[molpro-user] Reaction path following with approximate or no hessians
Dr Seth Carlton OLSEN
s.olsen1 at uq.edu.au
Wed Nov 10 06:41:56 GMT 2004
I am interested in constructing an approximate reaction path on the excited state of a 26 atom molecule using a state averaged CAS(10,10) wavefunction. I am concerned that the numerical hessian evaluation at the beginning of a OPTG,METHOD=QSDPATH will be too costly if the hessian is evaluated with the SA2-CAS wavefunction with that active space. I note that the first point corresponds to the S0 minimum - it is not a transition state on S1, where I want to find the path. I was wondering:
1. Is is possible to use an approximate starting hessian or a hessian calculated with a different wavefunction? (for example, one with a reduced active space). Does anyone have any experience with this?
2. Does the QSD procedure for straight optimization follow the procedure outlined in the third of the 3-part series by Sun and Ruedenberg in JCP,v. 99 (1993)? It seems in this case that it could be adapted to my purpose with a small step length.
3. It seems that a simple gradient-based procedure would work, in the context of an infinitely damped langevin dynamics approach. In this case, it would be necessary to mass-weight the appropriate gradient for the step. Does MolPro store the atomic masses in an array that can be addressed within the program to do the necessary mass weighting? Are the gradient elements stored in an accessible array?
I welcome any further insight in to this issue beyond the considerations framed in the above questions.
Dr Seth Olsen, PhD
Postdoctoral Fellow, Computational Systems Biology Group
Centre for Computational Molecular Science
The University of Queensland
Qld 4072, Brisbane, Australia
tel (617) 33653732
fax (617) 33654623
email: s.olsen1 at uq.edu.au
More information about the Molpro-user