VASP Forum

Hi everyone,

I am using the VASP linear response DFPT implementation to get the piezoelectric coefficient in R3m BaTiO3 (using the LEPSILON tag and IBRION = 8).

This is the INCAR I am using:

SYSTEM=relax

ISTART=0

ENCUT=500 eV
ALGO=Normal
NELM=100
EDIFF=1e-8
ISMEAR= 0
SIGMA = 0.02
PREC    = Accurate
EDIFFG=-0.0005
IBRION=8
NSW=50
ISIF=3
LMAXMIX=6
LEPSILON=.TRUE.
LREAL=.FALSE.
LCHARG=.FALSE.
LWAVE=.FALSE.
GGA=PS

I keep getting OOM error even with 256GB -
LRF_COMMUTATOR internal error: the vector H(1)-e(1) S(1) |phi(0)> is not orthog
onal to |phi(0)> (-1.093619832611588E-004,-5.396724733763847E-014)
[2026-02-23T22:53:58.659] error: Detected 1 oom_kill event in StepId=3427325.0. Some of the step tasks have been OOM Killed.

Is increasing the memory the right way to solve this or is there something wrong in the INCAR?

Thanks
Dominic

Hi Dominic,

BaTiO₃ in R3m is a very small system, so a genuine memory issue is unlikely — especially with 256 GB available. The key message in your output is actually

LRF_COMMUTATOR internal error: the vector H(1)-e(1) S(1) |phi(0)> is not orthogonal to |phi(0)>

This usually indicates a ground-state consistency / convergence problem in the DFPT step, rather than a true OOM situation. The Slurm OOM message is likely a consequence of the crash, not the root cause.

Can you please check:

1. for IBRION = 8 the structure must already be fully relaxed. NSW is ignored in DFPT as far as I know, and the calculation assumes a well-converged ground state structure. Make sure the structure is very well relaxed beforehand, e.g.

   Code: Select all

      IBRION = 2
      ISIF   = 3
      EDIFF  = 1E-8
      EDIFFG = -1E-3

   until forces are clearly below 1E-3 eV/Å.

2. Then perform a separate static SCF calculation (NSW=0) to ensure the wavefunctions and density are fully converged.

3. Try again the DFPT step:

   Code: Select all

      IBRION   = 8
      LEPSILON = .TRUE.

Since ferroelectric R3m BaTiO₃ is known to have soft modes, DFPT can become unstable if the reference state is not perfectly minimized.

If the problem persists, it would be helpful to see the POSCAR, KPOINTS, POTCAR details to try to reproduce the issue.

Best regards,
Alex

Hi Alex,

It worked. I used the larger tolerance values and now I have piezoelectric tensor for R3m BTO.

Thanks for the suggestions.