关键字：SYNAPSE

摘要：Metal oxide-based memristors usually exhibit robust resistive switching characteristics but poor mechanical tolerance, limiting their applications in wearable intelligent electronics. Here, we report a highly flexible and robust hafnium oxide-based memristor by using the ultrathin substrate for wearable in-memory computing. The fabricated memristor can display reliable resistive switching behaviors, including low switching voltage, good endurance, and excellent uniformity, under an extremely bending state with the radius of 0.8 mm. The mechanical behavior of the flexible memristor with varying substrate thickness is systematically analyzed using the finite element method. Moreover, typical synaptic plasticity including long-term potentiation and long-term depression was verified under the extremely bending state. Based on the highly flexible and robust memristor device, a three-layer neural network was constructed using a software simulator, achieving a recognition accuracy of 90.49% for handwritten digit recognition under the bending state. These results demonstrate our robust memristor synapse shows high mechanical flexibility, paving a promising way to realizing wearable in-memory computing.

卷号：123

期号：25

是否译文：否