Firefly and PC GAMESS-related discussion club

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

Re^2: SCF convergency in UHF-DFT mode

Alex Nest

On Sat Nov 23 '13 9:23pm, Alex Granovsky wrote

>you can try:

>1)Increase fshift in $scf to 0.15 or 0.20

>2) Use larger maxdii in $scf e.g. maxdii=20 or 30

Thank you very much! This two options (in cooperate with finer grid $dft nrad=96 $end) boost convergency of my molecular system!

>3) Use mom-default value of diimod in $scf. Valid values are 0 to 5, the default is 0.

this option didn't help in my case.

>See the recently updated manual for up to date description of $SCF group.

>Kind regards,
>Alex Granovsky
>On Sat Nov 23 '13 6:36pm, Alex Nest wrote
>>Good morning!

>>We have optimized a heme with 2 coordinational amino acids (above and under the heme plane), it is a big chromophore molecule with a Fe atom in the center and containing more than 100 atoms. The heme in high-spin reduced state was used for calculations. We used LANL2-DZ external basis-ecp combination, d5 was turned on, maxit,icut,itol and inttype was set up to improove convergency. In the SCF group following parameters was set:
>>   fdiff=.f. diis=.t. soscf=.f. dirscf=.t. damp=.t.  shift=.t. nconv=6
>>All this settings allows us to procceed geometry optimization in delocolized coordinates as well. Every step takes about 40-70 iterations of SCF for convergency. But in some conformations convergency takes more than 200 steps and the question is how to improve with Firefly abilities.

>>In GAMESS-US (december 2011) convergency appears to be slightly better. May be because of step-by-step procedures: pure UHF -> UDFT in coarse grid -> UDFT in fine grid. I didn't find this ability in Firefly. May be firefly have some alternative techniques to improve the convergency in our case?

>>DLC coordinates for optimization works in Firefly better than in GAMESS-US so we tend to use Firefly-only for energy minimization if UDFT-convergency can be improoved.

[ Previous ] [ Next ] [ Index ]           Sun Nov 24 '13 1:56pm
[ Reply ] [ Edit ] [ Delete ]           This message read 687 times