[gmx-users] sd integrator and P-R barostat/ACETONITRILE with CHARMM ff

[Justin Lemkul]

On 7/17/17 12:17 PM, Sonia Milena Aguilera Segura wrote:
> 
>>>>
>>>> I increased the size of the simulation box to 4 nm. Indeed, the values of
>>>> pressure improved (averages around -1 bar or 1.5, 2 bar, or so). However,
>>>> the T keeps being overestimated (302 K). During NVT I got the right value,
>>>> so I decided to run the 200 ns NPT equilibration with Berendsen barostat
>>>> instead of P-R. I got the right T value, around 298.3 and a pressure of
>>>> 1.5 more or less. Then I launched a continuation test with 200 ps MD with
>>>> P-R (I couldn't use the -e option because it gives me the error Could not
>>>> find energy term named 'Box-Vel-XX', moreover, I didn't use P-R before so
>>>> I guess I shouldn't expect stored values for it?. But I am using -t ).
>>>> Despite I got low pressure averages around 0.5 bar, the temperature raised
>>>> again to 301-302. This happens very early in the simulation, which seems
>>>> to indicate that for sd integrator/thermostat and P-R barostat there is
>>>> something going on. I am getting practically the same density in between
>>>> simulation, so I guess I can say t
>>>    ha
>>>>     t in term of P, the system has been equilibrated. However, what can I do
>>>>     to get the right T for this system? If I was able to get the right T
>>>>     with Berendsen barostat, I don't understand what's wrong when I change
>>>>     to P-R.
>>>>
>>>
>>> Sounds buggy.  What version of GROMACS are you using?  There were temperature
>>> issues with the Langevin integrator in previous versions, but those should
>>> have
>>> been long since solved.
>>
>>> I am using version 5.1.2. I checked the release notes for version 5.1.3 and 5.1.4 and the only thing related with sd integrator (none) or P-R barostat is this one 'Fixed Parrinello-Rahman with nstpcouple > 1'. Can this be the cause?
>>
> 
>> I don't know.  Upgrade to 2016.3 and try again.
> 
>> -Justin
> 
> Dear Justin,
> 
> I upgraded to 2016.3 and did several tests. In all cases I ran first a NVT with sd and 200 ps NPT with Berendsen. I always got a T of 297.6-298.6 for equilibrations. Then, for instance,  I ran a 4 nm box with isopropanol, which is bigger than acetonitrile, and got a T of 300 K instead of 302 K, for the same system with the 5.1.2 version. For a 4 nm box of acetonitrile I got also a small improvement: 299.5 vs 300-301 for the previous version. So, I guess there was something wrong with the version. Still, I am wondering if the level of precision is enough (an error of +-1.5 K), considering that I am getting very precise values for Berendsen barostat, even for the 3 nm boxes for both solvents. Also, my average P is -1.5 and 1.5.. which is giving me very similar results for density in all cases. Can I assume that those values are ok (even the negative) considering the expected fluctuations during the MD simulation?
> 

Yes.  Rarely do you get exactly the specified pressure.  The fluctuations should 
be quite large.  Is -1.5 +/- several hundred statistically indistinguishable from 1?

-Justin

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

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

Department of Pharmaceutical Sciences
School of Pharmacy
Health Sciences Facility II, Room 629
University of Maryland, Baltimore
20 Penn St.
Baltimore, MD 21201

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

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