Precision optics for molecular fingerprinting and AI computing

The advance of coherent control in optical fields has enabled unprecedented precision and sensitivity in optical measurements. In the past two decades, laser frequency combs revolutionized the field of optical metrology. A comb-enabled technique, dual-comb interferometry is emerging as a broadband, high-resolution, rapid spectroscopic solution. However, preserving the interferometric phase stability was a challenge due to phase jitters. Our work on acousto-optic feed-forward stabilization has enabled a full-stabilized dual-comb system from mid-infrared, near-infrared, visible, to ultra-violet spectral coverage at the quantum noise level. The coherence also allows photon statistics at < 1 photon per pulse. In the next part, I will talk about precision control in photonic circuits for efficient AI computing (matrix operation). This is motivated by the need for processing power in the rapid expansion of data volume from artificial intelligence (AI), internet of things (IoT) and 5G/6G mobile networks. Based on detector-based photoelectric multiplication and time multiplexing, our architecture indicates a path towards femtojoule per operation (fJ/OP) and with high computing density, which offers a new solution to tackle the fundamental bottlenecks in AI processing.

Speaker: Zaijun Chen, University of Southern California