This action might not be possible to undo. Are you sure you want to continue?
Department of Mathematics
F
F
o
o
u
u
r
r
i
i
e
e
r
r
T
T
h
h
e
e
o
o
r
r
y
y
B
B
.
.
M
M
.
.
N
N
.
.
C
C
l
l
a
a
r
r
k
k
e
e
Table of Contents
1. Introduction................................................................................................ ................................. 1
2. Linear differential operators ................................................................................................ ...... 3
3. Separation of variables ................................................................................................ ............... 5
4. Fourier Series................................................................................................ .............................. 9
5. Bessel's inequality................................................................................................ ..................... 14
6. Convergence results for Fourier series .................................................................................... 16
7. Differentiation and Integration of Fourier Series ................................................................... 20
8. Halfrange Fourier series ................................................................................................ ......... 23
9. General Intervals ................................................................................................ ..................... 25
10. Application to Laplace's equation ............................................................................................ 27
11. SturmLiouville problems and orthogonal functions .............................................................. 33
12. Bessel Functions ................................................................................................ ....................... 37
13. Fourier Transforms ................................................................................................ .................. 41
14. Inverse Fourier Transforms ................................................................................................ ..... 47
15. Applications to Differential Equations .................................................................................... 50
16. Plancherel's and Parseval's Identities ..................................................................................... 54
17. Band Limited Functions and Shannon's Sampling Theorem................................................. 56
18. Heisenberg’s Inequality ................................................................................................ ............ 59
1. Introduction. 1
1. Introduction.
Fourier theory is a branch of mathematics first invented to solve certain problems in
partial differential equations. The most wellknown of these equations are:
Laplace's equation,
∂
2
u
∂x
2
+
∂
2
u
∂y
2
= 0, for u(x, y) a function of two variables,
the wave equation,
∂
2
u
∂t
2
– c
2
∂
2
u
∂x
2
= 0, for u(x, t) a function of two variables,
the heat equation,
∂u
∂t
– κ
∂
2
u
∂x
2
= 0, for u(x, t) a function of two variables.
In the heat equation, x represents the position along the bar measured from some origin, t
represents time, u(x, t) the temperature at position x, time t. Fourier was initially
concerned with the heat equation. Incidentally, the same equation describes the
concentration of a dye diffusing in a liquid such as water. For this reason the equation is
sometimes called the diffusion equation.
In the wave equation, x, represents the position along an elastic string under tension,
measured from some origin, t represents time, u(x, t) the displacement of the string from
equilibrium at position x, time t.
In Laplace's equation, u(x, y) represents the steady temperature of a flat conducting plate
at the position (x, y) in the plane.
Since both the heat equation and the wave equation involve a single space variable x, we
sometime refer to them as the one dimensional heat equation and the one dimensional
wave equation respectively.
Laplace's equation involves two spatial variables and is therefore sometimes called the
twodimensional laplace equation. Laplace's equation is connected to the theory of analytic
functions of a complex variable. If f(z) = u(x, y) + iv(x, y), the real and imaginary parts
u(x, y), v(x, y) satisfy the CauchyRiemann equations,
∂u
∂x
=
∂v
∂y
,
∂u
∂y
= –
∂v
∂x
,
Then
∂
2
u
∂x
2
=
∂
2
v
∂x∂y
= –
∂
2
u
∂y
2
or
∂
2
u
∂x
2
+
∂
2
u
∂y
2
= 0.
Similarly,
∂
2
v
∂x
2
+
∂
2
v
∂y
2
= 0.
Heat conduction and wave propagation usually occur in 3 space dimensions and are
described by the following versions of Laplace's equation, the heat equation and the wave
equation;
∂
2
u
∂x
2
+
∂
2
u
∂y
2
+
∂
2
u
∂z
2
= 0,
1. Introduction. 2
∂u
∂t
– κ
¸
¸
,
_
∂
2
u
∂x
2
+
∂
2
u
∂y
2
+
∂
2
u
∂z
2
= 0,
∂
2
u
∂t
2
– c
2
¸
¸
,
_
∂
2
u
∂x
2
+
∂
2
u
∂y
2
+
∂
2
u
∂z
2
= 0.
2. Linear differential operators. 3
2. Linear differential operators.
All of the above mentioned partial differential equations can be written in the form
L[u] = F
where
L[u] ≡ ∇
2
u =
∂
2
u
∂x
2
+
∂
2
u
∂y
2
+
∂
2
u
∂z
2
in Laplace's equation,
L[u] ≡
∂u
∂t
– κ
¸
¸
,
_
∂
2
u
∂x
2
+
∂
2
u
∂y
2
+
∂
2
u
∂z
2
=
∂u
∂t
–κ ∇
2
u in the heat equation,
and
L[u] ≡
∂
2
u
∂t
2
– c
2
¸
¸
,
_
∂
2
u
∂x
2
+
∂
2
u
∂y
2
+
∂
2
u
∂z
2
=
∂
2
u
∂t
2
– c
2
∇
2
u in the heat equation.
L [u ] is in each case, a linear partial differential operator. Linearity means that for any
two functions u
1
, u
2
, and any two constants c
1
, c
2
,
L[c
1
u
1
+ c
2
u
2
] = c
1
L[u
1
] + c
2
L[u
2
].
In other words, L is linear if it preserves linear combinations of u
1
, u
2
. This definition
generalises to
L[c
1
u
1
+
....
+ c
n
u
n
] = c
1
L[u
1
] +
...
+ c
n
L[u
n
]
for any functions u
1
,..., u
n
and constants c
1
,..., c
n
.
Let u(x
1
,x
2
,...,x
n
) be a function of n variables x = (x
1
,x
2
,...,x
n
). Then the most general linear
partial differential operator is of the form
L[u] =
∑
i=1
n
∑
j=1
n
a
ij
(x)
∂
2
u
∂x
i
∂x
j
+
∑
i=1
n
b
i
(x)
∂u
∂x
i
+ c(x)u
where a
ij
(x), b
i
(x), c(x) are given coefficients.
The highest order partial derivative appearing is the order of the partial differential
operator. Henceforth we will consider only second order partial differential operators of the
form
L[u] =
∑
i=1
n
∑
j=1
n
a
ij
(x)
∂
2
u
∂x
i
∂x
j
+
∑
i=1
n
b
i
(x)
∂u
∂x
i
+ c(x)u.
The general linear second order partial differential equation is of the form
L[u] = F(x)
where F(x) is a given function. When F(x) ≡ 0, the equation L[u] = 0 is called homogeneous.
If F(x) ≠ 0, the equation L[u] = F(x) is called nonhomogeneous.
Linearity of L is essential to the success of the Fourier method. There are usually
(infinitely) many solutions of a linear partial differential equation. The number of
2. Linear differential operators. 4
solutions may be restricted by imposing extra conditions. Often these extra conditions are
given as linear equations involving u and its derivatives on the boundary of some region Ω
⊂ R
n
. These equations are written as boundary conditions
B[u(x)] = φ(x)
where B is a partial differential operator defined on the boundary ∂Ω of the region Ω.
As an example let u(x, t) be the temperature at x ∈ Ω, at time t, of a conducting body. Then
u satisfies the heat equation
∂u
∂t
–κ ∇
2
u = 0, x ∈ Ω, t > 0.
Suppose the temperature at the boundary is maintained at a given temperature φ, then
u(x, t) = φ(x, t), x ∈ ∂Ω, t > 0.
Also the initial temperature of the body is given by
u(x, 0) = f(x), x ∈ Ω.
Suppose u
1
,..., u
k
satisfy the linear partial differential equations
L[u
j
] = F
j
, j = 1, ... , k
and boundary conditions
B[u
j
] = φ
j
, j = 1, ... , k,
then the linear combination u = c
1
u
1
+ .... + c
k
u
k
satisfies the linear partial differential
equation
L[u] = c
1
F
1
+ .... + c
k
F
k
and the boundary condition
B[u] = c
1
φ
1
+ .... + c
k
φ
k
.
This result is commonly referred to as the principle of superposition and it is of paramount
importance for the Fourier method. It shows that by taking linear combinations of
solutions of related linear partial differential equations, other solutions can be constructed
for source and boundary terms F, φ which are linear combinations of simpler terms.
3. Separation of variables. 5
3. Separation of variables.
Example 1. To motivate Fourier series, consider a heat conducting bar of length l,
insulated along its length, so that heat can flow only along the bar. The temperature u(x,t)
along the bar, satisfies the heat equation
u
t
– κu
xx
= 0, 0 < x < l, t > 0
and boundary conditions
u(0, t) = 0, u(l, t) = 0, t > 0.
Let the initial temperature along the bar at t = 0 be given by
u(x, 0) = f(x), 0 < x < l.
Assume a solution of the form
u(x, t) = X(x)T(t).
Such a solution is called a separationofvariables solution. Substituting into the heat
equation,
X(x)T'(t) = κX"(x)T(t).
Dividing by κX(x)T(t), leads to
T'(t)
κT(t)
=
X"(x)
X(x)
.
Since the variables x, t appear on separate sides of this equation, each side of this equation
can only be equal to a constant, say λ. Then
T' = κλT, and X" = λX.
These are constant coefficient ordinary differential equations. λ is real but could be
positive, negative or zero. Assume for the moment that λ > 0. The solutions are
T(t) = Ce
κλt
, X(x) = Ae
x
√
λ
+ Be
–x
√
λ
for A, B, C constants. Then u(x, t ) = X(x)T(t) satisfies u(0, t) = 0, u(l, t) = 0, t > 0, if and
only if X(0) = 0, X(l) = 0. That is
0 = X(0) = A + B
0 = X(l) = Ae
l
√
λ
+ Be
–l
√
λ
.
Eliminating A, B leads to the condition
e
l
√
λ
– e
–l
√
λ
= 0
or
3. Separation of variables. 6
sinh ( ) l
√
λ = 0.
There are no values of λ > 0 which satisfy this condition. So λ cannot be positive.
Suppose that λ < 0 and let λ = – µ
2
for some real µ . Then
√
λ = iµ and hence µ satisfies
e
iµl
– e
–i µl
= 0
or
sin( ) µl = 0.
This has solutions µl = nπ, n = t1, t2,..., and hence µ
n
=
nπ
l
, n = t1, t2,... ,
T
n
(t) = e
–
κn
2
π
2
t
l
2
, X
n
(x) = e
inπx
l
– e
–
inπx
l
= 2i sin
¸
¸
,
_
nπx
l
, n = 1, 2,...
For λ = 0,
T' = 0, and X" = 0
or
T(t) = C, and X(x) = A + Bx.
The boundary conditions u(0, t) = 0, u(l, t) = 0, t > 0, are satisfied if and only if X(0) = 0, X(l)
= 0. That is
0 = X(0) = A , 0 = X(l) = Bl
or
A = B = 0.
Therefore the nontrivial solutions of the heat equation
u
t
– κu
xx
= 0, 0 < x < l, t > 0
satisfying boundary conditions
u(0, t) = 0, u(l, t) = 0, t > 0
are
u
n
(x, t) = e
–
κn
2
π
2
t
l
2
sin
¸
¸
,
_
nπx
l
, n = 1, 2,... .
By superposition
u(x, t) =
∑
n=1
∞
b
n
u
n
(x, t) =
∑
n=1
∞
b
n
e
–
κn
2
π
2
t
l
2
sin
¸
¸
,
_
nπx
l
also satisfies the heat equation and the boundary conditions. For the initial condition to be
satisfied
u(x, 0) = f(x) =
∑
n=1
∞
b
n
sin
¸
¸
,
_
nπx
l
, 0 < x < l.
That is we must show that f(x) can be represented as a series expansion of sines. We are
mainly concerned wuth those functions f(x) which have this property.
A series expansion such as
3. Separation of variables. 7
f(x) =
∑
n=1
∞
b
n
sin
¸
¸
,
_
nπx
l
is called a Fourier series expansion.
Example 2. Suppose the conducting bar is insulated at each end, the temperature u(x,t)
satisfies the same heat equation and initial condition but different boundary conditions
u
x
(0, t) = 0, u
x
(l, t) = 0, t > 0.
Separation of variables u(x, t ) = X(x)T(t) in the heat equation leads to the same ordinary
differential equations
T' = κλT, and X" = λX
and assuming for the moment that λ > 0, solutions
T(t) = Ce
κλt
, X(x) = Ae
x
√
λ
+ Be
–x
√
λ
.
The boundary conditions are satisfied if and only if
0 = X'(0), 0 = X'(l).
That is
0 =
√
λ (A – B), 0 =
√
λ
( )
Ae
l
√
λ
– Be
–l
√
λ
.
Eliminating A, B again leads to the condition, λ ≠ 0,
e
l
√
λ
– e
–l
√
λ
= 0
or
sinh( ) l
√
λ = 0.
There are no values of λ > 0 which satisfy this condition. So λ cannot be positive.
Suppose that λ < 0 and let λ = – µ
2
for some real µ . Then
√
λ = iµ and hence µ satisfies
e
iµl
– e
–i µl
= 0
or
sin( ) µl = 0.
This has solutions µ
n
l = nπ, n = t1, t2,..., and hence µ
n
=
nπ
l
, n = t1, t2,... ,
λ
n
= – µ
n
2
= –
¸
¸
,
_
nπ
l
2
, n = 1, 2,... ,
T
n
(t) = e
–
κn
2
π
2
t
l
2
, X
n
(x) = e
inπx
l
+ e
–
inπx
l
= 2 cos
¸
¸
,
_
nπx
l
, n = 1, 2,...
If λ = 0, T(t) = C, and X(x) = A + Bx. The boundary conditions u(0, t) = 0, u(l, t) = 0, t >
0, if and only if X'(0) = 0, X'(l) = 0. That is
0 = X'(0) = B , 0 = X'(l).
3. Separation of variables. 8
or
X(x) = A
0
, T(t) = C
0
Therefore the nontrivial solutions of the heat equation
u
t
– κu
xx
= 0, 0 < x < l, t > 0
satisfying boundary conditions
u
x
(0, t) = 0, u
x
(l, t) = 0, t > 0
are
u
n
(x,t) = e
–
κn
2
π
2
t
l
2
cos
¸
¸
,
_
nπx
l
, n = 0, 1, 2,... .
By superposition
u(x,t) =
∑
n=0
∞
a
n
u
n
(x,t) =
∑
n=0
∞
a
n
e
–
κn
2
π
2
t
l
2
cos
¸
¸
,
_
nπx
l
also satisfies the heat equation and the boundary conditions. For the initial condition to be
satisfied
u(x, 0) = f(x) =
∑
n=0
∞
a
n
cos
¸
¸
,
_
nπx
l
, 0 < x < l.
This is also called a Fourier series expansion.
4. Fourier Series. 9
4. Fourier Series.
A function f(x), f: R→R is called periodic if f(x + P) = f(x) for all x ∈ R. P > 0 is called the
period of f.
Suppose f(x ) is periodic with period 2π, then an important question is whether f(x ) has a
Fourier series expansion of the form
f(x) =
a
0
2
+
∑
n=1
∞
( )
a
n
cos nx+b
n
sin nx , 0 < x < 2π.
The constant term is taken as
a
0
2
as a matter of convenience.
The formulas
e
ix
= cos x + i sin x, e
–ix
= cos x – i sin x
can be used to write the Fourier series expansion as
f(x) =
∑
– ∞
∞
c
n
e
inx
where the coefficients are
c
n
=
a
n
– ib
n
2
, c
–n
=
a
n
+ib
n
2
, n = 1, 2, ... , c
0
=
a
0
2
.
Assuming for the moment that the 2πperiodic function f has a Fourier series expansion,
the Fourier coefficients c
n
will be determined using the following orthogonality property of
the complex exponentials e
inx
, n = 0, t1, t2, ... ,.
Lemma.
∫
– π
π
e
inx
e
–imx
dx =
¹
¹
'
¹
¹
0; n≠m
2π; n=m
.
Proof. For n ≠ m,
∫
– π
π
e
inx
e
–imx
dx =
∫
– π
π
e
i (n–m)x
dx
=
1
i(n–m)
[ ] e
i (n–m)π
– e
–i (n–m)π
=
2
(n–m)
sin(n – m)π = 0.
If n = m,
∫
– π
π
e
inx
e
–inx
dx =
∫
– π
π
1 dx = 2π.
¯
4. Fourier Series. 10
Since f(x) =
∑
– ∞
∞
c
n
e
inx
,
∫
– π
π
f(x) e
–imx
dx =
∫
– π
π
¸
¸
,
_
∑
– ∞
∞
c
n
e
inx
e
–imx
dx
=
∑
– ∞
∞
c
n ∫
– π
π
e
–i (n–m)x
dx = 2π c
m
.
Hence
c
m
=
1
2π
∫
– π
π
f(x) e
–imx
dx , m = 0, t1, t2, ... ,
which are the Fourier coefficients of f(x).
Since
c
n
=
a
n
– ib
n
2
, c
–n
=
a
n
+ib
n
2
, n = 1, 2, ... , c
0
=
a
0
2
,
the Fourier cosine and sine coefficients are given by
a
0
= 2c
0
=
1
π
∫
– π
π
f(x) dx,
a
n
= c
n
+ c
–n
=
1
2π
∫
– π
π
f(x) ( ) e
–inx
+e
inx
dx =
1
π
∫
– π
π
f(x) cos nx dx
b
n
= i(c
n
– c
–n
) =
i
2π
∫
– π
π
f(x) ( ) e
–inx
– e
inx
dx =
1
π
∫
– π
π
f(x) sin nx dx.
Either of the expansions
∑
– ∞
∞
c
n
e
inx
,
a
0
2
+
∑
n=1
∞
( )
a
n
cos nx+b
n
sin nx ,
is called a Fourier series expansion of f(x). The first, the exponential form and the second
the trigonometric form.
Notice that if f(x) is an even periodic function, i.e. f(–x) = f(x) for all x, then
b
n
=
1
π
∫
– π
π
f(x) sin nx dx =
1
π
∫
– π
0
f(x) sin nx dx +
1
π
∫
0
π
f(x) sin nx dx
=
1
π
∫
π
0
f(–x) sin(– nx) (–dx) +
1
π
∫
0
π
f(x) sin nx dx
=
1
π
∫
0
π
(–f(–x) + f(x)) sin nx dx = 0, n = 1, 2, ... .
4. Fourier Series. 11
Thus the Fourier series expansion of an even function f(x) has only cosine terms
a
0
2
+
∑
n=1
∞
a
n
cos nx
Furthermore, the coefficients a
n
a
n
=
1
π
∫
– π
π
f(x) cos nx dx =
1
π
∫
– π
0
f(x) cos nx dx +
1
π
∫
0
π
f(x)cos nx dx
=
1
π
∫
π
0
f(–x) cos(– nx) (–dx) +
1
π
∫
0
π
f(x) cos nx dx
=
1
π
∫
0
π
(f(–x) + f(x)) sin nx dx
=
2
π
∫
0
π
f(x) cos nx dx, n = 0, 1, 2, ... .
Similarly if f(x) is an odd periodic function, f(–x) = – f(x), then
a
n
=
1
π
∫
– π
π
f(x) cos nx dx =
1
π
∫
– π
0
f(x) cos nx dx +
1
π
∫
0
π
f(x)cos nx dx
=
1
π
∫
π
0
f(–x) cos(– nx) (–dx) +
1
π
∫
0
π
f(x) cos nx dx
=
1
π
∫
0
π
(f(–x) + f(x)) sin nx dx = 0, n = 0, 1, 2, ... .
Thus the Fourier series expansion of an odd function f(x) has only sine terms
∑
n=1
∞
b
n
sin nx,
and the coefficients are
b
n
=
1
π
∫
– π
π
f(x) sin nx dx =
1
π
∫
– π
0
f(x) sin nx dx +
1
π
∫
0
π
f(x) sin nx dx
=
1
π
∫
π
0
f(–x) sin(– nx) (–dx) +
1
π
∫
0
π
f(x) sin nx dx
=
1
π
∫
0
π
(–f(–x) + f(x)) sin nx dx
=
2
π
∫
0
π
f(x) sin nx dx , n = 1, 2, ... .
It is worth noting that for a periodic function f(x), the constant term in the Fourier series
a
0
2
=
1
2π
∫
– π
π
f(x) dx
4. Fourier Series. 12
is the average value of f(x) over a period, – π < x < π.
Example (a). Let f(x) be periodic with period 2π,
f(x) =
¹
¹
'
¹
¹
–1; – π<x<0
1; 0< x<π
.
5 3 1 1 2 3 4 5 6
1
0.5
6.28 6.28
1.5
1.5
This is called a squarewave function. It is obviously an odd function, hence its Fourier
cosine coefficients a
n
are all zero and
b
n
=
2
π
∫
0
π
f(x) sin nx dx =
2
π
∫
0
π
1 sin nx dx
=
2
nπ
[ ] –cosnπ+cos0 =
2
nπ
[ ]
–(–1)
n
+1
=
¹
¹
'
¹
¹
0; neven
4
nπ
; nodd
.
Therefore b
2n
= 0, b
2n–1
=
4
(2n–1)π
, n = 1, 2, ... . The Fourier series of f(x) is
∑
n=1
∞
b
n
sin nx, =
∑
n=1
∞
b
2n–1
sin (2n – 1)x
=
∑
n=1
∞
4
(2n–1)π
sin (2n – 1)x
=
4
π
∑
n=1
∞
1
2n–1
sin (2n – 1)x.
(b). Let f(x) be periodic with period 2π, f(x) = sin x, – π < x < π,
f(x) =
¹
¹
'
¹
¹
–sinx; – π<x<0
sin x; 0< x<π
.
4. Fourier Series. 13
5 3 1 1 2 3 4 5 6
0.25
0.75
1.25
6.28 6.28
0.5
1.5
f(x) is an even periodic function, hence b
n
= 0 for all n.
a
n
=
2
π
∫
0
π
f(x) cos nx dx =
2
π
∫
0
π
sin x cos nx dx
=
2
π
∫
0
π
1
2
[ ]
sin(n+1)x+sin((–n+1)x) dx
=
1
π
∫
0
π
[ ]
sin(n+1)x–sin(n–1) x dx
=
1
π
¸
]
1
1
–
cos(n+1)π
n+1
+
cos(n–1)π
n–1
+
1
n+1
–
1
n–1
=
1
π
¸
]
1
1
n+1
(1–cos( n+1)π)–
1
n–1
(1–cos( n–1) π)
=
¹
¹
'
¹
¹
0; nodd
1
π
¸
]
1
2
n+1
–
2
n–1
;neven
=
¹
¹
'
¹
¹
0; nodd
–
4
(n
2
–1)π
;neven
.
Therefore a
2n
= –
4
((2n)
2
–1)π
, a
2n–1
= 0 and the Fourier series of f(x) is
a
0
2
+
∑
n=1
∞
a
n
cos nx =
2
π
–
4
π
∑
n=1
∞
1
4n
2
–1
cos 2nx.
5. Bessel's Inequality. 14
5. Bessel's inequality.
Theorem. Let f be 2π periodic and Riemann integrable on [–π, π]. Then
∑
n=–∞
∞
c
n

2
≤
1
2π
∫
– π
π
f(x)
2
dx.
Proof. Using z z = z
2
for complex z,
0 ≤
¹
¹
¹
¹
¹
¹
f(x)–
∑
n=–N
N
c
n
e
inx
2
=
¸
¸
,
_
f(x)–
∑
n=–N
N
c
n
e
inx
¸
¸
,
_
f(x)–
∑
m=–N
N
c
m
e
imx
= f(x)
2
–
∑
n=–N
N
c
n
e
inx
f(x) –
∑
m=–N
N
c
m
e
–imx
f(x) +
∑
n=–N
N
∑
m=–N
N
c
n
c
m
e
i (n–m)x
Dividing by 2π and integrating over [–π, π],
0 ≤
1
2π
∫
– π
π
f(x)
2
dx –
∑
n=–N
N
c
n
1
2π
∫
– π
π
e
inx
f(x)dx –
∑
m=–N
N
c
m
1
2π
∫
– π
π
e
–imx
f(x) dx
+
∑
n=–N
N
∑
m=–N
N
c
n
c
m
1
2π
∫
– π
π
e
i (n–m)x
dx
=
1
2π
∫
– π
π
f(x)
2
dx –
∑
n=–N
N
c
n
c
n
–
∑
m=–N
N
c
m
c
m
+
∑
n=–N
N
c
n
c
n
=
1
2π
∫
– π
π
f(x)
2
dx –
∑
n=–N
N
c
n

2
for any N.
Letting N→ ∞ leads to the result.
¯
Bessel's inequality will later be shown to be actually an equality but for now it implies that
the series
∑
n=–∞
∞
c
n

2
converges where c
n
are the Fourier coefficients of the Riemann
integrable function f.
Using the equations
c
n
=
a
n
– ib
n
2
, c
–n
=
a
n
+ib
n
2
, n = 1, 2, ... , c
0
=
a
0
2
,
Bessel's inequality can be written as
5. Bessel's Inequality. 15
∑
n=–∞
∞
c
n

2
=
a
0
2
4
+
∑
n=1
∞
¹
¹
¹
¹
a
n
– ib
n
2
2
+
∑
n=1
∞
¹
¹
¹
¹
a
n
+ib
n
2
2
=
a
0
2
4
+
1
2
∑
n=1
∞
[ ]
a
n

2
+b
n

2
≤
1
2π
∫
– π
π
f(x)
2
dx.
This implies that the series
∑
n=1
∞
a
n

2
,
∑
n=1
∞
b
n

2
also converge, where a
n
, b
n
are the Fourier cosine and
sine coefficients of f.
6. Convergence results for Fourier series. 16
6. Convergence results for Fourier series.
We will consider the question: For what functions f do the Fourier series
∑
– ∞
∞
c
n
e
inx
,
a
0
2
+
∑
n=1
∞
( )
a
n
cos nx+b
n
sin nx ,
converge?
Because we are dealing with functions, the concept of convergence must be made precise.
Do we mean the numerical series converging at every x ∈ [– π, π]? Can convergence be
different at different points x, indeed can we have convergence at some points and not at
others and if so, which are the points of convergence? Can we have some sort of average
convergence on [– π, π]?
A function f is called piecewise continuous on an interval [a, b] if it is continuous
everywhere except at finitely many points x
1
, x
2
, ... , x
k
∈ [a, b] and the lefthand and
righthand limits of f exist at each of the points x
1
, x
2
, ... , x
k
. The set of all piecewise
continuous functions on [a, b] is denoted by PC[a, b].
f is called piecewise smooth if f and its derivative f' are piecewise continuous on [a, b].
Consider the N
th
partial sum of the complex Fourier series of f,
S
N
f
(x) =
∑
n=–N
N
c
n
e
inx
=
∑
n=–N
N
¸
¸
,
_
1
2π
∫
– π
π
e
–iny
f(y)dy e
inx
=
∫
– π
π
¸
¸
,
_
1
2π
∑
n=–N
N
e
in(x–y)
f(y) dy.
We define D
N
(x) =
1
2π
∑
n=–N
N
e
inx
. Then
S
N
f
(x) =
∫
– π
π
D
N
(x – y)f(y) dy =
∫
– π
π
D
N
(y)f(x – y) dy .
[The last equality follows from a change of variable and the periodicity of the integrand
D
N
(x – y)f(y)]. The function D
N
(x) is called the Dirichlet kernel and
D
N
(x ) =
1
2π
e
–iNx
∑
n=0
2N
e
inx
=
1
2π
e
–iNx
¸
¸
,
_
e
i (2N+1) x
–1
e
ix
–1
6. Convergence results for Fourier series. 17
=
1
2π
¸
¸
,
_
e
i (N+1) x
– e
–iNx
e
ix
–1
=
1
2π
¸
¸
,
_
e
i (N+
1
2
)x
– e
–i (N+
1
2
)x
e
ix
2
– e
–
ix
2
=
1
2π
sin
( )
N+
1
2
x
sin( )
x
2
.
The question of convergence of the Fourier series reduces to question:
Does S
N
f
(x) converge as N → ∞ for x ∈[– π, π]? If so, to what does it converge?
We prove a preliminary result.
Lemma.
∫
– π
0
D
N
(x) dx =
∫
0
π
D
N
(x) dx =
1
2
.
Proof. D
N
(x) =
1
2π
+
1
2π
n ≠0
∑
n=–N
N
e
inx
=
1
2π
+
1
π
∑
n=1
N
cos nx
From the evenness of D
N
(x) and integrating,
∫
– π
0
D
N
(x) dx =
∫
0
π
D
N
(x) dx =
∫
0
π
¸
]
1
1
1
2π
+
1
π
∑
n=1
N
cosnx dx
=
¸
]
1
1 x
2π
+
1
π
∑
n=1
N
sin nx
n
π
0
=
1
2
.
¯
Theorem. Let f be a piecewise smooth 2πperiodic function on R. Then
lim
N→∞
S
N
f
(x) =
1
2
[ ]
f(x
–
)+f(x
+
)
for every x. Hence lim
N→∞
S
N
f
(x) = f(x) for each point x of continuity of f.
Proof. S
N
f
(x) –
1
2
[ ]
f(x
–
)+f(x
+
) =
∫
– π
π
D
N
(y)f(x – y) dy –
1
2
[ ]
f(x
–
)+f(x
+
)
=
∫
0
π
D
N
(y)[f(x – y) – (f(x
–
)]dy +
∫
– π
0
D
N
(y)[f(x – y) – f(x
+
)]dy
(from lemma above)
6. Convergence results for Fourier series. 18
=
∫
0
π
D
N
(y)[f(x – y) – f(x
–
)]dy +
∫
π
0
D
N
(–y)[f(x + y) – f(x
+
)](–dy)
=
∫
0
π
D
N
(y)[f(x – y) – f(x
–
) + f(x + y) – f(x
+
)]dy
=
∫
0
π
1
2π
sin
( )
N+
1
2
y
sin( )
y
2
[f(x – y) – f(x
–
) + f(x + y) – f(x
+
)]dy
=
1
π
∫
0
π
sin
( )
N+
1
2
y
¸
¸
,
_
y
2
sin( )
y
2
¸
]
1
f(x–y)–f(x
–
)
y
+
f(x+y)– f(x
+
)
y
dy
For fixed x define
g(y) =
¸
¸
,
_
y
2
sin( )
y
2
¸
]
1
f(x–y)–f(x
–
)
y
+
f(x+y)– f(x
+
)
y
,
which is an odd piecewise continuous function for [–π, π] by the condition of piecewise
smoothness on f.
Then
S
N
f
(x) –
1
2
[ ]
f(x
–
)+f(x
+
) =
1
π
∫
0
π
sin
( )
N+
1
2
y g(y)dy
=
1
π
∫
0
π
sin (Ny){ } cos( )
y
2
g(y) dy +
1
π
∫
0
π
cos (Ny){ } sin( )
y
2
g(y) dy
The last two terms are the Fourier sine and cosine coefficients B
N
, A
N
of
1
2
cos( )
y
2
g(y),
1
2
sin( )
y
2
g(y) respectively. By Bessel's inequality, B
N
, A
N
→ 0 as N →
∞.
Therefore
S
N
f
(x) –
1
2
[ ]
f(x
–
)+f(x
+
) = B
N
+ A
N
→ 0, as N → ∞.
¯
This result, that the partial sums of the Fourier series of a piecewise smooth 2π−periodic
function converges pointwise to the mean of the left and right hand limits at x. If x is a
point of continuity of f, then the partial sums converge pointwise to f(x).
6. Convergence results for Fourier series. 19
Fourier series provide a useful method for summing certain numerical series.
Example. The Fourier series of the continuous periodic function f(x) =sin x, x ∈ [– π, π],
is
2
π
–
4
π
∑
n=1
∞
1
4n
2
–1
cos 2nx
5 3 1 1 2 3 4 5 6
0.25
0.75
1.25
6.28 6.28
0.5
1.5
Because x = 0 is a point of continuity of f, the Fourier series converges at x = 0 to f(0) = 0,
0 =
2
π
–
4
π
∑
n=1
∞
1
4n
2
–1
or
∑
n=1
∞
1
4n
2
–1
=
1
2
.
At x =
π
2
, the Fourier series converges to
f
( )
π
2
= 1 =
2
π
–
4
π
∑
n=1
∞
1
4n
2
–1
cos(nπ)
=
2
π
–
4
π
∑
n=1
∞
(–1)
n
4n
2
–1
∑
n=1
∞
(–1)
n
4n
2
–1
= –
¸
¸
,
_
π–2
4
.
7. Differentiation and Integration of Fourier Series. 20
7. Differentiation and Integration of Fourier Series.
Fourier series can be differentiated termbyterm but the question is does the resulting
series converge and if so, to what does it converge? Similar concerns apply to the series
resulting from the termbyterm integration of a Fourier series.
Theorem. Let a
n
, b
n
, c
n
be the Fourier coefficients of a 2πperiodic piecewise smooth
function f and a'
n
, b'
n
, c'
n
be the Fourier coefficients of f'. Then
a'
n
= nb
n
, b'
n
= –na
n
, c'
n
= inc
n
.
Proof. By integration by parts,
c'
n
=
1
2π
∫
– π
π
f'(x) e
–inx
dx
=
1
2π
f(x)e
–inx
¹
¹
π
–π
–
1
2π
∫
– π
π
f(x) (–ine
–inx
)dx
= in
1
2π
∫
– π
π
f(x) e
–inx
dx = inc
n
.
A similar proof works for a'
n
= nb
n
, b'
n
= –na
n
.
¯
Theorem. Let f be 2πperiodic, piecewise smooth, with piecewise smooth derivative f'.
Then the Fourier series of f' is
∑
– ∞
∞
inc
n
e
inx
=
∑
n=1
∞
( )
nb
n
cos nx– na
n
sin nx
and converge for each x where f'(x) exists. If f' is not continuous at x, then the series above
converge to
1
2
[f'(x
–
) + f'(x
+
)].
Proof. This result follows by combining the previous two theorems.
¯
Integration of Fourier series is not so straightforward since the antiderivative of a
periodic function need not be periodic. For example, f(x) = 1 is periodic but its anti
derivative F(x) = x is not. However, since all but the constant term of a Fourier series has a
periodic antiderivative, the following result is true.
Theorem. Let f be 2πperiodic, piecewise continuous with Fourier coefficients a
n
, b
n
, c
n
and let F(x) =
∫
0
x
f(t) dt. If c
0
=
1
2
a
0
= 0, then for all x,
F(x) = C
0
+
n ≠0
∑
n=– ∞
∞
c
n
in
e
inx
=
A
0
2
+
∑
n=1
∞
¸
¸
,
_
¸
¸
,
_
–b
n
n
cos nx+
¸
¸
,
_
a
n
n
sin nx
7. Differentiation and Integration of Fourier Series. 21
where C
0
=
A
0
2
=
1
2π
∫
–π
π
F(x) dx the average value of F on [–π, π].
If c
0
≠ 0, then the series above converges to F(x) – c
0
x.
Proof. Since f(x) is piecewise continuous, F(x) =
∫
0
x
f(t) dt is continuous.
If c
0
= 0, F(x) is 2πperiodic since
F(x + 2π) – F(x) =
∫
0
x+2π
f(t) dt –
∫
0
x
f(t) dt =
∫
x
x+2π
f(t) dt =
∫
–π
π
f(t) dt = 2πc
0
= 0.
Therefore the Fourier series of F converges pointwise at each x ∈ [–π, π] to F(x). Since f(x) =
F'(x), by the two previous theorems,
a
n
= nB
n
, b
n
= –nA
n
, c
n
= inC
n
,
where a
n
, b
n
, c
n
are the Fourier coefficients of f and A
n
, B
n
, C
n
are the Fourier coefficients
of F. If n ≠ 0, this implies that
A
n
= –
b
n
n
, B
n
=
a
n
n
, C
n
=
c
n
in
.
The constant C
0
=
A
0
2
=
1
2π
∫
–π
π
F(x) dx is the constant term in the Fourier series of
F(x) =
∫
0
x
f(t) dt.
If c
0
≠ 0, f(x) – c
0
has zero mean value on [–π, π] and therefore its zeroth Fourier coefficient
is
1
2π
∫
–π
π
(f(x) – c
0
)dx =
1
2π
∫
–π
π
f(x) dx – c
0
= c
0
– c
0
= 0.
Applying the result just obtained to f(x) – c
0
and its antiderivative F(x) – c
0
x completes
the theorem.
¯
Integrating and differentiating known Fourier series using the above results is a useful
way of obtaining new Fourier series.
Example. The 2πperiodic function f :R→R, f(x) = x(π – x), x ∈ [–π, π], is continuous with
piecewise smooth derivative. Therefore its Fourier series converges at each x to f(x). The
Fourier series is by the table #17
f(x) = x(π – x) =
8
π
∑
n=1
∞
1
(2n–1)
3
sin (2n – 1)x.
By the above results,
f'(x) = π – 2x =
8
π
∑
n=1
∞
1
(2n–1)
2
cos (2n – 1)x
7. Differentiation and Integration of Fourier Series. 22
or
x =
π
2
–
4
π
∑
n=1
∞
1
(2n–1)
2
cos (2n – 1)x
which agrees with table #2.
Example. f(x) =
¹
¹
'
¹
¹
0; – π<x<0
1; 0<x<π
5 3 1 1 2 3 4 5 6
0.25
0.75
1.25
6.28319 6.28319
0.5
1.5
is piecewise continuous with Fourier series
f(x) =
1
2
+
2
π
∑
n=1
∞
1
2n–1
sin (2n – 1)x
(table #7). Since c
0
=
1
2
≠ 0, F(x) =
∫
0
x
f(t) dt. =
¹
¹
'
¹
¹
0; – π<x<0
x; 0<x<π
.
Therefore
F(x) –
x
2
=
A
0
2
–
2
π
∑
n=1
∞
1
(2n–1)
2
cos (2n – 1)x
Since
A
0
2
=
1
2π
∫
–π
π
F(x) dx =
π
4
, F(x) –
x
2
=
π
4
–
2
π
∑
n=1
∞
1
(2n–1)
2
cos (2n – 1)x.
But F(x) –
x
2
=
¹
¹
'
¹
¹
–
x
2
; – π<x<0
x
2
; 0<x<π
=
x
2
Therefore
x
2
=
π
4
–
2
π
∑
n=1
∞
1
(2n–1)
2
cos (2n – 1)x
or
x =
π
2
–
4
π
∑
n=1
∞
1
(2n–1)
2
cos (2n – 1)x.
This result agrees with the previous example
8. Halfrange Fourier series. 23
8. Halfrange Fourier series.
It is often convenient to represent a given function as a Fourier series which contains only
cosine terms or only sine terms, as in the initial examples.
If f(x) is piecewise continuous on [0, π], it can be extended to [–π, 0] as either an even
function or as an odd function.
Then f is periodically extended to the whole real line as a 2πperiodic function by
f(x + 2π) = f(x), x ∈ R.
Let f(x) be a piecewise smooth function given on the interval [0, π] and extended to [–π, 0]
as an even function. That is, f(x) = f(–x), x ∈ [–π, 0]. Then the periodic extension to R, is
piecewise smooth and has Fourier series
a
0
2
+
∑
n=1
∞
a
n
cos nx
where
a
n
=
2
π
∫
0
π
f(x) cos nx dx.
Similarly, extending f(x) to [–π, 0] as an odd function and periodically to R, the extension
has Fourier series
∑
n=1
∞
b
n
sin nx
where
b
n
=
2
π
∫
0
π
f(x) sin nx dx.
The Fourier series
a
0
2
+
∑
n=1
∞
a
n
cos nx,
∑
n=1
∞
b
n
sin nx are known as halfrange series for
f(x), x ∈ [0, π].
Example. Let f(x) = sin x, x ∈ [0, π]. Extend f to [–π, 0] as an even function.
That is f(x) =
¹
¹
'
¹
¹
sin(–x)=–sin x; – π<x<0
sin x; 0<x<π
= sin x
8. Halfrange Fourier series. 24
2 1 1 2 3
0.25
0.75
1.25
3.14159 3.14159
0.5
1.5
The halfrange Fourier series of f is given by
f(x) =
2
π
–
4
π
∑
n=1
∞
1
2n
2
–1
cos 2nx (table #8).
9. General Intervals. 25
9. General Intervals.
The theory of Fourier series can be expanded to include functions which have arbitrary
period. Let f :R→R have period 2l > 0. That is f(x + 2l) = f(x), x ∈ R. Also define φ : R→R by
φ(x) = f
¸
¸
,
_
lx
π
. Then
φ(x + 2π) = f
¸
¸
,
_
l(x+2π)
π
= f
¸
¸
,
_
lx
π
+2l = f
¸
¸
,
_
lx
π
= φ(x)
So φ has period 2π. Applying the earlier results to φ, results in the Fourier series
φ(x) =
∑
– ∞
∞
c
n
e
inx
=
a
0
2
+
∑
n=1
∞
( )
a
n
cos nx+b
n
sin nx ,
where
a
n
=
1
π
∫
– π
π
φ(x) cos nx dx, b
n
=
1
π
∫
– π
π
φ(x) sin nx dx, c
n
=
1
2π
∫
– π
π
φ(x) e
–inx
dx.
The change of variable y =
lx
π
leads to
a
n
=
1
l
∫
– l
l
φ
¸
¸
,
_
πy
l
cos
¸
¸
,
_
nπy
l
dy =
1
l
∫
– l
l
f(x) cos
¸
¸
,
_
nπy
l
dy,
b
n
=
1
l
∫
– l
l
φ
¸
¸
,
_
πy
l
sin
¸
¸
,
_
nπy
l
dy =
1
l
∫
– l
l
f(y) sin
¸
¸
,
_
nπy
l
dy,
c
n
=
1
2l
∫
– l
l
φ
¸
¸
,
_
πy
l
e
–
inπy
l
dy =
1
2l
∫
– l
l
f(y) e
–
inπy
l
dy.
Therefore
f(x) = φ
¸
¸
,
_
πx
l
=
∑
– ∞
∞
c
n
e
inπx
l
=
a
0
2
+
∑
n=1
∞
¸
¸
,
_
a
n
cos
¸
¸
,
_
nπx
l
+b
n
sin
¸
¸
,
_
nπx
l
is the Fourier series of a 2lperiodic function f.
It follows from the above calculations that if f(x) is an even function on [–l, l], its Fourier
series contains no sine terms and is of the form
a
0
2
+
∑
n=1
∞
a
n
cos
¸
¸
,
_
nπx
l
,
with
a
n
=
2
l
∫
0
l
f(ξ) cos
¸
¸
,
_
nπξ
l
dξ
9. General Intervals. 26
Likewise if f(x) is an odd function on [–l, l], its Fourier series contains no cosine terms and
is of the form
∑
n=1
∞
b
n
sin
¸
¸
,
_
nπx
l
,
with
b
n
=
2
l
∫
0
l
f(ξ) sin
¸
¸
,
_
nπξ
l
dξ.
A function f(x) defined on an interval [0, l] can be extended to [–l, l] either as an even or an
odd function and the extended periodically with period 2l to the real line. The Fourier
series of such functions are halfrange expansions on [0, l] and contain no sine terms in the
case of an even extension or no cosine terms for an odd extension.
10. Application to Laplace's equation. 27
10. Application to Laplace's equation.
Example 1. Let Ω be a rectangle 0 ≤ x ≤ a, 0 ≤ y ≤ b, and consider the boundaryvalue
problem
∂
2
u
∂x
2
+
∂
2
u
∂y
2
= 0, 0 ≤ x ≤ a, 0 ≤ y ≤ b
u(x, 0) = 0, u(x, b) = f(x), 0 ≤ x ≤ a
u(0, y) = u(a, y) = 0, 0 ≤ y ≤ b.
u = 0
u = f
u = 0
u = 0
x = a
y = b
Assuming a separation of variables solution of the form u(x, y) = X(x)Y(y) and substituting
into Laplaces equation,
X
"
Y + X Y
"
= 0
or
–
X
"
X
=
Y
"
Y
= constant = λ
For λ > 0, X
"
+ λX = 0, Y
"
– λY = 0, hence
X(x) = A cos
√
λx + B sin
√
λx , Y (y) = C cosh
√
λy + D sinh
√
λy .
To satisfy the boundary conditions u(0, y) = u(a, y) = 0, 0 ≤ y ≤ b, requires that
X(0) = A = 0, X(a) = B sin( ) a
√
λ = 0.
Therefore a
√
λ = nπ, λ =
¸
¸
,
_
nπ
a
2
, n = 1, 2, ... , and X(x) = B sin
¸
¸
,
_
nπx
a
.
To also satisfy the boundary condition u(x, 0) = 0, 0 ≤ x ≤ a, requires that
10. Application to Laplace's equation. 28
Y(0) = C = 0 and therefore Y (y) = D sinh
¸
¸
,
_
nπy
a
. The separation of variables solutions for
λ > 0 are
u(x, y) = X(x)Y(y) = BD sin
¸
¸
,
_
nπx
a
sinh
¸
¸
,
_
nπy
a
, n = 1, 2, ... ,
for an arbitrary constant BD.
For λ < 0, let λ = – µ, µ > 0. Then X
"
– µX = 0, Y
"
+ µY = 0, hence
X(x) = A cosh
√
µx + B sinh
√
µx , Y (y) = C cos
√
µy + D sin
√
µy .
To satisfy the boundary conditions u(0, y) = u(a, y) = 0, 0 ≤ y ≤ b, requires that
X(0) = A = 0, X(a) = B sinh( ) a
√
λ = 0.
But B = 0 gives a trivial solution and sinh( ) a
√
λ ≠ 0 for λ < 0. So there is only the trivial
separation of variables solution u = 0, for λ < 0.
Finally suppose that λ = 0, then X
"
= 0, Y
"
= 0, hence
X(x) = A + Bx, Y(y) = C + Dy.
To satisfy the boundary conditions u(0, y) = u(a, y) = 0, 0 ≤ y ≤ b, requires that
X(0) = A = 0, X(a) = Ba = 0.
That is, A = B = 0 and the only separation of variables solution for λ = 0 is the trivial
solution.
Summarising thus far, there is a sequence of nontrivial solutions of the form
u
n
(x, y) = B
n
sin
¸
¸
,
_
nπx
a
sinh
¸
¸
,
_
nπy
a
, n = 1, 2, ... ,
where B
n
are the constants BD for each n = 1, 2, ... . By superposition, a solution is also
given by
u(x, y) =
∑
n =1
∞
B
n
u
n
(x, y)
=
∑
n =1
∞
B
n
sin
¸
¸
,
_
nπx
a
sinh
¸
¸
,
_
nπy
a
.
The boundary condition u(x, b) =f(x), 0 ≤ x ≤ a, implies that
10. Application to Laplace's equation. 29
u(x, b) = f(x) =
∑
n =1
∞
B
n
sin
¸
¸
,
_
nπx
a
sinh
¸
¸
,
_
nπb
a
.
Since φ(x), 0 ≤ x ≤ a, has a Fourier halfrange series f(x) =
∑
n =1
∞
b
n
sin
¸
¸
,
_
nπx
a
, where
b
n
=
2
a
∫
0
a
f(ξ) sin
¸
¸
,
_
nπξ
a
dξ = B
n
sinh
¸
¸
,
_
nπb
a
.
Therefore
B
n
=
b
n
sinh
¸
¸
,
_
nπb
a
=
2
asinh
¸
¸
,
_
nπb
a
∫
0
a
f(ξ) sin
¸
¸
,
_
nπξ
a
dξ
u(x, y) =
∑
n =1
∞
B
n
sin
¸
¸
,
_
nπx
a
sinh
¸
¸
,
_
nπy
a
.
Suppose all four sides of the rectangle have nonzero boundary conditions, then we can
break up the problem into four subproblems each similar to the one we have just solved.
Each of these subproblems will have zero boundary conditions on three of the four sides
and can be solved as above. Then the solution to the boundaryvalue problem is the sum of
the four subproblems by superposition.
Example 2. We seek a solution to Laplaces' equation on a circle, satisfying Dirichlet
boundary conditions. It is natural to choose Ω a circle centre the origin, radius a, and
Laplaces' equation in polar coordinates,
∆u(r, θ) =
∂
2
u
∂r
2
+
1
r
∂u
∂r
+
1
r
2
∂
2
u
∂θ
2
= 0, r < a, 0 ≤ θ < 2π,
subject to boundary conditions
u(a, θ) = f(θ), 0 ≤ θ < 2π,
where φ is a continuous function on 0 ≤ θ < 2π.
u =
a
10. Application to Laplace's equation. 30
Using separation of variables, assume that
u(r, θ) = R(r)Θ(θ)
and substitute into the differential equation to obtain
R
"
Θ +
1
r
R
'
Θ +
1
r
2
RΘ
"
= 0,
or separating variables,
r
2
R
"
+rR
'
R
= –
Θ
"
Θ
= constant = λ
2
(say),
which leads to the two families of ordinary differential equations
r
2
R
"
+ rR
'
– λ
2
R = 0, Θ
"
+ λ
2
Θ = 0.
Consider the case λ = 0; then the solutions of the above equations are
Θ = A
0
+ B
0
θ,
R = C
0
+ D
0
log r
Since solutions u must be periodic in θ, and bounded for r ≤ a, we conclude that B
0
= 0, D
0
= 0.
In the case λ =/ 0 the solutions of the above equations are
Θ = A
λ
cos λθ + B
λ
sin λθ
R = C
λ
r
λ
+ D
λ
r
–λ
where the coefficients depend on λ. Again because solutions are periodic in θ with period
2π, we conclude that λ = n = 1,2,3,.., and because solutions are bounded for r ≤ a, D
n
= 0, n
= 1,2,3,... By superposition we can combine these solutions to obtain
u(r,θ) =
∑
n = 0
∞
C
n
r
n
( )
A
n
cos nθ+B
n
sin nθ
We may as well incorporate the constant C
n
into A
n
and B
n
, hence
u(r, θ) =
∑
n = 0
∞
r
n
( )
A
n
cos nθ+B
n
sin nθ
At the boundary r = a,
f(θ) = u(a, θ) =
∑
n = 0
∞
a
n
( )
A
n
cos nθ+B
n
sin nθ
10. Application to Laplace's equation. 31
=
a
0
2
+
∑
n = 1
∞
( )
a
n
cos nθ+b
n
sin nθ
where
a
n
=
1
π
∫
0
2π
f(t)cosntdt , b
n
=
1
π
∫
0
2π
f(t)sinntdt
are the Fourier coefficients of the function f. Equating coefficients of cos nθ, sin nθ, in
these two expressions for f(θ), we obtain that
A
n
=
a
n
a
n
, B
n
=
b
n
a
n
and hence
u(r, θ) =
a
0
2
+
∑
n = 1
∞
¸
¸
,
_
r
a
n
( )
a
n
cos nθ+b
n
sin nθ
This series converges uniformly for 0 ≤ θ < 2π, r < a.
We can get a closed form expression for u by substituting the formulae for the Fourier
coefficients as follows.
u(r, θ) =
1
2π
∫
0
2π
f(t)dt +
1
π
∑
n = 1
∞
¸
¸
,
_
r
a
n
∫
0
2π
f(t)( ) cos nt cosnθ+sinnt sinnθ dt
=
1
2π
∫
0
2π
f(t)dt +
1
π
∑
n = 1
∞
¸
¸
,
_
r
a
n
∫
0
2π
f(t)cos n(t–θ)dt
=
1
2π
∫
0
2π
f(t)dt +
1
π
∫
0
2π
f(t)
∑
n = 1
∞
¸
¸
,
_
r
a
n
cos n(t– θ) dt
where the interchange of the order of summation and integration is allowed by the uniform
convergence of the series.
Therefore u(r, θ) =
1
2π
∫
0
2π
f(t)
¸
]
1
1
1+2
∑
n = 1
∞
¸
¸
,
_
r
a
n
cos n(t–θ) dt.
Using the exponential form, 2 cos x = e
ix
+ e
–i x
the term in the square brackets reduces to
two infinite geometric series whose sum is
1 + 2
∑
n = 1
∞
¸
¸
,
_
r
a
n
cos n(t – θ) =
a
2
–r
2
a
2
–2arcos(t–θ)+ r
2
(exercise). Therefore
u(r, θ) =
1
2π
∫
0
2π
f(t)
a
2
–r
2
a
2
–2ar cos(t–θ)+r
2
dt,
10. Application to Laplace's equation. 32
which is Poissons' integral formula for the circle.
11. SturmLiouville problems and orthogonal functions. 33
11. SturmLiouville problems and orthogonal functions.
The functions sin nx , cos nx, e
inx
, e
–inx
, n = 0, 1, 2, ... , are examples of orthogonal
functions on [–π, π] Their orthogonality properties follow from the fact that they solutions
of linear second order ordinary differential equations.
For example sin nx , cos nx, e
inx
satisfy
– u"(x) = λu(x),
u(–π) = u(π), u'(–π) = u'(π)
for λ = n
2
. The ordinary differential equation together with the boundary conditions is
called a boundaryvalue problem.
The set {e
inx
, n = 0, t1, t2, ... } form a basis for the vector space L
2
[–π, π] consisting of
functions f :[–π, π] →R for which
∫
– π
π
f(x)
2
dx < ∞. A set of functions {φ
n
(x); n = 1, 2, ... } is
said to be a basis for L
2
[–π, π] if for any function f ∈ L
2
[–π, π] , there is a unique set of
scalars c
n
, n = 1, 2, ... such that
lim
N→∞
∫
– π
π
¹
¹
¹
¹
¹
¹
f(x)–
∑
n=1
N
c
n
φ
n
(x)
2
dx = 0.
Then we say that f has the expansion f =
∑
n=1
∞
c
n
φ
n
.
Furthermore, if the set {φ
n
(x); n = 1, 2,... } is orthonormal, that is
∫
– π
π
φ
n
(x) φ
m
(x) dx =
¹
¹
'
¹
¹
0; n≠m
1; n=m
then the coefficients c
n
are given by
c
n
=
∫
– π
π
f(x) φ
n
(x) dx, n = 1, 2, ... .
The interval [–π, π] can be replaced by [a, b] and the theory of Fourier series generalised
to the expansion of arbitrary functions f ∈ L
2
[a, b] in terms of an orthonormal basis {φ
n
(x);
n = 1, 2,... } consisting of functions which are solutions of a boundaryvalue problem for
certain second order linear ordinary differential equations.
Let p(x), q(x), w(x) be real continuous functions on the interval [a, b]. Let p(x) be
continuous and p(x) > 0, q(x) ≥ 0, w(x) ≥ 0 on [a, b]. Let α
0
, α
1
, β
0
, β
1
, be real constants.
A SturmLiouville problem is a boundary value problem
–
( )
p(x)u' ' + q(x)u = λw(x)u, x ∈ [a, b]
11. SturmLiouville problems and orthogonal functions. 34
α
0
u(a) + α
1
u'(a) = 0,
β
0
u(b) + β
1
u'(b) = 0.
The value of λ for which this boundaryvalue problem has nontrivial solutions are called
eigenvalues of the boundaryvalue problem and the corresponding solutions u(x) are called
eigenfunctions. It can be shown that the eigenvalues form a countable set {λ
n
; n = 1, 2,... }
and the corresponding eigenfunctions {φ
n
(x); n = 1, 2,... } are orthonormal with respect to
the weighted innerproduct
< >
f; g
w
=
∫
a
b
w(x) f(x) g(x) dx.
Let the vector space L
2
w
[a, b] consist of functions f :[a, b] →R for which
f
2
=
∫
a
b
w(x)f(x)
2
dx < ∞. A set of functions {φ
n
(x); n = 1, 2, ... } is said to be a basis for
L
2
w
[a, b] if for any function f ∈ L
2
w
[a, b] , there is a unique set of scalars c
n
, n = 1, 2, ... such
that
lim
N→∞
∫
a
b
w(x)
¹
¹
¹
¹
¹
¹
f(x)–
∑
n=1
N
c
n
φ
n
(x)
2
dx = 0.
Let the linear SturmLiouville differential operator L be defined as
L[f] Ê=
1
w(x)
[ ]
–
( )
p(x)f' '+q(x)f .
Let f(x), g(x) be C
2
functions on [a, b]. Then
< >
L[f]; g
w
=
∫
a
b
[ ] ( )
p(x)f' '+q(x)f g(x) dx
= –p(x)f' g
¹
¹
b
a
+
∫
a
b
( )
p(x)f' g' + q(x)f g dx
Also
< >
f; L[g]
w
=
∫
a
b
f(x)
[ ] ( )
p(x)g' '+q(x)g dx
= –p(x)f g'
¹
¹
b
a
+
∫
a
b
( )
p(x)f' g' + q(x)f g dx.
Subtracting these two expressions,
< >
L[f]; g
w
–
< >
f; L[g]
w
= –p(x)
( )
f' g – f g'
¹
¹
b
a
.
This is Lagrange's identity.
11. SturmLiouville problems and orthogonal functions. 35
Lemma. If f, g are C
2
functions on [a, b] which satisfy
α
0
f(a) + α
1
f'(a) = 0, α
0
g(a) + α
1
g'(a) = 0,
β
0
f(b) + β
1
f'(b) = 0, β
0
g(b) + β
1
g'(b) = 0,
then
< >
L[f]; g
w
=
< >
f; L[g]
w
.
Proof. Suppose α
0
≠ 0, β
0
≠ 0, then by Lagrange's identity,
< >
L[f]; g
w
–
< >
f; L[g]
w
= –p(x)
( )
f' g – f g'
¹
¹
b
a
= –p(b)
¸
]
1
1
f'(b)
¸
¸
,
_
–
β
1
β
0
g'(b) –
¸
¸
,
_
–
β
1
β
0
f' (b) g'(b)
+ p(a)
¸
]
1
1
f'(a)
¸
¸
,
_
–
α
1
α
0
g'(a) –
¸
¸
,
_
–
α
1
α
0
f' (a) g'(a)
= 0.
Similarly if α
1
≠ 0, β
1
≠ 0 or α
1
≠ 0, β
0
≠ 0 or α
0
≠ 0, β
1
≠ 0, the result follows with minor
changes to the argument.
¯
Lemma. The eigenvalues of the SturmLiouville operator L are real.
Proof. Let λ be an eigenvalue with corresponding eigenfunction φ. Then φ satisfies
L[φ] =
1
w(x)
[ ] –( ) p(x)φ' '+q(x)φ = λφ,
α
0
φ(a) + α
1
φ'(a) = 0,
β
0
φ(b) + β
1
φ'(b) = 0.
Then
λ φ
w
2
= λ
< >
φ;φ
w
=
< >
λφ; φ
w
=
< >
L[φ];φ
w
=
< >
φ;L[φ]
w
(above lemma)
=
< >
φ;λφ
w
= λ
< >
φ;φ
w
= λ φ
w
2
.
Therefore
(λ – λ )φ
w
2
= 0.
Since φ ≠ 0, λ = λ and λ is real.
¯
11. SturmLiouville problems and orthogonal functions. 36
Lemma. Let φ(x), ψ(x) be eigenfunctions of the SturmLiouville boundaryvalue problem
with corresponding eigenvalues λ, µ respectively. Then If λ ≠ µ, φ(x), ψ(x) are orthogonal on
[a, b] with respect to the weighted inner product
< >
f; g
w
.
Proof.
¹
¹
'
¹
¹
L[φ]=λφ
α
0
φ(a)+α
1
φ'(a)=0
β
0
φ(b)+β
1
φ'(b)=0 ¹
¹
'
¹
¹
L[ψ]=λψ
α
0
ψ(a)+α
1
ψ'(a)=0
β
0
ψ(b)+β
1
ψ'(b)=0
From a previous lemma,
< >
L[φ];ψ
w
=
< >
φ;L[ψ]
w
,
< >
λφ; ψ
w
=
< >
φ;µψ
w
,
(λ – µ)
< >
φ;ψ
w
= 0.
Since λ ≠ µ,
< >
φ;ψ
w
= 0.
¯
These results can be used to prove the following result:
Theorem. The eigenfunctions of a regular SturmLiouville problem are a countable set
{φ
n
(x); n = 1, 2, ... } and form an orthonormal basis for L
2
w
[a, b]. That is, each function f ∈
L
2
w
[a, b] has a series expansion f =
∑
n=1
∞
c
n
φ
n
where the c
n
=
< >
f; φ
n w
, n = 1, 2, ... , and
convergence is in the sense that
lim
N→∞
∫
a
b
w(x)
¹
¹
¹
¹
¹
¹
f(x)–
∑
n=1
N
c
n
φ
n
(x)
2
dx = 0,
or
lim
N→∞
f –
∑
n=1
N
c
n
φ
n

w
2
= 0.
If f(x) ∈ C
2
[a, b] and satisfies the boundary conditions α
0
f(a) + α
1
f'(a) = 0,
β
0
f(b) + β
1
f'(b) = 0, the series f(x) =
∑
n=1
∞
c
n
φ
n
(x) converges uniformly on [a, b]. That is
lim
N→∞
max
x ∈[a; b]
¹
¹
¹
¹
¹
¹
f(x)–
∑
n=1
N
c
n
φ
n
(x) = 0.
12. Bessel Functions. 37
12. Bessel Functions.
Consider the heat equation on a disc 0 ≤ r < a, 0 ≤ θ < 2π,,
∂u
∂t
– κ∆u(r, θ) =
∂u
∂t
– κ
¸
¸
,
_
∂
2
u
∂r
2
+
1
r
∂u
∂r
+
1
r
2
∂
2
u
∂θ
2
= 0,
0 ≤ r < a, 0 ≤ θ < 2π, t > 0, where u(r, θ, t) is the temperature, κ > 0 a constant.
Let the initial temperature at t = 0, be
u(r, θ, 0) = f(r, θ),
0 ≤ r < a, 0 ≤ θ < 2π, and boundary condition at r = a,
u(a, θ, t) = 0,
0 ≤ θ < 2π, t > 0.
Assuming a solution of the form u(r, θ, t) = R(r) Θ(θ) T(t), substituting in the heat equation,
R Θ T' – κ
¸
¸
,
_
R"ΘT+
1
r
R' ΘT+
1
r
2
RΘ"T = 0,
T'
κT
=
R"
R
+
R'
rR
+
Θ"
r
2
Θ
= constant = – λ
2
(say),
T' + λ
2
κT = 0,
r
2
R"
R
+
rR'
R
+ λ
2
r
2
= –
Θ"
Θ
= constant = ν
2
(say),
Θ" + ν
2
Θ = 0, r
2
R" + rR' + (λ
2
r
2
– ν
2
)R = 0.
The equations for T, Θ are familiar but the equation for R is Bessel's equation and has
nonconstant coefficients. We can write it in the SturmLiouville form,
–
( )
rR' ' +
ν
2
r
R = λ
2
rR,
where p(r) = r ≥ 0, q(r) =
ν
2
r
> 0, w(r) = r ≥ 0.
A change of variable ρ = λr in Bessel's equation results in
ρ
2
d
2
S
dρ
2
+ ρ
dS
dρ
+ (ρ
2
– ν
2
)S = 0,
where S(ρ) = R(r) = R
¸
¸
,
_
ρ
λ
. We denote a solution of this Bessel's equation as S(ρ) = J
ν
(ρ)
and then a solution of the original form of Bessel's equation is R(r) = S(λr) = J
ν
(λr). A
second solution of Bessel's equation is Y
ν
(λr), a Bessel function of the second kind, and
therefore the general solution is
R(r) = AJ
ν
(λr) + BY
ν
(λr) .
12. Bessel Functions. 38
If ν ≠ 0, 1, 2, ..., then J
–ν
(λr) is also a solution and
{ }
J
ν
(λr);J
–ν
(λr) is a linearly
independent set of solutions to Bessel's equation. The general solution is then
R(r) = AJ
ν
(λr) + BJ
–ν
(λr)
for constants A, B. The solution J
ν
(λr) is called a Bessel's function of the first kind and a
series representation can be found
J
ν
(λr) = (λr)
ν
¸
]
1
1
1–
λ
2
r
2
2(2+2ν)
+
λ
4
r
4
2.4.(2+2ν)(4+2ν)
+
....
.
For ν > 0, J
ν
(0) = 0, J
0
(0) = 1. Y
ν
(λr), J
–ν
(λr) are unbounded as r → ∞, hence the only
bounded solutions R(r) occur when B = 0.
The Bessel functions J
ν
(λr) are oscillatory for r > 0. For ν > 0, let the zeros of J
ν
(ρ) be
{ρ
νm
; m = 0, 1, ... }, where ρ
ν0
= 0 for all ν > 0.
Then R(r) = J
ν
(λr) solves the SturmLiouville problem
– (rR'(r))' +
¸
¸
,
_
ν
2
r
R(r) = λ
2
rR(r),
R(0) = J
ν
(0) = 0, R(a) = J
ν
(λa) = 0,
if λa = ρ
νm
, m = 1, 2, ... , for each ν > 0.
The Bessel functions
¹
'
¹
¹
)
¹
J
ν
¸
¸
,
_
ρ
νm
r
a
∞
m=1
, are orthogonal on the interval (0, a), with respect to
the weighted inner product
< >
f; g
w
=
∫
0
a
r f(r) g(r) dr. That is,
∫
0
a
r J
ν
¸
¸
,
_
ρ
νm
r
a
J
ν
¸
¸
,
_
ρ
νl
r
a
dr = 0, if m ≠ l.
Returning to the heat equation on the disc 0 ≤ r < a, 0 ≤ θ < 2π, the solutions of the form
u(r, θ, t) = R(r) Θ(θ) T(t) are given by
R(r) = AJ
ν
(λr) + BY
ν
(λr),
Θ(θ) = C cos (νθ) + D sin (νθ),
T(t) = E e
–λ
2
κt
.
For Θ(θ) to have period 2π, ν = n = 0, 1, 2, ... .
For R(r) to be bounded at r = 0, B = 0.
12. Bessel Functions. 39
For u(a, θ, t) = 0, R(a) = J
n
(λa) = 0 or λa = ρ
nm
, n = 0, 1, 2, ... , m = 1, 2, ... , the zeros of
J
n
(ρ). That is λ
nm
=
λ
nm
a
, n = 0, 1, 2, ... , m = 1, 2, ... .
By superposition, a solution of the heat equation on the disc satisfying the boundary
condition u(a, θ, t) = 0, is given by
u(r, θ, t) =
∑
n=o
∞
∑
m=1
∞
exp
¸
¸
,
_
–
ρ
nm
2
κt
a
2
J
n
¸
¸
,
_
ρ
nm
r
a
( )
C
nm
cos(nθ)+D
nm
sin(nθ) .
The orthogonality relations
∫
0
a
r J
n
¸
¸
,
_
ρ
nm
r
a
J
n
¸
¸
,
_
ρ
nl
r
a
dr = 0, if m ≠ l, and
∫
π
π
cos(nθ) sin (kθ) dθ, if n ≠ k,
imply the orthogonality of
¹
'
¹
¹
)
¹
J
n
¸
¸
,
_
ρ
nm
r
a
cos(nθ);J
n
¸
¸
,
_
ρ
nm
r
a
sin(nθ)
∞
;
∞
n=0; m=1
on the rectangle
(r, θ) ∈ (0, a)×(–π, π) with respect to the inner product
< >
f; g =
∫
0
a
∫
–π
π
r f(r, θ) g(r, θ) dθ dr.
That is,
∫
0
a
∫
–π
π
r
¸
¸
,
_
J
n
¸
¸
,
_
ρ
nm
r
a
cos(nθ)
¸
¸
,
_
J
k
¸
¸
,
_
ρ
kl
r
a
cos(kθ) dθ dr
=
¹
¹
'
¹
¹
0; n≠ k
0; n=kandm≠l
π
∫
0
a
rJ
k
¸
¸
,
_
ρ
kl
r
a
2
dr;n=kandm=l
∫
0
a
∫
–π
π
r
¸
¸
,
_
J
n
¸
¸
,
_
ρ
nm
r
a
sin(nθ)
¸
¸
,
_
J
k
¸
¸
,
_
ρ
kl
r
a
sin(kθ) dθ dr
=
¹
¹
'
¹
¹
0; n≠ k
0; n=kandm≠l
π
∫
0
a
rJ
k
¸
¸
,
_
ρ
kl
r
a
2
dr;n=kandm=l
∫
0
a
∫
–π
π
r
¸
¸
,
_
J
n
¸
¸
,
_
ρ
nm
r
a
cos(nθ)
¸
¸
,
_
J
k
¸
¸
,
_
ρ
kl
r
a
sin(kθ) dθ dr = 0 for all n, m, k and l.
The coefficients C
nm
, D
nm
are determined using these orthogonality properties. For
example,
12. Bessel Functions. 40
∫
0
a
∫
–π
π
r J
k
¸
¸
,
_
ρ
kl
r
a
cos(kθ) f(r, θ) dθ dr = C
kl
π
∫
0
a
r J
k
¸
¸
,
_
ρ
kl
r
a
2
dr,
hence
C
kl
=
1
π
∫
0
a
rJ
k
¸
¸
,
_
ρ
kl
r
a
2
dr
∫
0
a
∫
–π
π
r J
k
¸
¸
,
_
ρ
kl
r
a
cos(kθ) f(r, θ) dθ dr.
Similarly,
D
kl
=
1
π
∫
0
a
rJ
k
¸
¸
,
_
ρ
kl
r
a
2
dr
∫
0
a
∫
–π
π
r J
k
¸
¸
,
_
ρ
kl
r
a
sin(kθ) f(r, θ) dθ dr,
k = 0, 1, 2, ... , l = 1, 2, ... .
13. Fourier Transforms. 41
13. Fourier Transforms.
A function f : R→R is said to be integrable on R if
∫
– ∞
∞
f(x)dx < ∞. We call the class of all
such functions L
1
(R). Similarly, f : R→R is said to be square integrable on R if
∫
– ∞
∞
f(x)
2
dx
< ∞. and we call the class of all such functions L
2
(R).
Examples. Let f : R→R and g : R→R be defined by
f(x) =
¹
'
¹
x
–
2
3
;0<x<1
0;otherwise
,
g(x) =
¹
'
¹
x
–
2
3
; x>1
0;otherwise
.
Then
∫
– ∞
∞
f(x)dx =
∫
0
1
x
–
2
3
dx = 3x
1
3

1
0
= 3 < ∞
and
∫
– ∞
∞
f(x)
2
dx =
∫
0
1
x
–
4
3
dx = – 3x
–
1
3

1
0
= ∞.
Therefore f ∈ L
1
(R) but f ∉ L
2
(R). On the other hand,
∫
– ∞
∞
g(x)dx =
∫
1
∞
x
–
2
3
dx = 3x
1
3

∞
1
= ∞
and
∫
– ∞
∞
g(x)
2
dx =
∫
0
1
x
–
4
3
dx = – 3x
–
1
3

∞
1
= 3 < ∞.
Therefore g ∉ L
1
(R) but g ∈ L
2
(R).
¯
Given two function f ∈ L
1
(R) and g ∈ L
1
(R) , the product fg is not necessarily in L
1
(R).
Counterexamples are given by
f(x) = g(x) =
¹
'
¹
x
–
2
3
;0<x<1
0;otherwise
.
There is a product
*
for which f
*
g ∈ L
1
(R) whenever f ∈ L
1
(R) and g ∈ L
1
(R). We define
the convolution product
13. Fourier Transforms. 42
(f
*
g)(x) =
∫
– ∞
∞
f(x – y) g(y) dy.
Lemma. If f ∈ L
1
(R) and g ∈ L
1
(R), then f
*
g ∈ L
1
(R).
Proof.
∫
– ∞
∞
(f
*
g)(x)dx =
∫
– ∞
∞
¹
¹
¹
¹
¹
¹
∫
– ∞
∞
f(x– y)g(y) dy dx
≤
∫
– ∞
∞
∫
– ∞
∞
f(x – y) g(y) dydx
≤
∫
– ∞
∞
f(x – y)
∫
– ∞
∞
g(y) dydx
≤
∫
– ∞
∞
f(x) dx
∫
– ∞
∞
g(y) dx < ∞.
¯
The Fourier transform of a function f ∈ L
1
(R) is defined to be
f
^
(ξ) =
∫
– ∞
∞
f(x)e
–iξx
dx,
where ξ ∈ R. It is easy to see that the Fourier transform of a function f ∈ L
1
(R) is a
bounded continuous function on R. The boundedness follows easily from
f
^
(ξ) =
¹
¹
¹
¹
¹
¹
∫
– ∞
∞
f(x)e
–iξx
dx ≤
∫
– ∞
∞
f(x) dx < ∞.
The continuity follows from the following.
Theorem. (RiemannLebesgue). If f ∈ L
1
(R), then f
^
∈ C(R) and f
^
(ξ) → 0 as ξ→ ∞.
Proof. Let ξ ∈ R, ξ ≠ 0. Then
f
^
(ξ) =
∫
– ∞
∞
f(x)e
–iξx
dx = – e
πi
∫
– ∞
∞
f(x)e
–iξx
dx
= –
∫
– ∞
∞
f(x)e
–iξ
¸
¸
,
_
x+
π
ξ
dx = –
∫
– ∞
∞
f
¸
¸
,
_
x–
π
ξ
e
–iξx
dx.
Therefore
2f
^
(ξ) =
¹
¹
¹
¹
¹
¹
∫
– ∞
∞
f(x)e
–iξx
dx–
∫
– ∞
∞
f
¸
¸
,
_
x–
π
ξ
e
–iξx
dx
≤
∫
– ∞
∞
¹
¹
¹
¹
¹
¹
f(x)– f
¸
¸
,
_
x–
π
ξ
e
–iξx
 dx
13. Fourier Transforms. 43
≤
∫
– ∞
∞
¹
¹
¹
¹
¹
¹
f(x)– f
¸
¸
,
_
x–
π
ξ
dx → 0 as ξ → ∞.
To prove continuity of f
^
, let ε > 0 be given and a > 0 chosen such that
∫
x>a
f(x)dx <
ε
4
and
δ > 0 chosen such that 2aδ
∫
x<a
f(x)dx < ε. Then for η < δ ,
  f
^
(ξ+η)– f
^
(ξ) =
¹
¹
¹
¹
¹
¹
∫
– ∞
∞
( ) f(x)e
–i(ξ+ η)x
– f(x)e
–iξx
dx
≤
∫
– ∞
∞
f(x)
.
e
–iξx
(e
–iηx
– 1) dx = 2
∫
– ∞
∞
f(x)
 
sin
( )
ηx
2
dx
≤ 2
∫
x>a
f(x)
 
sin
( )
ηx
2
dx + 2
∫
x<a
f(x)
 
sin
( )
ηx
2
dx
≤ 2
∫
x>a
f(x) dx + 2
∫
x<a
f(x)
 
ηx
2
dx
≤
ε
2
+ aδ
∫
x<a
f(x)dx ≤
ε
2
+
ε
2
= ε.
Therefore f
^
(ξ) is uniformly continuous on R.
¯
We define the linear transformation F: f → f
^
defined by
(Ff)(ξ) = f
^
(ξ) =
∫
– ∞
∞
f(x)e
–iξx
dx.
F is called the Fourier integral transformation and f
^
the Fourier transform of f ∈ L
1
(R).
13. Fourier Transforms. 44
Properties of Fourier transforms.
1. Translation (a). Let f : R → R and h ∈ R. The translate of f by h is the function τ
h
f
defined by
(τ
h
f)(x) = f(x – h), x ∈ R.
The Fourier transform of τ
h
f is
( )
τ
h
f
^
(ξ) =
∫
– ∞
∞
(τ
h
)(x)e
–iξx
dx
=
∫
– ∞
∞
f(x – h)e
–iξx
dx
=
∫
– ∞
∞
f(x )e
–iξ(x+ h)
dx
= e
–iξh
∫
– ∞
∞
f(x )e
–iξx
dx
= e
–iξh
f
^
(ξ).
(b). For real c,
( ) e
icx
f(x)
^
=
∫
– ∞
∞
e
–iξx
( ) e
icx
f(x) dx
=
∫
– ∞
∞
f(x )e
–iξ(x–c)
dx
= f
^
(ξ – c) =
( )
τ
c
f
^
(ξ).
2. Dilation. Let λ ∈ R, λ > 0, the dilation of f by λ is defined as δ
λ
f where
(δ
λ
f)(x) = λ
–
1
2
f(λ
–1
x), x ∈ R.
The Fourier transform of δ
λ
f is
( )
δ
λ
f
^
(ξ) =
∫
– ∞
∞
(δ
λ
f)(x)e
–iξx
dx
= λ
–
1
2
∫
– ∞
∞
f(λ
–1
x) e
–iξx
dx
= λ
–
1
2
∫
– ∞
∞
f(y) e
–iξλy
λdy
= λ
1
2
f
^
(λξ) =
( )
δ
λ
–1
f
^
(ξ)
13. Fourier Transforms. 45
3. Differentiation. Let f and f' ∈ L
1
(R) and denote by D the differential operator D =
∂
∂x
.
The Fourier transform of Df is
( )
Df
^
(ξ) =
∫
– ∞
∞
(Df)(x)e
–iξx
dx
= f(x)e
–iξx

∞
–∞
–
∫
– ∞
∞
f(x)( ) De
–iξx
dx
= (iξ)f
^
(ξ).
By induction
( )
D
k
f
^
= (iξ)
k
f
^
, k = 1, 2, ... .
4. Multiplication. We denote by ∂ the differential operator ∂ =
∂
∂ξ
. If f and xf ∈ L
1
(R)
then
( )
∂f
^
(ξ) =
∂
∂ξ
¸
¸
,
_
∫
– ∞
∞
f(x)e
–iξx
dx
=
∫
– ∞
∞
(–ix)f(x)e
–iξx
dx.
Therefore
( )
–ixf
^
= ∂f
^
. By induction it follows that ∂
k
f
^
=
( )
(–ix)
k
f
^
, k = 1, 2, ... .
5. Convolution. Let f, g ∈ L
1
(R). Then f
*
g ∈ L
1
(R) and has Fourier transform
(f
*
g)
^
(ξ) =
∫
– ∞
∞
(f
*
g)(x)e
–iξx
dx
=
∫
– ∞
∞
¸
¸
,
_
∫
– ∞
∞
f(x– y)g(y) dy e
–iξx
dx
=
∫
– ∞
∞
∫
– ∞
∞
f(x) g(y) e
–iξ(x+ y)
dx dy
=
¸
¸
,
_
∫
– ∞
∞
f(x)e
–iξx
dx
¸
¸
,
_
∫
– ∞
∞
g(y)e
–iξy
dy
= f
^
(ξ)g
^
(ξ).
Example 1. Let f(x) = e
–
x
2
2
. Then f ∈ L
1
(R) and has Fourier transform
f
^
(ξ) =
∫
– ∞
∞
e
–
x
2
2
e
–iξx
dx
13. Fourier Transforms. 46
∂f
^
∂ξ
(ξ) =
∂
∂ξ
∫
– ∞
∞
e
–
x
2
2
e
–iξx
dx
=
∫
– ∞
∞
e
–
x
2
2
(–ix)e
–iξx
dx
= i
∫
– ∞
∞
∂
∂x
¸
¸
,
_
e
–
x
2
2
e
–iξx
dx
= ie
–iξx
e
–
x
2
2

∞
–∞
– i
∫
– ∞
∞
e
–
x
2
2
∂
∂x
( ) e
–iξx
dx
= – ξ
∫
– ∞
∞
e
–
x
2
2
e
–iξx
dx
= – ξf
^
(ξ).
Therefore f
^
satisfies a first order ordinary differential equation with solution f
^
(ξ) =
ce
–
ξ
2
2
.
Since f
^
(0) = c =
∫
– ∞
∞
e
–
x
2
2
dx =
√
2π,
¸
¸
,
_
e
–
x
2
2
^
=
√
2π e
–
ξ
2
2
.
¯
Example 2. Let f(x) =
¹
¹
'
¹
¹
1;x≤ a
0;otherwise
≡ χ
a
(x) ∈ L
1
(R) and has Fourier transform
( )
χ
a
^
(ξ) =
∫
– ∞
∞
χ
a
(x) e
–iξx
dx =
∫
–a
a
e
–iξx
dx
= –
1
iξ
e
–iξx

a
–a
= –
1
iξ
( )
e
–iξa– e
–iξa
=
2sinaξ
ξ
.
14. Inverse Fourier Transforms. 47
14. Inverse Fourier Transforms.
The Fourier transform of f ∈ L
1
(R) is continuous, bounded and f
^
(ξ)→ 0 as ξ → ∞.
However f
^
is not neccesarily in L
1
(R). Assuming f ∈ L
1
(R), f piecewise smooth, and
f
^
∈ L
1
(R), define
f
a
(x) =
1
2π
∫
– a
a
f
^
(ξ) e
iξx
dξ
=
1
2π
∫
– a
a
¸
¸
,
_
∫
– ∞
∞
f(y)e
–iξy
dy e
iξx
dξ
=
1
2π
∫
– ∞
∞
¸
¸
,
_
∫
– a
a
e
iξ(x–y)
dξ f(y) dy
=
1
2π
∫
– ∞
∞
¸
¸
,
_
e
ia(x–y)
– e
–ia(x–y)
i(x– y)
f(y) dy
=
∫
– ∞
∞
D
a
(x – y) f(y) dy =
∫
– ∞
∞
D
a
(y) f(x – y) dy
=
( )
D
a
*
f (x)
where D
a
(x ) =
sinax
πx
.
Now
f
a
(x) –
1
2
[ ]
f(x
–
)+f(x
+
) =
1
π
∫
– ∞
∞
f(x – y)
sinay
y
dy –
1
2
[ ]
f(x
–
)+f(x
+
)
=
1
π
∫
0
∞
[ ]
f(x– y)–f(x
–
)
sinay
y
dy
+
1
π
∫
– ∞
0
[ ]
f(x– y)–f(x
+
)
sinay
y
dy
¸
¸
,
_
since
∫
0
∞
sinay
y
dy=
π
2
=
1
π
∫
0
∞
[ ]
f(x– y)–f(x
–
)
sinay
y
dy
+
1
π
∫
0
∞
[ ]
f(x+y)– f(x
+
)
sinay
y
dy
14. Inverse Fourier Transforms. 48
In the second integral,
1
π
∫
0
∞
[ ]
f(x+y)– f(x
+
)
sinay
y
dy
=
1
π
∫
0
K
[ ]
f(x+y)– f(x
+
)
sinay
y
dy +
1
π
∫
K
∞
[ ]
f(x+y)– f(x
+
)
sinay
y
dy
If K ≥ 1,
¹
¹
¹
¹
¹
¹
∫
K
∞
f(x+y)
sinay
y
dy ≤
∫
K
∞
 
f(x+y) dy
and
∫
K
∞
f(x
+
)
sinay
y
dy = f(x
+
)
∫
K
∞
sinay
y
dy.
Since
∫
0
∞
f(x)dx ,
∫
0
∞
sinay
y
dy are both convergent integrals,
∫
K
∞
sinay
y
dy → 0,
∫
K
∞
 
f(x+y) dy → 0 as K → ∞.
For the integrals over [0, K],
∫
0
K
[ ]
f(x+y)– f(x
+
)
sinay
y
dy =
∫
– ∞
∞
¸
¸
,
_
e
iay
– e
– iay
2i
g(y) dy
=
1
2i
[ ]
g
^
(–a)–g
^
(a)
where g(y) =
¹
'
¹
f(x+y)– f(x
+
)
y
;0<y<K
0;otherwise
.
Since f is piecewise smooth, f'(x
+
) exists for all x ∈ R and lim
y→0
+
g(y) = f'(x
+
). Therefore
g is bounded on [0, K] and hence g ∈ L
1
(R). By the RiemannLebesgue lemma, g
^
exists, is
continuous, g
^
(ta) → 0 as a → ∞ and therefore
∫
0
K
[ ]
f(x+y)– f(x
+
)
sinay
y
dy → 0 as a →∞
for K ≥ 1. A virtually identical argument works for the first integral.
Therefore
 
f
a
(x)–
1
2
[ ]
f(x
–
)+f(x
+
) → 0 as a → ∞
or
lim
a→∞
f
a
(x) = lim
a→∞
1
2π
∫
–a
a
f
^
(ξ) e
iξx
dξ
=
1
2π
∫
– ∞
∞
f
^
(ξ) e
iξx
dξ =
1
2
[ ]
f(x
–
)+f(x
+
) .
Summarising
14. Inverse Fourier Transforms. 49
Theorem. Let f ∈ L
1
(R) and let f be piecewise smooth on R. Then for every x ∈ R,
1
2π
∫
– ∞
∞
f
^
(ξ) e
iξx
dξ =
1
2
[ ]
f(x
–
)+f(x
+
) .
If x is a point of continuity of f, then
1
2π
∫
– ∞
∞
f
^
(ξ) e
iξx
dξ = f(x).
¯
For φ∈L
1
(R), we call φ
˘
(x) =
1
2π
∫
– ∞
∞
φ(ξ) e
iξx
dξ the inverse Fourier transform of φ and
F
–1
defined by ( ) F
–1
φ (x) =
1
2π
∫
– ∞
∞
φ(ξ) e
iξx
dξ in called the inverse Fourier transformation.
If φ is the Fourier transform of a piecewise smooth function f ∈ L
1
(R), that is φ = f
^
, then
we define f(x) =
1
2
[ ]
f(x
–
)+f(x
+
) at a point of discontinuity of f.
Then f(x) = φ
˘
(x) =
1
2π
∫
– ∞
∞
f
^
(ξ) e
iξx
dξ. That is F
–1
Ff = f.
Example. The inverse Fourier transform is useful in computing Fourier transforms.
Since F
( )
e
–ax
=
2a
a
2
+ξ
2
, it follows that F
–1
F
( )
e
–ax
= F
–1
¸
¸
,
_
2a
a
2
+ξ
2
or
e
–ax
= F
–1
¸
¸
,
_
2a
a
2
+ξ
2
. Interchanging the roles of x and ξ, multiplying by 2π, leads to
F
¸
¸
,
_
2a
a
2
+x
2
= 2π e
–aξ
or F
¸
¸
,
_
1
a
2
+x
2
=
π
a
e
–aξ
. In a similar fashion, every Fourier
transform pair defines a dual pair using the inverse Fourier transform.
15. Applications to Differential Equations. 50
15. Applications to Differential Equations.
1. The wave equation. The equation
∂
2
u
∂t
2
– c
2
∂
2
u
∂x
2
= 0, – ∞ < x < ∞, t > 0,
describes the vertical vibrations of an infinite stretched elastic string, where u(x, t) is the
vertical displacement of the string from its rest position at position x, time t. Let the initial
displacement and velocity be given as
u(x, 0) = f(x),
∂u
∂t
(x, t) = g(x), – ∞ < x < ∞.
We take Fourier transforms of the wave equation and the initial conditions with respect to
the x variable and denote by u
^
(ξ, t ) the Fourier transform F
( )
u(x, t) . The using the
derivative properties of the Fourier Transform,
∂
2
u
^
∂t
2
– (ciξ)
2
u
^
= 0, – ∞ < ξ < ∞, t > 0,
or
∂
2
u
^
∂t
2
+ (cξ)
2
u
^
= 0.
Then
u
^
(ξ, t) = A(ξ) cos(cξt) + B(ξ) sin(cξt).
When t = 0, u
^
(ξ, t) = A(ξ) = f
^
(ξ),
∂u
^
∂t
(ξ, t) = g
^
(ξ) = cξB(ξ). Therefore
u
^
(ξ, t) = f
^
(ξ) cos(cξt) +
g
^
(ξ)
2c
2sin(cξt)
ξ
= f
^
(ξ)
¸
¸
,
_ e
i(ct)ξ
+e
– i(ct)ξ
2
+
1
2c
( )
χ
ct
*
g
^
(ξ).
Using the translation property and the convolution theorem,
u(x, t) =
1
2
[ ]
f(x+ct)+f(x–ct) +
1
2c
∫
– ∞
∞
χ
ct
(x – ct) g(y) dy
=
1
2
[ ]
f(x+ct)+f(x–ct) +
1
2c
∫
x–ct<ct
g(y) dy
=
1
2
[ ]
f(x+ct)+f(x–ct) +
1
2c
∫
x – ct
x + ct
g(y) dy.
This is d'Alemberts solution to the onedimensional wave equation.
15. Applications to Differential Equations. 51
2. Laplace's Equation.
Consider Laplace's equation in two variables on the upper halfplane y > 0. Then
∂
2
u
∂x
2
+
∂
2
u
∂y
2
= 0, – ∞ < x < ∞, y > 0.
Let the boundary condition
u(x, 0) = f(x), – ∞ < x < ∞
be given for a function f ∈ L
1
(R). Then taking the Fourier transform in the variable x,
(iξ)
2
u
^
(ξ, y) +
∂
2
u
^
∂y
2
(ξ, y) = 0, – ∞ < x < ∞, y > 0.
or
∂
2
u
^
∂y
2
(ξ, y) – ξ
2
u
^
(ξ, y) = 0,
which has solutions
u
^
(ξ, y) = A(ξ)e
ξy
+ B(ξ)e
– ξy
.
We need two conditions to determne the functions A(ξ), B(ξ). In addition to u(x, 0) = f(x), –
∞ < x < ∞, let
∂u
∂y
(x, 0) = g(x), – ∞ < x < ∞ for some function g ∈ L
1
(R). We will find g
such that the solution u(x, y) is bounded for y > 0, in fact such that u(x, y) → 0 as y → ∞.
Taking transforms,
u
^
(ξ, 0) = f
^
(ξ),
∂u
^
∂y
(ξ, 0) = g
^
(ξ),
or
A(ξ) + B(ξ) = f
^
(ξ), ξA(ξ) – ξB(ξ) = g
^
(ξ).
Solving for A(ξ), B(ξ),
A(ξ) =
1
2
¸
¸
,
_
f
^
(ξ)+
g
^
(ξ)
ξ
, B(ξ) =
1
2
¸
¸
,
_
f
^
(ξ)–
g
^
(ξ)
ξ
,
and
u
^
(ξ, y) =
1
2
¸
¸
,
_
f
^
(ξ)+
g
^
(ξ)
ξ
e
ξy
+
1
2
¸
¸
,
_
f
^
(ξ)–
g
^
(ξ)
ξ
e
– ξy
.
For ξ > 0, u
^
(ξ, y) → 0 as y → 0 if and only if f
^
(ξ) +
g
^
(ξ)
ξ
= 0, and for ξ < 0, u
^
(ξ, y) → 0 as
y → 0 if and only if f
^
(ξ) –
g
^
(ξ)
ξ
= 0. Therefore
g
^
(ξ) =
¹
¹
'
¹
¹
–ξf
^
(ξ);ξ>0
ξf
^
(ξ);ξ<0
= – ξ f
^
(ξ).
Hence
15. Applications to Differential Equations. 52
u
^
(ξ, y) =
¹
¹
'
¹
¹
f
^
(ξ)e
– ξy
;ξ>0
f
^
(ξ)e
ξy
;ξ<0
= f
^
(ξ)e
– ξy
.
Since e
– ξy
=
¸
¸
,
_
y
π(x
2
+y
2
)
^
, by the convolution theorem,
u(x, y) =
¸
¸
,
_
y
π(x
2
+y
2
)
*
f =
∫
– ∞
∞
y
π((x– s)
2
+y
2
)
f(s) ds.
3. The Heat Equation.
The heat equation,
∂u
∂t
– κ
∂
2
u
∂x
2
= 0, – ∞ < x < ∞, t > 0
for u(x, t) a function of two variables, with initial condition
u(x, t) = f(x), – ∞ < x < ∞
for f ∈ L
1
(R), f continuous and bounded, can be solved using Fourier transforms. Taking
transforms in the variable x,
∂u
^
∂t
(ξ, t) – κ (iξ)
2
u
^
(ξ, t) = 0,
or
∂u
^
∂t
(ξ, t) + κ ξ
2
u
^
(ξ, t) = 0.
Solving this first order ordinary differential equation,
u
^
(ξ, t) = u
^
(ξ, 0)e
– κξ
2
t
= f
^
(ξ)e
– κξ
2
t
.
Now ¸
¸
,
_
e
–
x
2
2
^
=
√
2π e
–
ξ
2
2
and using the dilation property of Fourier transforms
¸
¸
,
_
λ
–
1
2
e
–
(λ
–1
x)
2
2
^
=
√
2π λ
1
2
e
–
(λξ)
2
2
e
–
(λξ)
2
2
=
1
λ
√
2π
¸
¸
,
_
e
–
(λ
–1
x)
2
2
^
.
Let
λ
2
2
= κt or λ =
√
2κt , then
15. Applications to Differential Equations. 53
e
– κξ
2
t
=
1
√
2π
1
√
2κt
¸
¸
,
_
e
–
1
2
¸
¸
,
_
x
√
2κt
2
^
=
1
√
4πκt
¸
¸
,
_
e
–
1
2
¸
¸
,
_
x
√
2κt
2
^
.
Therefore from the convolution theorem, since
u
^
(ξ, t) = f
^
(ξ)
1
√
4πκt
¸
¸
,
_
e
–
1
2
¸
¸
,
_
x
√
2κt
2
^
u(x, t) =
¸
¸
,
_
f
*
1
√
4πκt
e
–
1
2
¸
¸
,
_
x
√
2κt
2
(x)
=
1
√
4πκt
∫
– ∞
∞
e
–
(x–y)
2
4κt
f(y) dy.
The function h(x, t) =
e
–
x
2
4κt
√
4πκt
, is called the heat kernel. We can then write
u(x, t) =
∫
– ∞
∞
h(x – y, t) f(y) dy.
16. Plancherels' and Parsevals' Identities.
54
16. Plancherels' and Parsevals' Identities.
By the convolution theorem, for f, g ∈ L
1
(R)∩L
2
(R),
( )
f
*
g
^
= f
^
g
^
.
Therefore
f
*
g =
( )
f
^
g
^
˘
or
∫
– ∞
∞
f(x – y) g(y) dy =
1
2π
∫
– ∞
∞
e
i ξx
f
^
(ξ) g
^
(ξ) dξ.
Set x = 0,
∫
– ∞
∞
f(– y) g(y) dy =
1
2π
∫
– ∞
∞
f
^
(ξ) g
^
(ξ) dξ
Replacing g(x) by g(–x) , the Fourier transform g
^
(ξ) is replaced by g
^
(ξ) , hence
∫
– ∞
∞
f(– y) g(–y) dy =
1
2π
∫
– ∞
∞
f
^
(ξ) g
^
(ξ) dξ
or
∫
– ∞
∞
f( y) g(y) dy =
1
2π
∫
– ∞
∞
f
^
(ξ) g
^
(ξ) dξ.
This is Plancherels' identity. When f = g we obtain Parsevals' identity,
∫
– ∞
∞
f( y)
2
dy =
1
2π
∫
– ∞
∞
f
^
(ξ)
2
dξ.
Examples.
1. Let f(x) = χ
a
(x) =
¹
¹
'
¹
¹
1;x< a
0;otherwise
.
Then f
^
(ξ) =
2sinaξ
ξ
, and by Parsevals' identity,
∫
– ∞
∞
χ
a
(x)
2
dx =
1
2π
∫
– ∞
∞
¹
¹
¹
¹
¹
¹
sinaξ
ξ
2
dξ
∫
– ∞
∞
¸
¸
,
_
2sinaξ
ξ
2
dξ = 4πa
or
∫
– ∞
∞
¸
¸
,
_
sinaξ
ξ
2
dξ = πa.
16. Plancherels' and Parsevals' Identities.
55
2. Let f(x) = e
– ax
, f
^
(ξ) =
2a
a
2
+ξ
2
and by Parsevals' identity
∫
– ∞
∞
( )
e
– ax 2
dx =
1
2π
∫
– ∞
∞
¸
¸
,
_
2a
a
2
+ξ
2
2
dξ
1
a
=
2a
2
π
∫
– ∞
∞
1
(a
2
+ξ
2
)
2
dξ
or
∫
– ∞
∞
1
(a
2
+ξ
2
)
2
dξ =
π
2a
3
.
17. Band Limited Functions and Shannon's Sampling Theorem.
56
17. Band Limited Functions and Shannon's Sampling Theorem.
The Fourier transform variable has the role of frequency and f
^
(ξ) is referred to as the
frequency representation of f(x).
If f
^
(ξ) = 0 for ξ > ξ
c
> 0, then f(x) is called a bandlimited function and ξ
c
is called the
cutoff frequency. Many functions from science and technology, are bandlimited. For
example, human hearing is assumed to be limited to frequencies below about 20 kHz.
Therefore the acoustic signals recorded on compact discs are limited to a bandwidth of 22
kHz.
A first step in the processing of signals, is sampling. A signal represented by a continuous
function f(x), is replaced by its samples at regular intervals, {f(nL), n = 0, t1, t2, ... }.
Shannons' Theorem shows that it is possible to exactly recover the bandlimited
continuous function f(x) from knowledge of its samples, provided the sampling interval L,
is sufficiently short.
Consider the functions φ
^
n
(ξ) given by
φ
^
n
(ξ) =
¹
¹
'
¹
¹
0;ξ > ξ
c
1
√
2ξ
c
e
–
inπξ
ξ
c
; ξ≤ξ
c
The inverse Fourier transforms of φ
^
n
(ξ) are given by
φ
n
(x) =
1
2π
∫
– ∞
∞
e
i xξ
φ
^
n
(ξ) dξ
2π φ
n
(x) =
∫
– ξ
c
ξ
c
e
i xξ
1
√
2ξ
c
e
–
inπξ
ξ
c
dξ
=
1
√
2ξ
c
∫
– ξ
c
ξ
c
e
i ξ
¸
¸
,
_
x–
nπ
ξ
c
dξ
=
1
√
2ξ
c
¸
]
1
1
1
e
i ξ
¸
¸
,
_
x–
nπ
ξ
c
i
¸
¸
,
_
x–
nπ
ξ
c
ξ
c
–ξ
c
=
√
2ξ
c
¸
¸
,
_
e
i (ξ
c
x–nπ)
– e
–i (ξ
c
x–nπ)
2i(ξ
c
x–nπ)
=
√
2ξ
c
sin(ξ
c
x–nπ)
ξ
c
x–nπ
17. Band Limited Functions and Shannon's Sampling Theorem.
57
=
√
2π
L
sinξ
c
(x–nL)
ξ
c
(x–nL)
where L =
π
ξ
c
.
Consider the inner products
< >
φ
^
n
;φ
^
m
=
∫
– ∞
∞
φ
^
n
(ξ) φ
^
m
(ξ) dξ
=
1
2ξ
c
∫
– ξ
c
ξ
c
e
–
inπξ
ξ
c
¸
¸
,
_
e
–
imπξ
ξ
c
dξ
=
1
2ξ
c
∫
– ξ
c
ξ
c
e
i(m–n) πξ
ξ
c
dξ
=
¹
¹
'
¹
¹
0; m≠n
1; m=n
.
So the functions {φ
^
n
(ξ) ; n = 0, t1, t2, ... }, form an orthogonal set in L
2
(–ξ
c
, ξ
c
). Since f
^
(ξ)
is band limited to ξ ≤ ξ
c
, it has a Fourier series
f
^
(ξ) =
∑
n=–∞
∞
c
n
φ
^
n
(ξ)
=
∑
n=–∞
∞
c
n
1
√
2ξ
c
e
–
inπξ
ξ
c
dξ
where the Fourier coefficients c
n
are given by
c
n
=
< >
f
^
; φ
^
n
=
∫
– ξ
c
ξ
c
f
^
(ξ)
1
√
2ξ
c
e
inπξ
ξ
c
dξ.
By Plancherel's theorem
c
n
=
< >
f
^
; φ
^
n
= 2π
< >
f; φ
n
so
f
^
(ξ) =
∑
n=–∞
∞
c
n
φ
^
n
(ξ) =
∑
n=–∞
∞
2π
< >
f; φ
n
φ
^
n
(ξ).
Taking inverse Fourier transforms,
f(x) =
∑
n=–∞
∞
< >
f; φ
n
2π φ
n
(x)
=
∑
n=–∞
∞
< >
f; φ
n
√
2π
L
sinξ
c
(x–nL)
ξ
c
(x–nL)
.
17. Band Limited Functions and Shannon's Sampling Theorem.
58
The samples at intervals of length L =
π
ξ
c
are
f(kL) =
∑
n=–∞
∞
< >
f; φ
n
√
2π
L
sinξ
c
(kL–nL)
ξ
c
(kL–nL)
=
∑
n=–∞
∞
< >
f; φ
n
√
2π
L
sin(k–n)π
(k–n)π
=
< >
f; φ
k
√
2π
L
.
So
< >
f; φ
k
= f(kL)
√
L
2π
, k = 0, t1, t2, ... .
Finally therefore
f(x) =
∑
n=–∞
∞
< >
f; φ
n
√
2π
L
sinξ
c
(x–nL)
ξ
c
(x–nL)
=
∑
n=–∞
∞
f(nL)
sinξ
c
(x–nL)
ξ
c
(x–nL)
.
Summarising we have Shannon's theorem,
Theorem. Let f ∈ L
1
(R)∩L
2
(R) be continuous and band limited to ξ ≤ ξ
c
. Then
f(x) =
∑
n=–∞
∞
f(nL)
sinξ
c
(x–nL)
ξ
c
(x–nL)
where L =
π
ξ
c
.
¯
The relationship ωL = 2π , where ω is the frequency of sampling in cycles per unit length,
shows that from Shannon's theorem, to reconstruct a bandlimited function, it suffices to
sample at a frequency ω =
2π
L
= 2ξ
c
, twice the cutoff frequency.
18. HeisenbergÕs Inequality.
59
18. Heisenberg’s Inequality.
Let f ∈ L
2
(R), xf ∈ L
2
(R). Then the quantity
∆
a
f ≡
¸
¸
,
_
∫
– ∞
∞
(x–a)
2
f(x)
2
dx
∫
– ∞
∞
f(x)
2
dx
is called the dispersion about the point x = a of f. The reasoning behind the definition is
that if f(x) is concentrated near x = a , then ∆
a
f is smaller than when f is not close to zero
far from x = a.
Example. Consider the characteristic function
χ
b
(x) =
¹
¹
'
¹
¹
1;x≤ b
0;otherwise
which has Fourier transform
(χ
b
)
^
(ξ) =
2sinbξ
ξ
.
Notice that χ
b
is concentrated near x = 0 for small b. The dispersion about the origin is
∆
0
χ
b
=
¸
¸
,
_
∫
– b
b
x
2
dx
∫
– b
b
dx
=
b
2
3
and it clear that the dispersion increases as b increases.
Notice that the Fourier transform χ
^
b
does not have a finite dispersion about the origin,
since
∫
– ∞
∞
ξ
2
χ
^
b
(ξ)
2
dξ =
∫
– ∞
∞
ξ
2
¸
¸
,
_
2sin(bξ)
ξ
2
dξ
= 4
∫
– ∞
∞
sin
2
(bξ) dξ = ∞.
This indicates that χ
^
b
is spread out away from x = 0.
The following result shows that there is a type of inverse relationship between the
dispersion of a function and that of its Fourier transform.
Theorem. Let f ∈ L
1
(R)∩L
2
(R). Then for all a, α ∈ R,
18. HeisenbergÕs Inequality.
60
( )
∆
a
f
( )
∆
α
f
^
=
¸
¸
,
_
∫
– ∞
∞
x
2
f(x)
2
dx
∫
– ∞
∞
f(x)
2
dx
¸
¸
,
_
∫
– ∞
∞
ξ
2
f
^
(ξ)
2
dξ
∫
– ∞
∞
f
^
(ξ)
2
dξ
≥
1
4
and equality holds if and only if f(x) = c e
–kx
2
for constants c ∈ R and k > 0.
Proof. We firstly prove the result for a = α = 0.
∫
– ∞
∞
x
2
f(x)
2
dx and
∫
– ∞
∞
ξ
2
f
^
(ξ)
2
dξ are
both assumed finite since otherwise the result is trivial.
Let f
*
(x) ≡ ((iξf
^
(ξ))
˘
or (f
*
)
^
(ξ) = iξf
^
(ξ). Then f
*
∈ L
2
(R), and
¸
¸
,
_
∫
– ∞
∞
x
2
f(x)
2
dx
¸
¸
,
_
∫
– ∞
∞
ξ
2
f
^
(ξ)
2
dξ =
¸
¸
,
_
∫
– ∞
∞
x
2
f(x)
2
dx
¸
¸
,
_
∫
– ∞
∞
(iξ)f
^
(ξ)
2
dξ
=
¸
¸
,
_
∫
– ∞
∞
x
2
f(x)
2
dx
¸
¸
,
_
∫
– ∞
∞
(f
*
)
^
(ξ)
2
dξ
= 2π
¸
¸
,
_
∫
– ∞
∞
x
2
f(x)
2
dx
¸
¸
,
_
∫
– ∞
∞
f
*
(x)
2
dx (by Parseval’s identity).
Since
¸
]
1
1
∫
– ∞
∞
1
2
x
( )
f
*
(x) f(x) + f
*
(x) f(x) dx
2
=
¸
]
1
1
∫
– ∞
∞
xRe
( )
f
*
(x) f(x) dx
2
=
¸
]
1
1
Re
∫
– ∞
∞
xf (x) f
*
(x) dx
2
≤
¹
¹
¹
¹
¹
¹
∫
– ∞
∞
xf (x) f
*
(x) dx
2
≤
¸
¸
,
_
∫
– ∞
∞
x
2
f(x)
2
dx
¸
¸
,
_
∫
– ∞
∞
f
*
(x)
2
dx
(by the CauchySchwartz inequality).
We will show that
–
∫
– ∞
∞
x
( )
f
*
(x) f(x) + f
*
(x) f(x) dx =
∫
– ∞
∞
f(x)
2
dx
from which the result follows, for then
¸
¸
,
_
∫
– ∞
∞
x
2
f(x)
2
dx
¸
¸
,
_
∫
– ∞
∞
ξ
2
f
^
(ξ)
2
dξ = 2π
¸
¸
,
_
∫
– ∞
∞
x
2
f(x)
2
dx
¸
¸
,
_
∫
– ∞
∞
f
*
(x)
2
dx
≥ 2π
¸
]
1
1
∫
– ∞
∞
1
2
x
( )
f
*
(x) f(x) + f
*
(x) f(x) dx
2
18. HeisenbergÕs Inequality.
61
=
π
2
¸
]
1
1
∫
– ∞
∞
f(x)
2
dx
2
=
π
2
¸
¸
,
_
∫
– ∞
∞
f(x)
2
dx
¸
¸
,
_ 1
2π
∫
– ∞
∞
f
^
(ξ)
2
dξ
=
1
4
¸
¸
,
_
∫
– ∞
∞
f(x)
2
dx
¸
¸
,
_
∫
– ∞
∞
f
^
(ξ)
2
dξ .
To complete the proof, we assume that f is continuous and piecewise smooth, This
assumption can be removed since functions in L
1
(R) are the uniform limit of such
functions.
Then from the property of Fourier transforms,
f
*
(x) = f'(x)
wherever the derivative exists. Then for any interval [a, b] ,
b f(b)
2
– a f(a)
2
=
∫
a
b
d
dx
( ) xf(x)
2
dx
=
∫
a
b
( )
xf' (x) f(x) +x f(x) f' (x) + f(x)
2
dx
=
∫
a
b
x
( )
f
*
(x) f(x) + f
*
(x) f(x) dx +
∫
a
b
f(x)
2
dx
The assumption f ∈ L
2
(R) implies that b f(b)
2
→ 0 as b → ∞ and a f(a)
2
→ 0 as a → – ∞
since otherwise f(x) > c x
–1/2
as x → ∞, which is not integrable. Taking the limit as
b → ∞ and a → – ∞,
0 =
∫
– ∞
∞
x
( )
f
*
(x) f(x) + f
*
(x) f(x) dx +
∫
– ∞
∞
f(x)
2
dx
as required.
As for the case of equality in Heisenberg’s inequality, this holds if and only if f(x) f
*
(x) is
real and f
*
(x) = K x f(x) for some complex constant K. That is,
f(x) f
*
(x) = f(x) K xf(x) = x f(x)
2
K is real.
Therefore K is real.
The differential equation
f'(x) = f’ (x) = K x f(x)
has solutions of the form
f(x) = c e
–
Kx
2
2
, c any real constant,
18. HeisenbergÕs Inequality.
62
and f(x) = c e
–
Kx
2
2
∈ L
2
(R) if and only if K > 0. Therefore equality holds in Heisenberg’s
inequality only if f(x) = c e
–kx
2
for constants c ∈ R and k > 0.
Conversely, let f(x) = e
–
Kx
2
2
for constant K > 0. Then
¸
¸
,
_
∫
– ∞
∞
x
2
f(x)
2
dx
¸
¸
,
_
∫
– ∞
∞
ξ
2
f
^
(ξ)
2
dξ = 2π
¸
¸
,
_
∫
– ∞
∞
x
2
f(x)
2
dx
¸
¸
,
_
∫
– ∞
∞
f
*
(x)
2
dx
= 2π
¸
¸
,
_
∫
– ∞
∞
x
2
f(x)
2
dx
¸
¸
,
_
∫
– ∞
∞
K xf(x)
2
dx
= 2π K
2
¸
¸
,
_
∫
– ∞
∞
x
2
f(x)
2
dx
2
= 2π K
2
¸
¸
,
_
∫
– ∞
∞
x
2
e
–Kx
2
dx
2
= 2π K
2
¸
]
1
1
∫
– ∞
∞
¸
¸
,
_
x
–2K
( )
–2Kxe
–Kx
2
dx
2
= 2π K
2
¸
]
1
1
∫
– ∞
∞
¸
¸
,
_
x
–2K
( )
–2Kxe
–Kx
2
dx
2
= 2π K
2
¸
]
1
1
∫
– ∞
∞
¸
¸
,
_
x
–2K
d
dx
( )
e
–Kx
2
dx
2
= 2π K
2
¸
]
1
1
∫
– ∞
∞
¸
¸
,
_
1
2K
e
–Kx
2
dx
2
(integration by parts)
=
π
2
¸
]
1
1
∫
– ∞
∞
e
–Kx
2
dx
2
=
π
2K
¸
]
1
1
∫
– ∞
∞
e
–t
2
dt
2
=
π
2
2K
.
Whereas,
1
4
¸
¸
,
_
∫
– ∞
∞
f(x)
2
dx
¸
¸
,
_
∫
– ∞
∞
f
^
(ξ)
2
dξ =
1
4
¸
¸
,
_
∫
– ∞
∞
f(x)
2
dx 2π
¸
¸
,
_
∫
– ∞
∞
f(x)
2
dx
=
π
2
¸
¸
,
_
∫
– ∞
∞
f(x)
2
dx
2
18. HeisenbergÕs Inequality.
63
=
π
2
¸
]
1
1
∫
– ∞
∞
e
–Kx
2
dx
2
and equality holds.
The case of a ≠ 0, α ≠ 0, follows by observing that F(x) = e
–iαx
f(x + a) satisfies the same
hypotheses as f(x) and ∆
a
f = ∆
0
F and ∆
α
f
^
= ∆
0
F
^
for any a ≠ 0, α ≠ 0.
As a consequence of the inequality
( )
∆
a
f
( )
∆
α
f
^
≥
1
4
, we see that it is impossible for
both ∆
a
f and ∆
α
f
^
to be simultaneously small. That is, if one of ∆
a
f or ∆
α
f
^
is very small
then the other must be large.
Table of Contents
1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18.
Introduction ................................................................................................ ................................. 1 Linear differential operators ................................................................................................ ...... 3 Separation of variables ................................................................................................ ............... 5 Fourier Series................................................................................................ .............................. 9 Bessel's inequality................................................................................................ ..................... 14 Convergence results for Fourier series .................................................................................... 16 Differentiation and Integration of Fourier Series ................................................................... 20 Halfrange Fourier series ................................................................................................ ......... 23 General Intervals ................................................................................................ ..................... 25 Application to Laplace's equation ............................................................................................ 27 SturmLiouville problems and orthogonal functions .............................................................. 33 Bessel Functions ................................................................................................ ....................... 37 Fourier Transforms ................................................................................................ .................. 41 Inverse Fourier Transforms ................................................................................................ ..... 47 Applications to Differential Equations .................................................................................... 50 Plancherel's and Parseval's Identities ..................................................................................... 54 Band Limited Functions and Shannon's Sampling Theorem ................................................. 56 Heisenberg’s Inequality ................................................................................................ ............ 59
1. Introduction. 1. Introduction. Fourier theory is a branch of mathematics first invented to solve certain problems in partial differential equations. The most wellknown of these equations are: Laplace's equation,
1
∂2u ∂2u + = 0, for u(x, y) a function of two variables, ∂x 2 ∂y 2 ∂2u ∂2u – c 2 2 = 0, for u(x, t) a function of two variables, ∂t2 ∂x ∂u ∂2u – κ 2 = 0, for u(x, t) a function of two variables. ∂t ∂x
the wave equation,
the heat equation,
In the heat equation, x represents the position along the bar measured from some origin, t represents time, u(x , t) the temperature at position x, time t. Fourier was initially concerned with the heat equation. Incidentally, the same equation describes the concentration of a dye diffusing in a liquid such as water. For this reason the equation is sometimes called the diffusion equation. In the wave equation, x, represents the position along an elastic string under tension, measured from some origin, t represents time, u(x, t) the displacement of the string from equilibrium at position x, time t. In Laplace's equation, u(x, y) represents the steady temperature of a flat conducting plate at the position (x, y) in the plane. Since both the heat equation and the wave equation involve a single space variable x, we sometime refer to them as the one dimensional heat equation and the one dimensional wave equation respectively. Laplace's equation involves two spatial variables and is therefore sometimes called the twodimensional laplace equation. Laplace's equation is connected to the theory of analytic functions of a complex variable. If f(z) = u(x, y) + iv(x, y), the real and imaginary parts u(x, y), v(x, y) satisfy the CauchyRiemann equations,
∂u ∂v ∂x = ∂y ,
Then
∂u ∂v = – ∂x , ∂y
∂2u ∂2v ∂2u =– 2 2 = ∂x ∂y ∂x∂y ∂2u ∂2u + = 0. ∂x 2 ∂y 2 ∂2v ∂2v + = 0. ∂x 2 ∂y 2
or
Similarly,
Heat conduction and wave propagation usually occur in 3 space dimensions and are described by the following versions of Laplace's equation, the heat equation and the wave equation; ∂2u ∂2u ∂2u + + = 0, ∂x 2 ∂y 2 ∂z2
1. ∂t ∂x ∂y ∂z ∂2u ∂2u ∂2u ∂2u – c 2 2 + 2 + 2 = 0. Introduction. ∂x ∂t2 ∂y ∂z . 2 ∂u ∂2u ∂2u ∂2u – κ 2 + 2 + 2 = 0.
..xn) be a function of n variables x = (x1 .. cn . Let u(x1 .x2 . ∂x ∂y ∂z and L[u] ≡ ∂2u ∂2u ∂2u ∂2u ∂ 2u – c 2 2 + 2 + 2 = 2 – c2 ∇2u in the heat equation. the equation L[u] = F(x) is called nonhomogeneous.. Linearity of L is essential to the success of the Fourier method. + cnun] = c1L[u1 ] + .. There are usually (infinitely) many solutions of a linear partial differential equation. un and constants c1.. c2. All of the above mentioned partial differential equations can be written in the form L[u] = F where 2 L[u] ≡ ∇ u = 3 ∂2u ∂2u ∂2u + + ∂x 2 ∂y 2 ∂z2 in Laplace's equation. If F(x) ≠ 0.. This definition generalises to L[c1u1 + .... When F(x) ≡ 0. Linear differential operators.x2 . L[c1u1 + c2u2 ] = c1L[u1 ] + c2L[u2 ].. The number of . a linear partial differential operator. The general linear second order partial differential equation is of the form L[u] = F(x) where F(x) is a given function... bi (x).. ∂x ∂t2 ∂y ∂z ∂t L [u ] is in each case.. Linear differential operators. u2 . the equation L[u] = 0 is called homogeneous... and any two constants c 1.2. c(x) are given coefficients. 2. Henceforth we will consider only second order partial differential operators of the form L[u] = ∑∑ i=1 j=1 n n ∂2u aij (x)∂x ∂x + i j ∑ bi(x) ∂x i i=1 n ∂u + c(x)u. In other words. The highest order partial derivative appearing is the order of the partial differential operator. + cnL[un] for any functions u1 . Then the most general linear partial differential operator is of the form L[u] = ∑∑ i=1 j=1 n n ∂2u aij (x)∂x ∂x + i j ∑ bi(x) ∂x i i=1 n ∂u + c(x)u where a ij (x).. ∂u ∂2u ∂2u ∂2u ∂u L[u] ≡ ∂t – κ 2 + 2 + 2 = ∂t –κ ∇2 u in the heat equation. Linearity means that for any two functions u1 ... L is linear if it preserves linear combinations of u1. u2 ..xn).
at time t. Often these extra conditions are given as linear equations involving u and its derivatives on the boundary of some region Ω ⊂ Rn.. x ∈ ∂Ω. uk satisfy the linear partial differential equations L[uj ] = Fj . This result is commonly referred to as the principle of superposition and it is of paramount importance for the Fourier method. . j = 1... x ∈ Ω. Linear differential operators. It shows that by taking linear combinations of solutions of related linear partial differential equations. t > 0.. of a conducting body. x ∈ Ω. Suppose the temperature at the boundary is maintained at a given temperature φ ... t > 0...2. Then u satisfies the heat equation 4 ∂u 2 ∂t –κ ∇ u = 0..... These equations are written as boundary conditions B[u(x)] = φ(x) where B is a partial differential operator defined on the boundary ∂Ω of the region Ω. Suppose u1 . k. other solutions can be constructed for source and boundary terms F. t) be the temperature at x ∈ Ω. .. t)... Also the initial temperature of the body is given by u(x. k and boundary conditions B[uj ] = φ j . . then the linear combination u = c1u1 + . + c kFk and the boundary condition B[u] = c1φ1 + . + c kuk satisfies the linear partial differential equation L[u] = c 1F1 + . j = 1. t) = φ(x. then u(x.. + c kφk .. φ which are linear combinations of simpler terms. 0) = f(x). solutions may be restricted by imposing extra conditions. . As an example let u(x. .
Separation of variables.t) along the bar. Separation of variables. Dividing by κX(x)T(t). say λ. Eliminating A. t) = 0.3. if and only if X(0) = 0. t) = X(x)T(t). These are constant coefficient ordinary differential equations. Assume for the moment that λ > 0. u (l. Then u(x. leads to T'(t) X"(x) = . and X" = λ X. so that heat can flow only along the bar. 0 < x < l. κ T(t) X(x) Since the variables x. each side of this equation can only be equal to a constant. X(l) = 0. The solutions are T(t) = Ce κλt. λ is real but could be positive. 0 < x < l. √ X(x) = Ae x √ λ + Be –x λ for A. X(x)T'(t) = κ X"(x)T(t). u(l. t) = 0. t > 0 5 Let the initial temperature along the bar at t = 0 be given by u(x. t) = 0. 0) = f(x). B leads to the condition √ e l λ – e –l √ λ = 0 or . C constants. consider a heat conducting bar of length l. Then T' = κλT. t > 0. Such a solution is called a separationofvariables solution. Substituting into the heat equation. That is 0 = X(0) = A + B √ 0 = X(l) = Ae l λ + Be –l √ λ. t appear on separate sides of this equation. t ) = X(x)T(t) satisfies u(0. Example 1. satisfies the heat equation ut – κ uxx = 0. 3. insulated along its length. To motivate Fourier series. t > 0. The temperature u(x. Assume a solution of the form u(x. t) = 0. and boundary conditions u(0. negative or zero. B.
0 = X(l) = Bl or A = B = 0. Therefore the nontrivial solutions of the heat equation ut – κ uxx = 0. t > 0. The boundary conditions u(0. Then √ λ = iµ and hence µ satisfies ei µ l – e –i µ l = 0 or sin (µl)= 0.. 2.. t) = n=1 bn e ∑ n=1 ∞ nπ x sin l also satisfies the heat equation and the boundary conditions. So λ cannot be positive.. n = 1. sinh 6 (l√ λ)= 0. and X(x) = A + Bx. We are mainly concerned wuth those functions f(x) which have this property. t) = e By superposition u(x. are – 0 < x < l. ∞ – κ n 2 π2 t l2 ∑ bn un(x. That is 0 = X(0) = A .. A series expansion such as .. n = ±1. . ±2. Separation of variables. For λ = 0. t>0 κ n 2 π2 t l2 un(x... Xn (x) = e –e = 2i sin l . For the initial condition to be satisfied u(x. u(l. n = ±1. t) = ∞ nπ x sin l . satisfying boundary conditions u(0. t) = 0. t > 0 u(l. are satisfied if and only if X(0) = 0. T' = 0... ±2. Suppose that λ < 0 and let λ = – µ2 for some real µ . inπ x κ n 2 π2 t inπx – – 2 l l nπ x l Tn (t) = e . That is we must show that f(x) can be represented as a series expansion of sines. and hence µn = l . There are no values of λ > 0 which satisfy this condition. 0 < x < l. nπ This has solutions µ l = nπ ... t) = 0. 2.3.. n = 1. X(l) = 0. and X" = 0 or T(t) = C.. t) = 0. 0) = f(x) = ∑ bn sin n=1 nπ x l . t) = 0. .
and X" = λ X 7 ∑ bn sin n=1 ∞ nπ x l and assuming for the moment that λ > 0. Suppose that λ < 0 and let λ = – µ2 for some real µ . t> . and X(x) = A + Bx. 2.. √λ (A – B). nπ This has solutions µ nl = nπ .. and hence µn = l .t) satisfies the same heat equation and initial condition but different boundary conditions ux (0. 2. t) = 0. u(l. √ e l λ – e – l √ λ = 0 or sinh (l√ λ)= 0. The boundary conditions u(0.. ux (l. Suppose the conducting bar is insulated at each end. .. inπ x κ n 2 π2 t inπx – – l l nπ x l2 Tn (t) = e . solutions T(t) = Ce κλt. Example 2. . X'(l) = 0. That is 0= 0 = X'(l). t) = 0. ±2. Xn (x) = e +e = 2 cos l . T(t) = C. t) = 0... Separation of variables.3. B again leads to the condition.. n = ±1.. t > 0. t) = 0. 0= √ √ λ (Ae l λ – Be – l √ λ ).. n = 1. n = ±1.. the temperature u(x. So λ cannot be positive. ±2. √ X(x) = Ae x √ λ + Be –x λ. nπ 2 λ n = – µn 2 = – l . The boundary conditions are satisfied if and only if 0 = X'(0). n = 1. λ ≠ 0.. 0. if and only if X'(0) = 0. There are no values of λ > 0 which satisfy this condition.. Eliminating A. f(x) = is called a Fourier series expansion. t ) = X(x)T(t) in the heat equation leads to the same ordinary differential equations T' = κλ T.. If λ = 0. 0 = X'(l). That is 0 = X'(0) = B . Then √ λ = iµ and hence µ satisfies iµ l – e – i µ l = 0 e or sin (µl)= 0. Separation of variables u(x.
satisfying boundary conditions ux (0. For the initial condition to be satisfied u(x.t) = n=0 an e ∑ n=0 ∞ nπ x cos l also satisfies the heat equation and the boundary conditions. . 2. ux (l.3. 0) = f(x) = ∑ an cos n=0 nπ x l . t > 0 are – 0 < x < l.t) = e By superposition u(x. or X(x) = A 0 .... . Separation of variables. t) = 0. T(t) = C0 8 Therefore the nontrivial solutions of the heat equation ut – κ uxx = 0. 0 < x < l. n = 0. t > 0 κ n 2 π2 t l2 un(x. This is also called a Fourier series expansion. 1. t) = 0. – κ n 2 π2 t l2 ∑ an un(x.t) = ∞ nπ x cos l ∞ .
n = 0. 2 a0 n = 1. the Fourier coefficients cn will be determined using the following orthogonality property of the complex exponentials e inx . ∞ 0 < x < 2π . . Suppose f(x ) is periodic with period 2π . P > 0 is called the period of f. π n≠m . n=m π Proof..4. π – ∫e π inx e – imx 0. e – ix = cos x – i sin x can be used to write the Fourier series expansion as f(x) = where the coefficients are cn = an – ibn . For n ≠ m. – ∫e π e – inx dx = – ∫ 1 dx = 2 π. f: R→R is called periodic if f(x + P) = f(x) for all x ∈ R... a0 The constant term is taken as 2 as a matter of convenience. Lemma. ±1. dx = 2π . π π inx If n = m. Fourier Series. A function f(x).. ±2. – ∫e π inx e – imx dx = – ∫e π i (n–m )x dx 1 = i(n – m) [e i (n–m )π – e –i (n–m )π ] 2 = (n – m) sin(n – m)π = 0.. . Fourier Series. 2 c–n = an + ib n . . The formulas e ix = cos x + i sin x. ∑ ∞ –∞ cne inx Assuming for the moment that the 2π periodic function f has a Fourier series expansion. 4. then an important question is whether f(x ) has a Fourier series expansion of the form a0 f(x) = 2 + 9 n=1 ∑ (an cos nx + b n sin nx ). . c0 = 2 . 2. π ¯ .
. f(–x) = f(x) for all x.. n = 1. . The first. ±1.4. c–n = an + ib n 2 . a0 2 + n=1 ∑ (an cos nx + b n sin nx ). .. i. Notice that if f(x) is an even periodic function. which are the Fourier coefficients of f(x). Since cn = an – ibn 2 . π Either of the expansions ∑ ∞ –∞ cne inx .. π π inx π 1 an = cn + c – n = 2 π – ∫ f(x) (e π – inx +e – inx ) 1 dx = π – ∫ f(x) cos nx dx π π – bn = i(cn – c– n ) = 2 π – i π ∫ f(x) (e π –e inx ) 1 dx = π ∫ f(x) sin nx dx. m = 0. ±2. . 2. c0 = 2 .. the exponential form and the second the trigonometric form. π π – ∫ f(x) e – imx dx = – π ∫ π inx – imx dx ∑ cne e – ∞ dx = 2 π c m .e.. . . ∞ is called a Fourier series expansion of f(x). then π 1 bn = π – ∫ π 1 f(x) sin nx dx = π – ∫ 0 π π 1 f(x) sin nx dx + π π 0 ∫ f(x) sin nx dx 1 = π 1 = π π ∫ 0 1 f(–x) sin(– nx) (–dx) + π 0 ∫ f(x) sin nx dx π 0 ∫ (–f(–x) + f(x)) sin nx dx = 0. ∞ = Hence ∑ ∞ π –∞ cn – ∫e π – i (n – m )x π 1 cm = 2 π – ∫ f(x) e π – imx dx . . Since f(x) = 10 ∑ ∞ –∞ cne inx . 2. Fourier Series. the Fourier cosine and sine coefficients are given by π 1 a0 = 2c0 = π – ∫ f(x) dx. a0 n = 1..
2 = π 0 ∫ f(x) cos nx dx. 0 1 Thus the Fourier series expansion of an odd function f(x) has only sine terms n=1 ∑ ∞ bn sin nx. 1.. . the coefficients a n π π 1 an = π – ∫ π 1 f(x) cos nx dx = π – ∫ π 1 f(x) cos nx dx + π π 0 ∫ f(x)cos nx dx 1 = π π ∫ 0 1 f(–x) cos(– nx) (–dx) + π π 0 ∫ f(x) cos nx dx 1 = π π 0 ∫ (f(–x) + f(x)) sin nx dx n = 0. Fourier Series. 1. 0 and the coefficients are π π 1 bn = π – ∫ π 1 f(x) sin nx dx = π – ∫ π 1 f(x) sin nx dx + π π 0 ∫ f(x) sin nx dx 1 = π π ∫ 0 1 f(–x) sin(– nx) (–dx) + π π 0 ∫ f(x) sin nx dx 1 = π π 0 ∫ (–f(–x) + f(x)) sin nx dx n = 1. 2. . = π ∫ (f(–x) + f(x)) sin nx dx = 0. then π 1 an = π – ∫ π 1 f(x) cos nx dx = π – ∫ 0 π π 1 f(x) cos nx dx + π π 0 ∫ f(x)cos nx dx 1 = π π π ∫ 0 1 f(–x) cos(– nx) (–dx) + π 0 ∫ f(x) cos nx dx n = 0.. .. the constant term in the Fourier series 1 2 = 2π – ∫ f(x) dx π . Thus the Fourier series expansion of an even function f(x) has only cosine terms ∞ a0 2 + ∑ an cos nx n=1 0 11 Furthermore.4.. 2. a0 π It is worth noting that for a periodic function f(x). 2 = π 0 ∫ f(x) sin nx dx . Similarly if f(x) is an odd periodic function. . . f(–x) = – f(x). 2. ...
. – π < x < π. 4 nπ . n even .28 2 3 4 5 6 12 This is called a squarewave function. Fourier Series. f(x) = sin x. – π < x < π.5 0. Let f(x) be periodic with period 2 π .4. π n=1 (b). n odd 4 .28 1.5 6. is the average value of f(x) over a period. – sin x. 0 < x < π 1.. 1. Example (a). Let f(x) be periodic with period 2 π. . The Fourier series of f(x) is (2n – 1) π n=1 ∑ ∞ bn sin nx..5 5 3 1 1 1 6. hence its Fourier cosine coefficients a n are all zero and π π ∫ f(x) sin nx dx = 2 = nπ [– cos n π + cos 0 ] = bn = 2 π 0 ∫ 1 sin nx dx 2 n nπ [– (–1) + 1 ] 2 π 0 = Therefore b 2n = 0. . – π < x < 0 f(x) = . b2n– 1 = 0. = ∞ n=1 ∑ ∞ b2n– 1 sin (2n – 1)x = ∑ n=1 (2n – 1) π 4 sin (2n – 1)x 4 ∞ 1 = ∑ 2n – 1 sin (2n – 1)x. It is obviously an odd function. – 1. f(x) = sin x. – π<x<0 0< x<π . n = 1. 2.
n odd 2 1 2 – n – 1 . n even π n + 1 0.28 3 1 1 2 3 4 5 6 0. π π an = 2 π 0 ∫ 1 2 f(x) cos nx dx = π 0 ∫ sin x cos nx dx π 2 = π ∫ 2 [sin(n + 1)x + sin((–n + 1)x) ]dx 0 π = 1 π 0 ∫ [sin(n + 1)x – sin(n – 1) x ]dx = 1 π cos(n + 1) π cos(n – 1)π 1 1 – + + n+1–n– 1 n+1 n–1 = 1 1 1 (1 – cos( n + 1) π ) – n – 1 (1 – cos( n – 1) π) π n + 1 = = 0. hence b n = 0 for all n. 4 .25 5 6.28 13 f(x) is an even periodic function. a2n– 1 = 0 and the Fourier series of f(x) is 2 4 ∞ 1 an cos nx = – ∑ 4n2 – 1 cos 2nx.5 1. Fourier Series.25 0. 1. n even (n2 – 1)π Therefore a 2n = – 4 ((2n) 2 – 1)π a0 2 + .75 0.4. – n odd . π π n=1 n=1 ∑ ∞ .5 6.
Using z z = z2 for complex z. Bessel's inequality can be written as .. π ]. c0 = 2 . π Proof. π ]. Using the equations cn = an – ibn .. Letting N→ ∞ leads to the result. Theorem. π 1 0 ≤ 2π – ∫ f(x)2 dx – π ∑ n=– N N π 1 cn 2 π – ∫e π inx f(x) dx – ∑ m=–N π N π 1 cm 2 π – ∫e π – imx f(x) dx + π ∑ ∑ n=– N m=–N N N N 1 cn cm 2 π – ∫e π i (n – m )x dx 1 = 2π – ∫ f(x)2 π dx – π n=– N ∑ cn cn – m=–N ∑ N cm cm + n=– N ∑ N cn cn 1 = 2π – ∫ f(x)2 dx – π ∑N cn2 n=– N for any N. Let f be 2π periodic and Riemann integrable on [– π . 2 c–n = an + ib n . . 5. 0 ≤ f(x) – = f(x)2 – ∑ cne n=– N inx N inx = f(x) – N 2 ∑ cne n=– N – imx N inx f(x) – N ∑ cme imx m=–N N N ∑ cne n=– N N f(x) – ∑ cm e m=–N f(x) + ∑N m=–Ncn cm e i (n – m )x ∑ n=– Dividing by 2π and integrating over [–π. Then 14 n=– ∞ ∑ ∞ π cn2 1 ≤ 2π – ∫ f(x)2 dx. . Bessel's Inequality. 2.5. Bessel's inequality. 2 a0 n = 1. ¯ Bessel's inequality will later be shown to be actually an equality but for now it implies that the series n=– ∞ ∑ ∞ cn2 converges where c n are the Fourier coefficients of the Riemann integrable function f.
π This implies that the series ∑ n=1 ∞ an2 . ∑ n=1 ∞ b n2 also converge. .5. bn are the Fourier cosine and sine coefficients of f. Bessel's Inequality. where an. 15 n=– ∞ ∑ ∞ cn2 = a02 4 + n=1 ∑ ∞ ∞ a + ib 2 an – ibn2 n n + ∑ 2 2 n=1 1 = 4 +2 a02 n=1 ∑ [an2 + b n2] π ∞ 1 ≤ 2π – ∫ f(x)2 dx.
x k . Then n=– π f SN(x) N π = – ∫ π D N(x – y)f(y) dy = – ∫ π D N(y)f(x – y) dy .. . b] and the lefthand and 1 righthand limits of f exist at each of the points x . π ]? A function f is called piecewise continuous on an interval [a . .6. Convergence results for Fourier series. . f is called piecewise smooth if f and its derivative f' are piecewise continuous on [a. Consider the N th partial sum of the complex Fourier series of f. n=– N 1 We define D N(x) = 2 π ∑N e inx. x 2. b]. [The last equality follows from a change of variable and the periodicity of the integrand DN(x – y)f(y)]. x k ∈ [a. b]. f SN(x) = ∑N cne inx n=– N = n=– N ∑ π N 1 π e – iny f(y) dy e inx 2π – ∫ π = – ∫ π 1 2π ∑N e in (x – y ) f(y) dy. the concept of convergence must be made precise. b] is denoted by PC[a. indeed can we have convergence at some points and not at others and if so. x 2... 6. . Convergence results for Fourier series.. π]? Can convergence be different at different points x. The function DN(x) is called the Dirichlet kernel and 1 DN(x ) = 2 π e – iNx ∑ n=0 2N e inx e 1 = 2 π e – iNx i (2 N + 1) x e ix –1 –1 . a0 2 + n=1 ∑ (an cos nx + b n sin nx ). Do we mean the numerical series converging at every x ∈ [– π . ∞ Because we are dealing with functions. b] if it is continuous everywhere except at finitely many points x . We will consider the question: For what functions f do the Fourier series 16 ∑ converge? ∞ –∞ cne inx . The set of all piecewise 1 continuous functions on [a. which are the points of convergence? Can we have some sort of average convergence on [– π .
2 x sin (2 ) The question of convergence of the Fourier series reduces to question: f Does SN(x) converge as N → ∞ for x ∈[– π. 1 Proof.6. π – ∫ 0 π D N(x) dx = ∫ 0 n=1 D N(x) dx = 2 . Lemma. f SN(x) 1 –2 π [ f(x–) + f(x+ ) ]= ∫ – π D N(y)f(x – y) dy – 2 [f(x–) + f(x+ ) ] 1 = ∫ 0 D N(y)[f(x – y) – (f(x–)]dy + – ∫ 0 π D N(y)[f(x – y) – f(x+ )]dy (from lemma above) . – ∫ 0 π π π D N(x) dx = ∫ 0 D N(x) dx = N ∫ 0 1 1 2π + π π ∑ n=1 N cos nx dx x 1 = 2π + π n=1 ∑ sin nx 1 = . Convergence results for Fourier series. N→ ∞ 1 π Proof. n 0 2 ¯ Theorem. π]? If so. DN(x) = 2 π + 2 π 1 1 ∑N e inx = 2π + π n=– 1 n≠0 N 1 ∑ N cos nx From the evenness of DN(x) and integrating. to what does it converge? We prove a preliminary result. Hence lim SN(x) = f(x) for each point x of continuity of f. Then f lim SN(x) = 2 [f(x–) + f(x+ ) ] N→ ∞ f for every x. Let f be a piecewise smooth 2 π periodic function on R. e 1 = 2π i (N + 1) x 17 – e – iN x ix e –1 1 i (N + 2)x – e – i (N + 2)x 1 e = 2π ix ix – 2 2 1 e –e 1 sin = 2π (N + 1) x.
A N → 0 as N → 2 cos 2 ∞. y which is an odd piecewise continuous function for [– π . Convergence results for Fourier series. . π 18 = 0 ∫ D N(y)[f(x – y) – f(x–)]dy + π π ∫ 0 D N(–y)[f(x + y) – f(x+ )](–dy) = π 0 ∫ D N(y)[f(x – y) – f(x–) + f(x + y) – f(x+ )]dy = 0 ∫ 1 sin 2π (N + 1) y [f(x – y) – f(x–) + f(x + y) – f(x+)]dy 2 y sin (2 ) )y y f(x – y) – f(x–) 2 y y sin (2 ) + + f(x + y) – f(x+ ) dy y π 1 =π ∫ 0 sin ( 1 N+2 For fixed x define y f(x – y) – f(x–) 2 g(y) = y y sin (2) f(x + y) – f(x+ ) .6. () () Therefore f SN(x) – 2 [f(x–) + f(x+ ) ]= BN + A N → 0. 2 sin 2 g(y) respectively. as N → ∞. BN . By Bessel's inequality. AN of y y 1 1 g(y). that the partial sums of the Fourier series of a piecewise smooth 2 π−periodic function converges pointwise to the mean of the left and right hand limits at x. then the partial sums converge pointwise to f(x). 1 ¯ This result. Then π f SN(x) 1 –2 [ f(x–) + f(x+ ) y ] 1 =π ∫ 0 sin π (N + 1) y g(y)dy 2 y 1 =π ∫ sin (Ny){ cos (2) g(y) } 0 π 1 dy + π ∫ cos (Ny){ sin(2) g(y) } dy 0 The last two terms are the Fourier sine and cosine coefficients BN . π ] by the condition of piecewise smoothness on f. If x is a point of continuity of f.
Fourier series provide a useful method for summing certain numerical series. the Fourier series converges at x = 0 to f(0) = 0. The Fourier series of the continuous periodic function f(x) =sin x. 2 n=1 At x = 2 . 2 4 ∞ 1 0= – ∑ 4n2 – 1 π π n=1 or ∞ 1 ∑ 4n2 – 1 = 1 . 2 –1=– 4 4n . is 2 4 ∞ 1 – ∑ 4n2 – 1 cos 2nx π π n=1 1. the Fourier series converges to 2 4 ∞ 1 π f 2 =1= – ∑ 4n2 – 1 cos(n π) π π n=1 π () 2 4 ∞ (–1)n = – ∑ π π n=1 4n2 – 1 n=1 ∑ ∞ (–1)n π – 2 .28 3 1 1 2 3 4 5 6 0. 19 Example.5 6.75 0. x ∈ [– π .5 1.6.25 5 6. π ]. Convergence results for Fourier series.25 0.28 Because x = 0 is a point of continuity of f.
Let an . 20 Fourier series can be differentiated termbyterm but the question is does the resulting series converge and if so. f(x) = 1 is periodic but its antiderivative F(x) = x is not.7. b n. then for all x. c n and let F(x) = ∫ 0 x f(t) dt. n Proof. A similar proof works for a' = nbn. piecewise smooth. Theorem. b' . 7. to what does it converge? Similar concerns apply to the series resulting from the termbyterm integration of a Fourier series. π bn = –nan. Let f be 2 π periodic. then the series above 1 converge to 2 [f'(x–) + f'(x+ )]. ' c' = inc n. Then the Fourier series of f' is ∑ ∞ –∞ incne inx = n=1 ∑ (nbn cos nx – nan sin nx ) ∞ and converge for each x where f'(x) exists. ¯ Integration of Fourier series is not so straightforward since the antiderivative of a periodic function need not be periodic. This result follows by combining the previous two theorems. the following result is true. with piecewise smooth derivative f'. Differentiation and Integration of Fourier Series. However. Then n n n a' = nbn. n 1 c' = 2 π n – ∫ f'(x) e π – inx dx 1 = 2 π f(x)e– inx π π –π π 1 – 2π – ∫ f(x) (–ine π – inx )dx 1 = in 2 π – ∫ f(x) e π – inx dx = incn. For example. Differentiation and Integration of Fourier Series. n bn = –nan. 1 F(x) = C0 + n=– ∞ n≠0 ∑ ∞ c n in e inx = 2 + A0 ∑ n=1 ∞ – bn an cos nx + n sin nx n . Theorem. c' be the Fourier coefficients of f'. Let f be 2πperiodic. since all but the constant term of a Fourier series has a periodic antiderivative. By integration by parts. If c 0 = 2 a0 = 0. piecewise continuous with Fourier coefficients an. If f' is not continuous at x. Proof. ' ¯ Theorem. b n . c n be the Fourier coefficients of a 2 πperiodic piecewise smooth function f and a' .
is continuous with piecewise smooth derivative. bn = –nAn . C n are the Fourier coefficients n of F. an Bn = n . f(x) = x( π – x). π n=1 (2n – 1) 3 By the above results. ¯ Integrating and differentiating known Fourier series using the above results is a useful way of obtaining new Fourier series. bn. π ]. cn are the Fourier coefficients of f and A . cn = inC n . B n. F(x) = If c 0 = 0. π] to F(x). where an. The Fourier series is by the table #17 8 ∞ 1 f(x) = x( π – x) = ∑ sin (2n – 1)x. 0 ∫ x f(t) dt is continuous. F(x) is 2 π periodic since x+2π –π ∫ π F(x) dx the average value of F on [– π . x ∈ [– π. 1 The constant C0 = 2 = 2 π A0 F(x) = 0 ∫ x –π ∫ π F(x) dx is the constant term in the Fourier series of f(t) dt. Since f(x) is piecewise continuous. Example. this implies that bn An = – n . then the series above converges to F(x) – c0x. π ] and therefore its zeroth Fourier coefficient is 1 2π –π ∫ π (f(x) – c0)dx = 1 2π –π ∫ π f(x) dx – c 0 = c0 – c 0 = 0. Differentiation and Integration of Fourier Series. If c 0 ≠ 0. an = nB n . f(x) – c0 has zero mean value on [–π .7. x+2π F(x + 2 π) – F(x) = 0 ∫ f(t) dt – 0 ∫ x π –π f(t) dt = x ∫ f(t) dt = ∫ f(t) dt = 2 π c0 = 0. f'(x) = π – 2x = 8 ∞ 1 ∑ (2n – 1) 2 cos (2n – 1)x π n=1 . Applying the result just obtained to f(x) – c0 and its antiderivative F(x) – c 0x completes the theorem. cn Cn = in . If n ≠ 0. Proof. 1 where C 0 = 2 = 2 π 21 A0 If c 0 ≠ 0. The 2π periodic function f :R→R. Therefore its Fourier series converges at each x to f(x). π ]. Therefore the Fourier series of F converges pointwise at each x ∈ [– π . Since f(x) = F'(x). by the two previous theorems.
But F(x) – 2 = Therefore x – π<x<0 0<x<π x π 2 x x = 2 2 = 4– π ∑ 2 n=1 (2n – 1) ∑ 2 n=1 (2n – 1) ∞ ∞ 1 cos (2n – 1)x or x = 2 – π π 4 1 cos (2n – 1)x. – π < x < 0 f(t) dt.7.25 5 6. 0.75 0. Example. = .25 0. f(x) = 1. 2. or x = 2 – π which agrees with table #2. x. ∑ 2 n=1 (2n – 1) ∞ 1 cos (2n – 1)x. F(x) = 1 2 ∑ 2n – 1 n=1 ∞ 1 sin (2n – 1)x 0 ∫ x 0. – π<x<0 0<x<π 1. Differentiation and Integration of Fourier Series.5 6. This result agrees with the previous example . Since c 0 = 2 ≠ 0.5 1.28319 22 π 4 ∑ 2 n=1 (2n – 1) ∞ 1 cos (2n – 1)x is piecewise continuous with Fourier series f(x) = 2 + π 1 (table #7). 0 < x < π Therefore ∞ x A0 1 2 F(x) – 2 = 2 – π ∑ 2 cos (2n – 1)x n=1 (2n – 1) Since A0 1 2 = 2π ∫ –π π x π π 2 F(x) dx = 4 . F(x) – 2 = 4 – π x –2.28319 3 1 1 2 3 4 5 6 0.
is piecewise smooth and has Fourier series ∞ a0 2 + ∑ an cos nx n=1 where π 2 an = π 0 ∫ f(x) cos nx dx.8. Let f(x) = sin x. = sin x – π<x<0 0<x<π That is f(x) = . Halfrange Fourier series. 0] as an even function. x ∈ [–π . Halfrange Fourier series. Similarly. sin x. Then the periodic extension to R . If f(x) is piecewise continuous on [0. x ∈ [0. Then f is periodically extended to the whole real line as a 2π periodic function by f(x + 2 π) = f(x). π ]. the extension has Fourier series n=1 ∑ 2 π ∞ bn sin nx π where bn = a0 0 ∫ f(x) sin nx dx. 0] as either an even function or as an odd function. π ]. x ∈ R. Extend f to [–π . π]. 0] as an odd function and periodically to R. 23 It is often convenient to represent a given function as a Fourier series which contains only cosine terms or only sine terms. n=1 ∑ bn sin nx are known as halfrange series for Example. 0]. it can be extended to [–π . π ] and extended to [–π . sin(–x) = – sin x . ∞ The Fourier series 2 + f(x). n=1 ∑ ∞ an cos nx. f(x) = f(–x). Let f(x) be a piecewise smooth function given on the interval [0. as in the initial examples. extending f(x) to [– π . 0] as an even function. x ∈ [0. That is. 8.
75 0.14159 1 1 2 3 0. 1.14159 24 The halfrange Fourier series of f is given by 2 4 f(x) = π – π ∑ 2 n=1 2n – 1 ∞ 1 cos 2nx (table #8).25 2 3.25 0.5 1.8. . Halfrange Fourier series.5 3.
9. 9. General Intervals. π x f(x) = φ l = a0 ∑ –∞ cn e = 2 + ∑ an cos n=1 ∞ nπ x nπ x l + bn sin l is the Fourier series of a 2lperiodic function f. nπξ l dξ ∫ f(ξ) cos 0 . –l l 1 c n = 2l Therefore – ∫ l l π y – φ l e inπ y l 1 dy = 2l ∞ – ∫ f(y) e l inπ x l l – inπ y l dy. 1 bn = l – ∫ l l π y nπ y nπ y 1 φ l sin l dy = l ∫ f(y) sin l dy. results in the Fourier series ∞ ∞ a0 φ(x) = ∑ cne inx = 2 + ∑ (an cos nx + b n sin nx ). Applying the earlier results to φ. Then π l(x + 2 π) lx φ(x + 2 π) = f = f π + 2 l = π lx f = φ(x) π 25 So φ has period 2π . General Intervals. The change of variable y = an = l 1 lx leads to π – ∫ l π y φ l cos nπ y 1 l dy = l – ∫ l nπ y f(x) cos l dy. π l 1 c n = 2π – ∫ φ(x) e π – inx dx. its Fourier series contains no sine terms and is of the form a0 with an = 2 l 2 + ∑ an cos n=1 l ∞ nπ x l . The theory of Fourier series can be expanded to include functions which have arbitrary period. That is f(x + 2l) = f(x). x ∈ R. Let f :R→R have period 2l > 0. It follows from the above calculations that if f(x) is an even function on [–l. π l 1 bn = π – ∫ φ(x) sin nx dx. Also define φ :R→R by lx φ(x) = f . l]. –∞ n=1 where π π π 1 an = π – ∫ φ(x) cos nx dx.
l] either as an even or an odd function and the extended periodically with period 2l to the real line. The Fourier series of such functions are halfrange expansions on [0. nπξ l d ξ. l] and contain no sine terms in the case of an even extension or no cosine terms for an odd extension. l]. General Intervals. ∫ f(ξ) sin 0 A function f(x) defined on an interval [0. Likewise if f(x) is an odd function on [–l. .9. l] can be extended to [–l. its Fourier series contains no cosine terms and is of the form 26 ∑ bn sin n=1 with bn = 2 l l ∞ nπ x l .
0) = 0. . X(a) = B sin (a√ λ)= 0. 0 ≤ y ≤ b. Application to Laplace's equation. y) = u(a. Application to Laplace's equation. Let Ω be a rectangle 0 ≤ x ≤ a. requires that . u(x. and X(x) = B sin a . Y" – λY = 0.10. Y (y) = C cosh√ λ y + D sinh√ λ y . y) = X(x)Y(y) and substituting into Laplaces equation. and consider the boundaryvalue problem 27 ∂2u ∂2u + = 0. u=f y=b u=0 u=0 u=0 x=a Assuming a separation of variables solution of the form u(x. hence X(x) = A cos √ λ x + B sin√ λ x . 0 ≤ y ≤ b ∂x 2 ∂y 2 u(x. 0 ≤ x ≤ a u(0. y) = 0. To satisfy the boundary conditions u(0. y) = u(a... Therefore a √ λ = nπ . To also satisfy the boundary condition u(x. n = 1. b) = f(x). 0 ≤ x ≤ a. 10. 0 ≤ y ≤ b. X" + λ X = 0. y) = 0. 2. Example 1. 0) = 0. nπ 2 nπ x λ = a . . 0 ≤ x ≤ a. 0 ≤ y ≤ b. requires that X(0) = A = 0. X" Y + X Y" = 0 or X" Y" – X = Y = constant = λ For λ > 0.
.. where Bn are the constants BD for each n = 1. let λ = – µ. X(a) = Ba = 0. nπ y Y(0) = C = 0 and therefore Y (y) = D sinh a . ( )≠ 0 for λ < 0.. Y (y) = C cos √ µy + D sin√ µy . y) = 0. requires that X(0) = A = 0. y) = ∑ Bn u n(x. Then X" – µX = 0. By superposition.. 2.. a solution is also given by ∞ u(x. for an arbitrary constant BD. b) =f(x). hence X(x) = A cosh√ µx + B sinh √ µx . hence X(x) = A + Bx. y) n=1 = ∑ Bn n=1 ∞ nπ x nπ y sin a sinh a . implies that . So there is only the trivial Finally suppose that λ = 0.. Summarising thus far. . y) = u(a. Y(y) = C + Dy. y) = u(a. The boundary condition u(x. The separation of variables solutions for λ > 0 are nπ x nπ y u(x. n = 1. But B = 0 gives a trivial solution and sinh a√ λ separation of variables solution u = 0. n = 1. 0 ≤ y ≤ b. y) = X(x)Y(y) = BD sin a sinh a . there is a sequence of nontrivial solutions of the form nπ x nπ y un(x. Y " + µY = 0. 0 ≤ x ≤ a. then X " = 0. Y" = 0. For λ < 0. . That is. requires that X(0) = A = 0. 2. 2. A = B = 0 and the only separation of variables solution for λ = 0 is the trivial solution. for λ < 0. X(a) = B sinh 28 (a√ λ)= 0. . To satisfy the boundary conditions u(0. To satisfy the boundary conditions u(0. Application to Laplace's equation. . y) = 0. 0 ≤ y ≤ b.10. . y) = Bn sin a sinh a . µ > 0. .
0 ≤ x ≤ a.10. ∂r2 r ∂r r2 ∂θ2 subject to boundary conditions u(a. radius a. and Laplaces' equation in polar coordinates. θ) = ∂2u 1 ∂u 1 ∂2u + + = 0. u= a . then we can break up the problem into four subproblems each similar to the one we have just solved. nπ x nπ b sin a sinh a . b) = f(x) = 29 ∑ Bn n=1 ∞ 2 nπξ nπ b bn = ∫ f(ξ) sin d ξ = Bn sinh a . ∆u(r. Application to Laplace's equation. It is natural to choose Ω a circle centre the origin. where n=1 u(x. ∞ nπ x Since φ (x). Example 2. satisfying Dirichlet boundary conditions. has a Fourier halfrange series f(x) = ∑ bn sin a . where φ is a continuous function on 0 ≤ θ < 2 π . 0 ≤ θ < 2 π. Each of these subproblems will have zero boundary conditions on three of the four sides and can be solved as above. 0 ≤ θ < 2 π. a0 a Therefore Bn = bn nπ b sinh a u(x. y) = = 2 a a sinh a ∞ ∫ f(ξ) sin nπ b 0 a nπξ a dξ ∑ Bn n=1 nπ x nπ y sin a sinh a . r < a. Suppose all four sides of the rectangle have nonzero boundary conditions. θ) = f(θ). We seek a solution to Laplaces' equation on a circle. Then the solution to the boundaryvalue problem is the sum of the four subproblems by superposition.
then the solutions of the above equations are Θ = A 0 + B 0θ. r r 30 Θ" r2R" + rR ' =– = constant = λ2 (say). f( θ ) = u(a. we conclude that λ = n = 1.3. R = C 0 + D0 log r Since solutions u must be periodic in θ..... Again because solutions are periodic in θ with period 2π. Θ " + λ 2Θ = 0. we conclude that B0 = 0. Dn = 0. D 0 = 0. θ) = R(r) Θ(θ ) and substitute into the differential equation to obtain R" Θ + or separating variables. assume that u(r. θ) = At the boundary r = a. R Θ which leads to the two families of ordinary differential equations r 2R" + rR' – λ 2R = 0.. Using separation of variables.10. hence u(r. n = 1. By superposition we can combine these solutions to obtain u(r. Application to Laplace's equation. θ ) = n=0 ∑ ∞ rn (A an n cos nθ + B n sin nθ ) ) n=0 ∑ ∞ (A n cos nθ + B n sin nθ . / In the case λ = 0 the solutions of the above equations are Θ = A λ cos λθ + B λ sin λθ R = C λ rλ + Dλ r –λ where the coefficients depend on λ .. 1 ' 1 R Θ + 2 RΘ " = 0. and because solutions are bounded for r ≤ a.3.2. Consider the case λ = 0. and bounded for r ≤ a. θ) = n=0 ∑ ∞ Cnrn (A n cos nθ + B n sin nθ ) We may as well incorporate the constant Cn into A n and Bn.2.
10. θ ) = 2 + a0 an a n . θ ) = 2π 2π n a2 – r 2 r cos n(t – θ) = a a2 – 2arcos(t – θ) + r 2 n=1 ∑ ∞ ∫ f(t) 0 a2 – r 2 dt. r < a. 2π 2π ∞ 1 1 r n u(r. we obtain that An = and hence u(r. We can get a closed form expression for u by substituting the formulae for the Fourier coefficients as follows. θ ) = 2π 2π Therefore ∫ 0 ∞ rn f(t) 1 + 2 ∑ cos n(t – θ) dt. in these two expressions for f(θ). θ ) = f(t) dt + ∑ f(t) ( cos nt cos nθ + sin nt sin nθ )dt 2π ∫ π n = 1 a 0∫ 0 1 1 = ∫f(t) dt + π 2π 0 2π 2π rn ∑ a 0∫f(t) cos n(t – θ) dt n=1 2π ∞ 2π 1 1 = ∫f(t) dt + π0∫ f(t) 2π 0 n=1 ∑ ∞ r cos n(t– θ) dt a n where the interchange of the order of summation and integration is allowed by the uniform convergence of the series. bn = π ∫f(t) sinnt dt 0 0 are the Fourier coefficients of the function f. Application to Laplace's equation. sin nθ. 2 cos x = eix + e –i x the term in the square brackets reduces to two infinite geometric series whose sum is 1 +2 (exercise). Therefore 1 u(r. Equating coefficients of cos n θ . a2 – 2ar cos(t – θ) + r 2 . 1 u(r. a0 = 2 + where 1 an = π 2π 31 n=1 ∑ ∞ (a n cos nθ + b n sin nθ 2π ) 1 ∫f(t) cosnt dt . a n=1 Using the exponential form. Bn = bn an n=1 ∑ ∞ r a n (a n cos nθ + b n sin nθ ) This series converges uniformly for 0 ≤ θ < 2 π .
Application to Laplace's equation. 32 . which is Poissons' integral formula for the circle.10.
e – inx ... w (x) be real continuous functions on the interval [a . } is said to be a basis for L2 [– π. Let α0 . x ∈ [a. π ] consisting of π functions f :[–π . 2. The set {e inx . n = 1.11. n = 0.. cos n x. w(x) ≥ 0 on [a. . einx.. Then we say that f has the expansion f = n=1 ∑ ∞ cnφn.. be real constants. } consisting of functions which are solutions of a boundaryvalue problem for certain second order linear ordinary differential equations. u(– π ) = u( π). n = 0. n = m then the coefficients cn are given by π cn = – ∫ f(x) φn(x) dx. A SturmLiouville problem is a boundary value problem – (p(x)u' )' + q(x)u = λ w(x)u. β0. SturmLiouville problems and orthogonal functions. n ≠ m φn(x) φm(x) dx = 1. b] . n = 1. α1 . π ] .. are examples of orthogonal functions on [– π. π ] Their orthogonality properties follow from the fact that they solutions of linear second order ordinary differential equations. there is a unique set of scalars c n ... . such that lim ∫ N→ ∞ – π f(x) – n=1 ∑ N 2 cnφn(x) dx = 0. SturmLiouville problems and orthogonal functions.. For example sin nx . π ] if for any function f ∈ L2 [– π. b] in terms of an orthonormal basis {φ n(x).. 1. ±1.. The interval [– π. . } is orthonormal. b]. . Furthermore. q(x) ≥ 0. π ] can be replaced by [a. 33 The functions sin nx . cos nx. β1. 2. Let p(x) be continuous and p(x) > 0. that is π – ∫ π 0. 2. b]. . q(x). Let p (x ). ±2. 11. 2 n = 1.. π n = 1.. u'(–π ) = u'(π) for λ = n 2. e inx satisfy – u"(x) = λ u(x). b] and the theory of Fourier series generalised to the expansion of arbitrary functions f ∈ L [a. if the set { φn(x). } form a basis for the vector space L2 [– π. . 2. π ] →R for which – ∫ f(x)2 dx < π π ∞. n = 1. . 2. A set of functions {φn(x)... 2.. The ordinary differential equation together with the boundary conditions is called a boundaryvalue problem.
The value of λ for which this boundaryvalue problem has nontrivial solutions are called eigenvalues of the boundaryvalue problem and the corresponding solutions u(x) are called eigenfunctions.11. g(x) be C 2 functions on [a. } and the corresponding eigenfunctions { φn(x)... SturmLiouville problems and orthogonal functions.. (f' g – f g' ) . 2. A set of functions {φn(x). a b . N→ ∞ a n=1 Let the linear SturmLiouville differential operator L be defined as 1 L[f] Ê= w(x) [–(p(x)f')' + q(x)f ]. } are orthonormal with respect to the weighted innerproduct <f . b] . such that b N 2 lim ∫ w(x) f(x) – ∑ cnφn(x) dx = 0. n = 1. 2. 2.. n = 1. β 0u(b) + β 1u'(b) = 0. It can be shown that the eigenvalues form a countable set { λn. Then <L[f] . = –p(x) <L[f] . n = 1. . b] →R for which 2 f 2 = ∫ w(x)f(x)2 dx < a ∞. g >w b = a ∫ w(x) f(x) g(x) dx... Let f(x).. b]. a ∫ (p(x)f' g' + q(x)f g )dx. there is a unique set of scalars c n .. 34 α 0u(a) + α 1u'(a) = 0.. b] consist of functions f :[a. g >w = b a ∫ [(p(x)f')' + q(x)f ] g(x) dx a b b = –p(x)f' g + a Also ∫ b (p(x)f' g' + q(x)f g )dx [(p(x)g' )' + q(x)g ] dx <f . L[g]>w This is Lagrange's identity. b Let the vector space Lw[a. n = 1. b] if for any function f ∈ L 2 [a. 2. .. g >w – <f . L[g]>w = b a ∫ f(x) b = –p(x)f g' + a Subtracting these two expressions. } is said to be a basis for w w L 2 [a.
L[g]>w = –p(x) (f' g – f g' ) a b β1 β1 = –p(b) f'(b) – g'(b) – – f'(b) g'(b) β0 β0 α1 α1 + p(a) f'(a) – g'(a) – – f'(a) g'(a) α0 α0 = 0. The eigenvalues of the SturmLiouville operator L are real. φ>w <φ . β 1 ≠ 0 or α1 ≠ 0. Since φ ≠ 0. SturmLiouville problems and orthogonal functions. <L[f] . φ>w = <λφ . Proof. λ = λ and λ is real. λφ>w = = λ <φ . L[g]>w Proof. g are C2 functions on [a. g >w = <f . Let λ be an eigenvalue with corresponding eigenfunction φ . β 0g(b) + β 1g'(b) = 0. φ>w 2 = λ φw . β 0 ≠ 0 or α0 ≠ 0. If f. <L[f] . g >w – <f . β 1 ≠ 0. ¯ . then by Lagrange's identity. then α 0g(a) + α 1g'(a) = 0. β 0f(b) + β1f'(b) = 0. L[ φ ]>w <φ . Lemma. β 0φ(b) + β1φ'(b) = 0. 2 (λ – λ )φw = 0. Then φ satisfies 1 L[ φ] = w(x) [–(p(x) φ')' + q(x) φ ] = λφ . b] which satisfy 35 α 0f(a) + α1 f'(a) = 0. α 0φ(a) + α 1φ'(a) = 0. Similarly if α1 ≠ 0. Suppose α0 ≠ 0. . ¯ Lemma. φ >w (above lemma) = = Therefore <L[ φ] . Then 2 λ φw = λ <φ . the result follows with minor changes to the argument.11. β0 ≠ 0.
. The eigenfunctions of a regular SturmLiouville problem are a countable set {φn(x). Let φ(x). . µ respectively. That is. ψ (x) be eigenfunctions of the SturmLiouville boundaryvalue problem with corresponding eigenvalues λ. b]. } and form an orthonormal basis for Lw[a. N→ ∞ n=1 2[a. . each function f ∈ 2 Lw[a. n = 1.. . If f(x) ∈ C 0 1 β 0f(b) + β1f'(b) = 0. the series f(x) = n=1 ∑ ∞ cnφn (x) converges uniformly on [a.. 36 Lemma. or 2 lim f – ∑ cnφn w = 0. Then If λ ≠ µ. and convergence is in the sense that N→ ∞ a lim ∫ b w(x) f(x) – N n=1 ∑ N 2 cnφn(x) dx = 0. Since λ ≠ µ . ψ>w = 0.. Proof. L[φ] = λφ α φ(a) + α φ'(a) = 0 β φ(b) + β φ'(b) = 0 0 1 0 1 L[ψ] = λψ α ψ(a) + α ψ'(a) = 0 β ψ(b) + β ψ'(b) = 0 0 1 0 1 From a previous lemma. <L[ φ] . ψ(x) are orthogonal on [a. That is lim max f(x) – N→ ∞ x ∈[a. . φ(x). ψ>w = 0. ψ >w = <φ . g >w. SturmLiouville problems and orthogonal functions. . ¯ These results can be used to prove the following result: Theorem. φ > n w . ψ>w = <φ . b] and satisfies the boundary conditions α f(a) + α f'(a) = 0. 2. . n = 1. b] has a series expansion f = ∑ cnφn where the cn = 2 n=1 ∞ <f . b] with respect to the weighted inner product <f . <φ . b] ∑ n=1 N cnφn(x) = 0. 2. µψ>w (λ – µ )<φ . L[ ψ ]>w <λφ . b].11.
Θ" + ν2 Θ = 0. q(r) = r > 0.12. κT rΘ T' + λ 2κ T = 0. r Θ" R' T' R" = R + rR + 2 = constant = – λ2 (say). dρ dρ ρ where S( ρ ) = R(r) = R . θ . We can write it in the SturmLiouville form. t) is the temperature. t) = R(r) Θ(θ) T(t). t > 0. w(r) = r ≥ 0. 0 ≤ θ < 2 π. substituting in the heat equation. 12. ν2 – (rR' )' + r R = λ 2rR. be u(r. Θ" r2 R" rR' + R + λ 2r2 = – = constant = ν2 (say). Bessel Functions. and boundary condition at r = a. θ. The equations for T. 0 ≤ θ < 2π . Bessel Functions. θ . 1 1 R Θ T' – κ R" ΘT + r R' ΘT + 2 RΘ"T = 0. 0 ≤ θ < 2 π . a Bessel function of the second kind. 37 ∂u ∂u ∂2u 1 ∂u 1 ∂2u – κ ∆u(r. κ > 0 a constant. Θ are familiar but the equation for R is Bessel's equation and has nonconstant coefficients. Assuming a solution of the form u(r. t > 0. We denote a solution of this Bessel's equation as S( ρ ) = J ν(ρ ) λ and then a solution of the original form of Bessel's equation is R (r) = S( λ r) = J ν(λ r). A change of variable ρ = λ r in Bessel's equation results in d2 S dS ρ2 2 + ρ + ( ρ 2 – ν 2)S = 0. 0 ≤ θ < 2 π. u(a. ∂r ∂t r ∂θ 0 ≤ r < a. Let the initial temperature at t = 0. Consider the heat equation on a disc 0 ≤ r < a. R Θ r2R" + rR' + ( λ 2r2 – ν 2)R = 0. A second solution of Bessel's equation is Y ν (λ r). θ). where u(r. and therefore the general solution is R(r) = AJ ν(λ r) + BY ν(λ r) . . t) = 0. 0) = f(r.. 0 ≤ r < a. ν2 where p(r) = r ≥ 0. θ. θ) = ∂t – κ 2 + r ∂r + 2 2 = 0.
. . }. . The Bessel functions J ν(λ r) are oscillatory for r > 0. r R(0) = J ν(0) = 0..12. <f . .. B. m = 1. .. . For Θ(θ) to have period 2 π.. J –ν (λ r) is a linearly If ν ≠ 0. 2. 0 ≤ θ < 2π . . for each ν > 0. . Jν (λ r) .4. Jν (λ r) = ( λr) 1 – 2(2 + 2ν) 2. θ . m = 0. . For R(r) to be bounded at r = 0... 1. Then R(r) = Jν (λ r) solves the SturmLiouville problem ν2 – (rR'(r))' + R(r) = λ 2 rR(r). ν = n = 0.. J –ν (λ r) are unbounded as r → ∞. hence the only bounded solutions R(r) occur when B = 0.. J 0 (0) = 1. The general solution is then 38 { } R(r) = AJ ν(λ r) + BJ –ν (λ r) for constants A. For ν > 0. 2. Returning to the heat equation on the disc 0 ≤ r < a. ρ νmr ∞ The Bessel functions Jν a m=1 . The solution Jν (λ r) is called a Bessel's function of the first kind and a series representation can be found λ 2r2 λ 4r4 ν + + . 1. let the zeros of Jν (ρ ) be {ρ νm . the solutions of the form u(r. then J –ν (λ r) is also a solution and independent set of solutions to Bessel's equation. if m ≠ l. Y ν(λ r). B = 0. if λ a = ρν m . ∫ a ρ νmr ρ νlr r J ν a Jν a dr = 0.. a). 1. with respect to the weighted inner product R(a) = Jν (λ a) = 0. Θ(θ) = C cos (νθ ) + D sin (νθ ). 2 T(t) = E e –λ κt . 2. g >w 0 = 0 ∫ r f(r) g(r) dr. t) = R(r) Θ(θ) T(t) are given by R(r) = AJ ν(λ r) + BY ν(λ r). are orthogonal on the interval (0. Bessel Functions... Jν(0) = 0. a That is. where ρν 0 = 0 for all ν > 0.(2 + 2 ν)(4 + 2ν) For ν > 0.
n = k and m = l 0 π 0 –π ρ nmr ρ kl r r Jn cos(n θ) Jk a sin(kθ ) d θ dr = 0 for all n. 1. is given by u(r. n = k and m = l 0 π 0 –π ρ nmr ρ kl r r Jn sin(nθ ) Jk a sin(k θ) d θ dr a = ∫ ∫ a 0.. θ) d θ dr. Bessel Functions. For example. t) = n=o m=1 ∑ ∑ ∞ ∞ 2 ρ nm κ t ρ nmr exp – J a2 n a (C π nm cos(n θ) + Dnm sin(nθ ) ). 2. m = 1. g > That is. t) = 0. a)×(– π . . ρ nmr ∞∞ ρ nmr Jn on the rectangle cos(n θ) . By superposition. n = 0. θ .. a n a and π ∫ cos(nθ) sin (kθ ) dθ . = 0 –π ∫ ∫ r f(r. n≠ k 0 . the zeros of n 39 λ nm Jn( ρ ). if m ≠ l. . t) = 0. n = k and m ≠ l a ρ kl r2 π ∫ r Jk a dr.. The orthogonality relations 0 ∫ a ρ nmr ρ nlr r J n J dr = 0.. . . .. R(a) = J( λ a) = 0 or λa = ρ nm . n = 0.. π ) with respect to the inner product π <f . θ) g(r. ∫ ∫ 0 a π –π ρ nmr ρ kl r r Jn cos(n θ) Jk a cos(k θ) dθ dr a = ∫ ∫ a 0. n≠ k 0 .. a solution of the heat equation on the disc satisfying the boundary condition u(a. . . . θ) ∈ (0. 2. m=1 a (r.. k and l. θ. 1. That is λ nm = a . 2. .12. D nm are determined using these orthogonality properties. Jn a sin(nθ ) a n=0. m = 1. For u(a. m. imply the orthogonality of . θ . if n ≠ k. 2. n = k and m ≠ l a ρ kl r2 π ∫ r Jk a dr. a The coefficients C nm .
a 0 1 hence Ckl = 0 a π∫ Similarly. θ) dθ dr. a k = 0. a π π∫ a ρ kl r2 r Jk dr a 0 –π ∫ ∫ ρ kl r r J k sin(kθ) f(r. 40 a 0 –π ∫ ∫ a π ρ kl r ρ kl r2 r J k cos(k θ) f(r.. 2. .12. Bessel Functions. θ ) dθ dr = C kl π ∫ r J k a dr. θ ) dθ dr. .... l = 1. Dkl = 0 ρ kl r2 r Jk dr a 1 0 –π ∫ ∫ a a π ρ kl r r J k cos(k θ) f(r. . . . 1. 2.
We define the convolution product . 0 < x < 1 f(x) = g(x) = 0 . Let f : R→R and g : R→R be defined by x– 3 . x– 3 . Counterexamples are given by x– 3 . We call the class of all ∞ – ∞ such functions L (R). Fourier Transforms. 0 < x < 1 f(x) = 0 . ¯ ∞ Therefore g ∉ L1 (R) but g ∈ L2 (R). x > 1 g(x) = 0 . 1 Therefore f ∈ L1 (R) but f ∉ L2 (R). and we call the class of all such functions L (R). ∞ – ∫ ∞ ∞ 2 1 –3 ∞ g(x) dx = ∫ x dx = 3x 3 1 = ∞ 1 and – ∫ ∞ ∞ g(x) 2dx = 0 ∫x 1 4 –3 dx = – 3x 1 –3  1 = 3 < ∞. otherwise 2 . There is a product * for which f *g ∈ L1(R) whenever f ∈ L1(R) and g ∈ L 1(R). otherwise 2 . 13.13. 1 ∫ f(x) 2dx ∞ ∞ Examples. Then – ∫ ∞ – ∞ f(x) dx = ∞ 0 ∫ 1 x 2 –3 dx = 3x 3 = 3 < ∞ 0 1 –3 1 1 and ∫ ∞ f(x) 2dx = 0 ∫x 1 4 –3 dx = – 3x 0 = ∞. f : R→R is said to be square integrable on R if 2 < ∞. On the other hand. A function f : R→R is said to be integrable on R if – 41 ∫ f(x) dx < ∞. otherwise 2 . Similarly. the product fg is not necessarily in L1(R). Fourier Transforms. Given two function f ∈ L1 (R) and g ∈ L1 (R) .
then f ^ ∈ C(R) and f ^(ξ) → 0 as ξ → ∞. ∞ ∞ ∞ f^(ξ) = ∫ f(x)e –iξx dx ≤ – ∞ ∞ – ∫ f(x)  dx < ∞. If f ∈ L (R). ¯ The Fourier transform of a function f ∈ L1 (R) is defined to be f^(ξ) = – 1 where ξ ∈ R. ∞ ∞ Lemma. The boundedness follows easily from ∫ f(x)e –iξxdx. Then f^(ξ) = – ∫ ∞ ∞ f(x)e –iξxdx = – e πi – ∫ f(x)e –iξxdx ∞ ∞ =– – ∫ f(x)e ∞ ∞ –iξ x + π ξ dx = – – ∫ f x – ξ e –iξxdx. ∞ The continuity follows from the following. ∞ e –iξxdx ∞ π Therefore 2f^(ξ) = – ∫ ∞ ∞ f(x)e –iξxdx – π ∫ f x – ξ –∞ ∞ ≤ – ∫ f(x) – f x – ξ e –iξx dx ∞ ∞ π . If f ∈ L1 (R) and g ∈ L1 (R). Let ξ ∈ R. 1 Theorem. Proof. It is easy to see that the Fourier transform of a function f ∈ L (R) is a bounded continuous function on R. then f *g ∈ L1 (R).13. Fourier Transforms. ξ ≠ 0. ∫ (f*g)(x) dx = – ∫ – ∫ f(x – y) g(y) dydx –∞ ∞ ∞ ≤ – ∞ ∞ ∞ ∫ ∫ f(x – y) g(y)  dydx ∞ ∞ – ∞ ∞ ≤ – ∫ f(x – y)  ∫ g(y)  dydx ∞ ∞ – ∞ ∞ ≤ – ∫ f(x)  dx ∫ g(y)  dx ∞ ∞ – ∞ ∞ < ∞. Proof. (f*g)(x) = – 42 ∫ f(x – y) g(y) dy. (RiemannLebesgue).
13. ∞ ∞ . To prove continuity of f^.e –iξx (e –iη x – 1) dx = 2 ηx ∫ f(x) sin ( 2 ) dx ∞ x < a ∞ ηx ≤ 2 x > a ∫ f(x)  sin ( 2 ) dx ≤ 2 x > a + 2 ∫ f(x)  sin ( 2 ) dx ηx ηx ∫ f(x)  dx + 2 x < a ∫ f(x)   2 dx Therefore f^(ξ) is uniformly continuous on R. let ε > 0 be given and a > 0 chosen such that x > a ∫ f(x) dx < 4 and ε δ > 0 chosen such that 2a δ η < δ . x < a ¯ We define the linear transformation F: f → f^ defined by (Ff)( ξ) = f^(ξ) = – F is called the Fourier integral transformation and f ^ the Fourier transform of f ∈ L1 (R). x < a  ≤ – f^(ξ + η) – f^(ξ) ∫ ∞ ∞  = ∫ (f(x)e –i( ξ + η)x – f(x)e –iξx)dx – ∞ – f(x) . ≤ – 43 ∫ f(x) – f x – ξ dx → 0 ∞ ∫ f(x) dx < ε . ε ε ε ≤ 2 + aδ ∫ f(x) dx ≤ 2 + 2 = ε. ∫ f(x)e –iξxdx. Then for ∞ ∞ π as ξ  → ∞. Fourier Transforms.
λ > 0.13. the dilation of f by λ is defined as δ λ f where (δλ f)(x) = λ – 2f(λ –1 x). Let f : R → R and h ∈ R. x ∈ R. 44 1. For real c. ( e icxf(x) ) ^ = – ∫ e –iξx (e icxf(x) )dx ∞ ∞ = – ∫ f(x )e –iξ(x – c )dx ∞ ∞ = f^(ξ – c) = (τ f ^ )(ξ). Translation (a). Properties of Fourier transforms. Dilation. The Fourier transform of δ λ f is ∞ 1 (δλ f) = λ = λ 1 ^(ξ) = 1 – –2 1 ∫ f(λ –1 x) e –iξxdx ∞ – ∞ – ∫ (δλ f)(x)e –iξxdx ∞ –iξλy –2 ^ ∫ f(y) e ∞ ∞ λdy = λ 2 f ( λξ) = (δλ f )(ξ) –1 ^ . x ∈ R. – ∞ ∫ f(x – h)e –iξxdx ∞ ∫ f(x )e –iξx dx ∞ ∞ = e –iξ h (b). c 2. Let λ ∈ R. Fourier Transforms. The Fourier transform of τhf is ∞ (τ f) h ^(ξ) = ∞ – ∫ (τh)(x)e –iξxdx ∞ = – = – ∫ f(x )e –iξ(x + h)dx ∞ = e –iξ hf^(ξ). The translate of f by h is the function τhf defined by (τhf)(x) = f(x – h).
k = 1. Multiplication. Example 1. . 2. Fourier Transforms. . . 45 ∂ 1 3. k = 1. The Fourier transform of Df is – (Df)^(ξ) = ∫ (Df)(x)e –iξxdx ∞ = f(x)e –iξx ∞ –∞ ∞ – – ∫ f(x) (De –iξx)dx ∞ ∞ = (iξ)f^(ξ). ∞ ((–ix)kf)^ . g ∈ L (R). Convolution. ∂ .. Let f(x) = e x2 –2 ∞ ∞ .13. 2. ∞ Therefore (–ixf )^ = ∂f ^. By induction it follows that ∂ kf^ = 1 1 5. Differentiation.... Then f*g ∈ L (R) and has Fourier transform (f*g) ^(ξ) = – ∫ (f*g)(x)e –iξxdx ∞ ∞ = ∫ ∫ f(x – y) g(y) dy e –iξxdx – ∞ – ∞ = – ∞ ∞ ∫ ∫ f(x) g(y) e –iξ(x + y)dx dy ∞ ∞ – ∞ ∞ = ∫ f(x)e –iξxdx ∫ g(y)e –iξydy – ∞ – ∞ = f^(ξ)g^(ξ). . By induction (Dkf)^ = (iξ)kf^. Let f and f' ∈ L (R) and denote by D the differential operator D = ∂x . Let f. Then f ∈ L1 (R) and has Fourier transform f^(ξ) = – ∫ ∞ ∞ x2 – 2 –i ξx e e dx . We denote by ∂ the differential operator ∂ = then ∂ (∂f ^)(ξ) = ∂ξ ∫ f(x)e –iξxdx – ∞ = – ∞ ∫ (–ix)f(x)e –iξxdx. If f and xf ∈ L 1(R) ∂ξ 4.
ξ . 46 ∂f ^ ∂ (ξ ) = ∂ξ ∂ξ = – – ∫ ∞ ∞ x2 – 2 –i ξx e e dx ∫e ∞ ∞ ∞ x2 –2 (–ix)e –iξxdx 2 x ∂ – 2 –iξx =i ∫ dx ∂x e e –∞ x2 –2 –iξxe ∞ ∞ x2 –2 = ie  –∞ – i ∫ e ∞ – – ∂ –iξx)dx ∂x (e =– ξ = – ξf^(ξ). Fourier Transforms. – x ^ e 2 = 2 2 √π e ξ2 –2 . otherwise (χa)^(ξ) = ∫ ∞ – ∞ χa(x) e –iξxdx = ∫ –a a e –iξxdx =– 1 1 –iξx a e  – a = – iξ iξ = (e –iξa – e –iξa) 2 sin a ξ . ≡ χa(x) ∈ L1 (R) and has Fourier transform Example 2. ¯ x ≤ a 1 . Since f^(0) = c = – ∫e ∞ ∞ x2 –2 dx = 2 √π . Let f(x) = 0 . Therefore f^ satisfies a first order ordinary differential equation with solution f ^(ξ) = ce ξ2 –2 ∫ ∞ ∞ x2 – 2 –i ξx e e dx .13.
define 47 1 fa(x) = 2 π a ∞ – ∫ f^(ξ) e iξxdξ a a 1 = 2π – ∫ a –iξydy e iξxdξ ∫ f(y)e – ∞ a 1 = 2π ∞ ∫ – ∫ e iξ(x – y )dξ f(y) dy –∞ a ia(x – y ) – e –ia(x – y ) f(y) dy ∞ 1 = 2π ∞ e ∫ –∞ i(x – y) = – ∫ Da(x – y) f(y) dy = ∫ Da(y) f(x – y) dy ∞ ∞ – ∞ = where Da (x ) = Now 1 fa(x) – 2 (Da*f )(x) ∞ sin ax .14. 1 The Fourier transform of f ∈ L (R) is continuous. πx [ f(x–) + f(x+ ) 1 = π 1 + π 1 = π 1 ]=π ∞ – ∫ f(x – y) ∞ sin ay 1 dy – 2 [f(x–) + f(x+ ) ] y sin ay y dy since ∫ [f(x – y) – f(x–)] 0 0 sin ay ∫ [f(x – y) – f(x+)] y dy –∞ ∫ 0 ∞ sin ay π y dy = 2 0 ∫ [f(x – y) – f(x–)] 1 + π ∞ 0 ∞ sin ay y dy sin ay y dy ∫ [f(x + y) – f(x+)] . Assuming f ∈ L (R). bounded and f ^(ξ)→ 0 as ξ → ∞ . f piecewise smooth. Inverse Fourier Transforms. and 1 f^ ∈ L (R). 14. Inverse Fourier Transforms. 1 1 However f^ is not neccesarily in L (R).
14. Inverse Fourier Transforms. In the second integral, ∞ sin ay 1 [f(x + y) – f(x+)] y dy π 0∫ 1 = π If K ≥ 1,
∞ ∞ sin ay ∫ f(x + y) ≤ ∫ f(x + y) dy y dy K K
48
∫[ 0
K
f(x + y) – f(x+ )
sin ay 1 ] y dy + π
∫ [f(x + y) – f(x+)] K
∞
sin ay y dy
and
K ∞ ∞
∫
∞
f(x + )
sin ay sin ay + ∫ y dy. y dy = f(x ) K
∞
∞
Since
∞
0
∫ f(x) dx , 0∫
sin ay sin ay dy are both convergent integrals, ∫ y y dy → 0, K
K
∫ f(x + y) dy → 0 as K → ∞. ∫[
K
For the integrals over [0, K], f(x + y) – f(x+ ) sin ay ] y dy =
1
0
e ∫ –∞
∞
iay – e – iay g(y) dy 2i
= 2i [g^(–a) – g ^(a) ]
where g(y) =
f(x + y) – f(x+ ) ;0<y<K y 0 ; otherwise
.
Since f is piecewise smooth, f'(x+ ) exists for all x ∈ R and lim g(y) = f'(x+ ). Therefore g is bounded on [0, K] and hence g ∈ L (R). By the RiemannLebesgue lemma, g ^ exists, is continuous, g ^(±a) → 0 as a → ∞ and therefore sin ay y dy → 0 as a →∞ for K ≥ 1. A virtually identical argument works for the first integral.
1 y →0 +
∫ [f(x + y) – f(x+)] 0
K
Therefore
f (x) – 1 [f(x–) + f(x+) ]→ 0 as a → ∞ 2
a
or
1 lim fa(x) = lim 2 π f^(ξ) e iξxdξ a→∞ a→∞ –a
∫
a
1 = 2π
–
∫ f^(ξ) e iξxdξ = 2[f(x–) + f(x+) ]. ∞
1
∞
Summarising
14. Inverse Fourier Transforms. Theorem. Let f ∈ L1 (R) and let f be piecewise smooth on R. Then for every x ∈ R,
1 2π 1 ∫ f^(ξ) e iξxdξ = 2[f(x–) + f(x+) ]. ∞
∞
49
–
1 If x is a point of continuity of f, then 2 π
∞
–
∫ f^(ξ) e iξxdξ = f(x). ∞
¯ the inverse Fourier transform of φ and in called the inverse Fourier transformation.
∞
1 1 For φ ∈L (R), we call φ˘(x) = 2 π
–
F defined by (F φ)
1
–1
–1
1 (x) = 2 π
–
1 If φ is the Fourier transform of a piecewise smooth function f ∈ L (R), that is φ = f ^, then
∫ φ(ξ) e iξxdξ ∞
∞
∫ φ(ξ) e iξxdξ ∞
we define f(x) = 2 [f(x–) + f(x+ ) ]at a point of discontinuity of f.
1 Then f(x) = φ˘(x) = 2 π
–
∫ f^(ξ) e iξxdξ. ∞
∞
That is F –1 Ff = f.
Example. The inverse Fourier transform is useful in computing Fourier transforms. 2a 2a Since F (e –ax ) = 2 , it follows that F–1 F(e –ax )= F–1 2 2 a + ξ 2 or a +ξ 2a e –ax = F –1 2 a + ξ 2 . Interchanging the roles of x and ξ, multiplying by 2π, leads to π 2a 1 Fa2 + x 2 = 2 π e –aξ  or F a2 + x 2 = a e –aξ  . In a similar fashion, every Fourier transform pair defines a dual pair using the inverse Fourier transform.
15. Applications to Differential Equations. 15. Applications to Differential Equations. 1. The wave equation. The equation
50
∂2u ∂2u – c 2 2 = 0, – ∞ < x < ∞, t > 0, ∂t2 ∂x
describes the vertical vibrations of an infinite stretched elastic string, where u(x, t) is the vertical displacement of the string from its rest position at position x, time t. Let the initial displacement and velocity be given as u(x, 0) = f(x),
∂u ∂t (x, t) = g(x), – ∞ < x < ∞.
We take Fourier transforms of the wave equation and the initial conditions with respect to the x variable and denote by u^ (ξ, t ) the Fourier transform F (u(x, t) ). The using the derivative properties of the Fourier Transform,
∂2u^ – (ci ξ)2u^ = 0, – ∞ < ξ < ∞, t > 0, ∂t2
or
∂2u^ + (cξ)2u^ = 0. ∂t2
u^ (ξ, t) = A( ξ) cos(c ξt) + B( ξ) sin(c ξt).
Then
∂u^ When t = 0, u^ (ξ, t) = A( ξ) = f^(ξ), ∂t (ξ, t) = g^(ξ) = cξB(ξ). Therefore
g^(ξ) 2 sin(cξt) u^ (ξ, t) = f^(ξ) cos(cξ t) + 2c ξ e i(ct) ξ + e – i(ct) ξ 1 = f^(ξ) + 2c 2
(χ *g)^(ξ).
ct
Using the translation property and the convolution theorem,
1 u(x, t) = 2 1
[f(x + ct ) + f(x – ct )]
1 + 2c 1 2c
–
∫ ∞
∞
χ ct(x – ct) g(y) dy
= 2
[f(x + ct ) + f(x – ct )] +
1
x – ct < ct
∫
g(y) dy
= 2
[f(x + ct ) + f(x – ct )]
1 + 2c
x + ct x – ct
∫
g(y) dy.
This is d'Alemberts solution to the onedimensional wave equation.
Hence g ^(ξ) = 0. ∂y 2 u^ (ξ. Taking transforms. ∂y (ξ. – ∞ < x < ∞ for some function g ∈ L1 (R). (iξ)2u^ (ξ. B(ξ). Solving for A(ξ). and for ξ < 0. Laplace's Equation. – ∞ < x < ∞. B(ξ). y) – ξ 2u^ (ξ. B(ξ) = 2 f^(ξ) – ξ . 0) = f ^(ξ). y) = 2 f^(ξ) + e + 2 f (ξ ) – e . 0) = f(x). – ∂u ∞ < x < ∞. In addition to u(x. y) → 0 as y → ∞. u^ (ξ. ξ ξ For ξ > 0. Consider Laplace's equation in two variables on the upper halfplane y > 0. let ∂y (x. y) = 0. Then taking the Fourier transform in the variable x. ξA( ξ) – ξB( ξ) = g^(ξ). Then 51 ∂2u ∂2u + = 0. in fact such that u(x. y) + or ∂2u^ (ξ. u^ (ξ. ξ > 0 ^ (ξ ) = g ξf^(ξ) . ∂y 2 – ∞ < x < ∞. Applications to Differential Equations. We will find g such that the solution u(x. y) → 0 as ξ . ∂2u^ (ξ. y) → 0 as y → 0 if and only if f ^(ξ) + g ^(ξ) = 0. 2. y > 0. ξ and g ^(ξ) ξy 1 ^ g ^(ξ) – ξy 1 u^ (ξ. ξ < 0 = – ξ f ^(ξ). y) = 0. or A( ξ) + B(ξ) = f^(ξ).15. 0) = f(x). 0) = g(x). y > 0. g ^(ξ) g ^(ξ) 1 1 A( ξ) = 2 f^(ξ) + . ∂u^ ^ u^ (ξ. which has solutions We need two conditions to determne the functions A(ξ). –∞< x< ∞ be given for a function f ∈ L1 (R). Therefore y → 0 if and only if f^(ξ) – ξ –ξf^(ξ) . 0) = g (ξ). ∂x 2 ∂y 2 Let the boundary condition u(x. y) = A( ξ)e ξy + B(ξ)e – ξy. y) is bounded for y > 0.
f^(ξ)e – ξy . by the convolution theorem. t) = u^ (ξ. ξ < 0 = f^(ξ)e – ξy. y) = f^(ξ)e ξy . – ∞ < x < ∞ for f ∈ L1 (R). The heat equation. t) = 0. f continuous and bounded. t) + κ ξ u (ξ. Taking transforms in the variable x. t > 0 for u(x. with initial condition u(x. ^ y . t) = f(x). Solving this first order ordinary differential equation. can be solved using Fourier transforms. π ((x – s) 2 + y 2) ∂u ∂2u – κ 2 = 0. Since e – ξy = π (x2 + y 2) 52 y u(x.15. – x ^ Now e 2 = 2 2 √π e ξ2 –2 and using the dilation property of Fourier transforms 1 (λ x) 2^ λ – 2e – 2 = –1 2 √π λ e –1 1 2 (λ ξ) 2 – 2 e – 2 (λ ξ) 2 = 1 λ √π 2 (λ x) 2^ e – 2 . t) = 0. – ∫ ∞ ∞ y f(s) ds. t) a function of two variables. 2 2 u^ (ξ. Applications to Differential Equations. ∂u^ 2 ^ ∂t (ξ. or ∂u^ 2 ^ ∂t (ξ. t) – κ (i ξ) u (ξ. y) = π (x2 + y 2)*f = 3. ∂t ∂x – ∞ < x < ∞. The Heat Equation. λ2 Let 2 = κ t or λ = 2κ √ t . ξ > 0 u^ (ξ. then . 0)e – κξ t = f^(ξ)e – κξ t .
. 2 e – κξ t = 53 1 1 2 2κ √π √t 1 – 2 e 1 x 1 2κ √t 2 ^ x 2 ^ = 4πκ √ t – 2 e 2κ √t . t) = e 4πκ √ t . since u^ (ξ. – The function h(x. is called the heat kernel. t) = f^(ξ) 1 4πκ √ t – 2 e 1 2κ √t x 2 ^ – 2 1 u(x. t) = f* e √ t 4πκ 1 1 2 2κ √t (x) x = x2 4κt 4πκ √ t – ∫ ∞ ∞ – e (x – y )2 4 κ t f(y) dy. We can then write u(x. Applications to Differential Equations. t) f(y) dy. Therefore from the convolution theorem. t) = – ∫ ∞ ∞ h(x – y.15.
∞ ∞ ∞ ∞ – ∫ f( ∞ ∞ y) 2 1 dy = f ^(ξ)2dξ . g ∈ L (R)∩L (R). 1 2 By the convolution theorem. ξ 2 . Plancherels' and Parsevals' Identities. hence 1 f(– y) g(–y) dy = f^(ξ) g^(ξ) dξ ∫ 2π – ∫ –∞ ∞ or 1 ∫ f( y) g(y) dy = 2π – ∫ f^(ξ) g^(ξ) dξ. Let f(x) = χ a(x) = 0. 1. –∞ ∞ Set x = 0. When f = g we obtain Parsevals' identity. ξ ∞ ∞ sin a ξ2 2 dx = 1 dξ ∫ χa(x)  2π – ∫ ξ –∞ ∞ – ∫ ∞ ∫ ∞ ∞ 2 sin a ξ dξ = 4 πa ξ 2 or – ∞ sin a ξ dξ = π a. the Fourier transform g^(ξ) is replaced by g^(ξ) . .16. for f. –∞ ∞ This is Plancherels' identity. ∞ ∞ 1 ∫ f(– y) g(y) dy = 2π – ∫ f^(ξ) g ^(ξ) d ξ –∞ ∞ Replacing g(x) by g(–x) . 2π – ∫ ∞ ∞ Examples. 54 (f*g)^ = f^ g^. x < a 1. Plancherels' and Parsevals' Identities. 16. otherwise 2 sin a ξ Then f^(ξ) = . and by Parsevals' identity. Therefore f *g = or ∞ (f^ g^)˘ ∞ 1 ∫ f(x – y) g(y) dy = 2π – ∫ e i ξx f ^(ξ) g ^(ξ) d ξ.
16. Let f(x) = e – ax. Plancherels' and Parsevals' Identities. 2a (a2 + ξ 2)2 . 2. f^(ξ) = ∞ 55 a2 2a + ξ2 and by Parsevals' identity 1 = 2π 2a 2 ∫ 2 2 dξ –∞ a + ξ ∞ – ∫( ∞ e – ax ) 2dx 1 2a2 = a π or – – ∫ ∞ ∞ 1 dξ (a2 + ξ 2)2 ∫ ∞ ∞ π 1 dξ = 3 .
human hearing is assumed to be limited to frequencies below about 20 kHz. Band Limited Functions and Shannon's Sampling Theorem. 17. then f(x) is called a bandlimited function and ξ c is called the cutoff frequency. provided the sampling interval L. Band Limited Functions and Shannon's Sampling Theorem. Therefore the acoustic signals recorded on compact discs are limited to a bandwidth of 22 kHz. are bandlimited.17. If f ^(ξ) = 0 for ξ  > ξc > 0. For example. ±1. 56 The Fourier transform variable has the role of frequency and f ^(ξ) is referred to as the frequency representation of f(x). A signal represented by a continuous function f(x). is sufficiently short. }. A first step in the processing of signals.. {f(nL). . ξ  > ξ ^ inπξ φ (ξ ) = – ξ 1 2ξ e . n = 0. Shannons' Theorem shows that it is possible to exactly recover the bandlimited continuous function f(x) from knowledge of its samples. ±2. is sampling. is replaced by its samples at regular intervals.. ξ ≤ ξ √ c n c c The inverse Fourier transforms of φ n (ξ) are given by 1 φn(x) = 2π – ^ c ∫e ∞ ξc c ∞ i xξ ^ φ n (ξ ) d ξ inπξ – ξ c 2π φ n(x) = – ∫e ξ i xξ 1 √ 2 ξc e dξ = 1 2 √ξc – ∫ ξ ξc c e i ξ x – nπ ξc dξ i ξx – nξπ ξ 1 e = nπ 2 √ξc i x – ξ –ξ c c c c = 2 √ξc √ 2 ξc e i (ξcx – n π) – e – i (ξcx – n π) 2i(ξcx – nπ ) sin (ξcx – nπ ) = ξcx – nπ . ^ Consider the functions φ (ξ) given by n 0. Many functions from science and technology.
Band Limited Functions and Shannon's Sampling Theorem. φn ^ ^ f . Since f^(ξ) n is band limited to ξ  ≤ ξc.17. ξc). }. L ξc(x – nL) . < > n=– ∞ = 2π <f . f(x) = n=– ∞ ∑ ∞ <f . φ >φ (ξ). m ≠ n . = 1. φ >2π φ (x) n n = n=– ∞ ∑ ∞ < f .. ±2. φn >√ 2π sin ξc(x – nL) . φ > <f . form an orthogonal set in L2 (– ξc. m = n ^ So the functions {φ (ξ) . φm > = – ∫ ∞ ^ ^ φ n (ξ) φm(ξ) d ξ 1 = 2 ξc – ξc ∫ ξc e inπξ – ξ c – e imπξ ξc dξ 1 = 2 ξc – ∫ ξ ξc e i(m – n) πξ ξc dξ c 0. φn = – ∫ ξ ξc f^(ξ) 1 c 2 √ξc n e inπξ ξc dξ . ±1.. n = 0. it has a Fourier series f^(ξ) = n=– ∞ ∑ ∞ cn φn (ξ) cn 1 e inπξ – ξ c ^ = n=– ∞ ∑ ∞ √ 2 ξc dξ where the Fourier coefficients cn are given by cn = By Plancherel's theorem cn = so < > ^ ^ f . 57 = Consider the inner products √ 2π sin ξc(x – nL) π where L = . L ξc(x – nL) ξc ∞ < ^ ^ φ n . n n f^(ξ) = ∑ ∞ ∞ ^ cn φ n (ξ) = ∑ 2π n=– ∞ ^ Taking inverse Fourier transforms. .
Let f ∈ L (R)∩L (R) be continuous and band limited to ξ  ≤ ξ c. Band Limited Functions and Shannon's Sampling Theorem. to reconstruct a bandlimited function. . 2π sin ξc(x – nL) L ξc(x – nL) . 1 2 Theorem. = n=– ∞ ∑ ∞ f(nL) sin ξc(x – nL) ξc(x – nL) Summarising we have Shannon's theorem. it suffices to 2π sample at a frequency ω = L = 2 ξc. shows that from Shannon's theorem. . ξc ¯ The relationship ωL = 2π . φn >√ 2π . Then f(x) = n=– ∞ ∑ ∞ f(nL) sin ξc(x – nL) ξc(x – nL) where L = π .. where ω is the frequency of sampling in cycles per unit length. φn √ >√ L . φ >√ L = f(kL) < Finally therefore f . ±2. ±1. φk > f(x) = n=– ∞ ∑ ∞ < f . . = So <f . twice the cutoff frequency. 2π k = 0. φn f(kL) = n=– ∞ ∑ ∞ < >√ 2π sin ξc(kL – nL) L ξc(kL – nL) 2π sin(k – n)π L (k – n)π = n=– ∞ ∑ ∞ < k f ..17. The samples at intervals of length L = 58 π are ξc f .
^ . Then the quantity 59 ∫ (x – a) f(x) dx ∆ f ≡ f(x)  dx ∫ ∞ 2 2 – ∞ a ∞ 2 – ∞ is called the dispersion about the point x = a of f. xf ∈ L (R). Heisenberg’s Inequality. Notice that the Fourier transform χb does not have a finite dispersion about the origin. The reasoning behind the definition is that if f(x) is concentrated near x = a . α ∈ R. Example. The following result shows that there is a type of inverse relationship between the dispersion of a function and that of its Fourier transform. 0 . since ∞ ∞ 2 sin(bξ) 2 2χ ^(ξ)2 d ξ = ∫ξ b ∫ ξ 2 ξ dξ –∞ –∞ = 4 – ^ ∫ ∞ ∞ sin (bξ) dξ = ∞. then ∆af is smaller than when f is not close to zero far from x = a. Consider the characteristic function χb(x) = which has Fourier transform x ≤ b 1 .18. 2 2 Let f ∈ L (R). otherwise 2 sin b ξ . 18. Then for all a. HeisenbergÕs Inequality. 2 This indicates that χ b is spread out away from x = 0. Let f ∈ L1 (R)∩L2 (R). Theorem. The dispersion about the origin is ∆0 χb ∫ x dx b = = 3 ∫ dx b 2 – b 2 b – b and it clear that the dispersion increases as b increases. ξ (χb)^(ξ) = Notice that χb is concentrated near x = 0 for small b.
60 2 2 ∫ x f(x) dx ∫ ξ f^(ξ) dξ 1 (∆ f )(∆ f^ ) = ^ ≥4 ∫ f(x) dx ∫ f (ξ) dξ ∞ ∞ 2 2 – a α ∞ ∞ – ∞ ∞ 2 2 – ∞ – ∞ and equality holds if and only if f(x) = c e – kx for constants c ∈ R and k > 0. ∫ ∞ ∞ 1 2x ( f*(x) f(x) + ∞ f*(x) f(x) ) dx 2 = ∫ x Re – ∞ ∞ ( f*(x) f(x) ) 2 dx = Re – ∫ ∞ 2 2 x f (x) f*(x) dx ≤ ∫ x f (x) f*(x) dx – ∞ ∞ ∞ ≤ ∫ x2 f(x) 2 dx ∫ f*(x) 2 dx – ∞ – ∞ (by the CauchySchwartz inequality). We firstly prove the result for a = α = 0. and ∫ x2 f(x) 2 dx ∫ ξ 2f^(ξ)2dξ = ∫ x2 f(x) 2 dx ∫ (iξ)f^(ξ) 2dξ – ∞ – ∞ – ∞ – ∞ = ∫ x2 f(x) 2 dx ∫ (f*) ^(ξ) 2dξ – ∞ – ∞ = 2π ∫ x2 f(x) 2 dx ∫ f*(x) 2 dx – ∞ – ∞ Since – ∞ ∞ ∞ ∞ ∞ ∞ ∞ ∞ (by Parseval’s identity). Let f *(x) ≡ ((i ξf^(ξ)) ˘ or (f*) ^(ξ) = iξf^(ξ). We will show that – – ∞ ∫ ∞ ∞ x (f*(x) ∞ f(x) + f*(x) f(x) )dx = ∞ – ∫ f(x) 2 dx ∞ ∞ ∞ from which the result follows. Then f* ∈ L2(R).18. HeisenbergÕs Inequality. Proof. for then 2 2 2 2 2 2 2 ∫ x f(x)  dx ∫ ξ f^(ξ) dξ = 2π ∫ x f(x)  dx ∫ f*(x)  dx – ∞ – ∞ – ∞ – ∞ ≥ 2π ∞ – ∫ ∞ ∞ 1 2x ( f*(x) f(x) + f*(x) f(x) ) dx 2 . – 2 ∫ ∞ ∞ x2 f(x) 2 dx and – ∫ ∞ ∞ ξ2f^(ξ)2dξ are both assumed finite since otherwise the result is trivial.
As for the case of equality in Heisenberg’s inequality. Taking the limit as b → ∞ and a → – ∞ . ∞ ∞ To complete the proof. . f(x) f*(x) = f(x) K x f(x) = x f(x) 2 K Therefore K is real. we assume that f is continuous and piecewise smooth. Then from the property of Fourier transforms. That is. this holds if and only if f(x) f*(x) is real and f*(x) = K x f(x) for some complex constant K. 0= – ∫ ∞ ∞ x ( f*(x) f(x) + f*(x) f(x) )dx + – ∫ f(x) 2 dx ∞ ∞ as required. This assumption can be removed since functions in L 1 (R) are the uniform limit of such functions. HeisenbergÕs Inequality. b] . which is not integrable. The differential equation f'(x) = f’(x) = K x f(x) has solutions of the form f(x) = c e – 2 is real. c any real constant. f*(x) = f'(x) wherever the derivative exists. Kx2 . Then for any interval [a. b f(b)2 b – a f(a)2 = ∫ dx (x f(x)  2 )dx a b d = ∫ (x f ' (x) a b f(x) + x f(x) f '(x) + f(x)  2 b )dx 2 = ∫ a x (f*(x) f(x) + f*(x) f(x) )dx + ∫ f(x) a dx The assumption f ∈ L2 (R) implies that b f(b)2 → 0 as b → ∞ and a f(a)2 → 0 as a → – ∞ since otherwise f(x)  > c x–1/2 as x → ∞.18. π =2 61 – ∫ ∞ ∞ f(x) 2 2 π dx = 2 ∫ f(x) 2 dx – ∞ ∞ 1 2π – ∫ f^(ξ) 2dξ ∞ ∞ 1 = 4 – ∫ ∞ ∞ f(x) 2 dx – ∫ f^(ξ) 2dξ .
and f(x) = c – e 2 62 Kx2 ∈ L2(R) if and only if K > 0.18. – ∞ Whereas. Kx2 Conversely. Then ∞ ∞ ∞ 2 2 2 2 2 2 2 ∫ x f(x)  dx ∫ ξ f^(ξ) dξ = 2π ∫ x f(x)  dx ∫ f*(x)  dx – ∞ – ∞ – ∞ – ∞ = 2π ∫ x2 f(x) 2 dx ∫ K x f(x) 2 dx – ∞ – ∞ 2 = 2π K 2 ∫ x2 f(x) 2 dx – ∞ = 2π ∞ ∞ ∞ ∞ K2 2 2 2 – Kx dx ∫x e – ∞ ∞ = 2π K2 x ∫ –2K – ∞ x ∫ –2K – ∞ ∞ ∞ (–2Kxe (–2Kxe (e – Kx 2 ) dx ) dx 2 2 = 2π K 2 – Kx 2 2 = 2π K2 x d ∫ –2K dx – ∞ ∞ – Kx 2 ) dx = 2π K2 1 – Kx 2 2 ∫ 2K e dx – ∞ π 2 – Kx 2 =2 ∫ e dx – ∞ ∞ (integration by parts) π 2 π2 –t2 = 2K ∫ e dt = 2K . 1 4 – ∞ ∫ ∞ ∞ f(x) 2 1 dx ∫ f ^(ξ) 2dξ = 4 ∫ f(x) 2 dx 2π ∫ f(x) 2 dx – ∞ – ∞ – ∞ π 2 = 2 ∫ f(x) 2 dx – ∞ ∞ ∞ ∞ ∞ . HeisenbergÕs Inequality. let f(x) = e ∞ for constant K > 0. Therefore equality holds in Heisenberg’s – kx2 inequality only if f(x) = c e – 2 for constants c ∈ R and k > 0.
if one of ∆ af or ∆ αf^ is very small then the other must be large. α ≠ 0. follows by observing that F(x) = e –iα x f(x + a) satisfies the same hypotheses as f(x) and ∆a f = ∆ 0F and ∆α f^= ∆0 F^ for any a ≠ 0. That is.18. ∞ 63 As a consequence of the inequality (∆ f )(∆ f^ ) ≥ 1 . α ≠ 0. . π 2 – Kx 2 =2 ∫ e dx – ∞ and equality holds. we see that it is impossible for 4 a α both ∆af and ∆αf^ to be simultaneously small. HeisenbergÕs Inequality. The case of a ≠ 0.
This action might not be possible to undo. Are you sure you want to continue?
We've moved you to where you read on your other device.
Get the full title to continue reading from where you left off, or restart the preview.