Firefly and PC GAMESS-related discussion club

 
Learn how to ask questions correctly  
 
 

 

Solntsev Pasha
solntsev@univ.kiev.ua

I would suggest you increase maxit in $contrl up to 300-400 and

$scf diis=.t. soscf=.f. $end

Try also use internal coordinates:

$CONTRL NZVAR=1 $END
$ZMAT DLC=.T. AUTO=.T. $END

If you can't reach convergence, try to remove some number of electron by making an cation and run runtyp=energy. Then use optimized orbitals as an initial orbitals for your final calculation. You should be patient. You can remove 2 electrons or sometimes even 10. Then you can step-by-step pt electrons back and every time reuse orbitals from previous run.  

You can also try to use pure HF to converge you orbitals at first step and then reuse orbitals for DFT calculations.

Good luck.

Pavel.


On Mon Nov 12 '12 3:21pm, Diver wrote
-------------------------------------
>Dear Thom,

>   I have used SOSCF=.true. FDIFF=.true.

>GAMESS manual suggests to reduce ICUT by one for basis with diffuse functions.

>Your suggestions made energy jump +2000 Hartree.
>with SOSCF the oscilations were "just" +/- 0.1 Hartree

>I have attached a output file with the best convergence so far.
>Vadim
>
>
>On Mon Nov 12 '12 12:06pm, Thomas wrote
>---------------------------------------
>>Dear Vadim,

>>There currently is no option in Firefly that allows one to project orbitals converged with a small basis onto a larger basis. If you're experiencing convergence problems it would be best to post the output of your calculation so one of us can have a look at it and see what kind of convergence problem you're experiencing.

>>As you're using diffuse functions, I for now will assume that the problem is a partial linear dependence in your basis set (which is a common problem with diffuse basis functions). Most likely, the SCF is close to convergence but is slightly oscillating rather than reaching the convergence criteria. If this is the case, I advise the following:

>>- Increase the accuracy of the calculation by specifying $CONTRL INTTYP=HONDO ICUT=11 $END
>>- Turn of FDIFF by specifying $SCF FDIFF=.F. $END
>>- Increase the amount of SCF iterations using the MAXIT keyword in $CONTRL. For example, use $CONTRL MAXIT=100 $END
>>- If you're not yet using the DIIS converger, enable its use by $SCF DIIS=.T. SOSCF=.F. $END

>>If this doesn't help, you probably need to remove one or more of the diffuse functions that cause problems. Look for the message saying "THE OVERLAP MATRIX HAS   x EIGENVALUES BELOW 1.0E-05", followed by a list of shells. The shells that are causing the problem are at the bottom of the list.

>>But, again, I give you this advise because I suspect you're encountering partial linear dependence. If this is not the case, or if my advise didn't help you, please post the output on this forum so we can be of more help.
>>
>>
>>Kind regards,
>>Thom
>>
>>
>>
>>On Sat Nov 10 '12 3:40pm, Diver wrote
>>-------------------------------------
>>>Dear Alex,

>>>My job converges at basis with SVP basis set,
>>>but it is failing to converge with SVPD basis set
>>>(SVPD = the same SVP part + diffuse basis).
>>>How could I use MO generated in SVP job for SVPD job?

>>>or
>>>What could you suggest for convergence in case of diffuse basis.

>>>Thank you in advance,
>>>Vadim
>>>

>>>