NVT MD

[acharyachethan]

Hi,

I'm running a canonical MD simulation. The temperature in the simulation is increasing. But, after every 40 steps the thermostat tries to control the temperature. Is there a way to decrease the time step period for the thermostsat to come into play, say 5 time step?

Thanks,

Chethan

[admin]

set SMASS=-1 and choose NBLOCK
please also have a look at
http://cms.mpi.univie.ac.at/vasp/vasp/node100.html (NBLOCK)
and
http://cms.mpi.univie.ac.at/vasp/vasp/node100.html (SMASS)

[yilmaz]

HI,
1) what is the difference between SMASS=-1 and SMASS=0? are both canonical MDs? in manual it says SMASS>=0 are for canonical MDs.i dont know if scaling velocities according to tempurature in a specific periods (with NBLOCK) mean the same.
SMASS=-1 and NBLOCK=5 works fine for me to be around a constant tempurature (300 K). but i am not sure if i am doing right. because in this case i am not considering the syncronization between tempurature frequencies and latice frequencies.

2) somebody can please also explain how to set SMASS>0 walues and what it means (say SMASS=2. )

thanks alot

[graeme]

The velocity scaling (Berendsen, smass=-1) thermostat is a poor choice, because it does not give a canonical distribution. It would be better if the Andersen (BKS) thermostat were implemented in vasp. Here, instead of scaling velocities (changing their magnitude), they are drawn from a Boltzmann distribution (changing both magnitude and direction). It is as simple to implement, and gives the correct statistical distribution.

The Nosé thermostat (smass>=0) does give the canonical distribution if the coupling strength (smass) can be chosen appropriately for the system. I have noticed that the Nosé thermostat has a hard time getting energy into a system started from a minimum geometry. What seems to work is heating the system up to temperature with velocity scaling, and then collecting proper statistics with a subsequent run using the Nosé thermostat.