Key points are not available for this paper at this time.
Memristors have garnered significant attention due to their potential for multilevel storage and synaptic learning capabilities. However, the variability in forming and breaking conductive pathways in conventional memristors restricts their application in simulating neural synapses. The resistive switching in ferroelectric memristors is driven by stable ferroelectric switching, effectively avoiding the issues. Herein, we proposed a ferroelectric memristor based on yttrium-doped hafnium oxide, in which the resistive mechanism depends on the interaction between ferroelectricity and vacancies. It exhibits high on/off ratio (>103), long retention time (>104 s), stable endurance (100 cycles), and multilevel resistive memory. For information manipulation, diverse Boolean logic functions can be demonstrated, validating reconfigurable memory logic processing. Additionally, leveraging the unique properties of the memristor, an image encryption function is implemented. For neuromorphic computing, a high recognition accuracy of 96.36% is achieved for the handwritten digit data set. These results mark a significant step forward in the advancement of information processing and neuromorphic computing.
Cheng et al. (Fri,) studied this question.