From: Bjoern Olausson (namdlist_at_googlemail.com)
Date: Fri Jul 08 2011 - 09:07:43 CDT
On Friday 08 July 2011 15:09:57 you wrote:
> On Fri, Jul 8, 2011 at 7:30 AM, Bjoern Olausson <namdlist_at_googlemail.com>
> > On Friday 08 July 2011 12:44:50 Molecular Dynamics wrote:
> >> Dear Namd users,
> >> I have run a 30 ns md simulation at 300 K in NVT using langevin method.
> >> After the simulation I made a plot for temperature-TS with NAMDplot in
> >> VMD. But I saw that temperature was fixed above 300 K . What is the
> >> reason for this ?
> >> Please examine the graph of temperature-TS and the main parts of conf
> >> file from the link below
> >> http://uploading.com/files/38719fc4/namd_temp-ts.rar/
> > Looks like the mean is of by ~10 K. Looks pretty much to me.
> > For my simulations the mean is pretty close to the set temperature.
> > For example:
> > Target: 303K
> > MIN: 199.2046 @1100
> > Max: 304.8586 @332100
> > MEAN: 302.27269461
> > STD: 2.4002367513
> > To counteract your Problem I would suggest to increase the friction
> > (Langevin damping coefficient) to something around 5/ps or at least 2/ps
> > instead of 1/ps.
> this doesn't solve the problem of bad energy conservation,
> it merely hides it. a well set up MD simulation should conserve
> energy for an extended period _without_ a thermostat. the use
> of thermostat algorithms is (was) primarily meant to make a
> small system behave like it was embedded in (and in equilibrium
> with) a much larger environment for being able to correctly derive
> observables from the simulation data. now, a _little_ drift is
> unavoidable due to the truncation errors resulting from using
> floating point numbers and a numerical integrator, and this
> can be handled by the thermostat as well. but anything beyond
> that turns your MD simulation from MD to something that is
> more like a "guided MC" simulation.
> there are many possible sources that will lead bad energy
> conservation: too long a time step, very steep potentials,
> high temperature, improper use of SHAKE (= rigidbonds)
> or lack of it, bad multi-timestepping settings, bad choice
> of PME grid and so on. each of these need to be investigated
> for what causes the bad energy conservation and that needs
> to be corrected. as noted before, this is easiest done by
> running without a thermostat, after the system has been
> properly equilibrated with it.
thanks for your answer, it was, as allays, very informative!
I must admit, that I never thought of running a MD simulation without a
thermostat but in fact I never had to hunt down energy problems which were not
related to a simple failure in the system setup (PSF/PDB) or parametrization.
The drift, acceptable for the system you are simulating, clearly depends on
what you want and what your system can tolerate, but what is acceptable as
"_little_ drift" when we are talking about current MD-Methods?
I am just curious.
-- Bjoern Olausson Martin-Luther-Universitšt Halle-Wittenberg Fachbereich Biochemie/Biotechnologie Kurt-Mothes-Str. 3 06120 Halle/Saale Phone: +49-345-55-24942
This archive was generated by hypermail 2.1.6 : Mon Dec 31 2012 - 23:20:33 CST