Precise weight tuning in quantum dot-based resistive-switching memory for neuromorphic systems

Abstract

In this study, nonvolatile bipolar resistive switching and synaptic emulation behaviors are performed in an InGaP quantum dots (QDs)/HfO2-based memristor device. First, the physical and chemical properties of InGaP QDs are investigated by high-resolution transmission electron microscopy and spectrophotometric analysis. Through comparative experiments, it is proven that the HfO2 layer improves the variations in resistive switching characteristics. Additionally, the Al/QDs/HfO2/ITO device exhibits reversible switching performances with excellent data retention. Fast switching speeds in the order of nanoseconds were confirmed, which could be explained by trapping/detrapping and quantum tunneling effects by the trap provided by nanoscale InGaP QDs. In addition, the operating voltage is decreased when the device is exposed to ultraviolet light for low-power switching. Biological synapse features such as spike-timing-dependent plasticity are emulated for neuromorphic systems. Finally, the incremental step pulse using proven algorithm method enabled the implementation of four-bit states (16 states), markedly enhancing the inference precision of neuromorphic systems.