Resistive Random Access Memory Research Articles

Engineering of ZrO2-based RRAM devices for low power in-memory computing

This work evaluates the impact of applying hydrogen plasma (H-plasma) either after the deposition of the ZrO2 layer or as an intermediate step during the deposition on the performance of resistive random-access memory devices. Devices treated with H-plasma exhibited lower power consumption during the forming process and a higher Ron/Roff ratio over 50 cycles of SET and RESET pulses compared to untreated devices. The position of the plasma treatment significantly influenced the device's performance. We measured the leakage current, which correlates well with the forming process. Devices with higher leakage current required less power during the forming process. It was observed that a thicker capping layer following plasma insertion reduced forming power and improved the conductance quantization for multilevel cell characteristics.

Graphene Oxide-Doped Indium Oxide Buffer Film Sandwiched between Titanium Oxide Layers for the Development of Photosensitive Resistive Memory Devices.

Over the past decade, metal oxide semiconductors have attracted considerable attention because of their transparency, high intrinsic charge carrier mobility, and charge carrier density. Metal oxide semiconductors also provide a promising route to develop resistive memory devices because of the tunability of their conductivity via the removal of oxygen ions, forming oxygen vacancies that can act as electron donors. Here, this paper reports the fabrication of a resistive random-access memory (ReRAM) device with TiO2 and TiO2-x layers and introduces a solution-processed In2O3-graphene oxide (GO) buffer layer from a water-based precursor solution to tune the switching characteristics. The devices were compared with a ReRAM device with no buffer layer, and the device composition was optimized by tuning the GO concentration in the layers. The devices showed hysteresis under cyclically scanned bias between positive and negative voltages with clear switching between the low resistance state (LRS) and the high resistance state (HRS). The optimized device also showed a more than 3 orders of magnitude difference in resistance between the LRS and HRS and a stable current retention. The devices also showed photoresponsive behaviors. In the LRS, the current increased significantly under illumination, while in the HRS, the current deteriorated slowly when constantly illuminated. These results highlight the dual role of the GO flakes in the buffer layer by increasing the resistance in the HRS and providing a photoresponsive switching capability. The ReRAM device was connected to a TFT device, and its feasibility as a memory component in a digital circuit was successfully demonstrated. These results provide a new avenue for developing metal-oxide-based ReRAM devices with GO-containing active layers.

Effect of annealing conditions on resistive switching in hafnium oxide-based MIM devices for low-power RRAM

Abstract This work presents resistive switching (RS) behaviour in HfO2-based low-power resistive random-access memory (RRAM) devices. A metal-insulator-metal (MIM) structure (Au/HfO2/Pt) was fabricated by sandwiching a thin insulating layer of HfO2 between Pt and Au electrodes. HfO2 films deposited by RF sputtering at room temperature were rapid thermally annealed in N2 ambient at 400 °C and 500 °C. Grazing angle x-ray diffraction (GIXRD), Field emission gun-scanning electron microscopy (FEG-SEM), atomic force microscopy (AFM), and x-ray photoelectron spectroscopy (XPS) were employed to analyse the phase, crystal structure, morphology, surface roughness and chemical composition of the HfO2 films. The bipolar RS could be observed in both as-deposited and annealed HfO2 film-based devices from I–V characteristics measured using a source meter. We have investigated the effect of annealing temperature and annealing ambient on the phase formation of HfO2 as well as the RS characteristics and compared with as-deposited film-based device. Annealed HfO2 film-based devices exhibited improved electrical characteristics, including stable and repeatable RS at significantly lower switching voltages (<1 V) which indicates low power consumption in these devices. The relatively lower processing temperature of the HfO2 films and that too in the films deposited by physical vapor deposition (PVD) technique-RF magnetron sputtering makes this study significantly useful for resistive switching based non-volatile memories.

Enhancing Uniformity, Read Voltage Margin, and Retention in Three-Dimensional and Self-Rectifying Vertical Pt/Ta2O5/Al2O3/TiN Memristors.

This study introduces a Ta2O5-based self-rectifying memristor (SRM) with an Al2O3 interfacial layer adopted to improve switching uniformity, read voltage margin, and long-term retention. The Pt/Ta2O5/Al2O3/TiN (PTAT) device exhibits a 105 rectification ratio, 104 on/off ratio, 2 × 106 endurance, and retention of 104 s at 150 °C. A 3-layer 4 × 4 vertical resistive random access memory structure exhibits uniform switching parameters. The coefficient of variation (CV) for device-to-device measurements is 0.23 for the low resistance state (LRS) and 0.22 for the high resistance state (HRS), while for cycle-to-cycle measurements, the CV is 0.38 for the LRS and 0.11 for the HRS. Finally, the present study demonstrates the superior performance of the PTAT devices in the context of hardware-aware training for a fully connected neural network implementation. These advancements position the PTAT device as a promising candidate for high-density three-dimensional storage class memory and low-power neural networks, offering the consistent performance and reliability necessary for future high-density storage applications.

Investigation of self-selective RRAM based on V/ITO structure with rapid thermal annealed ITO for synapse emulation

This paper focuses on the investigation of V/ITO (O2 Rapid Thermal Annealing)-based Self-Selective Resistive Random-Access Memory (RRAM) device. In this study, the natural oxidation of Vanadium top electrode and the rapid thermal annealing (RTA) process greatly simplified the device fabrication process. The VO2-based Selector, formed by the natural oxidation of the Vanadium electrode, effectively suppresses current and is directly integrated with the ITO layer, eliminating the need for additional serial Selector. The oxygen content of the ITO film is significantly increased by the RTA process, enabling the previously conductive ITO material to be used as the RRAM insulating layer without the need for additional deposition of insulating layer. This V/ITO (O2 RTA) structure not only exhibits highly uniform resistance distributions (σ/μ<2.8 %) and endurance stability (10000 cycles), but also effectively simulates synaptic plasticity, exhibiting both short-term and long-term memory behaviors. Notably, potentiation and depression characteristics are displayed by the device when continuous pulse voltage is applied. These advancements underscore the innovation and applicability of RTA-treated ITO films in next-generation memory technologies.

A semiconducting supramolecular Co(II)-metallogel based resistive random access memory (RRAM) design with good endurance capabilities.

A highly efficient approach for synthesizing a supramolecular metallogel of Co(II) ions, denoted as CoA-TA, has been established under room temperature and atmospheric pressure conditions. This method employs the metal-coordinating organic ligand benzene-1,3,5-tricarboxylic acid as a low molecular weight gelator (LMWG) in DMF solvent. A comprehensive analysis of the mechanical properties of the resulting supramolecular Co(II)-metallogel was conducted through rheological investigation, considering angular frequency and thixotropic study. The hierarchical rocky network structure of the supramolecular Co(II)-metallogel was unveiled using field emission scanning electron microscopy (FESEM). Transmission electron microscopic (TEM) analysis showed rod-shaped structures via low-magnification high angle annular dark field (HAADF) bright field scanning transmission electron microscopic (STEM) imaging, while energy dispersive X-ray (EDX) elemental mapping confirmed its primary chemical constituents. The formation mechanism of the metallogel was examined via fourier transform infrared spectroscopy (FTIR) spectroscopy. The nature of the synthesized CoA-TA metallogel was affirmed through powder X-ray diffraction (PXRD) analysis. Furthermore, this study involved fabrication of Schottky diode structures in a metal-semiconductor-metal geometry based on cobalt(II) metallogel (CoA-TA), enabling observation of charge transport behavior. Remarkably, a resistive random access memory (RRAM) device utilizing cobalt(II) metallohydrogel (CoA-TA) demonstrated bipolar resistive switching at room temperature and under ambient conditions. The switching mechanism was investigated, revealing the formation and rupture of conductive filaments between metal electrodes that govern the resistive switching behavior. This RRAM device exhibited an impressive ON/OFF ratio (~ 414) and exceptional endurance over 5000 switching cycles. These structures offer great potential for diverse applications such as non-volatile memory design, neuromorphic computing, flexible electronics and optoelectronics. Their advantages lie in their fabrication process, reliable resistive switching behavior and overall performance stability.

Neuronal Multi Unit Activity Processing with Metal Oxide Memristive Devices

AbstractIntra‐cortical brain‐machine interfaces (BMIs), able to decode neural activity in real‐time, represent a revolutionary opportunity for treating medical conditions. However, traditional systems focusing on single‐neuron spike detection require high processing rates and power, hindering the up‐scaling for neurons‐population monitoring in clinical application. An intriguing proposition is the memristive integrating sensor (MIS) approach, which uses resistive RAM (RRAM) for threshold‐based neural activity detection. MIS leverages analogue multi‐state switching properties of metal‐oxide RRAM to compress neural inputs by encoding above‐threshold events in resistance displacement, facilitating efficient data down‐sampling in the post‐processing, enabling low‐power, high‐channel systems. Initially tested on spikes and local field potentials, here MIS is adapted to process multi‐unit activity envelope (eMUA)—the envelope of entire spiking activity—which has recently been proposed as crucial input for real‐time neuro‐prosthetic control. Prior necessary modifications to the MIS for effective operation, this adaptation achieved over 95% sensitivity across two types of metal‐oxide devices: Pt/TiOx/Pt and TiN/HfOx/TiN, proving its platform‐agnostic capabilities. Furthermore, towards the integration of MIS with silicon chips, it is shown that it can reduce total system power consumption to below 1 µW, as RRAM encoding stage relaxes the signal preservation and noise requirements that challenge traditional complementary metal‐oxide‐semiconductor (CMOS) front‐ends. This eMUA‐MIS adaptation offers a viable pathway for developing more scalable and efficient BMIs for clinical use.

An Efficient ReRAM-based Accelerator for Asynchronous Iterative Graph Processing

Graph processing has become a central concern for many real-world applications and is well-known for its low compute-to-communication ratios and poor data locality. By integrating computing logic into memory, resistive random access memory (ReRAM) tackles the demand for high memory bandwidth in graph processing. Despite the years’ research efforts, existing ReRAM-based graph processing approaches still face the challenges of redundant computation overhead . It is because the vertices of many subgraphs are ineffectively and repeatedly processed over the ReRAM crossbars for lots of iterations so as to update their states according to the vertices of other subgraphs regardless of the dependencies among the subgraphs. In this paper, we propose ASGraph , a dependency-aware ReRAM-based graph processing accelerator that overcomes the aforementioned performance bottlenecks. Specifically, ASGraph dynamically constructs the subgraph based on the dependencies between vertices’ states and then detects constructed subgraph that owns high value (it is likely that it has accumulated many state propagations from its neighbors and is able to affect more other neighbors) to be preferentially processed. In this way, it makes the vertex states propagate along the dependencies between vertices as much as possible to reduce the redundant computation. Besides, ASGraph employs a hybrid processing scheme to accelerate the state propagations of the tightly connected subgraph, thereby minimizing the redundant computations. Experimental results show that ASGraph achieves 25.5 × and 4.8 × speedup and 70.8 × and 2.2 × energy saving on average compared with the state-of-the-art ReRAM-based graph processing accelerators, i.e., GraphR and GaaS-X, respectively.

Enhanced resistive switching performance in TiN/AlOx/Pt RRAM by high-temperature I-V cycling

Set voltage is a key parameter for the application of Resistance Random Access Memory (RRAM). In this paper, based on TiN/AlOx/Pt RRAM synthesized by the magnetron sputtering method, we have studied the influence of I-V cycling at high temperatures on resistive switching performance. The results show that the treatment can significantly reduce the set voltage in resistive switching cycles. Moreover, the treatment can also enhance endurance effectively. Further studies indicated that a higher compliance current in treatment can induce a smaller and more uniform set voltage. We ascribe the improvements in resistive switching performance to the generation and accumulation of oxygen vacancies in the treatment. This research provides new ideas for synthesizing RRAM devices with low power consumption.

Demonstration of bipolar resistive memory fabricated using an ultra-thin BaTiOx resistive switching layer with a thickness of ∼5 nm

In this study, a high-performance non-volatile bipolar resistive random-access memory (RRAM), which was fabricated using an ultra-thin barium titanate (BaTiOx, BTO) film as a resistive switching (RS) layer, was demonstrated. The BTO RS layers, whose thicknesses were only about 5 nm, were prepared using radio-frequency sputtering under different oxygen flow rates. Introduction of oxygen was used to modify the chemical compositions of the BTO films. Copper was used as the top electrode material to realize a Cu/BTO/n+-Si electrochemical metallization memory, where the resistance switching was triggered by electrochemical reaction of Cu electrode. The RRAM could repeatedly and consistently switch between a high-resistance state and a low-resistance state over 300 cycles. A large memory window of 105 and a long data retention time of >1.5 × 104 s both at room temperature and 85 °C were observed. Moreover, the Cu/BTO/n+-Si RRAM showed high switching speeds of 60–70 ns.

Improving the Nonvolatile Memory Characteristics of Sol-Gel-Processed Y2O3 RRAM Devices Using Mono-Ethanolamine Additives.

In this study, Y2O3-based resistive random-access memory (RRAM) devices with a mono-ethanolamine (MEA) stabilizer fabricated using the sol-gel process on indium tin oxide/glass substrates were investigated. The effects of MEA content on the structural, optical, chemical, and electrical characteristics were determined. As the MEA content increased, film thickness and crystallite size decreased. In particular, the increase in MEA content slightly decreased the oxygen vacancy concentration. The decreased film thickness decreased the physical distance for conductive filament formation, generating a strong electric field. However, owing to the lowest oxygen vacancy concentration, a large electrical field is required. To ensure data reliability, the endurance cycles across several devices were measured and presented statistically. Additionally, endurance performance improved with the increase in MEA content. Reduced oxygen vacancy concentration can successfully suppress the excess formation of the Ag conductive filament. This simplifies the transition from the high- to the low-resistance state and vice versa, thereby improving the endurance cycles of the RRAM devices.

Resistive Switching Characteristics of NiO Thin Films Influenced by Changes in the Diameter of Nanometer-Scale Top Electrodes.

We investigated the forming, set, and reset voltages affected by the area of the top electrodes of a resistive random access memory (RRAM) capacitor fabricated by epitaxial NiO thin films. NiO RRAM capacitors with Au top electrode with a diameter of 100 μm showed typical unipolar switching characteristics. Au top electrodes with diameters of 10, 20, and 30 nm were formed on the surface of epitaxial NiO thin films by e-beam lithography. The forming, set, and reset voltages tended to decrease as the diameter of NiO RRAM capacitors with nanosized Au top electrodes decreased. As the area of the Au top electrodes increases, the volume of space in which the conductive filaments formed within the NiO RRAM capacitors can be formed becomes larger. This causes the conductive filament to be formed at relatively low energy, giving low forming voltages, while increasing the diversity of paths, causing a wide forming voltage distribution.

Photo‐synaptic Memristor Devices from Solution‐processed Ga2O3 Thin Films

AbstractHardware integration with biological synaptic function is the key to realizing brain‐like computing. Resistive Random Access Memory (RRAM), with a similar structure to biological synapses, are important candidate for the simulation of biological synaptic function. In this work, Ga2O3 film as a functional layer of RRAM is prepared by the solution method, and an RRAM‐based photo‐synaptic device with an Ag/Ga2O3/Si structure is constructed subsequently. The device exhibits excellent bipolar resistive switching characteristics, with the merits of a large storage window and long retention time. Furthermore, the devices generated excitatory postsynaptic currents (EPSC) and paired‐pulse facilitation (PPF) behaviors under light pulse stimulation, enabling the simulation of synaptic plasticity. The transformation of synaptic behavior from short‐term memory (STM) to long‐term memory (LTM) is achieved by observing the spike‐duration dependent plasticity (SDDP), spike‐intensity dependent plasticity (SIDP), spike‐number dependent plasticity (SNDP) and spike‐rate dependent plasticity (SRDP) characteristics of photonic synapses under different conditions. The device also simulates the process of successive “learning‐forgotten‐remembering”, revealing that RRAM‐based photonic synapses have great potential in the fields of artificial visual perception and memory storage.

A Cascaded ReRAM-based Crossbar Architecture for Transformer Neural Network Acceleration

Emerging resistive random-access memory (ReRAM) based processing-in-memory (PIM) accelerators have been increasingly explored in recent years because they can efficiently perform in-situ matrix-vector multiplication (MVM) operations involved in a wide spectrum of artificial neural networks. However, there remain significant challenges to apply existing ReRAM-based PIM accelerators to the most popular Transformer neural networks. Since Transformers involve a series of matrix-matrix multiplication (MatMul) operations with data dependencies, they should write intermediate results of MatMuls to ReRAM crossbar arrays for further processing. Conventional ReRAM-based PIM accelerators often suffer from high latency of ReRAM writes and intra-layer pipeline stalls. In this paper, we propose ReCAT, a ReRAM-based PIM accelerator designed particularly for Transformers. ReCAT exploits transimpedance amplifiers (TIAs) to cascade a pair of crossbar arrays for MatMul operations involved in the self-attention mechanism. The intermediate result of a MatMul generated by one crossbar array can be directly mapped to another crossbar array, avoiding costly analog-to-digital conversions. In this way, ReCAT allows MVM operations to overlap with the corresponding data mapping, hiding the high latency of ReRAM writes. Furthermore, we propose an analog-to-digital converter (ADC) virtualization scheme to dynamically share scarce ADCs among a group of crossbar arrays, and thus significantly improve the utilization of ADCs to eliminate the performance bottleneck of MVM operations. Experimental results show that ReCAT achieves 207.3 ×, 2.11 ×, and 3.06 × performance improvement on average compared with other Transformer acceleration solutions—GPUs, ReBert, and ReTransformer, respectively.

Investigation of the Versatile Utilization of Three-Dimensional Vertical Resistive Random-Access Memory in Neuromorphic Computing.

The continuous advancement of computing technologies such as the Internet of Things and artificial intelligence emphasizes the need for innovative approaches to data processing. Faced with limitations in processing speed due to the exponentially increasing volume of data, conventional von-Neumann computing systems are transforming into a new paradigm called neuromorphic computing, inspired by the efficiency of the human brain. We fabricate three-dimensional vertical resistive random-access memory (VRRAM), which is highly suited for neuromorphic computing, and demonstrate its value as an artificial synapse. Beyond simulating simple synaptic functionalities such as spike-rate-dependent plasticity, spike-timing-dependent plasticity, and paired-pulse facilitation, we propose specific applications and experimentally implement them. In pattern recognition simulations based on the weight update characteristics of the fabricated VRRAM, the accuracy achieved in pattern recognition using appropriate pulse schemes reaches 90.4%. Additionally, we demonstrate adaptive learning behavior on the device by mimicking Pavlov's dog experiment with combinations of applied voltage pulses. Finally, we employ suitable write/erase pulse trains to implement binary representations for decimal numbers ranging from 0 to 15, thereby illustrating the significant potential of local devices for edge computing applications.

One-Selector-One-Resistor Integrated Memory Cells Based on Two-Dimensional Heterojunction Memory Selectors.

Selectors are critical components for reducing the sneak path leakage currents in emerging resistive random-access memory (RRAM) arrays. Two-dimensional (2D) materials provide a rich choice of materials with van der Waals stacking to form the heterostructure selectors with controllable energy barriers. Here, we experimentally demonstrate 2D-material-based heterostructure selectors with exponential current-voltage (I-V) relationships and integrate them with hafnium oxide (HfOx)-based RRAMs, forming one-selector-one-resistor (1S1R) cells. The multilayer graphene (MG)/tungsten disulfide (WS2)/platinum (Pt) selector contains two asymmetric heterojunctions with different Schottky barriers, which lead to highly nonlinear and asymmetric I-V characteristics. The 2D selectors in 1S1R cells can successfully drive RRAMs, reduce sneak path leakage current by more than 100 times, and provide the set compliance current. The 1S1R cells are further modeled and integrated into both planar and 3D memory arrays, with circuit-level simulations demonstrating that the presence of 2D selectors in large memory arrays can reduce the power consumption by up to 86%, improve the read/write margin by up to 31%, and avoid write failure. Such a platform holds high potential for constructing 3D high-density memories and performing in-memory computing.

Large resistive switching in ultrathin BiFeO3 thin films

Electrically induced resistive switching (RS) effects have been proposed as the basis for future non-volatile memories. In this work, 9 nm-thick BiFeO3 (BFO) epitaxial thin films were deposited on (001)-oriented SrTiO3 substrates by pulsed laser deposition and their resistive switching (RS) behaviors were investigated. A large resistive switching with ON/OFF ratio of ∼106 is observed, surpassing the performance of most resistive random access memories ever reported. The conducting filament is proposed to dominate the RS behavior in the positive voltage region, while the modulation of ferroelectric polarization is suggested to play a significant role in the negative voltage region. Our study significantly deepens the understanding of the physical origin of RS and could provide a reference for designing high-performance memories and memristors based on ultrathin ferroelectric films.

Reset-Voltage Controlled Resistance-State and Applications of Forming-Free Fe-Doped SrTiO3 Thin-Film Memristor.

In this study, we prepared a strontium ferrite titanate (STF) thin film using a sol-gel process to insulate resistive random-access memory (RRAM) applications. Compared to the typical strontium titanate (STO) RRAM, the improvement in the resistive switching characteristics in STF RRAM is obvious. The Al/STO/ITO/Glass RRAM set/reset voltages of -1.4 V/+3.3 V and the Al/STF/ITO/Glass RRAM set/reset voltages of -0.45 V/+1.55 V presented a memory window larger than 103, a low operating voltage and device stability of more than 104 s. In this study, the influence of Fe on the conducting paths and the bipolar resistive switching properties of Al/STF/ITO/Glass RRAM devices is investigated.

The Role of Hydrogen in ReRAM.

Previous research on transistor gate oxides reveals a clear link between hydrogen content and oxide breakdown. This has implications for redox-based resistive random access memory (ReRAM) devices, which exploit soft, reversible, dielectric breakdown, as hydrogen is often not considered in modeling or measured experimentally. Here quantitative measurements, corroborated across multiple techniques are reported, that reveal ReRAM devices, whether manufactured in a university setting or research foundry, contain concentrations of hydrogen at levels likely to impact resistance switching behavior. To the knowledge this is the first empirical measurement depth profiling hydrogen concentration through a ReRAM device. Applying a recently-developed Secondary Ion Mass Spectrometry analysis technique enables to measure hydrogen diffusion across the interfaces of SiOx ReRAM devices as a result of operation. These techniques can be applied to a broad range of devices to further understand ReRAM operation. Careful control of temperatures, precursors, and exposure to ambient during fabrication should limit hydrogen concentration. Additionally, using thin oxynitride or TiO2 capping layers should prevent diffusion of hydrogen and other contaminants into devices during operation. Applying these principles to ReRAM devices will enable considerable, informed, improvements in performance.

A flexible and fast digital twin for RRAM systems applied for training resilient neural networks

Resistive Random Access Memory (RRAM) has gained considerable momentum due to its non-volatility and energy efficiency. Material and device scientists have been proposing novel material stacks that can mimic the “ideal memristor” which can deliver performance, energy efficiency, reliability and accuracy. However, designing RRAM-based systems is challenging. Engineering a new material stack, designing a device, and experimenting takes significant time for material and device researchers. Furthermore, the acceptability of the device is ultimately decided at the system level. We see a gap here where there is a need for facilitating material and device researchers with a “push button” modeling framework that allows to evaluate the efficacy of the device at system level during early device design stages. Speed, accuracy, and adaptability are the fundamental requirements of this modelling framework. In this paper, we propose a digital twin (DT)-like modeling framework that automatically creates RRAM device models from device measurement data. Furthermore, the model incorporates the peripheral circuit to ensure accurate energy and performance evaluations. We demonstrate the DT generation and DT usage for multiple RRAM technologies and applications and illustrate the achieved performance of our GPU implementation. We conclude with the application of our modeling approach to measurement data from two distinct fabricated devices, validating its effectiveness in a neural network processing an Electrocardiogram (ECG) dataset and incorporating Fault Aware Training (FAT).