Feedforward photonic optical circuits or meshes of programmable interferometers have recently been successfully demonstrated for potential in commercial applications ranging from sensing to machine learning to quantum computation. In this talk, I describe how to program and calibrate any feedforward circuit in presence of error via self-configuration. Incorporating this programming approach, I compare phase shift tolerance, loss tolerance, and fabrication error sensitivity as a function of circuit size and topology given an interferometer design. Our simulations and theory show that for a fixed circuit size, interfering modes in widely spaced (nonlocally interacting) waveguides increases error tolerance by orders of magnitude compared to interfering only modes in neighboring waveguides. Our results also explain why, in signal processing and sensing applications, this error-tolerant class of circuits can potentially scale fast Fourier transform, permutations, and low-rank matrix-vector products to thousands of input modes with minimal reduction in systematic error. Finally, I will present preliminary experimental results on a 4x4 mode photonic circuit to demonstrate a minimal example of programmable optics.

Sunil Pai received the B.S. degree in physics and the M.S. degree in computer science from Stanford University in 2015 and 2016, respectively. He is currently a Ph.D. Student in Electrical Engineering at Stanford University studying photonic networks, coadvised by Olav Solgaard, David A.B. Miller, and Shanhui Fan. His research interests include machine learning, photonics, and quantum optics.