[gmx-users] Re: Diagnosing an explosion

[Steven Kirk]

Apologies for replying to myself - maybe I can sharpen up some of my 
original questions below:

> Reply below
> 
>> > 
>>> >> I am having a problem with my BD simulation either crashing with a Range 
>>> >> error or locking up the mdrun process completely.
>>> >> 
>>> >> I am running Gromacs 3.3.1 (must use this version ATM because HPC 
>>> >> provider has not upgraded yet) on a quad core Intel machine, w/ Fedora 
>>> >> Core 7.
>>> >> 
>>> >> I have 1000 'particles' in a cubic box (configuration here:
>>> >> 
>>> >> http://datavet.hv.se/personal/SK/publicfiles/colloid.pdb
>>> >> 
>>> >> where the particles are placed randomly with no two particle centres 
>>> >> closer than 4 nm.
>>> >> 
>>> >> The particles are defined in the topology here:
>>> >> 
>>> >> ; Complete topology file
>>> >> ;
>>> >> ; particles.itp
>>> >> 
>>> >> [ defaults ]
>>> >> 1  1  no  1.0  1.0
>>> >> 
>>> >> [ atomtypes ]
>>> >> AR   28699.81134   0.00   A   1.0   1.0
>> > 
>> > Looks like you are using dimensionless eps and sigma while using 
>> > otherwise MD units (e.g. T). Check chapter 2 of the manual. You want to 
>> > multiply your T by boltz (0.008314 or something like that) probably.
> 
> Aha! Thank you David.
> 
> My goal was to use MD units in everything, since my tabulated data file 
> was extracted from another GROMACS simulation and is hence already in MD 
> units (kJ/mol vs. nm), and I would prefer to specify temperatures in K.
> 
> My reason for specifying 1.0 for the last two parameters was that I 
> thought this would allow me to use, unscaled, the tabulated potential 
> (which is already in MD units).
>
> If I want to do this, what alternative values should I use for the last 
> two parameters on the [ atomtypes ] line above (given that I am really 
> only using the g(r) and g''(r) columns to hold my potential data: the 
> h(r) and h''(r), which I don't care about, are all zeroes) ?

If I change the combination type to 2 or 3 in the  [ defaults ] line, 
will this remove the unit scaling problem ?
> 
> I have checked Chapter 2 in the manual and I am still unsure how to 
> handle this.
> 
> I took your advice (which matches with manual Section 2.4 on *reduced* 
> units) and scaled my temperature to 2.494353 =  300K * 0.008314. This 
> did not give a crash, but increasing the timestep and decreasing bd_fric 
> made the problem return.
> 
> Scaling the temperature value in this way did not fix the 'equilibration 
> around 2x requested temperature' problem mentioned below - g_energy 
> still shows the temperature settling at a numerical value around 4.988.

Can anyone suggest a reason for this strange thermostat behaviour ?

Many thanks,
Steve Kirk

-- 
Dr. Steven R. Kirk           <Steven.Kirk at hv.se, S.R.Kirk at physics.org>
Dept. of Technology, Mathematics & Computer Science  (P)+46 520 223215
University West                                      (F)+46 520 223299
P.O. Box 957 Trollhattan 461 29 SWEDEN       http://taconet.webhop.org