Iodide-Coated CsPbBr₃ Perovskite Nanowires for Resistive-Switching Memory in Neuromorphic Systems and Edge Computing

Area of Science:

• Materials Science
• Nanotechnology
• Computer Engineering

Background:

• The von Neumann architecture struggles with large-scale parallel processing demands of data-driven technologies.
• Neuromorphic computing, utilizing synaptic memory elements, presents a viable alternative for efficient computation.
• Perovskite materials offer potential for advanced electronic devices due to their tunable properties.

Purpose of the Study:

• To develop a high-performance memristor platform using engineered perovskite nanowires (PNWs) for neuromorphic computing.
• To investigate the impact of surface engineering on the stability and functionality of CsPbBr3 PNW memristors.
• To demonstrate the potential of these memristors for brain-inspired computing and edge AI applications.

Main Methods:

• Surface engineering of CsPbBr3 perovskite nanowires (PNWs) by removing insulating ligands and applying iodide passivation.
• Fabrication and characterization of memristor devices based on the engineered PNWs.
• Evaluation of device performance, including switching characteristics, multilevel conductance, and synaptic plasticity.
• Testing of brain-inspired functions like time-dependent plasticity and Pavlovian learning.
• Assessment of performance on the EMNIST dataset and for low-latency edge computing.

Main Results:

• The surface-engineered CsPbBr3 PNWs exhibited reliable bipolar switching and stable filament formation.
• Devices demonstrated multilevel conductance, robust potentiation/depression, and key brain-inspired functions.
• Achieved 91.6% recognition accuracy on the EMNIST dataset.
• Enabled low-latency 4-bit edge computing, showcasing energy efficiency.
• Suppressed bromide-related defects through surface modification, enhancing device stability.

Conclusions:

• Colloidal CsPbBr3 PNWs, through surface engineering, provide a scalable and high-performance platform for memristors.
• These memristors are suitable for next-generation neuromorphic applications, offering energy-efficient intelligent hardware solutions.
• The developed platform addresses limitations of current architectures for advanced AI processing.