|
Despite a great progress in photonics over the past few decades, we are nowhere near the level of integration and complexity in photonic systems that would be comparable to those of electronic circuits, which prevents use of photonics in many applications. This lag in integration scale is in big part a result of how we traditionally design photonics: by combining building blocks from a limited library of known designs, and by manual tuning a few parameters. Unfortunately, the resulting photonic circuits are very sensitive to errors in manufacturing and to environmental instabilities, bulky, and often inefficient. We show how a departure from this old fashioned approach can lead to optimal photonic designs that are much better than state of the art on many metrics (smaller, more efficient, more robust). This departure is enabled by development of inverse design approach and computer software which designs photonic systems by searching through all possible combinations of realistic parameters and geometries. We also show how this inverse design approach can enable new functionalities for photonics, including compact particle accelerators on chip which are 10 thousand times smaller than traditional accelerators, chip-to-chip on on-chip optical interconnects with error free terabit per second communication rates, and quantum technologies.
|
