CCL:G: geometry optimization convergence problem

[Sebastian Kozuch <kozuchs-.-yahoo.com>]

I've never tried GDIIS, but what sometimes work with flat potential surfaces is making small steps. You can write OPT(MAXSTEP=N), with N=5 or something like that (the default is N=30, equivalent to 0.3 Bohr). With this it's harder to oscillate around a minimum.

 

xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
..........Sebastian Kozuch...........
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
...The Lise Meitner-Minerva Center...
.for Computational Quantum Chemistry.
...Hebrew University of Jerusalem....
.....kozuchs(_)yfaat.ch.huji.ac.il.....
http://yfaat.ch.huji.ac.il/kozuch.htm
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

---

From: Alexander Hoepker achoepker]*[gmail.com <owner-chemistry(_)ccl.net>
To: "Kozuch, Sebastian " <kozuchs(_)yahoo.com>
Sent: Thursday, April 16, 2009 4:08:53 PM
Subject: CCL:G: geometry optimization convergence problem

Dear CCL community,

I am having difficulty with a particular geometry optimization using B3LYP level of theory using a 6-31G(d) basis set while constraining a bondlength using the ModRedundant keyword in Gaussian. The geometry sequence energies drop nicely but then plateau at an average value while oscillating endlessly without ever converging. The Maximum and RMS Forces have converged whereas the Displacements have not. Would GDIIS help with this?

Any suggestions would be very much appreciated.

Thanks,

Alexander Hoepker

Cornell University

Department of Chemistry and Chemical Biology