From: Jérôme Hénin (jerome.henin_at_uhp-nancy.fr)
Date: Fri Apr 23 2004 - 14:19:48 CDT
Hi Brian,
I am not sure what you mean by "stochastic", between quotes. Most usual
thermostats, spanning from velocity rescaling to a Gaussian isokinetic
constraint, are certainly not stochastic (without quotes). They are indeed
completely deterministic. On the other hand, I agree that they do not
accurately model actual heat transfers (for good reasons), so you may say
they are incorrect in some way. But classical MD simulations are incorrect in
so many ways that a well-used thermostat is often not the most problematic
issue.
If you're interested in statistics, any thermostat that samples from the
correct ensemble should do the job. And if you want very accurate dynamics...
why not just run NVE?
Jerome
Le jeudi 22 Avril 2004 19:02, Brian Bennion a écrit :
> Hi Alfredo
>
> How is any thermostat not "stochastic" ? At some point you have to decide
> to fiddle with an atoms velocity/momentum. Anytime energy/temperature
> needs to be adjusted something is wrong or at least modeled incorectly,
> no mater how many fancy names you put into a algorithm.
>
> Brian
>
> On Thu, 22 Apr 2004, Juan Alfredo Freites wrote:
> > Hi Brian,
> >
> > I think that we should clarify that STOCHASTIC temperature coupling is
> > "bad", in the sense you've indicated, because it precludes energy
> > conservation. Let's not forget that one could run a simulation in the
> > canonical ensemble by using a Nose-Hoover-chain thermostat, or in the
> > isokinetic ensemble using a Gaussian isokinetic thermostat. These I
> > would call "good" temperature controls. Of course, these options are not
> > available (yet) in NAMD.
> >
> > Alfredo.
> >
> > On Thu, 22 Apr 2004, Brian Bennion wrote:
> > > My training taught me that coupling, especially temperature
> > > coupling was "bad", in that you technically start a new simulation
> > > everytime you bleed energy off the system.
> > >
> > > So thats my $0.02
> > >
> > > Brian
> >
> > --
> > J. Alfredo Freites
> > Department of Physics
> > University of California, Irvine
> > Irvine, CA 92697-4575
> > (949) 824-9921
>
> *****************************************************************
> **Brian Bennion, Ph.D. **
> **Computational and Systems Biology Division **
> **Biology and Biotechnology Research Program **
> **Lawrence Livermore National Laboratory **
> **P.O. Box 808, L-448 bennion1_at_llnl.gov **
> **7000 East Avenue phone: (925) 422-5722 **
> **Livermore, CA 94550 fax: (925) 424-6605 **
> *****************************************************************
This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:37:33 CST