Memristor Arrays Formed by Reversible Formation and Breakdown of Nanoscale Silica Layers on Si–H Surfaces

Nonvolatile resistive switching, also known as the memristor effect, has emerged as an important concept in the development of neuromorphic computing. Memristive operation shares similarities to the mechanism of biological synapses, making it a promising technology for future artificial intelligence. To date, most memristor platforms are based on the resistive switching of an insulating material separating two metals. For future electrical circuitry that resembles those of synapses and neurons, memristors are likely to be integrated with materials already used in the microelectronics industry such as silicon. Here, we demonstrate a silicon-based memristor array based on creating nanoscale silicon oxide layers on silicon hydride surfaces, followed by a localized and reversible breakdown of the silicon oxide to form conducting channels. The size of an individual memristor is 30 nm, with a conductance ON/OFF ratio greater than 104 at room temperature. This work opens a new platform for realizing neuromorphic systems directly on silicon.