CCL: End of a frequency job

 Sent to CCL by: Toomas Tamm [tt-ccl2=yki.ttu.ee]
 On Mon, 2008-08-25 at 13:46 -0400, Close, David M. CLOSED^mail.etsu.edu
 wrote:
 > Now a more basic question.  If you are taking a course on doing
 > calculations, you are always told you must use the same basis sets in
 > frequency calculations as on the original optimizations.  But people
 > seem to be getting away from this are very large calculations.  If it
 > takes a week to do an optimization, it might take 3-4 weeks to do a
 > frequency calculation.  So people are now using smaller basis sets to do
 > faster frequency calcs.
 On several occasions, we have re-optimized the structure with a smaller
 basis set, checked that no significant changes in geometry have occurred
 (usually the gradients at initial optimization step, and the number of
 additional steps needed, are good diagnostics, too), and then performed
 a frequency calculation for the re-optimized geometry, using the smaller
 basis set. We have even used the zero-point energies from such
 calculations together with electronic energies from the large-basis
 calculations.
 I believe this is theoretically correct, just less accurate than
 calculating the frequencies with the larger basis set. Of course there
 remains the slight possiblity that the large-basis-set geometry was a
 transition state while the small-basis-set geometry is a minimum. We
 have considered the risk small enough given the savings in computation
 time.
 Any comments on this procedure?
 --
 Toomas Tamm                                 e-mail: tt-ccl2 (at) kky . ttu . ee
 Chair of Inorganic Chemistry                voice:  INT+372-620-2810
 Tallinn University of Technology            fax:    INT+372-620-2828
 Ehitajate tee 5, EE-19086 Tallinn, Estonia  http://www.kk.ttu.ee/toomas/