From owner-chemistry@ccl.net Sat Aug  6 13:34:50 2005
From: "CCL" <owner-chemistry * server.ccl.net>
To: CCL
Subject: CCL: RESPONSE SUMMARY: Gaussian 03, Opt. Changing ''maximum allowed number of steps''.
Message-Id: <-28951-050806132601-15310-UGblJhHCRhLlm/aY0tUb9g * server.ccl.net>
X-Original-From: "Niels Johan Christensen" <s991418 * student.dtu.dk>
Content-Transfer-Encoding: 8bit
Content-Type: text/plain;
	charset="iso-8859-1"
MIME-Version: 1.0


Sent to CCL by: "Niels Johan Christensen" <s991418 * student.dtu.dk>
Dear All,

Some weeks ago I asked for advice on how to change the maximum allowed 
number of optimisation steps in g03 -- that is to say, how to 
set the MaxCycles variable to any desired value.

I got the following response from D. J. Fox (help * gaussian.com):

>G03 imposes an absolute limit on the number of optimization cycles
>based on the size of the molecule. The thinking is that if you are well
>beyond the normal number of cycles there may be something more serious
>wrong and the amount of numerical error built up in the force constant
>matrix will become large enough to be a problem by itself.
>
>The first thing to do is to examine the last few structures at
>this point. Is the structure still close to the local minimum you wanted
>to explore? Is the gradient converged but the displacements are still
>not? Are the approximate force constants of a reasonable size, say no
>smaller than 0.01? Also is this the ultimate level of theory you plan to
>report or is this preliminary to later calculations?
>
>If the structure is close then FREQ at this point may be sufficient
>to confirm it really is a minimum or give you a good start for going on
>to the next level of theory.
>
>If it is a good structure but clearly not the minimum OPT=CalcFC from
>this geometry will likely get you there quickly, much more quickly than
>trying to continue with gradients only and poor force constants.
>
>If it is not clearly the minimum you meant to explore look back along
>the optimization, did it diverge at some point and you might want to
>start from that structure. Perhaps you need to add a constraint and
>reoptimize from the beginning to keep it from wandering on a flat potential
>energy surface.
>
>In any case the code does not permit increasing the number of optimization
>cycles without limit.

Thank you Dr. Fox, OPT=CalcFC worked very well in my case.

I found out, however, that g03 will actually permit the desired value of
MaxCycles in a second run. One simply loads the checkpoint file of the failed
optimisation. Thanks to Dr. Steen Hammerum for this hint.


Best Regards,
Niels Johan Christensen