Charge-Selective 2D Heterointerface-Driven Multifunctional Floating Gate Memory for In Situ Sensing-Memory-Computing

Authors:

Abstract
Flash memory, dominating data storage due to its substantial storage density and cost efficiency, faces limitations such as slow response, high operating voltages, absence of optoelectronic response, etc., hindering the development of sensing-memory-computing capability. Here, we present an ultrathin platinum disulfide (PtS2)/hexagonal boron nitride (hBN)/multilayer graphene (MLG) van der Waals heterojunction with atomically sharp interfaces, achieving selective charge tunneling behavior and demonstrating ultrafast operations, a high on/off ratio (108), extremely low operating voltage, robust endurance (105 cycles), and retention exceeding 10 years. Additionally, we achieve highly linear synaptic potentiation and depression, and observe the reversibly gate-tunable transitions between positive and negative photoconductivity. Furthermore, we employed the VGG11 neural network for in situ trained in-sensor-memory-computing to classify the CIFAR-10 data set, pushing accuracy levels comparable to pure digital systems. This work could pave the way for seamlessly integrated sensing, memory, and computing capabilities for diverse edge computing.
Funding Information
  • Research Grants Council, University Grants Committee (17205921, 17212923, 27206321)
  • National Key Research and Development Program of China (2022YFA1405600)
  • Science, Technology and Innovation Commission of Shenzhen Municipality (SGDX2022053011405040)
  • Natural Science Foundation of Beijing Municipality (Z210006)
  • National Natural Science Foundation of China (12104051, 62122004)

This publication has 55 references indexed in Scilit: