[gmx-users] rlist- group scheme

Mark Abraham mark.j.abraham at gmail.com
Tue Jan 23 18:05:13 CET 2018


Hi,

On Tue, Jan 23, 2018 at 1:31 PM Faezeh Pousaneh <fpoosaneh at gmail.com> wrote:

> >
> >
> >
> > > For any range of cut off it gives wrong results for rlist>rcut-off.
> > >
> > > Now I understand where the problem comes.  I tried to see the
> collisions
> > of
> > > two molecules with a table (user potential).
> >
> >
> > Over what range is your interaction defined? If the potential isn't zero
> by
> > the time you get to the relevant cutoff, then you will produce artefacts.
> >
> >
> > I use same 6,12 table  in Gromacs top folder, which goes to 1+rc.
>

Great.

> > From the resulted LJ potential
> > > between two particles I see that it takes the potential to zero at
> rlist,
> > > not at rcut-off.
> > > how to deal with it now?
> > >
> >
> > If the interaction isn't in the list, it doesn't get computed.
> >
>
> >
> I hope you got my point. I assume when I define rcut-off in .mdp file, the
> resulted potential should be zero after this value, so the larger rlist >
> rcut-off wont make difference.


Indeed. Unfortunately the group scheme wasn't designed for this use case
(which is now common) so support for it is a) slow, and b) not present in
many method combinations.


> But what I see from resulted potential vs
> displacement of two molecules, is that Gromacs has taken the potential to
> zero at rlist not *''rcut-off''*. That is why for rcut-off > rlist the
> result is not correct.
>

You're seeing the consequence of the particles at displacement > rlist not
being put into the list, so not making a contribution.


> So the solution for me now is to put the potential in the table file to
> zero manually for x>cut-off.


Glancing at the kernels that implement user tables, that's what you should
do. That means the forces are strange and you would wish to consider a form
where the potential smoothly reaches zero at the cutoff.

But it is better to solve the problem in
> Gromacs.
>

It looks like the group scheme with user tables doesn't have support for
buffered lists with a correct cutoff. That correct cutoff is only available
in the group scheme for potentials that do go to zero. Likely we concluded
that because you have the ability to express your wishes entirely in your
table, you should do so. Or use rlist == rvdw == rcoulomb.

Mark

> Mark
> >
> >
> > >
> > >
> > > Best regards
> > >
> > >
> > > On Sun, Jan 21, 2018 at 12:57 AM, Mark Abraham <
> mark.j.abraham at gmail.com
> > >
> > > wrote:
> > >
> > > > Hi,
> > > >
> > > > Yes, you said that before. But I asked what range it worked over, ie
> > what
> > > > range of cutoff.
> > > >
> > > > Mark
> > > >
> > > > On Sat, Jan 20, 2018, 17:15 Faezeh Pousaneh <fpoosaneh at gmail.com>
> > wrote:
> > > >
> > > > > Hi Mark,
> > > > >
> > > > > For rlist=rcut-off (max one) it works correctly, but as soon as
> > rlist>
> > > > > rcut-off it fails and gives different energies. Which I do not
> > > understand
> > > > > why. In principle, there is nothing going on after cut-off in
> > > potentials.
> > > > >
> > > > >
> > > > >
> > > > >
> > > > > Best regards
> > > > >
> > > > >
> > > > > On Sat, Jan 20, 2018 at 1:13 AM, Mark Abraham <
> > > mark.j.abraham at gmail.com>
> > > > > wrote:
> > > > >
> > > > > > Hi,
> > > > > >
> > > > > > There could be. But over what range does your table work
> correctly?
> > > > What
> > > > > do
> > > > > > you get when you have a two particle system at various
> > displacements?
> > > > > >
> > > > > > Mark
> > > > > >
> > > > > > On Fri, Jan 19, 2018, 23:59 Faezeh Pousaneh <fpoosaneh at gmail.com
> >
> > > > wrote:
> > > > > >
> > > > > > > Dear Mark,
> > > > > > >
> > > > > > > Thank you, however, the problem occurs when 'table' is used.
> For
> > > > > > > vdWtype=not-user the energies are OK and the same for both
> cases,
> > > as
> > > > I
> > > > > > > mentioned. So I guess there should be something wrong in the
> way
> > > > > Gromacs
> > > > > > > interpolates the table with cutoffs and rlist.
> > > > > > >
> > > > > > > Best regards
> > > > > > >
> > > > > > >
> > > > > > >
> > > > > > > Best regards
> > > > > > >
> > > > > > >
> > > > > > > On Fri, Jan 19, 2018 at 8:09 PM, Mark Abraham <
> > > > > mark.j.abraham at gmail.com>
> > > > > > > wrote:
> > > > > > >
> > > > > > > > Hi,
> > > > > > > >
> > > > > > > > With the group scheme, when you choose nstlist > 1 then you
> > get a
> > > > > list
> > > > > > of
> > > > > > > > all the groups that satisfy rlist, whatever it is, and
> continue
> > > to
> > > > > use
> > > > > > > that
> > > > > > > > list for the whole lifetime. If you chose rlist==rcoulomb,
> then
> > > > > > rcoulomb
> > > > > > > is
> > > > > > > > never considered again, because you chose to use an
> unbuffered
> > > > list.
> > > > > > But
> > > > > > > > with rlist > rcoulomb, you chose the list to have
> interactions
> > > that
> > > > > are
> > > > > > > not
> > > > > > > > within rcoulomb, so every interaction is checked at every
> step
> > > for
> > > > > > > whether
> > > > > > > > it is within rcoulomb, before including it. That's slower to
> > run,
> > > > > but a
> > > > > > > > better approach. And of course you can't expect them to
> compute
> > > the
> > > > > > same
> > > > > > > > numbers, particularly if things move fast relative to the
> > buffer
> > > > size
> > > > > > and
> > > > > > > > list lifetime.
> > > > > > > >
> > > > > > > > The Verlet scheme defaults do a good job of automating all
> > this,
> > > > but
> > > > > as
> > > > > > > you
> > > > > > > > know, there's not yet support for tabulated interactions.
> > > > > > > >
> > > > > > > > Mark
> > > > > > > >
> > > > > > > > On Fri, Jan 19, 2018, 14:25 Faezeh Pousaneh <
> > fpoosaneh at gmail.com
> > > >
> > > > > > wrote:
> > > > > > > >
> > > > > > > > > Hi,
> > > > > > > > >
> > > > > > > > > I run a simple example using Table,6,12 (for LJ) with vdW
> > type=
> > > > > user.
> > > > > > > > Then
> > > > > > > > > group scheme should be chosen.
> > > > > > > > >
> > > > > > > > > In results, I got different LJ energies when I set
> > > rlist=rcoulomb
> > > > > and
> > > > > > > > > rlist>rcoulomb (which is allowed in this scheme). I see in
> > the
> > > > > manual
> > > > > > > > that
> > > > > > > > > with group scheme that can be happen. But I do not find how
> > to
> > > > get
> > > > > > rid
> > > > > > > of
> > > > > > > > > the differences in energies? or which rlist is the case.
> > > > > > > > >
> > > > > > > > > However, when I run with vdW type= cutoff for both rlist
> > > > > > > rlist=rcoulomb
> > > > > > > > > and rlist>rcoulomb with group scheme I get similar results.
> > > > > > > > >
> > > > > > > > > Best regards
> > > > > > > > > --
> > > > > > > > > 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.
> > > > > > >
> > > > > > --
> > > > > > 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.
> > > >
> > > --
> > > 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.
>


More information about the gromacs.org_gmx-users mailing list