Page 1 of 1

Restarting a dimer

Posted: Fri Jan 06, 2017 1:45 pm
by dacaoyang2008
Hi All,
Now, I'm using dimer to search TS and I have two problems:
1.Can this method find the right saddle point for a reaction? For almost all of the references I read, the author used CI-NEB or NEB/dimer. But when I use neb, it's time-consuming, however, the dimer locate the TS quickly.
2.The first time I used EDIFFG=-0.1 and find the TS, now I want to set the EDIFFG=-0.05. Do I need to run the dimer all again or there is a way to use the first saddle point to reduce the computation.

Thanks a lot!

Re: Restarting a dimer

Posted: Mon Jan 09, 2017 3:52 pm
by graeme
The NEB and dimer methods are complimentary. The NEB is good when you know the final state of a reaction and want to find the reaction pathway and saddle for that specific final state. As you say though, it requires a number of images and so that calculation can be expensive. The dimer method is a single-point method which can be used to search for saddle points around an initial state. It can also be used to efficiently converge on a saddle from a nearby starting point. We have a neb2dim.pl script which interpolates from an NEB calculation to set up an initial geometry and dimer orientation for a dimer search.

You can restart a dimer calculation. Just as with any other calculation, you need to copy the CONTCAR to POSCAR. For the dimer, you should also copy the NEWMODECAR file to MODECAR. Again, we have a script "vfin.pl dir_name" which will move the results of your calculation to the dir_name subdirectory and leave things ready to restart in the current directory. This script will detect what kind of job was run (e.g. NEB, dimer, or regular) and move/copy the appropriate files.