[gmx-users] Re: Diatomic in MeCN NPT (NH and PR) simulation segfaults after 1 us

[Justin Lemkul]

On 11/5/12 11:25 PM, benjfitz wrote:
> Justin, thank you for reminding me that I need to adhere to sane parameters.
> I do have a few clarifying points, as it were. I tried these simulations
> again, but with a larger box and 0.9 nm cutoffs, and received a segfault at
> the same simulation time.
>
> The model I am using, from the user contributed downloads section, does make
> use of virtual sites (virtual2). The choice for the masses and diatomic bond
> length always seemed strange to me (9 and 33 amu) b/c one could use almost
> identical masses. This would make setting up the virtual sites far more
> intuitive. I did this, after going through your tutorial for CO2, and this
> adaptation of the MeCN model suffers the same instability.
>
> I then took your CO2 model and modified it to CS2 using the proper LJ
> parameters from Madden's 1981 model that doesn't have charges. I ran this
> using coulombtype=cutoff and the 2 us simulation completed without any
> problems. I gave small (-0.01 for S and 0.02 for C) charges to the atoms and
> ran using PME, which crashed at 1.07 us.
>
> Do these results shed any new light on what is happening with my NPT
> simulations?
>

The fact that any system crashes at 1.07 us (either CS2 or MeCN) suggests that 
the problem is external to your molecule's topology.  It seems too coincidental 
that the times are the same, almost as if your system is running out of memory 
or there is some nefarious error accumulating a problem.  Unfortunately, I 
suspect this will be really hard to debug, especially given the fact that it 
takes a very long time for the error to occur.  If you're interested in trying 
to debug it yourself, change algorithms one by one, i.e. tcoupl and pcoupl 
methods, and see if the situation improves.  This may not be how you want to do 
your data collection, but it would either rule out or identify which algorithm 
might be failing.

-Justin

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

Justin A. Lemkul, Ph.D.
Research Scientist
Department of Biochemistry
Virginia Tech
Blacksburg, VA
jalemkul[at]vt.edu | (540) 231-9080
http://www.bevanlab.biochem.vt.edu/Pages/Personal/justin

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