From: Mayne, Christopher G (cmayne2_at_illinois.edu)
Date: Sat May 04 2013 - 09:37:11 CDT

Prof. Suman de Araujo,

For relatively simple systems, our strategy is to iteratively run a few cycles of simulated annealing mode followed by downhill. This allows us to sample a reasonable amount of the charge space, and ultimately arrive at the best fit as for the target data. For complicated systems, one needs to pay more attention to charge constraints and weighting factors for dominant interactions, and we strongly suggest becoming familiar with the COLP utility (accessible from the "Results" section) to monitor charge performance and provide insight on how to best modify the various parameters that control the optimization.

Based on the few pieces of information I can glean from your email about your molecule, I would suggest that it falls into the "complicated" category. Without more information, I can't say much more than that.

Regards,
Christopher Mayne

On May 3, 2013, at 3:40 PM, Alexandre Suman de Araujo wrote:

> Hi all
>
> I'm modeling a small molecule using FFTK plugin in VMD version 1.9.2a14.
>
> I have some questions about the charge optimization procedure shown in the screencasts tutorial. There, after QM calculation of the interaction energy between water molecules and hydrogen bond donor/acceptor groups, it goes to the charge optimization step.
>
> Basically, this step consists in loading of structure information, loading of force field parameters already obtained, guess of charge constraints and values, loading of QM log files and, finally, the adjustment of the partial charges using an optimization procedure (simulated annealing or downhill).
>
> In the "Results" tab, there is an option to set the calculated charges as initial charges to perform, in my acknowledgment, another optimization calculation. However, after this new optimization the charges are very different from the first ones obtained and used as new initial charges. The partial charges values only converge after some cycles of optimization, and my question is about the obtained results:
>
> - If I compare these converged charges with Mulliken charges from SP calculation, I found strong disagreements between them. Including signal inversion of some charges (QM = -1.4, Final optimized partial charges = 1.3).
> - Another inconsistency are positive partial charges of hydrogen bond acceptors and negative of donors.
>
> The procedure to adjust the charges in several cycles is the correct approach? If yes, where I am doing mistakes? If no, how should I proceed?
>
> Best regards
>
> --
> Prof. Dr. Alexandre Suman de Araujo
> Departamento de Física
> Instituto de Biociências, Letras e Ciências Exatas - IBILCE
> Universidade Estadual Paulista
> e-mail: asaraujo_at_ibilce.unesp.br
> Phone: +55 (17) 3221-2566/3221-2240
>