the amplitude in temperature fluctuation of MD

[luke419]

I'd like to run MD of graphene using VASP.

1. I used SMASS 0 with the relevant other parameters determined appropriately as follows.

PPREC = Normal
ENCUT = 400
LREAL = A
IBRION = 0
SMASS = 0
NSW = 200
ALGO = Fast
MDALGO = 2
TEBEG = 300
TEEND = 300
ISYM = 0
POTIM = 0.5
LWAVE = .FALSE.
LCHARG = .FALSE.
ISMEAR = 0
SIGMA = 0.05

It works well, however, it showed too large amplitude in the temperature fluctuation of +-60K or ~ +-100K.
I tested it by changing SMASS value but it showed the same problem.
I also tested the atom number of 60, 120, and 200 but it does not solve this issue.
Is it common result of quantum mechanical MD since the atom number is so small?
If not, would you let me know what I should do for it?

2. I've runned VASP MD for the system blended with small polymers previously (around 200 atoms) and it showed a good result about it.
Is there a big difference between the systems of non-bonded interaction such as liquid or the blending of small polymers and the covalent bonded system such as graphene for this problem?

3. Incidentally, it is written that default of MDALGO of VASP is 0 (standard MD).
What does it mean by standard MD? Is it Nose-Hoover thermostat?
Is it right that in standard MD, the change of SMASS is not meaningful?

Thanks a lot in advance for your help.

Regards,

Luke

[admin]

Points 1 and 2: The atomic velocities in NVT MD are given by
Maxwell-Boltzmann distribution
(http://en.wikipedia.org/wiki/Maxwell%E2 ... stribution).
Relatively large fluctuations of the instant temperature for systems
containing small number of atoms are therefore natural. In fact,
correctly set thermostat only guarantees that the average temperature
(over long period of time) is correct (+ atomic velocities should be
Maxwell-Boltzmann distributed - this condition can be, however, partly
violated by using Nose-Hoover thermostat, which is known to fail to
generate proper NVT ensembles in some cases). It is rather difficult to
judge about the reason of your problem without knowing further details -
if you think the above description does not answer your questions,
please contact Dr. Bucko (tomas.bucko@univie.ac.at) who is involved in
developing MD in VASP.

As to the point 3 - there are currently two independent implementations
of MD in VASP - the "standard" one with basic MD routines (basically
only NVE and NVT with Nose-Hoover thermostat, see
http://cms.mpi.univie.ac.at/vasp/vasp/IBRION_0.html) and the alternative
implementation in which also some advanced MD techniques such as
metadynamics and constrained dynamics are available. As described here:

http://cms.mpi.univie.ac.at/vasp/vasp/A ... iques.html

in order to activate the advanced MD, vasp has to be compiled with cpp
flag -Dtbdyn . The "standard" MD routines can be still invoked in that
case - this can be done be setting MDALGO=0. For more details about our
advanced MD, please check the user guide:

http://cms.mpi.univie.ac.at/vasp/vasp/A ... iques.html