Physics-Informed Semiconductor Digital Twins

Developing a new semiconductor technology is a complex and iterative process that spans material synthesis, device fabrication, electrical characterization, physics-based modeling, and circuit design. Traditionally, each stage has been optimized independently through extensive trial-and-error experiments, making technology development slow and resource intensive. Our vision is to create semiconductor digital twins—virtual representations of semiconductor technologies that continuously learn from experiments, predict device behavior, and guide the next generation of discoveries. By tightly integrating physics, data, and artificial intelligence, we aim to transform semiconductor research from an empirical process into a predictive science.

Materials Discovery

Can artificial intelligence accelerate the discovery of new semiconductor materials? We develop closed-loop optimization frameworks that combine materials synthesis, characterization, and machine learning to efficiently navigate vast composition and process spaces. Using Bayesian optimization, active learning, and uncertainty quantification, our research identifies optimal growth conditions and material compositions while dramatically reducing the number of experiments required to develop emerging semiconductor technologies.

Device Digital Twins

How can we accurately predict the behavior of semiconductor devices before they are fabricated? Our research develops physics-informed digital twins that integrate electrical characterization, TCAD, compact modeling, reliability physics, and artificial intelligence into unified predictive models. These digital twins capture the relationships between material properties, fabrication processes, device performance, and long-term reliability, enabling rapid evaluation of new device architectures while preserving physical interpretability.

Technology Co-Optimization

Future semiconductor innovation requires simultaneous optimization across the entire technology stack rather than treating materials, devices, and circuits independently. Our research develops digital twin frameworks that connect materials, fabrication, device physics, reliability, and circuit-level performance into a unified optimization platform. By capturing these cross-scale interactions, we enable rapid exploration of complex design spaces, accelerate technology development, and establish a predictive pathway from material discovery to functional electronic systems.