Phonon transport and thermoelectric properties of Heusler-like Li2CuSb enabled by machine-learning interatomic potentials.

Li2CuSb is a Heusler-like ternary compound that combines potentially favorable electronic transport characteristics with intrinsically low lattice thermal conductivity, yet its phonon transport mechanism and thermoelectric performance remain insufficiently understood. In this work, we systematically investigate the structural, phononic, and thermoelectric properties of Li2CuSb by combining first-principles calculations, machine-learning interatomic potentials, self-consistent phonon theory, and Boltzmann transport analysis. The results show that Li2CuSb crystallizes in the cubic F4̄3m structure and exhibits mixed ionic-metallic bonding characteristics. HSE06+SOC calculations reveal a direct band gap of approximately 0.809 eV, indicating its narrow-gap semiconducting nature. The trained moment tensor potential well reproduces the DFT phonon dispersion, supporting its use in finite-temperature phonon and thermal transport calculations. Li2CuSb exhibits pronounced phonon renormalization while maintaining dynamical stability over the investigated temperature range. Its lattice thermal conductivity decreases continuously with increasing temperature and is further significantly reduced when four-phonon scattering is included, reaching 1.08 W m-1 K-1 at 500 K within the SCP+3,4ph framework. Detailed analysis reveals that the low lattice thermal conductivity originates from the combined effects of low phonon group velocities, pronounced anharmonicity, abundant scattering phase space, and strong three-phonon/four-phonon scattering processes. On the electronic side, p-type doping exhibits superior thermoelectric performance at elevated temperatures, reaching a maximum power factor of 11.378 mW m-1 K-2 at 900 K and a carrier concentration of 4 × 1020 cm-3. The maximum thermoelectric figure of merit (ZT) reaches 3.02 under p-type doping at 900 K. These findings suggest that Li2CuSb is a Heusler-like thermoelectric system with low calculated lattice thermal conductivity and favorable p-type transport characteristics at elevated temperatures, and demonstrate that combining machine-learning interatomic potentials with high-order anharmonic phonon transport analysis provides an effective route for investigating complex Heusler-like thermoelectric systems.