Machine learning interatomic potential prediction of thermal and mechanical characteristics of Si nanosheet transistors for low power applications

This work investigates the thermal conductivity and elastic moduli of silicon nanosheets of varying thicknesses using machine learning interatomic potentials (MLIPs). The training dataset was effectively generated. The interatomic forces and energy of each sample of the training dataset were calculated using the density functional theory (DFT). The Moment Tensor Potential (MTP) was utilized to build the MLIP model. The findings revealed that for nanosheets thinner than 6 nm, the thermal conductivity decreased to approximately 7% of the bulk value, whereas certain elastic constants decreased to approximately 3% of the bulk values. This study also examined how these reduced parameters affect the performance of nanosheet field-effect transistors (NS-FETs). This analysis is conducted using fully calibrated TCAD device simulations to evaluate the technological implications.