[gmx-developers] Group cutoff-scheme simulation temperature affected by nh-chain-length

[Michael R Shirts]

If you can file a bug report in redmine (http://redmine.gromacs.org/) with example files, that would be useful. Before you file it can you check if problem happens if you use standard cutoffs, instead of tabulated cutoffs? I wonder if there is some sort of code path that it’s going down.  I will bet that combination of inputs has only very rarely been used.  Also, you said the temperature jumps. Does it come back down to 310, or does it stay there?

Generally, I would recommend using v-rescale instead of Nose-Hoover.  Nose-Hoover has some known flaws; increasing the chain length is supposed to help things, but doesn’t really help that much for realistic systems, just with very simple ones.

However, any Nose-Hoover change length should in principle give the same average, and did when the code was written.  If it doesn’t, the code was broken somewhere, possibly because of lack of coverage in the regression tests and something getting out of sync.

Best,
~~~~~~~~~~~~~~~~
Michael Shirts
Associate Professor
michael.shirts at colorado.edu<mailto:michael.shirts at colorado.edu>
http://www.colorado.edu/lab/shirtsgroup/
Phone: (303) 735-7860
Office: JSCBB C123
Department of Chemical and Biological Engineering
University of Colorado Boulder

From: <gromacs.org_gmx-developers-bounces at maillist.sys.kth.se> on behalf of "Du, Yu" <duyu at sioc.ac.cn>
Reply-To: "gmx-developers at gromacs.org" <gmx-developers at gromacs.org>
Date: Monday, October 23, 2017 at 12:48 AM
To: GMX-developer <gromacs.org_gmx-developers at maillist.sys.kth.se>
Subject: [gmx-developers] Group cutoff-scheme simulation temperature affected by nh-chain-length

Dear GMX Developer and Prof. Hess,

I think this question is not common to GMX users, so I post it here.

Prof. Hess is very familiar with the tabulized potential and group cutoff scheme, so I also CC this email to you. Any advice will be appreciated.

The Backgroud:
I simulated protein-ligand system (ref_t=300) with amber99sb-ildn ff. I used the normal table6-12.xvg tabulized potential to reproduce this problem.

The Problem:
In NVT simulation, integrator=md-vv, Tcoupl=nose-hoover, coulombtype=pme-user, vdwtype=user. I found that at the start of the simulation, there is a temperature jump to around 310K (average 310K in 100ps) higher than the ref_t (300K) under nh-chain-length=10.

1)I tuned and spotted the nh-chain-length parameter. If I used nh-chain-length=1 there is no temperature jump and temperature can converge very well to 300K.

2)I replicated the same experiment in verlet scheme, the temperature converged to ref_t 300K in 100ps and no jump in nh-chain-length=2, which can 50K temperature jump (to 350K).

The Question:
1)How will the nh-chain-length parameter affect the simulation temperature under the group cut-off scheme?
2)Is it valid to use NH (nh-chain-length=1) not NH chain (nh-chain-length>1) to simulate the protein-ligand system?

Thanks for any advice.

--
Du, Yu
PhD Student,
Shanghai Institute of Organic Chemistry
345 Ling Ling Rd., Shanghai, China.
Zip: 200032, Tel: (86) 021 5492 5275
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-developers/attachments/20171023/8797be99/attachment-0003.html>