Dear ADF-list members,

Summary:
--------
This advice from SCM concerns only those ADF users who are using
the COSMO solvent model on symmetric molecules. Such calculations
can be done more safely by specifying ALLPOINTS in the input file.
The calculations will then take longer.

Full story:
-----------
Some months ago, Dr. Atanasov posted an E-mail to this list concerning
problematic convergence for COSMO geometry optimization for a symmetric
molecule, using ADF.

After discussions with Dr. Atanasov, Dr. Pye (original developer of COSMO
in ADF) and some testing, we have come to the following conclusions:

- it is currently not enforced that the COSMO surface has exactly the
  same symmetry as the molecule. Because of the way the COSMO surface
  is constructed, one would expect it not to deviate much from the true
  symmetry in the bulk of practical cases. However, there can be some
  side effects from this in special cases. We are thinking of small,
  highly symmetric molecules in particular.

- For small, highly symmetric molecules, the COSMO contributions to the
  Cartesian gradients currently calculated by ADF may be slightly different
  depending on the symmetry which is specified in the input file. We consider
  this a bit unfortunate and have therefore switched on the INPUT keyword
  ALLPOINTS for all COSMO calculations that are not single-point calculations.
  As a result, more integration points will be used for COSMO geometry
  optimizations on symmetric molecules. This will become effective in the
  next release. At the moment, you can reach the same effect by specifying
  ALLPOINTS in the input file.

- In principle, specifying ALLPOINTS may also have slight influence on
  single point COSMO calculations. However, in the cases we tested the
  effects were negligible. We have not switched on ALLPOINTS by default for
  single point COSMO calculations for this reason. If you are in doubt,
  you can specify ALLPOINTS in the input file to check its effect.

- Molecules without symmetry are not affected by all this. Specifying
  ALLPOINTS does not lead to a different integration grid there.

- For a symmetric molecule, the COSMO contribution to the force may have
  a small component that is inconsistent with the symmetry of the molecule.
  This is true whether or not ALLPOINTS is specified.
  If the full symmetry if the molecule is used during the optimization,
  the Cartesian forces are converted into forces for the degrees of freedom.
  In this step the nonsymmetric part of the COSMO forces will be removed.
  For this reason, the geometry step will comply with the symmetry of the
  system. If the optimization is done in a lower symmetry, the small spurious
  term in the COSMO forces may lead to a symmetry breaking of the system. We
  consider this to be an artifact of the current COSMO surface generation
  and advise not to lower the symmetry of the system.

- In Dr. Atanasov's case, a COSMO optimization which did not converge in
  200 geometry iterations without ALLPOINTS, the optimization converged in
  4 cycles with this keyword inserted.

- In general we expect this whole issue to be of minor importance in the
  vast majority of cases, otherwise we would have expected many more
  problem reports than the just one by Dr. Atanasov. Nevertheless, we
  thought the information in this E-mail might be of use to some ADF-COSMO
  users.

- A COSMO-FAST option will be made available to overrule the new default
  for COSMO geometry optimizations and use only the symmetry unique part
  of the numerical integration grid.

Please send any comments or questions to support_at_scm.com. If you consider
your reply to be of braod interest you can of course post it to the ADF-LIST
as well.

Best regards,
Stan van Gisbergen, on behalf of the SCM team.

------------------------------------------------------------------------
Dr. S.J.A. van Gisbergen SCIENTIFIC COMPUTING & MODELLING NV
Phone: +31-20-4447626 Vrije Universiteit, Theoretical Chemistry
Fax: +31-20-4447629 De Boelelaan 1083
E-mail: vangisbergen_at_scm.com 1081 HV Amsterdam, The Netherlands
                               http://www.scm.com
------------------------------------------------------------------------
Received on 2002-11-22 09:44:42