Density functional theory and machine learning guided search for RE <sub>2</sub> Si <sub>2</sub> O <sub>7</sub> with targeted coefficient of thermal expansion
Mukil V. Ayyasamy, Jeroen A. Deijkers, H.N.G. Wadley, Prasanna V. Balachandran
Abstract
Abstract Density functional theory (DFT) calculations and machine learning (ML) methods are used to establish a relationship between the crystal structures of rare‐earth (RE) disilicates (RE 2 Si 2 O 7 ) and their coefficient of thermal expansion (CTE). The DFT total energy data predict the presence of several energetically competing crystal structures, which is rationalized as one of the reasons for observing polymorphism. An ensemble of support vector regression models is trained to rapidly predict the CTE as a function of RE 2 Si 2 O 7 crystal chemistry. Experiments subsequently validated the structure and CTE predictions for Sm 2 Si 2 O 7 .
Topics & Concepts
Thermal expansionDensity functional theoryCrystal structureCrystal (programming language)Materials scienceSupport vector machineThermodynamicsCondensed matter physicsChemistryComputational chemistryPhysicsComputer scienceCrystallographyMachine learningProgramming languageNuclear materials and radiation effectsCrystal Structures and PropertiesAdvanced Condensed Matter Physics