[lammps-users] Difference between Berendsen and temp/rescale

LAMMPS LAMMPS Mailing List Mirror
1

Hey,

I have a crystal system of 1680 atoms (70 molecules) which I want to heat above the melting point (> 350K).
After the first step which was minimization of potential energy of the crystal at 0K, I have to heat it up.
Now, I have couple of choices:

  1. use a fix temp/berendsen rescaling scheme
  2. use a temp/rescale scheme
  3. fix nvt
    (I know that only the last one (nvt) does time integration to positions, but that’s a rather technical issue)

I’d like to ask what is the conceptual difference between the methods ?

and what is the relation of Tstart, Tstop to the desired temperature ?

In the above example:
My initial temp is 0K (pot energy minimzation) - so does that imply my Tstart should be 0K when using one of the above methods ?

My desired temperature is, say, 400K (above melting point), does that mean Tstop should be 400K ?

what if make Tstart equal Tstop ? while the “true” temp of the system is other than that ? what does that mean ?

Thanks in advance,

2

My initial temp is 0K (pot energy minimzation) - so does that imply my Tstart should be 0K when using one of the above methods ?
My desired temperature is, say, 400K (above melting point), does that mean Tstop should be 400K ?

yes to both, and you should change the Temp gradually (over a long run)

what if make Tstart equal Tstop ? while the "true" temp of the system is other than that ? what does that mean ?

if you do this, you are asking the system to change immediately from
0K to 350K, which will induce
a rapid change (in a time of order the damping constant) - which may
cause oscillations, etc

Re: which thermostat to use - that's up to you - they all have different
theoretical motivations, which the papers behind them explain.

OTOH, if your simulation results depend sensitively on what
kind of thermostatting you do, that's not generally a good thing.

Steve