Using a machine learning approach to determine the space group of a structure from the atomic pair distribution function

Abstract

A method is presented for predicting the space group of a structure given a calculated or measured atomic pair distribution function (PDF) from that structure. The method utilizes machine learning models trained on more than 100 000 PDFs calculated from structures in the 45 most heavily represented space groups. In particular, a convolutional neural network (CNN) model is presented which yields a promising result in that it correctly identifies the space group among the top‐6 estimates 91.9% of the time. The CNN model also successfully identifies space groups for 12 out of 15 experimental PDFs. Interesting aspects of the failed estimates are discussed, which indicate that the CNN is failing in similar ways as conventional indexing algorithms applied to conventional powder diffraction data. This preliminary success of the CNN model shows the possibility of model‐independent assessment of PDF data on a wide class of materials.

There's no place like real-space: elucidating size-dependent atomic structure of nanomaterials using pair distribution function analysis

“…246 For example, machine learning methods have already been applied to PDF in regards to component analysis 123,[247][248][249][250] and in identifying symmetry and extracting distance lists. 251,252 This development is likely to aid in maximizing the information that can be extracted from the PDF of complicated nanostructures. Automated modelling, where large numbers of structures (mined from databases or algorithmically generated) are tted to experimental PDFs, can improve the structural characterization workow and the discovery of new and improved models.…”

A cloud platform for atomic pair distribution function analysis: PDFitc

“…(ii) spacegroupMining: given a PDF, the app will use a pretrained convolutional neural network to predict the most likely space group of the structure that produces the PDF (Liu et al, 2019).…”

Structure determination of organic compounds by a fit to the pair distribution function from scratch without prior indexing

“…However, these methods -regardless of whether an organic or inorganic sample is investigated -require at least a rather well matching crystal structure model(s) (Farrow et al, 2007;Neder & Proffen, 2008;Yang et al, 2020) or at least the knowledge of the unit cell and space group (Prill et al, 2016) in order to succeed in a reasonable fit. Remarkable work was recently published describing the determination of the space group from the PDF data (Liu et al, 2019). Nevertheless, the identification of the lattice parameters is challenging for nanocrystalline compounds and often ends without an outcome.…”