Relaxing layered materials with LAMMPS

Hi all,

I know the present question might not exactly be strictly a LAMMPs question, but more a general MD question, so feel free to ignore it. Unless I am missing something specific about lammps.

I am relaxing a series of layers of G and BN stacked on top of each other. I’m using my own parametrization of the drip potential for the interlayer interactions (tersoff and rebo for intralayer interactions).

Using 4 BN layers, they are supposed to mimic the bulk substrate. I take a large vacuum of 100 Ang. I’ve noticed that these 4 layers do not relax to their equilibrium interlayer distance, but stay very close to the initial guess. If I start at 3.2 they will oscillate around 3.2 Ang, if I start at 3.3, they will oscillate around 3.3 Ang. Yet the equilibrium distance is 3.27 Ang. My total number of layers is 6 (including the overlayer and substrate), so there is still plenty of vacuum left to allow for full relaxation. I’ve been using the cg minimizer.

My experience on DFT tells me they should relax no matter what (as long as the starting configuration is reasonable). Is there something I am missing about relaxing layered materials in lammps (or MD in general) with weak interlayer potentials?


Hello Nicolas,

Do not use cg for this, I have the same problem. There are too many local minima to get trapped in, which is why you will want to use damped dynamics. Use min_style fire or min_style quickmin, I personally use fire. I also manually relax the lattice vector normal to the layers.


Thank you Shafat and Charlie. I’m recalculating with fire and will see if it solves my problem.


Just to confirm that indeed it works now and relaxes as expected using the fire minimizer (instead of CG).
Thanks for the responses I got (both on and off-mailing list) suggesting the same change. Seems like it is common knowledge :slight_smile: