`REFLN` category has no `Set` category links: is this a problem?

[jamesrhester]

Currently the REFLN category has no Set category links, and therefore is relying on implicit linking with other categories in order to be interpreted. Other dictionaries (e.g. modulated structures, powder) use this category, and so any new keys that are defined in the multiblock dictionary must be compatible with them.

Given that REFLN contains both model and observation, there is a dependence on the structure category (model) and (somehow) diffrn. However, for single crystal, multiple crystals can contribute, at the extreme end liquid flow experiments might contribute thousands, so there is no single _diffrn.id that can be associated with a particular hkl. In single crystal, separate observed reflections (diffrn_refln) are merged to a final reflection, thus removing the unique link to diffrn (and exptl_crystal).

As a first suggestion:

1. We add _refln.structure_id as a key data name linked to _structure.id.
2. We add _diffrn_refln.merged_id as a non-key data name pointing to _refln.id

Given the above, I think we have all the information needed to calculate/find values of all non-key datanames in REFLN, particularly F_meas and F_calc: for a given row we have a structure via _refln.structure_id. We need wavelength to do the calculation, which we find by backtracing the _diffrn_refln.merged_id -> _refln.id link, then _diffrn.id for that particular _diffrn_refln.id gives us our radiation source via _diffrn.id -> _diffrn_radiation.id. The wavelength used is the weighted sum of those listed in diffrn_radiation_wavelength - as long as the peaks haven't been spread out so far that they are considered distinct. That's a topic for the core DMG.

[rowlesmr]

I agree with point 1, and defer to your judgement on point 2.

Just for reference, the powder dictionary redefines REFLN to have keys _refln.id, _pd_refln.phase_id, and _refln.diffractogram_id. There is non-key linked _pd_refln.peak_id, _pd_refln.peak_overall_id, and _pd_refln.phase_id*, along with _refln.wavelength_id

The powder definitions seem to break down if more than one diffractogram contributes to a reflection.

.

• this phase ID was the original use of _pd_phase_id

[jamesrhester]

The powder definitions seem to break down if more than one diffractogram contributes to a reflection.

I think this is true. The REFLN category was always envisaged as a list of each peak in each diffractogram, with no concept of merging values from separate diffractograms, which would anyway be against the spirit of Rietveld refinement. We could adjust the definition of the REFLN category in cif_pow to note this - in essence it is identical to the single crystal case if each reflection is only observed once.

So I think my suggestion (1) is unaffected.

My suggestion (2) would seem to make no difference to powder as a non-key data name in a different category is being proposed that would make no sense in a powder context as diffrn_refln is not used (well, it would make sense in a pre-Rietveld world where peak intensities were extracted before data modelling, but in that case the data names work as advertised).