The abnormal and negative NEB energy profile

[randome]

Deal all,
I obtained an abnormal and negative energy profile of the phase transition from beta-Ti to alpha-Ti. In general, there is no energy barrier for this phases transitions. However, the flat and descend energy profile may be changed when some additive elements are added.

I want to obtain the flat and descend profile energy profiles of these phase transitions without any additive element. In following figure, the ‘neb-no-CI’ line is the much negative energy profile of the transition using no-climbing-image NEB method, while the ‘imgs-scf’ line is the the static energy of each images. It can be noted that the neb-no-CI line is very strange because the alpha-Ti is most stable Ti (only omega-Ti is slightly lower in energy than alpha-Ti at 0 K). Though I don’t know the precise transition profile, I think it may be as like as the imgs-scf line.

I tried different Optimizer methods, but all the obtained profiles are negative. I relaxed the atom positions of image 4 and 5 after the NEB calculation, as shown the blue triangle, since they are very lower in NEB energy profile. However, it can be seen that the relaxed image 4 and 5 are higher in energy than the alpha-Ti (image 9) and very different from the result of NEB. So what’s the problem of the negative NEB profile? How can I obtain the reasonable NEB profile of this phase transition.

The required input files are compressed and uploaded as 'no-CI.7Z'. Thanks a lot!

[graeme]

Set ISYM=0 in your INCAR file.

mep.pdf

(964 Bytes) Downloaded 1195 times

[lukewithyou]

I have the same problem, but I set ISYM = 0, also I have negative energy. So I have a question, what's mean of negative energy?

[graeme]

It is always possible to have an intermediate minimum which is more stable then the reactants and/or products. In that case, you should minimize to find the intermediate minimum and construct separate bands from initial -> intermediate and intermediate -> final.

The case shown here is different because there is no barrier to the low energy intermediate state. This is not consistent with having minimized endpoints. I have shown (see my attached MEP) that the problem is due to an imposed symmetry constraint. The endpoints have a symmetry which is detected and then imposed by vasp, but the NEB images can move to break that symmetry, leading to an inconsistency between the ionic geometry and the symmetrized electronic structure.

Another very common cause of having NEB images with lower energies than the endpoints is that people do calculations of the endpoints with different settings (primarily ENCUT and KPOINTS) than they do for the NEB. Any changes like this will shift the absolute energies to that the endpoints can not be compared to the NEB images.

[randome]

Dear graeme, I highly appreciate for your comments. I learned much from your comments and finally obtain the reasonable energy profile. Thanks again!