Key points are not available for this paper at this time.
Abstract With the boom of artificial intelligence (AI) and big data, electronics demand faster computing speed and lower power consumption, however, von Neumann architecture of current devices feature severe drawbacks for the further improvement of computing capability due to its design with separated memory and central processing unit (CPU). Fortunately, emerging nonvolatile memory devices, especially memristors, exhibit tremendous advantages in breaking the “memory wall” between memory and CPU by virtue of their in‐computing and neuromorphic computing abilities. Here, a WO 3 /HfO 2 heterojunction‐based memristor is proposed, and the device exhibits extraordinary resistive switching behaviors (e.g., high ON/OFF ratio, stable endurance, long retention time, and multilevel resistance states) and neuromorphic characteristics (long‐term/short‐term synaptic activities). Further, the mechanism underlying the electrical performances of this device is studied. Silver conductive filaments and Schottky barrier models are proposed and explained successfully. Additionally, a multilayer layer perceptron neural network is constructed in terms of the memristor model, and variables embracing learning rate, algorithm, and training epochs, are explored to enhance the recognition accuracy of the network. Undoubtedly, the proposed high‐quality WO 3 /HfO 2 heterojunction‐based memristor contributes to promoting the development of high‐density storage and neuromorphic computing technology, showing fascinating prospects in the era of AI.
Liu et al. (Mon,) studied this question.