Dear Fernando,

Often the force constant matrix used in a geometry or transition state optimization is approximate - for example, the optimization starts from some guess FC Matrix that has the proper curvature and updates that guess iteratively using something like BFGS or Powell (for geometry opt, transition state opt, respectively) using only the gradient information.  To actually calculate the Frequencies, you will of course calculate the "true" force constant matrix at the given geometry, which will be different than the approximate one calculated iteratively during the optimization.

If the true force constant matrix does not appear to satisfy the convergence criteria for your stationary point of interest, a few steps could help:  1)  Make your convergence criteria more stringent… iterative updates to the approximate FC Matrix generally get better as you accumulate more gradient information over more and more iterations.

2)  It is almost certainly possible in whatever program you are using to read in the FC Matrix generated from your frequency calculation.  You can use this instead of the guess FC matrix along with the improved convergence criteria for the optimization and it might give you better results.  

Hope this helps!

Best,

Jay

Dear, Jonathan

Did the TS I found is correct if only one convergence criterion is fulfilled in the calculation of frequencies?, Taking intoaccount that all criteria optimization of TS were met and I had the imaginary frequency corresponding to the normal mode of vibration that I was looking for.

Dear Fernando, 

Well, that is a tricky question because the criteria you are referring to from the frequency calculation is also not sufficient to guarantee that the TS is correct.  However, the criteria from the frequency calculation is more reliable than from the TS optimization itself, due to the fact that the TS optimization is not using the full FC matrix.

Best,

Jay

Dear Jonathan

What keywords or methodology do you recommend that I use in Gaussian 09 to have a better convergence in the optimization and frequencies, to be sure that the TS is correct?

if I have a typing error please correct me.

Thanks