# All-optical temporal integration mediated by subwavelength heat antennas

## TL;DR

Researchers developed an all-optical computing system using nanoantennas to process large vectors for AI, enabling fast signal integration and nonlinear functions entirely with light.

## Contribution

The system introduces a novel all-optical neuromorphic computing framework using subwavelength heat antennas for time integration and nonlinear activation.

## Key findings

- The system processes input vectors with over 250,000 elements using time division multiplexing.
- Thermal time dynamics enable integration of 50 GHz signals and programmable nonlinear activation functions in the optical domain.
- The platform demonstrates a unified framework for large-scale photonic computing suitable for AI workloads.

## 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 (50 GHz) 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.

The authors demonstrate the use of standing waves to generate localised heat using nanoantennae in integrated photonics. They utilise this heat to mediate all-photonic integration of 50 GHz optical signals, showing experimentally, the potential for achieving programmable nonlinear activation functions. This is a step towards full optical encoding in integrated photonic processors.