Beruflich Dokumente
Kultur Dokumente
• In our case:
– electron density vs. diffraction pattern
What is a Fourier transform?
• A function can be described by a summation
of waves with different amplitudes and
phases.
Fourier Transform
H ( f ) = h(t ) exp2ift dt
−
h(t ) = H ( f ) exp− 2ift df
−
If h(t) is real: H (− f ) = H ( f )
*
Discrete Fourier Transforms
• Function sampled at N discrete points
– sampling at evenly spaced intervals
– Fourier transform estimated at discrete values:
hn = h(n ) N N n
n=− ,..., fn =
n = ..., −3,−2,−1,0,1,2,3,... 2 2 N
– e.g. Images
• Almost the same symmetry properties as the
continuous Fourier transform
DFT formulas
N −1
H ( f n ) = h(t ) exp2if nt dt hk exp2if nt n
− k =0
N −1
= hk exp2ikn / N
k =0
N −1
H n hk exp2ikn / N H ( f n ) H n
k =0
N −1
1
hk = H n exp− 2ikn / N
N n =0
Examples
Properties of Fourier Transforms
• Convolution Theorem
• Correlation Theorem
• Wiener-Khinchin Theorem (autocorrelation)
• Parseval’s Theorem
Convolution
As a mathematical formula:
=
Convolution illustrated
Convolution Theorem
Gauss function:
B = 8 u 2 2
Convolution with a delta function
The delta function:
C ( g , g ) G ( f )G ( f ) *
2
C ( g , g ) = G( f )
Calculation of the electron density
F ( S ) = f j exp 2ir j S
j
F (S) = (r ) exp2ir
cell
j
S dv
0<x<1
Calculation of the electron density
F (S) = (r ) exp2ir
cell
j
S dv
dv = V dxdydz
= hx + kl + yz
r S = (a x + b y + c z ) S = a S x + b S y + c S z
1 1 1
F (hkl) = V ( xyz) exp2i(hx + ky + lz)dxdydz
x =0 y =0 z =0
Calculation of the electron density
1 1 1
F (hkl) = V ( xyz) exp2i(hx + ky + lz)dxdydz
x =0 y =0 z =0
1
( xyz) = F (hkl) exp− 2i(hx + ky + lz)
V h k l
Suggested reading
• http://www.yorvic.york.ac.uk/~cowtan/fouri
er/fourier.html and links therein
• http://www.bfsc.leidenuniv.nl/ for the
lecture notes