1256 B Some Useful Mathematics
integrals converge. For the integral of Eq. (B-118) to converge, the following integral
must converge:
∫∞
−∞
|f(x)|^2 dx <∞ (B-119)
We say that the functionf(x) must besquare integrable. The functionf(x) must
approach zero asx→−∞and asx→∞to be square integrable. If the Fourier
transformF(k) exists, it will also be square integrable.
If the functionf(x) is an even function, its Fourier transform is aFourier cosine
transform:
F(k)
1
√
2 π
∫∞
−∞
f(x)cos(kx)dx
√
2
π
∫∞
0
f(x)cos(kx)dx (B-120)
The second version of the transform is called aone-sided cosine transform.Iff(x)is
an odd function, its Fourier transform is aFourier sine transform:
F(k)
1
√
2 π
∫∞
−∞
f(x)sin(kx)dx
√
2
π
∫∞
0
f(x)sin(kx)dx (B-121)
There is a useful theorem for the Fourier transform of a product of two func-
tions, called theconvolution theoremor theFaltung theorem(Faltungis German for
“folding”). Theconvolutionof two functionsf(x) andg(x) is defined as the integral
1
√
2 π
∫∞
−∞
f(x)g(x−y)dy (B-122)
This integral is a function ofx, and its Fourier transform is equal toF(k)G(k) where
F(k) is the Fourier transform off(x) andG(k) is the Fourier transform ofg(x).^1
Since the Fourier transform is nearly the same going in both directions, the analogous
convolution
1
√
2 π
∫∞
−∞
F(1)G(k−1)dl (B-123)
has as its Fourier transform the productf(x)g(x).
(^1) Philip M. Morse and Herman Feshbach,Methods of Theoretical Physics, Part I, McGraw-Hill,
New York, 1953, p. 464ff.