[gmx-users] Simulation crashes: issue with [pair] potential

Fri Mar 3 08:33:40 CET 2017

Hi,

When you're using a version of the software that's more than three years
old, and a bug was fixed less than two years ago, you need to get an
updated version :-)

Once you do, your problem reduces to whether you can identify few enough
groups of atoms whose pairwise group-group interatomic interactions have
the same shape. They can use the same tables. Trying to use a different
table for eg each atom pair would run out of energy groups, and be an utter
disaster for performance.

Mark

On Wed, 1 Mar 2017 10:18 <poma at ifpan.edu.pl> wrote:

> Dear Mark,
>
> thanks for your comment regarding the pairs interaction. I did know about
> the tabulated choice for nb interaction, but the number of interaction
> will reach more than >200 pair. Even though, I go for the energy group
> description it will need a larger number of groups, as each entry will be
> assigned to one pair (i,j) with i and j index of particles. I have a
> Martini ff underneath, and sure I have several Martini types which reflect
> the amino acid character: (P)Polar, (A)Apolar, (I)Intermediate Polar and
> (C)Charge and in each different types(Qda,Nda,P1,P5,etc). If I use groups
> with P,A,I and C, then the issue will be, that I need to define
> interactions between four groups to describe specific interaction between
> CA atoms in each group.
>
> For instance,
> pair 1 and 34: involves groups (C)-(I) with types Qda and Nda
> pair 1 and 36: involves groups (C)-(I) with types Qda and Nda.
>
> the issue comes that one may need to use different sigmas which are
> defined by the distance between the pairs.
>
> The other pairs are also non-redundant due to atom indexes and different
> sigma parameters which are defined by the distance as well.
>
> Recenlty, I found another choice the [pairs_nb]. Guess you had a comment
> in the past regarding a deprecate directive see at
>
> https://mailman-1.sys.kth.se/pipermail/gromacs.org_gmx-users/2015-May/097965.html
>
> Currently, I have problem to set it up. Look at a piece of my .itp file
>
> [pairs_nb ]
>     1    34      1  0.0 0.0  0.34962  0.00974
>     1    36      1  0.0 0.0  0.29890  0.00712
>
> with error:
> ********************************************************
> Generated 0 of the 780 non-bonded parameter combinations
> case: 20
> -------------------------------------------------------
> Program grompp_mpi, VERSION 4.6.5
> Source code file:
> /home/HDD3000/HOME_poma/poma/Software/gromacs-4.6.5/src/gmxlib/gmx_fatal.c,
> line: 597
>
> Fatal error:
> Invalid case in switch statement, file
> /home/HDD3000/HOME_poma/poma/Software/gromacs-4.6.5/src/kernel/topio.c,
> line 1007
> ********************************************************
>
> initially I used [pairs] without the zeros entries for  qi, qj and the
> last two terms are combination rules for V** and W** parameters. It worked
> as the system was in the minimum energy state (folded state). Now I
> started the simulation in the unfolded state and the distance between
> pairs may reach lengths longer than molecular cutoff. This is the reason
> why I do want to shift the pairs in context of non-bonded picture. Any
> insights will be highly appreciated, maybe I really miss the concept of
> the tabulate scheme & groups definition.
>
> best,
>
> Adolfo
> > Hi,
> >
> > Tabulated non-bonded potentials are possible (see manual), but currently
> > only implemented between pairs of energy groups, and only in the group
> > cut-off scheme. So by e.g. putting each type of particle in its own
> energy
> > group you can arrange for whatever inter-group interactions suit you. But
> > there's only a small number of groups possible, maybe 64 or 128? I
> forget.
> >
> > Mark
> >
> > On Tue, Feb 28, 2017 at 10:22 AM <poma at ifpan.edu.pl> wrote:
> >
> >> Hi,
> >> We had employed the pair potentials in unconventional way.  They were
> >> used
> >> to describe nonbonded specific Lennard Jones (LJ) Potential between a
> >> pair
> >> of particles. Each pair has a different set of sigma's parameters, but
> >> just one epsilon. Although they are LJ potentials, they are considered
> >> as
> >> bonded interactions by definition and a table is made internally. When
> >> bonded interactions exceed the minimum convention image distance (MCID),
> >> then the simulation generally crashes due to a distance which is above
> >> the
> >> MCID.
> >> Obviously, Gromacs starts complaining about this issue with a warning
> >> regarding a distance between two atoms which is reasonably large and it
> >> assumes that we are performing a free energy calculation. Has anybody
> >> encounter a similar issue before, honestly I would like to find another
> >> way to define the interaction between pair of particles as purely
> >> nonbonded potential (with different sigma's parameters) and not as
> >> pairs,
> >> any help will be appreciate!
> >>
> >> kind regards,
> >>
> >>
> >>
> >> --
> >> Dr. Adolfo Poma
> >> Institute of Physics
> >> Polish Academy of Science
> >> Al. Lotnikow 32/48,
> >> 02-668 Warsaw, Poland
> >>
> >> --
> >> Gromacs Users mailing list
> >>
> >> * Please search the archive at
> >> http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List before
> >> posting!
> >>
> >> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
> >>
> >> * For (un)subscribe requests visit
> >> https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or
> >> send a mail to gmx-users-request at gromacs.org.
> >>
> > --
> > Gromacs Users mailing list
> >
> > * Please search the archive at
> > http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List before
> > posting!
> >
> > * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
> >
> > * For (un)subscribe requests visit
> > https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or
> send
> > a mail to gmx-users-request at gromacs.org.
> >
>
>
> --
> Gromacs Users mailing list
>
> * Please search the archive at
> http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List before
> posting!
>
> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>
> * For (un)subscribe requests visit
> https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or
> send a mail to gmx-users-request at gromacs.org.
>