Unconventional Inorganic‐Based Memristive Devices for Advanced Intelligent Systems

Abstract

The latest progresses in software engineering such as cloud computing, big data analysis, and machine learning have accelerated the emergence of advanced intelligent systems (AIS). However, the current computing system has significant challenges in dealing with unstructured data (e.g., image, voice, physiological signals) because the von‐Neumann bottleneck induces latency and power consumption issues. Neuromorphic computing, which imitates the behaviors of neuron and synapse within the biological neural network, is considered a promising solution beyond von‐Neumann architecture, since its collocated structure of processor and memory enables parallel processing of unstructured data with remarkable efficiency. Memristors are considered as next‐generation nonvolatile memory devices due to fast speed, low power, and excellent scalability. However, a low reliability and leakage current issues remain as obstacle to the commercialization of memristors. Memristive devices have been widely investigated as a strong candidate for artificial synapses since their resistance modulation characteristics under electrical stimulus are analogous to the plasticity of the brain synapse. Although emulation of synaptic behavior by single memristor cells is demonstrated by many researchers, the development of fully functional memristive neural network requires further investigations. This paper introduces the recent advances and developments in the field of inorganic‐based unconventional memristive devices for future AIS applications.

Semiconductor Quantum Dots for Memories and Neuromorphic Computing Systems

“…Solution-processable inorganic semiconductor quantum dots (QDs) offer an appealing path to nonvolatile memory and neuromorphic computing since they can be processed in solution while retaining the excellent electronic performance and structural stability of crystalline inorganic materials. − These nanometer-scale semiconductor crystals (typically 2–20 nm) can be tailored in composition (II–VI, III–V, and IV–VI semiconductors), size, shape, and surface ligands, ensuring the ability to easily engineer properties including gap energy, self-assembly capability, photoluminescence efficiency, and quantum confinement effect. Especially, electrophotoactive self-assembled semiconductor QD films can be integrated as floating gates, trapping sites, or channel layers in three-terminal flash memories and as resistive-switching media in two-terminal memristive devices (Figure b). ,− By doping the QD surface with ions, atoms, and molecular ligands or wrapping the QD core with an epitaxial semiconductor shell, desired merits such as high mobility for closely packed QDs films, low barriers at QD/electrode or QD/insulator interfaces, and charge-confinement capacity can be obtained . Recent studies have demonstrated that high surface-to-volume ratio of semiconductor QDs generates an influential role of the surface composition and structure to provide novel physical mechanisms in flash and RRAM devices, ensuring fast switching speed, low-energy operation, long state retention, and high reproducibility.…”

Memristor-Based Artificial Chips

“…High-quality manufacturing of memristor devices and large-scale integrated memristive hardware systems is expected from the collaboration of researchers with industry partners. Nowadays, some typical approaches, i.e., nanoimprinting, solution preparation, and electron beam lithography, are not ideal in practice to provide the high-quality and/or multifunctional-integration requirement of memristor devices and/or memristor arrays. − Meanwhile, most AI is realized based on software, and its high power consumption becomes an obstacle on the road to intelligent terminal development. Thus, the preparative technique and promising solutions of memristor-based nanodevice integrated chip systems require further investigation and development.…”

Memory-centric neuromorphic computing for unstructured data processing

“…However, the current memory devices are now experiencing inherent limitations related to device scaling, power consumption, and data processing complexity [54,[57][58][59]. Numerous next generation nonvolatile memories (NVMs) have been proposed with impressive architectures and mechanisms to redesign the current memory hierarchy [60][61][62][63][64][65][66][67]. NVMs such as RSMs, PCMs, and FTJs are expected to overcome the performance limit of conventional memories devices in speed (< 10 ns for SRAM and 200 μs for Flash memory), cell size (4-5 F 2 (feature size) for Flash memory and > 100 F 2 for SRAM), endurance (> 10 15 for DRAM and 10 4 for Flash memory), and retention (10 years in ambient temperature) [68].…”