Unable to recalculate Hessian in IRC

[shoubhikraj]

Description of the run environment:

• operating system Windows 10
• pre-compile binary
• communication library MS-MPI

• $CONTRL SCFTYP=RHF RUNTYP=IRC MULT=1 NZVAR=51 DFTTYP=B3LYP $END $SYSTEM MWORDS=5 PARALL=.TRUE. $END $BASIS GBASIS=N31 NGAUSS=6 NDFUNC=1 DIFFSP=.t. $END $IRC PACE=GS2 OPTTOL=0.0004 MXOPT=300 STRIDE=0.10 NPOINT=300 TSENGY=.t. SADDLE=.t. FORWRD=.f. $END $PCM SOLVNT=WATER SMD=.t. $END $ZMAT NONVDW(1)=3,7, 5,10 DLC=.t. AUTO=.t. $END

I am trying to run an IRC from a transition state. The IRC run takes $HESS group for the initial hessian input, but there seems to be no way to request for recalculation of the hessian for either the predictor or the corrector step (for the GS2 in my case). Putting IHREP in $STATPT does not do anything. (Also, there is no way of calculating the hessian at the beginning of the IRC instead of inputting.)

This is a problem in cases where convergence is slow, and recalculating hessian periodically along the IRC is necessary. Currently, the only to do this would be to recalculate hessian where the IRC stops and then restart it.

Since IRC run is basically a constrained optimization, I am guessing that it shares the same algorithms as RUNTYP=OPTIMIZE or SADPOINT, so running IHREP or HESS=CALC in IRC should be possible.

[saromleang]

You are correct, that capability is not available in GAMESS for RUNTYP=IRC.

Suggestions (each are independent of each other):

1. Switch convergers with $SCF DIIS=.T. SOSCF=.F. $END
2. Reduce STRIDE value to 0.05.
3. When geometry optimizing your structure try find a saddle point try a lower OPTTOL value. Like $SADPOINT OPTTOL=1E-5 $END or $SADPOINT OPTTOL=1E-5 $END. You could also apply #1 to help with convergence of the saddle point search.
4. This is more for performance, consider using DIRSCF=.T. in $SCF DIRSCF=.T. $END