Firefly and PC GAMESS-related discussion club

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

Re^9: Problems with SA-MCSCF Geometry optimization

Alex Granovsky


sorry I cannot reproduce your output file with the input file
you have sent. I'm attaching an archive with two files, the first
one is copy-paste from your sample output, the second one contains
the results I have obtained on my computer. Note the difference in
the final MCSCF energies.

Could you please double-check if you attached the correct input?

Kind regards,
Alex Granovsky

On Tue May 28 '13 8:17pm, lello wrote

>I am attaching my full input.

>If I remember correctly, the starting orbitals were obtained from
>a previous MCSCF  calculation without solvent, and the guess of
>that MCSCF was taken from a ROHF calculation.
>On Tue May 28 '13 5:39pm, Alex Granovsky wrote

>>sorry I cannot reproduce your calculations. The self-consistent
>>PCM reaction field at the first geometry dramatically differs
>>between what I see in your output and the results of my calculations.
>>Evidently in this case PCM is very sensitive to starting orbitals.
>>This sensitivity could explain the problems you faced.

>>Could you please attach the exact input file you used and provide
>>a brief description on how orbitals was generated?

>>By the way, with the orbitals I use geometry optimization behaves
>>quite logical. It converges to CI point and then oscillates in its
>>vicinity as it should if there is no minimum on the S1 PES.

>>Kind regards,
>>Alex Granovsky
>>On Sat May 25 '13 10:14pm, lello wrote
>>>Just to close my feedback on this point. I used the same input
>>>posted on the site but without PCM and the geometry optimization
>>>at SA-MCSCF level converged without problems in a about 8 cycles.

>>>On Mon May 20 '13 3:14pm, Alex Granovsky wrote

>>>>the problem was not specific to state-averaged calculations,
>>>>it was pertained to state-specific MCSCF + PCM gradients as well.

>>>>Kind regards,
>>>>Alex Granovsky
>>>>On Mon May 20 '13 2:59pm, lello wrote
>>>>>Ok thanks, I was running other calculations of the same type
>>>>>and I guess I must resubmit the jobs. Just to be sure,
>>>>>was it a problem only for state averaged or also for state
>>>>>specific calculations?


>>>>>On Mon May 20 '13 2:25pm, Alex Granovsky wrote

>>>>>>this problem (a missed factor of 0.5 in gradient contribution)
>>>>>>was specific to combination of PCM + MCSCF + gradients only.

>>>>>>Kind regards,
>>>>>>Alex Granovsky

>>>>>>On Mon May 20 '13 2:13pm, lello wrote
>>>>>>>Thanks a lot.
>>>>>>>I'll try the new release right away.
>>>>>>>Could you please tell me if all SA-MCSCF calculations
>>>>>>>were affected by this problem, or if it was specific to this case?

>>>>>>>On Mon May 20 '13 1:40pm, Alex Granovsky wrote
>>>>>>>>Dear Lello,

>>>>>>>>we have fixed the problem. Attached please find a sample input and
>>>>>>>>output files. You need to update to the latest Firefly v. 8.0.0 RC
>>>>>>>>build # 7697 binaries (I have updated them today).

>>>>>>>>Kind regards,
>>>>>>>>Alex Granovsky

>>>>>>>>On Thu May 16 '13 8:56am, lello wrote
>>>>>>>>>I have attached the input.
>>>>>>>>>On Wed May 15 '13 11:31pm, Alex Granovsky wrote

>>>>>>>>>>could you please send me your input file?

>>>>>>>>>>Kind regards,
>>>>>>>>>>Alex Granovsky
>>>>>>>>>>On Wed May 15 '13 10:55pm, lello wrote
>>>>>>>>>>>Dear all,

>>>>>>>>>>>I am trying to perform a SA-MCSCF geometry optimization
>>>>>>>>>>>of the ground state of a system, and I have the following behaviour
>>>>>>>>>>>of the energy of the target state:

>>>>>>>>>>> NSERCH=  0     ENERGY= -915.9734212   RMS GRAD = 0.0095239
>>>>>>>>>>>  NSERCH=  1     ENERGY= -915.9501045   RMS GRAD = 0.0196205
>>>>>>>>>>>  NSERCH=  2     ENERGY= -915.9576756   RMS GRAD = 0.0058206
>>>>>>>>>>>  NSERCH=  3     ENERGY= -915.9634734   RMS GRAD = 0.0091716
>>>>>>>>>>>  NSERCH=  4     ENERGY= -915.9626773   RMS GRAD = 0.0056877
>>>>>>>>>>>  NSERCH=  5     ENERGY= -915.9597202   RMS GRAD = 0.0036023
>>>>>>>>>>>  NSERCH=  6     ENERGY= -915.9586388   RMS GRAD = 0.0022706
>>>>>>>>>>>  NSERCH=  7     ENERGY= -915.9581777   RMS GRAD = 0.0020863
>>>>>>>>>>>  NSERCH=  8     ENERGY= -915.9575078   RMS GRAD = 0.0033306
>>>>>>>>>>>  NSERCH=  9     ENERGY= -915.9566032   RMS GRAD = 0.0036036
>>>>>>>>>>>  NSERCH= 10     ENERGY= -915.9559209   RMS GRAD = 0.0035010
>>>>>>>>>>>As you can see the first point is quite low in energy but after that
>>>>>>>>>>>the energy increase a lot. I have let the optimization run for more than 100 cycles but the energy is always high and never reaches
>>>>>>>>>>>that of the first point.
>>>>>>>>>>>I have observed this behaviour in a several
>>>>>>>>>>>cases. Is it just a problem of coordinate choice for the optimization
>>>>>>>>>>>or am I doing something wrong?
>>>>>>>>>>>Below is my input:

>>>>>>>>>>>         D5=.T. FSTINT=.T. GENCON=.T. ICHARG=1 MULT=2 INTTYP=HONDO  
>>>>>>>>>>>         ICUT=11   ITOL=20 $END
>>>>>>>>>>> $SCF DIRSCF=.TRUE. $END
>>>>>>>>>>> $SYSTEM TIMLIM=1000000 MEMORY=100000000 KDIAG=0 NOJAC=100 $END
>>>>>>>>>>> $P2P P2P=.T. DLB=.T. $END
>>>>>>>>>>> $BASIS EXTFIL=.T. GBASIS=cc-pVDZ $END
>>>>>>>>>>> $MCSCF NTRACK=2 ISTATE=1 $END
>>>>>>>>>>> $MCSCF NPFLG(9)=1 $END
>>>>>>>>>>> $DET NCORE=66 NACT=4 NELS=3 NSTATE=4 ITERMX=400
>>>>>>>>>>>     WSTATE(1)=1.0,1.0 CVGTOL=1.0E-7 $END
>>>>>>>>>>>Thanks in advance,

This message contains the 4 kb attachment
[ firstcas.rar ] Results of the first CASSCF procedure

[ Previous ] [ Next ] [ Index ]           Wed May 29 '13 1:53am
[ Reply ] [ Edit ] [ Delete ]           This message read 883 times