Modification to the implementation of the dimer method

[Jaap Louwen]

Hi All,

I am considering to do some work on the dimer method implementation in CP2K 
to see if I can increase the speed of overall convergence to the transition 
state. To avoid duplicate efforts, I'd like to ask if someone else has 
perhaps already done something similar.

To be more precise:
Kaestner and Sherwood (JCP, 128, 014106) found that convergence of the 
dimer vector can be greatly enhanced by using the (L-)BFGS method. Further, 
I may want to play around with using a BOFILL updated approximate Hessian 
(possibly modified such that the directional vector is a true eigenvector) 
to speed up convergence to the TS.

If code for such modifications already exists and it can be shared that 
would be great. Otherwise I'll just proceed.

Best regards, Jaap
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20180419/8d541cb5/attachment.htm>