Molecular Dynamics - Bug Report

[marekmihalkovic]

REPORT Bug in VASP: vasp.5.3.3

Error description:

We would like to report an error discovered while attempting to run constant volume MD simulation on a supercell of face centered diamond-cubic Silicon lattice (Pearson symbol cF8). We enclose three setup folders packed into archive file data.tgz. Each setup folder contains files INCAR, POSCAR, POTCAR, KPOINTS, OUTCAR, job.out. The last file is the standard output.

1. The initial problem had occured before the first ionic step finished:

file setup1/job.out:

RSPHER: internal ERROR: 624 -43 576 -43

This behaviour was repreduced for several setups:

* Si, Ge: cF8, the same structure
* 2x2x2 (64 atoms) as well as 3x3x3 (216 atoms) supercell
* PBE and GGA PAW pseudopotential

2. Afterwards we set ISYM = 0 and after several ionic steps we obtained different error:

file setup2/job.out:

ERROR FEXCP: supplied Exchange-correletion table
is too small, maximal index : 11915

3. folder setup3: Finally, we have set ISYM to its default value and randomly displaced several atoms out of their symmetric positions and subsequently simulation ended without error.

We suspect that this erroneous behaviour is caused by implementation of symmetry constraints.

[ferenc_karsai]

This is a very old version of VASP, the problems are most likely resolved in the newest versions of VASP.
Also setup 1 is a bad setup, since running an MD with ISYM>0 is not good.