Combining state-of-the-art optimization and machine learning techniques with high speed electromagnetic solvers offers a new approach to “inverse” design and implement classical and quantum photonic circuits with superior properties, including robustness to errors in fabrication and environment, compact footprints, novel functionalities, and high efficiencies. We illustrate this with a number of demonstrated devices in silicon, diamond, and silicon carbide, including wavelength and polarization splitters and converters, power splitters, couplers, nonlinear optical isolators, on chip laser driven particle accelerators, and efficient quantum emitter-photon interfaces.
|