Molecule explodes upon S1 geometry optimization

[TdeJ]

I am performing a geometry optimization of the first excited state with XMS-CASPT2(6,6) using 3 states. After the 5th step it yields
"ERROR: EXCEPTION RAISED: max size reached in AugHess".

Checking the geometry reveals that the bond distances have grown from 1-2 to over 10 angstrom in one step. According to the opt.log, the displacement was no more than other steps, and the gradient also looks fine. Could this be something that is done every X steps and now breaks? Any help would be greatly appreciated.

Best regards,
Tjeerd de Jong

opt.log
caspt2.json
casscf.log
opt_traj.txt

[FreemanTheMaverick]

Bagel is not good at geometry optimization. I wrote a shell script to let G16 optimize molecular structures by reading the energy and its nuclear derivatives calculated by Bagel.

The key procedure is like this:

1. Carry out a HF to obtain the initial orbitals.
2. Carry out a CASSCF with the keyword active to select the active space and put the active orbitals at the frontiers.
3. Check whether the frontiers orbitals are what you want to be active.
4. Copy the wavefunction file of the last step to a file named reference.archive. This will be the working wavefunction file of the following optimization job. In each job, Bagel will read reference.archive for inital guess and export the converged wavefunction of the current step into it.
5. Make a gjf file with the headers below and run.

%nprocshared=1 ! One core for G16 is enough.
%chk=c4h4s.bag.chk ! Save the hessian of the last step to chk
# opt=(ts,noeigen,calcfc,nomicro) freq nosym ! "nomicro" and "nosym" are necessary.
external='bash ~/gau_bag.sh 9 cc-pvdz cc-pvdz-jkfit 2 0 4 12' ! Call the gau_bag.sh in the HOME directoty; 9 cores for Bagel; cc-pvdz and cc-pvdz-jkfit as the basis set and the auxiliary basis set; 2 states in total; the geometry of the first state (denoted as 0) to optimize; 4 active orbitals in the frontier; 12 fully occupied inactive orbitals.

Title Card Required

0 1 ! Charge and spin

With this script you can also carry out other nuclear-derivative-related jobs such as IRC and BOMD.

[TdeJ]

Thank you for your reply FreemanTheMaverick. Would you also be able to put the shell script up for download here? I am unable to reach the referenced website.

[FreemanTheMaverick]

Thank you for your reply FreemanTheMaverick. Would you also be able to put the shell script up for download here? I am unable to reach the referenced website.

This is it. Just remove the txt suffix.
gau_bag.sh.txt