Course 231: Equations of Mathematical Physics

Notes by Chris Blair

These notes cover the 2007-2008 Methods course given by
Dr. Conor Houghton, up until the end of Hilary term
(excluding power series).
Contents
I Vector Calculus 2
1 2 and 3-Dimensional Integration 2
1.1 2-Dimensional Integration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 3-Dimensional Integration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Changing Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2 Vector Operators 4
2.1 Grad, Div and Curl . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.2 Vector Identities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
3 Line Integrals and Conservative Fields 5
3.1 Line Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
3.2 Conservative Fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
4 Surface Integrals 6
5 Integrating Scalars 7
6 The Integral Theorems 7
6.1 Greens Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
6.2 Stokes Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
6.3 Gauss Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
7 Vector Potentials 10
II Fourier Analysis 10
8 Fourier Series 10
8.1 Real Fourier Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
8.2 Complex Fourier Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
8.3 Dirichlets and Parsevals Theorems . . . . . . . . . . . . . . . . . . . . . . . . . 11
9 Fourier Integrals 12
1
10 The Dirac Delta Function 13
10.1 Denition and Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
10.2 Delta Function and Fourier Integrals . . . . . . . . . . . . . . . . . . . . . . . . . 14
III Ordinary Dierential Equations 15
11 First Order ODEs 15
12 Second Order ODEs 16
12.1 Homogeneous Constant Coecients . . . . . . . . . . . . . . . . . . . . . . . . . . 16
12.2 Solving Using Fourier Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
12.3 Cauchy-Euler Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
Part I
Vector Calculus
1 2 and 3-Dimensional Integration
1.1 2-Dimensional Integration
We wish to integrate a scalar eld (x, y) over a 2-dimensional region R
_ _
R
dA(x, y)
To do so we write the integral as a double integral, iterated either rst in the y and then the x
direction, or vice versa.
y
x
a b
y = d(x)
y = c(x)
y
x
d
c
x = a(y)
x = b(y)
If the top boundary of R can be written as y = d(x) and the bottom as y = c(x) then we can
write
_ _
R
dA(x, y) =
_
b
a
dx
_
d(x)
c(x)
dy (x, y)
2
i.e. we integrate rst over y from c(x) to d(x), and then over x from a to b . Similarly in the
other direction,
_ _
R
dA(x, y) =
_
d
c
dy
_
b(y)
a(y)
dx(x, y)
1.2 3-Dimensional Integration
The idea in 3-dimensional integration is similar; we write the top surface as z = f(x, y) and the
bottom surface as z = e(x, y), and treat the projection of the region in the xy plane as in the
2-dimensional case, hence
_ _ _
R
dV (x, y, z) =
_
b
a
dx
_
d(x)
c(x)
dy
_
f(x,y)
e(x,y)
dz (x, y, z)
1.3 Changing Coordinates
Jacobian
It is often convenient to integrate using a dierent set of coordinates. Change of coordinates
involves a scaling factor known as the Jacobian. In two-dimensions for old coordinates x, y and
new coordinates u, v we have
dA = dxdy = Jdudv
J =
(x, y)
(u, v)
=

x
u
x
v
y
u
y
v

and in 3-dimensions, for old coordinates x, y, z and new coordinates u, v, w we have

dV = dxdv dz = Jdudv dw
J =
(x, y, z)
(u, v, w)
=

x
u
x
v
x
w
y
u
y
v
y
w
z
u
z
v
z
w

Examples
Cylindrical Polar Coordinates
x = r cos y = r sin z = z
J = r
where r [0, ), [0, 2).
Spherical Coordinates
x = r sin cos y = r sin sin z = r cos
J = r
2
sin
where r [0, ), [0, 2), [0, ].
3
2 Vector Operators
2.1 Grad, Div and Curl
There are some important vector operators related to the operator

=

x

i +

y

j +

z

k.
grad
The gradient of a scalar eld is the vector eld
grad =

=
_

x
,

y
,

z
_
div
The divergence of a vector eld

F = (F
1
, F
2
, F
3
) is
div

F =

F =
F
1
x
+
F
2
y
+
F
3
z
If div

F = 0 then we say

F is solenoidal.
curl
The curl of a vector eld

F = (F
1
, F
2
, F
3
) is
curl

F =

F =

i

j

k

z
F
1
F
2
F
3

If curl

F = 0 we say

F is irrotational.
Laplacian
The Laplacian of a scalar eld is the scalar eld
= div grad =

=

2

x
2
+

2

y
2
+

2

z
2
2.2 Vector Identities
Some important vector identities related to these operators are:
1.

() =

2.

(

F) =



F +

F
3.

(

F) =



F +

F
4.

(

F

G) = (

F)

G

F (

G)
5.

(

F

G) = (

G)

F + (

G

)

F (

F)

G(

F

)

G
6.

(

F

G) =

F (

G) +

G(

F) + (

F

)

G + (

G

)

F
7.

(

F) = 0
8.

(

) = 0
9.

=
4
3 Line Integrals and Conservative Fields
3.1 Line Integrals
To integrate a vector eld

F along a curve C we parameterise the curve C in terms of t as r(t)
and use
_
C

F d

l =
_

F(r(t))
dr
dt
dt
3.2 Conservative Fields
A vector eld

F is called conservative if there exists a scalar eld such that

F =

A vector eld

F is called path independent if the line integral
_
C

F d

l between any two points

is the same for all paths C between those two points.
The following are equivalent:

F is conservative,

F is path independent, and
_
C

F d

l = 0.
Proof. i) Conservative path independence: let

F =

=
_

x
,

y
,

z
_
, and consider
restricted to a curve r(t), so that on the curve we have = (r(t)) and
d
dt
=

x
dx
dt
+

y
dy
dt
+

z
dz
dt
Now consider
_
C

F d

l =
_
t
2
t
1

F(r(t))
dr
dt
dt
=
_
t
2
t
1

dr
dt
dt
=
_
t
2
t
1
_

x
dx
dt
+

y
dy
dt
+

z
dz
dt
_
dt
=
_
t
2
t
1
d
dt
dt
= (t
1
) (t
2
)
by the Fundamental Theorem of Calculus, showing that

F is path independent.
ii) Path independence
_
C

F d

l = 0: let C
a
and C
b
be two curves with the same endpoints P
1
and P
2
, then
_
C
a

F d

l =
_
C
b

F d

l
and now consider the closed curve C = C
a
C
b
, then
_
C

F d

l =
_
C
a

F d

l
_
C
b

F d

l = 0
for all closed loops C.
5
iii)
_
C

F d

l = 0 path independence: let

_
C

F d

l = 0 for all closed loops C, and let C

1
and
C
2
be two paths between two points a and b. Then we have
_
C

F d

l =
_
C
1

F d

l
_
C
2

F d

l = 0

_
C
1

F d

l =
_
C
2

F d

l
for all C
1
and C
2
, so path independence.
iv) Path independence conservative: let p be some point and let
(x) =
_
C(p,x)

F d

l
where C(p, x) is any curve from p to x = (x, y, z). We will show that

F = . Componentwise,
we have to prove that
F
1
=

x
=

x
_
C(p,x)

F d

l
We choose a path C that goes from p to a point p

= (x

, y, z), so that the path from p

to x is
a straight line. We then have

x
_
C(p,x)

F d

l =

x
_
C(p,p

F d

l +

x
_
C(p

,x)

F d

l
where the rst integral does not depend on x and so is zero when dierentiated. We now
parameterise C(p

, x) as
r(t) = t

i +y

j + z

dr
dt
=

i
and so we have

x
_
C(p

,x)

F d

l =

x
_
x
x

F
dr
dt
dt
=

x
_
x
x

i
=

x
_
x
x

F
1
dt
= F
1
using the Fundamental Theorem of Calculus. Similarly for the other components.
4 Surface Integrals
To integrate a vector eld

F over a surface we parameterise the surface as r(u, v) and use
_ _
S

F d

A =
_ _
D
dudv

F(r(u, v))
r
u

r
v
where D is the domain in R
2
of u, v. Note that the choice
r
u

r
v
gives one orientation of the
surface;
r
v

r
u
is the other.
6
5 Integrating Scalars
We can dene two- and three-dimensional integrals of scalars by
_
C
d

l =
_
C

=
_
t
2
t
1

dr
dt

dt
_ _
S
d

S =
_ _
S

=
_ _
D

r
u

r
v

dudv
6 The Integral Theorems
6.1 Greens Theorem
Let D be a region in the xy plane bounded by a piecewise smooth curve C oriented anti-clockwise.
Then if f(x, y) and g(x, y) have continuous rst derivatives
_ _
D
dA
_
g
x

f
y
_
=
_
C
(fdx + gdy)
Proof. Consider D a simple region (i.e. a region where a double integral can iterated in either
order). Then,
_ _
D
dA
_
g
x

f
y
_
=
_
d
c
dy
_
b(y)
a(y)
dx
g
x

_
b
a
dx
_ _
d(x)
c(x)
dy
f
y
=
_
d
c
dy g(x, y)
_
b(y)
a(y)

_
b
a
dxf(x, y)
_
d(x)
c(x)
=
_
d
c
dy g(b(y), y)
_
d
c
dy g(a(y), y)
_
b
a
dxf(x, d(x)) +
_
b
a
dxf(x, (c(x))
=
_
C
dy g(x(y), y) +
_
C
dxf(x, y(x))
=
_
C
(fdx + gdy)
as required. For an arbitrary region, we can divide the region up into many simple regions and
sum.
6.2 Stokes Theorem
Let S be a piecewise smooth orientable surface with boundary C a piecewise smooth curve
oriented so that nd

l points into the surface. Let

F be a continuously dierentiable vector eld
in the neighbourhood of S, then
_ _
S
curl

F d

S =
_
C

F d

l
Proof. Consider rst the simple case

F = F
3
(x, y, z)

k, and a simple region that can be parame-

terised by x, y, that is z = h(x, y), so that
r = x

i + y

j + h(x, y)

k
7
is the parameterised surface. We then have
r
x
=

i +
h
x

k
r
y
=

j +
h
y

r
x

r
y
=

i

j

k
1 0
h
x
0 1
h
y

=
h
x

i
h
y

j +

k
and
curl

F =

i

j

k

z
0 0 F
3

=
F
3
y

i
F
3
x

j
so
curl

F
_
r
x

r
y
_
=
F
3
y
h
x
+
F
3
x
h
y
=

y
_
F
3
h
x
_
+

x
_
F
3
h
y
_
as

x
_
F
3
h
y
_
=
F
3
x
h
y
+
F
3
h
h
x
h
y
+ F
3

2
h
xy

y
_
F
3
h
x
_
=
F
3
y
h
x
+
F
3
h
h
y
h
x
+ F
3

2
h
yx
so then
_ _
S
curl

F d

S =
_ _
D
dxdy
_

x
_
F
3
h
y
_

y
_
F
3
h
x
__
and we now apply Greens theorem with f = F
3
h
x
and g = F
3
h
y
, hence
_ _
S
curl

F d

S =
_
D
F
3
h
x
dx +F
3
h
y
dy
=
_
D
F
3
_
z
x
dx +
z
y
dy
_
=
_
D
F
3
dz
=
_
C

F d

l
as required. We can perform similar calculations for

F = F
1

i and

F = F
2

j and sum them to give

the general result. For a more general surface S we can split S into a number of simple surfaces
and integrate over each of them.
An application of Stokes theorem is to show that on a simply connected domain curl

F = 0

F
conservative. A simply connected domain is a domain such that any smooth curve can be shrunk
to a point. Given some curve C we can shrink it to a point and let S be the surface traced by
8
the shrinking, hence any closed curve can be expressed as the boundary of some surface. Then
if curl

F = 0,
_
C

F d

l =
_ _
S
curl

F d

S = 0
and hence

F is conservative.
6.3 Gauss Theorem
Let D be a connected three-dimensional region in R
3
whose boundary is a closed piecewise smooth
surface S. Then if

F is a vector eld with continuous rst derivatives in a domain containing D
_ _ _
D
div

F dV =
_ _
S

F d

S
Proof. Consider just D simple, and let

F = F
3

k, so div

F =
F
3
z
. We now write
_ _ _
D
div

F dV =
_ _
D
2
_
f(x,y)
e(x,y)
dz
F
3
z
where D
2
is the parameter region in the xy plane, so
_ _ _
D
div

F dV =
_ _
D
2
_
F
3
(x, y, f(x, y)) F
3
(x, y, e(x, y))
_
and we parameterise the top surface by
r = x

i + y

j + f(x, y)

k
then
_ _
top

F d

S =
_ _
D
2
dxdy

F(x, y, f(x, y))
r
x

r
y
and we have
r
x

r
y
=

i

j

k
1 0
f
x
0 1
f
y

=
f
x

i
f
y

j +

k
hence for

F = F
3

k we get
_ _
top

F d

S =
_ _
D
2
dxdy F
3
(x, y, f(x, y))
and similarly for the bottom surface,
_ _
bot

F d

S =
_ _
D
2
dxdy F
3
(x, y, e(x, y))
hence
_ _ _
D
div

F dV =
_ _
S

F d

S
and again the proof is similar for other components, and by summing these components works
for all

F. As before, we can treat more general domains as a collection of simple domains.
9
7 Vector Potentials
If div

F = 0 then

F = curl

A for some

A called the vector potential. The converse is true only
on domains in R
3
with no 3-dimensional obstructions (i.e. domains where any closed surface can
be shrunk to a point).
A star-shaped domain is a region D in which there exists a point a such that the line segment
between a and any point x D lies in D. On such domains if div

F = 0 then we can obtain a
vector potential for

F using the formula

A(r) =
_
1
0
dt

F(tr) tr
The Hodge decomposition of a vector eld

F neither solenoidal nor irrotational on D a simply
connected domain with no obstructions to 2-spheres is

F = curl

A + grad
Part II
Fourier Analysis
8 Fourier Series
8.1 Real Fourier Series
Consider a function f(x) with period l, then the Fourier series expansion of f(x) is
f(x) =
a
0
2
+

n=1
a
n
cos
_
2nx
l
_
+

n=1
b
n
sin
_
2nx
l
_
We can nd the Fourier coecients using the following properties of sin and cos:
_ l
2

l
2
dx cos
_
2nx
l
_
=
_ l
2

l
2
dx sin
_
2nx
l
_
= 0
_ l
2

l
2
dx cos
_
2nx
l
_
sin
_
2mx
l
_
= 0
_ l
2

l
2
dx cos
_
2nx
l
_
cos
_
2mx
l
_
=
_ l
2

l
2
dx sin
_
2nx
l
_
sin
_
2mx
l
_
=
1
2

mn
where m, n are positive integers and
mn
is the Kronecker delta,

mn
=
_
1 m = n
0 m = n
To nd a
0
we integrate both sides over a period
_ l
2

l
2
dxf(x) =
_ l
2

l
2
dx
a
0
2
+

n=1
a
n
_ l
2

l
2
dx cos
_
2nx
l
_
. .
=0
+

n=1
b
n
_ l
2

l
2
dx sin
_
2nx
l
_
. .
=0
10
hence
a
0
=
2
l
_ l
2

l
2
dxf(x)
To nd a
n
we multiply across by cos
_
2mx
l
_
and integrate over a period
_ l
2

l
2
dxf(x) cos
_
2mx
l
_
=
_ l
2

l
2
dx
a
0
2
cos
_
2mx
l
_
. .
=0
+

n=1
a
n
_ l
2

l
2
dx cos
_
2nx
l
_
cos
_
2mx
l
_
. .
=
1
2

mn
+

n=1
b
n
_ l
2

l
2
dx sin
_
2nx
l
_
cos
_
2mx
l
_
. .
=0
hence
a
n
=
2
l
_ l
2

l
2
dxf(x) cos
_
2nx
l
_
and similarly
b
n
=
2
l
_ l
2

l
2
dxf(x) sin
_
2nx
l
_
8.2 Complex Fourier Series
The complex Fourier series of a function f(x) with period l is
f(x) =

n=
c
n
exp
_
2inx
l
_
where the complex Fourier coecients are given by
c
n
=
1
l
_ l
2

l
2
dxf(x) exp
_

2inx
l
_
which comes from a similar method to before, using
_ l
2

l
2
dx exp
_
2ix
l
(n m)
_
= l
mn
One feature of the complex Fourier series is that if f(x) is real then c
n
= c
n
, where the overline
denotes the complex conjugate.
8.3 Dirichlets and Parsevals Theorems
Dirichlets Theorem: If f(x) is periodic with a nite number of minima and maxima in one
period, and with a nite number of discontinuities in one period, and
_ l
2

l
2
dx|f(x)|
2
<
11
then the Fourier series for f is convergent, and converges to f(x) for all points where f is
continuous. For points a where f(x) is discontinuous it converges to
1
2
_
lim
xa
+
f(x) + lim
xa

f(x)
_
i.e. it extrapolates across the discontinuity.
Parsevals Theorem: The L
2
norm of f(x) is
1
l
_ l
2

l
2
dx|f(x)|
2
=
1
4
a
2
0
+
1
2

n=1
(a
2
n
+ b
2
n
) =

n=
|c
n
|
2
Proof. (complex case) We have that
_ l
2

l
2
dx|f(x)|
2
=
_ l
2

l
2
dxf(x)f(x)
=
_ l
2

l
2
dx

n=
c
n
exp
_
2inx
l
_

n=
c
m
exp
_

2imx
l
_
=

n,m
c
n
c
m
_ l
2

l
2
dx exp
_
2ix
l
(n m)
_
. .
=
mn
=

n,m
c
n
c
m

mn
l
= l

n=
|c
n
|
2
9 Fourier Integrals
Consider the complex Fourier series,
f(x) =

n=
c
n
exp
_
2inx
l
_
c
n
=
1
l
_ l
2

l
2
dxf(x) exp
_

2inx
l
_
f(x) =

n=
2
l
_
1
2
_ l
2

l
2
dxf(x) exp
_

2inx
l
_
_
exp
_
2inx
l
_
and as l gets large we have k =
2
l
, and can write k =
2
l
n, and argue that in the limit l
the sum over n becomes an integral over k, hence
f(x) =
_

dk
_
1
2
_

dxf(x) exp(ikx)
_
exp(ikx)
the Fourier integral of f. We may write this as
f(x) =
_

dk

f(k) exp(ikx)
12
where

f(k) =
1
2
_

dxf(x) exp(ikx)
is known as the Fourier transform of f(x). The Fourier integral formula holds if f(x) is L
1
, that
is, if it satises
_

dx|f(x)| <
10 The Dirac Delta Function
10.1 Denition and Properties
The Dirac delta function (which is not, strictly speaking, a function) may be dened by
(x) =
_
0 x = 0
x = 0
_

(x)dx = 1
with the characteristic property
_

(x)f(x)dx = f(0)
We then have
_

(x a)f(x)dx = f(a)
established using the substitution y = x a.
We can consider the derivative of the delta function by integrating by parts
_

(x)f(x)dx = (x)f(x)
_

(x)f

(x)dx
=
_

(x)f

(x)dx
= f

(0)
We consider (ax) by letting y = ax dy = adx so
_

(ax)f(x)dx =
_

1
a
(y)f
_
y
a
_
=
1
a
f(0)
if a > 0, and
_

(ax)f(x)dx =
_

1
a
(y)f
_
y
a
_
=
_

1
a
(y)f
_
y
a
_
=
1
a
f(0)
if a < 0, so then in general
_

(ax)f(x)dx =
1
|a|
f(0)
Now let h(x) be a smooth function, and consider
_

(h(x))f(x)dx
13
If h(x) has no zeros this is identically zero. Suppose h(x) has one zero, h(x
1
) = 0 and suppose
h

(x
1
) > 0. Then we can write
_

(h(x))f(x)dx =
_
b
a
(h(x))f(x)dx
where h

(x) > 0 on (a, b), with x

1
(a, b). This means we can invert h(x) on the interval; let
y = h(x) then x = h
1
(y), and also
dy = h

(x)dx dx =
dy
h

(h
1
(y))
giving
_
b
a
(h(x))f(x)dx =
_
h(b)
h(a)
(y)f(h
1
(y))
dy
h

(h
1
(y))
and y = 0 for x = x
1
, so this integrates to
f(h
1
(0))
h

(h
1
(0))
=
f(x
1
)
h

(x
1
)
If h

< 0 then everything is the same except one of the limits of integration will change giving a
minus, so then
_

(h(x))f(x)dx =
f(x
1
)
|h

(x
1
)|
If h has multiple zeros then we can split the integral up into multiple intervals to get
_

(h(x))f(x)dx =

x
i
:h(x
i
)=0
f(x
i
)
|h

(x
i
)|
so
(h(x)) =

x
i
:h(x
i
)=0
(x x
i
)
|h

(x
i
)|
10.2 Delta Function and Fourier Integrals
Consider
(x) =
_

dk

(k) exp(ikx)
where

(k) =
1
2
_

dx(x) exp(ikx) =
1
2
(x) =
1
2
_

dk exp(ikx)
giving us the orthogonality relation
(x x

) =
1
2
_

dk exp
_
ik(x x

)
_
14
We also have that the Fourier transform of the constant function f(x) = 1 is

f(k) =

1(k) =
1
2
_

dxexp(ikx) = (k)
Parsevals Theorem for Fourier Integrals: also called Plancherels formula,
_

dx|f(x)|
2
= 2
_

dk |

f(k)|
2
Proof.
_

dx|f(x)|
2
=
_

dxf(x)f(x)
=
_

dx
_

dx

f(x)f(x

)(x x

)
=
1
2
_

dx
_

dx

dk f(x)f(x) exp
_
ik(x x

)
_
=
1
2
_

dk
__

dxf(x) exp(ikx)
___

dx

f(x

) exp(ikx

)
_
=
1
2
_

dk 2

f(k)2

f(k)
= 2
_

dk |

f(k)|
2
as required.
Part III
Ordinary Dierential Equations
11 First Order ODEs
These are equations of the form
y

+ p(x)y = f(x)
with general solution
y = y
p
+Cy
h
where y
p
is a particular solution to the inhomogeneous equation, and y
h
is the solution to the
homogeneous version. We can solve this equation by multiplying across by an integrating factor
such that

= p, as then we have
y

+ py = f
y

y =
(y)

= f
15
which we can solve by integrating. It follows that we must have
= exp(px)
if p constant, or in general,
= exp
__
x
a
p(z)dz
_
where a is an arbitrary constant.
12 Second Order ODEs
These are equations of the form
a(x)y

+ b(x)y

+c(x)y = f(x)
with general solution
y = C
1
y
1
+ C
2
y
2
+ y
p
where y
1
and y
2
are solutions of the homogeneous equation, and y
p
is the particular solution to
the inhomogeneous version.
12.1 Homogeneous Constant Coecients
In the homogeneous constant coecient case
ay

+ by

+ cy = 0
we guess solution y = e
x
which gives
a
2
+ b +c = 0
known as the auxiliary equation, with roots
1
and
2
. The general solution is then
y = C
1
e

1
x
+ C
2
e

2
x
In the case that
1
=
2
= then the general solution is
y = C
1
e
x
+ C
2
xe
x
We can solve the inhomogeneous case easily in the case that f(x) = Fe
rx
, by guessing y = Ce
rx
which gives
ar
2
C + brC + cC = F C =
F
ar
2
+ br +c
and the general solution is
y = C
1
e

1
x
+ C
2
xe

2
x
+Ce
rx
In the case that r coincides with one of the roots of the auxiliary equation, then we instead guess
y = Cxe
rx
(and in the exceptional cases that both roots are equal and equal to r, y = Cx
2
e
rx
).
16
12.2 Solving Using Fourier Analysis
We can analyse periodic driving forces using Fourier analysis by writing f(x) as a Fourier series,
hence
ay

+ by

+ cy =

n=
c
n
exp
_
2inx
l
_
and then solve
ay

n
+by

n
+ cy
n
= c
n
exp
_
2inx
l
_
giving general solution
y =

n=
y
n
Similarly, if f(x) is non-periodic but decays at we can write it as a Fourier integral
f(x) =
_

dk

f(k)e
ikx
and then solve
ay

k
+ky

n
+ cy
k
=

f(k)e
ikx
and integrate over the solutions
y =
_

dk y
k
12.3 Cauchy-Euler Equations
These are equations of the form
x
2
y

+xy

+ y = 0
which can be solved using the substitution
x = e
z
z = log x
so that
y

=
d
dx
y =
dz
dx
dy
dz
=
1
x
dy
dz
y

=
d
dx
y

=
d
dx
1
x
dy
dz
=
1
x
2
dy
dz
+
1
x
d
dx
dy
dz
=
1
x
2
dy
dz
+
1
x
2
d
2
y
dz
2
The equation becomes

dy
dz
+
d
2
y
dz
2
_
+
dy
dz
+y = 0
or

d
2
y
dz
2
+ ( )
dy
dz
+y = 0
hence we have auxiliary equation

2
+ ( ) + = 0
and solution
y = C
1
e

1
z
+C
2
e

2
z
= C
1
x

1
+ C
2
x

2
(A more direct approach would be to substitute y = x

in the original equation, though in the

case of repeated roots you need to know to multiply the second solution by log x.)
17