An Optoelectronic Artificial Synapse Based on CuIn_0.7Ga_0.3Se₂/ Al-doped ZnO p-n Heterojunction for Bioinspired...

Area of Science:

• Materials Science
• Neuroscience
• Computer Science

Background:

• The von Neumann architecture presents limitations for artificial intelligence (AI) development.
• Memristors offer a promising pathway to overcome these architectural constraints.
• Optoelectronic synapses are crucial for developing efficient neuromorphic computing systems.

Purpose of the Study:

• To engineer an optoelectronic synapse utilizing a novel heterojunction material.
• To evaluate the synaptic behaviors and learning capabilities of the fabricated device.
• To demonstrate the device's potential in AI applications, specifically image recognition.

Main Methods:

• Fabrication of a CuIn_0.7Ga_0.3Se₂ (CIGS)/Al-doped ZnO (AZO) p-n heterojunction via radio-frequency magnetron sputtering.
• Integration into an Au/CIGS/AZO/ITO p-n heterojunction artificial synapse configuration.
• Simulation of synaptic plasticity and learning-forgetting-relearning processes.
• Implementation within a convolutional neural network (CNN) for dataset recognition.

Main Results:

• The Au/CIGS/AZO/ITO heterojunction successfully emulated synaptic behaviors, including learning and forgetting.
• Achieved high recognition accuracies: 97.36% on MNIST and 83% on Fashion-MNIST datasets using a CNN.
• Demonstrated the feasibility of CIGS/AZO heterojunctions for high-performance optoelectronic devices.

Conclusions:

• The developed optoelectronic synapse based on CIGS/AZO p-n heterojunctions shows significant potential for neuromorphic computing.
• This work provides a viable approach for creating advanced artificial synapses for AI.
• The findings pave the way for next-generation AI hardware beyond the von Neumann architecture.