Ternary semiconductors with tunable band gaps from machine-learning and crystal structure prediction

Computational tools are being employed at an increasing rate to discover and design novel materials with tailored properties to tackle global environmental challenges. Besides the two most common approaches based on high-throughput density functional theory (DFT) calculations and crystal structure prediction schemes, novel methods based on materials informatics and machine learning (ML) models have recently emerged to assist the search for materials with improved properties in industrially relevant applications.

Here, we present a computational investigation of a series of ternary X₄Y₂Z compounds with X={Mg, Ca, Sr, Ba}, Y={P, As, Sb, Bi}, and Z={S, Se, Te}, which we identify by a combined search using a machine learning model and the minima hopping crystal structure prediction method. Accroding to our ab initio results, these compounds are thermodynamically stable and semiconducting with band gaps in the range of 0.3 to 1.8 eV, well suited for various energy applications. We show that several candidate compounds exhibit good photo absorption in the visible range, and excellent thermoelectric performance due to high power factors and extremely low lattice thermal conductivities.