Temporal information encoding in dynamic memristive devices

Abstract

We show temporal and frequency information can be effectively encoded in memristive devices with inherent short-term dynamics. Ag/Ag2S/Pd based memristive devices with low programming voltage (∼100 mV) were fabricated and tested. At weak programming conditions, the devices exhibit inherent decay due to spontaneous diffusion of the Ag atoms. When the devices were subjected to pulse train inputs emulating different spiking patterns, the switching probability distribution function diverges from the standard Poisson distribution and evolves according to the input pattern. The experimentally observed switching probability distributions and the associated cumulative probability functions can be well-explained using a model accounting for the short-term decay effects. Such devices offer an intriguing opportunity to directly encode neural signals for neural information storage and analysis.

Document Details

Document Type
Pub Defense Publication
Publication Date
Nov 09, 2015
Source ID
10.1063/1.4935220

Entities

People

  • Chao Du
  • Lin Chen
  • Wei Lu
  • Wen Ma

Organizations

  • Defense Advanced Research Projects Agency
  • National Science Foundation
  • University of Michigan

Tags

Fields of Study

  • Physics

Readers

  • Mathematical Modeling and Probability Theory.
  • Optical Physics and Photonics.
  • Thin Film Deposition Science.