Room temperature exciton-polariton neural network with perovskite crystal

TL;DR

This paper demonstrates the first room-temperature exciton-polariton neural network using perovskite crystals, enabling practical optical neuromorphic computing with high efficiency and speed.

Contribution

It introduces a novel room-temperature polariton neural network based on perovskite waveguides, overcoming previous cryogenic limitations.

Findings

Successfully performed binary classification tasks.

Achieved object detection with polariton neural network.

Validated high-speed, energy-efficient optical computing capabilities.

Abstract

Limitations of electronics have stimulated the search for novel unconventional computing platforms that enable energy-efficient and ultra-fast information processing. Among various systems, exciton-polaritons stand out as promising candidates for the realization of optical neuromorphic devices. This is due to their unique hybrid light-matter properties, resulting in strong optical nonlinearity and excellent transport capabilities. However, previous implementations of polariton neural networks have been restricted to cryogenic temperatures, limiting their practical applications. In this work, using non-equillibrium Bose-Einstein condensation in a monocrystalline perovskite waveguide, we demonstrate the first room-temperature exciton-polariton neural network. Its performance is verified in various machine learning tasks, including binary classification, and object detection. Our result is a crucial milestone in the development of practical applications of polariton neural networks and provides new perspectives for optical computing accelerators based on perovskites.