From chemistry-request@ccl.net Tue Jul 14 13:40:18 1992
Date: Tue,  14 Jul 92 16:44:59 +0200
From: "Prof. Benzion Fuchs - Tel-Aviv Univ."
Subject: CAMEO
To: CHEMISTRY@ccl.net
Status: R

Hi there,
Can anyone provide the EM address of the CAMEO group of
W. L. Jorgensen at Yale? Thanks, Benzion Fuchs <bfuchs@taunos>

From chemistry-request@ccl.net Tue Jul 14 20:27:54 1992
Date:     Tue, 14 Jul 92  15:44 GMT
From: RWOODS%VAX.MOLECULAR-BIOPHYSICS.OXFORD.AC.UK@OHSTVMA.ACS.OHIO-STATE.EDU
To: CHEMISTRY@ccl.net
Subject:  Geom Optimization with G90
Status: RO


I have been using the Berny optimizer (default) with G90 and recently
I have noticed some peculiarities.  Briefly, at the HF/6-31G* level
with approximately 82 variables (out of 83) being optimized in several
cases the algorithm required over 50 steps to converge!  The geometries
were initially chosen based on previously optimized structures (same basis)
in which one variable (a torsion angle) was adjusted to create the new
geometry.  I might not have been extremely concerned except that the latest
calculation failed to optimize altogether and actually took a 20Hartree
step before failing.

These are not particularly bizarre structures and converge readily
when run with g88.  I re-ran the failed optimization specifying
OPT=NEWTON (avoiding the RFO step in Berny which is a new default)
and the structure optimized in 11 Steps!

Has anyone else run into similar problems with G90?  Does anyone
have any comments on the RFO modification to the Berny optimizer?

Rob Woods
Glycobiology Institute
University of Oxford
South Parks Road
Oxford, OX1 3QU