Dear BerkeleyGW developers,
I'm currently trying to perform metallic GW calculation. Since the material is metallic on the DFT level, epsmat.h5 (on a 20 20 1 kgrid) and eps0mat.h5 are calculated in two steps (on a 40 40 1 kgrid). Since the k grid is already kind of dense, parabands is used when obtaining WFN files. The epsmat.h5 matrix is obtained by three epsilon runs on three subsets of the 20 20 1 kgrid; epsmat_hdf5_merge.py is then used to generate the full epsmat.h5 file.
Then, in the sigma step, the following error occurs:
epsmat: ekold(isrtold(i)) = 0.0000000000000000 ekin = 1373351376603.6543
ERROR: Incorrect kinetic energies in epsmat.
I wonder what are possible reasons behind it; the output files imply that all the epsilon runs ended successfully (with "Job done" as the last line). Is there some caveats pertaining to epsmat_hdf5_merge.py?
Below is where sigma.out stops:
GPU acceleration is : DISABLED
Algorithms used:
- mtxel : CPU_ALGO
- sigma_gpp : CPU_ALGO
Total number of frequencies in the dielectric matrix: 1
Number of imag. frequencies in the dielectric matrix: 0
Started reading eps0mat.h5 with 1 q-point(s) at 06:57:51.
[ 06:57:51 | 0% ] q-point 1 / 1.
Finished reading eps0mat.h5 at 06:58:16.
Elapsed time: 25 s.
Started reading epsmat.h5 with 120 q-point(s) at 06:58:16.
[ 06:58:16 | 0% ] q-point 1 / 120.
[ 06:58:19 | 2% ] q-point 3 / 120, remaining: 148 s.
Thanks for any help you are willing to offer.
Best,
Jinyuan