Hopfield-like open channel retrieval for disordered optical media

TL;DR

This paper introduces a novel, reference-free method to retrieve open channels in disordered optical media by mapping interference terms into a bi-dyadic matrix, enabling high-efficiency light transmission and fast laser scanning.

Contribution

It presents the first complete, reference-less approach to retrieve open channels using a Hopfield-like neural network analogy in disordered optical systems.

Findings

Achieves nearly 100% of the theoretical intensity transmission.

Demonstrates high-speed laser scanning through disordered media.

Validates the neural network analogy for optical channel retrieval.

Abstract

The measurement of the optical Transmission Matrix (TM) enables to access "open channels": input patterns, specific to each scattering structure, capable to deliver very high transmission. Various approaches, based either on multiple interferometric measurements or on systematic random testing of incident wavefronts, enable to estimate the inputs required to excite these open channels. Here, we provide for the first time an approach enabling the complete and reference-less retrieval of the open channels. It is based on the full mapping all the pairwise interference terms resulting from all the input modes couples. We show that these interference terms are organized into a bi-dyadic coupling matrix whose eigenvalues enables to access the open channel. A disordered optical system, is thus behaving exactly like an Hopfield neural network, where a specific input vector (an eigenvalue of the neurons' coupling matrix) enables to retrieve a specific memory pattern. The proposed Hopfield like open-channel-retrieval approach, enables to reach almost 100$\%$ of the theoretically expected value of the Intensity. Moreover employing a digital micromirror device to modulate light, we demonstrate high speed laser scanning at the back of a disordered medium.