Zaijun Chen
Zaijun Chen is a Professor in the Electrical Engineering & Computer Sciences Department at UC Berkeley.
Project Description
Wafer-Scale Heterogeneous Integration of Lithium-Niobate-on-Silicon Optoelectronics for Ultralow-Energy AI Computing
As artificial intelligence (AI), autonomous systems, and cloud-edge computing grow in complexity and scale, the demand for energy-efficient, high-speed data processing is outpacing the capabilities of traditional electronic systems. Silicon-based electronics face physical and architectural bottlenecks, including capacitive power dissipation and data movement limitations. This project proposes a fundamentally optical solution: leveraging wafer-scale heterogeneous integration (HI) of thin-film lithium niobate (TFLN) with silicon photonics to enable ultra-lowenergy, high-bandwidth optical computing for AI workloads.
At the heart of the team’s innovation is the use of light for data modulation, transport, and computation. By exploiting the ultrafast electro-optic properties of TFLN, they demonstrate optical modulators exceeding 50 GHz bandwidth and energy efficiencies below 10 fJ/symbol. Their integrated photonic circuits harness space-time-wavelength multiplexing to achieve high-throughput, parallel processing directly in the optical domain. Co-packaged optical digital-to-analog converters (ODACs) further enable real-time optical signal generation and conversion, reducing reliance on inefficient electronic memory access.
The approach is scalable, CMOS-compatible, and directly aligns with existing semiconductor manufacturing workflows, positioning it for rapid commercialization. Beneficiaries span AI hardware developers, photonic chip manufacturers, biomedical imaging, and academic research communities. The team’s commercialization roadmap advances from proof-of-concept optical AI processors (TRL 3–4) to full-system optical computing modules validated in real-world applications (TRL 7–8) over three years.
Support from the Bakar Fellows Program will accelerate photonic integration, system co-design, and workforce development. This project pioneers a new computing paradigm grounded in integrated optics, establishing a scalable path toward photonic AI accelerators that combine speed, efficiency, and manufacturability.