Tunable lattice thermal transport properties in high-entropy alloys through manipulating chemical short-range order guided by machine learning

High-entropy alloys (HEAs) exhibit low lattice thermal conductivity due to extreme chemical disorder. However, chemical short-range order (CSRO), a prominent yet not fully resolved structural feature in high-entropy materials, may mitigate the disordered effects. Taking the TaNbMoW HEAs as a representative system, this study systematically investigates the role of CSRO on thermal transport properties using a combination of molecular dynamics (MD), non-equilibrium MD (NEMD), and machine learning (ML) techniques. Our results revealed that CSRO can significantly enhance the lattice thermal conductivity of the considered HEA. To uncover the underlying mechanisms, a novel ML model focusing on local structure and chemical environments was developed to efficiently predict thermal conductivity influenced by varied elemental arrangements. From this analysis, the key local environments responsible for the observed thermal conductivity trends are identified. Further phonon analyses indicate that favorable chemical bonds reduce interactions among phonon modes, leading to a decreased phonon participation ratio and an increase in group velocities and mean free path, especially in the low-frequency regime, which promotes thermal conduction. These findings underscore the pivotal role of CSRO in controlling thermal behavior in HEAs and pave the way for designing materials with tailored thermal properties through precise structural and chemical modulation. © 2025 The Authors