[gmx-users] Tabulated potentials

[Francesco Mambretti]

 Dear GROMACS users and developers, I am using GROMACS 5.0.1.

I would like to use tabulated potentials (Aziz ones for argon and krypton in particular); as a starting test, I began with LJ tabulated potential. I followed the instructions on the manual, and I did well because my .xvg file of Lennard-Jones 6-12 was substantially identical to the one provided in /share/gromacs/top/table6-12.xvg.

I set the .top file and .mdp file according to the instructions on the website and on the manual, besides following the attached pdf in http://www.gromacs.org/Documentation/How-tos/Tabulated_Potentials?highlight=tabulated.

.top [ defaults ] section has nbfunc and comb-rule equal to 1; C and A parameters in the [ atomtypes ] sections are set equal to 4 epsilon sigma^6 and 4 epsilon sigma^12 respectively.

.mdp parameters are: cutoff-scheme = group, coulombtype = user and vdwtype = user.

My system is a molecular system of 256 argon atoms. PBC = xyz, no long-range dispersion corrections. 
Until now, all worked fine. The problems began when I decided to freeze a group of atoms in my system. I added the right atom numbers to the index file, and add energy group exclusions in the .mdp file. Moreover, I specified in the .mdp that I wanted the motion frozen in the 3 directions.

The problem is that ordinary LJ of GROMACS (not tabulated) goes on well, and the system solidifies around the frozen seed (an fcc-like seed of 13 atoms); but, with the same conditions, LJ tabulated (also with the table given in the share/top folder) behaves differently. In particular, frozen atoms seem to repulse very strongly the others, in such a way that a hole creates around the frozen nucleus.
Also the g(r) are very different, and while in the first case I obtain a more or less ordered solid, in the latter I obtain a strongly disordered solid.
What am I missing? Should I apply particular switches or corrections?
Why does GROMACS behaves differently if I give it the same potential, but tabulated in the latter case and not tabulated in the first?
The only thing that changed between the two cases is the use of a tabulated interaction instead that LJ GROMACS default.
Do you know anything about similar problems with freeze groups?

Thanks a lot

Francesco Mambretti