I am constructing a model in LAMMPS using two approaches: (1) I create a supercell and then remove atoms from specific regions to form the final model, and (2) I directly construct the model and then fill it with unit cells. I am unable to explain why running a model created using approach 2 has failed, even after the minimization process. Additionally, the first approach presents challenges with significant computation time at the outset in very large models.
This is a very vague description, so I can only make vague statements.
Method 1 seems most robust to me, as filling space with voxels can be tricky --to say the least.
I think it is premature to conclude that the approach how you create a specific system determines its viability. After all, the physics of a model is a state function, i.e. it doesn’t matter how you get there. When you have the same system, you have the same results.
Thus if there are different approaches to do the same give different results, then it is safe to assume that you are not really creating the same system or that you are making mistakes somewhere.
“the physics of a model is a state function, i.e. it doesn’t matter how you get there”
I agree witht thist, however in lammps it is different to create a complex model from different regions and groups using the “union” or “subtract” for connecting them to create the main model.
It is better to use “region abstract” rather than “region union”, especially when there is a fixed part connecting to the moving part. otherwise, during minimization, they will be taken apart.
Such a blanket statement without proof is a rather unscientific way to discuss something in a scientific context. I maintain that if you are not getting the same results you are not creating the exact same system and that is - most likely - due to mistakes or bad choices. From that you cannot draw a conclusion that one approach is better than the other.
A common mistake with an “additive” system building is that people choose their region boundaries to coincide with lattice points. That can easily lead to atoms being created on the exact same position in two different regions that share a boundary. This ambiguity can be easily avoided by either shifting the lattice or choosing regions so that the boundaries are away from lattice points. Of course this cannot show by a “subtractive” approach, but that doesn’t make the additive approach unusuable. It just needs to be applied correctly. And that is not so much a matter of the LAMMPS code but of how it is used. There are other issues that require attention of the user, but that is just something that needs to be done anyway for any research project.