Neuromorphic floating-gate memory based on 2D materials
image: Firstly, a brief introduction was given to commonly used two-dimensional materials, including graphene, transition metal dichalcogenides (TMDC), black phosphorus (BP), and hexagonal boron nitride (h-BN). Their structural differences and unique physicochemical properties, such as electron mobility, bandgap, flexibility, and mechanical properties, were analyzed. The potential applications of these materials in FG transistors were also explored. Then, the charge tunneling mechanism of FG transistors was introduced, and the development of FG transistors classified into five common structures was summarized. In addition, various 2D FG devices and their applications in materials, devices, and applications in memory devices and neural morphology computing were systematically reviewed. Finally, issues related to material preparation, device structure design, device stability control, and integration of 2D FG transistors were raised, and potential development trends for exploration and innovation in this field were outlined.
Credit: Chao Hu, Beijing Institute of Graphic Communication.
The rapid expansion of artificial intelligence and the Internet of Things has intensified the demand for advanced computing capability and robust data-storage solutions. In conventional systems, the physical separation between the central processing unit and memory forces extensive data shuttling, which increases power consumption and reduces computational speed—commonly framed as the von Neumann bottleneck or “memory wall.” These limitations become more pressing as silicon scaling faces growing constraints, motivating the exploration of new architectures grounded in novel materials and operating principles. Neuromorphic computing, inspired by the brain’s parallelism and adaptive learning, aims to emulate synaptic information-processing mechanisms and to colocate computation with storage, thereby reducing energy dissipation while improving efficiency. However, existing software/hardware routes toward neuromorphic systems still encounter challenges such as high power consumption, limited integration density, and insufficient reliability, underscoring the need for device- and materials-level innovation. Two-dimensional materials, characterized by atomic thickness, electrical tunability, and high integrability, are regarded as a pivotal platform for synaptic devices because their interfaces are highly sensitive to charge transfer and electrostatic modulation. From a device-architecture perspective, floating-gate transistors—widely used in NOR/NAND flash—offer strong retention via charge storage and can potentially emulate synaptic plasticity. Yet, as device dimensions shrink and tunneling layers thin, conventional floating-gate memories suffer increased leakage, severe charge loss, and degraded reliability. “Leveraging the high mobility and tunability of 2D materials, integrating them with floating-gate structures is therefore viewed as a promising route toward advanced nonvolatile memory and neuromorphic hardware.” said the author Chao Hu, a researcher at Beijing Institute of Graphic Communication, “Accordingly, this review introduces key 2D materials and floating-gate fundamentals, surveys progress in 2D-material/FG device integration for memory and neuromorphic computing, and outlines major constraints and future research directions.”
This review centers on the emerging intersection between two-dimensional (2D) materials and floating-gate (FG) transistors, and organizes the discussion along a materials–mechanisms–architectures–applications–outlook narrative. It first frames the motivation by highlighting scaling-induced limitations in conventional FG memories—most notably increased leakage, severe charge loss, and degraded reliability as devices shrink and tunnel layers thin—thereby motivating the adoption of atomically thin, highly tunable 2D materials to advance nonvolatile memory and neuromorphic hardware. Regarding the core technical content, the article introduces commonly used 2D material families (e.g., graphene, TMDCs, black phosphorus, and h-BN) and discusses why their physicochemical properties make them attractive for FG integration. It then summarizes FG operating principles based on charge capture/release and program/erase operations, and reviews key charge-injection pathways, including channel hot-electron injection, Fowler–Nordheim tunneling, and direct tunneling. Building on this foundation, the review classifies 2D-material-based FG transistors into five representative architectures—back/top gate, dual gate, semifloating gate, two-terminal FG, and extended FG—and uses representative studies to compare trade-offs and capabilities across nonvolatile memory, logic/multivalued functionalities, and photomemory (e.g., nanosecond-scale writing and multilevel storage). At the application level, the review emphasizes near-sensor/edge neuromorphic computing, where synaptic devices are coupled with sensing to enable multifunctional perceptual intelligence, and it summarizes typical implementations spanning visual processing, auditory recognition, and tactile perception. It further consolidates major bottlenecks in materials preparation, device-structure design, stability control, and integration consistency, and outlines forward directions such as new synthesis routes, more integrable functional architectures, improved stability and scalable integration, and expanded application spaces.
In the concluding section, the authors emphasize that integrating two-dimensional (2D) materials with floating-gate (FG) device concepts opens an expanded design space for both nonvolatile memory and neuromorphic synaptic hardware. The atomic thickness and tunable electrical/optical properties of 2D materials can synergize with the FG charge trapping–detrapping mechanism, enabling controllable modulation of channel conductance to realize low-power, fast-response memory and synaptic functionalities with promising stability. The conclusion further notes that such 2D-material/FG devices have already shown potential for perception-oriented tasks, including vision-, audition-, and touch-related processing, consistent with the broader goal of enabling efficient, brain-inspired information handling. The authors then distill the primary barriers to engineering translation and scalable deployment. First, reproducible preparation of large-area, high-quality 2D films remains difficult, and defects or thickness nonuniformity can induce substantial device-to-device variability. Second, challenges become more pronounced when moving from single devices to arrays and system-level integration, where consistency and long-term stability dominate performance limitations. Addressing these issues requires coordinated optimization of contacts, dielectric integration, and interface quality to reduce variability and preserve reliable operation after integration. “Therefore, future research should broaden synthesis and fabrication routes and develop device architectures that better support miniaturization and multifunctional integration, thereby enabling wider applications in flexible/stretchable electronics, VLSI, sensing, and integrated optoelectronics.” said Chao Hu.
Authors of the paper include Chao Hu, Lijuan Liang, Jinran Yu, Liuqi Cheng, Nianjie Zhang, Yifei Wang, Yichen Wei, Yixuan Fu, Zhong Lin Wang, and Qijun Sun.
This work was supported by the Youth of Excellence Project of Beijing Institute of Graphic Communication (Ea202401), Beijing Natural Science Foundation (2202018), National Natural Science Foundation of China (21604005 and 52073031), and the National Key Research and Development Program of China (2023YFB3208102).
The paper, “Neuromorphic Floating-Gate Memory Based on 2D Materials” was published in the journal Cyborg and Bionic Systems on Apr. 22, 2025, at DOI: 10.34133/cbsystems.0256.