D.8 Completeness of Fourier modes

The purpose of this note is to show completeness of the
Fourier modes

for describing functions that are periodic of period . It is to be shown that

allthese functions can be written as combinations of the Fourier modes above. Assume that is any reasonable smooth function that repeats itself after a distance , so that . Then you can always write it in the form

or

for short. Such a representation of a periodic function is called a “Fourier series.” The coefficients are called “Fourier coefficients.” The factors 1 can be absorbed in the definition of the Fourier coefficients, if you want.

Because of the Euler formula, the set of exponential Fourier modes
above is completely equivalent to the set of real Fourier modes

so that -periodic functions may just as well be written as

The extension to functions that are periodic of some other period than
is a trivial matter of rescaling . For a period
, with any half period, the exponential Fourier
modes take the more general form

and similarly the real version of them becomes

See [40, p. 141] for detailed formulae.

Often, the functions of interest are not periodic, but are required to be zero at the ends of the interval on which they are defined. Those functions can be handled too, by extending them to a periodic function. For example, if the functions relevant to a problem are defined only for 0 and must satisfy 0, then extend them to the range 0 by setting and take the range to be the period of a -periodic function. It may be noted that for such a function, the cosines disappear in the real Fourier series representation, leaving only the sines. Similar extensions can be used for functions that satisfy symmetry or zero-derivative boundary conditions at the ends of the interval on which they are defined. See again [40, p. 141] for more detailed formulae.

If the half period becomes infinite, the spacing between the discrete values becomes zero and the sum over discrete values turns into an integral over continuous values. This is exactly what happens in quantum mechanics for the eigenfunctions of linear momentum. The representation is now no longer called a Fourier series, but a “Fourier integral.” And the Fourier coefficients are now called the “Fourier transform” . The completeness of the eigenfunctions is now called Fourier’s integral theorem or inversion theorem. See [40, pp. 190-191] for more.

The basic completeness proof is a rather messy mathematical
derivation, so read the rest of this note at your own risk. The fact
that the Fourier modes are orthogonal and normalized was the subject
of various exercises in chapter 2.6 and will be taken for
granted here. See the solution manual for the details. What this
note wants to show is that any arbitrary periodic function
of period that has continuous first and second order
derivatives can be written as

in other words, as a combination of the set of Fourier modes.

First an expression for the values of the Fourier coefficients
is needed. It can be obtained from taking the inner product
between a
generic eigenfunction and the
representation for function above. Noting that all the inner
products with the exponentials representing will be zero except
the one for which , if the Fourier representation
is indeed correct, the coefficients need to have the values

a requirement that was already noted by Fourier. Note that and are just names for the eigenfunction number and the integration variable that you can change at will. Therefore, to avoid name conflicts later, the expression will be renotated as

Now the question is: suppose you compute the Fourier coefficients
from this expression, and use them to sum many terms of the
infinite sum for , say from some very large negative value
for to the corresponding large positive value ;
in that case, is the result you get, call it ,

a valid approximation to the true function ? More specifically, if you sum more and more terms (make bigger and bigger), does reproduce the true value of to any arbitrary accuracy that you may want? If it does, then the eigenfunctions are capable of reproducing . If the eigenfunctions are not complete, a definite difference between and will persist however large you make . In mathematical terms, the question is whether .

To find out, the trick is to substitute the integral for the
coefficients into the sum and then reverse the order of
integration and summation to get:

The sum in the square brackets can be evaluated, because it is a geometric series with starting value and ratio of terms . Using the formula from [40, item 21.4], multiplying top and bottom with , and cleaning up with, what else, the Euler formula, the sum is found to equal

This expression is called the

Dirichlet kernel. You now have

The second trick is to split the function being integrated into the two parts and . The sum of the parts is obviously still , but the first part has the advantage that it is constant during the integration over and can be taken out, and the second part has the advantage that it becomes zero at . You get

Now if you backtrack what happens in the trivial case that is just a constant, you find that is exactly equal to in that case, while the second integral above is zero. That makes the first integral above equal to one. Returning to the case of general , since the first integral above is still one, it makes the first term in the right hand side equal to the desired , and the second integral is then the error in .

To manipulate this error and show that it is indeed small for large
, it is convenient to rename the -independent part
of the integrand to

Using l’Hôpital's rule twice, it is seen that since by assumption has a continuous second derivative, has a continuous first derivative. So you can use one integration by parts to get

And since the integrand of the final integral is continuous, it is bounded. That makes the error inversely proportional to , implying that it does indeed become arbitrarily small for large . Completeness has been proved.

It may be noted that under the stated conditions, the convergence is uniform; there is a guaranteed minimum rate of convergence regardless of the value of . This can be verified from Taylor series with remainder. Also, the more continuous derivatives the -periodic function has, the faster the rate of convergence, and the smaller the number of terms that you need to sum to get good accuracy is likely to be. For example, if has three continuous derivatives, you can do another integration by parts to show that the convergence is proportional to 1 rather than just 1/. But watch the end points: if a derivative has different values at the start and end of the period, then that derivative is not continuous, it has a jump at the ends. (Such jumps can be incorporated in the analysis, however, and have less effect than it may seem. You get a better practical estimate of the convergence rate by directly looking at the integral for the Fourier coefficients.)

The condition for to have a continuous second derivative can be relaxed with more work. If you are familiar with the Lebesgue form of integration, it is fairly easy to extend the result above to show that it suffices that the absolute integral of exists, something that will be true in quantum mechanics applications.