From chemistry-request@ccl.net Mon Jul 22 14:57:52 1991
Date: Mon, 22 Jul 91 14:41 EDT
From: "Scott Le Grand" <SML108@PSUVM.PSU.EDU>
Subject: Conformational Entropy Summary
To: chemistry@ccl.net
Status: R
Here is a repost of my original question about conformational entropy and
the various replies which described methods of estimating it that
I received. Thanks for the help everyone!
Scott
Date: Tue, 25 Jun 91 00:56 EDT
From: "Scott Le Grand" <SML108@PSUVM.PSU.EDU>
Subject: Conformational Entropy
To: chemistry@ccl.net
Sender: chemistry-request@ccl.net
Errors-To: owner-chemistry@ccl.net
Precedence: bulk
Is there any way to crudely estimate the conformational entropy of a
particular conformation of a molecule that does not involve the O(n"3)
matrix inversion of the Hessian matrix of the potential energy function?
I'm unfortunately betting that there isn't...
Scott Le Grand
Date: Tue, 25 Jun 1991 05:42:35 PDT
Sender: Sundar_Sundararajan.XRCC@xerox.com
From: sundar.XRCC@xerox.com
Subject: Blank Mail Note
To: sml108@psuvm.psu.edu
Reply-to: "sundar.xrcc-ns@xerox.com".XRCC@xerox.com
Message-ID: <"25-Jun-91 8:42:28".*.Sundar_Sundararajan.XRCC@Xerox.com>
Dear Scott,
As for the conformational entropy, in the case of polymers, it is often
calculated with the energy and the partition function. See for example, P.R.
Sundararajan, Macromolecules, 23, 2600 (1990)
Regards
Sundar
Date: Tue, 25 Jun 91 10:57:31 -0500
From: "Richard A. Caldwell" <caldwell@utdallas.edu>
To: SML108%PSUVM.PSU.EDU@vm.utdallas.edu
Subject: Re: Conformational Entropy
how crude is crude? The Benson group equivalent technique (Benson, S. W.,
"Thermochemical Kinetics," 2nd Edition, John WIley & Sons, N. Y., 1976)
provides a CNQM (figure it out!) technique which gives heats of formation
to +/- 1 kcal/mol in many cases and also provides a way to estimate
standard entropies of formation. I ahve no personal experience with
the latter but expect it to be comparably good. It isn't generally used
for isolated conformations but seems to me adaptable with a little
ingenuity. Dick Caldwell
09:45:02 EDT
Date: Wed, 26 Jun 91 09:45:02 EDT
From: jacque@isadora.albany.edu (Jacque Fetrow)
Message-Id: <9106261345.AA24623@isadora.albany.edu>
To: sml108@psuvm.psu.edu
Subject: entropy
scott - i don't know anything about a matrix inversion of the hessian
matrix, so this suggestion may not make any sense, but...
couldn't you estimate the entropy of a residue using the good old
equation S=k(lnW) where W is the number of ways of arranging the
system. you can estimate W from a rotamer library--you know the
possible allowed conformations. i don't know if this will work
for what you want...
Date: Thu, 27 Jun 91 08:05:22 -0400
Message-Id: <9106271205.AA05285@esds01.es.dupont.com>
From: fredvc@esvax.DNET.dupont.com
To: "sml108@psuvm.psu.edu"@ESDS01.DNET.dupont.com
Cc: FREDVC@esds01.es.dupont.com
Subject: Conformational entropy: part II
Conformational entropy, as I understand it, is simply
{sum over i} Xi*ln(Xi)
where Xi is the mole fraction of the i-th conformation. This, in turn, is
given by
exp[-Ei/RT]/{sum over j} exp[-Ej/RT]
where Ei is the (relative) energy of the i-th conformation. Is there something
special in the realm of protein/peptide chemistry that I am missing???
Frederic A. Van-Catledge Office: (302) 695-1187
Scientific Computing Div. FAX: (302) 695-9658
Central Res. & Dev. Dept.
The Du Pont Company
P. O. Box 80320
Wilmington DE 19880-0320 Internet address: fredvc@esvax.dnet.dupont.com
~~~~~~~~~~~~~~~~~~~~~~~~~~~~
NEW E-MAIL ADDRESS
From chemistry-request@ccl.net Mon Jul 22 16:17:04 1991
Date: Mon, 22 Jul 91 16:00:52 EDT
From: Shi Yi Yue <shiyi@bobino.bri.nrc.ca>
To: chemistry@ccl.net
Subject: Summary of Force Constant Calculations
Status: R
Hi,
Here is a repost of my original question about force constant
calculation and the various replies I received.
Thanks for the help everyone!
!
! Hi,
! I will be very appreciated if someone could point out that
! there has been any approvements from Gaussian 80, 82, 86, 88
! to Gaussian 90 for the calculation of the Force Constant.
! The reason for asking this question is that what I have in
! hand is Gaussian 80. But what I want to carry out is some bond
! constants from the combination of {C, N, O}. I hope Gaussian
! 80 is not too old for this job.
! By the way, I think STO-3G is OK for this calculation
! because of the size of the molecules I have to calculate.
! Any comments, any suggestions?
! Thank you in advance!
!
---------------------------------------------------------------------
>From rbw@msc.edu Fri May 10 21:46:05 1991
You might wish to read sections 6.3.1 through 6.3.10 in the book
Ab Initio Molecular Theory where 3-21G basis sets are recommended.
Richard Walsh
Minnesota Supercomputer Center
--------------------------------------------------------------------
>From cpaulse@magnus.acs.ohio-state.edu Fri May 10 22:27:19 1991
For hartree fock calculations, gaussian 82 and 86, 88, and 90 most
definitely use the method of analytic second derivatives for force
constant evaluation. As for gaussian 80, I'm not sure. You should
be able to check the documentation quite easily in order to find
this out. As for improvements, other than a more efficient algorithm
found in gaussian 88-up for evaluation of second derivative integrals,
there should be no difference in the numbers you obtain using any of
the newer programs, provided the calculation is analytic.
Hope this helps.
Chris
-------------------------------------------------------------------
>From oles@ulrik.uio.no Sat May 11 08:57:21 1991
You suggest:
> By the way, I think STO-3G is OK for this calculation
>because of the size of the molecules I have to calculate.
> Any comments, any suggestions?
Indeed. Bond constants for [C,N,O] containing molecules
will probably be better modeled by a modern semiempiric method
such as AM1. Minimal basis Hartree-Fock calculations cost more
and will probably give you less.
------------------------------------------------------------------
>From ryanmd%phvax.dnet@smithkline.com Sat May 11 14:12:33 1991
I concur with the recent posting about sto3g, but several additional
cautions are needed. Semi-empirical calculations can in some cases
get the optimized geometries quite wrong in torsion angles, and
therefore you need to check for this in the literature.
If you need force constants for force field calculations you also
need to worry about values that will fit with the other terms of
the field. I do not feel that sto-3g will do this well at all,
in fact 6-31G** would be more suitable, and then not even force
constants may be correct but rather a set of discrete points along
the bond stretch you are interested in could be better. You will
also have to deal with the transformation from normal modes (what a
frequency calculation will give you) to internal coordinate
representation if you do not carry out many calculations along the
bond stretch desired. Normal modes of vibration are *not* the same
as internal coordinates. If you need this for a force field you may
also have to worry about balancing non-bonded terms and other angle
or coupling terms to your new ones. Charges in particular can be
troublesome, and you should use atom charges fitted to a potential
field using at least a split-valence basis set.
Finally, G80 is a dinosaur. G90 will probably be 10 to 50 times
faster and has features that will permit you to calculate much
larger molecules as well. I have found G90 to be ~3-5x G88, which
was another 3-10x G86 which was in turn faster than predecessors.
There are also alternatives, I would recommend Spartan, faster and
tied to the graphical interface, very nice. Call Wavefunction at
714-955-2120.
There is much to be considered if you hope to get numbers that won't
lead you astray, I strongly suggest spending a large chunk of time
in the literature.
M. Dominic Ryan, Ph.D