[gmx-users] Simulations in NVE ensemble with CHARMM36 - update

Justin Lemkul jalemkul at vt.edu
Tue May 13 16:57:04 CEST 2014



On 5/13/14, 10:44 AM, ABEL Stephane 175950 wrote:
> Hello Justin
>
> Thanks for your reply. I have finally found why I obtained a large drift and decrease of the temp of my system during the NVE : I stupidly mixed different cpt and tpr files. If I use the right files and the following parameters below suggested by M. Shirts in  [1] for my tests:
>
> rlist           = 1.3
> ;vdW
> vdwtype         = Switch
> rvdw            = 1.1
> rvdw_switch     = 1.0
>
> ; Electrostatics
> coulombtype     = PME
> rcoulomb        = 1.3
>

CHARMM parameters can be very sensitive to the value of rvdw.  It's something we 
consider as close to an absolute rule as you get, so the settings above should 
be modified to read:

rlist = 1.4
rvdw = 1.2
rvdw-switch = 1.0
rcoulomb = 1.4

Or use the settings in my previous message in version 5.0-rc1.

> The energy and the temp of my system remain stable during the 100 ps of my simulation after  initial small decrease of the energy and the temp to 295.5 K (initial temp was 297.904 K)
>
> [1] http://comments.gmane.org/gmane.science.biology.gromacs.user/42686
>
> Please also note that with the parameters for Verlet options you gave I obtain the following error with 4.6.5 for simulations with NVE,
>
>   Temperature coupling is required for calculating rlist using the energy
>    drift with verlet-buffer-drift > 0. Either use temperature coupling or
>    set rlist yourself together with verlet-buffer-drift = -1
>

Good to know; I've never had a need to run NVE.

> If I add the verlet-buffer-drift = -1 option in my mdp (not sure it is correct), grompp gives only two notes
>
> NOTE 1 [file run_CH_1_NVE_pot_mod.mdp]:
>    With Verlet lists the optimal nstlist is >= 10, with GPUs >= 20. Note
>    that with the Verlet scheme, nstlist has no effect on the accuracy of
>    your simulation.
>
>
> NOTE 2 [file run_CH_1_NVE_pot_mod.mdp]:
>    rlist is equal to rvdw and/or rcoulomb: there is no explicit Verlet
>    buffer. The cluster pair list does have a buffering effect, but choosing
>    a larger rlist might be necessary for good energy conservation.
>

I don't believe the options I provided will work in 4.6.5; they definitely 
should in 5.0-rc1, so a test with verlet-buffer-drift = -1 would be potentially 
useful.

-Justin

>
> Stéphane
>
>
> On 5/7/14, 9:05 AM, ABEL Stephane 175950 wrote:
>> Dear GMX-users
>>
>> I have done a 90 ns long simulation of a heterogeneous  system (SPCE water, phospholipids and a hydrophobic solvent) in NPT ensemble (298 K at 1 bar) with GROMACS4.6.5. to compute related dynamical properties, I would like to continue this simulation in NVE. But I am puzzled to choose the mdp parameters to have a good energy conservation. Indeed, I have checked the gromacs 2004 JCTC or the following and follow the advices given in the following threads
>>
>> http://comments.gmane.org/gmane.science.biology.gromacs.user/42686
>> https://www.mail-archive.com/gmx-users@gromacs.org/msg60393.html
>>
>> But I always obtain for my system a large drift and the temp decreases significantly (20 K) after few dozens of ps (the initial temp was around 297.3 K). For this simulation I used the double precision of gromacs with a timestep of 1 fs.
>>
>
> Can you provide a full .mdp file so we can look at all your settings?  Are these
> drifts during NPT or NVE?
>
>> More, since I have never used the potential-modifier features, so I not sure how to setup  the correct parameters in the context of CHARMM simulations.
>>
>
> With the caveat that I have NOT thoroughly tested the following, my
> understanding of the new Verlet options indicates that these settings should
> perform pretty well using CHARMM36:
>
> cutoff-scheme = Verlet
> rlist = 1.2
> vdwtype = cutoff
> vdw-modifier = Potential-switch
> rvdw = 1.2
> rvdw-switch = 1.0
> coulombtype = PME
> rcoulomb = 1.2
>
> Again, barely tested on my part, and never tested for NVE.
>
> -Justin
>
> --
> ==================================================
>
> Justin A. Lemkul, Ph.D.
> Ruth L. Kirschstein NRSA Postdoctoral Fellow
>
> Department of Pharmaceutical Sciences
> School of Pharmacy
> Health Sciences Facility II, Room 601
> University of Maryland, Baltimore
> 20 Penn St.
> Baltimore, MD 21201
>
> jalemkul at outerbanks.umaryland.edu | (410) 706-7441
> http://mackerell.umaryland.edu/~jalemkul
>
> ==================================================
>

-- 
==================================================

Justin A. Lemkul, Ph.D.
Ruth L. Kirschstein NRSA Postdoctoral Fellow

Department of Pharmaceutical Sciences
School of Pharmacy
Health Sciences Facility II, Room 601
University of Maryland, Baltimore
20 Penn St.
Baltimore, MD 21201

jalemkul at outerbanks.umaryland.edu | (410) 706-7441
http://mackerell.umaryland.edu/~jalemkul

==================================================


More information about the gromacs.org_gmx-users mailing list