Dealing with corrugation in lammps

Hi all,

I have a trilayer system in 2 dimensions that I properly set up with a triclinic cell. The central layer is rotated leading to moire type of physics, while the outer bottom and top layer are aligned with each other.

I am using a Kolgomorov Cresti (KC) type of potential to account for the registry dependence in the potential as well as a rebo potential for atoms in the layer.

My purpose is to find the shift value between the top and bottom layer that is most stable.

I have tried a series of initial shifts. All of them relax to the same sliding, which in this specific case turns out to be zero. They are also all the same when seen looking at the xy-plane: the sizes of the local stacking registry have same ratios for all initial condition parameters.

All is good, except that depending on the initial sliding, the final system contains corrugation in the z-direction on the order of magnitude of 1 Ang.

I guess this would be fine, indicating that corrugation stabilizes the system (in absence of a substrate), were it not that with or without corrugation, the total energy is the same (different by about 10 micro eV per atom, which seems negligible to me).

So the system does not seem to differentiate between the presence and absence of corrugation, as long as the interlayer distances have properly adjusted between layers.

I can get rid of corrugation if I put the dimension to 2, but then I don’t relax to the ideal interlayer distances.

I have optimized the lattice vector both using a series of discrete values as well as by using fix 1 all box/relax iso 0.0 to allow optimization of the simulation box. I did my initial tests with 2.46 and the optimized value is 2.4602 Ang (for the constituent graphene system, not the super-cell). Yet, for the size of my systems/supercells, this could account of 1-2 Angstrom total mismatch, but not for the corrugation with an amplitude of 1 Ang at various points in the system. Also, I re-did the tests with 2.4602. While slightly reducing the corrugation, it definitely does not get rid of it.

My questions:

  • Is a combination of KC type and rebo pair-wise potential able to correclty capture corrugation physics?
  • Before considering anything physical, is there something I am missing from a simulation point of view to properly simulate large 2d systems that are not expected to show any corrugation?

Thanks in advance.