BSE in supercells with VCA

[kaynat_kalvi]

Hi,

I am trying doing a mBSE calculation for 128 atom system, using VCA approximation. i did the first dft run and exported the WAVEDER and WAVECAR files to mBSE, but my calculations stops without giving errors/reasons.
--------------------------------------------
found WAVECAR, reading the header
POSCAR, INCAR and KPOINTS ok, starting setup
FFT: planning ...
reading WAVECAR
the WAVECAR file was read successfully
initial charge from wavefunction

The Fermi energy was updated, please check that it is located mid-gap
values below the HOMO (VB) or above the LUMO (CB) will cause erroneous energies
E-fermi : 4.7456

the WAVEDER file was read successfully
energies w=

responsefunction array rank= 43264
LDA part: xc-table for Pade appr. of Perdew
allocating 0 responsefunctions rank= 43264
Doing 0 frequencies on each core in blocks of 0
allocating two-electron 4 orbital integral table 64 256 64 2
BSE (scaLAPACK) attempting allocation of 34.360 Gbyte rank= 524288
BSE setting up matrix

===================================================================================
= BAD TERMINATION OF ONE OF YOUR APPLICATION PROCESSES
= RANK 0 PID 26890 RUNNING AT mel0150
= KILLED BY SIGNAL: 9 (Killed)
===================================================================================

===================================================================================
= BAD TERMINATION OF ONE OF YOUR APPLICATION PROCESSES
= RANK 1 PID 26891 RUNNING AT mel0150
= KILLED BY SIGNAL: 9 (Killed)
===================================================================================

===================================================================================
= BAD TERMINATION OF ONE OF YOUR APPLICATION PROCESSES
= RANK 2 PID 26892 RUNNING AT mel0150
= KILLED BY SIGNAL: 9 (Killed)
===================================================================================

===================================================================================
= BAD TERMINATION OF ONE OF YOUR APPLICATION PROCESSES
= RANK 3 PID 26893 RUNNING AT mel0150
= KILLED BY SIGNAL: 9 (Killed)
===================================================================================
-------------------------------------------------------------------------------------------------------------------------------------------------------------------
The job description is as follows,

seff 3308357
Job ID: 3308357
Cluster: meluxina
User/Group: u102344/hpcusers
State: FAILED (exit code 255)
Nodes: 2
Cores per node: 256
CPU Utilized: 04:07:50
CPU Efficiency: 22.51% of 18:20:48 core-walltime
Job Wall-clock time: 00:02:09
Memory Utilized: 488.73 GB
Memory Efficiency: 50.91% of 960.00 GB

-------------------------------------------------------------------------------------------------------

please helpe me out here.

[alexey.tal]

Dear kaynat_kalvi,

I moved your post here as a separate topic as I think it is a different issue.

The error indicates that your calculation ran out of memory. How much memory do you have available per node? Storing the matrix of rank 524288 would required 4.4 Tb of memory with vasp_std and double precision. You need to make sure that such calculation can fit on your nodes.

Also, from the output it looks to me that you are not using the most recent version of VASP, in the version 6.5.0 the parallelization over bands has been largely improved so if you have access to a newer version, you should use it instead.

Regarding the VCA approach, I am very skeptical that this method works in the context of BSE. The first problem that comes to mind is that using VCA leads to fractional occupation of states, i.e., metallic bands, and at the moment BSE in VASP does not properly treat the intraband transitions.

Best wishes,
Alexey Tal