Structure maps and crystal growth of ternary ThCr2Si2-type rare-earth transition-metal silicides and germanides

To classify ternary rare-earth silicides REM 2 Si 2 and germanides REM 2 Ge 2 adopting the ThCr 2 Si 2 -type structure, a two-dimensional map based on radius ratios and valence electron counts was developed. This map suggested that the transition metal M plays a dominant role, which was confirmed independently by applying a machine learning algorithm called the sure independence screening and sparsifying operator (SISSO) method. In this way, a simple one-dimensional descriptor based solely on properties of the metal component M was identified in which ThCr 2 Si 2 -type phases are more likely to be formed if this descriptor meets a minimum threshold of 1.68 for silicides and 2.27 for germanides. Although arc-melting is typically used to prepare these compounds, it does not usually afford suitably sized crystals for further characterization. Flux growth of ternary germanides was investigated, with the use of indium yielding crystals of RE Co 2 Ge 2 ( RE = Ce, Eu, Yb) and other compounds. Two-dimensional maps were developed to classify ternary rare-earth silicides and germanides with ThCr2Si2-type structure, but simpler one-dimensional maps were also sought by machine learning. • Models were sought to classify ThCr 2 Si 2 -type silicides and germanides. • Radius ratios and valence electron counts are effective descriptors. • A simpler one-dimensional descriptor was developed by machine learning. • Single crystals of ternary germanides were grown in indium flux.