Multi-dimensional optical computing.

The development of deep neural networks is witnessing fast growth in network size, which requires novel hardware computing platforms. Optical computing has been a potential candidate for next-generation computing systems. Specifically, wavelength-division multiplexing (WDM) has been adopted in optical computing architecture to increase the computation bandwidth. Although existing WDM architectures have shown promise, they face challenges in the integration of light sources and further increase of the computing bandwidth. We introduce a mode-division multiplexing (MDM) strategy, offering what we believe to be a new degree of freedom in optical computing based on the micro-ring resonator platform. We propose an MDM approach and a multi-dimensional architecture that augments WDM with MDM to enhance channel capacity for computation. We design and experimentally demonstrate key components of the proposed architectures, including a multimode beam splitter, a thermo-optical tuner for the high-order mode, and a multimode waveguide bend. A proof-of-principle matrix multiplexing system, fabricated in a foundry and working for both MDM and MDM-WDM computing, is demonstrated.