[gmx-users] simulation with non-zero total charge

[Mark Abraham]

seunghwan lee wrote:
> Hi,
> 
> I am trying to simulate lipid bilayer with protein, which has net positive
> charge. If I run simulations after neutralizing, I don't have any problem,
> whether I use PME or reaction-field. But, when I try to run it with net
> positive charge using reaction-field, I got domain-decomposition error
> after 1800000 steps. In log file, I have
> 
>> DD  load balancing is limited by minimum cell size in dimension Z
>> DD  step 1878999  vol min/aver 0.840! load imb.: force 18.6%
> 
> and the actual error message was
> 
> DD cell 0 0 5: Neighboring cells do not have atoms: 4484
> 
> DD cell 0 0 3: Neighboring cells do not have atoms: 4477
> 
> -------------------------------------------------------
> Program mdrun, VERSION 4.0.5
> Source code file: domdec_con.c, line: 679
> 
> Fatal error:
> DD cell 0 0 5 could only obtain 51 of the 52 atoms that are connected via constraints from the neighboring cells. This probably means your constraint lengths are too long compared to the domain decomposition cell size. Decrease the number of domain decomposition grid cells or lincs-order or use the -rcon option of mdrun.
> 
> My questions are:
> (1) Is this error anything to do with the fact that the system has net charge?

Potentially. Something's causing it to explode, but it can be hard to 
tell why after the fact.

> (2) How can I fix this with "-rcon" option? I don't want to change LINCS order, etc.

Reading mdrun -h is the first port of call, however changing this option 
is probably just fixing the symptoms, not the cause.

> (3) Is there any parameter that I need to be careful when I simulate systems with non-zero net charge (especially lipid system)?

With RF, I would have guessed that there's probably not much to worry 
about. Probably nobody's studied it in any detail, because it's so easy 
to neutralize, and so easy to defend a neutral simulation as a realistic 
model of real physics...

Mark