Re: [AMBER-Developers] thoughts on gamma_ln?

From: Alexey Onufriev <alexey.vt.edu>
Date: Wed, 9 Mar 2016 14:11:44 -0500

Hi, All.

I think the default should depend on whether implicit (GB) or explicit
model is used.
For the GB, we want a pretty low gamma, e.g gamma_ln=0.01 for optimal
sampling (we tested that extensively, e.g. our paper with Ramu and Ross).

But otherwise, 1.0 should be just fine as many have already pointed out.

regards, Alexey


On Wed, Mar 9, 2016 at 12:32 PM, Jason Swails <jason.swails.gmail.com>
wrote:

> A few comments here. First, gamma_ln!=0 is an error when ntt!=3 unless I'm
> mistaken, so the default gamma_ln needs to either be 0 or dependent on
> ntt=3. However, there *is* a use-case for ntt = 3 and gamma_ln = 0 -- it's
> a good test for testing and validation. In this case, the Langevin
> integrator should be equivalent to the standard integrator.
>
> +1 for the change in principle, though.
>
> All the best,
> Jason
>
> On Wed, Mar 9, 2016 at 11:13 AM, Carlos Simmerling <
> carlos.simmerling.gmail.com> wrote:
>
> > I agree that an error might be better than changing it away from what the
> > input says. if it does that, we should make it the default so it's
> > documented that way.
> >
> > On Wed, Mar 9, 2016 at 10:56 AM, Ross Walker <ross.rosswalker.co.uk>
> > wrote:
> >
> > > > Didn't the code use to quit with an error telling you gamma_ln needed
> > to
> > > be set if you set ntt=3 and didn't change it. Did this change at some
> > point?
> > >
> > > >From mdin_ctrl_dat.F90 in pmemd
> > >
> > > gamma_ln = 0.d0 !Default is zero but if ntt=3 we set it to 1.0d0
> > >
> > > if (gamma_ln .gt. 0.d0 .and. ntt .ne. 3) then
> > > write(mdout, '(a,a)') error_hdr, 'ntt must be 3 if gamma_ln > 0!'
> > > inerr = 1
> > > end if
> > >
> > > if ( ntt == 3 .and. ntb == 2) then
> > > !Require gamma_ln > 0.0d0 for ntt=3 and ntb=2 - strange things
> happen
> > > !if you run NPT with NTT=3 and gamma_ln = 0.
> > > if ( gamma_ln <= 0.0d0 ) then
> > > write(mdout,'(a,a)') error_hdr, 'gamma_ln must be > 0 for ntt=3
> > with
> > > ntb=2.'
> > > inerr = 1
> > > end if
> > > end if
> > >
> > > So it requires it for NTP runs and the comment implies it is set to 1.0
> > > for ntt=3 but never actually is. So close to what you want...
> > >
> > >
> > > > On Mar 9, 2016, at 07:55, Carlos Simmerling <
> > carlos.simmerling.gmail.com>
> > > wrote:
> > > >
> > > > Sorry that wasn't clear - I support changing it.
> > > > On Mar 9, 2016 9:34 AM, "Carlos Simmerling" <
> > carlos.simmerling.gmail.com
> > > >
> > > > wrote:
> > > >
> > > >> I would agree, users could I tend to turn it on (ntt) but have it
> not
> > > >> really turn on....
> > > >> On Mar 9, 2016 9:33 AM, "Adrian Roitberg" <roitberg.ufl.edu> wrote:
> > > >>
> > > >>> Hi
> > > >>>
> > > >>> The default for gamma_ln for langevin thermostat (ntt=3) is =0,
> which
> > > >>> basically means constant energy. Obviously, when you use shake and
> a
> > > >>> dt=0.002 energy is not really conserved, so there is trouble.
> > > >>>
> > > >>> Should we keep it as 0 or change it to something small but useful,
> > > like 1?
> > > >>>
> > > >>> adrian
> > > >>>
> > > >>> --
> > > >>> Dr. Adrian E. Roitberg
> > > >>> Professor.
> > > >>> Department of Chemistry
> > > >>> University of Florida
> > > >>> roitberg.ufl.edu
> > > >>> 352-392-6972
> > > >>>
> > > >>>
> > > >>> _______________________________________________
> > > >>> AMBER-Developers mailing list
> > > >>> AMBER-Developers.ambermd.org
> > > >>> http://lists.ambermd.org/mailman/listinfo/amber-developers
> > > >>>
> > > >>
> > > > _______________________________________________
> > > > AMBER-Developers mailing list
> > > > AMBER-Developers.ambermd.org
> > > > http://lists.ambermd.org/mailman/listinfo/amber-developers
> > >
> > >
> > > _______________________________________________
> > > AMBER-Developers mailing list
> > > AMBER-Developers.ambermd.org
> > > http://lists.ambermd.org/mailman/listinfo/amber-developers
> > >
> > _______________________________________________
> > AMBER-Developers mailing list
> > AMBER-Developers.ambermd.org
> > http://lists.ambermd.org/mailman/listinfo/amber-developers
> >
> _______________________________________________
> AMBER-Developers mailing list
> AMBER-Developers.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber-developers
>
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Wed Mar 09 2016 - 11:30:03 PST
Custom Search