From chemistry-request@server.ccl.net Wed May 8 16:33:43 2002
Received: from freyr.chem.washington.edu ([128.95.128.138])
by server.ccl.net (8.11.6/8.11.0) with ESMTP id g48KXhV11053
for ; Wed, 8 May 2002 16:33:43 -0400
Received: from localhost (fer@localhost)
by freyr.chem.washington.edu (8.11.6/8.11.6) with ESMTP id g48KWaj16412;
Wed, 8 May 2002 13:32:36 -0700
Date: Wed, 8 May 2002 13:32:36 -0700 (PDT)
From: "Fernando D. Vila"
To: Computational Chemistry List
cc:
Subject: Finite Field PT in Gaussian
Message-ID:
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Hi..
I don't send questions to the CCL very often, but it looks like this
week I'm going to send a couple.. And this one is looong.. :-)
I have been trying to calculate the dipole-quadrupole (A) and
quadrupole-quadrupole (C) polarizabilities of water by means of Finite
Field Perturbation Theory (FFPT) using Gaussian 98 Rev A.7. I have
previous experience with this method, I've used it to calculate dipole
moments and dipole polarizabilities. Now I've run into some trouble
because I need to transform the Cartesian-based properties that Gaussian
generates into the traceless ones. Right now I'm pretty confused, so I'll
try to explain the problem as clearly as I can.
I think the main question is: what is the expression of the perturbation
Hamiltonian used by Gaussian??
I have found that when a dipole perturbation is applied (lets say, in the
z direction), the perturbed energy can be written as:
E = E0 + Mz * Vz
where E0 is the unperturbed energy, Mz is the Cartesian z component of the
dipole moment and Vz is the magnitude of the perturbation. I have checked
and the Mz moment is identical to the one reported at the end by Gaussian.
This is reasonable, the problem comes when you check that for a quadrupole
perturbation the SAME formula is valid. Now, if the standard Cartesian
perturbation Hamiltonian was used, the formula should be
E = E0 + 1/2 * Mzz * Vzz
I did the same for the octupole and the hexadecapole and the apparent
Hamiltonian used by Gaussian is
H' = Mi * Vi + Mij * Vij + Mijk * Vijk + Mijkl * Vijkl
where the repeated indices are summed. This series is missing the 1/n!
coefficients in front of each term.
I tried checking if these coefficients are folded into the Cartesian
moments, but this is not the case: when the Cartesian moments are used to
calculate the traceless moments, the results are exactly correct, so no
coefficients are included (the moments are simple the Cartesian
components).
Since I don't know what perturbation is being used, I can't derive the
transformation relation that would take the Cartesian components of, say,
the dipole-quadrupole polarizability and give me the ones associated with
the traceless operators.
What is really frustrating is that the Gaussian manual is extremely terse
in the explanation of what is being done. Moreover, I've found that the
warning in the manual regarding the sign ("be careful of the choice of
sign convention when interpreting the results") is quite misleading:
according to usual standards, what Gaussian is using are not electric
fields but potential gradients.
Let me say, finally, that I have actually gone into the code to see if
there were any more explanatory comments, without success..
Hope anybody can help, thanks in advance.. Fer.
*******************************************************************************
Fernando D. Vila Voice (206)616-3207
Department of Chemistry Fax (206)685-8665
University of Washington E-mail fdv@u.washington.edu
Seattle, WA 98195, USA WWW http://faculty.washington.edu/fdv
*******************************************************************************