Reconfigurable Counterion Gradient around Charged Metal Nanoparticles Enables Self-Rectifying and Volatile Artificial Synapse

Abstract

Developing a device that faithfully replicates the characteristics of biological synapses is a prerequisite for the hardware implementation of neuromorphic computing. An electronic device fabricated with metallic materials, instead of traditional semiconductors, is “genetically” impossible since the field applied to the metal is screened, leading to insensitive modulation in material conductance. Herein, these two independent or uncorrelated “challenges” can be addressed by using metal nanoparticles decorated with charged molecules. Specifically, we have developed a self-rectifying and volatile metal nanoparticle artificial synapse whose conductance can be continuously modulated. Theoretical calculations indicate that the current rectification is due to the reconfigurable asymmetric counterion gradients within the nanoparticle channel with asymmetric electrode areas. The synaptic functions based on nanoparticle devices are subsequently emulated. Finally, we demonstrate that our synapse array can accurately classify handwritten digits with a newly developed computational structure that combines CNN and reservoir computing.