Technological progress is advancing at a rapid pace, and the rate of chip performance improvements and transistor density increases is gradually diverging from Moore's Law predictions. Simultaneously, the "memory wall" and "power wall" challenges inherent in the von Neumann architecture have become increasingly prominent. Today, neuromorphic computing—inspired by the human brain—is emerging as one of the most promising computational paradigms. To realize efficient brain-like hardware, a high-performance, scalable, non-volatile memory device is essential. Among these, ferroelectric field-effect transistors (FeFETs) based on hafnium oxide (HfO₂) stand out as the most representative. It effectively mimics biological synaptic plasticity while offering high compatibility with CMOS processes, excellent scalability, fast low-power switching characteristics, and inherent controllable three-terminal field-effect properties.

The paper systematically reviews the research progress of HfO₂-based ferroelectric devices in the field of neuromorphic computing. Starting from the ferroelectric physical properties of HfO₂, it first discusses engineering regulation of the ferroelectric phase through methods such as doping and strain, and elaborates on the associated material-level challenges. Building upon this foundation, this review discusses representative HfO₂-based ferroelectric device architectures, including FeFETs, ferroelectric tunnel junctions (FTJs), ferroelectric diodes (Fe-diodes), and advanced fin field-effect transistor (FinFET)/nanowire ferroelectric transistors. The primary focus is on their applications in simulating biological synapses and neurons, along with the impact on performance optimization. The study also extends from devices to system-level integration, addressing issues such as the non-idealities in crossbar arrays, power consumption in peripheral circuits, and challenges in hardware-software co-design. Finally, this research comprehensively summarizes the significant challenges currently facing science and technology from multiple perspectives and proposes several promising solutions. Through comprehensive research, this paper identifies contemporary development directions: integrated applications of two-dimensional ferroelectric materials, exploration of novel computational paradigms, and convergence of sensing, storage, and computation. It aims to provide a comprehensive reference framework for future research in this field.

Traditional computing structures are blocked by the von Neumann bottleneck, and neuromorphic computing devices inspired by the human brain which integrate storage and computation have received more and more attention. Here, a flexible organic device with 2,7-dioctyl[1] benzothieno [3,2-b][1] benzothiophene (C8-BTBT) and 2,9-didecyldinaphtho [2,3-b:2′,3′-f] thieno [3,2-b] thiophene (C10-DNTT) heterostructural channel having excellent synaptic behaviors was fabricated on muscovite (MICA) substrate, which has a memory window greater than 20 V. This device shows better electrical characteristics than organic field effect transistors with single organic semiconductor channel. Furthermore, the device simulates organism synaptic behaviors successfully, such as paired-pulse facilitation (PPF), long-term potentiation/depression (LTP/LTD) process, and transition from short-term memory (STM) to long-term memory (LTM) by optical and electrical modulations. Importantly, the neuromorphic computing function was verified using the Modified National Institute of Standards and Technology (MNIST) pattern recognition, with a recognition rate nearly 100% without noise. This research proposes a flexible organic heterojunction with the ultra-high recognition rate in MNIST pattern recognition and provides the possibility for future flexible wearable neuromorphic computing devices.

Brain-inspired neuromorphic computing can overcome the energy and throughput limitations of traditional von Neumann-type computing systems, which requires analog updates of their artificial synaptic strengths for the best recognition performance and low energy consumption. Here, we report synaptic devices made from highly insulating ferroelectric LiNbO₃ (LNO) thin films bonded to SiO₂/Si wafers. Through the creation/annihilation of periodically arrayed antiparallel domains within LNO nanocells, which are stimulated using positive/negative voltage pulses (synaptic plasticity), we can modulate the synaptic conductance linearly by controlling the number of the conducting domain walls. The multilevel conductance is nonvolatile and reproducible with negligible dispersion over 100 switching cycles, representing much better performance than that of random defect-based nonlinear memristors, which generally exhibit large-scale resistance dispersion. The simulation of a neuromorphic network using these LNO artificial synapses achieves 95.6% recognition accuracy for faces, thus approaching the theoretical yield of ideal neuromorphic computing devices.

The explosively developed era of big-data compels the increasing demand of nonvolatile memory with high efficiency and excellent storage properties. Herein, we fabricated a high-speed photoelectric multilevel memory device for neuromorphic computing. The novel two-dimensional (2D) MoSSe with a unique Janus structure was employed as the channel, and the stack of Al₂O₃/black phosphorus quantum dots (BPQDs)/Al₂O₃ was adopted as the dielectric. The storage performance of the resulting memory could be verified by the endurance and retention tests, in which the device could remain stable states of programming and erasing even after 1, 000 cycles and 1, 000 s. The multibit storage could be realized through both different voltage amplitudes and pulse numbers, which could achieve 6 bits (64 distinguishable levels) under pulse width of 50 ns. Furthermore, our memory device also could realize the simulations of synapses in human brain with optical and electric modulations synergistically, such as excitatory post-synaptic current (EPSC), long-term potentiation/depression (LTP/LTD), and spike-timing-dependent plasticity (STDP). Neuromorphic computing was successfully achieved through a high recognition of handwritten digits up to 92.5% after 10³ epochs. This research is a promising avenue for the future development of efficient memory and artificial neural network systems.

Atomic layer deposition (ALD) can be used for wafer-scale synthesis of 2D materials. In this paper, a novel, reliable, secure, low-cost, and high-efficiency process for the fabrication of MoS₂ is introduced and investigated. The resulting 2D materials show high carrier-mobility as well as excellent electrical uniformity. Using molybdenum pentachloride (MoCl₅) and hexamethyldisilathiane (HMDST) as ALD precursors, thickness-controlled MoS₂ films are uniformly deposited on a 50 mm sapphire and a 100 mm silica substrate. This is done with a high growth-rate (up to 0.90 Å/cycle). Large-scale top-gated FET arrays are fabricated using the films, with a room-temperature mobility of 0.56 cm²/(V·s) and a high on/off current ratio of 10⁶. Excellent electrical uniformity is observed in the whole sapphire wafer. Additionally, logical circuits, including inverters, NAND, AND, NOR, and OR gates, are realized successfully with a high-k HfO₂ dielectric layer. Our inverters exhibit a fast response frequency of 50 Hz and a DC-voltage gain of 4 at V_DD = 4 V. These results indicate that the new method has the potential to synthesize high quality MoS₂ films on a large-scale, with hypo-toxicity and enhanced efficiency, which can facilitate a broader range of applications in the future.