Firefly and PC GAMESS-related discussion club



Learn how to ask questions correctly


Re: SCF convergence

Alex Granovsky
gran@classic.chem.msu.su


Dear Masoud,

a few comments in the hope they are useful.

1. With Firefly, you can set MAXIT to be as large as you like,
as there is no upper limit on it.

2. It seems your job was about to converge, so you could try
to use vectors from the punch file to restart it with guess=moread

3. Try to change $scf fshift - this could help as well. The threads
of possible interest are:

http://classic.chem.msu.su/cgi-bin/ceilidh.exe/gran/gamess/forum/?C3c317208dSX4-7576-898-00.htm

http://classic.chem.msu.su/cgi-bin/ceilidh.exe/gran/gamess/forum/?C3d2b8e708S8y-7292-171+00.htm

http://classic.chem.msu.su/cgi-bin/ceilidh.exe/gran/gamess/forum/?C35413cca25e3-7543-1398+00.htm

Regards,
Alex Granovsky



On Sun Jan 23 '11 5:23pm, Masoud Nahali wrote
---------------------------------------------
>Dear Firefly Users

>My study system contains 174 atoms (C, H, and Ni) and as proposed by Sanya in the discussion with David G. , I have used 6-31g*, 6-31+g*, and Lanl2dz for C, H, and Ni respectively to overcome the possible linear dependence problem. After about 200 steps, the system was not converged. It seems that I must increase the MAXIT but it takes a long time to be converged. How can I increase the efficiency of the SCF procedure to get the convergence in a shorter time ?

>These are the variables of my input file :

>$CONTRL SCFTYP=RHF RUNTYP=OPTIMIZE DFTTYP=B3LYP ECP=read MAXIT=200 MULT=1
> ICHARG=0 COORD=unique d5=.t. NZVAR=516 $END
> $SYSTEM TIMLIM=6897500000 MWORDS=200 $END
> $SCF DIRSCF=.TRUE. NCONV=4.5 $END
> $STATPT OPTTOL=0.0001 NSTEP=150 $END
> $zmat dlc=.true. auto=.true. $end
>! to activate P2P inteface and DLB:
> $P2P P2P=.T. DLB=.T. $END
>! to speed up Huckel guess:
> $GUESS GUESS=HUCKEL KDIAG=0 $END
> $smp call64=.t. mkl64=8 load=0 ssg32=0 msu32=0 mkllvl=3 $end
> $BASIS gbasis=LANL2DZ extfil=.t. $END

>and the SCF steps was attached.


[ Previous ] [ Next ] [ Index ]           Mon Jan 31 '11 10:07pm
[ Reply ] [ Edit ] [ Delete ]           This message read 1063 times