[gmx-users] Re: charges

[Erik Lindahl]

Christoph Freudenberger wrote:

> Erik Lindahl wrote:
>
>> The nice thing with gaussian is that newer versions allow you to just 
>> enter a set cartesian coordinates and the number of electrons for 
>> each atom, and then it can either optimize the structure or just 
>> derive partial charges (e.g. with CHELPG).
>
> The standart method of gaussian for calculating charges (Mulliken) is
> known to be quite bad.
> Which one should be used to produce relieable results for MD?

CHELPG is stable and good to very good even for large molecules. 
Merz-Kollman  might be slightly better for small molecules,
but it is reported to be much more sensitive to the conformation, so it 
can be unstable for large molecules.

For the single Gaussian run I've done I used CHELPG, simply because it's 
the suggested standard in OPLS, and my molecules were large (lipids).

You select the method to use with pop=chelpg or pop=mk

Cheers,

Erik