[gmx-users] rvdw parameters

[Berk Hess]

>>Also, does anyone know what are the 'ideal' parameters for rcoulomb and 
>>rvdw
>>for simulations where electrostatics is computed using the PME method?
>
>No ideal, just only one correct option: the parameters must be exactly
>what they were during forcefield development. This is 0.8/1.4 (twin-range)
>for the G42* forcefields, both for coulomb and rvwd. Thechnically, using
>PME in stead of twin-range is not correct either, but it won't hurt and
>for some (heavily charged) systems it is inevitable to use PME.
>Especially the short-range cutoff (0.8) *SHOULD* *NEVER* *BE* *CHANGED*.
>For the long range it is not so critical, but you would rather use 0.8/PME
>in stead of increasing the 1.4 long range cut-off.

This is nonsens.
Only the long-range cut-off is important for parametrization. For gromos96
forcefields the rvdw should be >= 1.4 nm.
The rlist cut-off can be optimized for performance and/or accuracy. Also the
neighborlist update frequency (nstlist) and the timestep play a role here.
When rlist is short, you need to update more often, since the forces change
more rapidly at a short distance of rlist.
In general the gromacs pme implementation works most efficient with an rlist
and rcoulomb of around 0.9 or 1.0.
The gromos rlist of 0.8 and nstlist of 5 is not very accurate for non-PME or
non-PPPM electrostatics, but for pme it should be fine.
With rlist 0.9 or 1.0 with PME one could even use an nstlist of 10,
especially for a bilayer simulation in octane where motions are propably 
slower
than the motion of water molecules.

Berk.



_________________________________________________________________
Send and receive Hotmail on your mobile device: http://mobile.msn.com