Insights of BDAPbI₄-Based Flexible Memristor for Artificial Synapses and In-Memory Computing

Area of Science:

• Materials Science
• Neuroscience
• Computer Science

Background:

• Neuromorphic computing aims to reduce energy consumption in artificial intelligence (AI) by mimicking brain-like spiking frameworks.
• Hybrid perovskites offer potential for developing advanced computing devices.

Purpose of the Study:

• To investigate the potential of butane-1,4-diammonium based low-dimensional Dion-Jacobson hybrid perovskite (BDAPbI4) memristor devices for artificial synapses and neuromorphic computing.
• To demonstrate the learning and memory capabilities of these memristor devices.

Main Methods:

• Memristor devices based on BDAPbI4 were fabricated and tested for spike-dependent plasticity, validating Hebbian learning rules.
• An artificial neural network (ANN) was utilized to evaluate handwritten image recognition accuracy using the memristor devices.
• A flexible 4x4 crossbar array was designed and programmed to showcase in-memory computing capabilities.

Main Results:

• Memristors exhibited Hebbian learning rules within a 10 ± 2 ms timeframe under both flat and bending conditions.
• The ANN achieved a high recognition accuracy of approximately 94% for MNIST handwritten images within 50 epochs.
• The flexible crossbar array demonstrated data retention up to 10^3 s with 26 multilevel resistance states and successful programming for image recognition.

Conclusions:

• The BDAPbI4 memristor devices show significant promise for artificial synapses and neuromorphic computing.
• The integration of supervised, unsupervised, and associative learning in these devices can accelerate learning and memory consolidation.
• These findings pave the way for future technologies including spiking neural networks, brain-machine interfaces, and adaptive control systems.