Combined synthesis and computational search for new correlated electronic materials

点击翻译按钮获取中文摘要

Worldwide efforts are being made to develop new transition metal oxides due to their wide array of interesting behaviours such as superconductivity, magnetoresistance, ferroelectricity, photoluminescence, thermoelectricity, etc. More specifically, perovskites oxides, having the general ABO3 structure with large B cations sitting inside a framework of A cations and oxygen ions, have the ability and flexibility to accommodate a large variety of elements and are thus widely studied in order to carefully design new functional materials. A library of these oxides is built up generally by the substitution of cations. However, the anion or oxide chemistry can be vastly expanded through changes in the anionic lattice which is difficult to accomplish through conventional thermodynamic synthesis as only the most stable configuration or a mixture of configurations is formed. The varying stoichiometries of anions can be explored through kinetic control with topochemical reactions, which are reactions that are locally confined with crystal lattices. For example, the reduction of LaSrNiRuO6 to LaSrNiRuO4 leads to the formation of novel Ru2+ centers, allowing its electronic structure to be studied in extended oxide frameworks. It is, however, time-consuming to explore the chemical space of perovskite oxides through experimentation only.Recently, a wide range of functional materials have been under study by machine learning, allowing the exploration of vast configuration landscapes with high efficiency. For instance, machine learning has been applied to aid materials discovery such as the prediction of the critical temperature of superconducting materials, Curie temperature of ferromagnets, electronic structure features of photovoltaics, etc. While computational methods have been used to study topochemically modified structures, their simulations involve solving complex quantum mechanical equations, requiring exponentially increasing computational power with system size. Therefore, simulating numerous systems with a large number of atoms to elucidate material behaviour is extremely difficult with conventional methods. This project aims to combine computational modelling including machine learning and synthesis to expedite the discovery of these new topochemically altered perovskite oxides. Ultimately, this will provide an opportunity to create a feedback loop in which a new material is predicted by machine learning from existing databases, synthesising the material and feeding the information back to the learning model allowing the exploration of a myriad of material structures.This project falls within the EPSRC Functional Ceramics and Inorganics research area.