Optofluidic memory and self-induced nonlinear optical phase change for reservoir computing in silicon photonics

TL;DR

This paper demonstrates a novel optofluidic silicon photonics system that uses self-induced phase changes in a thin liquid film for neuromorphic computing, achieving high nonlinearity and miniaturization.

Contribution

It introduces a new physical mechanism based on liquid film deformation for optical memory and computation, enabling nanoscale reservoir computing in silicon photonics.

Findings

Nonlinear effect magnitude exceeds traditional thermo-optical effects by over ten times.

Supports optically-driven periodic deformation at several kHz.

Achieves reservoir computing in a region five orders of magnitude smaller than existing systems.

Abstract

Implementing optical-based memory and utilizing it for computation on the nanoscale remains an attractive but still a challenging task. While significant progress was achieved in nanophotonics, allowing to explore nonlinear optical effects and employ light-matter interaction to realize non-conventional memory and computation capabilities, light-liquid interaction was not considered so far as a potential physical mechanism to achieve computation on nanoscale. Here, we experimentally demonstrate self-induced phase change effect which relies on the coupling between geometry changes of thin liquid film to optical properties of photonic modes, and then employ it for neuromorphic computing. In particular, we employ optofluidic Silicon Photonics system in order to demonstrate thermocapillary-based deformation of thin liquid film capable of operating both as a nonlinear actuator and memory element, both residing at the same compact spatial region, thus realizing beyond von Neumann computational architecture. Our experimental results indicate that the magnitude of the nonlinear effect is more than one order of magnitude higher compared to the more traditional heat-based thermo-optical effect, capable to support optically-driven periodic deformation of frequencies of several kHz, and furthermore allows to implement Reservoir Computing at spatial region which is approximately five orders of magnitude smaller compared to state-of-the-art experimental liquid-based systems.