5
$\begingroup$

I would like to form the (numerical) gradient and the Hessian of the total electronic energy with respect only to a few internal coordinates (namely, the six intermolecular coordinates of a weakly bound dimer). In those cases I saw, when second derivatives were calculated, Gaussian always determined the full Hessian. Do you know an option capable of controlling the set of active variables used to calculate the gradient and the Hessian of reduced size?

$\endgroup$
8
  • 1
    $\begingroup$ +1 Gaussian (and most other softwares) determine the analytic hessian for the full wavefunction. If you want hessian for particular coordinates, it can only be done numerically I believe (and it won't be as accurate). If this is what you want, this can be done in Gaussian (gaussian.com/opt look at the "Options" tab). Are you open to using other QM softwares or only Gaussian? $\endgroup$
    – S R Maiti
    Aug 11 at 8:31
  • $\begingroup$ @SRMaiti Thanks for your detailed comments. It is not a problem for me to have numerical partial derivatives: they are accurate enough to make a reduced dimensional optimization for (higher-order) saddle points. I also saw the option "Diff" and tried to use it, but I did not recognize the derivative value in the .log file or the .fchk file. Furthermore, I do not see a similar options for the calculation of gradient components. Another problem is that the option "Diff" determines all the Hessian entries contained in the row/column of a particular internal coordinate. $\endgroup$
    – TobiR
    Aug 11 at 12:35
  • $\begingroup$ @SRMaiti (As mentioned in the question,I do not need for the "coupling" between inter- and intramolecular coordinates".) I do not insist on using Gaussian, but the other software is expected to calculate the MP2/aug-cc-pVTZ energies in parallel and as fast as Gaussian does. $\endgroup$
    – TobiR
    Aug 11 at 12:35
  • $\begingroup$ if you can calculate the numerical hessian, then you can transform the Hessian into internal coordinates and project into the space of the ICs that you care about. From there you can back transform to Cartesians or get force constants etc. $\endgroup$
    – Cody Aldaz
    Aug 11 at 22:18
  • 1
    $\begingroup$ You should also check out Orca. It can also calculate hessians for particular sets of internal coordinates, and it does a two-sided displacement i.e. is slightly more accurate. However, one limitation is that it only takes atom numbers, and it will calculate all internal coordinates involving those atoms (i.e. you cannot specify bonds, angles etc. you can only select atoms). Orca can do MP2 calculations in parallel, not sure if it's faster or slower than Gaussian though. $\endgroup$
    – S R Maiti
    Aug 12 at 3:18

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.