3D Monte Carlo Device Simulation Integrating Machine Learning and Physical Laws

Novel device structures are being introduced to overcome short-channel effects in ultra-small MOSFETs. Simulating such devices requires the development of next-generation simulators that go beyond conventional empirical approaches. The Monte Carlo (MC) method presents a promising solution for this purpose; however, traditional MC simulations often require repeated spatial discretization and continuation, which is computationally inefficient. This study proposes a machine learning-based approach for predicting the electric field distribution at arbitrary electron positions. To improve generalization across untrained device structures and varying bias conditions, the model incorporates physical law-based constraints, with the aim of building a more robust and reliable simulation framework.