Numerical Methods in Quantum Mechanics - Dipartimento di Fisica
Numerical Methods in Quantum Mechanics - Dipartimento di Fisica
Numerical Methods in Quantum Mechanics - Dipartimento di Fisica
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
Fast Fourier Transform<br />
In the simple case that will be presented, the matrix elements of the Hamiltonian,<br />
Eq.(9.13), can be analytically computed by straight <strong>in</strong>tegration. Another<br />
case <strong>in</strong> which an analytic solution is known is a crystal potential written as a<br />
sum of Gaussian functions:<br />
This yields<br />
V (x) =<br />
N−1 ∑<br />
p=0<br />
〈b i,k |V |b j,k 〉 = 1 a<br />
v(x − pa), v(x) = Ae −αx2 . (9.14)<br />
∫ L/2<br />
−L/2<br />
Ae −αx2 e −iGx dx (9.15)<br />
The <strong>in</strong>tegral is known (it can be calculated us<strong>in</strong>g the tricks and formulae given<br />
<strong>in</strong> previous sections, extended to complex plane):<br />
∫ ∞<br />
√ π<br />
e −αx2 e −iGx /4α<br />
dx =<br />
α e−G2 (9.16)<br />
−∞<br />
(remember that <strong>in</strong> the thermodynamical limit, L → ∞).<br />
For a generic potential, one has to resort to numerical methods to calculate<br />
the <strong>in</strong>tegral. One advantage of the plane-wave basis set is the possibility to<br />
exploit the properties of Fourier Transforms (FT), <strong>in</strong> particular the Fast Fourier<br />
Transform (FFT) algorithm.<br />
Let us <strong>di</strong>scretize our problem <strong>in</strong> real space, by <strong>in</strong>troduc<strong>in</strong>g a grid of n po<strong>in</strong>ts<br />
x i = ia/n, i = 0, n − 1 <strong>in</strong> the unit cell. Note that due to perio<strong>di</strong>city, grid po<strong>in</strong>ts<br />
with <strong>in</strong>dex i ≥ n or i < 0 are “refolded” <strong>in</strong>to grid po<strong>in</strong>ts <strong>in</strong> the unit cell (that<br />
is, V (x i+n ) = V (x i ), and <strong>in</strong> particular, x n is equivalent to x 0 . Let us <strong>in</strong>troduce<br />
the function f j def<strong>in</strong>ed as follows:<br />
f j = 1 ∫<br />
V (x)e −iGjx dx, G j = j 2π , j = 0, n − 1. (9.17)<br />
L<br />
a<br />
Apart from factors, this is the usual def<strong>in</strong>ition of FT (and is noth<strong>in</strong>g but matrix<br />
elements). We can exploit perio<strong>di</strong>city to show that<br />
f j = 1 a<br />
∫ a<br />
0<br />
V (x)e −iG jx dx. (9.18)<br />
Note that the <strong>in</strong>tegration limits can be translated at will, aga<strong>in</strong> due to perio<strong>di</strong>city.<br />
Let us write now such <strong>in</strong>tegrals as a f<strong>in</strong>ite sum over grid po<strong>in</strong>ts (with<br />
∆x = a/n as f<strong>in</strong>ite <strong>in</strong>tegration step):<br />
f j = 1 a<br />
= 1 n<br />
= 1 n<br />
n−1 ∑<br />
m=0<br />
n−1 ∑<br />
m=0<br />
n−1 ∑<br />
m=0<br />
V (x m )e −iG jx m<br />
∆x<br />
V (x m )e −iG jx m<br />
V m exp[−2π jm n i], V i ≡ V (x i). (9.19)<br />
71