[gmx-users] CHELPG charges from gamess very large

[Douglas Ridgway]

On Wed, 10 Sep 2003, Lakshmi Padmavathi wrote:

> i calculated the CHELPG charges by using GAMESS for a molecule which contains 
> megnesium. but the charges are very large. for example, MG has 10.000 
> charge.i used STO as basis set.

I don't know much about this, but that doesn't sound right. You might try
a larger basis set: 6-31G* seems to be popular, but it may depend on your 
molecule. With GAMESS you use

 $BASIS NGAUSS=6 GBASIS=N31 NDFUNC=1 $END

Hartree-Fock at 6-31G* will overestimate dipole moments, leading to 
larger partial charges, but apparently this is good because it helps 
compensate for a lack of polarizability in the classical model. (See eg 
Cieplak JCC 22(10) 1048 2001.)

There are various methods to compute charges from the electrostatic
potential. The comparison of Sigfridsson and Ryde (JCC 19(4) 377-395 1998)  
has been recommended on this list; CHELPG gets criticized for being too
sensitive to too many points too close to the atoms. The GAMESS default or
Merz-Kollman methods are weighted further out, and might be better.

doug.
ridgway at dridgway.com