Interfacial Engineering of Metal Ion Injection in Threshold Switching Memristor for Neuronal Applications

Neural networks composed of artificial neurons and synapses mimicking the biological nervous systems have governed much attention because of their promising potential in high-density memory storage and large-scale neuromorphic computing. In this context, various resistive switching phenomena in memristors such as ionic charge transfer, oxygen vacancy ordering, and electrochemical metallization (ECM) have been utilized to replicate neuronal dynamics. Among them, ECM threshold switching (TS) devices are favorable for realizing complex neural networks due to their simple structure, large on/off ratio, and low operation voltages. However, ECM TS device uniformity is relatively poor, which may affect the accuracy of artificial neural networks, limiting the practical applications of ECM TS memristors. In this work, we experimentally demonstrate a highly-uniform Ta₂O₅- based TS device with nanoporous Pt interfacial layer as the high-performance selector, which shows low leakage current (&lt; 1 pA), high on/off ratio (&gt; 10⁸), and high endurance (&gt; 10⁸ cycles). Furthermore, the Ta₂O₅/Pt-nanoporous layer TS device exhibits self-oscillation behaviour at low voltage (&lt;1 V), where the oscillation frequency increases with the applied voltage and decreases with the load resistance.<br/>Based on the studies of the nanoporous Pt interfacial layer with Ta₂O₅ - based TS devices, the overall variability of the operating voltages of TS devices with nanoporous Pt interfacial layer were reduced to levels below that achieved in TS devices without nanoporous Pt interfacial layer. Furthermore, the nanoporous Pt interfacial layer enables the TS devices to operate at higher current compliance levels (&lt; 100 uA). The insertion of the nanoporous Pt interfacial layer offers a simplistic technique of metal ions manipulation in ECM TS devices.