If you want to calculate a diffusion coefficient, don't thermostat your
system. The diffusion coefficient is related to the area under the
velocity-velocity auto-correlation function via a Green-Kubo relationship
and thermostats artificially affect the velocities of particles and hence
give unrealistic results for the velocity auto-correlation funciton.
To get a good value of a diffusion coefficient, run your simulations at
constant energy. If you want the diffusion coefficient at a given
temperature then you should run an ensemble of NVE systems. You can
sample this ensemble by running a long NVT simulation, then selecting
configurations from this simulation that are separated in time by at least
the autocorrelation time of the kinetic energy of the system, and then run
each of those configurations at NVE and average your results.
Please note that the first sentence of the abstract of the above paper (
http://pubs.acs.org/doi/abs/10.1021/ct400109a) is incorrect. Systems at
a given temperature can be studied by sampling an ensemble of NVE systems.
TK
On Sat, Jan 23, 2016 at 8:33 AM, Christina Bergonzo <cbergonzo.gmail.com>
wrote:
> Hi Jinfeng,
>
> The difference you see (which I don't think is huge) is most likely due to
> an imperfect translation of the MD input from DL POLy to Amber. Diffusion
> properties of water are very sensitive to many input parameters (as Hannes
> indicated). The best discussion of this was published here:
>
> http://pubs.acs.org/doi/abs/10.1021/ct400109a
>
> -Christina
>
> On Saturday, January 23, 2016, <hannes.loeffler.stfc.ac.uk> wrote:
>
> > I'm afraid I can't answer this either. I was just wondering how
> sensitive
> > the diffusion constant would be to statistics.
> >
> > Other points to consider: is the constant sensitive to the pressure
> > algorithm that is if you use, say, the MC barostat would you have to
> expect
> > (very) different results? The MSD computation should be pretty
> > straightforward but any chance of a wrong conversion factor? You could
> > counter-check to compute the MSD via the velocity autocorrelation
> function.
> >
> > ________________________________________
> > From: jinfeng liu [jinfengliu1119.gmail.com <javascript:;>]
> > Sent: 23 January 2016 11:49
> > To: AMBER Mailing List
> > Subject: Re: [AMBER] Water diffusion constant obtained in Nose-Hoover
> > thermostat in Amber14+AmberTools15
> >
> > Yes, I tried as much as possible to replicate his simulation in that
> paper.
> > But he ran the simulation using DL_POLY package, not Amber. So I am not
> > sure where the problem is. According to my knowledge, the Nose-Hoover
> > thermostat is used for the first time in AmberTools15, I don't know
> anyone
> > tested it before.
> >
> > 2016-01-23 19:35 GMT+08:00 <hannes.loeffler.stfc.ac.uk <javascript:;>>:
> >
> > > So I assume it is a simulated result then? If so what was his
> simulation
> > > setup/code that is have you tried to replicate his as much as possible?
> > > Anything special about the analysis of the results e.g. use of multiple
> > > time origins, etc.?
> > >
> > > ________________________________________
> > > From: jinfeng liu [jinfengliu1119.gmail.com <javascript:;>]
> > > Sent: 23 January 2016 11:27
> > > To: AMBER Mailing List
> > > Subject: Re: [AMBER] Water diffusion constant obtained in Nose-Hoover
> > > thermostat in Amber14+AmberTools15
> > >
> > > The expected diffusion constant is 0.232, according to Gregory A.
> Voth's
> > > SPC/FW paper.
> > >
> > > 2016-01-23 19:23 GMT+08:00 <hannes.loeffler.stfc.ac.uk
> <javascript:;>>:
> > >
> > > > Have you looked into the literature to see what the expected
> diffusion
> > > > behaviour of the water _model_ you use is?
> > > >
> > > > ________________________________________
> > > > From: jinfeng liu [jinfengliu1119.gmail.com <javascript:;>]
> > > > Sent: 23 January 2016 11:10
> > > > To: AMBER Mailing List
> > > > Subject: [AMBER] Water diffusion constant obtained in Nose-Hoover
> > > > thermostat in Amber14+AmberTools15
> > > >
> > > > Dear Amber users,
> > > >
> > > > I performed liquid water dynamics using the latest
> > > > Amber14+AmberTools15. I used the SPC/fw water model in Nose-Hoover
> > > > thermostat (I found this thermostat in the latest Amber manual). But,
> > the
> > > > water diffusion constant calculated in 1 ns MD simulation was 0.15,
> > > > comparing to the experimental value of 0.23 A^2/ps. My MD input is
> like
> > > the
> > > > following:
> > > >
> > > > &cntrl
> > > > imin=0,irest=1,ntx=5,
> > > > dt=0.001,nstlim=1000000,
> > > > ntb=2,ntp=1,taup=2.0,
> > > > igb=0, cut=8.0, jfastw=4,
> > > > ntt=9,gamma_ln=10.0,nkija=1,temp0=300.0,
> > > > ntpr=200,ntwr=200,ntwx=200,
> > > > &end
> > > >
> > > > The trajin file calculating the diffusion constant is like:
> > > >
> > > > trajin md.crd
> > > > diffusion 0.2 :WAT.O average
> > > >
> > > > According to the MSD calculated above, I obtained the diffusion
> value.
> > > >
> > > > So, Can anyone help to explain why the calculated diffusion constant
> is
> > > > much smaller than the experimental value ?
> > > > _______________________________________________
> > > > AMBER mailing list
> > > > AMBER.ambermd.org <javascript:;>
> > > > http://lists.ambermd.org/mailman/listinfo/amber
> > > >
> > > > _______________________________________________
> > > > AMBER mailing list
> > > > AMBER.ambermd.org <javascript:;>
> > > > http://lists.ambermd.org/mailman/listinfo/amber
> > > >
> > > _______________________________________________
> > > AMBER mailing list
> > > AMBER.ambermd.org <javascript:;>
> > > http://lists.ambermd.org/mailman/listinfo/amber
> > >
> > > _______________________________________________
> > > AMBER mailing list
> > > AMBER.ambermd.org <javascript:;>
> > > http://lists.ambermd.org/mailman/listinfo/amber
> > >
> > _______________________________________________
> > AMBER mailing list
> > AMBER.ambermd.org <javascript:;>
> > http://lists.ambermd.org/mailman/listinfo/amber
> >
> > _______________________________________________
> > AMBER mailing list
> > AMBER.ambermd.org <javascript:;>
> > http://lists.ambermd.org/mailman/listinfo/amber
> >
>
>
> --
>
> ---------------------------------------------------------------------------------------
> Christina Bergonzo, PhD
> Postdoctoral Researcher
> Department of Medicinal Chemistry, University of Utah
> http://home.chpc.utah.edu/~cheatham/
>
> ---------------------------------------------------------------------------------------
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
--
************************************************
Tom Kurtzman, Ph.D.
Assistant Professor
Department of Chemistry
Lehman College, CUNY
250 Bedford Park Blvd. West
Bronx, New York 10468
718-960-8832
http://www.lehman.edu/faculty/tkurtzman/
<http://www.lehman.edu/faculty/tkurtzman/index.html>
************************************************
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Sat Jan 23 2016 - 06:00:06 PST
