UT theoretical chemistry code forum

Hello All,

I have noticed that literatures on dimer methods has specified their parameters such as dimer separation, maximum move distance, and the values of the finite difference steps for rotation and translation. However I am not sure where the finite difference steps for rotation (delta theta) is claimed. Is there a INCAR tag for this? and if there is, what would be the default value?

Also, I noticed that ppl ran hundreds of dimer method with the variance in unit vector to find saddle points. Say If I am only interested in the lowest transition barrier between two minima, would you recommend to run with greater value of the finite difference steps for rotation and translation? FYI, CNEB method is not working well in my system which is amorphous.

Any reference or suggestion would be helpful.
Thank you in advance .

Tim

Using a trick that was introduced by Andreas Heyden (JCP 123, 224101, 2005) we use a 45 deg rotation instead of a finite-difference value. This removes one parameter from the method without introducing any additional uncertainty.

Finding the lowest saddle from a minimum is a harder problem than converging to a nearby saddle. If you know the final state, I strongly recommend using the climbing image NEB. If you don't know the final state, you can make displacements from the minimum and run the dimer method to find saddles. These displacements can be made at random or by targeting active regions (e.g. undercoordinated atoms) where you expect reactions to occur. Determining the optimal magnitude of the displacement is a bit of an art. You want to displace enough so that you quickly find a negative mode; very small displacements will lead to following soft modes which are not generally relevant to low barrier reactions. Very large displacements will lead to a wider variety of saddles found, but at the cost of more failed searches and finding saddles which do not nessecarily directly connect back to the initial state.