Bias Sweep-Induced Analog Memristor Behavior, Using a Cuprous Iodide Thin Film, for Neuromorphic Computing

Mixed ionic-electronic conductors (MIECs) are known to show analog resistive switching (RS) behavior due to their coupled electronic and ionic transport properties. This study introduces a memristor device, made up of a well-known MIEC cuprous iodide (CuI), for artificial synaptic applications. A cross-point structured Cu/CuI/Pt device initially shows digital bipolar RS under bias voltage sweeping, which is characterized by well-separated SET and RESET voltages with a high ON/OFF resistance ratio of ∼105. After 100 bias sweeping cycles, the device completely changes, showing analog RS behavior without any well-defined SET and RESET voltage, and exhibits a continuous current trajectory under bias sweeps. In comparison to the digital RS mode, the analog RS mode exhibits minimal cycle-to-cycle variability with a reduced ON/OFF ratio of ∼10. The current conduction mechanism underlying the digital switching behavior is ascribed to the formation and dissolution of a Cu filament. The analog RS behavior arises from charge trapping/detrapping at defect sites created during digital RS cycles. The device showing analog RS exhibits long-term and short-term plasticity, similar to biological synapses under voltage pulse applications. Utilizing the long-term plasticity data, artificial neural network simulations demonstrate an image recognition accuracy of ∼93% for handwritten digits. Furthermore, the device successfully replicates paired-pulse facilitation/depression and spike timing-dependent plasticity. These findings indicate the great potential of the CuI-based analog memristor to serve as artificial synapses for neuromorphic computing applications.