PC GAMESS/Firefly-related discussion club

Learn how to ask questions correctly

Alex Granovsky
gran@classic.chem.msu.su

I completely agree with Sanya.

However, I'd like to address your initial question on how to enforce SOSCF to proceed...

There exist (undocumented as of yet) $soscf group.
One of its parameters allows one to avoid abnormal termination if the rotation is huge.

Namely:

 $soscf crit=value $end

The default critical value to abort soscf is rather small (and depends on the Firefly's version).

The following value:

 $soscf crit=1d4 $end

allows your job to proceed (and I guess it will not terminate even fantastically ill-behaved SOSCF whatsoever).

Attached please find input file as well as converged output.
There are three quasi-degenerate orbitals to be partially
filled... and this is the origin of the slow convergence.

Hope this helps.

Alex



On Mon Nov 9 '09 2:42am, sanya wrote
------------------------------------
>Sorry for the delay, your system was rather difficult indeed.

>First, my experience shows that it is better to turn SOSCF off on DFT runs.

>Second, DFT is not good for open-shell systems (Ni²⁺ is d⁸, a difficult open shell case), so poor convergence is not surprising.

>Third, are you sure you are interested in a system with charge +2, i.e., Ni²⁺...urea? For a neutral system SCF converges (with SOSCF=.f.) in less than 50 iterations.

>>I have used $VEC from the same input but with MULT=1

>Did your SCF converge for MULT=1 ICHARGE=2? I tried MULT=3 ICHARGE=0 and it converged. Unfortunately, the orbitals from these runs were poor guess for MULT=3 ICHARGE=2.

>So, if you are sure you really need the system with Ni²⁺, you should use state-averaged MCSCF instead of DFT. It's a skilled work.

This message contains the 224 kb attachment [ urea_soscf.rar ] Ni-Urea complex... soscf