[gmx-developers] vdW-tabulated potential in charge-group kernel

[Tristan Bereau]

Dear all,

I've been toying with the charge-group kernel
(gmxlib/nonbonded/nb_generic_cg.c) as a means to easily implement a
somewhat exotic interaction potential. That implementation runs well in
Gromacs 4.5.4.

Since that, I've tried porting it to 4.6.6. I've found inconsistencies
between the two versions affecting the vdw-tabulated potentials:
- the C6 and C12 coefficients (in the vdwparam array) differ by a factor
of 6 and 12, respectively.
- the vdW tables (previously nblists->vdwtab, now
kernel_data->table_elec_vdw->data) are now very different--I can't tell
what's the relationship between the old and new ones, unfortunately.

As a result, the exact same simulation using a vdW-tabulated force field
with gromacs 4.5.4 and 4.6.6 yield very different energies. Has there
been a change in convention in the input force field for the
charge-group kernel? There seems to be an inconsistency somewhere.

Many thanks for your help.

Best,
Tristan