[gmx-users] freezing/position restraints and minimization

[David Mobley]

> OK, further info on this. Looks like it's the minimizer, at least in
> the case of freeze groups. I'm using l-bfgs; it seems the minimizer
> doesn't respect the constraints imposed by freeze groups.
>
> Now that I think of it, I seem to remember that most of the minimizers
> in GROMACS don't respect constraints. I think I'll try again with only
> a steepest descents minimization (I think steep works OK with
> constraints, if I remember correctly) and see if this helps.

Sorry to e-mail so much. But it seems this is, indeed, the problem.
For freeze groups to work in minimization you need to use a minimizer
which is can handle constraints properly. In GROMACS the only
minimizer that is set up to handle constraints properly is steepest
descents.

Note that this also has implications for anyone using rigid water
models (most of us): If you're "minimizing" with L-BFGS or other
minimizers aside from steepest descents, you're probably deforming
your water molecules, as well.

David


> Thanks,
> David
>
> On 3/16/07, David Mobley <dmobley at gmail.com> wrote:
> > Bob,
> >
> > > Maybe this is a stupid suggestions but....did you check if your two
> > > atoms are centered in the box and that the box dimensions are big
> > > enough to accomodate your system? Maybe the atoms appear to be "flying
> > > apart" because periodic boundary conditions are being applied.
> >
> > They're not "flying apart" in a ridiculous sense -- i.e.  box
> > dimensions are a couple nm, and the distance between them goes from
> > 0.14 nm up to 0.25-0.28 nm. Also, I used g_dist to measure the
> > distance between them; it takes care of all the PBC wrapping. So no,
> > that's not what's going on.
> >
> > One possible complicating issue is that the atoms in question right
> > now are two carbon atoms and I'm trying to hold them at the typical
> > carbon-carbon bond distance, so they are sort of up against each
> > other. So I can buy it might take a big force constant to keep them
> > from moving apart -- I just can't figure out why freeze groups isn't
> > working. (Incidentally, if I look at a movie of the trajectories run
> > with freeze groups, it does look like they stay frozen -- just not at
> > the positions I've specified).
> >
> > Thanks,
> > David
> >
> > >
> > >
> > > On 3/16/07, David Mobley <dmobley at gmail.com> wrote:
> > > > All,
> > > >
> > > > I'm trying to be ultimately able to fix the position of several
> > > > arbitrary atoms (think of them just as spheres) in a simulation. One
> > > > would think this should be straightforward, but I'm having big
> > > > problems.
> > > >
> > > > In particular, right now I'm just using two atoms, which I initially
> > > > have at a separation of 0.14 nm (i.e. one is at 0,0,0; the other at
> > > > 0.14, 0, 0). I have tried freezing the positions in two ways. First, I
> > > > tried using freeze groups, with the two atoms defined as a single
> > > > group, specified in my mdp file, using freezedim= y y y.
> > > > Unfortunately, as soon as I even minimize the system (with freeze
> > > > groups on!) they fly apart to about twice the desired separation and
> > > > stay further apart throughout dynamics (at least as measured by
> > > > g_dist).
> > > >
> > > > I've also tried using a position restraint file with the default
> > > > spring constant to hold them at the desired positions, but that
> > > > doesn't seem to work either (again, they end up at about twice the
> > > > desired separation).
> > > >
> > > > Any ideas? Especially on why freeze groups might not be working --
> > > > ideally I would just fix these in place using freeze groups, but for
> > > > some reason that doesn't seem to be working properly.
> > > >
> > > > This is in 3.3.
> > > >
> > > > Thanks,
> > > > David
> > > > _______________________________________________
> > > > gmx-users mailing list    gmx-users at gromacs.org
> > > > http://www.gromacs.org/mailman/listinfo/gmx-users
> > > > Please don't post (un)subscribe requests to the list. Use the
> > > > www interface or send it to gmx-users-request at gromacs.org.
> > > > Can't post? Read http://www.gromacs.org/mailing_lists/users.php
> > > >
> > > _______________________________________________
> > > gmx-users mailing list    gmx-users at gromacs.org
> > > http://www.gromacs.org/mailman/listinfo/gmx-users
> > > Please don't post (un)subscribe requests to the list. Use the
> > > www interface or send it to gmx-users-request at gromacs.org.
> > > Can't post? Read http://www.gromacs.org/mailing_lists/users.php
> > >
> >
>