PALTO: Physics-Informed Active Learning for Tri-Gate FinFET Design Optimization for Vertical Power Delivery
Researchers have developed PALTO, a physics-informed active learning framework to optimize the design of Gallium Nitride (GaN) tri-gate FinFETs for vertical power delivery systems. This machine learning approach significantly accelerates the design process compared to traditional TCAD methods by intelligently guiding simulations. The framework identified two optimized devices, with one outperforming industrial benchmarks by achieving 3.3A at 0.49 ohm on-resistance and demonstrating 2x greater switching efficiency. AI
IMPACT Accelerates discovery of optimized semiconductor devices, potentially leading to more efficient power electronics.