Davidson calculations
Closed this issue · 4 comments
It seems from the paragsm1 file that the GSM program finished by running a Davidson diagonalization of the lowest (3?) eigenvalues of the transition state Hessian to verify that there is only one negative eigenvalue. Is this just a double-check, or is this step important (if more or less than one negative eigenvalue was found, would it go back to changing the string and TS)?
Please refer to this paper for more details:
https://onlinelibrary.wiley.com/doi/epdf/10.1002/jcc.23833
Okay I see now. I will follow up with a question about restarting a calculation: I have to restart my calculation periodically due to runtime constraints, saving a restart0001.xyz string file. It seems like it will go back and repeat calculations that it already had in the last run, maybe since it only uses a set of structures to restart. For example, I had a run that made it to the Davidson method portion at the end before timing out, and upon restarting the calculation began back at the climbing image search. Is there a way to restart without completing these processes again?
You can turn off Davidson in the code and recompile it, you actually don't need it.
In the gstring.cpp file, set the following to zero:
#define USE_DAVID 0
Got it, thank you!