Nanopore confined resistive switches for neuromorphic applications
Two-terminal resistance change devices, so-called memristors are scalable below 5 nm due to the filamentary nature of their resistive switching, offer multi-bit operations via the analog tunability of their resistance state and directly facilitate low-power in-memory computing due to their non-volatile nature. The objectives of the proposed project are the fabrication, experimental characterization, microscopic understanding and optimization of strongly confined, nanometer scale memristor devices for computational purposes. We explore the structural properties, the microscopic conduction mechanisms and emergent functionalities upon downscaling toward the ultimate atomic limit.
Figure (a) In a biological neural network information is stored and transmitted via synapses. (b) Crossbar network of artificial synapses implemented by memristors. The inset shows a typical device layout where an insulator layer is sandwiched between two metal electrodes. (c) The operation cycle of a filamentary memristor. The electric field induced formation and disruption of a conducting filament across the insulator results in a tuneable device resistance, hallmarked by the pinched hysteresis loop in the current-voltage characteristic.