# Gaussian: Transition state convergence question

I am running a TS computation in Gaussian 16. It performs several steps.

At the end of each there is a summary on whether Maximum Force, RMS Force, Maximum Displacement and RMS Displacement have converged. The thing is that in several of my computations the second-last step reports all 4 converged and in the last step always one or two are not longer converged.

Why is that? Does that mean I can take the second last step parameters and use them?

Are you doing an opt and freq calculation? In Gaussian, the opt keyword uses an approximate Hessian to determine the displacement at each step. The freq keyword will recalculate these displacements using the actual Hessian for that level of theory and sometimes you will find that they are no longer below the threshold. This can mean that your structure is not fully converged.
To ensure that the optimization is actually fully converged, you can include the keyword opt(RecalcFC=N) in your calculation to make it use the actual Hessian in the optimization every $$N$$ steps.