In your "free postoptimization", have you tried playing with the convergence criteria, both for the SCF and the geometry? Eventually, that may be helpful.

A transition state which has been calculated with fewer constraints will surely find more acceptance than a constrained one.

Jürgen Weippert

Thanks for your response

I was wondering how I should play with the geometry and SCF. Is there a rule of thumb I can use for geometry? I tried running scf=qc, is this what u meant by SCF manipulation?

There are convergence criteria with respect to the energy deviation from optimization cycle-to-cycle. That should be a number. In Turbomole, which is my "home" QC system, I could tell you where you can do that, in Gaussian I am not that proficient and my last usage dates back to 2012.

Looking at the manual, I would speculate the CONVER option might be useful for you.

https://gaussian.com/scf/

I've seen you asked a second question:

https://www.researchgate.net/post/Why_my_transiton_state_does_not_converge

Is this the same calculation or a different problem? If it is, it would indeed speak in favor of convergence criteria, e.g. try getting it to converge with loose geometry convergence criteria and then narrow it down step by step. Eventually, the geometry and SCF criteria must be tightened simultaneously.

Jürgen Weippert

Yes, it is the same problem. I tried lowering the SCF criteria to 7 (the default is 8)

# opt=(calcfc,ts,noeigen) freq mpw1pw91/6-31+g(d,p) SCF=(XQC,Conver=7)

However, the job terminated promptly with the error

ERROR!!!!

SCF has not converged. Gradients and post-SCF results would be GARBAGE!!

SCF Error SCF Error SCF Error SCF Error SCF Error SCF Error SCF Error SCF Error

When locating the transition state (TS), freezing of any parameter of molecular geometry, which changes with moving along the reaction coordinate, does not permit to reach the TS. This is obvious: a sufficiently accurate numerical value of such a parameter is a priori unknown for a concrete molecular system. Especially, for protonation/deprotonation processes.

Moreover, even energy minimization without geometry constraints may afford the molecular structure with one imaginary frequence, which, in fact, is not a TS: it corresponds to an inflection point in the PES. The use of the GEDIIS algorithm (Gaussian) leads to true saddle points.

Correction (see above)

The use of the GEDIIS algorithm (Gaussian) leads to true minimum energy points.