[molpro-user] CASSCF fails to converge

[Hans-Joachim Werner]

Dear Terry,
I think this problem is due to the fact that there are 3 nearly  
degenerate
states (the 3 states arising from C(3P)). There might then be
root flipping problems, or the program gets into other convergence  
problems
in the CI step. Even though your calculation converged with the current
development version 2008.3, the energy at the initial geometry was  
slightly higher than in the calculation with 2008.1 (which did not  
converge), and thus it appears
that another state was found (this may be due to slighly different  
starting
orbitals). You should use state-averaged casscf for this case!
Best regards
Joachim
Am 02.07.2009 um 18:39 schrieb Terry Frankcombe:

> Can someone point me to how to get a FV CASSCF calculation on C +  
> CH2 to
> converge tightly?  If I only want the energy at a particular  
> geometry, the
> CASSCF converges in something like 6 iterations.  If I want to  
> optimise
> anything, though, the incipient optg is detected and the convergence
> thresholds tightened (despite any global or local options).   
> Depending on
> the geometry, the gradients and/or the steplength level off at a value
> above the thresholds and stay there.  Increasing the number of  
> iterations
> doesn't help.
>
> How can I either a) get the CASSCF to converge, or b) loosen the
> convergence criteria (I really only want a loose optimisation)?
>
> (2008.1-16)
>
> Input below, output attached.
>
>
> basis=AVDZ
> geometry={Angstrom
> 	NoSym
>        C1
>        H1 C1 rCH1
>        H2 C1 rCH2 H1 a1
>        C2 C1 R H1 a2 H2 a3}
>
> rCH1=1.07
> rCH2=1.10
> a1=132
> a2=140
> a3=150
> R=9
>
> {HF; WF 14 1 0}
> CASSCF
> optg
> <MEP.DZ.1.out.gz>_______________________________________________
> Molpro-user mailing list
> Molpro-user at molpro.net
> http://www.molpro.net/mailman/listinfo/molpro-user