Nudged elastic band simulations with variable unit cells in LAMMPS

Dear Lammps users,

I want to calculate the energy barrier between two metastates of a periodic solid structure (periodic in x, y, and z directions).

The size of the unit cell (simulation box) of these metastases is different. However, when I read the documentation of NEB on the LAMMPS website (, it seems the present version of NEB implemented in LAMMPS is only valid for a fixed unit cell. Is it correct?

If I use this version to calculate the energy barrier for my case, will I get a reasonable result?

I found several papers about generalized NEB which includes the effect of variable unit cells, see e.g., and

If I want to implement this generalized NEB, do you have any suggestions? One main issue I noticed is that NEB is using a damped dynamics minimization (I guess fire) but relaxing the box during minimization with fire is not implemented. Do I need to implement this first?

I also noticed that the new version fire included the option to relax the box initially but finally is removed, is it because it’s too complicated?

Thanks in advance!



You posted the exact same e-mail on thursday. What is the point of sending it again?

If you didn’t get the response you wanted the first time, it won’t make it more likely to get a (better) response by resending the identical e-mail.

On the contrary, you will project the appearance that you are not willing to start a dialog with people, but rather only will accept a response that pleases you and will ignore everything else. Considering that people on mailing lists like this one volunteer their time, this will rather turn people off than it would make it more attractive to respond.


Hi Axel,
Thanks for the response. I’m resending the email again because I didn’t find my email in my inbox (I already subscribed to receive the emails from LAMMPS) and I thought it didn’t reach the mailing list. But later I found this is a problem with my email system. I should check more carefully before resending the email. Sorry for the convenience.

In the meantime, I indeed have some new information. I checked for my system that the energy barrier calculated from the NEB in Lammps is sensitive to the choice of the simulation box when its size is different between the initial and final configurations. Therefore, although LAMMPS does not check the size of the simulation box and can give some number for the energy barrier, this value seems to be inaccurate, or at least not the real one.

I also read more documentations on the website of LAMMPS, it seems it’s essential to implement the fire with the option to relax the box before implementing the NEB for variable unit cells, which is very complicated. I’m struggling to find a way out and I’m grateful if anyone has experience with this question can give some suggestions. Thanks in advance!