Hi Dan (and Francois),
I'll throw my two cents in. For molecules that do not have a permanent dipole moment, determining classical atom-centered point charges that are representative of the quantum charge distribution is difficult to do. As far as I know, such a problem is still unsolved.
One would have to validate the determined charges (as always) using experimental/QM observable that is dependent upon the nonbonded forces. The graphene sheet you attached shows a slight asymmetry in the structure, so it should have a "very" small permanent dipole. However, I don't think it would be enough to produce reliable partial atomic charges.
Like Dr. Case mention, the carbon atoms should predominantly have charges of zero for an large (i.e. infinite) sheet. In your model there are many carbons that are attached to hydrogen atoms (i.e. a relatively high ratio of edge-to-embedded carbons compared to reality), which might give rise to artificial MD behavior due to partial atomic charges and Lennard-Jones parameters. This would be true if you are trying to model a large sheet. However, if your model is representative of an experimental structure, then this would not be the case.
Cheers,
Karl
----- Original Message -----
From: "FyD" <fyd.q4md-forcefieldtools.org>
To: dpiraner.caltech.edu, "fan wang" <fan.wang.q4md-forcefieldtools.org>
Cc: "AMBER Mailing List" <amber.ambermd.org>
Sent: Tuesday, June 18, 2013 9:20:52 AM
Subject: Re: [AMBER] Graphene topology from GAFF
Dear Dan,
I do not think the method of the graphs is suitable to determine
chemical equivalencing here: I would _not_ use Antechamber (from what
I understand more there are cycles more you need memory; the problem
you encountered; my guess is that upgrading to the last version will
change nothing) to determine the charges here, besides the problem
related to geometry optimization.
As underlined by Dr Case the atomic charges of the hydrogen atoms are
supposed to be slightly different; thus I would derive the atomic
charges of this molecule (i.e. I would not set all of them to zero).
I just ran R.E.D. Python on your molecule to test how goes chemical
equivalencing in this case. It should work with our algo. The problem
is that your molecule is quite big for our cluster, and it might take
'some' time to get an optimized geometry. Then, you will be able to
decide if you want RESP vs ESP; Connolly surface vs CHELPG, etc... The
RRMS of the fit is important in this case; if it is not good we will
propose you alternatives to improve the fit; all that is quite stable
in R.E.D. Python by now...
I let you know what we get...
regards, Francois
> I'm trying to obtain .top and .crd files for a graphene sheet using
> AMBERtools 1.4. I generated the graphene pdb file using VMD's nanotube
> builder plugin, and added terminal hydrogens using PyMOL. I then replaced
> the VMD atom names (C and H01) with what I believe to be the appropriate
> GAFF atom names: CA and HA. The structure was then used as an input for
> Antechamber, to be then loaded into tLEaP:
>
> antechamber -i graphene.pdb -fi pdb -o graphene.prepin -fo prepi -c bcc -j
> 4 -at gaff
>
> I received a huge amount of warnings from Antechamber about exceeding 10
> residues, a final warning about reallocating memory due to exceeding
> MAXBOND, and then the program appeared to freeze. I can't tell if it's
> still calculating slowly or simply stuck in a loop.
>
> Is the procedure I'm following correct for defining a graphene molecule?
> Is the format of the PDB file, attached, correct? I wasn't able to find
> much information about the setup online, other than that it has previously
> been done. Any help would be greatly appreciated.
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Tue Jun 18 2013 - 01:30:03 PDT
