Streitz-Mintmire vacuum surface

The example (LAMMPS/examples/streitz/data.streitz) is a bulk Al2O3 structure – how can you create a stable surface by simply cleaving the XY plane?

The correct procedure is to first minimize a bulk structure, then create a valid surface (which is 0001 in your case). Please see the attached screenshot of a properly minimized Al2O3 (0001) surface.

By the way, almost for all oxide surfaces O-termination yields more stable surfaces.

Al2O3_0001.png

The example (LAMMPS/examples/streitz/data.streitz) is a bulk Al2O3 structure – how can you create a stable surface by simply cleaving the XY plane?

The correct procedure is to first minimize a bulk structure, then create a valid surface (which is 0001 in your case). Please see the attached screenshot of a properly minimized Al2O3 (0001) surface.

By the way, almost for all oxide surfaces O-termination yields more stable surfaces.

What I am saying is that you can’t just “put a gap” – that is what I meant simply cleaving the XY plane. You can create metal surfaces with this method without any problems but not with oxide surface. You need to start from the minimized supercell, carefully translate the atoms, put a gap in the XY plane, then carefully delete some atoms on both sides of the surfaces. This can be facilitated by visualizing the structure or use some external builders.

Sorry Ray, I see what you mean, but what I meant with my last message is that even with a bad surface, shouldn’t I expect the NVE integrator to conserve energy? Before I worry about creating a more correct surface to get the dynamics I want, I want to make sure I am setting up everything else correctly and that I am not missing some more fundamental issue.

So I guess the meat of my question is, based on the setup I described last time, why would energy not be conserved?

Thanks again

Seth

You are mistaken in that the NVE integrator doesn’t auto-magically conserve energy for any given system and potential parameters – it just doesn’t alter the equation of motion but the system is allowed to evolve to anything. With an unstable initial structure, the total energy is difficult to be conserved.

Ray

Sorry Ray, I see what you mean, but what I meant with my last message is that even with a bad surface, shouldn’t I expect the NVE integrator to conserve energy?

​you may expect it to conserve energy only when your system is in equilibrium. ​especially, when you have primarily cutoff based interactions and a non-homogeneous system, you will have unbalanced changes in the number and strength of interactions when you start from a non-equilibrated, high potential energy initial configuration.

axel.

Hey guys, sorry for the long delay between posts, my wife just gave birth last week, so couldn’t follow up as quickly as I would have liked.

Sorry for the naïve questions, but with the REAX force field, the cleaving method of just adding a gap and running dynamics to relax it worked fine, so I hoped it would work for this potential as well.

I expected that I wouldn’t need major changes beyond cleaving as the single aluminum terminated surface in every ab-initio minimization and LEED studies show only slight contractions of the two or three topmost layers away from their bulk positions with no significant restructuring, so I assumed that this very minor restructuring would lead to not especially large forces relative to the relaxed surface. Even after performing a minimization of the bare surface, there seem to be surprising large forces present.

Just to check my understanding of what is actually happening, I have a scenario that I tried, and I want to see if I am misunderstanding something about what LAMMPS is doing.

If I take the bulk structure from the example, cleave it between two aluminum layers, and add a gap; this should leave me with two UNrelaxed (0001) surfaces.

If I then run a minimize command with fairly tight tolerances, I should end up at the bottom of a potential well, local or otherwise (barring that I landed at a saddle point) even if it isn’t the most stable surface.

If I then run an NVE simulation without initializing any velocities (confirmed by the fact that the initial temperature is 0 in the log), it should just hover around that minimum configuration due to the very small forces from the minimized configuration; but the simulation does not hover near that configuration, it starts shaking (almost like a phonon vibration) and then eventually falls apart.

Am I just totally missing something obvious?

Thanks again for all your help.

Seth