Propose unit cell concept

[rartino]

No description provided.

[jamesrhester]

So I'd note that this definition will be easier to write if we have agreed on the concept of a crystal structure model which we have a draft of here. Can we agree on that first, and then the unit cell concept should fall out of it?

I note that the present draft for unit cell includes ideas of disorder, whereas the simple crystal structure model described above doesn't explicitly allow for that. We could either adjust the crystal structure model description to allow for the unit cell to have "disordered" contents while having fixed dimensions, or we could have the basic unit cell with well-defined contents and then a separate concept that allows for various forms of disorder.

[sauliusg]

I have summarized thoughts on the "unit cell" (and, incidentally, of the "crystal") definitions from today's discussions:

crystal-unit-cell-concept.txt

[jamesrhester]

After reviewing the EMMO ontologies for data and semiosis, and bearing in mind last weeks' discussion, I think perhaps it makes more sense to pursue "unit cell" as the key concept instead of "crystal structure model" (see #12 and a discussion topic #17). How I think this might work:

1. The unit cell parameters listed in CIF files and derived from experiments are "secondary data", that is, data derived from primary data. We can use the data ontology to describe these values.
2. Separately, we use the "reductionistic" perspective to describe crystalline materials as being composed of many non-overlapping unit cells.
3. We then use semiosis to state that the unit cell parameters found in CIF files ("secondary data" from point 1 above) are interpreted by the publisher of the CIF data to be parameters of the unit cell we have described in the reductionistic perspective.
4. The same process can be used to capture the other information that arises from a crystallographic experiment.

I had initially understood that we should not think about working with the reductionistic perspective because we weren't 100% sure that a material really was a crystal, so could not properly assign any individual to the entity "crystalline material". However, if we are simply interpreting a piece of data as a unit cell parameter, then we are not committing the ontology to assert that any given material is crystalline / aperiodic / periodic. A search for a crystalline material could be expressed as "find a material to which at least one publisher has assigned unit cell parameters".

My question to @emanueleghedini, @jesper-friis and @francescalb as EMMO experts is, does that sound like a reasonable way to proceed?

[rartino]

I'm a bit worried that this is taking us in a direction of an ontology that will be difficult for our use cases of materials/crystallography databases that are primarily built on theory and computation, and thus work entirely with models (in the arguably most common use of that word in science).

Nevertheless, with a completely reductionist perspective, isn't "unit cell" also a model abstraction? Don't we need to get to the cell parameters bottom-up through something like this?:

• An experimental x-ray diffraction pattern is-a semiotic sign; which has been created via an x-ray experiment performed on a sample.
• An essentially sharp diffraction pattern is-a experimental x-ray diffraction pattern; where most of the intensity of the diffraction is concentrated in relatively sharp Bragg peaks.
• A diffraction 2*theta peak value is a semiotic sign; which is interpreted from an essentially sharp diffraction pattern.
• A lattice constant or angle is-a semiotic sign; which is the result of a mathematical equation on a set of diffraction 2*theta peak values.

But this would leave us completely without the abstract concepts of "unit cells" and "periodic crystals". So, can we get there via "semiotic representations", or, is that just another word for models? E.g.,

• A crystalline solid represents a sample with an essentially sharp diffraction pattern.
• A unit cell is-a constituent part of a crystalline solid.

[jamesrhester]

I don't think "unit cell" has to be seen as a model abstraction. If the EMMO physicalistic perspective is able to posit things like "gluons", "gravitons", and "fields" and so on, then I think we can happily introduce a "crystalline solid" composed of many non-overlapping "unit cells".

I believe that all of the examples above ("diffraction pattern" through to "lattice constant") are simply types of data in EMMO. These numbers then have to be interpreted by somebody in order to see them as relating to the lattice parameters of a unit cell in a material, at least, that is how I think EMMO wants to describe the process.

I agree that it seems strange that we are sidelining models, when essentially all the experimental results come from applying a model, but on the other hand that means we don't actually have to reproduce the models in all their detail and variation. Instead we only need to introduce those aspects that the "interpreter" cares about (e.g. the choice of extinction model can be ignored). And the fact that models posit an infinite crystal is irrelevant, as the people interpreting the results of the model are happy to see the resulting lattice parameters as facts about an actual space-limited object.

Another "win" as I see it is that the whole aperiodic/periodic thing goes away. We construct in the physicalistic (reductionistic?) perspective distinct aperiodic and periodic crystals, and whether or not a particular material is seen as one or the other just depends on the model from which the parameters in the data base are derived.

At least that's how I see it but I would dearly love the ontologists to (a) confirm that this is viable (b) provide a pointer to how other domains have implemented models, where a "model" is a mathematical equation that has free parameters whose values can be chosen so that the result of calculations using the equation best match observed values.