Firefly and PC GAMESS-related discussion club

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

Re^2: SCF convergence problems for CdS nanocluster in DFT-BLYP framework


Dear Andrei

Thank you for your suggestions and sorry for delay.

I’ve already tried to use the Hartree-Fock scheme (I used fshift,SOSCF,DIIS,HONDO,...) to prepare a better guess in order to run a GGA calculation,
but unfortunately neither HF converges.
Now I’m trying to change the basis set,using a larger basis set to improve the description.

Kind regards, emiliano

On Fri Mar 23 '12 1:09am, Andrei V Scherbinin wrote
>Dear Emiliano,

>I would suppose that your system is far from trivial, and probably
>your calculation suffers from narrow-bandgap problems due to rather
>large size of the cluster. In addition, plain GGA DFT (BLYP in your
>case), tends to underestimate bandgaps in materials.

>I can not guarantee that the following advices help, but try the

>1) try to converge pure HF first (perhaps to a large treshold), then
>start DFT SCF from converged HF orbitals. Usually, RHF performs
>better because it tends to overestimate bandgaps as opposed to DFT.

>2) Try diferent SCF convergence options (level shift, SOSCF vs DIIS, etc)
>during both HF/DFT stages if problems still appear.

>2) If RHF converges but pure DFT still fails to converge starting from
>the RHF MOs on input, may be it is reasonable to perform a sequence of
>hybrid DFT calculations with gradually decreasing HF exchange fraction
>(say, HFLYP, then BHHLYP, B3LYP, BLYP). As a rule of thumb: the larger
>the HF fraction, the better the convergence usually is - see above.

>3) If all that doesn't work, re-consider your system again. May be,
>the closed shell description fails, or some surface-localized states
>emerge, or the minimal basis set is a poor choice (d-AOs are usually
>rather important for 3rd row elements), etc...
>Best wishes, Andrei
>On Mon Mar 19 '12 4:10pm, eburresi wrote
>>Dear firefly users

>>I'm trying to perform DFT-BLYP calculation on CdS nanocluster with 48 Cd atoms and 48 S atoms, using 3-21g and sto-3g basis set,
>>but I'm finding some convergence problems during SCF optimization. I tried to use combo,
>>building an intial guess from  optimized molecular orbitals of sulphur and cadmium fragments separately but in my case didn’t help a lot.
>>The atom coordinates are from CPMD ab initio molecular dinamics simulations.
>>I hope that someone can help me.

>> $SYSTEM TIMLIM=600 MEMORY=70000000 $END
>> $SCF    DIRSCF=.True. DIIS=.T. FDIFF=.F.
>>          SHIFT=.T. SOSCF=.F.  $END
>> $DATA
>>CdS Cluster 96 Temperature 340K
>>Thank you in advance
>>  Emiliano

[ Previous ] [ Next ] [ Index ]           Tue Mar 27 '12 3:33pm
[ Reply ] [ Edit ] [ Delete ]           This message read 1080 times