Retention control of Ag nanoparticle-based memristor via laser-material interaction

Abstract

Next-generation neuromorphic computing aims to create a highly integrated and energy-efficient computing network by emulating the way human brain processes information. Artificial synapses and artificial neurons which are implemented through nonvolatile and volatile memristors are getting attention due to their hysteretic and resistive switching properties and spiking characteristics. Since the retention of the memristor is significantly determined by the material properties, previous studies have focused on implementing artificial neuron and synapse with different materials. Methods of connecting memristor-based artificial synapses with bulky CMOS neurons or connecting neurons and synapses with different materials have been mainly reported. However, these methods not only increase process complexity and cost, but also have disadvantages in that retention of each unit cannot be precisely controlled. In this paper, a method to control retention via laser irradiation is reported based on Pt/SiO2:Ag/Au-structured single device. This retention control resulted in the transition from volatile to nonvolatile switching of memristor, which is suitable for simultaneously emulating neurons and synapses. Leaky integrate-and-fire (LIF) of biological neurons and synaptic plasticity of biological synapses were both emulated by using artificial memristive neural components. In addition, the retention control mechanism was investigated through various material analyses.