Abstract
The lattice thermal conductivity () is a key property for many potential applications of compounds. Discovery of materials with very low or high remains an experimental challenge due to high costs and time-consuming synthesis procedures. High-throughput computational prescreening is a valuable approach for significantly reducing the set of candidate compounds. In this article, we introduce efficient methods for reliably estimating the bulk for a large number of compounds. The algorithms are based on a combination of machine-learning algorithms, physical insights, and automatic ab initio calculations. We scanned approximately 79,000 half-Heusler entries in the AFLOWLIB.org database. Among the 450 mechanically stable ordered semiconductors identified, we find that spans more than 2 orders of magnitude—a much larger range than that previously thought. is lowest for compounds whose elements in equivalent positions have large atomic radii. We then perform a thorough screening of thermodynamical stability that allows us to reduce the list to 75 systems. We then provide a quantitative estimate of for this selected range of systems. Three semiconductors having are proposed for further experimental study.
- Received 26 August 2013
DOI:https://doi.org/10.1103/PhysRevX.4.011019
This article is available under the terms of the Creative Commons Attribution 3.0 License. Further distribution of this work must maintain attribution to the author(s) and the published article’s title, journal citation, and DOI.
Published by the American Physical Society
Popular Summary
The lattice thermal conductivity is a key property for many potential applications of a compound, such as thermoelectric and heat scavenging technologies. The discovery of materials with very low or high values remains an experimental challenge, however, because of the high costs and time-consuming synthesis procedures. In this article, we fill that need with a new, high-throughput computational approach that provides reliable estimates of the thermal conductivity for a large number of Heusler compounds with a huge potential for different energy and spintronics applications.
Our computational approach combines machine-learning algorithms and physical insights with automated ab initio quantum-mechanical calculations. Based on this highly efficient approach, we have scanned the approximately 80,000 half-Heusler compounds collected in the online integrated repository http://www.aflowlib.org and found that the conductivity of this family of compounds spans more than 2 orders of magnitude, a much larger range than the previously accepted range of 10−20 Wm−1 K−1. Some interesting and practically useful rules for pinpointing ultralow values are also extracted from regression.
Our approach circumvents the computationally demanding task of calculating the heat conductivity of many materials by finding descriptors through machine-learning techniques. Within this methodological framework, different classes of materials can be rapidly tackled for the goal of accelerating materials development.
