[CPMD-list] Problem doing transition state search

[Salomon Billeter]

Dear Christian,

you can try the linear-scaling geometry optimizer which is new in CPMD. It
can do both linear-scaling geometry optimization and transition-state
search (P-RFO, cubic scaling, and microiterative based on P-RFO and L-BFGS,
linear scaling). You'll find details in

Comput. Mat. Sci. 27, 437-445, 2003.

The manual tells you the options. Please see also the hints section. P-RFO
partitions your configurational space into two subspaces, one spanned by
the Hessian eigenvector along which you want to maximize the energy and the
orthogonal space for which the energy is to be minimized. You can either
follow an eigenmode (based on overlap), or you can select a fixed mode.

The P-RFO implementation in CPMD has a dynamic trust radius scheme based on
energy prediction which helps ensure the stability of the optimization.
However, it requires you to choose the right mode at the beginning.

Best regards,
Salomon