Re: Regarding thermostats

From: Marcos Sotomayor (
Date: Thu Feb 23 2012 - 16:37:03 CST

Hi Felipe,

Different ensembles and thermostat have different
advantages/disadvantages, but I think the main point to take into account
is that if your results depend too much on the ensemble you are using,
then they are unlikely to be physiologically relevant or you may have
to make a serious effort to explain why that happened (the results might
be interesting for other reasons though).

I believe Alek Aksimentiev tested the effect of the Langevin thermostat on
induced ionic currents with external fields. It seems that a Langevin
constant gamma of 0.1 has minimum effects (I repeated the same
simulations with minor modifications and got the same results). SMD
simulations also seem OK with a gamma of 0.1. at reasonable pulling

I personally prefere to have the correct ensemble and
usually run simulations in both the NpT and NVE ensembles to check that
what I see is ensemble-independent.

Hope this helps,

On Mon, 13 Feb 2012, wrote:

> Dear NAMD users,
> Lately in the lab we have been discussing about which thermostat use in
  our simulations. I know this can be just a detail, but it is quite
  tricky. The velocity rescaling method will not give he correct ensemble,
  but the langevin dynamics will affect the dynamics of the system and in
  principle quantities such as diffusion constants will be modified. For
  NVT simulations the Lowe-Andersen thermostat should be the option to
  avoid that, but we are doing NPT and the Langevin barostat already
  modifies the dynamics (it contains a "thermostat on its own right?).
> Besides, as far as i remember the TIP3 water model (the one we are
  using) already has a higher diffusion constant and the langevin damping
  coefficient is used to somehow match it to the experimental value, so i
  do not know if it is possible to obtain the "correct" dynamics with TIP3
  water and any “Galilean invariant” thermostat.
> Does anyone have any insights on this topic?
> regards
> Felipe

This archive was generated by hypermail 2.1.6 : Tue Dec 31 2013 - 23:21:41 CST