Coinage metals have always attracted significant interest due to their distinctive chemical and physical properties. Indeed, these properties are governed by their ability to form an assortment of complicated crystal structures like high-disordered polyanionic networks, quasicrystals, and complex bonding patterns containing a mixture of ionic and polar covalent bonding. Therefore, to further understand the fundamental chemistry of compounds containing coinage metals, the relationship between the composition and crystal structures must be studied. In this project, we employ a new machine-learning method that is capable of predicting formation energies by training our support vector regression algorithm using the reaction energies determined from density functional theory based on the chemical composition. The resulting tool allows us to screen unexplored phase-space by predicting compositions that fall near the convex hull. More specifically, we locate areas in the La-Au-Bi and Y-Ag-In ternary phase diagrams with a high probability of new phase formation and then target those compositions using solid-state synthesis. The crystal structures in the targeted regions of the phase diagram predicted by machine learning are solved using single crystal X-ray diffraction while the composition confirmed by Energy-dispersive X-ray spectroscopy (EDX). In addition, the electronic structure and corresponding chemical bonding were analyzed based on COHP (crystal orbital Hamilton population). The electron localization function (ELF) analysis and Bader charge analysis indicates the charge redistribution among the atoms. Our experimental results support our proposed machine learning approach to detect the favorable regions on an isosection of the ternary phase diagram thereby guiding out synthetic efforts.<br/>

Establishing the complex crystal chemistry of the coinage metals through synthesis, computation, and data-driven analysis

2019 Fall National Meeting

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.