A high-throughput descriptor for prediction of lattice thermal conductivity of half-Heusler compounds

• Published in: Journal of physics. D, Applied physics Applied physics, 2020-07, Vol.53 (31), p.315301
• Main Authors: Liu, Jianghui, Han, Shihao, Cao, Guohua, Zhou, Zizhen, Sheng, Caiyu, Liu, Huijun
• Format: Article
• Language: English
• Subjects: big-data analytics; half-Heusler compounds; high-throughput prediction; lattice thermal conductivity
• ISSN: 0022-3727; 1361-6463
• DOI: 10.1088/1361-6463/ab898e

Lattice thermal conductivity is a key property of materials for applications ranging from thermal barrier coatings to thermoelectric conversion. In principle, the thermal conductivity can be accurately predicted by solving the phonon Boltzmann transport equation, which is however very time-consuming, especially for complex systems with large unit cell. Here, using half-Heusler compounds as prototypical examples, we apply a compressed-sensing approach to rapidly evaluate the lattice thermal conductivity with very good accuracy, as realized by a physically interpretable descriptor. Beyond the initial 86 training data, the descriptor is employed to predict the thermal conductivities of 75 half- and 15 full-Heusler compounds, which shows good agreement with explicit first-principles results. Moreover, the descriptor is further optimized by including only the fundamental properties of the constituent atoms, which could accelerate materials discovery with desirable thermal conductivity.