## Firefly and PC GAMESS-related discussion club

Learn how to ask questions correctly

Re^2: SCF Convergence Problem in modelling Ag cluster

Davide Vanossi
vanossi.davide@unimore.it

Dear Sanya, thank you very much for your interest. Actually my system is made-up of 37 Ag atoms and 1 Co (not Ni) atom. Anyway I try to start from a higher-spin state (even if one of my colleague was able to converge the same system with multiplicity=1 using Gaussian package) setting guess=hcore as you suggested.
Thank you again.
Have a nice day,

Davide

On Sat Oct 8 '11 3:53am, sanya wrote
------------------------------------
>Dear Davide,

>Metal clusters are really a challenge both for SCF-like methods and for the researcher.

>Are you sure that the multiplicity is 1? Low-spin states are rare in metal clusters. Probably, you should start with the highest possible spin. Ni atom alone has at least 3 unpaired electrons, and each Ag atom has 1 unpaired electron. One cannot even guess how all these spins are paired in the cluster. Probably, 40 unpaired electrons (37 from Ag and 3 from Ni, mult=41) will be a good start. After converging the highest-spin state, you may use these orbitals as a start for lower-spin states. Alternatively, try mult=5 (3 unpaired electrons from Ni and 1 unpaired electron from Ag cluster, supposing other spins of Ag cluster are paired).
>Well, MULT=41 is a little bit too strong, because silver is not ferromagnetic, and such a state will be rather unfavorable... but MULT=5 is OK. Another possibility is MULT=3 (1 electron from Ni and the only unpaired electron of Ag are now paired, and two last electrons from Ni are still unpaired). Try both MULT=5 and MULT=3.

>In addition, guess=huckel is not a good start for metal clusters, try guess=hcore instead.

>Finally, I see that you're trying to reproduce the surface of Ag by such a cluster. This is not good at all. Periodic models are more adequate for these purposes.

Mon Oct 10 '11 11:15am