optimizer, EDIFF, and DdR in dimer

cenwanglai · Post by **cenwanglai** » Sat Nov 12, 2011 1:51 am

(1) In the dimer example provided in the website, CG (OPTIM = 2) was used.

Could I change to LBFGS or other NEB optimizer?

In my dimer run when CG optimizer employed, it did not seem to be aggressive because only 0.001 ev or even lower energy decreases for steps.

(2) It seems that the EDIFF affects the Torque dramastically. Setting EDIFF from 1E-5 to 1E-6 makes the Torque decreases aggressively. But, why? Because for geo-optimization or static total energy calculation, EDIFF is only set to be 1E-4 routinely.

(3) How does the DdR effect the convergence process? What is the difference between DdR= 0.01 and 0.005 in convergence of electronic and ion steps?

Thanks for any advice!