Reconfigurable neuromorphic memristor network for ultralow-power smart textile electronics

Area of Science:

• Materials Science
• Neuroscience
• Electrical Engineering

Background:

• Neuromorphic computing memristors are crucial for low-power electronic textiles and wearable devices.
• Current artificial synapses and neurons require distinct materials and configurations, limiting device multifunctionality.

Purpose of the Study:

• To develop a single, reconfigurable fiber-based memristor capable of both synaptic and neuron functions for electronic textiles.
• To reduce the complexity and energy consumption of neuromorphic computing circuits in wearable systems.

Main Methods:

• Fabrication of a textile memristor network using Ag/MoS2/HfAlOx/carbon nanotube materials.
• Demonstration of nonvolatile synaptic plasticity and volatile neuron functions within a single device.
• Integration of memristive neuron, synapse, and heating resistor into a smart textile system.

Main Results:

• The reconfigurable memristor achieved both synaptic plasticity and integrate-and-fire neuron functions.
• The fiber-based memristive neuron exhibited an ultra-low firing energy consumption of 1.9 fJ/spike, three orders of magnitude lower than existing artificial neurons.
• A smart textile system for warm fabric applications was successfully constructed.

Conclusions:

• The developed textile memristor offers a significant advancement in creating multifunctional, low-power electronic textiles for neuromorphic computing.
• The ultra-low energy consumption paves the way for electronic neural networks rivaling the efficiency of the human brain.
• This work provides a pathway towards next-generation in-memory computing textile systems.