Posted by Björn Roos on April 15, 2004 at 12:26:54:

In Reply to: CASPT2 analytic gradient was so bad??? posted by Evgeny K. Dolgov on April 15, 2004 at 09:32:20:

There are, however, numerical gradients in MOLCAS-6. They work quite well for geometry optimizations on systems with at least up to 10 atoms. I have used such technigues myself to study the vibronic spectrum of pyrrol (JCP, 116, 7526 (2002)) with very nice results. Luis Serrano in Valencia is also using this method for vibronic spectroscopy (Luis.Serrrano@uv.es) and so does M. Olivucci in Siena. What you do to compute a vibronic spectrum is to use the numerical gradients to find the geometry of the ground and excited states. You then ise a grid of points around that minimum. Compute the energies and fit a second order polynomia, from which you obtain the force constants and can computed the vibrational energies. You also need to fit the transition dipole operator on the same grid. You then use the program MULA of MOLCAS-6 to compute the vibronic spectrum. A shortcut is to use the CASSCF method for the geometry and force constants, but this leads to lower accuracy. It is a good start though!

Good Luck!

Björn Roos