Firefly and PC GAMESS-related discussion club

Learn how to ask questions correctly  
We are NATO-free zone

Re^6: SCF Convergence Problem in modelling Ag cluster

Alex Granovsky

Dear Davide,

For MCSCF, I'd suggest to try using either semi-converged
UHF orbitals, orbitals of positively charged system or system
of different multiplicity. This may need some orbital reordering
directives. With Firefly, SOSCF usually converges well even for
problematic systems. If it does not, try to perform several FOCAS
steps before SOSCF. The simplest way is to use e.g.:

 $mcscf numfo=5 $end

This will force five FOCAS steps before launching SOSCF.

Best wishes,

On Tue Nov 8 '11 11:00am, Davide Vanossi wrote
>Dear Alex,
>          thank you very much for your kind reply.
>Actually I know very well the need to perform an MCSCF description for Co isolated atom. But how can I obtain, for my cluster (Ag37Co), an acceptable (converged) initial guess for using it in the next MCSCF calculation? Using fshift=0.15-0.2 didn’t help a lot.
>Thank you again.
>Have a nice day

>   Davide

>On Sun Nov 6 '11 11:41pm, Alex Granovsky wrote
>>Dear Davide,

>>actually, Firefly' DFT code has problems converging just an isolated
>>Co atom with e.g., mult=4. The reason is very simple. As Sanya has already
>>mentioned, this state actually requires MCSCF for correct description.
>>Indeed, there are both empty and occupied degenerated d orbitals so SCF oscillates
>>between solutions having different occupation numbers for different
>>orbitals. Increasing $scf fshift to say 0.15 or so helps. Nevertheless,
>>one should generally realize that DFT can be inadequate and that
>>there exist multiple (quasi)degenerated solutions of KS equations for this system.
>>As to system with Co near Ag cluster, my guess is that as Co is far
>>away the cluster, the convergence problems are caused by the same
>>reason. Try to move Co atom closer to the surface.

>>Best regards,
>>On Fri Nov 4 '11 2:04pm, Davide Vanossi wrote
>>>Dear Amir,
>>>         I’m absolutely agree with you and, in fact,  I tried the lanl2dz from the beginning; unfortunately this introduced, on my system, a quasi linear dependence on the basis set which is not a good companion “to cure” the SCF convergence problem. For the moment I have “solved” the problem using another quantum chemistry package (NWChem) which permits to obtain a feasible SCF convergence even for the singlet multiplicity case.
>>>Thank you very much for your interest.
>>>Have a nice day.

>>>    Davide
>>>On Fri Oct 28 '11 11:56am, Amir Nasser Shamkhali wrote
>>>>Dear Davide
>>>>I think that relativistic effects can be important for Ag atom. So, it's better to perform calculations with an ECP basis set which has been included relativistic effects (spin-orbit coupling) such as LANL2DZ and etc. Also these basis sets reduces the cpu time and memory considerably. You can download them form this site:
>>>>Please note that spin-orbit coupling reduces spin multiplicity and do not use high spin cases. The 2S+1=3 or 5 or 7 is sufficient.

>>>>Best wishes

[ Previous ] [ Next ] [ Index ]           Thu Nov 10 '11 6:06pm
[ Reply ] [ Edit ] [ Delete ]           This message read 749 times