High-Throughput First-Principles Calculations and Machine Learning of Grain Boundary Segregation in Metals

Abstract

The segregation of solute elements to defects in metals plays a fundamental role for microstructure evolution and the material performance. However, the available computational data are scattered and inconsistent due to the use of different simulation parameters and methods. A high-throughput study is presented on grain boundary and surface segregation together with their effect on grain boundary embrittlement using a consistent first-principles methodology. The data are evaluated for most technologically relevant metals including Al, Cu, Fe, Mg, Mo, Nb, Ni, Ta, Ti, and W with the majority of the elements from the periodic table treated as segregating elements. Trends among the solute elements are analyzed and explained in terms of phenomenological models and the computed data are compared to the available literature data. The computed first-principles data are used for a machine learning investigation, showing the capabilities for extrapolation from first-principles calculation to the whole periodic table of solutes. The present work allows for comprehensive screening of new alloys with improved interface properties.