Unveiling the impact of intrinsic defects on thermal conductivity in CuInTe₂ using neural network potential

Defect engineering is a powerful strategy for tailoring the thermoelectric performance of materials. While experiments indicate that intrinsic point defects influence the thermal properties of CuInTe₂, the underlying physical mechanisms remain unclear. In this study, we systematically investigate the effects of intrinsic defects and higher-order phonon interactions on the lattice thermal conductivity κ_L of CuInTe₂ using neural network potentials combined with the phonon Boltzmann transport equation. Our results reveal that V_Cu vacancies significantly reduce the κ_L of CuInTe₂, whereas Cu_In antisite defects conversely increase it. Further analysis demonstrates that the κ_L reduction in V_Cu-containing CuInTe₂ is primarily due to decreased phonon group velocity, as well as increased phonon scattering phase space caused by the upward shift of low-frequency optical branches. Meanwhile, the unexpected κ_L enhancement induced by Cu_In antisites is attributed to reduced phonon linewidths resulting from a decreased lattice distortion coefficient η. Furthermore, including four-phonon (4ph) scattering significantly reduces κ_L in all cases, with reductions of 21.0%, 26.7%, and 20.1% for pristine CuInTe₂, V_Cu-containing, and Cu_In-containing CuInTe₂, respectively, highlighting the substantial impact of 4ph scattering on their phonon thermal transport. This work provides fundamental insights into the roles of intrinsic point defects and higher-order phonon interactions in the thermal transport properties of CuInTe₂.