Type Of Non-volatile Memory Research Articles

Analysis of conductive filament heat transfer in TiO2-based RRAM device

Abstract Resistive random access memory (RRAM) has developed into a new type of non-volatile memory that has attracted much attention for its high density and low power; TiO2 has been studied to make RRAM because of its high conductivity. Based on the COMSOL Multiphysics finite element method and domain decomposition method, an electrically-thermally coupled model of TiO2-based RRAM with oxygen vacancy (Vo) conduction mechanism is constructed, the formation and cutting process of RESET and SET conductive filament (CF) under different voltages is simulated. Moreover, the characteristics of the internal temperature distribution of the CF are explored. The results show that the RESET process is more sensitive to heat changes, the transverse direction thermal value of the CF is more obvious than the longitudinal direction, and the rate of change is faster. Additionally, the electric field induces the migration of the Vo in filament, which affects the enthalpy change of the device’s heat transfer and resistive properties. This is an important reference for a deeper understanding of the switching behavior of RRAM devices and the control mechanisms for thermal studies.

Global Stability of Phase-Change Neural Networks With Mixed Time Delays.

Phase-change memory (PCM) is a novel type of nonvolatile memory and is suitable for artificial neural synapses. This article investigates the Lagrange global exponential stability (LGES) of a class of PCNNs with mixed time delays. First, based on the conductivity characteristics of PCM, a piecewise equation is established to describe the electrical conductivity of PCM. By using the proposed piecewise equation to simulate the neural synapses, a novel PCNN with discrete and distributed time delays is proposed. Then, using comparative theory and fundamental inequalities, the LGES conditions based on the M -matrix are proposed in the sense of Filippov, and the exponential attractive set (EAS) is obtained based on M -matrix and external input. Moreover, the Lyapunov global exponential stability (GES) conditions of PCNNs without external input are obtained by using the inequality technique and eigenvalue theory, which is a form of M -matrix. Finally, two simulation examples are given to verify the validity of the obtained results.

Multi-level Storage Characteristics of MoSe2 Resistive Random Access Memory

Resistive Random Access Memory (RRAM) is a type of non-volatile memory (NVM) device that stores information by switching between high and low resistance values. It has attracted widespread attention due to its promising potential for miniaturization. In this study, molybdenum diselenide (MoSe2) was successfully synthesized via the hydrothermal method, and the RRAM was fabricated with MoSe2 as the resistance change layer. Furthermore, the MoSe2 samples were characterized by scanning electron microscopy (SEM) and transmission electron microscopy (TEM). The results demonstrate that the prepared MoSe2 forms flower-like nanospheres assembled by nanosheets, with a particle size is about 200 nm. In addition, RRAM has a low operating voltage (< 1V), a high OFF/ON-state resistance ratio (> 102), a good endurance (103 cycles), and its resistance switching mechanism is dominated by the trap-controlled space charge limited current (SCLC) mechanism in the high resistance state (HRS) and by the ohmic mechanism in the low resistance state (LRS). Furthermore, multi-level storage is achieved by adjusting the compliance currents (Icc) and the stop voltage (Vstop).

Neuromorphic synaptic applications of HfAlOx-based ferroelectric tunnel junction annealed at high temperatures to achieve high polarization

HfO2-based ferroelectric tunnel junctions (FTJs) are promising nonvolatile memory types for neural network applications because of their speed, low power, and excellent complementary metal-oxide semiconductor compatibility. Specifically, HfAlOx (HAO) has led to extensive research efforts owing to its outstanding ferroelectric performance. This is a result of the fact that the atomic radius of Al is smaller than that of Hf. In this study, we investigate the metal–ferroelectric–semiconductor device with an Al doping concentration of 2% that was annealed at 900 °C. A high-remnant polarization (Pr) value of 39.85 µC/cm2 and endurance were achieved by using the polarization switching positive-up-negative-down measurement method at this annealing condition. Our device shows long-term potentiation and depression properties, including high linearity and multiple conductance states for neuromorphic applications. Moreover, paired-pulse facilitation was implemented to mimic human synaptic functions. The construction of 16 states comprising four bits was achieved by employing reservoir computing with the FTJ device functioning as a physical reservoir. Finally, the results obtained from the experiment show promising outcomes for the ferroelectric memory characteristics and synaptic properties of the manufactured HAO device.

Improving resistive switching effect by embedding gold nanoparticles into ferroelectric thin films

The ferroelectric memristor, as a new type of nonvolatile memory, has a broad prospect in the fields of information storage, exchange, and neural computing. Nowadays, it is still a challenge to achieve ferroelectric memristor with high resistive switching effect. In this work, epitaxial Ca-doped Pb(Zr0.40Ti0.60)O3 films embedded various concentrations Au nanoparticles (NPs) (Abbreviated as Au-PCZT) were deposited on the Nb:SrTiO3 (NSTO) substrate to form Au-PCZT/NSTO heterostructures. On the one hand, adding appropriate Au NPs into the PCZT films can improve the leakage current. When Au NPs capture electrons, the current in the high resistance state will be decreased due to the coulomb blocking effect. On the other hand, the ferroelectric polarization still maintains at a good level. Eventually, the resistive switching (RS) on/off ratio can reach 106 by embedding 3 mol% Au NPs. Compared with the pure PCZT thin films, the resistance-variable switching ratio is improved by two orders of magnitude. In addition, multi-level data storage can be realized under different bias voltages. Our results provide a feasible way to achieve high-on/off-ratio ferroelectric memristors with multi-level data storage capabilities.

Enhanced memristor performance via coupling effect of oxygen vacancy and ferroelectric polarization

As a new type of nonvolatile memory, the resistive memristor has broad application prospects in information storage and neural computing based on its excellent resistive switching (RS) performance. At present, it is still a great challenge to improve both ferroelectric polarization and leakage current to achieve a high RS on/off ratio of ferroelectric memristors. Herein, epitaxial Pb(Zr0.40Ti0.60)O3 (PZT) thin films with low content Ca doping were deposited on the Nb:SrTiO3 substrate to prepare PCZT/NSTO heterostructures and their RS behaviors were studied. The research findings show that compared with pure PZT film, the ferroelectric polarization of 1-mol%-Ca-doped PZT film is slightly improved, while the leakage current is increased by three orders of magnitude. Therefore, the RS on/off ratio reaches 2.5 × 105, about three orders of magnitude higher than pure PZT films. The theoretical analysis reveals that the RS behavior of PCZT/NSTO heterostructures is controlled by the PCZT/NSTO interfacial barrier and the space charge-limited current mechanism. Our results demonstrate that the ferroelectricity and electricity of ferroelectric thin films can be improved simultaneously by doping low-content Ca ions to increase the RS performance, which provides a good reference for the development of high-performance ferroelectric memristor devices.

Security Assessment of Nonvolatile Memory Against Physical Probing

Abstract This article describes how physical attacks can be launched on different types of nonvolatile memory (NVM) cells using failure analysis tools. It explains how the bit information stored inside these devices is susceptible to read-out and fault injection attacks and defines vulnerability parameters to help quantify risks associated with different modalities of attack. It also presents an in-depth security analysis of emerging NVM technologies and discusses potential countermeasures.

Evolution of Nb-Ta Anodic Memristors Identified By Combinatorial Screening

The conventional memory technology, based on von Neumann architecture, has challenged technical and scaling limitations. The architectures with separated memory and computing units have resulted in the low efficiency of computing systems. Hence, huge scientific attention has been recently drawn into the development of novel systems using the in-memory computing concept. Memristors are foreseen as a type of non-volatile memories that can fulfill such requirements but also as electronic components that can solve sneak path currents problems while consuming low power. The information storage mechanism is based on the resistance change of memristors which is driven by an applied external electric field. With this regard, the memory medium plays a crucial role, being an oxide layer sandwiched between the metallic bottom and top electrode. Valve metals and their oxides have shown promising switching properties when used as elements of metal-insulator-metal memristive (MIM) structures.The switching mechanism still depends on the nano-dimensional conductive filaments formation inside the oxide. Even though devices based on Hf, Nb, Ti, and Ta demonstrated a long lifetime, high data retention or resistive state ratio, devices are suffering from low lifetimes in the order of several thousand switching cycles, mainly due to the unpredictable size and position of conductive filaments being affected by the movement of cations or O species. Therefore, the development of materials for defect-engineered memristive applications is crucial. This was the main motivation for this study supported by the fact that memristors based on pure valve metals exhibited promising switching behavior. In this work, devices based on Ta, Nb, and their alloys were investigated. Since the goal was to improve the properties of devices based on pure metals, Nb and Ta were mixed and their alloys were used as bottom electrode for anodic memristors. It can be inferred that by mixing Ta and Nb, multifunctional devices can be produced.High number of Nb-Ta alloys were co-deposited by sputtering onto previously oxidized Si wafers. Afterwards, the samples were anodized for growing an insulating active layer for memristive devices. Onto the so-formed mixed oxides, Pt was deposited as top electrode. The compositional gradient was mapped by a scanning energy-dispersive X-ray spectroscopy (SEDX) system. The total compositional spread of the Nb-Ta library was ranging from Nb - 13 at.%Ta to Nb – 80 at% Ta over three Si wafers. Memristive devices based on Nb or Ta were also fabricated in identical conditions and used as reference samples. By using a combinatorial approach, the investigation of various electrical parameters could be established for more than 300 memristive devices produced per Si wafer. Electrical characterization was performed by a self-developed Gantry robot controlled via Labview software specifically designed for basic tests required for memristive characterization (I-U sweeps, endurance, retention tests). Electrical properties were screened with a compositional resolution of 1 at.%.The results have shown that memristive devices based on Ta reached a long lifetime and high resistive state ratio, whereas devices based on Nb indicated multi-level switching as well as non-volatile and threshold switching characteristics. The formation of Ta-oxyphosphate, responsible for conductive filaments pining, was confirmed by hard X-ray photoelectron spectroscopy. Similarly, multi-level switching characteristics were explained by the incorporation of electrolyte species in Nb2O5. However, Nb devices did not show reversible switching from unipolar to bipolar mode and vice versa. The improvement regarding reversible switching between unipolar and bipolar mode has been observed for memristive devices based on alloys containing Nb -35 at. % Ta. These memristors have also demonstrated longer endurance and data retention as compared to devices based on pure Nb and Ta. The switching mechanism was investigated by conductive filaments imaging using transmission electron microscopy.Conclusively, anodic memristors based on Nb and Ta exhibited non-volatile and threshold characteristics which can spread the range of memristive applications. Reversible switching behavior and multi-level switching possibilities are relevant for the development of memristive synapses and neuromorphic systems. The anodic growth of an insulating layer for devices based on Nb, Ta, and Nb/Ta alloys has simplified, speed up and decreased the cost of fabrication processes. The performance of anodic memristors based on metallic thin films containing Nb -35 at. % Ta indicated a substantial improvement. The switching mechanisms could be suggested based on the nanostructures found in oxide revealing the position and size of conductive filaments.I. Zrinski et al., Appl. Surf. Sci., 548, 149093 (2021) https://doi.org/10.1016/j.apsusc.2021.149093.I. Zrinski, et al., Appl. Mater. Today, accepted, (2021).I. Zrinski, et. al, J. Phys. Chem. Lett., 12, 8917 (2021) https://doi.org/10.1021/acs.jpclett.1c02346 Figure 1

Bipolar Resistive Switching in Lanthanum Titanium Oxide and an Increased On/Off Ratio Using an Oxygen-Deficient ZnO Interlayer.

The present study pioneered an oxygen migration-driven metal to insulator transition Mott memory, a new type of nonvolatile memory using lanthanum titanium oxide (LTO). We first show the reset first bipolar property without an initial electroforming process in LTO. We used oxygen-deficient ZnO as an interlayer between LTO and a W electrode to clarify whether oxygen migration activates LTO as the Mott transition. ZnO oxygen deficiency provides oxygen ion migration paths as well as a reservoir, facilitating oxygen migration from LTO to the W electrode. Thus, including the ZnO interlayer improved oxygen migration between LTO and the W electrode, achieving a 10-fold increased on/off current ratio. The current research contributes to a better understanding of valence change Mott memory by exploring the LTO resistive switching mechanism and ZnO interlayer influences on the oxygen migration process.

Resistive properties of CuInS&lt;sub&gt;2&lt;/sub&gt; quantum dots regulated by niobium-doped lead zirconate titanate ferroelectric films

As a new type of non-volatile memory, quantum dot resistive random access memory (RRAM) has attracted much attention for its easy preparation, fast responding time, high storage density, and smaller device size. CuInS&lt;sub&gt;2&lt;/sub&gt; quantum dot (CuInS&lt;sub&gt;2&lt;/sub&gt; QD) is a kind of excellent resistive functional material with abundant electron capture sites, high optical absorption coefficient, and high carrier mobility. In this work, CuInS&lt;sub&gt;2&lt;/sub&gt; QDs/Nb:Pb (Zr&lt;sub&gt;0.52&lt;/sub&gt;Ti&lt;sub&gt;0.48&lt;/sub&gt;)O&lt;sub&gt;3&lt;/sub&gt; (PNZT) films are prepared by spin-coating CuInS&lt;sub&gt;2&lt;/sub&gt; QDs on PNZT films. The results show that the resistive properties of CuInS&lt;sub&gt;2&lt;/sub&gt; QDs RRAMs can be effectively improved by introducing PNZT films and can be controlled by changing the polarization direction. The CuInS&lt;sub&gt;2&lt;/sub&gt; QDs/PNZT film in the negative polarization state promotes the interfacial electrons to enter into the PNZT film, which will reduce the height of the interfacial barrier and the thickness of the interfacial depletion region. And it will reduce the resistance of the composite film at the low resistance state (LRS). Compared with the switching voltage and resistive switching ratio of the pure CuInS&lt;sub&gt;2&lt;/sub&gt; QDs film (10&lt;sup&gt;3&lt;/sup&gt;), the switching voltage of the device decreases to –4.1/3.4 V and the resistive switching ratio increases to 10&lt;sup&gt;6&lt;/sup&gt;. Furthermore, it maintains good stability in the 10&lt;sup&gt;3&lt;/sup&gt; cycle durability test. In contrast, the CuInS&lt;sub&gt;2&lt;/sub&gt; QDs/PNZT film interface has a larger barrier height and depletion-layer thickness when the PNZT is in the positive polarization state, which increases the resistance of the composite film in the LRS state. As a result, the switching voltage of the device increases to –6.4/5.7 V with a resistive switching ratio of 10&lt;sup&gt;4&lt;/sup&gt;. The resistive properties of the CuInS&lt;sub&gt;2&lt;/sub&gt; QDs/PNZT film can be tuned by changing the polarization direction, as the polarization direction of the PNZT changes the interfacial energy band structure and affects the conduction mechanism. This work reveals the feasibility of using ferroelectric thin films to improve the resistive properties of quantum dots RRAMs and thus providing an approach to further developing RRAMs.

Comparison of hexagonal boron nitride and MgO tunnel barriers in Fe,Co magnetic tunnel junctions

Magnetic tunnel junctions (MTJ) with MgO/Fe based interfaces and out-of-plane spin direction form the basis of present-day spin-transfer-torque magnetic random-access memory (STT-MRAM) devices. They are a leading type of nonvolatile memory due to their very long endurance times and lack of reliability problems. Many semiconductor devices, such as the field effect transistor or nonvolatile memories, have undergone fundamental changes in materials design as dimensional scaling has progressed. Here, we consider how the future scaling of the MTJ dimensions might affect materials choices and compare the performance of different tunnel barriers, such as 2D materials like h-BN with the existing MgO tunnel barriers. We first summarize key features of MgO-based designs of STT-MRAM. We then describe general aspects of the deposition of 2D materials and h-BN on metals. We compare the band structures of MgO and h-BN with their band gaps corrected for the GGA band error. The different absorption sites of h-BN on Fe or Co are compared in terms of physisorbtive or chemisorbtive bonding sites and how this affects their spin-polarized bands and the transmission magneto-resistance (TMR). The transmission magneto-resistance is found to be highest for the physisorptive sites. We look at how these changes would affect the overall TMR and how scaling might progress.

Phase-Change Memory by GeSbTe Electrodeposition in Crossbar Arrays

Phase change memories (PCM) is an emerging type of non-volatile memory that has shown a strong presence in the data-storage market. This technology has recently attracted significant research interest in the development of non-Von Neumann computing architectures such as in-memory and neuromorphic computing. Research in these areas has been primarily motivated by the scalability potential of phase change materials and their compatibility with industrial nanofabrication processes. In this work, we are presenting our development of crossbar phase change memory arrays through the electrodeposition of GeSbTe (GST). We show that GST can be electrodeposited in microfabricated TiN crossbar arrays using a scalable process. Our phase switching test of the electrodeposited materials have shown that a SET/RESET resistance ratio of 2-3 orders of magnitude is achievable with a switching endurance of around 80 cycles. These results represent the first phase switching of electrodeposited GeSbTe in microfabricated crossbar arrays. Our work paves the way towards developing large memory arrays involving electrodeposited materials for passive selectors and phase switching devices. It also opens opportunities for developing a variety of different electronic devices using electrodeposited materials.

Review on resistive switching mechanisms of bio-organic thin film for non-volatile memory application

Abstract Bio-organic, as one of the sustainable and bioresorbable materials, has been used as an active thin film in producing resistive switching random access memory (RRAM) due to its specialized properties. This type of nonvolatile memory consists of a simple unit structure with the processed and solidified bio-organic-based thin film sandwiched between two electrodes. Its memory characteristics are significantly affected by the resistive-switching mechanism. However, to date, the reported mechanisms are very diverse and scattered, and to our best knowledge, there is no literature that reviewed comprehensively the mechanisms of resistive switching in bio-organic-based thin films. Therefore, the objective of this article is to critically analyze data related to the mechanisms of the bio-organic-based RRAM since it was first reported. Based on the pool of literature, three types of mechanisms are categorized, namely electronic, electrochemical, and thermochemical, and the naming is well justified based on the principle of operation. The determining factors and roles of bio-organic material and the two electrodes in governing the three mechanisms have been analyzed, reviewed, discussed, and compared.

Non-volatile memory system design of edge server and cloud centralized server for multiple-tier 5G network

This paper co-designs a network and non-volatile memory system for a fifth generation (5G) multiple-tier network. Because a 5G network has shorter latency and wider bandwidth compared with a 4G network, the non-volatile memory system becomes a bottleneck. From the bottom to the top of the multiple-tier network, edge devices/machines, edge servers, and cloud centralized servers play different types of roles and handle different types of data. Thus, different types of non-volatile memory are required for cloud centralized servers, edge servers, and edge devices/machines. This work employs three-dimensional single-level cell (3D-SLC) NAND flash memory for edge servers and resistive RAM (ReRAM) for cloud centralized servers. To achieve short latency and low cost of the non-volatile memory system, distributed multiple write for 3D-SLC NAND flash memory and hierarchical multiple-deck ReRAM are proposed. By using an evaluation platform of the network and non-volatile memory co-design system, the proposed non-volatile memory system is defined as a network bottleneck or memory bottleneck.

Understanding the Memory Window Overestimation of 2D Materials Based Floating Gate Type Memory Devices by Measuring Floating Gate Voltage.

The memory window of floating gate (FG) type non-volatile memory (NVM) devices is a fundamental figure of merit used not only to evaluate the performance, such as retention and endurance, but also to discuss the feasibility of advanced functional memory devices. However, the memory window of 2D materials based NVM devices is historically determined from round sweep transfer curves, while that of conventional Si NVM devices is determined from high and low threshold voltages (Vth s), which are measured by single sweep transfer curves. Here, it is elucidated that the memory window of 2D NVM devices determined from round sweep transfer curves is overestimated compared with that determined from single sweep transfer curves. The floating gate voltage measurement proposed in this study clarifies that the Vth s in round sweep are controlled not only by the number of charges stored in floating gate but also by capacitive coupling between floating gate and back gate. The present finding on the overestimation of memory window enables to appropriately evaluate the potential of 2D NVM devices.

A monolithic artificial iconic memory based on highly stable perovskite-metal multilayers

Artificial iconic memories, also called photomemories, are new types of nonvolatile memory that can simultaneously detect and store light information in a monolithic device. Several approaches have been proposed to construct artificial iconic memories, such as three-terminal field effect transistors, which can achieve an effective control of the gate voltage and external light terminals. The drawbacks in constructing these memories involve complicated fabrication processes, and the resulting performance of, for example, perovskite transistor-type photomemories is limited by the low carrier mobilities and poor ambient stabilities, whereas architectures based on floating gate modulations entail strict interface engineering and poor device reliability. In this paper, we propose a novel monolithic artificial iconic memory with a multilayer architecture of indium tin oxide/perovskite/gold/perovskite/silver, which combines the memory and photodetector functionalities of perovskites in an integrated device. The bottom perovskite layer plays the role of a photodetector, modulating the voltage bias on the top perovskite layer that serves as a resistive switching memory. This multilayer perovskite device can store photo-sensing data in its resistive states, with a memory retention of 5 × 103 s and ambient stability longer than sixty days. As a prototype demonstration, a 7 × 7 artificial iconic memory array is constructed to detect and store data on light intensity distribution, enabling a nonvolatile imaging functionality. Our work provides a new platform for designing perovskite-based architectures with simultaneous light detection and data storage capabilities.

Hole trapping capability of silicon carbonitride charge trap layers

We have evaluated the hole trapping capability of the silicon carbonitride (SiCN) dielectric film for application in metal-oxide-nitride-oxide-silicon (MONOS)-type non-volatile memory devices. After a great number of holes were injected to the SiCN charge trap layer of memory capacitors at high applied voltages, the flat-band voltage shift ΔV fb,h of the capacitors was saturated and the charge centroid location of holes trapped in the SiCN layer was found to reach at 1.8–2.0 nm from the blocking oxide-charge trap layer interface. Using the obtained ΔV fb,h and charge centroid values, the maximum density of holes trapped in the SiCN layer was estimated to be 1.2 × 1013 holes/cm2, which was higher than that trapped in a silicon nitride charge trap layer (=1.0 × 1013 holes/cm2). It is concluded that the high density of trapped holes caused large ΔV fb,h in the memory capacitors with the SiCN layer.

Optimization of MIS type Non-Volatile Memory Device with Al-Doped HfO2 as Charge Trapping Layer

The reduction in physical size of the Non-volatile memory (NVM) demands the use of high-k dielectrics due to loss in charge trapping behavior. The large bandgap of Al2O3 (∼7.0 eV) proves it suitable for blocking and tunneling layer, while the high dielectric constant of HfO2 proves it suitable for charge trapping layer (CTL). In this paper, we propose the application of Al doped HfO2 used as CTL. The doping concentration has been varied by varying the number of sub-cycles (sequential cycles of HfO2, Al2O3) by ALD system. The optical energy bandgap of Al doped HfO2 observed to be increases from 4.91−5.20 eV as the Al incorporation content increases with 1:9−1:4 ALD sub-cycles. Oxygen vacancy defect in the thin film decreased to 17.87% for 1:4-Al doped HfO2 sample as verified from the XPS. The increase in Al incorporation leads to charge loss towards silicon substrate due to high bandgap energy. In addition, charge traps are not well performed due to reduction of internal defects in the film. After 10 years, the NVM device with Al doped HfO2 (1:9) CTL has improved charge retention characteristics by 14% as compared to a normal HfO2 charge trap layer NVM device.

The vacuum arc ion source for indium and tin ions implantation into phase change memory thin films.

One of the most prospective electrical and optical nonvolatile memory types is the phase change memory based on chalcogenide materials, particularly Ge2Sb2Te5. Introduction of dopants is an effective method for the purposeful change of Ge2Sb2Te5 thin film properties. In this work, we used the ion implantation method for the introduction of In and Sn into Ge2Sb2Te5 thin films by a Multipurpose Test Bench (MTB) at the National Research Center "Kurchatov Institute"-Institute for Theoretical and Experimental Physics. For Sn and In ion implantation into Ge2Sb2Te5, the following MTB elements were used: a vacuum arc ion source, an electrostatic focusing system, and a system for current and beam profile measurements. The MTB parameters for Sn and In ion implantation and its effect on the material properties are presented. Implanted Ge2Sb2Te5 thin films were irradiated by femtosecond laser pulses. It was shown that the ion implantation resulted in a decrease in the threshold laser fluence necessary for crystallization compared to the undoped Ge2Sb2Te5.

All-room-temperature processed high flexibility resistive switching memory by ultraviolet-radioactive induction

Alumina-based materials are widely used in semiconductor industry due to the high reliability. However, they are usually employed as high-k gate dielectric rather than memory layer, and high temperature annealing process is also required to maintain the electrical characteristics. These factors enormously limit their further applications in new semiconductor devices such as new type non-volatile memory and flexible electronic devices. In this work, a one-step ultraviolet irradiation method was applied to rapidly achieve AlO–OH thin film at room temperature for flexible resistive memory. This flexible memory device shows satisfied electrical characteristics such as typical unipolar resistive switching behaviors, stable retention, good electrical reliability and mechanical endurance both in flat and bent states. Anion vacancies in AlO–OH layer were contributed to the formation of conductive filament. This technical route with low cost and efficient fabrication process paves a feasible way for developing flexible non-volatile memory and wearable electronics.