Firefly and PC GAMESS-related discussion club

Learn how to ask questions correctly

Alexey A. Popov
alxy.popov@gmail.com

there are different way how one can improve convergence, it is difficult to say without seeing your files.

You can try fshift=XXX in $SCF group, by default the value is 0.1, you can try to increase it (or decrease it).

You can also change the number of DIIS vectors by maxdii=XX command in $scf (default is 10, larger number can be sometimes better)

If molecule is symmetric, symmetrization often helps

$moorth symden=1 symvec=1 syms=1 $end
$guess purify=1 symden=1 $end

There are other ways you can play, check other threads on this forum

Finally, 70 iterations is not a large number. I mean, in difficult situations you may need more to converge - at least, at the first step. If your starting geometry is not very bad, then at subsequent optimization steps the number of scf iterations needed is usually much smaller.
-------------------------------------------
>Hi,
>Alexey, thanks for help. I did not know that it is impossible to run with $Vec group from run with another basis set.
>I cannot run some tasks with optimize geometry coordinates from run with smaller basis set (without any $Vec). FF writes in outs that "SCF IS UNCONVERGED" (with maxit=70, I cannot set more value, otherwise calculation on first step will be too long). Standard options like INTTYP=HONDO, ICUT=11, ITOL=30; FDIFF=F, DIIS=T, SOSCF=F, etc do not help. Problem…
>        Vyacheslav

>--------------------------------------------------------------------------
>On Tue Aug 24 '10 11:29pm, Alexey A. Popov wrote
>------------------------------------------------
>>>As for second problem I think you are right: I used the optimized >geometry from another run with little basis set. Probably, I should >take not optimized geometry from this run.

>>Probably there is some misunderstanding. Using geometry optimized with the smaller basis set for computations with extended basis sets is a standard reasonable way. The problem with your computation was that somehow the code has found a gradient from the previous in your input file (maybe you just copied it accidentally?) - at least, it seems so after checking your output.  For this reason, it did not compute the gradient again, it just took the gradient from the previous run - and since in the previous run you really optimized the structure, gradient was "zero".