Fourier Series.: V IV JV KV I J I I

FOURIER SERIES
Waves of any shape are possible. Harmonic waves are just a particular case.
However, they are an especially interesting case because any wave can be
represented as a sum of harmonic waves. This representation is called the
Fourier series.
The basic idea is already familiar to you from your study of vectors. You
know that any vector can be represented as a sum of basis vectors. In three
dimensions, these are sometimes called i, j, and k. Then we can write
v = iv1 + jv2 + kv3 . The basis vectors are orthonormal; for example, i ⋅ j = 0
and i ⋅ i = 1 . We will generalize this idea in two steps---first to vector spaces
of arbitrary dimension and then to function spaces, which are vector spaces
of infinite dimension.
Consider a vector space of dimension d. The unit vectors will now be

called e i with i = 1, ,d . They are orthonormal so that e i ⋅ ej = ij . The
Kronicker delta on the right hand side is defined to be one if i and j are the
same and zero otherwise. Now an arbitrary vector can be written
d
v = ∑ e i vi . We say that the basis vectors are complete when any vector can
i=1
be written as such a sum. Conversely, we can get the components of v via

vi = e i ⋅ v . These are the two properties that we need of a basis---
orthonormality and completeness.
The second, and much bigger step, is to generalize this idea from a space of
vectors to a space of functions. The case we are working on is the waves on
a string of finite length L with the boundary condition that the string is
clamped at the ends. So the function space is all the possible shapes of the
string---all functions f(x) on [0,L] with f(0)=f(L)=0. One such function is
the analogue of a vector in the previous discussion. We would like to have
a basis of functions so that any such function f can be represented as a sum
over the basis functions with some coefficients. As it happens, a very nice
orthonormal and complete basis is the harmonic, standing waves that we
have already discussed! Let’s see how that works.
Recall that the spatial part of the standing waves is given by the functions
sin(n x L) . These are essentially the basis functions. However, to save
writing and to make these orthonormal, it is convenient to define the basis
functions via
2 n x
Sn ( x) = sin .
L L
Direct integration gives the result
L
∫ dxS m ( x )Sn ( x ) = mn .
0
This is the generalization to functions of the notion of orthonormal for

vectors. If m and n are not the same, the integral is zero, and we say that
the functions are orthogonal. If m and n are the same, then the integral is
one, and we say that the functions are normalized to one. Since n can be
any positive integer, there are an infinite number of basis functions. We
say that the vector space of functions is infinite dimensional.
Just as the basis vectors were complete, so too are the basis functions
complete. That means that any function f(x) that satisfies the boundary
conditions f(0)=f(L)=0 can be written as a sum over the basis functions
with coefficients f n .
∞ ∞ ∞
2  n x  n x
f ( x) = ∑ fn Sn ( x) = ∑ fn L
sin = ∑
L  n =1
cn sin L .
n= 1 n =1
In the last form, cn = fn L2 is introduced for convenience. The proof of this

is a little too much to put here. If you would like to discuss it, please come
see me. At this point, it’s not clear that we have done anything useful since
the coefficients have not been given. However, with f(x) given, the
coefficients can be obtained by an integral with the basis functions, which is
the analogue of finding the components of a vector by computing its dot
product with a basis vector.
L L
f n = ∫ dx Sn (x ) f ( x) = 2 dx sin n x  f ( x)
L∫  L 
0 0
or
( )
L
cn = 2L ∫ dx sin nL x f ( x) .
0
Thus, if we know f(x), then we can find the cn from this inversion formula
and then use them to give the representation of f as a Fourier series. This
can be very useful because the harmonic basis functions are simple and
have very nice properties.
Now with this new tool, let’s return to the physics problem at hand. The
problem is to describe all the possible wave forms on the string of length L
with the ends clamped. The standing wave solutions are of the form
( )
cos( nt )sin nLx . The angular frequency is n = v nL . You can easily check ( )
( )
that another solution is sin( nt )sin nLx . The superposition principle tells us
that any linear combination of these
D(x,t ) = ∑[ an cos(
n
n t ) + bn sin( n ( )
t)] sin nLx = ∑c ( t) sin( nLx )
n
n
[where cn (t ) ≡ an cos( nt ) + bn sin( n t ) ] is also a solution. Thus the function

D(x,t) is a possible motion of the string, and at each t, this wave is a
fourier series. The fourier coefficients cn (t ) change with time as indicated.
Since the basis functions are complete, any wave can be represented in this
form! This includes all kinds of complicated solutions that don’t look
anything like standing waves. Nevertheless, we now know that they are
linear combinations of standing waves.
This is great except for the problem that we now have a huge number of
solutions to the wave equation. The problem manifests itself in the
constants an ,bn . Any values for those give a valid solution D. The problem
is resolved by using additional information. The constants are determined
by the initial conditions, which are given in the two functions that give the
shape of the string and the velocity of each piece of the string at t=0: D( x,0 )
and D(x,t ) . Setting t=0 in the form above or differentiating with
t t=0
respect to t and than setting t=0 gives
n x D( x,0 )  n x
D(x,0) = ∑ a n sin = ∑ n bn sin
 L   L 
and .
n t n
Now the a n and the bn can be determined by using the inversion formula
given above
 n x  D( x,0 ) .
L L
2  n x 1 2
an = ∫ dx sin ∫ dx sin  
 D( x,0 ) and bn =
L0 L n L0 L t
Thus, if you know the initial conditions, you can do these integrals to get
the an and b n coefficients. Then they are put back into the general
expression for D(x,t) to give the evolution of the disturbance with time.
To close, I would like to emphasize that this method of fourier series is
very general. It can be used to advantage in many other problems. You will
certainly encounter it many more times in your studies.
FOURIER TRANSFORM
A closely related method is the fourier transform a.k.a. the fourier

integral. It is the appropriate representation when the length of the interval
L becomes infinite, i.e. we return to the case of waves on an infinite string.
In that limit, the wave numbers k n = nL become very closely spaced. The
sum on n becomes an integral on k . The representation of D(x,t) is as an
integral over all the harmonic waves. This can be written in terms of trig
functions, but it is more convenient and compact to use the complex
exponential notation, which comes from the identity e i = cos + isin . (If
you do not recall this, or otherwise want more info on complex numbers,
see sections 11.6 and 11.7 of your calculus text by Stein and Barcellos, 5th
ed.)
[ ]
D(x,t ) = ∫ dk a(k )ei t + b( k )e−i t eikx with = (k ) = vk .
This looks a lot different from the trig functions. You can verify that it is
equivalent by using the identity to write out the exponentials. I do not
recommend this. It is much easier to just stick this form into the wave
equation and verify that it is a solution. As with the fourier series, this
representation has wide application in mathematics, physics, other
quantitative sciences, and engineering.

Fourier Series.: V IV JV KV I J I I

Uploaded by

Copyright:

Available Formats

Fourier Series.: V IV JV KV I J I I

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Fourier Series.: V IV JV KV I J I I

Uploaded by

Copyright:

Available Formats

FOURIER SERIES

Consider a vector space of dimension d. The unit vectors will now be

be written as such a sum. Conversely, we can get the components of v via

This is the generalization to functions of the notion of orthonormal for

In the last form, cn = fn L2 is introduced for convenience. The proof of this

[where cn (t ) ≡ an cos( nt ) + bn sin( n t ) ] is also a solution. Thus the function

respect to t and than setting t=0 gives

A closely related method is the fourier transform a.k.a. the fourier

You might also like