[gmx-users] Positive Coul. recip. term

[Mark Abraham]

On 24/02/2012 7:35 AM, Denny Frost wrote:
> Dear all,
> I am trying to equilibrate a solvent of pure ionic liquid.  The system 
> keeps exploding (after 2-5 ns) and I am not sure why, though I believe 
> coulombic interactions are to blame.  This is because the Coul-SR term 
> is negative, but the Coul. recip term is very positive throughout the 
> entire run (giving the entire system a positive potential energy).  I 
> think this means that the short-range electrostatics are okay, but the 
> long range electrostatics (calculated with PME) are not.  Does anybody 
> have any suggestions as to why this would happen?  I have used the 
> exact same PME input parameters for another ionic liquid that works 
> just fine.  They are listed below.
>
> rcoulomb = 1.2
> fourierspacing = 0.12
> pme_order = 4
> ewald_rtol = 1e-5
>

The most likely explanation for this is a problem linking the 
executable, perhaps dynamically. The FFT process is then garbage which 
normally generates inappropriate values for Coul. recip. terms. I'd 
suggest re-compiling GROMACS, perhaps using static linking. You could 
use the GROMACS built-in fftpack for simplicity while diagnosing.

Mark