Energy mismatch between NEB/Single Point

[ng00822]

Hi Prof Henkelman,

Firstly, thanks for all of your (and your team's) work on enabling, supporting, and improving VTST over the years! I had a problem recently about differences in energy values being reported from an NEB run, and the corresponding single point energy calculation for the same images. The offset in energy was fairly consistent, about 0.46 eV, but I couldn't figure what would cause that change. I've had the same problem with VASP 4.6/VTST 2.03 and VASP 5.3.3/VTST 3.0. The endpoints (and all images) are fully relaxed to < 0.02 eV/A, but the volume/shape is not completely relaxed, there is a residual stress of ~10kbar. Do you have any idea what could be causing this? I've attached the directories for each version. BTW, this is a vacancy migration path in an alloy. Also, I just realized in this example that NEB is being run with medium precision, while the SinglePoint is with Normal, but that doesn't make any difference that I found. Thanks in advance!

-Niraj

[graeme]

I did not check this, but I think that the LREAL options, AUTO vs TRUE, which are also switched by default in the change between PREC Medium and Normal are causing this energy difference. These settings must be the same in order to compare energies between calculations.

[ng00822]

That was indeed the case, I wouldn't have ever thought of that, thank you!