occupations out of bounds in wfn

[Yaze Wu]

Dear Developers

I'm getting the error message "occupations out of bounds in wfn" when I try to run "wfn_rho_vxc_info.x WFN_out". Here, WFN_out is the Binary file converted from WFN_out.h5, which is the output from parabands.x. The conversion is performed by "hdf2wfn.x BIN WFN_out.h5 WFN_out"

However, when I run "wfn_rho_vxc_info.x WFN_out.h5", I am able to obtain sensible output.

This problem by itself if benign; however when I'm doing CSI subsampling, setup_subsampling_csi.x does not accept hdf5 format for the wfn file, and "setup_subsampling_csi.x WFN_out .........." runs into the "same occupations out of bounds in wfn" error.

I did some tracing back in the source code and the error message comes from line 396 of wfn_rho_vxc_io_inc.f90:

      if(any(occupations(1:nbands, 1:nk, 1:ns) < -0.5 .or. occupations(1:nbands, 1:nk, 1:ns) > 1.5)) then
        call die("occupations out of bounds in wfn")

Further checking of the output from "wfn_rho_vxc_info.x WFN_out.h5" reveals that certain elements of the output were in fact "*********" or "NaN". I'm not sure if this is acceptable or already a sign of trouble. Also, I'm curious how after processing by "hdf2wfn.x, WFN_out cannot pass the if test.

May I know if there's any way I can resolve/circumvent this issue? Also, is this related to the fact that my pw.x calculation was run with spin orbit coupling considerations turned on? If so, should I forget about SOC altogether or is there still some way to run the GW+BSE calculations with SOC?

Many thanks

Yaze

[Yaze Wu]

In the case of the original post which encountered the abovementioned issue, I ran the parabands calculation with pseudobands. I've just rerun the parabands calculation without pseudobands and the problem disappeared.

So is this some problem with the pseudobands routine?

I'll let everyone know if the wavefunction files with pseudobands can still be used at Epsilon when I come to that.

Best

Yaze