All-optical temporal integration mediated by subwavelength heat antennas

TL;DR

This paper presents an all-optical neuromorphic computing system capable of processing extremely large input vectors using thermo-optic modulation and subwavelength heat antennas, advancing AI-compatible photonic computing.

Contribution

It introduces a novel all-optical computing platform that handles high-dimensional vectors and performs non-linear activation entirely in the optical domain, surpassing previous low-dimensional implementations.

Findings

Processed input vectors over 250,000 elements

Enabled simultaneous ultra-fast signal integration and non-linear activation

Utilized thermo-optic modulation with titanium nano-antennas

Abstract

Optical computing systems deliver unrivalled processing speeds for scalar operations. Yet, integrated implementations have been constrained to low-dimensional tensor operations that fall short of the vector dimensions required for modern artificial intelligence. We demonstrate an all-optical neuromorphic computing system based on time division multiplexing, capable of processing input vectors exceeding 250,000 elements within a unified framework. The platform harnesses optically driven thermo-optic modulation in standing wave optical fields, with titanium nano-antennas functioning as wavelength-selective absorbers. Counterintuitively, the thermal time dynamics of the system enable simultaneous time integration of ultra-fast (50GHz) signals and the application of programmable, non-linear activation functions, entirely within the optical domain. This unified framework constitutes a leap towards large-scale photonic computing that satisfies the dimensional requirements of AI workloads.