Machine Learning Force Field Accelerated Prediction of Low Lattice Thermal Conductivity in Strained WS ₂ Monolayer

Due to its excellent electronic properties, the WS2 monolayer is promising for electronic and optoelectronic devices. However, its high lattice thermal conductivity (κL) limits its application in thermoelectric and thermal insulation. One effective strategy to reduce κL is the application of biaxial tensile strain. In this work, we perform an efficient approach that integrates on-the-fly machine learning force fields with the Peierls-Boltzmann transport equation to accurately and efficiently evaluate κL of the WS2 monolayer. This method reproduces density functional theory results with substantial computational savings. Extending this, we predicted the κL of the WS2 monolayer under 2–20% biaxial tensile strain, beyond which dynamical instability arises. The room temperature κL shows a nonmonotonic decrease with strain, reaching a minimum of 11.18 W/mK at 16%. Thermal transport analysis reveals that reduced phonon velocity and strong phonon–phonon scattering at 16% strain led to such low κL. The 16% strained WS2 monolayer is confirmed to be dynamically, thermally, and mechanically stable for practical applicability. The finding suggests the 16% strained WS2 monolayer as a potential energy-conversion material, offering an efficient approach to explore and engineer thermal transport in 2D materials.