91̽��

Abstract:

Mode-sorting is a procedure that decomposes a light field into a basis of transverse modes, directing each mode into a separate spatial location, allowing the constituent mode intensities to be measured simultaneously. We demonstrate a mode-sorter based on a diffractive optical neural network and show that it is advantageous to include the output detection regions in the trainable set of parameters of that network. This approach outperforms traditional mode-sorting methods, achieving lower crosstalk levels for the same efficiency. For example, in sorting 25 Hermite-Gaussian modes with a 3 plate sorter, at 12 % efficiency, the experimentally measured crosstalk decreases from 37.5 % for fixed detection to 8.7 % for flexible detection.

Abstract:

Large arrays of optical parametric oscillators can solve combinatorial optimization problems with potential quantum advantage but are challenging to realize. Gray et al. developed a photonic chip with this capability and elaborated a method to bring these oscillators into controllable interaction, opening new possibilities in quantum and classical optical computing.

Abstract:

A widely tested approach to overcoming the diffraction limit in microscopy without disturbing the sample relies on substituting widefield sample illumination with a structured light beam. This gives rise to confocal, image scanning, and structured illumination microscopy methods. On the other hand, as shown recently by Tsang and others, subdiffractional resolution at the detection end of the microscope can be achieved by replacing the intensity measurement in the image plane with spatial mode demultiplexing. In this work, we study the combined action of Tsang’s method with image scanning. We experimentally demonstrate superior lateral resolution and enhanced image quality compared to either method alone. This result paves the way for integrating spatial demultiplexing into existing microscopes, contributing to further pushing the boundaries of optical resolution.

Abstract:

Optics is an exciting route for the next generation of computing hardware for machine learning, promising several orders of magnitude enhancement in both computational speed and energy efficiency. However, reaching the full capacity of an optical neural network necessitates the computing be implemented optically not only for inference, but also for training. The primary algorithm for network training is backpropagation, in which the calculation is performed in the order opposite to the information flow for inference. While straightforward in a digital computer, optical implementation of backpropagation has remained elusive, particularly because of the conflicting requirements for the optical element that implements the nonlinear activation function. In this work, we address this challenge for the first time with a surprisingly simple scheme, employing saturable absorbers for the role of activation units. Our approach is adaptable to various analog platforms and materials, and demonstrates the possibility of constructing neural networks entirely reliant on analog optical processes for both training and inference tasks.