VASP Forum

Hi,

I am running MD simulations with ML-FF, but no ionic positions are written to the files. The XDATCAR file remains empty, 'grep POSITIONS OUTCAR' also yields nothing, and the hdf5 file also contains no trajectory. It happens consistently every time I start a simulation.

This is the INCAR file:

SYSTEM = H2O
ENCUT = 700
IBRION = 0
ISIF = 2
ISYM = 0
NSW = 20000
EDIFF = 1E-6
ISMEAR = 0
SIGMA = 0.2
PREC = Accurate
ALGO = Normal
LREAL = Auto
LWAVE = .FALSE. !write WAVECAR (def T)
LCHARG = .FALSE. !write CHGCAR (def T)
NCORE = 4
NELM = 300

POTIM = 0.5
TEBEG = 300
TEEND = 300
MDALGO = 3
LANGEVIN_GAMMA = 10 10
LANGEVIN_GAMMA_L = 3
PSTRESS = 0.001
PMASS = 100

ML_LMLFF = .TRUE.
ML_MODE = run
#ML_MB = 6000
ML_LBASIS_DISCARD = .TRUE.
ML_DESC_TYPE = 1
ML_ISCALE_TOTEN = 1
#ML_EATOM_REF = -.11293374E+00 -.32094164E+00
ML_IERR = 100
ML_OUTBLOCK = 10

I wonder what I am doing wrong, as this is new behavior.

I forgot to add: the OSZICAR file and the slurm output file both show that the simulation progresses to the end. It is just the positions that are not written anywhere.

Hi,

The input looks reasonable so I don't know of any particular reason why you would not get the ionic positions during the run.
Could you share all the input files so we can try to reproduce the issue on our side?

I have uploaded the input files here:

https://tucloud.tuwien.ac.at/index.php/ ... pHiCWQ4biW

Thank you for sharing the files, I quickly ran your calculation locally and I got a XDATCAR file as expected.
Could you perhaps try to run this same calculation with a different version of vasp that was built with a different compiler?
Or even possibly just a different machine.

It is a bit hard for me to make a concrete suggestion when I cannot reproduce this issue.