1 September 2024 in Fundamental Research
In recent years, the emergence of numerous applications of artificial intelligence (AI) has sparked a new technological revolution. These applications include facial recognition, autonomous driving, intelligent robotics, and image restoration. However, the data processing and storage procedures in the conventional von Neumann architecture are discrete, which leads to the “memory wall” problem. As a result, such architecture is incompatible with AI requirements for efficient and sustainable processing. Exploring new computing architectures and material bases is therefore imperative. Inspired by neurobiological systems, in-memory and in-sensor computing techniques provide a new means of overcoming the limitations inherent in the von Neumann architecture. The basis of neural morphological computation is a crossbar array of high-density, high-efficiency non-volatile memory devices. Among the numerous candidate memory devices, ferroelectric memory devices with non-volatile polarization states, low power consumption and strong endurance are expected to be ideal candidates for neuromorphic computing. Further research on the complementary metal–oxide–semiconductor (CMOS) compatibility for these devices is underway and has yielded favorable results. Herein, we first introduce the development of ferroelectric materials as well as their mechanisms of polarization reversal and detail the applications of ferroelectric synaptic devices in artificial neural networks. Subsequently, we introduce the latest developments in ferroelectrics-based in-memory and in-sensor computing. Finally, we review recent works on hafnium-based ferroelectric memory devices with CMOS process compatibility and give a perspective for future developments.
1 January 2025 in The Innovation Materials
Reservoir computing has emerged as an efficient computational paradigm for processing temporal and dynamic data, driving advancements in neuromorphic electronics for physical implementation. This review covers the advancements in neuromorphic devices for implementing physical reservoir computing, emphasizing device-level innovations that address the challenges of low-latency, energy-efficient, multimodal physical reservoir computing implementations. The advantages, disadvantages, and core challenges of various spatial architectures for building physical reservoir computing systems are discussed. Realistic paths on algorithmic and physical implementations of the input and output layers of the system are investigated, and issues such as heterogeneous device integration, consistent readout, and system stability are analyzed. This topical review emphasizes the reconfigurability and scalability of fully analogized physical reservoir computing architectures and adaptive dynamic nodes. We discuss challenges and future directions of physical reservoir computing across algorithmic, device, architectural, and application domains. This review establishes a foundational framework and provides strategic guidance for implementing physical reservoir computing in neuromorphic edge artificial intelligent systems.
10 February 2025 in Nano Letters
26 January 2024 in Journal of Applied Physics
Information processing and memorizing in the brain take place in a neural network consisting of neurons connected with each other by synapses. Meanwhile, the neural network is immersed in a common electrochemical environment with global parameters regulating the overall functions, which is barely discussed in neuromorphic devices. In this study, organic/inorganic hybrid transistors with sodium alginate as the gate dielectric layer and indium tin oxide as the channel were successfully prepared. We have not only simulated the basic properties of synapses in a single device, but, on top of that, also simulated the global regulation of information processing in the brain due to the incorporation of global grids, achieving excitatory and inhibitory synaptic weight. Moreover, the construction of a 3 × 3 synaptic array enables image learning and memorizing functions. These results demonstrate the significant advantages of electrolyte-gated transistors in enabling complex neural network connectivity and offer a promising opportunity for future artificial synapses.
25 March 2024 in Applied Physics Letters
The influence of electrodes on antiferroelectricity and fatigue endurance of 15 nm thick Hf0.2Zr0.8O2 thin films has been studied by a metal–antiferroelectric–metal capacitor structure using TiN and W as electrodes. The W|Hf0.2Zr0.8O2|W capacitor shows significantly enhanced antiferroelectricity and better endurance compared to the capacitor using TiN as the electrode. Assisted by grazing incidence x-ray diffraction and scanning transmission electron microscopy, the different electrical properties are discussed based on the contents of different phases and the diffusion of oxygen from the thin film into electrodes.