LECTURE 42

FOURIER SERIES PART I

Suppose that a function f has period T = 2. Then f may be approximated by the Fourier
series
f(x) = a
0
+

n=1
(a
n
cos nx + b
n
sin nx)
where the Fourier coecients a
0
, a
n
and b
n
are given by
a
0
=
1
2

f(x) dx
a
n
=
1

f(x) cos nxdx (n = 1, 2, 3, . . . )

b
n
=
1

f(x) sin nxdx (n = 1, 2, 3, . . . )

.
Suppose that we are dealing with a periodic function f of period 2. Periodic functions
are complex in their own special way. We saw earlier that the innite repetition made
the calculation of their Laplace transforms quite a drama. By using Fourier series we
can bust such functions up as a sum of much simpler periodic trigonometric components
sin(nx) and cos(nx). This is of enormous value when dealing with both dierential and
partial dierential equations.
In this lecture we will start o with the simpler case of functions with period T = 2.
Later on the theory will be generalised to functions of arbitrary period T.
We begin with the periodic function
f(x) =

x, 0 x < ;
, x < 2.
f(x + 2) otherwise
This is (by denition) is a function of period 2 and the graph over 6 x 6 looks
like
3
2
0
2.5
1.5
15 10 5 0 -10
0.5
1
-5 -15
1
The graphs of sin x and cos x are also of period T = 2:
1
0
0.5
-0.5
-1
x
10 5 0 -5 15 -10 -15
1
0
0.5
-0.5
-1
x
15 10 5 -15 -10 0 -5
It is fairly clear that there is no way that f could possibly be approximated by taking
linear combinations of sin(x) and cos(x)....... f looks nothing at all like these two functions!
But we have a trick up our sleeves! We can make the trig functions busier (by reducing
their periods). Consider the graphs of sin 3x and cos 3x:
1
0
0.5
15
-0.5
x
10 0
-1
-15 -10 -5 5
1
0
0.5
15
-0.5
x
10 0
-1
-15 -10 -5 5
(Note that in general the period of sin nx and cos nx is T =
2
n
).
What we then do to create a Fourier series is to use all possible functions sin nx and
cos nx together with a constant term a
0
to approximate f as an innite series:
f(x) = a
0
+

n=1
(a
n
cos nx + b
n
sin nx)
2
The coecients a
0
, a
n
and b
n
are referred to as the Fourier coecients. The bad news
is that their calculation is usually a tedious process involving the integral formulae
a
0
=
1
2

f(x) dx
a
n
=
1

f(x) cos nxdx (n = 1, 2, 3, . . . )

b
n
=
1

f(x) sin nxdx (n = 1, 2, 3, . . . )

.
We will prove these formulae in the next lecture but for today lets look at what needs
to be done to calculate the Fourier series of our function.
Example 1 Find the Fourier series of the function f above on page 1.
Some equations (amongst others) that you must have at your ngertips for Fourier series
are:

cos nxdx =
1
n
sin nx + C

sin nxdx =
1
n
cos nx + C
sin n = 0
cos n = (1)
n
sin(x) = sin x
cos(x) = cos x
sin 0 = 0
cos 0 = 1
3
4
a
0
=
3
4
a
n
=
(1)
n
1
n
2

b
n
=
1
n

f(x) =
3
4
+

n=1

(1)
n
1
n
2

cos nx +

1
n

sin nx
f(x) =
3
4
+
2

cos x sin x
1
2
sin 2x
2
9
cos 3x
1
3
sin 3x + . . .
Observe that both a
n
and b
n
tend to zero as n . This always happens under normal
circumstances and guarantees that the Fourier series will converge for standard functions.
5
It is fascinating to look at how the Fourier series steps up in accuracy as the number
of terms increases. The three graphs below show the rst partial sum
3
4
+
2

cos x sin x
the third partial sum
3
4
+
2

cos x sin x
1
2
sin 2x
2
9
cos 3x
1
3
sin 3x
and nally the tenth partial sum.
3.5
2.5
0.5
3
2
15 -5
1
1.5
5 0 -15
0
10 -10
3.5
2.5
0.5
3
2
15 10
1
1.5
-5 5
0
-10 -15 0
3
2
0
2.5
1.5
15 10
0.5
1
-15 5 -5 0 -10
If we were to take innitely many terms then the Fourier series would sit right over the
original function. Observe that the nth partial sums are continuous and yet they do a ne
job of approximating discontinuous objects. We will examine carefully at a later stage
what actually happens at the points of discontinuity, a situation known as the Gibbs
phenomenon.
It is fair to say that the construction of a Fourier series is a major undertaking, but
the payos are huge. In the next lecture we will still restrict our attention to functions
with a period of 2 but will develop some shortcuts which will help us out on occasion.
42
You can now do Q 108
6