> I've been trying to do a single point calculation on a 22 atom, 19 atom, and 18
> atom Pd cluster. I haven't been able to get a calculation to converge. I
> tried turning off the diis and lowering the mixing factor to 0.02, and have
> gotten calculations moderately converged, but the scf cycles won't converge any
> further. I've also tried using a low mixing factor and turning the diis on in
> later scf cycles when there is moderate convergence, but that doesn't appear to
> be working either. Any help would be greatly appreciated.

Dear Dylan,
Have you tried to play with the Occupations smearq=<value> setting?
This introuduces an electronic temperature in the system that in many
cases will help scf convergence. In most cases this will results in a
non-ground state density, but this can often be solved by gradually
reducing the temperature using a previously scf converged calculation
at a higher temperature as a restart. In difficult cases you can start
with a high value for the smearq parameter, say 0.05 to 0.1 (default units
is Hartree). Then gradually "cool" the electrons by reducing smearq
in small (or large) steps (depending on how diffcult your system is
to treat). For a metallic cluster it might be hard to get to the
"real" ground state, but in my experience you can get very close,
i.e., that the binding energy do not change by more than 0.01-0.05 eV
for the lowest sets of smearq values (say 0.001-0.005).

[The annoying thing at the moment is that you'll have to do all the
gradual cooling by hand or by scripts - adf does not allow you to
give a list of values to the smearq parameter. However, if you do
find that it works maybe you could help "motivate" SCM to introduce
this change in the code by writing them an email (like I did) ;-) ]

Good luck!
-Roar.

--
Roar A. Olsen
(r.a.olsen "at" chem "dot" leidenuniv "dot" nl)
Leiden Institute of Chemistry
Gorlaeus Laboratory
Leiden University
P.O. Box 9502
2300 RA Leiden
The Netherlands
Tel: +31 71 527 5569
Fax: +31 71 527 4397
Web: http://rulgla.leidenuniv.nl