Researchers propose neural computing with optically driven nonlinear fluid dynamics

الحوسبة العصبية مع ديناميكيات السوائل غير الخطية المدفوعة بصريًا

The simulation result of light affecting the geometry of the liquid, which in turn affects the reflection and transmission properties of the optical mode, thus forming a bidirectional interaction mechanism between light and liquid. The degree of distortion acts as an optical memory that allows storing the amount of energy for the previous optical pulse and using fluid dynamics to influence the subsequent optical pulse in the same operating region, thus forming a structure where the memory is part of the computation. Credit: Gao et al. , Advanced Photonics (2022). DOI: 10.1117/1.AP.4.4.046005

Sunlight shimmering on water evokes the rich phenomena of liquid-light interaction, spanning across spatial and temporal scales. While fluid dynamics has fascinated researchers for decades, the advent of neural computing has led to major efforts to develop new, unconventional computational schemes based on recurrent neural networks, which are essential to support a wide range of modern technological applications, such as pattern recognition and autonomous driving. . Since biological neurons also depend on a fluid environment, convergence can be achieved by bringing nanoscale nonlinear fluid dynamics to neuronal computing.

Researchers from the University of California, San Diego recently proposed a new model in which liquids, which do not normally interact strongly with light at the micro or nanoscale, support the important nonlinear response of optical fields. as stated in Advanced Photonicsthe researchers predicted a significant effect of the light-liquid interaction through a proposed nanoscale gold patch that acts as a photoheater and generates thickness changes in liquid film Covering the waveguide.

The liquid film acts as a file optical memory. Here’s how it works: Light in the waveguide affects the surface geometry of the liquid, while changes in the shape of the liquid’s surface affect the properties of the optical mode in the waveguide, forming a reciprocal coupling between the optical mode and the liquid film. Importantly, as the fluid geometry changes, the optical mode properties undergo a nonlinear response; After the optical pulse stops, the magnitude of the liquid film deformation indicates the strength of the previous optical pulse.

Nonlinear phase change in a single waveguide with a gold patch as the heat source. Credit: Gao et al. , Advanced Photonics (2022). DOI: 10.1117/1.AP.4.4.046005

Remarkably, in contrast to traditional computational methods, the nonlinear response and memory are located in the same spatial region, indicating the realization of a compact structure (beyond von Neumann) where memory and computational unit occupy the same space. The researchers explained that the combination of memory and nonlinearity allows for “backup computing” capable of performing both digital and analog tasks, such as nonlinear logic gates and handwriting image recognition.

Their model also exploits another important liquid advantage: UN-signed. This enables them to predict computational optimization that is simply not possible in solid material platforms with a non-local finite spatial scale. Although not centered, the model does not fully achieve the levels of modern solid-state optics tank computing However, the work nevertheless provides a clear roadmap for future experimental work aimed at validating the expected effects and exploring the complex coupling mechanisms of different physical processes in a computational fluid environment.

Use of multi-physics simulation to examine the coupling of light, fluid dynamics, heat transfer, and surface tension effects, the researchers anticipate a new set of nonlinear, non-local optical effects. They go further by pointing out how they can be used to achieve versatile and unconventional computing platforms. Taking advantage of the mature silicon photonics platform, they propose improvements to the latest fluid-assisted computing platforms by about five orders of magnitude in space and at least two orders of magnitude in terms of speed.

Intrinsic optical nonlinearity and InSe . carrier dynamics

more information:
Chengkuan Gao et al, Thin liquid film as a nonlinear optical medium and memory element in an integrated optical fluid reservoir computer, Advanced Photonics (2022). DOI: 10.1117/1.AP.4.4.046005

the quote: Researchers propose neural computing with optically driven nonlinear fluid dynamics (2022, July 25) Retrieved on July 25, 2022 from

This document is subject to copyright. Notwithstanding any fair dealing for the purpose of private study or research, no part may be reproduced without written permission. The content is provided for informational purposes only.