AMBER Archive (2006)

Subject: RE: AMBER: ab initio terminology query

From: Hayden Eastwood (
Date: Wed Nov 01 2006 - 14:12:49 CST

Hi Guys

Thanks for the feedback. The example question

"Data was computed at the B3LYP/6-31++G(2d,2p)//HF/6-31G(d) level of theory"

actually comes from the AMBER manual (sorry, should have mentioned this!),
page 23 under the Glycam-04 force field section. Indeed, my own
understanding of the problem has been that:

1. Optimisation should be carried out at whatever level of theory is
computationally feasible (ie better optimized structure = better final
partial charges).

2. ESP charges for solution phase residues should be derived at the
HF/6-31G(d) level of theory for
the reason that Ross stated (ie fortuitous cancellation of errors that
permit the modelling of a solvent polarized environment).

3. ESP charges for gas-phase should be calculated at the B3LYP/cc-pVTZ level
of theory because this level (being more accurrate than HF/6-31G*) doesn't
exagerate the dipole moment and hence better represents a non-polarizing
gas-phase environment.

Please correct me if I'm wrong on this...

At the risk of making this discussion laboured I have a few final
clarifications to make:

1. I am developing some sugar residues (for use within an implicit solvent
environment) using the gaff parameter set. Am I right in presuming that I
should use HF/6-31G* to get the partial charges? I chose this charge model
for the reason that Ross gave (fortuitous cancellation blah blah...) and
because it's consistent with the charge model of the GLYCAM force field
sugar residues. However, Ross's earlier comment:

"In all cases the single point ESP should be calculated using the same level
of theory as the rest of the force field was fit to."

Confuses me...according to the gaff paper (Development and Testing of a
General Amber Force Field. J Comput Chem 25: 1157-1174, 2004) the gaff bond
parameters are calculated at (MP2/6-31G*), yet the charges are derived at
HF/6-31G*. This is one example of a few in which the charges are obtained at
different levels of theory to the other parameters.

Sorry, please bear with me on this, I'm trying to get to the bottom of the
many (apparently) contradicting views on this issue that I've heard over the
last 8 months....

2. Is it safe to assume that I can use the ff02.pol.r1 force field to
perform explicit water calculations? Presumably it doesn't matter that the
B3LYP/cc-pVTZ derived charges don't account for the average polarizing
nature of the solvent because the ff02 parameter set introduces the required
polarization arising from the explicit solvent.

3. If I wish to run the gaff parameterized sugar molecule in explicit water,
do I need to change the charge model from HF/6-31G* to something else? If I
believe Ross's statement from earlier:

"The FF03 force field uses a different level of theory [to HF/6-31G*] for
the charge calculation that includes an implicit solvent model to obtain
solvent phase charges. If you use a higher level of theory for the ESP
calculation you will over polarise your molecule when simulating it in

Again I'm confused here...according to the reference given on page 20 of the
amber manual (the ff03 united atom force-field section):

 "...this force field [The united atom one] uses the same charging scheme as

When I looked up the reference for this, I saw that the charges were derived
at the B3LYP/cc-pVTZ level, which is the same as that mentioned in the
manual for the ff02 force field. If the ff02 force field describes gas-phase
charges and the ff03 force field approximates solvent charges (each of which
will be different in nature), then why do they have exactly the same basis
set for deriving the charges?

Please help me out here!


Hayden Eastwood

The AMBER Mail Reflector
To post, send mail to
To unsubscribe, send "unsubscribe amber" to