Page 1 of 1

problem in the BSE step of optical calculation

Posted: Fri Apr 15, 2022 8:41 pm
by alok_shukla1
Hi all,
I am trying to perform the BSE GW calculation using the following link:

wiki/index.php/Dielectric_properties_of_Si_using_BSE

Everything is working fine, but last step 5 is not working. I am following the correct procedure, but calculations are stopped while performing the last BSE step. I am not able to figure out why this is happening. Here is the output last part


LDA part: xc-table for Pade appr. of Perdew
found WAVECAR, reading the header
POSCAR, INCAR and KPOINTS ok, starting setup
WARNING: small aliasing (wrap around) errors must be expected
FFT: planning ...
reading WAVECAR
the WAVECAR file was read sucessfully
initial charge from wavefunction
the WAVEDER file was read successfully
energies w=

responsefunction array rank= 176
allocating 0 responsefunctions rank= 176
allocating 0 responsefunctions rank= 176
Doing 0 frequencies in blocks of 0
allocating two-electron 4 orbital integral table 4 8 4 8
BSE (scaLAPACK) single prec attempting allocation of 0.019 Gbyte rank= 6170
BSE setting up matrix
reading now WFULL0001.tmp
reading now WFULL0003.tmp
reading now WFULL0006.tmp
reading now WFULL0008.tmp
reading now WFULL0011.tmp
reading now WFULL0012.tmp
reading now WFULL0007.tmp
reading now WFULL0015.tmp
| reading now WFULL0005.tmp
reading now WFULL0014.tmp
reading now WFULL0016.tmp
reading now WFULL0009.tmp
reading now WFULL0002.tmp
reading now WFULL0004.tmp
. reading now WFULL0010.tmp
........|......... BSE redistributing all elements

BSE diagonalizing matrix (scaLAPACKaware)
BSE calculating oscillator strength

After this calculation are not proceeding.DFT and RPA results are matching but BSE step is not proceeding. Please sugget.

Re: problem in the BSE step of optical calculation

Posted: Tue Apr 19, 2022 11:58 am
by merzuk.kaltak
Dear Alok Shukla,

I ran the tutorial with vasp-5.4.4 and vasp-6.3.0 without problems.
The last step is the most expensive one and takes about 2 minutes using 8 MPI ranks on a modern CPU.
In total the doall.sh script takes about 3 minutes to complete.

The last step uses a specific LAPACK routine (divide and conquer) which might cause problems on specific architectures.
Although I do not think this is the reson, however, just to be sure please let us know which vasp version you use and how it was installed (makefile.include and used toolchain).