Paper Spotlight; Structural Rugosity in Crystallisation
September 11, 2020
The field of Crystal Structure Prediction (CSP) has grown dramatically over the last few decades, but today I want to highlight a recent paper which takes a new approach. The recent publication from Montis et al makes use of new informatics tools to assess surface roughness (rugosity) and proposes a relationship between this and ease of crystallisation.
The CSP challenge – finding diamonds in the rough
Although the many potential applications and benefits of being able to predict a crystal structure from a simple 2D molecular structure are known (see our introduction to CSP video here) there are many challenges that stand in the way. As the authors of this paper note, one of the major challenges is over-generation; a CSP landscape will return many possible results, but identifying which of these polymorphs will be experimentally observed is not simple.
The authors of this paper posit that surface roughness of potential crystal faces could be key in linking theoretical predictions to observed structures. This is a novel approach, building on the methods developed by Mat Bryant, Andy Maloney and Richard Sykes at the CCDC during the ADDopT project.
Morphology & Rugosity
Crystals grow in very defined ways. Look in your pot of sugar under a microscope; you’ll easily see that the crystals are essentially pretty much the same shape to a degree. This is because when a crystal grows there are certain directions that grow more quickly or slowly due to the ease (or difficulty) for new molecules to get attached to the surface. The consequences are certain crystal faces feature in all the crystals of a given form. The faces that make up the surfaces of the crystal define the Morphology of the crystal and due to the nature of crystal growth, crystals of the same polymorphs will have the same set of indexable faces.
The face rugosity of a surface measures, in effect, how bumpy a surface is by looking at how far atoms in molecules cross a given plane. This gives a simple and fast metric for characterising a given surface. Building on the methods described by Bryant et al, the team generated a new crystalline rugosity measure by averaging the face rugosity across the BFDH Morphology faces, weighting for face importance to get a measure of crystalline rugosity.
Montis et al reasoned that due to interfacial tension, they might expect rougher surfaces to have slower growth than smooth surfaces and so reasoned that this crystalline rugosity might be useful for interpreting polymorphic forms.
The first interesting results came when they started looking at how this crystalline rugosity can be used to characterise pairs of polymorphs. They calculated the crystal rugosity on a number of datasets, showed that the difference in the calculated crystalline rugosity betwen forms was rarely greater then 0.2, and that for a selected number of cases where the pairs of polymorphs were ranked as easy to form or hard to form, the easy to form polymorph often had a lower crystalline rugosity than the hard to form one, particularly when there was no other cause known for polymorphism (such as conformational polymorphism.)
The authors then took this further and assessed all the structures in a crystal structure prediction landscape which contained observed and unobserved polymorphs. They found (in 2 different cases) that the observed meta-stable polymorphs in the landscape were generally smoother in nature. Also intriguingly, the method seemed to allow a user to focus in on the important part of the landscape. In one of the 2 landscapes (the enigmatically named system Compound X) the metric suggested that some energetically more favourable forms, as yet unobserved, would be very hard to form due to their crystalline rugosity.
Rugosity, of course, is in practice only part of the picture; the authors note that there are many other reasons why a polymorph may not form, but this method does look like a really interesting addition to the arsenal of metrics one might use to interpret the crystal structure prediction landscape of an organic compound.
Improving morphology prediction?
There are many methods already in the literature for calculating morphologies that are more sophisticated than BFDH, but these generally only consider the energetics of a given face. It will be interesting to see if rugosity can also help in these predictions too, perhaps by down weighting faces that are energetically favourable but deemed too rough for growth.
Driving CSP forward
CSP has come a long way, and initiatives like the CSP Blind Test continue to push the methods forwards – but there is a long way to go. This work is of particular note as it takes a new approach to thinning the CSP landscape; it will be intriguing to see how such methods could have impact in future.
The seventh CSP Blind Test administered by the CCDC will begin in late 2020 – find out more here.
Crystal Structure Prediction (9)
Crystalline Rugosity (1)
CSP Blind Test (16)