Dear Jesse,

I believe the attached link describes exactly what you are looking for. If the displacements indicated after the frequency calculation are not outrageously large, you may be fine without any further action. In your case, my guess is that it is really a non-issue, because the forces are very well-converged. But if you don't know what all of this really tells you and if you can ignore it or not, try to follow the instructions in the link for re-optimizing the structure using the exact Hessian matrix at the starting point ("opt=readfc") and see what happens.

(Hopefully you retained the checkpoint file from the frequency calculation. If you didn't, then this is the right time to start making this standard practice from now on. You will save a lot of time and resources in cases like these. Every optimization step that calculates an exact Hessian matrix will take as long as the frequency calculation, because they both have to do the same thing. The disk space for preemptively storing all your chk files is much, much cheaper in comparison.)

http://dev.gaussian.com/faq3/

Dear Jesse Quinn,

If you like to calculate the frequency, the final scf energy(freq job) should match with the scf energy of the optimized structure. In order to check this, you search for the scf Done in both the cases opt and freq. If you dont get the same energy in both opt and freq calculations then care must be taken and you need to redo the calculation.

you please use int=acc2e=11 or int-acc2e=12 along with the int=grid=ultrafine. acc2e=11 or 12 are the two subroutines which are included in the g09 program for more accuracy.

You please redo the calculation using the last optimized structure so that it will easily be converged. 

And one more thing I dont have an idea about CPHF more. so better you remove in case of optimization and check.

All the best.

Best regards,

Ranga Santosh

Dear Jesse,

I believe the attached link describes exactly what you are looking for. If the displacements indicated after the frequency calculation are not outrageously large, you may be fine without any further action. In your case, my guess is that it is really a non-issue, because the forces are very well-converged. But if you don't know what all of this really tells you and if you can ignore it or not, try to follow the instructions in the link for re-optimizing the structure using the exact Hessian matrix at the starting point ("opt=readfc") and see what happens.

(Hopefully you retained the checkpoint file from the frequency calculation. If you didn't, then this is the right time to start making this standard practice from now on. You will save a lot of time and resources in cases like these. Every optimization step that calculates an exact Hessian matrix will take as long as the frequency calculation, because they both have to do the same thing. The disk space for preemptively storing all your chk files is much, much cheaper in comparison.)

http://dev.gaussian.com/faq3/

Dear Jesse,

I think the structure you have is likely not reliable (not a true stationary point) that lead to non-convergence in frequency calculations. It may belong to transition structure or a saddle point. 

I, therefore, suggest you to please check your geometry optimization results.

Hallo, Jesse!

 Using numerical methods of optimizations you can obtain stable structures within some accuracy only (i.e. not exactly). If you repeat your optimization from another start, you should obtain only very similar structure but not exactly the same (maybe except very simple molecules). Another reason of your problem is that the number of valid digits in your coordinates is restricted as well (i.e. there are small differences in the coordinates between Opt and Freq jobs). In general, atomic coordinates are presented with 0.001 Angstrom accuracy only (for standard optimization parameters).

 Nevertheless, in the case of poor SCF convergence you need not obtain the same electronic states in Opt and Freq jobs but this may be checked by comparing the SCF energies (and usually any negative frequency in Freq job is observed).

According to my experience your result is OK and you do not need any further improvement.

   Best wishes

                    mb

Thank you all for the quick replies. From what I understood using GAMESS-US in the past is that if no negative frequency is found I should be ok for the optimized structure as Martin suggests.

I tried Matthias' suggestion; however, I ran into the same problem.

The punch line was:

# b3lyp/6-31g(d) OPT=(ReadFC,Tight) Freq Geom=AllCheck Guess=Read Int=(Grid=Ultrafine)

I will also try Ranga's suggestion; however, from what I understand he had issues with the " int=acc2e=11" or " int=acc2e=12" with calculations for charged compounds? At some point I would like to calculate EA and IP. In addition, isn't the default in Gaussian acc2e=12 or is it 10 for G09?