[gmx-users] desired usage of verlet-buffer-tolerance with charmm ff?

[Christopher Neale]

Dear Justin, Mark, and Szilárd:

Thank you for all the help. I understand this a lot better now.

My intended usage is REMD with PLUMED in gromacs 5.1.2 to do some replica exchange of the Hamiltonian. In this usage, the exchange attempts have to occur at steps when the neigborlist is updated. Therefore I would like to avoid having mdrun changing nstlist. What I have done is to run a short simulation at the highest temperature and see what mdrun does to nstlist and rlist. I then set these explicitly in the .mdp file for REMD and set verlet-buffer-tolerance=-1. 

I am happy to see that mdrun tells me what kind of kernels it uses ( "Using AVX_256 4x4 non-bonded kernels") so that I can make sure that I've got the same type of lists with a given rlist and nstlist.

Thank you,
Chris.
________________________________________
From: gromacs.org_gmx-users-bounces at maillist.sys.kth.se <gromacs.org_gmx-users-bounces at maillist.sys.kth.se> on behalf of Szilárd Páll <pall.szilard at gmail.com>
Sent: 08 April 2016 06:42
To: Discussion list for GROMACS users
Subject: Re: [gmx-users] desired usage of verlet-buffer-tolerance with  charmm ff?

On Thu, Apr 7, 2016 at 9:55 PM, Christopher Neale
<chris.neale at alum.utoronto.ca> wrote:
> Dear users:
>
> Charmm mdp options are listed here: http://www.gromacs.org/Documentation/Terminology/Force_Fields/CHARMM with the following values:
>
> constraints = h-bonds
> cutoff-scheme = Verlet
> vdwtype = cutoff
> vdw-modifier = force-switch
> rlist = 1.2
> rvdw = 1.2
> rvdw-switch = 1.0
> coulombtype = PME
> rcoulomb = 1.2
> DispCorr = no
>
> ;; verlet-buffer-tolerance is not included in that list , so I presume the suggestion is to accept the default value of 0.005 kJ/mol/ps
>
> However, I would have expected:
> rlist = 1.4
> verlet-buffer-tolerance = -1
>
> I realize that issue #1413 ( http://redmine.gromacs.org/issues/1413 ) is still open, but when I use the approach suggested at http://www.gromacs.org/Documentation/Terminology/Force_Fields/CHARMM I get a surprisingly small increase in rlist as nstlist is increased:
>
> Reading file MD.tpr, VERSION 5.1.2 (single precision)
> Changing nstlist from 10 to 25, rlist from 1.2 to 1.207
>
> I read somewhere that there is buffering in the verlet scheme, but 0.007 nm sure doesn't seem like a lot... I must be missing something?

What you are missing is likely that the Verlet scheme uses a
particle-cluster based approach where the pair ("neighbor") list is
built of such clusters. This provides an implicit buffer which is
always present even when rcut=rlist and therefore the explicit buffer
can be kept shorter.
See e.g. Fig 3 of [1], the blue non filled atoms represent the implicit buffer.

1. http://www.sciencedirect.com/science/article/pii/S2352711015000059


> Thank you,
> Chris.
>
>
>
> --
> 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.