Dear Carlos,
Everything Francois and Karl have written is correct, but there is an
overarching issue that might be easy to miss. That is, it is critical to
be clear that the goal of the RESP fitting is to generate a set of atomic
charges that can work together with all of the other atomic charges in the
force field to result in reasonably accurate intermolecular forces. The
'working together' part is the critical piece. Thus, it is more important
that the charges be BALANCED with respect to the other charges in the force
field than that they be accurate with respect to experimental MEPs.
If one is working with one of the Cornell-type force fields (e.g. ff12SB,
ff99, parm94) this means that one must use a computational method that is
consistent with what was used in those force fields (e.g. HF/6-31G(d)).
The reasons Cornell et al. chose to use HF/6-31G(d) are described in the
papers referenced by Francois.
Cheers,
Brent
On Wed, Jul 17, 2013 at 4:18 AM, Karl N. Kirschner <
kkirsch.scai.fraunhofer.de> wrote:
> Dear Carlos,
>
> I second Francois's suggestions for where to start looking for knowledge
> concerning this. I can only add my perspective and experiences.
>
> The MEP is computed directly from the wavefunction, and is thus an
> experimental observable. However, experimentally it is difficult to measure
> the MEP, which would allow for a comparison to the calculated one. The
> final optimized wavefunction, which is used to provide the MEP, is
> determined by the basis set and by the theory that is employed (e.g. HF,
> MP2, DFT).
>
> Alternatively, one could compare to dipole and quadrupole moments for
> which a wealth of experimental small molecule data exists. I have done this
> in a study on methanol (doi:10.1016/j.cpc.2011.05.018). I looked at how 18
> different approaches for generating partial atomic charges effect the the
> molecule's dipole moment's magnitude and orientation, using theories
> ranging from OPLS charges to HF/6-31G(d) to MP2/ACCT. With certain
> theories, one can reproduce the experimental gas-phase dipole moment well.
> For example, a 0.6% error was obtained using HF/ACCD with differentiating
> the methyl hydrogens into trans versus guache; forcing the methyl hydrogen
> charges to be equivalent value (as would be needed in an MD simulation)
> results in an error of 17.8%. However, this resulting error is suggested to
> be advantageous (see the Cornell papers) for modeling systems in solution
> due to polarization effects.
>
> The end game for developing partial atomic charges and Lennard-Jones
> parameters, is that they reproduce some intermolecular forces dependent
> experimental observable (e.g. enthalpy of vaporization, diffusion rate,
> density). They also have some effect on the relative potential energy curve
> of torsion angles, which is arising from 1-4 interactions. However, in my
> research I have not seen this to be significant in that I believe the
> errors within the bonded parameters are greater in most force fields.
>
> If you want to check the performance of the parameters given by
> antechamber, with regards to reproducing QM-generated bond, angle, and
> torsion curves, you are welcomed to use Wolf2Pack.
>
> Best regards,
> Karl
>
> ----- Original Message -----
> From: "FyD" <fyd.q4md-forcefieldtools.org>
> To: "AMBER Mailing List" <amber.ambermd.org>
> Sent: Tuesday, July 16, 2013 6:11:26 PM
> Subject: Re: [AMBER] Resp fiiting method and theory
>
> Dear Carlos T. Nieto,
>
> I would start by:
> http://ambermd.org/doc6/html/AMBER-sh-19.4.html#sh-19.4
>
> then see: http://q4md-forcefieldtools.org/RED/reference.php
> read:
> W.D. Cornell, P. Cieplak, C.I. Bayly & P.A. Kollman J. Am. Chem.
> Soc. 1993, 115, 9620-9631. [Abstract]
> C.I. Bayly, P. Cieplak, W.D. Cornell & P.A. Kollman J. Phys. Chem.
> 1993, 97, 10269-10280, [Abstract]
> P. Cieplak, W.D. Cornell, C.I. Bayly & P.A. Kollman J. Comput. Chem.
> 1995, 16, 1357-1377. [Abstract]
>
> Finally you might be interested in:
> http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2918240/
>
> regards, Francois
>
> > The question of quantum theory used and resp fitting method have been
> > asked previously in the forum.
> >
> > I´m planning to obtain parameters to a series of organic molecules (no
> > metal atoms) through antechamber.
> >
> > I´ve read several considerations about basis set dependence of the mep
> map
> > obtained, but i don´t find clear the idea regarding to quantum theories,
> > specially MP2, DFT and HF.
> >
> > I´ll apreciate if anyone could share their opinions about the use of this
> > theories and if there is some papers that mention some comparative study.
>
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
--
_______________________________________________
Brent P. Krueger.....................phone: 616 395 7629
Associate Professor................fax: 616 395 7118
Hope College..........................Schaap Hall 2120
Department of Chemistry
Holland, MI 49423
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Jul 17 2013 - 06:00:05 PDT
