A DFT and Machine Learning-Assisted Study on the Lattice Thermal Conductivity of LiCdSb for Thermoelectric Applications

Abstract

By using first-principles density functional theory (DFT) and the Boltzmann transport equation, we have calculated the corresponding electronic and thermoelectric properties of LiCdSb. For calculating electron transport properties, accurate band-structure estimation is crucial. Hence, for the precise band gap calculation, we have implemented a hybrid functional HSE06, which is widely known for its high accuracy. To evaluate the thermoelectric performance of a material, the calculation of lattice thermal conductivity (Kl) is a key parameter. However, from a theoretical perspective, the calculation of lattice thermal conductivity is very complex and demands huge computational resources. Therefore, in this work, we have opted for an alternative method of machine-learning interatomic potentials (MLIPs) for the calculation of Kl. Our result of Kl=0.24 Wm^-1K^-1 at room temperature is in qualitative agreement with the available theoretical and experimental data. The figure of merit (ZT) with Kl estimated from Slack+TDEC ZT is 0.18 at 300 K, and machine learning (ML) models ZT is 0.17 at 300K, combining with HSE06-based electronic transport properties agreed well with the available experimentally reported value of ZT is 0.10 at 300K. However, we report the ZT value well above the benchmark value of 1 beyond 600K. The ZT value exceeding 1 at higher temperatures makes LiCdSb a promising material for high-temperature energy conversion.