Dear LAMMPS community,
I am modelling a molecular system with both bonded interactions (within the molecule) and non-bonded interactions (between atoms of different molecules). I found the functionality of using neigh_modify exclude molecule/intra particularly handy for computational efficiency, and also for the easy parameterization of the bonded interactions.
As I understand it, this command (neigh_modify exclude molecule/intra) enables exclusion of atoms from the neighbor list which are part of the same molecule, as the neighbor list are needed only to compute non-bonded interactions. Thus the exclusion of atoms is based on whether their molecular IDs match.
However, if I have a large molecule in a periodic box, such that it starts near the one end of the box and ends at the other end of the box (but does not cross the periodic boundary), I would expect that the interaction of this molecule with its periodic image should be treated as if they are different molecules. But, I think in its current implementation, the two molecules (the original and its periodic image) retain the same molecular IDs. I assume there are no "ghost molecules"?
My question therefore is, can I use this command and still be able to have interaction between a molecule with its periodic image? Should I be using some other method to achieve this result, or is this just a limitation? If it is a limitation, can it be fixed by simple changes to the source code?
Best regards,
Nitant Gupta
PhD Candidate
Department of Materials Science And Nano-engineering
Rice University
for a large molecule you should not be excluding all intramolecular non-bonded interactions.
that would be wrong. rather you should follow the “special_bonds” settings required by your force field.
in most cases (and that is the default setting in LAMMPS) this will exclude all directly bonded (1-2) pairs, and pairs connected via a central atom (1-3 pairs) and pairs connected via a central bond (1-4 pairs). non-bonded interactions beyond that should be included, as at that point, there is no significant difference between atoms from the same or a different molecule.
axel.
Hello Axel,
Thank you for your reply. Big fan!
I agree that for typical situations (where the non-bonded interaction represent simple van-der Waals type interaction) I should not exclude intramolecular interactions beyond 1-4 neighbors.
However, the system I am modelling is using a granular potential for non-bonded interactions (actually to model friction between the molecules), which is why such interactions within the molecule are kind-of meaningless for the purposes of my problem. If I don't exclude them, then I get uncharacteristic behavior (I can explain this in more detail if needed) within the molecule, which is undesirable .
So, from your answer then should I understand that it is not possible to be able to exclude intramolecular interactions, but still have interaction between a molecule and its periodic image?
To be more clear, when I said large, I only meant that its size is comparable to the dimensions of the box. This is only needed so that I can construct a unit cell of the "bulk solid representation" of such molecular assemblies (which have relative friction between the molecules).
This analysis can be analogous to having a unit cell of atoms in a many-body potential. From common experience, if such a unit-cell is relaxed in LAMMPS, then due to ghost interactions, lattice properties (lat.const, elastic moduli,etc.) are easily reproduced, and a large super-cell is not needed. I want to do similar analysis for a molecular solid.
Best regards,
Nitant
Quoting Axel Kohlmeyer <[email protected]>:
Hello Axel,
Thank you for your reply. Big fan!
I agree that for typical situations (where the non-bonded interaction
represent simple van-der Waals type interaction) I should not exclude
intramolecular interactions beyond 1-4 neighbors.
However, the system I am modelling is using a granular potential for
non-bonded interactions (actually to model friction between the
molecules), which is why such interactions within the molecule are
kind-of meaningless for the purposes of my problem. If I don’t exclude
them, then I get uncharacteristic behavior (I can explain this in more
detail if needed) within the molecule, which is undesirable .
you are not making sense here. consider a geometry where the molecule is bent back on itself, e.g. forming a U shape.
now, when you exclude all intramolecular non-bonded interactions, the particles at the end of the U can move on top of each other without any repulsion, which is very obviously unphysical. your argument doesn’t count, then where is a difference between two “legs” of an U rubbing against each other? they need to be treated as “a leg” regardless of whether they belong to the same molecule or a different one.
So, from your answer then should I understand that it is not possible
to be able to exclude intramolecular interactions, but still have
interaction between a molecule and its periodic image?
i have not said that. the best way to get decisive answer is to set up a test system. but please consider my concern about your issue. if you have problems without the exclusion, perhaps that is a different problem and the exclusion is not really fixing it, but rather hiding it.
axel.
Thanks Axel.
I see that by not-excluding intramolecular non-bonded interactions, my primary problem (interaction of molecule with its periodic image) will be solved. I will try to think about this option more carefully (by potentially modifying the model).
The reason I wanted to exclude such interactions is because the bond-lengths inside the molecule are much smaller compared to the length scale for non-bonded interactions.
More details: Since I use a granular potential model, the non-bonded interactions are between beads which are much larger than bond lengths. This results in a single bead significantly overlapping with several of its bonded neighbors. The large size of the bead is required to have correct interaction with neighboring molecules, since granular interactions are based on contact between beads and inter-molecular bead distance does not need to be of the same length scale as intra-molecular bead-distance (or bond-length). As for your earlier point, I actually never have to worry about having "U" shape of the molecule, because they are practically rigid, and can not bend too much.
This then brings me to your suggestion to use special_bonds command. However I realize that the current implementation of special_bonds is incompatible with granular potential (or any pairwise potential besides lj and coul). Just as an example, if I have morse-potential for the non-bonded interactions, will I be able to use special_bonds to selectively exclude 1-2, 1-3 and 1-4 atom pairs?
Best regards,
Nitant
Quoting Axel Kohlmeyer <[email protected]>:
Thanks Axel.
[…]
This then brings me to your suggestion to use special_bonds command.
However I realize that the current implementation of special_bonds is
incompatible with granular potential (or any pairwise potential
besides lj and coul). Just as an example, if I have morse-potential
for the non-bonded interactions, will I be able to use special_bonds
to selectively exclude 1-2, 1-3 and 1-4 atom pairs?
you need to study the documentation more carefully before making such incorrect claims.
axel.