The idea of this feature would be to have an automated search of rings of a particular number of atoms (that is, squares, pentagons, hexagons, etc) within a sisl.Geometry() object. Such a feature would be extremely useful to further analyze/categorize the inner fragments of a given geometry. For example, say you have a nanoporous graphene structure, such as in this open-access paper (Fig. 1 in https://onlinelibrary.wiley.com/doi/full/10.1002/adfm.202104031), which is formed as a 2D array of (1D) graphene nanoribbons (GNRs). By getting the hexagons within the NPG, one could probably find a relatively simple (and general) manner to classify the atoms within the NPG as belonging to one GNR or the other, which would be something much harder to do (in a general way) without having that prior classification of atoms as within hexagons.
The way that I implemented this thing (outside sisl, of course) was focused exclusively on hexagonal rings in organic systems (i.e. benzene or aryl rings). I find that using the connections (or bonds) information is the most general way to do this. I attach a snipped of that implementation I made, where this is done within a polymer class that I made myself, so the "self" is a polymer instance.
rings_snipped.py.txt
As you will see, this is just 7 for loops over all atomic indices, and for each index we loop over the bonds of that atom (via "for id_3 in self.bonds[id_2]", for example) and we keep this going on until we find a closed 6-atoms ring. Of course, this is not a very efficient implementation, but for small unit cells of a few hundreds of atoms this takes absolutely nothing in a regular laptop. Of course, such implementation would be more problematic for large-scale structures, I guess, but for regular periodic or molecular structures this works fine.
Another important question is how to treat the hexagon (or pentagon, or whatever) structure (i.e. atomic indices): that is, if you treat them as a list of atoms, or as a geometry object in itself, or, if I understand it correctly, as a geometry.category - obviously, the more things we could do with that object the better, however, the complexity of the thing could explode quite easily. But that might be subject for another issue/discussion.
The idea of this feature would be to have an automated search of rings of a particular number of atoms (that is, squares, pentagons, hexagons, etc) within a sisl.Geometry() object. Such a feature would be extremely useful to further analyze/categorize the inner fragments of a given geometry. For example, say you have a nanoporous graphene structure, such as in this open-access paper (Fig. 1 in https://onlinelibrary.wiley.com/doi/full/10.1002/adfm.202104031), which is formed as a 2D array of (1D) graphene nanoribbons (GNRs). By getting the hexagons within the NPG, one could probably find a relatively simple (and general) manner to classify the atoms within the NPG as belonging to one GNR or the other, which would be something much harder to do (in a general way) without having that prior classification of atoms as within hexagons.
The way that I implemented this thing (outside sisl, of course) was focused exclusively on hexagonal rings in organic systems (i.e. benzene or aryl rings). I find that using the connections (or bonds) information is the most general way to do this. I attach a snipped of that implementation I made, where this is done within a polymer class that I made myself, so the "self" is a polymer instance.
rings_snipped.py.txt
As you will see, this is just 7 for loops over all atomic indices, and for each index we loop over the bonds of that atom (via "for id_3 in self.bonds[id_2]", for example) and we keep this going on until we find a closed 6-atoms ring. Of course, this is not a very efficient implementation, but for small unit cells of a few hundreds of atoms this takes absolutely nothing in a regular laptop. Of course, such implementation would be more problematic for large-scale structures, I guess, but for regular periodic or molecular structures this works fine.
Another important question is how to treat the hexagon (or pentagon, or whatever) structure (i.e. atomic indices): that is, if you treat them as a list of atoms, or as a geometry object in itself, or, if I understand it correctly, as a geometry.category - obviously, the more things we could do with that object the better, however, the complexity of the thing could explode quite easily. But that might be subject for another issue/discussion.