge ρ
ρ
∇2 φ = − ,
ε
and Laplace’s Equation in empty regions

∇2 φ = 0 .
Theorem 9.2 (Deformation). If div F = ∇ · F ≡ 0 in a region V between two
closed surfaces S and S 0 then
‹ ‹
F · n̂dS = F · n̂dS.
S S0

Theorem 9.3 (Conservation of Flux). A closed surface S may be deformed through

regions in which div F ≡ 0 without changing the flux integral of F.
Physical Law 2 (Gauss’ Law of Magnetostatics). The flux of the B field ΦB is
zero through every closed surface S
‹
ΦB = B · n̂dS = 0.
S

54
Note. This implies there are no isolated magnetic north or south poles, that is,
there are no magnetic monopoles.

Maxwell’s 2nd Equation (Gauss’ Law of Magnetostatics). The magnetic B sat-

isfies
div B = ∇ · B = 0 (M2)

Proof. Let S be a closed surface enclosing the volume V , then by definition of

flux and Gauss’ Divergence Theorem
‹
ΦB = B · n̂dS = 0 by Gauss’ Law of Magnetostatics
˚∂V
⇒ ∇ · BdV = 0 by Gauss’ Divergence Theorem
V

But V is arbitary, so the integrand itself must be zero: ⇒ ∇ · B ≡ 0 .

Corollary 9.1. The flux of B is the same through all surfaces S having the same
boundary curve C.

Proof. Let S and S 0 be two surfaces with the same boundary curve C. Orient
S and S 0 consistent with the orientation of C. Let V = S ∪ S 0 be the volume
enclosed by the closed surface S ∪ S 0 , then applying Corollary 9
‹ ¨ ¨
0= B · n̂dS = B · n̂dS + B · (−n̂0 )dS
S∪S 0 S S0

¨ ¨
⇒ B · n̂dS = B · n̂0 dS .
S S0

Corollary 9.2 (Magnetic Vector Potential). There exists a magnetic vector poten-
tial A such that B = curl A = ∇ × A.

Proof. This is a consequence of Gauss’ Law of Magnetostatics, Maxwell’s 2nd

Equation (M2) div B = ∇ · B = 0.

Definition 9.3 (Flux of E). The flux of E through a surface S with unit normal n̂
is ¨
ΦE ≡ E · n̂dS.
S

55
Definition 9.4 (Flux of B). The flux of B through a surface S with unit normal n̂
is ¨
ΦB ≡ B · n̂dS.
S

Definition 9.5 (Magnetic Force). The magnetic force F on a charge q moving

with velocity v is
F = qv × B.

Definition 9.6 (Permeability). The permeability of the medium µ = µ0 µr , where

µ0 ≡ 4π × 10−7 H/m is the permeability of free space and µr is the relative
permeability of the medium.

Physical Law 3 (Faraday’s Law of Induction). A varying magnetic field B in-

duces an electric field E, relating changing flux of B, ΦB , through a surface S to
the circulation of E around ∂S
˛ ¨
d dΦB
E · dr = − B · n̂dS ≡ − .
∂S dt S dt

Maxwell’s 3rd Equation (Faraday’s Law of Induction). The electric and magnetic
fields E and B satisfy
∂B
curl E = ∇ × E = − . (M3)
∂t
Proof.
¨ ˛
∇ × E · n̂dS = E · dr by Stokes’ Theorem
S
¨
∂S
d
= − B · n̂dS by Faraday’s Law of Induction
dt S
¨
∂B
= − · n̂dS by Leibniz’s formula
S ∂t
¨  
∂B
⇒ ∇×E+ · n̂dS ≡ 0.
S ∂t
∂B
But S is arbitary, so the integrand itself must be zero: ⇒ ∇ × E = − .
∂t

56
Note. So the scalar potential equation E = −∇φ has to be modified in the pres-
ence of varying magnetic fields, since ∇ × E = −∇ × ∇φ = 0 = ∂B ∂t
⇒
B = constant.
Corollary 9.3 (Electric Scalar Potential). There exists a electric scalar potential
φ such that
∂A ∂A
E = − grad φ − = −∇φ − .
∂t ∂t
Proof. This is a consequence of Gauss’ Law of Electrostatics, Maxwell’s 1st and
3rd Equations (M1) and (M3).
Physical Law 4 (Ampère’s Law). A varying electric field E induces a magnetic
field B, relating changing flux of E, ΦE , through a surface S and the electric
current I to the circulation of B around ∂S
˛ ¨
1 d dΦE
B · dr = I + ε E · n̂dS ≡ I + ε .
µ ∂S dt S dt

Note. The dependence on the change in electric flux ε dΦdtE in Ampère’s Law was
introduced by Maxwell.
Maxwell’s 4th Equation (Ampère’s Law). The electric field E, magnetic field B,
and current density J satisfy
 
∂E
curl B = ∇ × B = µ J + ε . (M4)
∂t
Proof.
¨ ˛
∇ × B · n̂dS = B · dr by Stokes’ Theorem
S ∂S
¨
d
= I +ε E · n̂dS by Ampère’s Law
dt S
¨ ¨
∂E
= J · n̂dS + ε · n̂dS by Leibniz’s formula
S S ∂t
¨  
∂E
⇒ ∇×E−J− · n̂dS ≡ 0.
S ∂t
1 ∂E
But S is arbitary, so the integrand itself is zero: ⇒ ∇×B=J+ε .
µ ∂t

57
Lemma 9.1 (Conservation of Charge). Let ρ be the charge density and v be the
charge velocity, so J ≡ ρv is the current density, then
∂ρ ∂ρ
+ div(ρv) = + ∇ · (ρv) = 0
∂t ∂t
∂ρ ∂ρ
+ div J = +∇·J=0
∂t ∂t
Proof. Taking the divergence of both sides of Ampère’s Law, (M4) we obtain
1 ∂E
div curl B = div J + div ε
µ ∂t
1 ∂
∇ · ∇ × B = ∇ · J + ∇ · εE
µ ∂t
Substituting Gauss’ Law of Electrostatics (M1) ∇ · εE = ρ and ∇ · ∇ × B ≡ 0,
∂ρ
we find 0 = ∇ · J + .
∂t
Lemma 9.2 (Wave Equation). Any function of the form f (x − ct) satisfies the
wave equation
∂ 2f 2
2∂ f
=c ,
∂t2 ∂x2
in 1D where c2 = ε0 µ0 are travelling waves.
Definition 9.7 (Coulomb Gauge).
div A = ∇ · A ≡ 0.
Definition 9.8 (Lorentz Gauge).
∂φ ∂φ
div A + µε = ∇ · A + µε = 0.
∂t ∂t
Definition 9.9 (Lorentz Force). The total force on a charge q moving with velocity
v due to both electric and magnetic fields E and B is the Lorentz force
F = q(E + v × B)
 
∂A
F = q −∇φ − +v×∇×A
∂t
˚
F = ρ (E + v × B) dV
V

for distributed charges.

58
Note. For a point charge q at the origin of an otherwise empty region,
˚
ρ(r)dV = q
V

for an arbitrary volume V of R3 centered on the origin, if V is the sphere of radius

α, then taking the limit α → 0+
¨ ˚ ˚
q q 2
ρ(r)dV ≈ dV = 4 3
r sin ϕdrdϕdθ
V V V V 3 πα
ˆ 2π ˆ π ˆ α 3 α
3qr2 π qr
= 3
sin ϕdrdϕdθ = 2π[− cos ϕ] 0 =q
0 0 0 4πα 4πα3 0

Definition 9.10 (Dirac δ Function). If f : R3 → R, then the Dirac δ-function

satisfies ˚ ∞
f (r)δ(r − r0 )dV = f (r0 ).
−∞

Corollary 9.4.
˚ ∞ ˆ ∞ ˆ ∞ ˆ ∞
δ(r)dV = δ(x)dx δ(y)dy δ(z)dz = 1
−∞ −∞ −∞ −∞

Corollary 9.5. 
∞ if r = 0
δ(r) =
0 otherwise
 2
 √
Exercise 9.1. Show the normalized Gaussian function exp − |r−r 0|
2∆2
/ 2π∆2
behaves as a Dirac δ-function in the limit ∆ → 0+ , so that
|r−r0 |2
e− 2∆2
√ → δ(r − r0 ) as ∆ → 0+ .
2π∆
Definition 9.11 (Poynting Vector). The Poynting vector S measures the power
flow in the electromagnetic wave
1
S≡ E×B
µ0

in W/m2 .

59
10 The Heat Equation
Definition 10.1 (Specific Heat). The specific heat of a body C is the heat energy
required to raise one kilogram of material by one degree, that is, in J/(kg ◦ C).
Definition 10.2 (Thermal Energy). The thermal energy of a body E is
˚
E= CT ρdV
V

in J where C is the specific heat, T is the absolute temperature in K, and ρ is the

mass density in kg/m3 .
Definition 10.3 (Thermal Conductivity). The thermal conductivity of the material
k is the rate of heat flow in J/(s m ◦ C).
Definition 10.4 (Thermal Diffusivity). The thermal diffusivity of a material a is
k
a2 =
Cρ
in m2 /s where C is the specific heat, k is the thermal conductivity, and ρ is the
mass density of the material.
Definition 10.5 (Temperature Gradient). The temperature gradient ∇T is the
change in temperature with position in the direction of greatest increase of T .
Theorem 10.1 (Time-Dependent Heat Equation in R3 ). The time dependent heat
equation in three dimensions for a body with thermal conductivity k, specific heat
C, and mass density ρ is
∂(CρT )
− ∇ · (kT ) ≡ 0.
∂t
k
For a homogeneous body with thermal diffusivity a2 = Cρ
, this simplifies to

∂T
− a2 ∇2 T ≡ 0.
∂t
Proof. Consider the flow of heat energy in a region of volume V in R3 . The ther-
mal energy associated with an object is proportional to its absolute temperature
T in K. If thermal energy is conserved, then the heat flow out of V per unit time
corresponds to the temperature decrease of the object.

60
 Let C be the specific heat of the body, ρ be the mass density of the material, T
be the temperature, and E be the thermal energy of the body. Then by Definition
10.2, ˚
E= CT ρdV.
V
The rate of change of heat energy with respect to time is
˚
dE d
Ė = = CρT dV.
dt dt V

This rate of change must correspond to the heat flow accross the boundary of V ,
∂V , that is the heat flow into/out of V .
The outward flux through ∂V is
‹
k (−∇T ) · n̂dS
∂V

per unit time. Here ∇T is the temperature gradient, and heat flows in the opposite
direction, from hot to cold, i.e., in the direction of greatest rate of decrease of T ,
and k is the thermal conductivity of the material.
Conservation of heat energy requires
‹
dE
= − k(−∇T ) · n̂dS
dt
˚ ¨ ∂V
d
⇒ CρT dV = k∇T · n̂dS
dt
˚ V
˚ ∂V
∂
CρT dV = ∇ · (k∇T )dV by Gauss’ Divergence Theorem
V ∂t V
˚  
∂
⇒ (CρT ) − ∇ · (k∇T ) dV ≡ 0.
V ∂t
∂
Since V is arbitrary ⇒ (CρT ) − ∇ · (k∇T ) ≡ 0 throughout the body.
∂t
If C, ρ, k are constants, independent of position in the body, that is, the body
is homogeneous
∂T
Cρ − k∇ · ∇T = 0
∂t
∂T k 2
⇒ − ∇T ≡ 0
∂t Cρ

61
 ∂T
Using the definition of thermal diffusivity a2 = k
cρ
, ⇒ − a2 ∇2 T ≡ 0 .
∂t
Note. The heat equation is the diffusion equation for diffusion of heat energy in a
body. It is a parabolic, 2nd -order partial differential equation for T (x, y, z, t).

Lemma 10.1 (Heat Equation under Steady-State Conditions). The heat equation
under steady-state conditions in a volume V is Laplace’s Equation ∇2 T ≡ 0.

Proof. Under steady-state conditions, T is no longer time dependent, and thermal

equilibrium has been reached. This mean ∂T∂t
≡ 0, and the Heat Equation reduces
to the time independent equation ∇2 T ≡ 0 in V .
Exercise 10.1. Find the general solution to the Heat Equation in one space dimen-
sion x for heat diffusion along a rod

∂T ∂ 2T 0≤x≤L
= a2 2
∂t ∂x 0≤t

for T = T (x, t) in the rod, subject to the initial conditions:

IC: T (x, 0) = f (x) = initial temperature distribution in rod 0 ≤ x ≤ L.

and boundary conditions:

T (0, t) = T1
BC: t ≥ 0.
T (L, t) = T2

This is a well posed boundary value and initial value (BV/IV) problem for the
partial differential equation (PDE). It has a unique solution.

Definition 10.6 (Separable). A function T (x, t) is separable if there exist func-

tions X(x) and Y (t) such that T (x, t) = X(x)Y (t).

Definition 10.7 (Fourier Sine Series). The Fourier sine series of a function f (x)
for 0 ≤ x ≤ L is given by
∞
X  nπx 
f (x) = Bn sin .
n=1
L

where Bn are the Fourier sine coefficients of f (x).

62
Definition 10.8 (Complex Conjugate). Let z = x + iy where x, y ∈ R and i2 ≡
−1, so that z ∈ C, then the complex conjugate of the complex number z, denoted
z ∗ , is z ∗ = x − iy.
Lemma 10.2. Let z = x + iy where x, y ∈ R and i2 ≡ −1, then zz ∗ = |z|2 .
Proof. zz ∗ = (x + iy)(x − iy) = x2 + ixy − ixy + i(−i)y 2 = x2 + y 2 = |z|2 .
Definition 10.9 (Inner Product). An inner product hf, gi of f and g has conjugate
symmetry, is linear, and is positive definite, that is

hf, gi = hg, f i∗ (16)

hαf, gi = αhf, gi where α ∈ R (17)
hf + g, hi = hf, hi + hg, hi (18)
hf, f i ≥ 0 (19)
hf, f i = 0 ⇔ f = 0. (20)

Lemma 10.3 (Inner Product of Functions).

ˆ L
hf, gi ≡ f ∗ (x)g(x)dx
0

is an inner product of functions f and g on [0, L].

Proof.
ˆ L ∗ ˆ L
∗ ∗
hg, f i = g (x)f (x)dx = f ∗ (x)g(x)dx = hf, gi
0 0
ˆ L ˆ L
hαf, gi = (αf (x))∗ g(x)dx = α f ∗ (x)g(x)dx = αhf, gi
0 0
ˆ L
hf + g, hi = (f (x) + g(x))∗ h(x)dx
0
ˆ L ˆ L
∗
= f (x)h(x)dx + g ∗ (x)h(x)dx = hf, hi + hg, hi
0 0
ˆ L
hf, f i = f ∗ (x)f (x)dx ≥ 0
0
hf, f i = 0 ⇔ f (x) = 0 for all x ∈ [0, L] ⇔ f = 0 on [0, L]

63
Definition 10.10 (Kronecker δ). The Kronecker delta δnm is defined as

0 if n 6= m,
δnm =
1 if n = m.
where n, m ∈ Z are integers.
Definition 10.11 (Orthonormal Basis). A set of functions {ψn } forms an or-
thonormal basis if hψn , ψm i = δnm .
nq o
2 nπx
Lemma 10.4 (Sine Orthonormal Basis). The set of sine functions L
sin L
forms an orthonormal basis.
Proof.
*r r +
2 nπx 2 mπx
hψn , ψm i = sin , sin
L L L L
ˆ
2 L nπx mπx
= sin sin
L 0 L L
ˆ L
1 (m − n)πx (m + n)πx
= cos − cos dx
L 0 L L
where we have used the sine product identity sin A sin B = 12 [cos(A−B) – cos(A+
B)].
 L
1 L (m − n)πx L (m + n)πx
hψn , ψm i = sin − sin
L (m − n)π L (m + n)π L 0

sin(m − n)π 0 if m 6= n,
= = ,
m−n 1 if m = n.
*r r +
2 nπx 2 mπx
so that sin , sin = δnm as required.
L L L L

Lemma 10.5 (Fourier Sine Series Coefficients). The coefficients of the Fourier
sine series Bn of a function f (x) for 0 ≤ x ≤ L are
ˆ
2 L  nπx 
Bn = f (x) sin dx,
L 0 L
where n ∈ N.

64
Proof. From the definition of the Fourier sine series
∞
X  nπx 
f (x) = Bn sin . (21)
n=1
L

Taking the inner product of f (x) with sin mπx

L
, we find
* ∞
+
D mπx E mπx X nπx
sin , f (x) = sin , Bn sin
L L n=1 L
∞ D mπx
X nπx E
= Bn sin , sin
n=1
L L
∞
X L L
= Bn δnm = Bm ,
n=1
2 2
ˆ L
2 mπx
⇒ Bm = sin f (x)dx .
L 0 L

Exercise 10.2. Suppose f (x) = T0 , a “square wave” on the rod, is the initial
condition for the standard homogeneous heat problem in 1D (T (0, t) = T (L, t) =
0). Find the Fourier sine series for f (x) and the temperature T (x, t) for 0 ≤ x ≤ L
and t ≥ 0, and show that
4T0 −( πa 2 πx
T (x, t) ≈ e L ) t sin .
π L
Definition 10.12 (Periodic Extension). If the Fourier series of a function f (x) for
0 ≤ x ≤ L converges, it converges to the periodic extension of f (x), fˆ(x) : R →
R.
Lemma 10.6 (Continuity of Fourier Sine Series). The Fourier sine series fˆ(x) of
f (x) for 0 ≤ x ≤ L is continuous if f (0) = 0 = f (L).
Definition 10.13 (Pointwise Convergence). A series ∞
P
n=1 fn (x) converges point-
wise on (a, b) if for any  > 0 and x ∈ (a, b) there exists an N (x) and δ such that
P∞ PN
|x − x0 | < δ implies n=1 f n (x) − n=1 fn (x0 ) <  for all N ≥ N (x).

Definition 10.14 (Uniform Convergence). A series ∞

P
n=1 fn (x) converges uni-
formly on [a, b] if for any  > 0 there exists an N and δ such that |x − x0 | < δ
P∞ PN
implies n=1 fn (x) − n=1 fn (x0 ) <  for all N ≥ N and x ∈ [a, b].

65
 P∞
Definition 10.15 (Absolute
P∞ Convergence). A series n=1 fn (x) converges abso-
lutely for x ∈ [a, b] if n=1 |fn (x)| also converges.

Theorem 10.2 (Alternating Series Test). Let {Bn } be a sequence for which Bn ≥
0, Bn+1 ≤ Bn , and Bn → 0 as n → ∞, then the alternating series
∞
X
(−1)n Bn = −B1 + B2 − B3 + B4 − B5 + · · ·
n=1

converges, and an upper bound on the error for the N th partial sum is BN +1 , i.e.,
∞
X N
X
n
(−1) Bn − (−1)n Bn ≤ BN +1 .
n=1 n=1

Theorem 10.3 (Weierstrass M-Test). Suppose {fn (x)} is a sequence of functions

defined for x ∈ [a, b] and {Mn } is a sequence P∞of positive numbers such that for
all
P∞ n ≥ 1 and x ∈ [a, b] |f n (x)| ≤ Mn and n=1 Mn converges, then the series
n=1 fn (x) converges absolutely and uniformly.

Definition 10.16 (General Fourier Series). The general Fourier series of a func-
tion f (x) for −L ≤ x ≤ L is
∞
X  nπx   nπx 
f (x) = A0 + An cos + Bn sin ,
n=1
L L

where
ˆ ˆ
1 L  nπx  1 L
An = cos f (x)dx, A0 = f (x)dx,
L −L L 2L −L
ˆ
1 L  nπx 
Bn = sin f (x)dx
L −L L

are the Fourier coefficients of f (x).

Lemma 10.7
n (Cosine and Sine Orthonormal
o Basis). The set of cosine and sine
1 nπx 1 mπx
functions √L cos L , √L sin L form an orthonormal basis on [−L, L]

66
 n o
Proof. To demonstrate √1L cos nπx L
, √1 sin mπx
L L
form an orthonormal basis on
[−L, L] it is sufficient to show
 
1 nπx 1 mπx
√ cos , √ sin = 0 (22)
L L L L
 
1 nπx 1 mπx
√ cos , √ cos = δnm , (23)
L L L L
 
1 nπx 1 mπx
√ sin , √ sin = δnm (24)
L L L L

Beginning with (22)

 
1 nπx 1 mπx
√ cos , √ sin
L L L L
ˆ L
1 nπx mπx
= cos sin dx
L −L L L
ˆ L
1 (n + m)πx (n − m)πx
= sin − sin dx
2L −L L L

where we have used the cosine sine product identity cos A sin B = 12 [sin(A + B)
– sin(A − B)].
 L
1 1 (n + m)πx 1 (n − m)πx
= − cos + cos
2π n+m L n−m L
 n+m n+m n−m n−m
 −L
1 (−1) − (−1) (−1) − (−1)
= − + = 0,
2π n+m n−m

as required.
Next demonstrating (23)
 
1 nπx 1 mπx
√ cos , √ cos
L L L L
ˆ L
1 nπx mπx
= cos cos dx
L −L L L
ˆ L
1 (n + m)πx (n − m)πx
= cos + cos dx
2L −L L L

67
where we have used the cosine product identity cos A cos A = 12 [cos(A + B) +
cos(A − B)].
 L
1 1 (n + m)πx 1 (n − m)πx
= sin + sin
2π n + m L n−m L −L
 
1 sin(n + m)π − sin(n + m)π sin(n − m)π − sin(−n + m)π
= +
2π n+m n−m
sin(n − m)π
=
(n − m)π
( 0
(n−m)π
= 0 if n 6= m
= (n−m)π
(n−m)π
= 1 if n = m

where we have used the Taylor expansion for sin(n − m)π ≈ (n − m)π.
Finally, demonstrating (24)
 
1 nπx 1 mπx
√ sin , √ sin
L L L L
ˆ L
1 nπx mπx
= sin sin dx
L −L L L
ˆ L
1 (n − m)πx (n + m)πx
= sin − cos dx
2L −L L L

where we have used the cosine product identity sin A sin A = 12 [cos(A − B) –
cos(A + B)].
 L
1 1 (n − m)πx 1 (n + m)πx
= sin − sin
2π n − m L n+m L −L
 
1 sin(n − m)π − sin(−n + m)π sin(n + m)π − sin(n + m)π
= −
2π n−m n+m
sin(n − m)π
=
(n − m)π
( 0
(n−m)π
= 0 if n 6= m
= (n−m)π
(n−m)π
= 1 if n = m

where we have again used the Taylor expansion for sin(n−m)π ≈ (n−m)π.

68
Definition 10.17 (Even and Odd Functions). A function f : R → R is even if
f (−x) = f (x) for all x ∈ R, and is odd if f (−x) = −f (x) for all x ∈ R.

Lemma 10.8 (Even and Odd Decomposition). A function f : R → R can be

decomposed into an even part fe (x) and odd part fo (x), so f (x) = fe (x) + fo (x),
where
f (x) + f (−x) f (x) − f (−x)
fe (x) = and f0 (x) = .
2 2
Proof. First, show an arbitrary function can be expressed as a sum of the so-
defined fe (x) and fo (x)

f (x) + f (x) + f (−x) − f (−x)

f (x) =
2
f (x) + f (−x) f (x) − f (−x)
= +
2 2
= fe (x) + fo (x).

Next, we must show that fe (x) is an even function of x.

f (−x) + f (− − x) f (x) + f (−x)

fe (−x) = = = fe (x),
2 2
so fe (x) is an even function of x by definition. Next we must show that fo (x) is
an odd function of x.
f (−x) − f (− − x) f (x) − f (−x)
fo (−x) = =− = −fo (x),
2 2
so fo (x) is an odd function of x by definition. Thus, since f (x) = fe (x) + fo (x),
we can decompose an arbitrary function into even and odd parts.

Lemma 10.9 (Integrals of Even and Odd Functions). Let fe (x) and fo (x) be the
even and odd parts of a function f (x), respectively. Then the integrals of fe (x)
and fo (x) over a symmetric interval [−L, L] satisfy
ˆ L ˆ L ˆ L
fe (x)dx = 2 fe (x)dx and fo (x)dx = 0,
−L 0 −L

so that ˆ ˆ
L L
f (x)dx = 2 fe (x)dx.
−L 0

69
Proof.
ˆ L ˆ L
f (x) + f (−x)
fe (x)dx = dx
−L −L 2
ˆ 0 ˆ L
f (x) + f (−x) f (x) + f (−x)
= dx + dx
−L 2 0 2
ˆ L ˆ L
f (−x) + f (x) f (x) + f (−x)
= dx + dx
0 2 0 2
ˆ L
= 2 fe (x)dx.
0

ˆ L ˆ L
f (x) − f (−x)
fo (x)dx = dx
−L −L 2
ˆ 0 ˆ L
f (x) − f (−x) f (x) − f (−x)
= dx + dx
−L 2 0 2
ˆ L ˆ L
f (−x) − f (x) f (x) − f (−x)
= dx + dx = 0
0 2 0 2

Then, since f (x) = fe (x) + fo (x) by construction,

ˆ L ˆ L ˆ L
f (x)dx = fe (x) + fo (x)dx = 2 fe (x)dx
−L −L 0

as required.
Example 10.1. The 1⁄2-wave rectified sine wave

sin πt, 0 ≤ 1
f (t) =
0, −1 < t < 0

70
 1.0
0.5
f(t)

0.0
0.5
1.0
2 1 0 1 2 3 4
1.0
0.5
fe (t)

0.0
0.5
1.0
2 1 0 1 2 3 4
1.0
0.5
fo (t)

0.0
0.5
1.0
2 1 0 1 2 3 4
t
is neither even nor odd, with even part the full-wave rectified sine wave fe (t) =
|1⁄2 sin πt|, and odd part the pure sine wave fo (t) = 1⁄2 sin πt.
Lemma 10.10 (Even and Odd Decomposition of the Fourier Series). If f : R →
R is a function with even and odd parts fe (x) and fo (x), the Fourier coeffients of
f (x) are
ˆ ˆ
2 L  nπx  1 L
An = cos fe (x)dx, A0 = fe (x)dx,
L 0 L L 0
ˆ
2 L  nπx 
Bn = sin f0 (x)dx
L 0 L
Proof. Using Definition 10.16 of the Fourier coefficients, the Even-Odd decom-
position theorem f (x) = fe (x) + fo (x), and employing Lemma 10.9, we find,
ˆ L ˆ L ˆ
1 1 1 L
A0 = f (x)dx = fe (x) + f0 (x)dx = fe (x)dx.
2L −L 2L −L L 0
ˆ ˆ
1 L  nπx  1 L  nπx 
An = cos f (x)dx = cos (fe (x) + fo (x))dx,
L −L L L −L L
ˆ
1 L  nπx   nπx 
= cos fe (x)dx, since cos fo (x) is odd,
L −L L L
ˆ
2 L  nπx 
= cos fe (x)dx.
L 0 L

71
 ˆ ˆ
1 L  nπx  1 L  nπx 
Bn = sin f (x)dx = sin (fe (x) + fo (x))dx
L −L L L −L L
ˆ
1 L  nπx   nπx 
= sin fo (x)dx, since sin fe (x) is odd,
L −L L L
ˆ
2 L  nπx 
= sin fo (x)dx,
L 0 L
as required.
Exercise 10.3. Find the general Fourier series for the 1⁄2-wave rectified sine wave
f (x) = 1/2 (| sin πx| + sin πx) for −1 ≤ x ≤ 1.

11 Fourier Transform
Definition 11.1 (Fourier Transform). If f : R → R is piecewise smooth on the
interval [−L, L], and continuous at x ∈ [−L, L], then the Fourier transform of
f (x) is ˆ ∞
F{f (x)} = F (α) = f (x)e−iαx dx,
−∞

and the Fourier cosine and sine transforms of f (x) are

ˆ ˆ
1 ∞ 1 ∞
A(α) = f (x) cos αxdx, and B(α) = f (x) sin αxdx,
π −∞ π −∞
so that
ˆ ∞ ˆ ∞
−1 1 iαx
f (x) = F {F (α)} = F (α)e dα = A(α) cos αx+B(α) sin αxdα.
2π −∞ 0

Theorem 11.1 (Convergence of´ Fourier Transform). If f is piecewise-smooth on

∞
every interval [a, b] ∈ R and −∞ |f (x)|dx converges, then the Inverse Fourier
transform F −1 of the Fourier transform F of f , F −1 {F{f (x)}} = F −1 {F (α)}
converges to f (x) for all x at which f is continuous.
Exercise 11.1. Defining αn = nπL
and ∆α = αn+1 − αn = Lπ and substituting into
Definition 10.16 for the general Fourier series, derive the Fourier transform of f
by taking the limit L → ∞.
Lemma 11.1. Let δ(x) be the Dirac delta function, then F{δ(x)} = 1

72
Proof. ˆ ∞
F{δ(x)} = δ(x)eiαx dx = eiα0 = 1.
−∞

Definition 11.2 (Convolution). For two functions f, g : R → R the convolution

of f and g, f ∗ g, is defined as
ˆ ∞
f ∗g = f (u)g(x − u)du.
−∞

Theorem 11.2 (Convolution of Fourier Transform). If F (α) = F{f (x)} and

G(α) = F{g(x)}, then

F{f }F{g} = F{f ∗ g}

and ˆ ∞ ˆ ∞
1 iαx
F (α)G(α)e dα = f (u)g(x − u)du.
2π −∞ −∞

Proof.
ˆ ∞  ˆ ∞ ˆ ∞
F{f ∗ g} = F f (u)g(x − u)du = f (u)g(x − u)e−iαx dudx
ˆ ∞ ˆ−∞
∞
−∞ −∞

= f (u)e−iαu g(x − u)e−iα(x−u) dudx.

−∞ −∞

Since the limits of integration are over the entire real axis (−∞, ∞), we can re-
place dx with d(x − u) without loss of generality, so that letting x̃ = x − u we
find
ˆ ∞ ˆ ∞
−iαu
F{f ∗ g} = f (u)e du g(x̃)e−iαx̃ dx̃ = F{f }F{g}.
−∞ −∞

Applying the inverse Fourier transform to both sides, we find

ˆ ∞ ˆ ∞
−1 1
F {F{f ∗ g}} = f ∗ g = f (u)g(x − u)du = F (α)G(α)eiαx dα,
−∞ 2π −∞

as required.

73
Theorem 11.3 (Parseval’s Identity). If F (α) = F{f (x)} and G(α) = F{g(x)}
1
then hf, gi = 2π hF, Gi, that is
ˆ ∞ ˆ ∞
∗ 1
f (x)g (x)dx = F (α)G∗ (α)dα.
−∞ 2π −∞

Proof.
ˆ ∞
hF, Gi 1
= F (α)G∗ (α)dα
2π 2π −∞
ˆ ∞ˆ ∞ ˆ ∞ ∗
1 −iαx 0 −iαx0 0
= f (x)e dx g(x )e dx dα
2π −∞ −∞
ˆ ∞ˆ ∞ ˆ ∞−∞
1 −iαx 0
= f (x)e dx g ∗ (x0 )eiαx dx0 dα
2π
ˆ ∞ −∞
ˆ ∞ −∞ ˆ ∞ −∞
1 0
= eiα(x −x) dαf (x)g ∗ (x0 )dxdx0
−∞ 2π −∞
ˆ−∞∞ ˆ ∞
= δ(x − x0 )f (x)g ∗ (x0 )dxdx0 , since F −1 {1} = δ(x)
ˆ ∞
−∞ −∞

= f (x)g ∗ (x)dx = hf, gi

−∞

as required.
1
Corollary 11.1. If F (α) = F{f (x)} then hf, f i = 2π hF, F i, that is
ˆ ∞ ˆ ∞
2 1
|f (x)| dx = |F (α)|2 dα.
−∞ 2π −∞

12 Sturm-Liouville Theory
Definition 12.1 (Sturm-Liouville Problem). A Sturm-Liouville Problem is a sec-
ond order linear differential equation of the form
 
d dy
p(x) + q(x)y = −λw(x)y,
dx dx

where y is a function of x, and p(x), q(x), and w(x) > 0 are known functions of
x.

74
Theorem 12.1 (Method of Frobenius). To solve a Sturm-Liouville problem we
assume a power series solution around the regular singular point x = 0 of the
form ∞
X
u(x) = ak xk+α , a0 6= 0
k=0

and substitute in the derivatives of the solution

∞
X ∞
X
u0 (x) = ak (k + α)xk+α−1 and u00 (x) = ak (k + α)(k + α − 1)xk+α−2 ,
k=0 k=0

where α ∈ R is a constant.

Definition 12.2 (Legendre’s Equation). Legendre’s equation is a second order

linear differential equation of the form

(1 − x2 )y 00 − 2xy 0 + n(n + 1)y = 0.

Solutions of this equation are called Legendre polynomials of order n, Pn (x), for
n ∈ Z.

Lemma 12.1. Legendre’s equation can be rewritten in Sturm-Liouville form as

[(1 − x2 )y 0 ]0 + n(n + 1)y = 0.

Lemma 12.2 (Rodrigue’s Formula). The Legendre polynomials Pn (x) are given
by Rodrigue’s formula
1 dn 2
Pn (x) = (x − 1)n ,
2n n! dxn
so that P0 (x) = 1 and P1 (x) = x.

Definition 12.3 (Legendre’s Associated Equation). Legendre’s Associated equa-

tion is a second order linear differential equation of the form

m2
 
2 00 0
(1 − x )y − 2xy + n(n + 1) − y = 0.
1 − x2

Solutions of this equation are called associated Legendre polynomials Pnm (x),
where 0 ≤ m ≤ n.

75
Lemma 12.3. Legendre’s associated equation can be rewritten in Sturm-Liouville
form as
m2
 
2 0 0
[(1 − x )y ] + n(n + 1) − y = 0.
1 − x2
Lemma 12.4. The associated Legendre polynomial are related to the Legendre
polynomials by
dm (1 − x2 )m/2 dm+n 2
Pnm (x) = (1 − x2 )m/2
P n (x) = (x − 1)n ,
dxm 2n n! dxm+n
so that Pn0 (x) = Pn (x) and Pnm (x) = 0 if m > n.
Definition 12.4 (Spherical Harmonics). Spherical harmonics Y`m (θ, ϕ) are eigen-
functions of Laplace’s equation in spherical coordinates, and may be expressed in
terms of associated Legndre polynomials P`m in the form
Y`m (θ, ϕ) = N eimθ P`m (cos ϕ).
Definition 12.5 (Bessel’s Equation). Bessel’s equation is a second order linear
differential equation of the form
x2 y 00 + xy 0 + (x2 − ν 2 )y = 0.
Solutions of this equation are called Bessel functions of order ν, Jν (x).
Example 12.1 (Bessel Functions). Bessel functions of the first kind Jn (x) and
Bessel functions of the second kind Yn (x) for n = 0 (blue), 1 (green), 2 (red), 3
(cyan).
1.0 0.5

0.5 0.0
Jn (x)

Yn (x)

0.0 0.5

0.5 1.0

0 5 10 15 20 0 5 10 15 20
x x

76
Definition 12.6 (Bessel’s Modified Equation). Bessel’s modified equation is a sec-
ond order linear differential equation of the form
x2 y 00 + xy 0 − (x2 + m2 )y = 0.
Solutions of this equation are called modified Bessel functions of order m of the
first and second kind Im (x) and Km (x).
Example 12.2 (Modified Bessel Functions). Modified Bessel functions of the first
kind Im (x) and modified Bessel functions of the second kind Km (x) for m = 0
(blue), 1 (green), 2 (red), 3 (cyan).
3.5 3.5

3.0 3.0

2.5 2.5

2.0 2.0
Km (x)
Im (x)

1.5 1.5

1.0 1.0

0.5 0.5

0.0 0.0
0 1 2 3 4 5 0 1 2 3 4 5
x x

Lemma 12.5. Bessel’s equation can be rewritten in Sturm-Liouville form as

ν2
 
0 0
(xy ) + x − y = 0.
x
Exercise 12.1. Find the general solution to the Wave Equation on a circular do-
main D = {0 ≤ r ≤ b, −π ≤ θ ≤ π}
∂ 2u
= c2 ∇ 2 u
∂t2
subject to the initial conditions:
∂u(r, θ; t = 0)
IC: u(r, θ; t = 0) = f (r, θ), = g(r, θ),
∂t
77
and boundary conditions:

∂u(r, π, t) ∂u(r, −π, t)

BC: u(b, θ, t) = 0; u(r, π, t) = u(r, −π, t); = .
∂t ∂t
Exercise 12.2. Using separation of variables, show that the set of eigenfunctions
to Laplace’s equation in spherical coordinates, that is,

∂ 2f 2 ∂f 1 ∂ 2f 1 ∂ 2f cot ϕ ∂f
∇2 f (r, θ, ϕ) = 2
+ + 2 2
+ 2 2
+ = 0,
∂r r ∂r r2 sin ϕ ∂θ r ∂ϕ r2 ∂ϕ

where f (r, θ, ϕ) = R(r)Θ(θ)Φ(ϕ), has the form f (r, θ, ϕ) = (Ar` +B −`−1 )Y`m (θ, ϕ),
where 0 ≤ m ≤ ` and Y`m (θ, ϕ) = eimθ P`m (cos ϕ) are the spherical harmonics.

13 Green’s Functions
Definition 13.1 (Linear Differential Operator). Let Lr [f (r)] be a linear differen-
tial operator on the function f : R3 → R, then Lr satisfies

Lr [αf (r) + βg(r)] = αLr [f (r)] + βLr [g(r)],

where f, g : R3 → R and α, β ∈ R.

Definition 13.2 (Green’s Function). Let Lr be a linear differential operator, then

the two-point Green’s function G(r, r0 ) of Lr satisfies

Lr [G(r, r0 )] = δ(r − r0 ),

where δ(r − r0 ) is the Dirac delta function.

Theorem 13.1 (Green’s Function Solution). Let Lr be a linear differential opera-

tor with two-point Green’s function G(r, r0 ), then the general solution to the linear
inhomogeneous differential equation

Lr [f (r)] = h(r)

in terms of the Green’s function G(r, r0 ) is

ˆ
f (r) = G(r, r0 )h(r0 )dr0 .

78
Proof. Inserting the solution into the linear inhomogeneous differential equation,
we find
ˆ 
0 0 0
Lr [f (r)] = Lr G(r, r )h(r )dr ,
ˆ
= Lr [G(r, r0 )]h(r0 )dr0 , since Lr is a linear differential operator,
ˆ
= δ(r − r0 )h(r0 )dr0 , by Definition 13.2 for the Green’s function,
= h(r), by Definition 9.10 for the Dirac δ function,
as required.
Note. The Green’s function G(r, r0 ) reflects the action of u(r) due to the applica-
tion of f (r).
Example 13.1. Determining a potential Φ(r) generated by the charge density n(r).
Under the application of the Poisson equation and Coulomb’s law ∇2 Φ = n(r),
and we obtain the solution for Φ(r),
ˆ
n(r0 )
Φ(r) = 0
dr0 .
|r − r |
The Green’s function G(r, r0 ) to this equation is then the Coulomb kernel
1
G(r, r0 ) = ,
|r − r0 |
which is the integrand of the integral operator converting n(r) in Φ(r),
ˆ
Φ(r) = G(r, r0 )n(r0 )dr0 .

Exercise 13.1. Show that the Green’s function G(r, r0 ) for Poisson’s equation in
cylindrical coordinates, ∇2r,θ,z Φ(r, θ, z) = n(r, θ, z), for a cylindrical shell of ra-
dius r0 satisfies
 2
1 ∂2 ∂2

∂ 1 ∂ 1
2
+ + 2 2 + 2 G(r, r0 ) = δ(r − r0 )δ(θ − θ0 )δ(z − z 0 ),
∂r r ∂r r ∂θ ∂z r
and is given by the Fourier-Bessel transformation
∞ ˆ ∞
0 1 X dk ik(z−z0 ) im(θ−θ0 )
G(r, r ) = 0
=− 2
e e Im (|k|r< )Km (|k|r> ),
|r − r | m=−∞ −∞ (2π)
where r< ≡ min(r, r0 ), r> ≡ max(r, r0 ), and Im (|k|r< ) and Km (|k|r> ) are mod-
ified Bessel’s functions of the first and second kind, respectively.

79
14 Solutions to the Exercises
1 Curves in R3
Exercise 1.1. Plot r = (α cos t, α sin t, βt), α ≥ 0, β constants for 0 ≤ t ≤ 4π.
Solution 1.1.

14
12
10
8
6
4
2
0
1.0
0.5
1.0 0.0
0.5
0.0 0.5
0.5
1.0 1.0

80
Exercise 1.2. Plot r = (αt cos t, αt sin t, βt), α ≥ 0, β constants for 0 ≤ t ≤ 4π.
Solution 1.2.

14
12
10
8
6
4
2
0
10
5
10 0
5 5
0
5 10
10
15 15

81
Exercise 1.3. Plot r = (t, t2 , t3 ) for 0 ≤ t ≤ 2.
Solution 1.3.

8
6
4
2
0
2
4
6
8
4.0
3.5
3.0
2.5
2.0 2.0
1.5
1.0 1.5
0.5
0.0 1.0
0.5 0.5
1.0
1.5
2.0 0.0
Exercise 1.4. For motion on a circle of radius α, and angular velocity ω, provide
the centripetal acceleration and centripetal force.
Solution 1.4. ρ = α ⇒ κ = α1 , v = ds dt
= αω, κv 2 = α1 (αω)2 = αω 2 . So
2
centripetal acceleration is αω 2 = vα = κv 2 . So centripetal force is mass times
2
centripetal acceleration = maω 2 = mva = mκv 2 .

82
Exercise 1.5. For a circular helix r = (α cos θ, α sin θ, βθ), α ≥ 0, and θ is the
polar angle, find T̂, N̂, B̂, κ, and τ as functions of θ.
Solution 1.5.
dr
ṙ = = (−α sin θ, α cos θ, b)
dθ
ds p
⇒ = |ṙ| = α2 + β 2
dθ
dr dθ 1
⇒ T̂ = = ṙ = p (−α sin θ, α cos θ, β)
ds ds α + β2
2

(−α sin θ, α cos θ, β)

T̂ = p .
α2 + β 2

!2
dT̂ ˙ dθ 1
κ = κN̂ = = T̂ = p (−α cos θ, −α sin θ, 0)
ds ds α + β2
2

α
= − 2 (cos θ, sin θ, 0)
α + β2

α
⇒ κ= = constant
α2 + β2
1 α2 + β 2
⇒ ρ= = = constant
κ α
⇒ N̂ = −(cos θ, sin θ, 0) which points directly towards the z-axis

(−α sin θ, α cos θ, β)

⇒ B̂ = T̂ × N̂ = p × (− cos θ, − sin θ, 0)
α2 + β 2
(β sin θ, −β cos θ, α)
B̂ = p
α2 + β 2
dB̂ ˙ dθ (β cos θ, β sin θ, 0)
⇒ −τ N̂ = = B̂ =
ds ds α2 + β 2
β
⇒ τ= = constant
α2 + β2

83
Note. The helix is the only space curve for which both κ(s) and τ (s) are constant.
Exercise 1.6. For motion on a curve r = (t, t2 , 32 t3 ) find v, s, T̂, N̂, B̂, κ, and τ
as functions of t.
√ p
Solution 1.6. v = ṙ = (1, 2t, 2t2 ) ⇒ |ṙ| = 1 + 4t2 + 4t4 = (1 + 2t2 )2 =
´ ´
1 + 2t2 . Since |ṙ| =
6 1, t is not arclength. Indeed s = |ṙ|dt = (1 + 2t2 )dt = t +
2
t +C. Measuring arclength from (0, 0, 0) at t = 0, then C = 0 ⇒ s = t + t2 .
2 3
3
3
2t2
 
dr dt 1 2 1 2t
T̂ = = ṙ = (1, 2t, 2t ) = , ,
ds ds 1 + 2t2 1 + 2t2 1 + 2t2 1 + 2t2

(1, 2t, 2t2 )

T̂ = .
1 + 2t2
 
dT̂ ˙ dt d 1 2 1
κN̂ = = T̂ = (1, 2t, 2t )
ds ds dt 1 + 2t2 1 + 2t2
 
−4t 2 1 1
= 2 2
(1, 2t, 2t ) + 2
(0, 2, 4t)
(1 + 2t ) 1 + 2t 1 + 2t2
(−4t, −8t2 , −8t3 ) + (0, 2 + 4t2 , 4t + 8t3 )
=
(1 + 2t2 )3
(−4t, 2 − 4t2 , 4t)
κN̂ =
(1 + 2t2 )3
So √ p
16t2 + 4 − 16t2 + 16t4 + 16t2 2 (1 + 2t2 )2
κ = |κN̂| = =
(1 + 2t2 )3 (1 + 2t2 )3
2
κ=
(1 + 2t2 )2
and
(−2t, 1 − 2t2 , 2t)
N̂ = .
1 + 2t2

1
B̂ = T̂ × N̂ = (1, 2t, 2t2 ) × (−2t, 1 − 2t2 , 2t)
(1 + 2t2 )2
(2t2 + 4t4 , −4t3 − 2t, 1 + 2t2 )
=
(1 + 2t2 )2

84
 (2t2 , −2t, 1)
B̂ = .
1 + 2t2

dB̂ ˙ dt
−τ N̂ = = B̂
ds ds 
4t 2 1 1
= − (2t , −2t, 1) + (4t, −2, 0)
(1 + 2t2 )2 1 + 2t2 1 + 2t2
(4t, 4t2 − 2, −4t)
=
(1 + 2t2 )3

2
τ=
(1 + 2t2 )2
Note. κ ≡ τ along C but they are not constants.
Question. How far is travelled along the curve for 0 ≤ t ≤ 1?
Answer. s(1) − s(0) = 1 + 32 (1)2 = 53 .
Exercise 1.7. Using the chain rule, find expressions for the curvature κ(s) and
torsion τ (s) along the curve C : r(t) in terms of derivatives of r with respect to
...
time t, that is, ṙ, r̈, and r .
Solution 1.7. Applying the definitions of T̂, N̂, and B̂, and using the chain rule
d
ds
= v1 dtd , we first express the moving triad in terms of ṙ, r̈, κ, and v

dr ṙ
T̂ ≡ =
ds v
dT̂ 1 d ṙ r̈ ṙ
κN̂ ≡ = = 2 − 3 v̇
ds v dt v v v
vr̈ − v̇ ṙ
⇒ N̂ =
κv 3
ṙ vr̈ − v̇ ṙ ṙ × r̈
B̂ ≡ T̂ × N̂ = × 3
= ,
v κv κv 3

Since B̂ is a unit vector, |B̂| = 1, and thus from the expression for the magnitude
of B̂ we may express κ in terms of ṙ, r̈, and v = |ṙ| as

|ṙ × r̈|
κ= .
|ṙ|3

85
 dN̂
Using the Frenet-Serret formula for ds
,

dN̂
= τ B̂ − κT̂,
ds
and taking the dot product of both sides with B̂, we find

dN̂
· B̂ = τ B̂ · B̂ − κT̂ · B̂ = τ, (25)
ds
d 1 d ˙
since T̂ ⊥ B̂. Using ds
= v dt
by chain rule, we may express τ in terms of N̂, B̂,
and v as
1 ˙
τ = N̂ · B̂. (26)
v
Employing Lemma 1.7 for accelaration r̈, and differentiating with respect to t, we
˙
may obtain an expression for N̂

r̈ = v̇ T̂ + κv 2 N̂,
... ¨ ˙
⇒ r = v̈ T̂ + v̇ T̂ + (κ̇v 2 + 2κv v̇)N̂ + κv 2 N̂
˙
T̂
Since N̂ = ˙ , by taking the dot product of both sides with respect to B̂, we find
|T̂|

... *0 ¨
 :0 :0 ˙
N̂· B̂ + (κ̇v 2 + 2κv v̇)
N̂· B̂ + κv 2 N̂ · B̂
  
B̂ · r = v̈ · B̂ + v̇|T̂|
 
T̂
˙ ...
B̂ · N̂ B̂ · r
⇒τ = = .
v κv 3
Substituting for B̂ and κv 3 into our expression for τ , that is, ṙ × r̈ = κv 3 B̂ and
|ṙ × r̈| = κv 3 , we find
...
(ṙ × r̈) · r
τ=
|ṙ × r̈|2
as required.

2 Surfaces in R3
Exercise 2.1. Sketch the level curves of f , and hence sketch the surface z =
f (x, y) for (a) f (x, y) = x2 + y 2 , (b) f (x, y) = x2 − y 2 , and (c) f (x, y) = x2 .

86
Solution 2.1. √ √
(a) x2 + y 2 = C ⇒ y = ± C − x2 concentric circles of radius C.
2

1
3

8
7
6
5
4
0
y

3
2
1
1

0
2.5
1 2.0
2

1.5
1.0
4

2.0 0.5
1.5 0.0
1.0
0.5 0.5
0.0
0.5 1.0
1.0 1.5
1.5
2 2.0
2.5 2.0
2 1 0 1 2
x
√
(b) x2 − y 2 = C ⇒ y = ± x2 − C hyperbola
2 -3
-1
-2

1 4
3

0
0

0
y

2.5
1
3

2.0
1.5
2

1.0
0

2.0 0.5
1.5 0.0
1.0
0.5 0.5
0.0
1

0.5 1.0
1.0
1.5 1.5
-3
-2

2 2.0
-1

2.5 2.0
2 1 0 1 2
x
√
(c) x2 = C ⇒ x = ± C parallel lines.

87
 2

2
3
4.5

1 4.0
3.5
3.0
2.5
2.0
0
y

1.5
1.0
0.5
0.0
2.5
1 2.0
1.5
1.0
2.0 0.5
1.5 0.0
1.0
0.5 0.5
0.0
0.5 1.0
1.0 1.5
1.5
2
3
2

2.0
2.5 2.0
2 1 0 1 2
x

Exercise 2.2. Calculate the surface area ofp a sphere x2 + y 2 + z 2 = α2 (a) using
x, y as parameters on one hemisphere z = a2 − x2 − y 2 (the hard way), and (b)
using spherical polar coordinates (r, θ, ϕ) with r = α (the easy way).
Solution 2.2.
(a) {Hard Way}
p
z = α2 − x2 − y 2 on the upper hemisphere
∂z x
⇒ = p
∂x α 2 − x2 − y 2
∂z y
⇒ = p
∂y α 2 − x2 − y 2
s  2  2
∂z ∂z
⇒ dS = 1+ + dxdy
∂x ∂y
s
x2 y2
= 1+ 2 + dxdy
α − x2 − y 2 α 2 − x2 − y 2
αdxdy
= p
α 2 − x2 − y 2

So
¨ ˆ x=α ˆ √
y= α2 −x2
αdxdy dy
S=2 p = 2α dx √
p
Rxy α 2 − x2 − y 2 x=−α y=− α2 −x2 α 2 − x2 − y 2

88
Using arcsin substitution1
d 1 du
arcsin u = √
dy 1 − u2 dy
y du 1
with u = √ , =√ , we find
α2 − x2 dy α 2 − x2
d y 1
arcsin √ =p
dy 2
α −x 2 α 2 − x2 − y 2
so that
ˆ α
√
α2 −x2
y
S = 2α dx arcsin √
−α α − x2 −√α2 −x2
2
ˆ α
= 2α dx[arcsin(1) − arcsin(−1)]
−α
 π  π 
= 2α(2α) − − = 4πα2
2 2
(b) {Easy Way}
r = r(θ, ϕ) = (α sin ϕ cos θ, α sin ϕ sin θ, α cos ϕ)
⇒ rθ = (−α sin ϕ sin θ, α sin ϕ cos θ, 0)
rϕ = (α cos ϕ cos θ, α cos ϕ sin θ, −α sin ϕ)
⇒ dS = |rθ × rϕ |dθdϕ
= |(−α2 sin2 ϕ cos θ, −α2 sin2 ϕ sin θ, −α2 sin ϕ cos ϕ)|
= | − α2 sin ϕ(sin ϕ cos θ, sin ϕ sin θ, cos ϕ)|
is an inward pointing normal of length α2 sin ϕ. So an outward pointing unit nor-
mal is n̂ = (sin ϕ cos θ, sin ϕ sin θ, cos ϕ) and dS = |rθ ×rϕ |dϕdθ = α2 sin ϕdθdϕ.
Geometrically, dS = (αdϕ) · (α sin ϕdθ) = α2 sin ϕdθdϕ.
¨ ¨
S= dS = α2 sin ϕdθdϕ
Rθϕ
ˆ 2π ˆ π
2
= α dθ sin ϕdϕ
0 0
= α2 2π[− cos ϕ]π0
= 4πα2 .
1
Equation 13.20 of Schaum’s Mathematical Handbook

89
Exercise 2.3. Find the surface area of the paraboloid r(u, v) = (u cos v, u sin v, ku2 )
for 0 ≤ z ≤ 1, using cylindrical coordinates (r, θ, z).
Solution 2.3. u2 = x2 + y 2 = r2 polar coordinates in plane, with (u, v) = (r, θ)
planar polar coordinates. z = ku2 = kr2 . Find the surface area for 0 ≤ z ≤ 1 ⇒
0 ≤ u ≤ √1k

dS = |ru × rv |dudv
= |(cos v, sin v, 2ku) × (−u sin v, u cos v, 0)|dudv
√
= u 1 + 4k 2 u2 dudv

So the surface area is given by

¨ ¨ √
S = dS = u 1 + 4k 2 u2 dudv
0≤z≤1 Ruv
ˆ 2π ˆ √1
√
k
= dv duu 1 + 4k 2 u2
0 0
3 √1
(1 + 4k 2 u2 ) 2 k
= 2π
12k 2
0
π 3
= [(1 + 4k) 2 − 1]
6k 2

3 Scalar & Vector Fields and the “∇” Operator

Exercise 3.1. In R3 , what is the gravitational potential function φ, where the grav-
itational force is
GmM
F(x, y, z) = − 3 r.
r
Find a φ such that
GmM
∇φ = − 3 r
r
1
and show that φ ∝ r is the gravitational potential.

90
Solution 3.1. Working in spherical coordinates, (x, y, z) → (r, θ, ϕ),
 
GmM
F = − 2 , 0, 0 = − grad φ
r
∂φ GmM 1 ∂φ 1 ∂φ
⇒ = 2
, = 0, =0
∂r r r sin ϕ ∂θ r ∂ϕ
ˆ
GmM GmM
⇒φ = 2
dr = − + C(θ, ϕ).
r r
∂φ ∂φ
Since ∂θ
= ∂ϕ
= 0, C(θ, ϕ) = 0, so that

GmM
φ == −
r

and hence φ ∝ 1r .
Exercise 3.2. Prove Lemma 3.6

curl curl V = grad div V − div grad V

∇ × (∇ × V) = ∇(∇ · V) − ∇2 V

where div grad V ≡ ∇2 V ≡ (∇2 V1 , ∇2 V2 , ∇2 V3 ).

Solution 3.2. Proof. Let V = (V1 , V2 , V3 )
 
∂V3 ∂V2 ∂V1 ∂V3 ∂V2 ∂V1
∇ × (∇ × V) = ∇ × − , − , −
∂y ∂z ∂z ∂x ∂x ∂y
    
∂ ∂V2 ∂V1 ∂ ∂V1 ∂V3
= − − − ,
∂y ∂x ∂y ∂z ∂z ∂x
   
∂ ∂V3 ∂V2 ∂ ∂V2 ∂V1
− − − ,
∂z ∂y ∂z ∂x ∂x ∂y
   
∂ ∂V1 ∂V3 ∂ ∂V3 ∂V2
− − −
∂x ∂z ∂x ∂y ∂y ∂z

by repeated application of the definition of ∇ × V. Simplifying and collecting

91
terms
∂ 2 V2 ∂ 2 V3 ∂ 2 V1 ∂ 2 V1

= + − − ,
∂x∂y ∂x∂z ∂y 2 ∂z 2
∂ 2 V3 ∂ 2 V1 ∂ 2 V2 ∂ 2 V2
+ − − ,
∂y∂z ∂y∂x ∂x2 ∂z 2
∂ 2 V1 ∂ 2 V2 ∂ 2 V3 ∂ 2 V3

+ − −
∂z∂x ∂z∂y ∂x2 ∂y 2
∂ 2 V1 ∂ 2 V1 ∂ 2 V1
  
∂ ∂V1 ∂V2 ∂V3
= + + − − − ,
∂x ∂x ∂y ∂z ∂x2 ∂y 2 ∂z 2
∂ 2 V2 ∂ 2 V2 ∂ 2 V2
 
∂ ∂V1 ∂V2 ∂V3
+ + − − − ,
∂y ∂x ∂y ∂z ∂x2 ∂y 2 ∂z 2
∂ 2 V3 ∂ 2 V3 ∂ 2 V3
  
∂ ∂V1 ∂V2 ∂V3
+ + − − −
∂z ∂x ∂y ∂z ∂x2 ∂y 2 ∂z 2
 
∂ 2 ∂ 2 ∂ 2
= ∇ · V − ∇ V1 , ∇ · V − ∇ V2 , ∇ · V − ∇ V3
∂x ∂y ∂z

by definition of ∇ · V and ∇2 V . By the definition of grad and ∇2 V we then

obtain
∇ × (∇ × V) = ∇(∇ · V) − ∇2 V

Exercise 3.3. Consider the vortex field

(−y, x, 0)
V= , (x, y) 6= (0, 0).
x2 + y 2

Show that V is irrotational, that is, curl V = ∇ × V ≡ 0 for all (x, y) 6= (0, 0),
so that in simply connected regions there exists φ such that V = ∇φ, and find
φ(x, y, z). Show that V is solenoidal, that is div V = ∇ · V = 0 for all (x, y) 6=
(0, 0), so that in simply connected regions there exists A such that V = curl A =
∇ × A, and find A by assuming (1) Az ≡ 0 and (2) A1 (x, y) = 0 = A2 (x, y).

92
Solution 3.3. To demonstrate V is irrotational,
   
∂ ∂ ∂ −y x
∇×V = , , × , ,0
∂x ∂y ∂z x2 + y 2 x2 + y 2
 
∂ ∂ −y ∂ x ∂ ∂ x ∂ −y
= 0− , − 0, −
∂y ∂z x2 + y 2 ∂z x2 + y 2 ∂x ∂x x2 + y 2 ∂y x2 + y 2
2x2 2y 2
 
1 1
= 0, 0, 2 − + −
x + y 2 (x2 + y 2 )2 x2 + y 2 (x2 + y 2 )2
x2 + y 2 2x2 x2 + y 2 2y 2
 
= 0, 0, 2 − + −
(x + y 2 )2 (x2 + y 2 )2 (x2 + y 2 )2 (x2 + y 2 )2

⇒ ∇ × V = (0, 0, 0) = 0
so that V is irrotational on simply connected regions of R2 /(0, 0). This means in
a simply connected region there exists φ such that V = grad φ
 
(−y, x, 0) ∂φ ∂φ ∂φ
⇒ 2 = , , .
x + y2 ∂x ∂y ∂z

Find φ(x, y, z)

∂φ −y
⇒ = 2 , (27)
∂x x + y2
∂φ x
= 2 , (28)
∂y x + y2
∂φ
= 0. (29)
∂z
Integrating (27) with respect to x keeping y and z fixed,
ˆ
−y y
φ(x, y, z) = 2 2
dx = arctan + g(y, z) (30)
x,yfixed x + y x

By (29), g is a function of y alone, i.e., g(y, z) = g(y). Substitute (30) into (28):
x 0 x
+ g (y) =
x2 + y 2 x2 + y 2
0
⇒ g (y) = 0
g(y) = C

93
 y
⇒ φ(x, y, z) = arctan +C
x
is a potential function for the vortex field, provided (x, y) 6= (0, 0). The whole
z-axis is a singularity.
To demonstrate V is solenoidal
   
∂ ∂ ∂ −y x
∇·V = , , · , ,0
∂x ∂y ∂z x2 + y 2 x2 + y 2
∂ −y ∂ x
= 2 2
+
∂x x + y ∂y x + y 2
2

−2xy 2xy
= 2 2 2
+ 2
(x + y ) (x + y 2 )2

⇒ ∇·V =0
so that V is solenoidal on simply connected regions of R2 /(0, 0). This means in a
simply connected region there exists a vector potential A such that V = curl A.
∂A3 ∂A2 −y
⇒ − = 2
∂y ∂z x + y2
∂A1 ∂A3 x
− = 2
∂z ∂x x + y2
∂A2 ∂A1
− = 0
∂x ∂y
These are three simultaneous partial differential equations for unknowns A1 ,
A2 , and A3 containing arbitrary functions of integration. This means there are
infinitely many vector potentials of totally different functional form. We thus try
some simplifying assumptions to help us integrate:

1. Assume A1 , A2 , A3 are functions of x, y only

∂A1 ∂A2 ∂A3
⇒ = = = 0.
∂z ∂z ∂z

2. Assume A1 (x, y) = A2 (x, y) = 0.

∂A3 y ∂A3 x
⇒ =− 2 , = −
∂y x + y 2 ∂x x2 + y 2

94
 1
⇒ A3 (x, y) = − ln(x2 + y 2 )
2 p
= − ln x2 + y 2
1
= ln p
x + y2
2
 
1
= ln = − ln r.
r
So a vector potential for the vector field V is
!
1
A = (0, 0, − ln r) = 0, 0, ln p ,
x + y2
2

for (x, y) 6= (0, 0).

4 Physical Interpretation of Grad, Div, and Curl

Exercise 4.1. Find the flow lines of V = (1, 2x) in R2 in parametric and non-
parametric form.
Solution 4.1. The flow lines are given by the ordinary differential equation (ODE):

1dx + 2xdy = 0
⇔ (1, 2x) · (dx, dy) = 0

which are curves orthogonal to the flow lines.

(a) Equations of flow lines

ṙ = (1, 2x)
dx dy
⇔ = 1, = 2x
dt dt
x = t + C ⇒ y = t2 + 2C1 t + C2

are the flow lines. Elimate t ⇒ y = x2 +C, parabolas in non-parametric form.

(b) Non-parametric form

dx dy
= ⇒ 2xdx − dy = 0
1 2x
95
 Separating and integrating
ˆ ˆ
2xdx = dy

⇒ x2 = y + C,
parabolas.
Exercise 4.2. Find the flow lines of V = (−ωy, ωx, 0) in parametric and non-
parametric form for rigid body motion where ω is the angular velocity about the
z-axis.
Solution 4.2. The flow lines are given by the ordinary differential equation (ODE):
−ωydx + ωxdy = 0
⇔ (y, x) · (dx, dy) = 0
(a) Parametric form of flow lines:
dx dy dx
= −ωy, = ωx, = 0 ⇒ z = C3
dt dt dt
dy x
= − ⇒ separable
dx y
ˆ ˆ
ydy = − xdx
x2 y 2
⇒ + = C
2 2
Flow lines are circles about the origin in the xy-plane.
(b) Non-parametric form of flow lines:
dx dy
− =
ωy ωx
Flow lines are again circles about the origin in the xy-plane.
Exercise 4.3. Try finding expressions for (1) div V and (2) ∇2 V in spherical
coordinates in terms of r, θ, ϕ derivatives.
x = r sin θ cos ϕ
y = r sin θ sin ϕ
z = r cos ϕ
using chain rule.

96
Solution 4.3. From Lemma 3.8, ∇ · (φA) = ∇φ · A + φ∇ · A so

∇ · V = ∇ · (V1 êr + V2 êθ + V3 êϕ )

= ∇ · (V1 êr ) + ∇ · (V2 êθ ) + ∇ · (V3 êϕ )
= ∇(V1 )êr · êr + V1 ∇ · êr + ∇(V2 )êθ · êθ + V2 ∇ · êθ + ∇(V3 )êϕ · êϕ + V3 ∇ · êϕ

∂V1 1 ∂V2 1 ∂V3

∇·V = + + + V1 ∇ · êr + V2 ∇ · êθ + V3 ∇ · êϕ (31)
∂r r sin ϕ ∂θ r ∂ϕ
Expressing r, θ, and ϕ in terms of x, y, and z
p
r = x2 + y 2 + z 2
y
tan θ =
x
z
cos ϕ = p
x2 + y 2 + z 2

Differentiating r with respect to x, y, and z we obtain

∂r x r cos θ sin ϕ
= p = = cos θ sin ϕ
∂x 2
x +y +z2 2 r
∂r y r sin θ sin ϕ
= p = = sin θ sin ϕ
∂y x2 + y 2 + z 2 r
∂r z r cos ϕ
= p = = cos ϕ
∂z 2
x +y +z2 2 r

Differentiating θ with respect to x, y, and z we obtain

∂θ y 1 sin θ sin ϕ sin θ
= − 2 2 =− 2 2 cos2 θ = −
∂x x sec θ r cos θ sin ϕ r sin ϕ
∂θ 1 1 1 cos θ
= 2
= cos2 θ =
∂y x sec θ r cos θ sin ϕ r sin ϕ
∂θ
= 0
∂z

97
Differentiating ϕ with respect to x, y, and z we obtain
∂ϕ xz −1 cos θ sin ϕ cos ϕ cos θ cos ϕ
= − 3 = =
∂x (x2 + y 2 + z 2 ) 2 sin ϕ r sin ϕ r
∂ϕ yz −1 sin θ sin ϕ cos ϕ sin θ cos ϕ
= − 3 = 2
=
∂y (x + y 2 + z 2 ) 2
2 sin ϕ r sin ϕ r
!
∂ϕ 1 z2 −1 cos2 ϕ 1 sin ϕ
= p − 3 = − =−
∂z x2 + y 2 + z 2 (x2 + y 2 + z 2 ) 2 sin ϕ r sin ϕ r sin ϕ r

êr = cos θ sin ϕêx + sin θ sin ϕêy + cos ϕêz

∂θ ∂ϕ ∂θ ∂ϕ ∂ϕ
∇ · êr = − sin θ sin ϕ + cos θ cos ϕ + cos θ sin ϕ + sin θ cos ϕ − sin ϕ
∂x ∂x ∂y ∂y ∂z
2 2 2 2 2 2 2
sin θ cos θ cos ϕ cos θ sin θ cos ϕ sin ϕ
= sin ϕ + + sin ϕ + +
r sin ϕ r r sin ϕ r r
1 1
= +
r r
2
∇ · êr = (32)
r

êθ = − sin θêx + cos θêy

∂θ ∂θ
∇ · êθ = − cos θ − sin θ
∂x ∂y
− sin θ cos θ
= − cos θ − sin θ
r sin ϕ r sin ϕ

∇ · êθ = 0 (33)

êϕ = cos θ cos ϕêx + sin θ cos ϕêy − sin ϕêz

∂θ ∂ϕ ∂θ ∂ϕ ∂ϕ
∇ · êϕ = − sin θ cos ϕ − cos θ sin ϕ + cos θ cos ϕ − sin θ sin ϕ − cos ϕ
∂x ∂x ∂y ∂y ∂z
2 2 2 2
sin θ cos θ cos ϕ cos θ sin θ cos ϕ cos ϕ sin ϕ
= cos ϕ − sin ϕ + cos ϕ − sin ϕ +
r sin ϕ r r sin ϕ r r

cot ϕ
∇ · êϕ = (34)
r

98
Substituting (32), (33), and (34) into (31) yields

∂V1 1 ∂V2 1 ∂V3 2 cot ϕ

∇·V = + + + V1 + V3 (35)
∂r r sin ϕ ∂θ r ∂ϕ r r

or alternatively

1 ∂ 2 1 ∂ 1 ∂
∇·V = r V1 + V2 + sin ϕV3 (36)
r2 ∂r r sin ϕ ∂θ r sin ϕ ∂ϕ

Substituting V = ∇V into (35) yields

∂ 2V 1 ∂ 2V 1 ∂ 2V 2 ∂V cot ϕ ∂V
∇2 V = 2
+ 2 2
+ 2 2
+ + (37)
∂r r2 sin ϕ ∂θ r ∂ϕ r ∂r r2 ∂ϕ

or alternatively

1 ∂ 2 ∂V 1 ∂ 2V 1 ∂ ∂V
∇2 V = 2
r + 2 2
+ 2
sin ϕ (38)
r ∂r ∂r r2 sin ϕ ∂θ r sin ϕ ∂ϕ ∂ϕ

5 Line Integrals
´
Exercise 5.1. Find C f (x, y, z)dx where f (x, y, z) = x + y + z and C : x = cos t,
y = sin t, z = t for 0 ≤ t ≤ π2 .
Solution 5.1.
ˆ B ˆ π
2 dx
T̂ = C f dx = C f dt
A t=0 dt
ˆ π
2
= (cos t, sin t, t)(− sin t)dt substituting C in the integral
0
  π2
1 2 t 1
= cos t + (− + sin 2t) + (t cos t − sin t)
2 2 4 0
1 π 3 π
= − −1=− −
2 4 2 4

´B
Exercise 5.2. Evaluate I = C A
(x + y)dx along each of

99
 1. circle x = cos θ, y = sin θ from (1,0) to (0,1).
   
y=0 x=0
2. broken line ∪ from (1,0) to (0,1).
0≤x≤1 0≤y≤1
Solution 5.2. This is an integral with F = (f, 0, 0) so F · dr = f dx.
π dx
1. Use θ : 0 ≤ θ ≤ 2
⇒ dx = dθ
dθ = − sin θdθ
ˆ π
2
I = (cos θ + sin θ)(− sin θ)dθ
θ=0
  π2
1 1 θ
= cos2 θ + sin 2θ −
2 4 2 0
1 π
= − − .
2 4
2.
ˆ x=0 ˆ y=1 0
I = (x + y)dx + (x + y)
dx
>

y=0,x=1 x=0,y=0
ˆ 0  0
1 2 1
= (x + 0)dx = x =− .
x=1 2 1 2

Note. Different answers imply this line integral is not path independent. Here
F = (x + y, 0, 0) ⇒ curl F = (0, 0, −1) 6= 0. Since curl F ≡ 6 0 we do not have
path independence.
Exercise 5.3. Find the work done by the force field F = (2xy, x2 , 0) in the xy-
plane when moving from (0,0,0) to (1,1,0) along
   
y=0 x=1
1. ∪ ,
0≤x≤1 0≤y≤1
2. y = x, 0 ≤ x ≤ 1,
   
x=0 y=1
3. ∪ .
0≤y≤1 0≤x≤1
Solution 5.3. curl F ≡ (0, 0, 2x − 2x) ≡ 0 on R3 ⇒ path independence on R3 .
ˆ (1,1,0) ˆ (1,1,0)
W = C F · dr = C (2xydx + x2 dy)
(0,0,0) (0,0,0)

100
 1.
ˆ x=1 ˆ y=1 ˆ 1
W = F · dr + F · dr = 0 + 1dy = [y]10 = 1.
y=0,x=0 x=1,y=0 y=0

2.
ˆ x=1 ˆ 1 ˆ 1
2 2
 1
W = (2xydx+x dy) = (2xxdx+x dx) = 3x2 dx = x3 0 = 1.
y=x,x=0 x=0 0

3.
ˆ y=1 ˆ x=1 0 ˆ 1  1
x2 dy)+ 2
2xdx = x2 0 = 1.
((
W = (2xydx
( (((+ (( (2x1dx+x dy)
 =
x=0,y=0 y=1,x=0 0

6 Green’s Theorem
Exercise 6.1. Verify Green’s Theorem for
˛
(xy + y 2 )dx + x2 dy
C

where C is the curve formed by y = x and y = x2 for 0 ≤ x ≤ 1.

∂Q ∂P
Solution 6.1. P = xy + y 2 , Q = x2 , ∂x
= 2x, ∂y
= x + 2y

1. As a line integral, C : y = x2 ∪ y = x,
ˆ (1,1) ˆ (0,0)
I = y=x2 P dx + Qdy + y=x P dx + Qdy
(0,0) (1,1)
ˆ (1,1) ˆ (0,0)
2 2
= y=x2 (xy + y )dx + x dy + y=x (xy + y 2 )dx + x2 dy
(0,0) (1,1)
ˆ 1 ˆ 0
= (x3 + x4 + x2 2x)dx + (x · x + x2 + x2 )dx
0 1
 4 1
3x x5  0
= + + x3 1
4 5 0
19 1
= −1=− .
20 20

101
 2. As a surface integral,
¨
∂Q ∂P
I = − dxdy
∂x ∂y
¨R
= (2x − (x + 2y))dxdy
R
ˆ ˆ x 1
= (x − 2y)dydx
x=0 y=x2
ˆ 1 x
= xy − y 2 x2
dx
0
ˆ 1
= x2 − x2 − (x3 − x4 )dx
0
 5 1
x x4
= −
5 4 0
1 1 1
= − =−
5 4 20

Exercise 6.2. Evaluate

˛
I= (y − sin x)dx + cos xdy
C

counter-clockwise around the triangle with vertices (0, 0), ( π2 , 0), and ( π2 , 1).
Solution 6.2.
Note. The field F = (Q, P, 0) is not irrotational, so the integral is not path inde-
pendent, and I is probably non-zero.

102
 ¨  
∂Q ∂P
I = − dxdy
∂x ∂y
¨
= (− sin x − 1) dxdy
R
ˆ π ˆ 2x
2 π
= − (1 + sin x)dydx
0 0
ˆ π
2 2x
= − (1 + sin x)dx
0 π
 π2
2 x2

= − − x cos x + sin x using integrations by parts
π 2 0
 
π 2
= − +
4 π

7 Surface Integrals
˜
Exercise 7.1. Evaluate I = S x2 zdS where S is the curved cylindrical surface
x2 + y 2 = 1, (0 ≤ z ≤ 1 using cylindrical coordinates (r, θ, z) with r = 1.
Solution 7.1.

x = r cos θ x = cos θ
y = r sin θ with r = 1 ⇒ y = sin θ
z=z z=z

⇒ r(θ, z) = (cos θ, sin θ, z)

rθ × rz = (− sin θ, cos θ, 0) × (0, 0, 1)

= (cos θ, sin θ, 0)
⇒ dS = |rθ × rz |dθdz = dθdz.
¨ ˆ 2π ˆ 1
2
⇒I = cos θzdθdz = cos2 θzdθdz
Rθz 0 0
 2π
θ 1 1 π
= + sin 2θ = .
2 4 0 2 2

103
Exercise 7.2. Find the centroid z, that is, the position of the center of mass, of a
hemispherical shell of constant density using spherical polar coordinates on S :
r = α.
Solution 7.2. By symmetry x = y = 0,
¨ ¨
1 1
z≡ zdS = zdS.
S S 2πα2 S
for a shell of radius r = α. Using spherical polar coordinates on S : r = α

r = r(θ, ϕ) = (α cos θ sin ϕ, α sin θ sin ϕ, α cos ϕ)

dS = |rθ × rϕ |dθdϕ
= |(−α sin θ sin θ, α cos θ sin ϕ, 0) × (α cos θ cos ϕ, α cos sin θ cos ϕ, −α sin ϕ)|dθdϕ
= |(−α2 cos θ sin2 ϕ, −α2 sin2 ϕ sin θ, −α2 sin ϕ cos ϕ sin2 θ − α2 sin ϕ cos ϕ cos2 θ)|dθdϕ
q
= α2 sin4 ϕ + sin2 ϕ cos2 ϕdθdϕ
= α2 | sin ϕ|dθdϕ
= α2 sin ϕdθdϕ for 0 ≤ ϕ ≤ π
¨
1
⇒z = α cos ϕα2 sin ϕdθdϕ
2πα2 Rθϕ
ˆ π ˆ 2π
α 2
= cos ϕ sin ϕdϕdθ
2π 0 0
 2  π2
sin ϕ α
= α =
2 0 2

Note. The centroid of a solid hemisphere is z = 3α

8
˜
Exercise 7.3. Evaluate the flux integral I = S A· n̂dS for A = 18z î−12ĵ+3y k̂,
where n̂ points upward and S is the plane 2x + 3y + 6z = 12 in the first octant
using (x, y) as surface parameters.
Solution 7.3.

dS = n̂dS = (rx × ry )dxdy

 
12 − 2x − 3y
where r = r(x, y) = x, y,
6

104
Geometrically

n̂ ∝ ∇f where f = 2x + 3y + 6z = (2, 3, 6)
(2, 3, 6)
⇒ n̂ =
7
⇒ dS = sec γdxdy
1 7
= dxdy = dxdy
n̂ · k̂ 6
¨
(2, 3, 6) 7
⇒I = (18z, −12, 3y) · dxdy
Rxy 7 6
¨
1
= (36z − 36 + 18y)dxdy.
6 Rxy
12−2x−3y
Using z = 6
, so
¨
1
I = (36 − 12x)dxdy
6 Rxy
ˆ 6 ˆ 12−2x
1 3
= (36 − 12x)dx dy
6 x=0 y=0
ˆ 6
1 12 − 2x
= (36 − 12x dx
6 0 3
ˆ 6
2
= (36 + 8x + 2x2 )dx = 24.
0 3

8 Orientable Surfaces
Exercise 8.1. Verify Gauss’ Divergence Theorem for F = 4xz î − y 2 ĵ + yz k̂ and
V the cube bounded by x = 0, x = 1, y = 0, y = 1, z = 0, z = 1.

105
Solution 8.1. (a) Evaluating the total divergence over V directly
˚
I = div FdV
˚ V

= (4z − 2y + y)dxdydz
ˆ 1ˆ 1ˆ 1
= (4z − y)dydzdx
0 0 0
ˆ 1 1
y2
= 4zy − dz
0 z 0
ˆ 1 
1
= 4z − dz
0 2
h
2 z i1 3
= 2z − =
2 0 2

(b) Evaluating the flux outward through the boundary of V , ∂V ,

‹
I = F · n̂dS
¨
∂V ¨ ¨ ¨
k = − F1 dydz + F1 dydz − F2 dxdz + F2 dxdz
¨ x=0
¨ x=1 y=0 y=1

− F3 dxdy + F3 dxdy
¨ z=0 ¨ z=1 ¨
2
= 4zdydz + −(1) dxdz + ydxdy
x=1 y=1 z=1
ˆ 1ˆ 1 ˆ 1 ˆ 1 ˆ 1 ˆ 1
= 4 zdzdy − dxdz + ydydx
0 0 0 0 0 0
 1  1
1 2 1 2
= 4 −z −1+1 y
2 0 2 0
1 3
= 2−1+ = .
2 2

Exercise 8.2. Verify Stokes’ Theorem for F = (2x − y)î − yz 2 ĵ − y 2 z k̂ where S

is the upper 21 hemisphere x2 + y 2 + z 2 = 1.
Solution 8.2. Choose the upper pointing normal n̂ on S, so that ∂S : x2 + y 2 = 1.

106
(a) Using polar coordinates (r = 1, θ) : r = (cos θ, sin θ, 0), 0 ≤ θ ≤ 2π,
ds = rdθ = dθ, T̂ = drds
dr
= dθ = (− sin θ, cos θ, 0)
˛ ˛
I = F · dr = F · T̂ds
∂S ∂S
ˆ 2π
= (2 cos θ − sin θ, 0, 0) · (− sin θ, cos θ, 0)dθ
θ=0
ˆ 2π
= (−2 sin θ cos θ + sin2 θ)dθ
0
 2π
2 θ 1
= − sin θ + − sin 2θ =π
2 4 0

is the circulation of F around ∂S.

(b) Using spherical polar coordinates (r = 1, θ, ϕ) : r = (cos θ sin ϕ, sin θ sin ϕ, cos ϕ),
n̂ = êr (radial), dS = r2 sin ϕdθdϕ = sin ϕdθdϕ
¨ ˆ ¨
I = ∇ × F · n̂dS = (0, 0, 1) · n̂dS = k̂ · n̂dS
S Rθϕ S

Converting to cartesian coordinates:

¨ ¨
I = cos γ sec γdxdy = dxdy
S S
= area of projection of S onto the xy-plane
¨
= dxdy = π
Rxy

Or using spherical polar coordinates (θ, ϕ) :

¨ ˆ 2π ˆ π   π2
2 1 2
I = cos ϕdS = cos ϕ sin ϕdϕdθ = 2π sin ϕ = π
Rθϕ 0 0 2 0

Exercise 8.3. Use Stokes’ Theorem to evaluate

¨
I= curl F · n̂dS,
S

where n̂ is the upward pointing normal, F = (x2 + y − 4)î + 3xy ĵ + (2xz + z 2 )k̂
and S is the curved surface of the paraboloid z = 4−(x2 +y 2 ) above the xy-plane,
so z = 4 − r2 .

107
Solution 8.3. By Stokes’ Theorem
˛
I= F · dr.
∂S

Using θ on ∂S:

r = (2 cos θ, 2 sin θ, 0)
dr
dr = dθ = (−2 sin θ, 2 cos θ, 0),
dθ
So that
ˆ 2π
I = (4 cos2 θ = 2 sin θ − 4, 12 cos θ sin θ, 0) · (−2 sin θ, 2 cos θ, 0)dθ
0
ˆ 2π
= 4 (4 cos2 θ sin θ − sin2 θ + 2 sin θ)dθ
0
 2π
4 3 θ 1
= 4 − cos θ − + sin 2θ − 2 cos θ = −4π
3 2 4 0

Exercise 8.4. Evaluate ¨

I= r · n̂dS
S

where S is the sphere of radius α with center (0, 0, 0).

Solution 8.4.

(a) By inspection r||n̂ and so r · n̂ = α

¨
⇒I=α dS = α4πα2 = 4πα3 .
S

(b) As a surface integral, I can be evaluated using cartesians or spherical polars.

(c) Using Gauss’ Divergence Theorem:

˚ ˚
4
I= div rdV = 3dV = 3 πα3 = 4πα3
V V 3

because div r = ∇ · r ≡ 3 in R3 .

108
Exercise 8.5. Evaluate ˚
I= div FdV
V

where F = (x2 + y 2 + z 2 )(x, y, z) and V : x2 + y 2 + z 2 ≤ α2 .

Solution 8.5.

(a) Switch to a surface integral x2 + y 2 + z 2 ≡ α2 on ∂V . By Gauss’ Divergence

Theorem
¨ ¨ ¨
2 3
I = F · n̂dS = α r · n̂dS = α dS = α3 (4πα2 ) = 4πα5 .
∂V ∂V ∂V

(b) As a volume integral div F = ∇ · F = 5r2 , so evaluate the triple integral in

spherical polar coordinates (0 ≤ r ≤ α, θ, ϕ):
ˆ 2π ˆ π ˆ α
5α5
I = sin ϕdϕ dθ 5r2 dr = 2(2π) = 4πα5
0 0 0 5

9 The Heat Equation

Exercise 9.1. Find the general solution to the Heat Equation in one space dimen-
sion x for heat diffusion along a rod

∂T ∂ 2T 0≤x≤L
= a2 2
∂x ∂x 0≤t

for T = T (x, t) in the rod, subject to the initial conditions:

IC: T (x, 0) = f (x) = initial temperature distribution in rod 0 ≤ x ≤ L.

and boundary conditions:

T (0, t) = T1
BC: t ≥ 0.
T (L, t) = T2

This is a well posed boundary value and initial value (BV/IV) problem for the
partial differential equation (PDE). It has a unique solution.

109
Solution 9.1. Assuming T (x, t) is a separable function of x and t, so that T (x, t) =
X(x)Y (t), then we may employ separation of variables to solve the heat equation,
which takes the form
Y 0 (t) X 00 (x)
X(x)Y 0 (t) = a2 X 00 (x)Y (t) ⇒ = a2 = λ = constant.
Y (t) X(x)
This is because a ratio of functions of t being equal to a ratio of functions of x
implies the two ratios are both equal to a constant, λ, which is independent of
x and t. The heat equation is thus broken into two separate ordinary differential
equations (ODEs) of the form:

Y 0 (t) = λY (t)
⇒ Y (t) = Ceλt

λ
X 00 (x) =X(x) X(0) = 0 = X(L)
a2 √ √
⇒ X(x) = Ae λax + Be− λax

Imposing the boundary condition X(0) = 0 = X(L) implies

Ae0 + Be0 = 0 ⇒ A √ = −B
√
λ
√
λ λ n2 π 2 a2
B(e a L − e− a L ) = 0 ⇒ L = inπ, ⇒ λ = −
a L2
where we have made use of Euler’s formula eiθ = cos θ + i sin θ so that einπ −
e−inπ = 0. Thus, functions of the form
n2 π 2 a2
 nπx 
ψn = Bn e− L2 t sin
L
are solutions to the heat equation. Since the differential operator
∂ ∂2
L≡ − a2 2
∂t ∂x

P∞ combination of the functions {ψn } is also a solution to

is linear, then any linear
the heat equation L[ n=1 Bn ψn ] = 0. Imposing the initial conditions, we obtain
∞
X  nπx 
T (x, t = 0) = f (x) = Bn sin .
n=1
L

110
To determine the coefficients Bn , that is, the Fourier sine series coefficients of
f (x), we may use the orthogonality of {ψn }:
ˆ L
mπx nπx
hψm , ψn i = sin sin dx
0 L L
ˆ
1 L (m − n)πx (m + n)πx
= cos − cos dx
2 0 L L
 L
1 L (m − n)πx L (m + n)πx
= sin − sin
2 (m − n)π L (m + n)π L 0

0, m 6= n L
= L = δnm
2
, m=n 2

1, n = m
where δnm is the Kronecker delta δnm ≡ . So applying the integral
0, n 6= m
´L
operator 0 dx sin mπx L
to our expression for the initial temperature T (x, t = 0) =
f (x), we obtain
ˆ L ˆ L ∞
mπx mπx X nπx
dx sin f (x) = dx sin Bn sin
0 L 0 L n=1 L
∞ ˆ L
X mπx nπx
= Bn sin sin dx
n=1 0 L L
∞ ˆ L
X L L
= Bn δnm = Bm
n=1 0 2 2

yielding the expression for the Fourier sine coefficients of f (x)

ˆ
2 L mπx
Bm = sin f (x)dx.
L 0 L
The general solution to the heat equation in one dimension, with the temperature
zero on the boundaries is then
∞ ˆ
2 L nπx0 0 −( nπa 2 nπx
dx e L ) t sin
X
T (x, t) = f (x0 ) sin .
n=1
L 0 L L

Exercise 9.2. Suppose f (x) = T0 , a “square wave” on the rod, is the initial con-
dition for the standard homogeneous heat problem in 1D (T (0, t) = T (L, t) = 0).

111
Find the Fourier sine series for f (x) and the temperature T (x, t) for 0 ≤ x ≤ L
and t ≥ 0, and show that
4T0 −( πa 2 πx
T (x, t) ≈ e L ) t sin .
π L
Solution 9.2. From Exercise ??, the solution to the standard homogeneous heat
problem in 1D (T (0, t) = T (L, t) = 0) is
∞
Bn e−( ) sin nπx
nπa 2
X
t
T (x, t) = L

n=1
L

where the Fourier sine coefficients of f are

ˆ
2 L nπx0 0
Bn = f (x0 ) sin dx .
L 0 L

Evaluating this integral for f (x) = T0 , we find

ˆ L
2 L nπx0 0 2T0

L nπx 2T0
Bn = T0 sin dx = − cos = [1 − cos nπ]
L 0 L L nπ L 0 nπ
2T0 (1 − (−1)n )

0 if n is even
= = 4T0
nπ nπ
if n is odd

Substituting for Bn into our expression for f (x) we find

∞ ∞ (2n−1)πx
X nπx 4T0 X sin L
f (x) = Bn sin =
n=1
L π n=1 2n − 1
 
4T0 πx 1 3πx
≈ sin + sin + ···
π L 3 L

Then ∞
4T0 X 1 (2n−1)πa 2 (2n − 1)πx
T (x, t) = e−( L ) t sin .
π n=1 2n − 1 L
for 0 ≤ x ≤ L and t ≥ 0. This “solves” the Boundary Value/Initial Value (BV/IV)
problem for T (x, t) in the rod as a function of x and t for t ≥ 0 and 0 ≤ x ≤ L.
πa 2
Note that because of the e−( L ) t factors whose exponents are proportional to
n2 , the terms decrease very rapidly in magnitude as t increases.

112
 (a) N = 10

6 6

sin[(2n − 1)πx] /(2n − 1)

20/ π sin[(2n − 1)πx] /(2n − 1)

4 4

2 2

X
n
0 0

20/ π
2 2

0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
x x
(b) N = 100

6 6
sin[(2n − 1)πx] /(2n − 1)
20/ π sin[(2n − 1)πx] /(2n − 1)

4 4

2 2
X
n

0 0
20/ π

2 2

0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
x x

L 2


For values of t ≥ πa
, all terms but the first are very small, and so the answer

113
 L 2


is dominated by just the first term. So for t & πa
the solution behaves as

4T0 −( πa 2 πx
T (x, t) ≈ e L ) t sin ,
π L
1
which exhibits exponential decay, and a 2
sinusoidal profile.
Exercise 9.3. Find the general Fourier series for the 1⁄2-wave rectified sine wave
f (x) = 1/2 (| sin πx| + sin πx) for −1 ≤ x ≤ 1.
Solution 9.3. Using Lemma 10.10, and that the odd part of f (x), fo (x) = 1/2 sin πx
while the even part of f (x), fe (x) = 1/2| sin πx|, the Fourier Sine Coefficients are
simply
ˆ 1  
1 1
Bn = 2 sin nπx sin πx dx = δn0 ,
0 2 2

while the Fourier Cosine Coefficients are

ˆ 1 ˆ 1
1
An = 2 cos nπx sin πx dx = cos nπx sin πxdx
0 2 0
ˆ
1 1
= sin(n + 1)πx − sin(n − 1)πxdx
2 0
 1
1 cos(n + 1)πx cos(n − 1)πx
= − +
2π n+1 n−1 0
n n
−((−1)n + 1)
 
1 (−1) + 1 (−1) + 1
= − =
2π n+1 n−1 π(n2 − 1)
−2
A2n =
π(4n2 − 1)
A2n−1 = 0.

and
ˆ 1 1
1 1 1
A0 = sin πxdx = − cos πx = ,
2 0 2π 0 π

so that ∞
1 1 2 X cos 2nπx
f (x) = + sin πx − .
π 2 π n=1 4n2 − 1

114
10 Fourier Transform
Exercise 10.1. Defining αn = nπL
and ∆α = αn+1 − αn = Lπ and substituting into
Definition 10.16 for the General Fourier Series, derive the Fourier transform of f
by taking the limit L → ∞.
Solution 10.1. If f : R → R is piecewise-smooth on the interval [−L, L], and
continuous at x ∈ [−L, L], then f (x) equals the sum of its general Fourier series,
that is,
X∞  nπx   nπx 
f (x) = A0 + An cos + Bn sin ,
n=1
L L
where
ˆ ˆ
1 L  nπx  1 L
An = cos f (x)dx, A0 = f (x)dx,
L −L L 2L −L
ˆ
1 L  nπx 
Bn = sin f (x)dx, B0 = 0.
L −L L
by Definition 10.16.
Letting αn = nπL
and ∆α = αn+1 − αn = Lπ , we find
ˆ
∆α L
An = cos αn xf (x)dx = A(αn )∆α
π −L
ˆ
∆α L
Bn = sin αn xf (x)dx = B(αn )∆α
π −L
X∞
f (x) = A0 + An cos αn x + Bn sin αn x
n=1
∞
X
= A0 + [A(αn ) cos αn x + B(αn ) sin αn x] ∆α.
n=1

Now we let the interval [−L, L] become arbitrarily large, that is, L → ∞, so
´that ∆α = Lπ → 0 and the points αn become increasingly finely spaced. If
∞
−∞
|f (x)|dx exists, then
ˆ L ˆ ∞
1 1
A0 = lim f (x)dx ≤ lim |f (x)|dx = 0,
L→∞ 2L −L L→∞ 2L −∞
ˆ L ˆ
1 −1 ∞
A0 = lim f (x)dx ≥ lim |f (x)|dx = 0.
L→∞ 2L −L L→∞ 2L −∞

115
so that A0 = 0.
We thus obtain the integral formula for f (x),
ˆ ∞
f (x) = [A(α) cos αx + B(α) sin αx] dα
0

where
ˆ ∞ ˆ ∞
1 1
A(α) = f (x) cos αxdx, B(α) = f (x) sin αxdx.
π −∞ π −∞

11 Sturm-Louiville Theory
Exercise 11.1. Find the general solution to the Wave Equation on a circular do-
main D = {0 ≤ r ≤ b, −π ≤ θ ≤ π} in polar coordinates

∂ 2u
2
= c2 ∇ 2 u
∂t
subject to the initial conditions:

∂u(r, θ; t = 0)
IC: u(r, θ; t = 0) = f (r, θ), = g(r, θ),
∂t
and boundary conditions:

∂u(r, π, t) ∂u(r, −π, t)

BC: u(b, θ, t) = 0; u(r, π, t) = u(r, −π, t); = .
∂t ∂t
Solution 11.1. Formulating the wave equation in polar coordinates, we obtain

∂ 2u
 2
1 ∂ 2u

2 ∂ u 1 ∂u
=c + + .
∂t2 ∂r2 r ∂r r2 ∂θ2

Assuming a separable form for the solution, so that u(r, θ, t) = R(r)Θ(θ)T (t),
we find by subsituting into the wave equation in polar coordinates, and dividing
by R(r)Θ(θ)T (t) we obtain

1 T 00 (t) R00 (r) 1 R0 (r) 1 Θ00 (θ)

= + + =λ (39)
c2 T (t) R(r) r R(r) r2 Θ(θ)

116
where λ is a constant. This is because the left hand side of (??) consists of func-
tions of time t alone, while the right hand side of (??) exhibits only spatial (r, θ)
dependence. Rearranging, we obtain the ODE for T (t)

T 00 (t) = λc2 T (t).

Since we are only interested in oscillatory solutions, we may express T (t) as a

sum of sinusoidal functions by specifying λ < 0, so that λ = −µ2 .

T (t) = A cos µt + B sin µt.

Rearranging the spatial dependent part of (??) into radial and angular depen-
dent parts, we find

R00 (r) R0 (r) Θ00 (θ)

r2 +r + µ2 r 2 = − = ν, (40)
R(r) R(r) Θ(θ)

where ν is a constant. This is because the left hand side of (??) consists of ra-
dial dependent functions of r alone, while the right hand side of (??) consists of
angular dependent functions of θ alone. Rearranging, we obtain the ODE for Θ(θ)

Θ00 (θ) = −νΘ(θ),

√ √
⇒ Θ(θ) = C cos νθ + D sin νθ.

Enforcing the boundary condition Θ(π) √ = Θ(−π), and √ disgarding the trivial√solu-
tions C = 0 and D = 0, we find sin νπ = 0 = − sin νπ which implies ν =
n ∈ Z+ . This implies basis functions for the angular dependent part of u(r, θ, t)
are Θ(θ) ∈ {1, cos θ, cos 2θ, . . . , cos nθ, . . . ; sin θ, sin 2θ, . . . , sin nθ, . . .}, so that
Θn (θ) ∈ {cos nθ, sin nθ}, where n ∈ Z+ .
Substituting v = n2 into (??) and rearranging, we obtain the ODE for R(r)

r2 R00 (r) + rR0 (r) + (µ2 r2 − n2 )R(r) = 0. (41)

Performing the substitution x = µr into (??), we obtain Bessel’s Equation 12.5

x2 R00 (x) + xR0 (x) + (x2 − n2 )R(x) = 0. (42)

Employing Theorem 12.1, the Method of Frobenius, and substituting a power

117
series expansions for R(x), R0 (x), and R00 (x) around x = 0, that is
∞
X
R(x) = ak xk+α
k=0
X∞
R0 (x) = ak (k + α)xk+α−1
k=0
∞
X
R00 (x) = ak (k + α)(k + α − 1)xk+α−2
k=0

into (??), we obtain

∞
X
ak (k + α)(k + α − 1)xk+α + (k + α)xk+α + (x2 − n2 )xk+α = 0,
 
k=0
∞
X
ak ((k + α)2 − n2 ) + ak−2 xk+α = 0.
 
k=0

Since these equations must be satisfied for all x, this implies

k=0: a0 ((0 + α)2 − n2 ) = 0 ⇒ α = ±n.

k=1: a1 ((1 ± n)2 − n2 ) = a1 (1 ± 2n) = 0 ⇒ a1 = 0.
k≥2: ak ((k ± n)2 − n2 ) + ak−2 = 0
ak−2 (−1)ak−2
⇒ ak = 2 = .
n − (k ± n)2 k(k ± 2n)

Since a0 6= 0 by assumption, and a1 = 0, only the even terms 2k of the power

series expansion for R(x) survive. Further, since the solution u(r, θ, t) is assumed
to be well defined over the region, we may assume α > 0, yielding

(−1)a2(k−1) (−1)k a0
a2k = ⇒ a2k = 2k .
4k(k + n) 2 k!(k + n)!

So that R(x) = Jn (x) may be expressed as a power series of the form

∞
X (−1)k  x 2k+n
Jn (x) = ,
k=0
k!(k + n)! 2

118
where we have assumed a0 = 1/2n . Enforcing the boundary condition u(r =
b, θ, t) = 0, we find Jn (µb) = 0.
The Bessel function Jn (x) has an infinite sequence of positive roots, denoted
jnm , for n ∈ N. It then follows that
jnm
µb = jnm ⇒ µ = .
b
The spatial dependent part of u(r, θ, t), that is, R(r)Θ(θ), has the set of eigen-
functions
     
j0,m r jnm r jnm r
J0 , Jn sin nθ, Jn cos nθ, n ∈ Z+ , m ∈ N,
b b b
and the time dependent part of u(r, θ, t), that is T (t), has the set of eigenfunctions
jnm jnm
cos t, sin t.
b b
The general solution for u(r, θ, t) on a disk of radius b may be expressed as

∞    
X jnm t jnm t jnm r
u(r, θ, t) = Anm cos + Bnm sin Jn (sin nθ + cos nθ) .
nm
b b b

Exercise 11.2. Using separation of variables, show that the set of eigenfunctions
to Laplace’s equation in spherical coordinates, that is,
∂ 2f 2 ∂f 1 ∂ 2f 1 ∂ 2f cot ϕ ∂f
∇2 f (r, θ, ϕ) = 2
+ + 2 2 2
+ 2 2
+ 2 = 0,
∂r r ∂r r sin ϕ ∂θ r ∂ϕ r ∂ϕ

where f (r, θ, ϕ) = R(r)Θ(θ)Φ(ϕ), has the form f (r, θ, ϕ) = (Ar` +B −`−1 )Y`m (θ, ϕ),
where 0 ≤ m ≤ ` and Y`m (θ, ϕ) = eimθ P`m (cos ϕ) are the spherical harmonics.
Solution 11.2. Substituting f (r, θ, ϕ) = R(r)Θ(θ)Φ(ϕ) into Laplace’s equation
in spherical coordinates, multiplying by r2 and dividing by R(r)Θ(θ)Φ(ϕ), yields
00
2R (r) R0 (r) 00
2 Θ (θ) Φ00 (ϕ) Φ00 (ϕ)
r + 2r = − csc ϕ − − cot ϕ =λ (43)
R(r) R(r) Θ(θ) Φ(ϕ) Φ(ϕ)
where λ is a constant. This is because the left hand side of (??) depends only on
the radial coordinate r, while the right hand side of (??) consists of only angularly
dependent functions of (θ, ϕ).

119
 Focusing on the angular dependence of (??), we have the ODE
Θ00 (θ) Φ00 (ϕ) Φ(ϕ)
csc2 ϕ + + cot ϕ +λ=0
Θ(θ) Φ(ϕ) Φ(ϕ)
Rearranging in terms of θ and ϕ, we find
Θ00 (θ) 00
2 Φ (ϕ) Φ0 (ϕ)
− = sin ϕ + cos ϕ sin ϕ + λ sin2 ϕ = ν, (44)
Θ(θ) Φ(ϕ) Φ(ϕ)
since the left hand side of (??) only has dependence on θ while the right hand side
only has dependence on ϕ. We then obtain the separate ODE for Θ(θ) of

Θ00 (θ) = −νΘ(θ),

√ √
⇒ Θ(θ) = A cos νθ + B sin νθ.
√
Enforcing the boundary condition Θ(π) = Θ(−π), we find sin νπ = 0 ⇒ ν =
m2 ∈ N, so that Θm (θ) ∈ {1, cos mθ, sin mθ}.
Rearranging (??) for Φ(ϕ) we find

sin2 ϕΦ00 (ϕ) + cos ϕ sin ϕΦ0 (ϕ) + (λ sin2 ϕ − m2 )Φ(ϕ) = 0.

 
sin ϕ ∂ ∂Φ
sin ϕ + λ sin2 ϕ = m2 .
Φ(ϕ) ∂ϕ ∂ϕ

Performing the substitution t = cos ϕ, we find

∂Φ ∂Φ ∂t
sin ϕ = sin ϕ = − sin2 ϕΦ0 (t) = (t2 − 1)Φ0 (t).
∂ϕ ∂t ∂ϕ
yielding
d  2
0 = (t2 − 1) (t − 1)Φ0 (t) + (λ(1 − t2 ) − m2 )Φ(t),

dt
m2
   
d 2 d
⇒0 = (1 − t ) Φ(t) + λ − Φ(t).
dt dt 1 − t2

Making the substitution λ = `(` + 1) we obtain the associated Legendre equation

of Defintion 12.3,
2
m2
 
2 d m d m
(1 − x ) 2 P` (x) − 2x P` (x) + `(` + 1) − P m (x) = 0.
dx dx 1 − x2 `

120
whose solutions are the Legendre polynomials P`m (x). Here, ` ∈ N and 0 ≤ m ≤
`. The substitution λ = `(` + 1) is required to ensure P`m (cos ϕ) are nonsingular
for ϕ = 0 and ϕ = π.
Employing the Method of Frobenius, we find
∞
X
Φ(t) = ak xk+α
k=0
∞
X
Φ0 (t) = ak (k + α)xk+α−1
k=0
∞
X
Φ00 (t) = ak (k + α)(k + α − 1)xk+α−2
k=0

so that by substituting in the associated Legendre equation we obtain

∞
m2
    
X 1
ak (k + α)(k + α − 1) 2
− 1 − 2(k + α) + `(` + 1) − 2
xk+α = 0
k=0
x 1−x

Focusing on the right hand side of (??), we obtain the ODE for R(r)
r2 R00 (r) + 2rR0 (r) = λR(r) = `(` + 1)R(r).
d  2 0 
r R (r) = `(` + 1)R(r).
dr
d2
r 2 [rR(r)] = `(` + 1)R(r),
dr
where we have used λ = `(` + 1), ` ∈ N. Letting X(r) = rR(r), we obtain the
simplified ODE for X(r)
r2 X 00 (r) = `(` + 1)X(r).
⇒ X(r) = Ar`−1 + Br−`−2
⇒ R(r) = Ar` + Br−`−1 .
Requiring R(r) to be regular in R3 would force B = 0, while requiring R(r) → 0
as r → ∞ would force A = 0.
The general form of the solution to Laplace’s equation in spherical coordinates
is then
R(r)Θ(θ)Φ(ϕ) = (Ar` + Br−`−1 )eimθ P`m (cos ϕ) = (Ar` + Br−`−1 )Y`m (θ, ϕ)
as required.

121
12 Green’s Functions
Exercise 12.1. Show that the Green’s function G(r, r0 ) for Poisson’s equation in
cylindrical coordinates, ∇2r,θ,z Φ(r, θ, z) = n(r, θ, z), for a cylindrical shell of ra-
dius r0 satisfies
 2
1 ∂2 ∂2

∂ 1 ∂ 1
2
+ + 2 2 + 2 G(r, r0 ) = δ(r − r0 )δ(θ − θ0 )δ(z − z 0 ),
∂r r ∂r r ∂θ ∂z r

and is given by the Fourier-Bessel transformation

∞ ˆ ∞
0 1 X dk ik(z−z0 ) im(θ−θ0 )
G(r, r ) = = − e e Im (|k|r< )Km (|k|r> ),
|r − r0 | m=−∞ −∞ (2π)2

where r< ≡ min(r, r0 ), r> ≡ max(r, r0 ), and Im (|k|r< ) and Km (|k|r> ) are mod-
ified Bessel’s functions of the first and second kind, respectively.
Solution 12.1. Suppose ∇2r G(r, r0 ) = 1r δ(r − r0 )δ(θ − θ0 )δ(z − z 0 ). This form is
symmetric in r and r0 , that is, G(r, r0 ) = G(r0 , r). If such a function is found, the
solution to Poisson’s equation is then
˚
Φ(r, θ, z) = G(r, r0 )n(r0 , θ0 , z 0 )r0 dr0 dθ0 dz 0 .

Checking this, we find

˚
∇2r,θ,z Φ(r, θ, z) = ∇2r,θ,z G(r, r0 )n(r0 , θ0 , z 0 )r0 dr0 dθ0 dz 0
˚
= ∇2r,θ,z G(r, r0 )n(r0 , θ0 , z 0 )r0 dr0 dθ0 dz 0
˚
1
= δ(r − r0 )δ(θ − θ0 )δ(z − z 0 )n(r0 , θ0 , z 0 )r0 dr0 dθ0 dz 0
r
= n(r, θ, z),

as required. Therefore, the Green’s function for Poisson’s equation in cylindrical

coordinates satisfies
 2
1 ∂2 ∂2

∂ 1 ∂ 1
2
+ + 2 2 + 2 G(r, r0 ) = δ(r − r0 )δ(θ − θ0 )δ(z − z 0 ). (45)
∂r r ∂r r ∂θ ∂z r

122
Taking the inverse Fourier transform of the Green’s function G(r, r0 ) = F −1 {gm (k; r, r0 )}
from Definition 11.1 with respect to z ∈ R, and the Fourier series of (??) with
respect to θ ∈ [−π, π], we obtain after inserting into (??)

∇2r,θ,z G(r, r0 )
= ∇2r,θ,z F −1 {gm (k, r, r0 )}
∞ ˆ ∞
X dk ik(z−z0 ) im(θ−θ0 )
2
= ∇r,θ,z 2
e e gm (k, r, r0 )
m=−∞ −∞
(2π)
∞ ˆ ∞
X dk 0 0
= 2
∇2r,θ,z eik(z−z ) eim(θ−θ ) gm (k, r, r0 )
m=−∞ −∞
(2π)
∞ ˆ ∞  2  2 
X dk ∂ 1 ∂ m 0 0
= 2 2
+ − 2
+k 2
eik(z−z ) eim(θ−θ ) gm (k, r, r0 )
m=−∞ −∞
(2π) ∂r r ∂r r
δ(r − r0 )
= δ(θ − θ0 )δ(z − z 0 )
r
∞ ˆ ∞
δ(r − r0 ) X dk ik(z−z0 ) im(θ−θ0 )
= 2
e e .
r m=−∞ −∞ (2π)

where gm (k; r, r0 ) is the Fourier transform with respect to z and the Fourier series
with respect to θ of the Green’s function G(r, r0 ). Rearranging, and noting that
since the integral is zero, the integrand must be zero, we find

δ(r − r0 )
 2  2 
∂ 1 ∂ m 2 0
+ − + k gm (k; r, r ) = .
∂r2 r ∂r r2 r

Multiplying by r2 , and making the substitution κ ≡ kr, so that

∂ ∂ ∂κ ∂
= =k ,
∂r ∂κ ∂r ∂κ
we find that
∂2 0 ∂ 0
2 2 2
gm (κ, κ0 ) = 0,


κ 2
g m (κ, κ ) + κ g m (κ, κ ) − m + κ (46)
∂κ ∂κ
when r 6= r0 . Equation (??) is Bessel’s Modified Equation, as provided in Def-
inition 12.6, and has solutions Im (κ) and Km (κ), which are the modified Bessel
functions of order m of the first and second kind, respectively. Moreover, as shown

123
in Example 12.2, Im (κ) diverges for large κ while Km (κ) diverges for small κ.
Mathematically, Im (κ) → ∞ as κ → ∞ and Km (κ) → ∞ as κ → 0+ . This
suggests that “inside” a cylindrical shell of charge of radius r0 , the electric poten-
tial Φ should behave as Im (|k|r), while “outside” the cylindrical shell, the electric
potential Φ should behave as Km (|k|r).
To determine the form of gm (κ, κ0 ), we must apply the boundary condition
on the cylindrical shell, with the “delta function” implying a discontinuity in the
slope of gm (k; r, r0 ) of

∂gm (k; r, r0 ) ∂gm (k; r, r0 ) 1

− = ,
∂r r0+ ∂r r0− r0
0
∂gm (κ, κ ) ∂gm (κ, κ0 ) 1
− = 0,
∂κ κ0+ ∂κ κ0− κ

Using the symmetric property of the Green’s function, and that the Wronskian of
Im (|κ|) and Km (|κ|) is

0 0 1
Im (|κ|)Km (|κ|) − Im (|κ|)Km (|κ|) = − ,
|κ|

The radial Green’s function gm (k; r, r0 ) = gm (κ, κ0 ) is given by

gm (k; r, r0 ) = −Im (|k|r< )Km (|k|r> ),

gm (κ, κ0 ) = −Im (κ< )Km (κ> ),

where r< = min(r, r0 ) and r> = max(r, r0 ). Substituting into our definition for
G(r, r0 ), we find
∞ ˆ ∞
0
X dk ik(z−z0 ) im(θ−θ0 )
G(r, r ) = − e e Im (|k|r< )Km (|k|r> ),
m=−∞ −∞ (2π)2

and the electronic potential Φ(r, θ, z) has the form

˚ Xˆ
dk ik(z−z0 ) im(θ−θ0 )
− 2
e e Im (|k|r< )Km (|k|r> )n(r0 , θ0 , z 0 )r0 dr0 dθ0 dz 0 .
m
(2π)

124