University Of Zimbabwe

MT205 Engineering Mathematics 2

Section: Fourier Series

Lecturer: Department:
S. Murambiwa Mathematics

Course Outline
∑
Fourier series: f (x) = a20 + ∞
n=1 (an cos nx + bn sin nx) .
Orthogonal relations and the Euler formula for the coefficients. Bessel’s inequality and
Parseval’s identity. Expansion in sine or cosine series. The Fourier transform, inverse
transform, convolutions. Application to partial differential equations.

1
Contents
1 Introduction to Fourier Series 3
1.1 Orthogonality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Euler Formula for the Coefficients . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Bessel’s Inequality and Parseval’s Identity . . . . . . . . . . . . . . . . . 6
1.4 Fourier Series of Even and Odd Functions . . . . . . . . . . . . . . . . . 7

2 The Fourier Transform 9

2.1 Fourier transforms as integrals . . . . . . . . . . . . . . . . . . . . . . . 9
2.2 Properties of the Fourier transform . . . . . . . . . . . . . . . . . . . . . 10
2.3 Fourier Transform of a Convolution . . . . . . . . . . . . . . . . . . . . . 11
2.4 Fourier Sine and Cosine Transforms . . . . . . . . . . . . . . . . . . . . . 11

3 Application to Partial Differential Equations 13

3.1 Heat Equation for an Infinite Rod . . . . . . . . . . . . . . . . . . . . . . 13
3.2 Half Range Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

2
1 Introduction to Fourier Series
A function or a real variable f is said to be periodic with period 2L if
f (x + 2L) = f (x)
holds for all x.
For any function with period 2L, we can describe the f (x) in terms of an infinite sum
of sines and cosines
a0 ∑ (
∞
nπx nπx
f (x) = + an cos + bn sin ). (1)
2 n=1
L L

When the period of the function is 2πthen the Fourier series is given by

a0 ∑(
∞
f (x) = + an cos nx + bn sin nx).
2 n=1

1.1 Orthogonality
There is a nice integral formula for finding the coefficients of any Fourier sum. This is
based on the orthogonality of the function cos nπx L
and sin nπx
L

Theorem 1.1 (Orthogonality Relations) If m, n ∈ N, then

∫ L ∫ L {
nπx mπx nπx mπx L, if m = n
cos cos dx = sin sin dx =
−L L L −L L L 0, otherwise,
and ∫ L
nπx mπx
sin cos dx = 0.
−L L L
Proof: Home work

1.2 Euler Formula for the Coefficients

To find the coefficients an , bn and a0 , we multiply equation (1) by cos nπx
L
or sin nπx
L
and
integrate it over the interval −L < x < L. By the orthogonality relations of sin and cos
functions, we can get
∫
1 L nπx
an = f (x) cos dx
L −L L
∫
1 L nπx
bn = f (x) sin dx
L −L L
∫
1 L
a0 = f (x)dx
L −L

3
Example

Find the Fourier series representation of the periodic function


 1, 0 < x < π
f (x) =

−1, −π < x < 0.

Solution:

∫
1 π
a0 = f (x)dx
π −π
∫ ∫ 0 ]
1[ π
= dx − dx
π 0 −π
= 0
∫
1 π
an = f (x) cos nxdx
π −π
∫ ∫ 0
1[ π ]
= cos nxdx − cos nxdx
π 0 −π
π 0
 
1 1 
= sin nx − sin nx
nπ 0 nπ 
−π
= 0
∫
1 π
bn = f (x) sin nxdx
π −π
∫ ∫ 0
1[ π ]
= sin nxdx − sin nxdx
π 0 −π
π 0
 
1 1 
= − cos nx + cos nx
nπ 0 nπ 
−π
= 0
1 ( ) 1 ( )
= − cos nπ − cos 0 + cos 0 − cos(−nπ)
nπ nπ
1 ( ) 1 ( )
= − (−1)n − 1 + 1 − (−1)n
nπ nπ
2 ( )
= 1 − (−1)n
nπ
 4
 nπ , when n is odd
=

0, otherwise.

4
Therefore
4 ( sin x sin 3x sin 5x )
f (x) = + + + ···
π 1 3 5
4 ∑ sin(2n − 1)x
∞
= .
π n=1 (2n − 1)

Example

Find the Fourier series representation of the periodic function

f (x) = x2 , − π < x < π.
Solution: ∫
1 π
1 3 π 2
a0 = x2 dx = x  = π2.
π −π 3π −π 3
∫ π
1
an = x2 cos nxdx. Applying integration by parts.
π −π

Let
1
u = x2 , du = 2xdx, dv = cos nxdx, v = sin nx.
n
Thus
π ∫ π
1 x2  2
an = sin nx − x sin nxdx
π n −π nπ −π
∫ π
2
= − x sin nxdx. Applying integration by parts again.
nπ −π
Let
1
u = x, du = dx, dv = sin nxdx, v = − cos nx.
n
Thus
π ∫
2 [ x  1 π ]
an = − − cos nx + cos nxdx
nπ n −π n −π
4
= (−1)n
n2

bn = 0
Thus the Fourier series for f is
π2 ∑
∞
4
f (x) = + (−1)n 2 cos nx
3 n=1
n
π2 cos 2x cos 3x
= − 4 cos x + 4 2 − 4 2 + · · · .
3 2 3

5
1.3 Bessel’s Inequality and Parseval’s Identity
Let us consider the nth partial sum

a0 ∑ ( rπx )
n
rπx
Sn (x) = + ar cos + br sin , (2)
2 r=1
L L

of the Fourier series for f (x) in (1) defined over the interval −L ≤ x ≤ L. Them
∫L
provided the integral −L [f (x)]2 dx exists and is finite, we have the obvious result
∫ L ∫ L ∫ L ∫ L
[f (x) − Sn (x)] dx =
2
[f (x)] dx − 2
2
f (x)Sn (x)dx + [Sn (x)]2 dx (3)
−L −L −L −L

From the definition of Sn (x) in (2), it follows that

∫ L ∫ L[ ∑n (
]
) 2
a 0 rπx rπx
[Sn (x)]2 dx = + ar cos + br sin dx,
−L −L 2 r=1
L L

but the orthogonality of the sine and cosine functions reduces this to
∫ L ∫ L 2 ∑n [ ∫ L ] ∑ n [ ∫ L ]
2 a0 2 2 rπx 2 2 rπx
[Sn (x)] dx = dx + ar cos dx + br sin dx
−L −L 4 −L L −L L
[ r=1
] r=1

a 2 ∑n
( 2 )
= L 0+ ar + b2r (4)
2 r=1

If f (x) is replaced by its Fourier series, a similar argument shows that

∫ L [ ]
a 2 ∑ n
( )
f (x)Sn (x)dx = L 0 + a2r + b2r (5)
−L 2 r=1

So combining (3) to (4) gives

∫ ∫ [ ]
L L
a 2 ∑n
( )
[f (x) − Sn (x)]2 dx = [f (x)]2 dx − L 0 + a2r + b2r (6)
−L −L 2 r=1

The integral on the left of (6) is non-negative, because its integrand is a squared quantity,
so it follows at once that for all n
∫
a20 ∑ ( 2 )
n
1 L
+ ar + b r ≤
2
[f (x)]2 dx.
2 r=1
L −L

This is Bessel’s Inequality for Fourier series, and the restriction to function f (x)
∫L a2 ∑
such that −L [f (x)]2 dx exists and is finite implies that the series 20 + nr=1 (a2r + b2r ) is
convergent, so the coefficients in the associated Fourier series (1) must be such that

lim an = 0 and lim bn = 0.

n→∞ n→∞

6
It is also a consequence of (6) that if the nth partial sum Sn (x) converges to f (x) in the
sense that ∫ L
lim [f (x) − Sn (x)]2 dx = 0,
n→∞ −L

then ∫
a20 ∑ ( 2 )
L n
1
[f (x)]2 dx = + ar + b2r .
L −L 2 r=1

This is the Parseval relation for the Fourier series.

1.4 Fourier Series of Even and Odd Functions

If f (x) is an even function defined on the interval −L ≤ x ≤ L, then

a0 ∑
∞
nπx
f (x) = + an cos , with
2 n=1
L
∫
2 L
a0 = f (x)dx
L 0
∫
2 L nπx
an = f (x) cos dx, for n = 1, 2, · · · .
L 0 L

If f (x) is an odd function, then

∑
∞
nπx
f (x) = bn sin , with
n=1
L
∫ L
2 nπx
bn = f (x) sin dx, for n = 1, 2, · · · .
L 0 L
Example

Find the Fourier series representation of

f (x) = |x|, in the interval − L ≤ x ≤ L.

Solution The Euler formula for the coefficients an of the even function |x| defined as

 −x, forx < 0
|x| =

x, forx ≥ 0,

7
gives
∫ L
2
a0 = xdx = L
L 0

and
∫ L
2 nπx
an = x cos dx
L 0 L
[ L
nπx ] 
2 L2 cos nπx Lnπx sin 
= L
+ L

L n2 π 2 n2 π 2 
0
2L [ ]
= (−1)n − 1 , forn = 1, 2, · · ·
n2 π 2
 4L
 n2 π2 , when n is odd
=

0, when n ̸= 0, is even.
Thus, the Fourier series representation of f (x) = |x| for −L ≤ x ≤ L is
[ ]
L 4L cos πx cos 3πx cos 5πx
f (x) = − 2 L
+ L
+ L
+ ··· .
2 π 12 32 52
The sequence of positive odd numbers can be written in the form 2n−1 with n = 1, 2, · · · ,
so the last result can be expressed more concisely as
L 4L ∑
∞
1 (2n − 1)πx
f (x) = − 2 cos , for − L ≤ x ≤ L.
2 π n=1 (2n − 1)2 L

Example

Find the Fourier series representation of f (x) = x on the interval −2 ≤ x ≤ 2

Solution Using the fact that L = 2,
∫ 2
nπx 4 4
bn = x sin dx = − cos nπ = (−1)n+1
0 2 nπ nπ
and as the function is odd all the coefficients an = 0. The required Fourier series
representation is
[ ]
4 πx 1 1 3πx
f (x) = sin − sin πx + sin − ··· ,
π 2 2 3 2
which can be written in the more concise form
4 ∑ (−1)n+1
∞
nπx
f (x) = sin , for − 2 ≤ x ≤ 2.
π n=1 n 2

8
2 The Fourier Transform
2.1 Fourier transforms as integrals
There are several ways to define the Fourier transform of a function f : R → C. In this
section, we define it using an integral representation and state some basic uniqueness
and inversion properties, without proof.
Definition 2.1 Let f : R → R. The Fourier transform of f (x), denoted by F (µ) =
F[f (x)], is given by the integral
∫ ∞
1
F (µ) = F[f (x)] = √ f (x)e−iµx dx,
2π −∞
for x ∈ R for which the integral exists.

We have the Dirichlet condition for inversion of Fourier integrals.

Theorem:
Let f : R → R. Suppose that
∫∞
1. −∞ |f (x)|dx converges and
2. in any finite interval, f, f ′ are piece-wise continuous with at most finitely many
maxima/minima/discontinuities.
Let F (µ) = F[f (x)]. Then if f (x) is continuous at x ∈ R, we have
∫ ∞
1
f (x) = √ F (µ)eixµ dµ.
2π −∞
Moreover, if f is discontinuous at x ∈ R and f (x + 0) and f (x − 0) denote the right and
left limits of f at x, then
∫ ∞
1
f (x + 0) + f (x − 0) = √ F (µ)eixµ dµ.
2π −∞
Example

Find the Fourier transform of f (x) = e−|x| and hence using inversion, deduce that
∫ ∞ ∫ ∞
dµ π µ sin µxdµ π
2
= and 2
= e−x , x > 0.
0 1+µ 2 0 1+µ 2
Solution:
We write
∫ ∞
1
F (µ) = √ f (x)e−iµx dx
2π −∞
[∫ 0 ∫ ∞ ]
1 −(1+iµ)x
= √ e (1−iµ)x
dx + e dx
2π −∞
√ [ ]
0

2 1
= .
π 1 + µ2

9
Now by the inversion formula,
∫ ∞
−x 1
e = √ F (µ)exµ dµ
2π −∞
∫
1 ∞ 1
= exµ dµ
π −∞ 1 + µ2
[∫ ∞ ixµ ]
1 e + e−ixµ
= dµ
π 0 1 + µ2
∫
2 ∞ cos µx
= dµ.
π 0 1 + µ2
now this formula holds at x = 0, so substituting x = 0 into the above gives the first
required identity. Differentiating with respect to x as we may for x > 0, gives the second
required identity.

2.2 Properties of the Fourier transform

Recall that ∫ ∞
1
F[f (x)] = F (µ) = √ f (x)e−iµx dx
2π −∞

1. F[.] is a linear operator. For a, b ∈ R we have

F[af1 (x) + bf2 (x)] = aF[f1 (x)] + bF[f2 (x)] = aF1 (µ) + bF2 (µ)

2. Fourier Transform exchanges differentiation with multiplication. F[f ′ (x)] = iµF (µ).
In general F[f n (x)] = (iµ)n F (µ).

Example:

Find the Fourier transform of the function

 −x
 e sin x, forx > 0
f (x) =

0, forx < 0.

Solution:

10
 ∫ ∞
1
F (µ) = √ f (x)e−iµx dx
2π −∞
∫ ∞
1
= √ e−x sin xe−iµx dx
2π −∞
=
.
.
. [ ]
1 1
= √ .
2π 1 + (1 + iµ)2

2.3 Fourier Transform of a Convolution

Definition 2.2 The convolution of f and g is the function f ∗ g defined by
∫ ∞
(f ∗ g)(x) = f (t)g(x − t)dt.
−∞

Note also that ∫ ∞

(f ∗ g)(x) = g(t)f (x − t)dt,
−∞

as can be shown by a change of variable.

Theorem [Convolution theorem]

√
F [(f ∗ g)(x)] = 2πF (µ)G(µ).

Proof: Home Work

2.4 Fourier Sine and Cosine Transforms

The Fourier Cosine Fc [f (.)] and Fourier Sine Fs [f (.)] of f : R → R are defined as follows,
√ ∫ ∞
2
Fc [f (.)] = Fc (µ) = f (x) cos µxdx
π 0
√ ∫ ∞
2
Fs [f (.)] = Fs (µ) = f (x) sin µxdx
π 0

Example

Find the Fourier Cosine and Sine transforms of the function f (x) = e−ax , a > 0, x > 0.

Solution

11
 √ ∫ ∞
2
Fc (µ) = f (x) cos µxdx
π 0
√ ∫ ∞
2
= e−ax cos µxdx
π
√ ∫0 ∞ ( iµx )
2 −ax e + e−iµx
= e dx
π 0 2
√ ∫
1 2 ∞ ( −(a−iµ)x )
= e + e−(a+iµ)x dx
2 π 0
√ [ ] ∞
1 2 1 1 
= − e−(a−iµ)x − e−(a+iµ)x 
2 π (a − iµ) a + iµ) 
√ [ ]
0
1 2 1 1
= +
2 π (a − iµ) a + iµ)
√ [ ]
1 2 2a
=
2 π a2 + µ 2
√ [ ]
2 a
=
π a2 + µ 2

Similarly, Show that

√ ∫ ∞
2
Fs (µ) = f (x) sin µxdx
π 0
√ ∫ ∞
2
= e−ax sin µxdx
π 0
√ ∫ ∞ ( iµx )
2 −ax e − e−iµx
= e dx
π 0 2i
=
.
.
. √
[ ]
2 µ
=
π a2 + µ 2

12
3 Application to Partial Differential Equations
Fourier transforms are used to convert the linear partial differential equation to an
ordinary differential equation.

3.1 Heat Equation for an Infinite Rod

Consider the Initial Value Problem (IVP)
ut (x, t) − kuxx = 0, − ∞ < x < ∞, t > 0, k > 0 (7)
u(x, 0) = f (x) (8)
Let ∫ ∞
1
F[u(x, t)] = U (µ, t) = √ u(x, t)e−iµx dx,
2π −∞
be the Fourier transform of u(x, t) with respect to the x-variable, and let
∫ ∞
−1 1
F [U (µ, t)] = u(x, t) = √ U (µ, t)eiµx dµ,
2π −∞
be its inversion formula.
Taking Fourier transform of equation (7),
F[ut (x, t)] − kF[uxx (x, t)] = F[0],
since F is a linear operator, we obtain
∂
U (µ, t) + kµ2 U (µ, t) = 0 (9)
∂t
Equation (9) for a fixed value µ is an ordinary differential equation in t.
Solving (9), we obtain
U (µ, t) = A(µ)e−kµ t , A(µ) arbitary function.
2

Taking Fourier transform of the initial condition and setting t = 0, we get

U (µ, 0) = A(µ) = F (µ)
Therefore
U (µ, t) = F (µ)e−kµ
2t
(10)
To determine u(x, t), we use the inversion formula
∫ ∞
u(x, t) = U (µ, t)eiµx dµ
∫−∞
∞
F (µ)e−kµ t eiµx dµ,
2
=
−∞
∫ ∞
1
where F (µ) = √ f (y)e−iµy dy, thus
2π −∞
∫ ∞∫ ∞
1
f (y)e−iµ(y−x) e−kµ t dydµ.,
2
u(x, t) = √
2π −∞ −∞

13
3.2 Half Range Problems
We now consider applications that involve the use of Fourier sine or cosine transforms

1. Fourier sine/cosine transforms can be used when the transformed variable is re-
stricted to a semi-infinite interval.

2. The choice of the cosine-sine transform is dictated by the boundary condition

Recall
√
′′ 2 ′
Fc [f (x) ] = − f (0) − µ2 Fc (µ) (11)
π
√
′′ 2
Fs [f (x) ] = µf (0) − µ2 Fs (µ) (12)
π
One can easily prove the following
√
2
Fc [uxx (x, t)] = − ux (0, t) − µ2 Uc (µ) (13)
π
√
2
Fs [uxx (x, t)] = µUs (0, t) − µ2 Us (µ) (14)
π
Now we not that [ ]
∂u(x, t) ∂Uc (µ, t)
Fc =
∂t ∂t
and [ ]
∂u(x, t) ∂Us (µ, t)
Fs =
∂t ∂t
Remarks:

1. From equation (13) we note that the Fourier cosine transform is useful when ux (0, t)
is defined in the problem.

2. From equation (14) we note that Fourier sine transform is useful when u(0, t) is
defined in the problem.

Example

Consider the diffusion equation (Heat Equation)

ut (x, t) − kuxx (x, t) = 0, 0 < x < ∞, t > 0 (15)

I.C u(x, 0) = f (x), 0 < x < ∞, t > 0 (16)
B.C u(0, t) = g(t), t > 0. (17)

14
Solution

We choose the Fourier sine transform in x. Taking the Fourier sine transform of equation
(15) we obtain
[√ ]
∂Us (µ, t) 2
−k µUs (0, t) − µ2 Us (µ) = 0
∂t π
√
∂Us (µ, t) 2
⇒ + kµ2 Us (µ) − kµU (0, t) = 0
∂t π
√
∂Us (µ, t) 2
⇒ 2
+ kµ Us (µ) = kµg(t) (18)
∂t π
Solving (18) we obtain
√ ∫
2 −kµt
g(t)ekµ t dt + A(µ)e−kµ t
2 2
Us (µ, t) = kµe
π
√ ∫
2
kµ g(t)e−kµ (τ −t) dt + A(µ)e−kµ t
2 2
=
π
setting t = 0, we obtain
Us (µ, 0) = A(µ) = Fs (µ)
√ ∫
−kµ2 t 2
kµ g(t)e−kµ (τ −t) dt
2
⇒ Us (µ, t) = Fs (µ)e +
π

The inversion formula of Us (µ, t) yields the solution u(x, t) as

∫ ∞
u(x, t) = Us (µ, t) sin µxdµ
0
∫ ∞ √ ∫ ∞∫
−kµ t 2
g(t)e−kµ (τ −t) sin µxdµ.
2 2
= Fs (µ)e sin µxdµ + kµ
0 π 0

15