Thank you Axel for your constant help! I assume in the case if I need to
assess the system at different temperatures (say 300K, 310K, 350K etc... )
what I will need to do is ramp up the temperature to the needed one and then
use the final state of that temperature equilibration as the first step of
my production run at that temperature.
While I am at it, my other worry is the damping factor in the langevin fix;
having read up on it from the mailing list I realise that it cannot be too
high or too low. In the past I did not worry about it much but now that I am
simulating peptides/proteins, I think I need to be more careful with the
choice of the damping factor, which I have set to 1 timestep (0.01 fs)... I
have noticed that Steve suggested 100 timesteps to someone who works in
polymers. I would like to know what the rationale is behind setting the damp
parameter? Again, Steve said that as long as one is looking at times longer
that the damp factor, the T should remain around target T, So I guess I
should not worry too much about it?
i think you need to spend some time reading up on thermostat
algorithms and how they impact properties computed from MD
simulations.
there is no single "do this, not that" rule in these cases. most of
the "rule of thumb" values work best for dense monoatomic systems with
periodic boundaries.
unless you want to model a specific property, i.e. use fix langevin to
represent interaction with an implied solvent, you usually want the
thermostat to have as little impact on the system as possible without
losing the desired impact, e.g. maintaining a constant temperature.
for most well equilibrated systems, there should be very little need
for this kind of correction. the larger the amount of energy
transferred, the more likely you are hiding a bad choice of simulation
parameters (e.g. too large a time step, bad potential parameters,
interactions that remove energy in an unphysical way)
also, you have to be careful to not mix up different thermostats. with
nose-hoover thermostats, the time constant essentially determines a
characteristic frequency. this typically is chosen to be near, but not
exactly on top of a characteristic mode of your system. that choice
determines how well energy is exchanged and how much the system is
impacted. whereas in fix langevin, the impact is much more drastic and
immediate (you apply friction and then add noise, how much is
determined by the time constant). based on this description, it should
be evident, that nose-hoover chain thermostat can have much less
impact with a shorter time constant, than fix langevin, but also that
nose-hoover may need much longer until a system reaches equilibrium
(if the coupling is very weak) than with langevin, where the exchange
is forced.
LAMMPS supports a whole zoo of thermostats that all have benefits and
drawbacks and are more suitable or less for specific cases.
axel.