Finding synthesisable structures: adapting the generalised convex hull to molecular crystals Jennie Martin 1 , Graeme Day 1 , Michele Ceriotti 2 1 University of Southampton, Southampton, UK 2 École Polytechnique Fédérale de Lausanne, Lausanne, Switzerland Crystal Structure Prediction (CSP) is a useful tool in materials discovery for identifying potential structures and their properties of interest ahead of synthesis. However, it suffers from a problem of overprediction, and techniques are needed to identify the synthesisable candidates from a set of structure predictions. One such technique is the recently developed Generalised Convex Hull 1 (GCH). The GCH identifies synthesisable structures as those close to the convex hull on a landscape of structures plotted in terms of energy and machine-learned structural descriptors - derived using the Smooth Overlap of Atomic Positions (SOAP) similarity kernel 2,3 and kernel Principal Component Analysis (kPCA). The approach has demonstrated positive results, with the ability to recover stable hydrogen-oxygen binary phases and identify potential organic semiconductor structures that are potentially stabilisable by nitrogen substitution 1 . However, its implementation is not optimised for applications to organic molecular crystals as the underlying SOAP kernel involves computing atom-atom comparisons that are not theoretically reasonable for defining their structural similarity. In this work, we discuss adaptation of the GCH approach to be more chemically reasonable for exploring molecular crystals. This involves adaptation of the input SOAP kernel – limiting atom-atom comparisons made in obtaining the kernel to a set of comparisons that can reasonably be said to define the similarity for molecular crystal structures. We then put these alterations to the test, applying the adapted GCH to organic molecular systems with varying applications, such as pharmaceuticals, organic semiconductors, and porous structures. References 1. A. Anelli, E. A. Engel, C. J. Pickard, and M. Ceriotti, Phys. Rev. Materials , 2018, 2 , 103804 2. A. P. Bartók, R. Kondor, and G. Csányi, Phys. Rev. B ,2013, 87 , 184115 3. S.De, A. P. Bartók, G. Csányi , and M. Ceriotti, Phys. Chem. Chem. Phys ., 2016, 18 , 13754
P38DD
Made with FlippingBook Learn more on our blog