Bidirectional All-Optical Synapses for Neuromorphic Computing and Vision

Area of Science:

• Optoelectronics
• Materials Science
• Neuroscience

Background:

• All-optical synapses (AOS) are crucial for optical computing and neuromorphic imaging.
• Current AOS often lack bidirectional weight modulation, limiting their functionality.
• Achieving both excitatory and inhibitory outputs optically remains a significant challenge.

Purpose of the Study:

• To develop a novel strategy for bidirectional neuroplasticity in all-optical synapses.
• To overcome the limitations of unidirectional weight modulation in existing AOS.
• To create a phosphorescent and photothermal carbon dot hybrid (CDH) neurotransmitter for advanced optical signal processing.

Main Methods:

• A dual-photon-mediated strategy was employed using ultraviolet and infrared light.
• Controllable exciton release and superposition of phosphorescent emission were utilized.
• A phosphorescent and photothermal carbon dot hybrid (CDH) was engineered as a neurotransmitter.

Main Results:

• Demonstrated bidirectional neuroplasticity (potentiation and depression) within the CDH neurotransmitter.
• Achieved controllable optical logic operations and neurovision capabilities.
• Realized real-time tracking and recognition of light objects with 97% accuracy, including trajectory, velocity, and direction.

Conclusions:

• The developed CDH neurotransmitter enables all-optical regulation for innovative phosphorescent paradigms.
• This work advances all-optical computing and neuromorphic imaging by overcoming unidirectional modulation limitations.
• The dual-photon strategy offers a new pathway for designing sophisticated optical neuro-mimetic devices.