Clay layer disintegration (explosion) during water relaxation in Montmorillonite using CLAYFF

LAMMPS LAMMPS Beginners
,
1

E0FF

Hello everyone.

I have a problem, I wanted to simulate hydrated montmorillonite. I created a montmorillonite model from a pyrophyllite lattice through a python script, replaced aluminum atoms with magnesium atoms through a randomizer and added sodium ions, tried to compensate for the charge and pulled apart the layers to place water in LAMMPS through create_atoms. After placing the water molecules and attempting to relax the system, the lattice bent and tore.

I suspect the issue originates either from a geometric distortion introduced during the triclinic matrix transformation in Python (when mapping manual Cartesian shifts along the tilt vector onto the expanded cell matrix), or from an incorrect minimization protocol in LAMMPS that fails to handle the initial water-clay overlaps safely.

I have attached a snapshot of the corrupted structure, along with my input files and the log.lammps file showing the extreme forces at the start of the simulation.

Could anyone review my workflow or suggest a foolproof minimization/restraint protocol for CLAYFF to prevent the clay layers from exploding under initial water overlap? Any guidance would be highly appreciated!

Best regards, E0FF

2

Most likely the problem is with the geometry of the initial structure. I am doing a similar work, using ClayFF to relax aluminosilicates with various defects and compositions. The protocol I am using is:

  1. minimize
  2. fix nvt
  3. fix npt tri

If the initial guess is particularly bad, you may want to do a thermal annealing at fixed volume, or use an isotropic barostat for the initial phase of the equilibration.

You didn’t attach any file, but bear in mind that this forum is not a free debugging service. Nor it is advisable to run computer simulations with instructions provided by random people from the internet. I had some bad experiences a looong time ago (thank you, telnet!)

1 Like
3

Hi, thank you very much for your reply. The problem was that I initially selected the wrong structure. The calculations were successful!

1 Like
4

I hope the new structure won’t fall apart when NIEL’s flaws are introduced.