Wavefunction is not normalized in file wfn%cg

Submitted by xuanfy on Thu, 08/24/2017 - 03:40

Forums 

User questions

Hi all,

I just finished running the example of GaAs of BGW, with coarse kgrid 2 by 2 by 2 and fine kgrid 4 by 4 by 4.

When I tried to change the coarse kgrid to 5 by 5 by 5 and fine kgrid to 10 by 10 by 10, the absorption code showed error. Below are parts of the absorption.out where error messages appeared:

Coarse-grid wavefunctions read from file WFN_co:
- Number of k-points in irreducible BZ: 14
- Number of k-points in full BZ: 125
Wavefunction is not normalized in file wfn%cg
Wavefunction is not normalized in file wfn%cg
abs(norm - 1) = 0.0000045
iband = 1 ispin = 1 ik = 4

abs(norm - 1) = 0.0000041
iband = 1 ispin = 1 ik = 3

Wavefunction is not normalized in file wfn%cg
abs(norm - 1) = 0.0000023
iband = 4 ispin = 1 ik = 2

Started wavefunction interpolation with 125 k-point(s) at 18:21:11.
[ 18:21:11 | 0% ] k-point 1 / 125.
Wavefunction is not normalized in file wfn%cg
Wavefunction is not normalized in file wfn%cg
Wavefunction is not normalized in file wfn%cg
Wavefunction is not normalized in file wfn%cg
abs(norm - 1) = 0.0000086
iband = 1 ispin = 1 ik = 7

abs(norm - 1) = 0.0000061
abs(norm - 1) = 0.0000049
iband = 1 ispin = 1 ik = 8

abs(norm - 1) = 0.0000094
iband = 1 ispin = 1 ik = 5

iband = 1 ispin = 1 ik = 6

It seems like the wavefunctions from WFN_co are not normalized. Anyone knows how to solve it?
Thanks !

Regards,
Fengyuan

jdeslip's picture

Submitted by jdeslip on Thu, 08/24/2017 - 14:33

Hi Fengyuan,

What code did you use to generate the wavefunctions? Quantum ESPRESSO? Can you try try increasing the cutoffs or convergence parameters?

You can also use "dont_check_norm" input flag. But, if your wavefuntion norm is off it could be a sign of something more off in the files.

Submitted by xuanfy on Mon, 08/28/2017 - 20:00

Hi,

Thank you for the reply. Yes, I use QuantumEspresso. I finally succeed running the absorption code.

The k points for QE bands calculation on shifted and unshifted fine grid (in my case 10*10*10) was before generated by kgrid.x which used symmetry. The shifted k grid has 220 k points in the irreducible BZ while unshifted k grid has 73 k points. Therefore I set "use_symmetries_shifted_grid" and "use_symmetries_fine_grid" in absorption.inp. In the output absorption.out, it showed that unshifted k grid unfolded to 1000 k points which is correct, but shifted k grid unfolded to 4000 k points which is wrong. So I used fullbz_replace in absorption.inp and got the corrected number of k points for both shifted and unshifted kgrid, but with the error message I posted earlier. After I increased the accuracy in the QE calculation, the error remained. Then I read the manual carefully and found that I should NOT use fullbz_replace. The 4000 k points unfolded from the shifted k grid could be due to the different symmetry between shifted and unshifted k grid. The absorption code might by default treat them with the same symmetry operations.

So I removed "fullbz_replace" and generated a shifted k grid without using any symmetry, which has 1000 k points. Then I set "use_symmetries_fine_grid" and "no_symmetries_shifted_grid" in the absorption.inp and run the QE bands calculation on the fine shifted kgrid again. This time the absorption code succeeded.