Re: [AMBER] About Langevin dynamics algarithm

From: Mo Chen <mc842.cornell.edu>
Date: Thu, 20 Sep 2012 15:49:50 -0400

Hi Brian,
Thank you for the discussion! Yes, by "tau", I meant the lag time. I agree
with the ensemble part, though with respect to "MSD vs. lag time plot
should only be linear in the long time limit", I am not quit sure about
this. I though at least for water, we are always expecting the whole range
to be linear.

Thanks,
Mo

On Thu, Sep 20, 2012 at 2:06 PM, Brian Radak <radak004.umn.edu> wrote:

> Mo,
>
> I'm not quite sure what you mean by "tau," is this the lag time between
> frames? The MSD vs lag time plot should only be linear in the long time
> limit (although it may appear linear throughout). In retrospect I suppose
> that the stochastic nature of Langevin dynamics means that this type of
> relation may not hold, even in this limit. My initial thought was that
> there would be well defined diffusion (although not in terms of physical
> time) and that it would be impacted by the collision parameter (which I
> believe can actually be related to the diffusion coefficient of the solvent
> if desired). Anyway, that is all pure speculation on my part.
>
> To answer the probably more important question here, yes, if you want a
> diffusion coefficient that is physically meaningful in the sense that it
> corresponds to "real" time then you need to run NVE dynamics. However, the
> averages in this case are over the canonically distributed *initial
> conditions,* not the microcanonical ensemble of configurations generated by
> the dynamics (if the NVE dynamics are ergodic then all initial
> configurations give equivalent, time reversible trajectories in the long
> time limit).* *That is, NVE dynamics preserves the canonical weight (since
> energy is conserved) and thus any single NVE run is only one part of the
> average. Of course, if you take one NVE trajectory with the highest
> canonical weight it will contribute the biggest part of the average and you
> are done. That's just basic ensemble equivalence. I digress.
>
> In my opinion, what you need is a long NVT trajectory (such your Langevin
> dynamics trajectory) to be assured that you have a converged ensemble of
> initial conditions and then perform multiple NVE trajectories from multiple
> initial conditions. Taking the MSD vs time plot from that ensemble of
> trajectories should give you the information you want.
>
> Regards,
> Brian
>
>
> On Thu, Sep 20, 2012 at 1:17 PM, Mo Chen <mc842.cornell.edu> wrote:
>
> > Hi Brian,
> > I have run the NVT Langevin dynamics for hundreds of nanoseconds, but the
> > MSD vs. tau curve of the organic ions was not linear. I understand now
> that
> > the the self-diffusion coefficient calculated from a Lagenvin dynamics
> > simulation is not physically meaningful. As you said, the diffusion could
> > be well defined. So I wonder if I should definitely redo my simulation
> > using NVE simulation (ntt=0) to get meaningful self-diffusion
> coefficient?
> > Or calculation-wise, the self-diffusion coefficient should not be much
> > different between the two calculations. Thank you!
> >
> > Best,
> > Mo
> >
> > On Thu, Sep 20, 2012 at 1:03 PM, Brian Radak <radak004.umn.edu> wrote:
> >
> > > What kind of problem are you looking for? I believe Professor Case's
> > point
> > > was that diffusion calculated from a Langevin dynamics simulation is
> not
> > > physically meaningful, but it is probably well defined and calculable
> so
> > > long as you do not perform wrapping.
> > >
> > > Regards,
> > > Brian
> > >
> > > On Thu, Sep 20, 2012 at 12:40 PM, Mo Chen <mc842.cornell.edu> wrote:
> > >
> > > > Hi David,
> > > > Sorry, actually I was running NVT simulations actually.
> > > >
> > > > Best,
> > > > Mo
> > > >
> > > > On Thu, Sep 20, 2012 at 12:10 PM, Mo Chen <mc842.cornell.edu> wrote:
> > > >
> > > > > Hi David,
> > > > > I have run NVE simulations for the system. Below is my input file.
> > May
> > > I
> > > > > ask if you think there is still a problem? Thank you!
> > > > >
> > > > > Sample pulling input
> > > > > &cntrl
> > > > > nstlim=10000000, cut=12.0,
> > > > > ntpr=1000, ntwx=1000, ntt=3, gamma_ln=1.0,
> > > > > imin=0, ntx=7, irest=1, ig = -1,
> > > > > ntb=1, ntc=2, ntf=2, tol=0.00001,
> > > > > dt=0.002, tempi=363., temp0=363.
> > > > > /
> > > > > &ewald vdwmeth=0 /
> > > > > /
> > > > > END
> > > > > eof
> > > > >
> > > > >
> > > > >
> > > > > Best,
> > > > > Mo
> > > > >
> > > > > On Thu, Sep 20, 2012 at 8:56 AM, David A Case <
> > > case.biomaps.rutgers.edu
> > > > >wrote:
> > > > >
> > > > >> On Tue, Sep 18, 2012, Mo Chen wrote:
> > > > >>
> > > > >> > I am running Langevin dynamics simulations (ntt=3) for organic
> > > solvent
> > > > >> > using Amber 11, and would like to calculate the self-diffusion
> > > > >> coefficient
> > > > >>
> > > > >> Just a note: diffusion properties will be messed up by the use of
> > > > (almost
> > > > >> any) thermostat. So you should not expect the results to be
> > > physically
> > > > >> meaningful. If you want to study dynamics, you need to run NVE
> > > > >> simulations.
> > > > >>
> > > > >> ...dac
> > > > >>
> > > > >>
> > > > >> _______________________________________________
> > > > >> AMBER mailing list
> > > > >> AMBER.ambermd.org
> > > > >> http://lists.ambermd.org/mailman/listinfo/amber
> > > > >>
> > > > >
> > > > >
> > > > _______________________________________________
> > > > AMBER mailing list
> > > > AMBER.ambermd.org
> > > > http://lists.ambermd.org/mailman/listinfo/amber
> > > >
> > >
> > >
> > >
> > > --
> > > ================================ Current Address
> =======================
> > > Brian Radak : BioMaPS
> > > Institute for Quantitative Biology
> > > PhD candidate - York Research Group : Rutgers, The State
> > > University of New Jersey
> > > University of Minnesota - Twin Cities : Center for
> > Integrative
> > > Proteomics Room 308
> > > Graduate Program in Chemical Physics : 174 Frelinghuysen Road,
> > > Department of Chemistry : Piscataway, NJ
> > > 08854-8066
> > > radak004.umn.edu :
> > > radakb.biomaps.rutgers.edu
> > > ====================================================================
> > > Sorry for the multiple e-mail addresses, just use the institute
> > appropriate
> > > address.
> > > _______________________________________________
> > > AMBER mailing list
> > > AMBER.ambermd.org
> > > http://lists.ambermd.org/mailman/listinfo/amber
> > >
> > _______________________________________________
> > AMBER mailing list
> > AMBER.ambermd.org
> > http://lists.ambermd.org/mailman/listinfo/amber
> >
>
>
>
> --
> ================================ Current Address =======================
> Brian Radak : BioMaPS
> Institute for Quantitative Biology
> PhD candidate - York Research Group : Rutgers, The State
> University of New Jersey
> University of Minnesota - Twin Cities : Center for Integrative
> Proteomics Room 308
> Graduate Program in Chemical Physics : 174 Frelinghuysen Road,
> Department of Chemistry : Piscataway, NJ
> 08854-8066
> radak004.umn.edu :
> radakb.biomaps.rutgers.edu
> ====================================================================
> Sorry for the multiple e-mail addresses, just use the institute appropriate
> address.
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Thu Sep 20 2012 - 13:00:03 PDT
Custom Search