Incoherent Optoelectronic Differentiation with Optimized Multilayer Films

TL;DR

This paper introduces a novel method for optical differentiation using optimized multilayer films with incoherent light, employing neural networks for design, enabling compact, lithography-free, and versatile imaging applications.

Contribution

It presents a new approach combining neural network optimization and multilayer films to perform spatial differentiation with incoherent light, overcoming limitations of previous coherent-only methods.

Findings

Achieved incoherent differentiation with high numerical aperture (0.8).

Demonstrated resolution of 6.2 μm in experiments.

Enabled lithography-free fabrication compatible with existing systems.

Abstract

Fourier-based optical computing operations, such as spatial differentiation, have recently been realized in compact form factors using flat optics. Experimental demonstrations, however, have been limited to coherent light requiring laser illumination and leading to speckle noise and unwanted interference fringes. Here, we demonstrate the use of optimized multilayer films, combined with dual color image subtraction, to realize differentiation with unpolarized incoherent light. Global optimization is achieved by employing neural networks combined with the reconciled level set method to optimize the optical transfer functions of multilayer films at wavelengths of 532 nm and 633 nm. Spatial differentiation is then achieved by subtracting the normalized incoherent images at these two wavelengths. The optimized multilayer films are experimentally demonstrated to achieve incoherent differentiation with a numerical aperture up to 0.8 and a resolution of 6.2 {\mu}m. The use of multilayer films allows for lithography-free fabrication and is easily combined with existing imaging systems opening the door to applications in microscopy, machine vision and other image processing applications.