On Sun, 2011-10-02 at 11:01 +0200, Guangping Zhang wrote: > I say non-local minimum to mean that the minimum is not the local > minimum that near the initial configuration. I want the system to go > over some barriers and find a lower minimum.
I don't think MD is a good choice in this case. There is metadynamics but as far as I know there is no implementation of it in SIESTA. Some variant of Monte Carlo could also be a good option. If you insist on SIESTA, you can try to set initial temperature relatively high, so the system is able to overcome the barriers, and cool it down during the run, so it settles in a "hopefully" more stable minimum - but you'll never be sure it really is the one you want. I guess the most reasonable approach is the one suggested by Yun-Peng - try different starting configurations and see what happens to your system. Good luck, Bartek
