[QE-users] NEB with variable cell?

[Andrii Shyichuk]

Dear Users,

I have two structures (A and B), on which I want to do a NEB 
calculation, using them as the initial and the final images.

The problem is: they differ in their cell size by about 1%.

My options:

1. Do NEB with lattice vectors from structure A, then another NEB with 
vectors from structure B, and then another one with the average vectors. 
That, obviously, would be an approximation, and I will see how bad it is 
from the results.

2. Do manual frozen "NEB", where I simply interpolate structure A into 
structure B in N steps and just do SCFs. Not a NEB per se, but that's 
something.

3. Interpolate A cell vectors into B cell vectors in N steps, and do 
atomic position relaxations at each set of vectors. Not sure if the 
moving atoms will move the way I expect them to move.

4. Get the reaction path from option 1, and then do vc-relax at each 
step, freezing the moving atoms at the path. However, constraints in PW 
do not work with vc-relax, as I can see from the manual. Thus, what I 
can do is option 3 with constraints.

So, what would you do? What do people usually do in situations like 
this?

Thank you in advance.
Best regards.
Andrii Shyichuk
Faculty of Chemistry, University of Wrocław, Poland.