resetting occupations of shifted grid according to Fermi level

[dou bei]

Dear BGW developers,

I got the below warning when doing epsilon calculation with the subsampling technique.

WARNING: resetting occupations of shifted grid according to Fermi level.
Make sure your shifted and unshifted WFNs are consistent. The following kpoints had their occupations reset (spin, kpoint, original band, energy, reset band, energy):
1   12   148  -4.556050   147     -4.578583

my system is metallic with a small indirect band overlapping near Fermi level according to QE DFT calculations. I used "occupations = 'smearing', degauss = 1.0D-8, smearing='gaussian' " when calculating WFN and WFNq. so this warning might be caused by some other issues according to the previous post in the forum.

my input file for epsilon_subsampling calculation is:

epsilon_cutoff 15.0
number_bands 501
cell_slab_truncation
frequency_dependence 0
degeneracy_check_override
subsample
begin qpoints
4.811102509182485E-03 0.000000000000000E+00 0.000000000000000E+00 1.0 1
1.924441003672994E-02 0.000000000000000E+00 0.000000000000000E+00 1.0 1
4.329992258264238E-02 0.000000000000000E+00 0.000000000000000E+00 1.0 1
7.697764014691977E-02 0.000000000000000E+00 0.000000000000000E+00 1.0 1
0.000000000 0.166666667 0.000000000 1.0 0
0.000000000 0.333333333 0.000000000 1.0 0
0.000000000 0.500000000 0.000000000 1.0 0
0.000000000 0.666666667 0.000000000 1.0 0
0.000000000 0.833333333 0.000000000 1.0 0
0.166666667 0.000000000 0.000000000 1.0 0
0.166666667 0.166666667 0.000000000 1.0 0
0.166666667 0.333333333 0.000000000 1.0 0
0.166666667 0.500000000 0.000000000 1.0 0
0.166666667 0.666666667 0.000000000 1.0 0
0.166666667 0.833333333 0.000000000 1.0 0
0.333333333 0.000000000 0.000000000 1.0 0
0.333333333 0.166666667 0.000000000 1.0 0
0.333333333 0.333333333 0.000000000 1.0 0
0.333333333 0.500000000 0.000000000 1.0 0
0.333333333 0.666666667 0.000000000 1.0 0
0.333333333 0.833333333 0.000000000 1.0 0
0.500000000 0.000000000 0.000000000 1.0 0
0.500000000 0.166666667 0.000000000 1.0 0
0.500000000 0.333333333 0.000000000 1.0 0
0.500000000 0.500000000 0.000000000 1.0 0
0.500000000 0.666666667 0.000000000 1.0 0
0.500000000 0.833333333 0.000000000 1.0 0
0.666666667 0.000000000 0.000000000 1.0 0
0.666666667 0.166666667 0.000000000 1.0 0
0.666666667 0.333333333 0.000000000 1.0 0
0.666666667 0.500000000 0.000000000 1.0 0
0.666666667 0.666666667 0.000000000 1.0 0
0.666666667 0.833333333 0.000000000 1.0 0
0.833333333 0.000000000 0.000000000 1.0 0
0.833333333 0.166666667 0.000000000 1.0 0
0.833333333 0.333333333 0.000000000 1.0 0
0.833333333 0.500000000 0.000000000 1.0 0
0.833333333 0.666666667 0.000000000 1.0 0
0.833333333 0.833333333 0.000000000 1.0 0
end

I can show more info if needed to identify the issue. I appreciate your help.

Thanks,

dou bei

[Zhenglu Li]

Hi Dou Bei,

It seems that the code is doing the following: it determines the Femi energy from unshifted WFN, and changes the occupation of WFNq using the already-determined Fermi energy.  This should be acceptable in your case if everything is correct since your case is a semimetal and the shifted WFN may have a "different" Fermi energy due to discretized sampling of BZ.  If the relevant energies are very close to the reported Fermi energy, it should be fine.

However, on the other hand, for metal, we recommend you compute epsilon 0 with a separate calculation alone, using a much finer uniform WFN grid, and set the q0 to be the smallest spacing of that WFN, and link WFNq to WFN.  Also, for metallic q0 in epsilon calculation, is_q0 should be 2, and you do not need to use subsampling.  Furthermore, physically, for metal q0, you do not need many bands, probably just few tens of empty bands are more than enough, and this alleviates heavy workload from the dense k point sampling. 

Best,

Zhenglu

--
You received this message because you are subscribed to the Google Groups "BerkeleyGW Help" group.
To unsubscribe from this group and stop receiving emails from it, send an email to help+uns...@berkeleygw.org.
To view this discussion on the web visit https://groups.google.com/a/berkeleygw.org/d/msgid/help/95510d60-889d-4db6-903b-27e2a75de189n%40berkeleygw.org.