vasp MD temperature fluctuation too big

[Felix]

Hi , guys

I am using VASP to do Molecular dynamics. However, the amplitude of temperature fluctuation is very large at +-200 K. I've checked the value of SMASS and it seems not effective. The incar is as follows.

SYSTEM = GST
NCORE=4;NSIM=4
ENCUT = 250
ISTART = 0
ICHARG = 2
ISIF = 2
NBLOCK = 1
KBLOCK = 100
NELMIN = 6
EDIFF = 1E-4
LREAL = Auto
APACO = 10
NPACO = 200
NSW = 2000
IBRION = 0
ISMEAR = 0
SIGMA = 0.2
POTIM = 3
IALGO = 48
LCHARG = .FALSE.
LWAVE = .TRUE.
ISYM = 0
SMASS = 0
TEBEG = 1075
TEEND = 1075
NWRITE = 1

If anybody is familiar with the MD, please help me find out what is the problem with my calculation. I'm wondering whether I miss some parameters to set. Thanks.

[nsolomat]

Hi, I have the same issue and a very similar INCAR. Were you able to resolve the T fluctuations?

[valln59]

Hi,

i am also facing the same issue. if anyone replies this question , it would help many users
I am trying with eh inputs file from vasp example site( for Si standard MD simulation). played with SMASS but no use at all

thank you

[ferenc_karsai]

First one needs to check if the structure is undergoing significant structure changes or not (via the pair correlation function PCDAT - wiki/index.php/PCDAT, or via the CONTCAR,XDATCAR files). In the "Liquid Si - Standard MD" example one starts from pristine Si and melts the structure drastically by setting a target temperature at 2000K.
So the structure will change quite for some time and one should wait to look at the mean temperature and temperature fluctuations only after proper equilibration of the structure (no more significant structure changes, only collective "vibrations" of the structure).

If the mean temperature after equilibration is in the desired region and if after that changing SMASS doesn't seem to have an effect, one can try to reduce the time step (POTIM). This way the thermostat has to deal with smaller changes of the structure at each MD step and has more "time" to equilibrate to the desired temperature.

Remark: Temperature fluctuations are not necessary something bad and depending on the structure (and temperature) one cannot go below some fluctuations without destroying the desired ensemble.