Search

With the rapid growth of computer science and big data, the traditional von Neumann architecture suffers the aggravating data communication costs due to the separated structure of the processing units and memories. Memristive in-memory computing paradigm is considered as a prominent candidate to address these issues, and plentiful applications have been demonstrated and verified. These applications can be broadly categorized into two major types: soft computing that can tolerant uncertain and imprecise results, and hard computing that emphasizes explicit and precise numerical results for each task, leading to different requirements on the computational accuracies and the corresponding hardware solutions. In this review, we conduct a thorough survey of the recent advances of memristive in-memory computing applications, both on the soft computing type that focuses on artificial neural networks and other machine learning algorithms, and the hard computing type that includes scientific computing and digital image processing. At the end of the review, we discuss the remaining challenges and future opportunities of memristive in-memory computing in the incoming Artificial Intelligence of Things era., (© 2022. The Author(s).)

Electrical synapses provide rapid, bidirectional communication in nervous systems, accomplishing tasks distinct from and complementary to chemical synapses. Here, we demonstrate an artificial electrical synapse based on second-order conductance transition (SOCT) in an Ag-based memristor for the first time. High-resolution transmission electron microscopy indicates that SOCT is mediated by the virtual silver electrode. Besides the conventional chemical synaptic behaviors, the biphasic plasticity of electrical synapses is well emulated by integrating the device with a photosensitive element to form an optical pre-processing unit (OPU), which contributes to the retinal neural circuitry and is adaptive to ambient illumination. By synergizing the OPU and spiking neural network (SNN), adaptive pattern recognition tasks are accomplished under different light and noise settings. This work not only contributes to the further completion of synaptic behaviour for hardware-level neuromorphic computing, but also potentially enables image pre-processing with light adaptation and noise suppression for adaptive visual recognition.

The field of neuromorphic computing systems has been through enormous progress in recent years, whereas some issues are still remaining to be solved. One of the biggest challenges in neuromorphic circuit designing is the lack of a robust device with functions comparable to or even better than the metal-oxide-semiconductor field-effect transistor (MOSFET) used in traditional integrated circuits. In this work, we demonstrated a MoS 2 neuristor using a dual-gate transistor structure. An ionic top gate is designed to control the migration of ions, while an electronic back gate is used to control electronic migration. By applying different driving signals, the MoS 2 neuristor can be programmed as a neuron, a synapse, or an n-type MOSFET, which can be seen as a fundamental building block in the neuromorphic circuit design. The MoS 2 neuristor provides viable solutions for future reconfigurable neuromorphic systems and can be a promising candidate for future neuromorphic computing.

Memristive devices based on electrochemical processes are promising candidates for next-generation memory and neuromorphic applications. The redox processes happening at the interfaces are crucial steps for the ionization as well as generation of counter charges, and are thus indispensable for successful resistive switching, but their detailed mechanism has not been fully clarified. Here, we study the interfacial redox reactions in the forming process of memristive devices based on valence change and electrochemical metallization, using high-resolution electron microscopy and electrostatic force microscopy observations. We show direct evidence for the anodic oxidation of oxygen ions and cathodic reduction of moisture in HfO2- and Ta2O5-based valence change cells, which could take place in different horizontal locations. We further found that the anodic reactions always led to more pronounced structural damage to the electrode, indicating the possibility of additional cathodic reactions without producing gaseous products. When an active electrode is present, oxidation of metal atoms takes place at the anodic interface instead. Further investigations on electrochemical metallization cells have identified Cu ionization and moisture reduction as the anodic and cathodic reactions, respectively, and formation of Cu nuclei at the cathodic interface was directly observed. These findings with microscopic evidence could facilitate future development of memristive devices.

Retina shows an extremely high signal processing efficiency because of its specific signal processing strategy which called computing in sensor. In retina, photoreceptor cells encode light signals into spikes and ganglion cells finish the shape perception process. In order to realize the neuromorphic vision sensor, the one-transistor-one-memristor (1T1M) structure which formed by one memristor and one MOSFET in serial is used to construct photoreceptor cell and ganglion cell. The voltage changes between two terminals of memristor and MOSFET can mimic the changes of membrane potential caused by spikes and illumination respectively. In this paper, the tunable memristive neurons with 1T1M structures are built. According to the concept of receptive field of ganglion cells (GCs) in the retina, the artificial shape perception retina network is constructed with these memristive neurons. The final results show that the artificial retina can extract shape information from the image and transfer it into spike frequency realizing the function of computing in sensor.

Neuromorphic computing represents an innovative technology that can perform intelligent and energy-efficient computation, whereas construction of neuromorphic systems requires biorealistic synaptic elements with rich dynamics that can be tuned based on a robust mechanism. Here, an ionic-gating-modulated synaptic transistor based on layered crystals of transitional metal dichalcogenides and phosphorus trichalcogenides is demonstrated, which produce a diversity of short-term and long-term plasticity including excitatory postsynaptic current, paired pulse facilitation, spiking-rate-dependent plasticity, dynamic filtering, etc., with remarkable linearity and ultralow energy consumption of ≈30 fJ per spike. Detailed transmission electron microscopy characterization and ab initio calculation reveal two-stage ionic gating effects, namely, surface adsorption and internal intercalation in the channel medium, causing different poststimulation diffusive dynamics and thus accounting for the observed short-term and long-term plasticity, respectively. The synaptic activity can therefore be effectively manipulated by tailoring the ionic gating and consequent diffusion dynamics with varied thickness and structure of the van der Waals material as well as the number, duration, rate, and polarity of gate stimulations, making the present synaptic transistors intriguing candidates for low-power neuromorphic systems., (© 2018 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim.)

Polychloro-para-xylylene (parylene-C) is a flexible and transparent polymer material which has excellent chemical stability and high biocompatibility. Here we demonstrate a polymer device based on single-component parylene-C with memory and temperature sensing functionalities. The device shows stable bipolar resistive switching behavior, remarkable storage window (>10⁴), and low operation voltages, exhibiting great potential for flexible resistive random-access memory (RRAM) applications. The I-V curves and conductive atomic force microscopy (CAFM) results verify the metallic filamentary-type switching mechanism based on the formation and dissolution of a metal bridge related to the redox reaction of the active metal electrode. In addition, due to the metallic properties of the low-resistance state (LRS) in the polymer device, the resistance in the LRS exhibits a nearly linear relationship at the temperature regime between 25 °C and 100 °C. With a temperature coefficient of resistance (TCR) of 2.136 × 10 -3 /°C, the device is also promising for the flexible temperature sensor applications., Competing Interests: The authors declare no conflict of interest.

A resistive switching device with inherent nonlinear characteristics through a delicately engineered interfacial layer is an ideal component to be integrated into passive crossbar arrays for the suppression of sneaking current, especially in ultra-dense 3D integration. In this paper, we demonstrated a TaO x -based bipolar resistive switching device with a nearly symmetrical bi-directional nonlinear feature through interface engineering. This was accomplished by introducing an ultra-thin interfacial layer (SiO 2-x ) with unique features, including a large band gap and a certain level of negative heat of oxide formation between the top electrode (TiN) and resistive layer (TaO x ). The devices exhibit excellent nonlinear property under both positive and negative bias. Modulation of the inherent nonlinearity as well as the resistive switching mechanism are comprehensively studied by scrutinizing the results of the experimental control groups and the extensive characterizations including detailed compositional analysis, which suggests that the underlying mechanism of the nonlinear behavior is associatively governed by the serially connected metallic conductive filament and Flower-Nordheim tunneling barrier formed by the SiO 2-x interface layer. The proposed device in this work has great potential to be implemented in future massive storage memory applications of high-density selector-free crossbar structure.

Animals' survival is dependent on their abilities to adapt to the changing environment by adjusting their behaviours, which is related to the ubiquitous learning behaviour, nonassociative learning. Thus mimicking the indispensable learning behaviour in organisms based on electronic devices is vital to better achieve artificial neural networks and neuromorphic computing. Here a three terminal device consisting of an oxide-based memristor and a NMOS transistor is proposed. The memristor with gradual conductance tuning inherently functions as the synapse between sensor neurons and motor neurons and presents adjustable synaptic plasticity, while the NMOS transistor attached to the memristor is utilized to mimic the modulatory effect of the neuromodulator released by inter neurons. Such a memristor-based multi-terminal device allows the practical implementation of significant nonassociative learning based on a single electronic device. In this study, the experience-induced modification behaviour, both habituation and sensitization, was successfully achieved. The dependence of the nonassociative behavioural response on the strength and interval of presented stimuli was also discussed. The implementation of nonassociative learning offers feasible and experimental advantages for further study on neuromorphic systems based on electronic devices.

Brain-inspired neuromorphic computing is expected to revolutionize the architecture of conventional digital computers and lead to a new generation of powerful computing paradigms, where memristors with analog resistive switching are considered to be potential solutions for synapses. Here we propose and demonstrate a novel approach to engineering the analog switching linearity in TaOx based memristors, that is, by homogenizing the filament growth/dissolution rate via the introduction of an ion diffusion limiting layer (DLL) at the TiN/TaOx interface. This has effectively mitigated the commonly observed two-regime conductance modulation behavior and led to more uniform filament growth (dissolution) dynamics with time, therefore significantly improving the conductance modulation linearity that is desirable in neuromorphic systems. In addition, the introduction of the DLL also served to reduce the power consumption of the memristor, and important synaptic learning rules in biological brains such as spike timing dependent plasticity were successfully implemented using these optimized devices. This study could provide general implications for continued optimizations of memristor performance for neuromorphic applications, by carefully tuning the dynamics involved in filament growth and dissolution.

Parylene is a Food and Drug Administration (FDA)-approved material which can be safely used within the human body and it is also offers chemically inert and flexible merits. Here, we present a flexible parylene-based organic resistive random access memory (RRAM) device suitable for wearable biomedical application. The proposed device is fabricated through standard lithography and pattern processes at room temperature, exhibiting the feasibility of integration with CMOS circuits. This organic RRAM device offers a high storage window (>10(4)), superior retention ability and immunity to disturbing. In addition, brilliant mechanical and electrical stabilities of this device are demonstrated when under harsh bending (bending cycle >500, bending radius <10 mm). Finally, the underlying mechanism for resistance switching of this kind of device is discussed, and metallic conducting filament formation and annihilation related to oxidization/redox of Al and Al anions migrating in the parylene layer can be attributed to resistance switching in this device. These advantages reveal the significant potential of parylene-based flexible RRAM devices for wearable biomedical applications.

Here we propose a novel encapsulated vertical 3D RRAM structure with each resistive switching cell encapsulated by dielectric layers, contributing to both the reliability improvement of individual cells and thermal disturbance reduction of adjacent cells due to the effective suppression of unwanted oxygen vacancy diffusion. In contrast to the traditional vertical 3D RRAM, encapsulated bar-electrodes are adopted in the proposed structure substituting the previous plane-electrodes, thus encapsulated resistive switching cells can be naturally formed by simply oxidizing the tip of the metal bar-electrodes. In this work, TaO x -based 3D RRAM devices with SiO2 and Si3N4 as encapsulation layers are demonstrated, both showing significant advantages over traditional unencapsulated vertical 3D RRAM. Furthermore, it was found thermal conductivity and oxygen blocking ability are two key parameters of the encapsulation layer design influencing the scalability of vertical 3D RRAM. Experimental and simulation data show that oxygen blocking ability is more critical for encapsulation layers in the relatively large scale, while thermal conductivity becomes dominant as the stacking layers scale to the sub-10 nm regime. Finally, based on the notable impacts of the encapsulation layer on 3D RRAM scaling, an encapsulation material with both excellent oxygen blocking ability and high thermal conductivity such as AlN is suggested to be highly desirable to maximize the advantages of the proposed encapsulated structure. The findings in this work could pave the way for reliable ultrahigh-density storage applications in the big data era.

A novel vertical 3D RRAM structure with greatly improved reliability behavior is proposed and experimentally demonstrated through basically compatible process featuring self-localized switching region by sidewall electrode oxidation. Compared with the conventional structure, due to the effective confinement of the switching region, the newly-proposed structure shows about two orders higher endurance (>10(8) without verification operation) and better retention (>180h@150 °C), as well as high uniformity. Corresponding model is put forward, on the base of which thorough theoretical analysis and calculations are conducted as well, demonstrating that, resulting from the physically-isolated switching from neighboring cells, the proposed structure exhibits dramatically improved reliability due to effective suppression of thermal effects and oxygen vacancies diffusion interference, indicating that this novel structure is very promising for future high density 3D RRAM application.