Neuromorphic Optoelectronics: Three Paths to Light-Powered AI Hardware

Multiple research groups have recently published findings on optoelectronic devices designed for neuromorphic computing, with each team demonstrating different approaches to integrating light sensing, memory, and processing functions.

Fully Optical Artificial Synapse

A team of researchers has developed an artificial synapse that operates entirely using light. The device uses optical signals both to receive information and to update its internal state. It is constructed from a rare-earth-doped crystal that emits a persistent afterglow after illumination, storing optical information via trapped charge carriers.

Under ultraviolet light, the device exhibits paired-pulse facilitation; under near-infrared light, it shows paired-pulse depression. These behaviors mimic short-term synaptic plasticity without requiring electrical control. The researchers developed a model to explain the device's behavior, which matched experimental measurements.

"The material was combined with a silicon imaging sensor to create a neuromorphic camera prototype that performs in-sensor processing, enhancing contrast and reducing noise."

In a neural network simulation using the device's measured behavior, image recognition accuracy on handwritten digits improved to 95.99% after denoising, compared to approximately 78% without.

The study was published in Advanced Photonics (doi: 10.1117/1.AP.8.4.046001).

Optoelectronic Synaptic Device Using Van der Waals Crystal Design

Researchers at Sungkyunkwan University have developed an optoelectronic synaptic device that mimics biological neurons and synapses. The device was created through a single-step sulfurization process using mixed plasma on van der Waals rhenium selenide (ReSe₂), forming a nano-crystalline layer atop a bulk single-crystalline layer.

Grain boundaries in the nano-crystalline layer confined sulfur ion migration, enabling synaptic weight control. The device demonstrated multi-level conductance modulation, long-term potentiation/depression, paired-pulse facilitation, and tunable short-term to long-term memory transition. Compared to bulk ReSe₂, the nano-crystalline device showed a 34.7% increase in retention efficiency during learning-forgetting-relearning cycles.

In system-level tests, the device performed edge detection on natural images and achieved 96.24% classification accuracy on the CIFAR-10 dataset.

Professor Taesung Kim, corresponding author, stated: "This study demonstrates a single-step method to design the structure of van der Waals crystals for optoelectronic synaptic devices that learn and store information using light."

He added: "By structurally resolving the random nature of ionic migration and interfacial issues inherent in conventional devices, this architecture can be applied to research on next-generation neuromorphic semiconductors and AI hardware."

The research received support from the National Research Foundation of Korea Leader Research Program, the Institute for Basic Science, and the Ministry of Trade, Industry and Energy. Collaboration involved Sungkyunkwan University, the Center for Quantum Nanoscience at IBS, and the Korea Institute of Machinery and Materials. The findings were published online in Advanced Materials on June 3, 2026.

Optoelectronic Device with Programmable Memory Decay

Researchers at Oregon State University have developed a light-sensitive device that integrates sensing, memory, and signal processing in a single phototransistor. The device is inspired by aspects of the human brain.

The device uses light to create stored electrical charges that act as memory. An oxide semiconductor serves as the transistor channel for current, while an organic photosensitive material absorbs light and generates charges. Applying a gate voltage moves trapped charges relative to the transistor channel, strengthening or weakening their influence, allowing memories to persist longer or fade more quickly.

Project leader Larry Cheng of the OSU College of Engineering stated that technology functioning more like the human brain could enable artificial intelligence systems to work faster while consuming less electricity.

Cheng stated the device introduces a new hardware capability that may enable more efficient processing of information directly at the sensor level.

The research was published in Advanced Functional Materials and supported by the National Science Foundation. Collaborators include researchers from OSU College of Engineering and OSU College of Science.