[Wien] eigenfunction integration in the interstitial

asubedi asubedi at gmail.com
Sat Jun 13 23:07:23 CEST 2009


Dear All,

I am interested in calculating various matrix elements between
eigenfunctions, but first I am trying to see if I can calculate the
norm of a eigenfunction. I am able to do the integration in the muffin
tins. I checked the values I get with the numbers in case.qtl, and
they match for each atom. However, I am having problems getting the
right number for the interstitial part. To do the integration in the
interstitial, I perform following steps:

1) Read the K-vectors and corresponding coefficients for a particular
k-point and band number.
2) Set up a FFT grid (using the fft dimensions from case.output2) and
put the coefficients in a FFT array with zero where coeff for a K
point is not specified.
3) Set up step function U using the algorithm specified in rean0.f
4) Fourier transform FFT and U to real space using c3fft.
5) set FFT(i1,i2,i3) = conjg(FFT(i1,i2,i3)) * FFT(i1,i2,i3) * U(i1,i2,i3)
6) Back Fourier transform FFT.
7) Renormalize by dividing by the dimension of FFT and volume of the unit cell.
8) FFT(1,1,1) should be the desired integral.

I think there is something wrong with renormalization because for
degenerate bands, I get the same value for the integral. But
renormalization constant is different for different set of degenerate
bands :-/

To check if the coefficients are stored only for one element of star
of K, I did calculation with no symmetry, too. In particular I did
CsCl with 1_P1 and Pm-3m symmetries. In both cases the number of
coefficients calculated were the same (347 coeffs when RK-max = 7 and
Gmax = 12).

I tried to look how the interstitial part is calculated for case.qtl.
The corresponding code seems to be in outp.f. But it is simply 1 -
spherical part (line 223 in outp.f: TCOUT=100.D0-TC)

I would be extremely grateful if anyone can point me to the right direction.

Thank you,
Alaska


More information about the Wien mailing list