Interpreting the van der Waals Energy Discrepancy between MoS2 and Graphene Layers in LAMMPS

I am conducting computational simulation tests and have observed that when simulating systems like molybdenum disulfide (MoS2), the interaction between layers (S-S interaction) is quite strong, preventing the system from ‘sliding’ in comparison to others, such as graphene, for instance. To substantiate this hypothesis, I thought about measuring the van der Waals energy between graphene layers (bilayer) and, separately, between MoS2 layers. When calculating the energy per atom, the vdW energy is lower for MoS2 than for graphene, both being positive. I am having difficulty interpreting this issue – does the lower energy imply a stronger bond in the system? Additionally, unlike graphene, MoS2 has half of the sulfur atoms in one layer interacting with half of the sulfur atoms in the other layer. Should I only consider the vdW energy for the groups of atoms that are interacting?

I selected this general discussion topic on LAMMPS. The documentation wasn’t entirely clear to me on how van der Waals energy is obtained. I apologize if the discussion of the physical interpretation of a simulation is not suitable for this forum, and I appreciate it if I could be directed to the appropriate place.

My LAMMPS simulation code is running smoothly.

It is difficult to comment on this since you didn’t say how you construct your model and specifically which potentials you are applying. It is also not clear if you are comparing absolute energies (and how obtained) or relative one.

It is borderline. For as long as you discuss this for a specific model and potential and provide reproducible LAMMPS input decks, people here might be able to tell syntactical or conceptual problems.

In the more general way, this might be something for the Science Talk category on MatSci.org or something like Physics Stackexchance, or ResearchGate or similar.

I apologize.

I am using a simulation with an NPT fix and ReaxFF interatomic potential.

I aim to demonstrate that a system, such as MoS2, encounters challenges in sliding (thinking about nano-scroll-type systems), unlike graphene, which slides easily.

Here is the simulation deck.
interaction_energy_forum.zip (1.3 MB)

Oh, I mentioned MoS2, but the actual issue pertains to WSe2.

The first thing that comes to my mind (being mostly ignorant about the details of the science of such systems) is that you may need to look at the Coulomb interactions. Coulomb because your materials are polar while graphene is not.

I would also be concerned about the accuracy of those calculations, if the ReaxFF parameters are only parameterized based on GGA level DFT calculations. Those are notorious for underestimating dispersion interactions. In general, I would probably look for some advanced meta-GGC DFT calculations. This is because until last summer the group of John Perdew was located on the same floor in my building and I know they were studying such layered systems and developing advanced DFT functionals (a family of functionals called SCAN, https://arxiv.org/pdf/1511.01089.pdf) to improve accuracy.

2 Likes

I appreciate the concerns regarding the parameterization issue.

On the other hand, you raised an interesting point. Perhaps the most compelling justification for the “non-sliding” behavior of TMDs lies in Coulombic energy. I just plotted it, and in the case of graphene, it is practically zero (-5E-10), whereas in WSe2, it is around -1.25 kcal/mol/atom.

The vdW energy term (as outputted by compute pair reaxff) in ReaxFF formalism is somehow a misnomer, because it also contains contribution from repulsion at short (<0.2 nm) distances. This is why the term is attractive.

There is no Coulomb energy in case of graphene, because the charge equilibration algorithm uses difference in electronegativity of atoms to assign electric charges. One type of atoms = no charge at all.

If you want to learn more about the energy terms in ReaxFF, read the SI to this article: https://pubs.acs.org/doi/10.1021/jp709896w

1 Like