Development and assessment of an improved powder-diffraction-based method for molecular crystal structure similarity

Abstract

Identifying whether two experimental crystal structures determined under different experimental conditions correspond to the same polymorph is a challenging problem in crystallography, and its solution has practical (and even legal) implications for various technological fields as well as for molecular crystal structure prediction. In this work, we assess the popular COMPACK method vis a vis powder X-ray diffraction (PXRD)-based comparison methods using a dataset of 44 939 structure pairs employed in a previous study [CrystEngComm, 2020, 22, 7170–7185]. We propose a new PXRD-based similarity index and comparison method (VC-PWDF, variable-cell powder difference) that substantially improves the agreement with COMPACK (2.84% total disagreement), compared to the CCDC packing similarity PXRD-based comparison tool (12.35%). By analysing the structure pairs for which COMPACK and VC-PWDF disagree, we evaluated the strengths and weaknesses of each method. COMPACK has a counter-intuitive dependence on its tolerance parameters, by which structures that are considered the same at a given tolerance are viewed as different at a looser tolerance. COMPACK's RMSD(N) can also increase with increasing tolerance values at fixed number of matching molecules (N). We demonstrated a few additional weaknesses of COMPACK: a) extremely costly or incorrect comparisons in molecules with highly-branched substituents (possibly due to its use of Ullmann's method), b) failure to match structures when the molecular connectivity is incorrectly determined, c) difficulties with molecules presenting helical chirality, and d) very large cluster sizes (up to 50 molecules) are sometimes needed to correctly identify unequal polymorphs. In turn, VC-PWDF has difficulty differentiating structures with similar packings, such as polytypes, and conformational and isomorphous phases. It is shown that the proposed VC-PWDF is at least as robust as COMPACK for comparing molecular crystal structures. Although a grey area of difficult-to-compare structures still exists, using both COMPACK and VC-PWDF in combination may be successful at narrowing it.