PHASE2 restart normally requires hessian from a previous run.
In addition, I'd suggest using small trmax and dxmax (these are
variables of $statpt group), normally ten times smaller than
their default values. Tighter cutoffs throughout are necessary.
Finally, are you using DLCs or cartesians?
On Wed Sep 7 '16 6:09pm, GrEv wrote
>I am attempting to find a MECI between S0 and S1 using SA-CASSCF(12,11). An attempt to localize MECI starting directly with SHIFT=0.0001 (default) failed. I decided then to search in a step-wise fashion, starting with SHIFT=0.01 and then decreasing it after every successful search. However, I couldn't optimize MECI for SHIFT less than 0.005. I wonder what the problem could be? I am using the following options for the search:
>$CONIC SHIFT=0.001 SHIFT0=0.001 STAGE2=.t. LSTART=0.8492530370 $END
>(LSTART is from the previous serach with SHIFT=0.005)
>$STATPT OPTTOL=0.00010 NSTEP=500 method=conic $END
>$TRACK TOL=1.4 FREEZE=.T. STICKY=.F. UPDATE=.T. RESET=.F. DELCIV=.F. $END
>$MCAVER JSTATE=2 CONIC=2 $END