UT theoretical chemistry code forum

Dear Sir,
I have utilized NEB to investigate the activation barrier of vacavcy(single) migration in Fcc-Zn(we all know that the stable configuration of Zn is Hcp_A3), a problem occurs: when I use different nodes and cpus, I get different activation barrier of vacavcy migration, sometimes it could be 0.461199eV, and sometimes it is 0.423174eV. I think it is a large difference, so what is the reason? (I have calculated the phonon dispersion of FCC-Zn, and find this configuration is mechanically stable).

Looking forward to your reply and suggestions! Many thanks!

Are these calculations all converged? This is best done by choosing a force criteria - e.g. ediffg=-0.01.

Is the box frozen, or allowed to relax?

Hi Prof.Graeme,
Two configurations are fully relax to get the initio-state and final-state, which were utilized to create the images!
I have checked that all the calculations are converged. And in my calculation, I utilize force criteria ediffg=0.01, both the cell size and the ionic coordinates are fully relaxed.
Question: If the initio-state and final-state are metastable, so is it reasonable to encounter this phenomena?
By the way, I have utilized the same procedure and settings to investigate activation barrier of vacavcy migration in fcc-Al(stable configuration), and get very good results!(0.598eV, the experimental data is 0.59+-0.03eV)
I will try as you suggested, many thanks!

Using a relaxed cell could cause the problems that you are seeing.

Try using a frozen cell (ISIF=0) and a force criteria (ediffg=-0.01). Note that you need the negative sign to make this a force criteria. The energy criteria that you are using is not appropriate.

Hi Prof. Graeme,
Many thanks for your helpful suggestions! I will have a try as you suggested!