You are on page 1of 2

Mathematical Tripos Part II Further Complex Methods

Michaelmas term 2007 Dr S.T.C. Siklos

Fourier Transforms
This handout covers the basic properties of Fourier Transforms. It should be mainly revision, except for the penultimate item, which is about transforms of functions that depend on three variables rather than the usual one. This is not dicult: you just transform with respect to each variable consecutively. Notation We write the Fourier transform of f (x) as f (k) where k is the conjugate variable to x, and dene f (x)eikx dx. f (k) =

Properties 1. Linearity: g (x) = a f 1 (x) + b f 2 (x) = g (k) = a f 1 (k) + b f 2 (k). This just follows from the linearity of the dening integrals. 2. Scaling: g (x) = f (ax) = g(k) = |a|1 f (k/a). This comes from changing variable x ax. If a < 0, the limits change sign, whence the mod signs. 3. Shifting: g (x) = f (x x0 ) = g (k) = eikx0 f (k); g (x) = eik0 x f (x) = g (k) = f (k k0 ). The rst of these requires a change of variable in the dening integral x x x0 . Note that these results are essentially conjugates of each other. 4. Derivatives:

df = g(k) = ik f (k); dx d g (x) = x f (x) = g (k) = i (k) f dk g (x) = The rst of these requires integration by parts. The second cannot be managed by integrating the left hand side; instead, start with the right hand side and dierentiate under the integral sign.

5. Convolution: h = f g = h (k) = f (k) g (k), where f g (x) =


f (u) g (x u)du.

To prove this, just take the Fourier transform of the convolution and use the shifting result from above:

h (k) =

f (u) g (x u)du eikx dx

f (u) g (x u)eikx dxdu

f (u)eiku g (k)du = f (k) g (k)

6. 3-dimensional transform:

f (k) =

eik1 x f (x)dx eik2 y dy eik3 z dz eikr cos f (x)r2 sin dr d d.


all space

eikx f (x)d3 x =
all space

For the last expression, polar coordinates in x-space have been used. The polar direction (North pole) is taken to be parallel to k, so that is the angle between k (considered xed) and x. 7. Transform of a constant (using the theory of distributions, giving a result that is consistent with the inversion theorem): 1 = 2(k) . Inversion Theorem If
| f

(x)|dx < and f (x) is continuous then f (x) = 1 2


eikx (k)dk. f

This is the Fourier representation of f. If f (x) is only piecewise continuous, then at a discontinuity the Fourier representation gives the average value of f . The conditions can be weakened. In three dimensions, we have f (x) = 1 2
3

eikx (k)d3 k, f

all k-space

which we might have to invert using polar coordinates in k-space, taking the polar direction to be parallel to x.