Versatile memristor for memory and neuromorphic computing

Nanoscale Horiz. 2022 Feb 28;7(3):299-310. doi: 10.1039/d1nh00481f.

Abstract

The memristor is a promising candidate to implement high-density memory and neuromorphic computing. Based on the characteristic retention time, memristors are classified into volatile and non-volatile types. However, a single memristor generally provides a specific function based on electronic performances, which poses roadblocks for further developing novel circuits. Versatile memristors exhibiting both volatile and non-volatile properties can provide multiple functions covering non-volatile memory and neuromorphic computing. In this work, a versatile memristor with volatile/non-volatile bifunctional properties was developed. Non-volatile functionality with a storage window of 4.0 × 105 was obtained. Meanwhile, the device can provide threshold volatile functionalities with a storage window of 7.0 × 104 and a rectification ratio of 4.0 × 104. The leaky integrate-and-fire (LIF) neuron model and artificial synapse based on the device have been studied. Such a versatile memristor enables non-volatile memory, selectors, artificial neurons, and artificial synapses, which will provide advantages regarding circuit simplification, fabrication processes, and manufacturing costs.

Publication types

  • Research Support, Non-U.S. Gov't

MeSH terms

  • Costs and Cost Analysis
  • Electronics
  • Neural Networks, Computer*
  • Neurons / physiology
  • Synapses*