UT theoretical chemistry code forum

Recently, I started to use the "single" image version of dimer.
The newer version is quite different from the previous one using "two" images.

I have two questions raised in using the newer version:

(1) I wander how CENTCAR differs from CONTCAR from a converged dimer run. Are they supposed to be virtually same?

(2) Suppose I stopped a dimer run before reaching to the convergence by creating STOPCAR in the work directory. Which one should I use as the POSCAR for the subsequent continuation run, CENTCAR or CONTCAR ?

Thanks a lot for your answer in advance.

HK

The CENTCAR contains the position of the center of the dimer, which is the current best estimation of the saddle. During the run, we make finite difference steps to evaluate curvatures. For these steps, the CONTCAR will be displaced from the CENTCAR by a small amount (DdR, or FDSTEP depending upon the optimizer).

Practically, these configurations are very similar, but for the sake of being able to continue a stopped run where you left off, the CENTCAR is written so that you can always use that as the POSCAR for continuation runs or for subsequent calculations that use the saddle geometry.

When a dimer run reaches convergence and stops on its own, the CENTAR and CONTCAR will be the same.