Researchers propose neuromorphic computing with optically driven nonlinear fluid dynamics

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

Researchers at the University of California, San Diego recently proposed a novel paradigm in which liquids, which do not normally interact strongly with light at the microscale or nanoscale, support a significant nonlinear response to optical fields. As reported in advanced photonicsThe researchers predict a substantial light-liquid interaction effect through a proposed nanoscale gold patch that functions as an optical heater and generates thickness changes in a liquid film covering the waveguide.

The liquid film works as a optical memory. This is how it works: light in the waveguide affects the geometry of the liquid surface, while changes in the shape of the liquid surface affect the properties of the optical mode in the waveguide, constituting a coupling mutual between the optical mode and the liquid film. . Importantly, as the geometry of the liquid changes, the optical mode properties experience a nonlinear response; after the optical pulse stops, the magnitude of the liquid film deformation indicates the strength of the previous optical pulse.

Surprisingly, unlike traditional computational approaches, nonlinear response and memory reside in the same spatial region, suggesting the realization of a compact (beyond von-Neumann) architecture where memory and computational unit occupy the same space. same space. The researchers show that the combination of memory and nonlinearity enables the possibility of “repository computing” capable of performing digital and analog tasks, such as nonlinear logic gates and recognition of handwritten images.

His model also exploits another important liquid feature: non locality. This allows them to predict computational enhancement that is simply not possible on solid-state material platforms with limited non-local spatial scale. Despite the non-locality, the model does not reach the levels of modern ones based on solid-state optics. reservoir computing However, the work presents a clear roadmap for future experimental work with the aim of validating the predicted effects and exploring the intricate coupling mechanisms of various physical processes in a liquid environment for computation.

Using multiphysics simulations to investigate the coupling between light, fluid dynamics, heat transport, and surface tension effects, the researchers predict a family of new nonlinear and nonlocal optical effects. They go a step further by indicating how they can be used to realize versatile and unconventional computing platforms. Leveraging a mature silicon photonics platform, they suggest improvements to next-generation liquid-assisted computing platforms by around five orders of magnitude in space and at least two orders of magnitude in speed.