A Review on Machine-Learning-Assisted Prediction of Lattice Thermal Conductivity and Anomalous Nernst Conductivity in Heusler Alloys

Abstract Heusler alloys are versatile intermetallic materials in which crystal symmetry, chemical substitution, magnetic order, spin–orbit coupling and electronic topology can be tuned, making them attractive for thermoelectric conversion, thermal management and transverse thermoelectric devices. This article presents a physics-informed machine-learning framework for predicting and interpreting lattice thermal conductivity in Heusler alloys, particularly half-Heusler thermoelectric and magnetic full-Heuslers. The framework employs physically meaningful descriptors including atomic mass, lattice parameter, cohesive-energy-related quantities, elastic properties, valence-electron count and structural information. Based on the phonon Boltzmann transport picture, lattice thermal conductivity depends on phonon heat capacity, group velocity and relaxation time. Reported room-temperature values for TiNiSn, ZrNiSn and HfNiSn are approximately 15.4, 13.3 and 15.8 Wm−1K−1, respectively, while alloying, disorder and microstructural engineering can reduce ZrNiSn-based thermal conductivity to approximately 3–6 Wm−1K−1. The article also examines the anomalous Nernst effect (ANE) in magnetic Heuslers. Co2MnGa exhibits anomalous Nernst thermoelectric power of approximately 6.0 μVK−1 at 300 K and 6.6 μVK−1 at 340 K, substantially exceeding conventional magnetization-scaling expectations. This large response is associated with Berry curvature near the Fermi energy, including gapped nodal lines and Weyl points. Recent Co2MnAl0.69Si0.31 results further demonstrate enhanced transverse thermoelectricity, with anomalous Nernst conductivity of 1.46 Am−1K−1 at 300 K.