Uniformity and Retention Improvement of TaOx-Based Conductive Bridge Random Access Memory by CuSiN Interfacial Layer Engineering

Zinc oxide (ZnO) is an amorphous oxide semiconductor (AOS) material valued for its high carrier mobility, transparency to visible light, and low-temperature manufacturing process compared to silicon-based semiconductors. These desired features make ZnO-based AOSs widely popular as the channel layer in TFTs and as key dielectric layers in electronic devices like light sensors, gas detectors, resistive random-access memory (RRAM), and more. Among them, Conductive Bridge Random Access Memory (CBRAM) with a dielectric switching layer sandwiched between the electrochemically active metal (Ag, Cu) top electrode and the inert bottom electrode, is of a class of RRAM that offers low-voltage operation, high-density storage device, non-volatile, and compatibility with standard CMOS processes [1]. The CBRAM operates based on the principle of conductive filament formation and rupture. During operation, conductive filaments will be formed in switching layer by connecting metal ions which diffuse from the electrode after applying a positive bias. The device will switch to a low resistance state (LRS) after the conductive filaments were formed. After this, applying a negative bias will break the conductive filaments and switch the device back to a high resistance state (HRS). The HRS and LRS are the 0 or1 state of the memory operation. Studies and works have shown that controlling the distribution and diffusion of electrode metal ions in the switching layer is challenging because the conductive filaments formed can vary significantly in strength, ranging from robust to fragile [2]. This leads to instability in the operation of memory devices and encountering difficulties in the integration with TFT pixel array for memory-in-pixel (MiP) power-saving displays [3] or CMOS-integrated compute-in-memory applications [4].In this work, we proposed a novel CBRAM device featuring ZnO nanostructures as the resistive switching layer. The combination of a sputter-deposited seed layer with the water bath method enables the production of high-quality ZnO nanostructures at temperatures below 90°C. These nanostructures play a crucial role in constraining metal ion diffusion and filament formation, leading to the formation of uniform conductive filaments. This improvement significantly enhances the operational stability of the CBRAM device. The experimental results have demonstrated superior memory characteristics with a highly uniform distribution achieved successfully in the ZnO nanostructure-embedded CBRAM, compared to its counterpart lacking ZnO nanostructures. Furthermore, post-treatments are implemented in this study to further promote electrical performance of the nanostructure-embedded CBRAM, like operation voltage, memory window, retention, and endurance. Reference: [1] C. Gopalan, Y. Ma, T. Gallo, J. Wang, E. Runnion, J. Saenz, F. Koushan, P. Blanchard, S. Hollmer, “Demonstration of Conductive Bridging Random Access Memory (CBRAM) in logic CMOS process”, Solid-State Electronics, vol. 58 (1), p. 54-61, 2011.[2] C. C. Hsu, P. T. Liu, K. J. Gan, D. B. Ruan, Simon M. Sze, “Investigation of deposition technique and thickness effect of HfO2 film in bilayer InWZnO-based conductive bridge random access memory”, Vacuum, vol. 201, 111123, 2022.[3] S. H. Lee, B. C. Yu, H. J. Chung, and S. W. Lee, “Memory-in-pixel circuit for low-power liquid crystal displays comprising oxide thin-filmtransistors,” IEEE Electron Device Lett., vol. 11 (11), pp. 1551–1554, 2017.[4] W. Wan, R. Kubendran, C. Schaefer, et al., “A compute-in-memory chip based on resistive random-access memory”, Nature, vol. 608, p. 504–512, 2022.

In resistive random access memory devices (RRAM), materials selection plays a critical role in the tuning of properties like switching voltages, retention, and endurance. Aside from choosing between transition metal-oxide matrices like HfOx, TiOx, TaOx, NiOx, and AlOx, doping in the oxide can also be a powerful tool for device optimization. However, only a handful of dopants have been studied in detail [1-3], and it is currently not known which will offer the greatest benefit for device performance. The filamentary switching mechanism in RRAM devices, in which charged oxygen vacancies (VO’s) coalesce into conductive filaments during switching, has been widely documented [4-9]. Dopants have been shown to affect the switching characteristics of RRAM devices by decreasing the formation energy of VO’s and potentially stabilizing conductive filaments [1-3]. In this investigation, first principles modeling is used to characterize 42 possible dopants in HfOx. In addition to quantifying their effects on HfOx devices, trends are extracted which can serve as general guidelines for doping in any transition metal-oxide, and which are applicable to conductive bridge random access memories (CBRAM) as well. This work identifies a critical new metric for predicting dopant behavior in RRAM and CBRAM: the relative favorability of the dopant forming on substitutional or interstitial sites in the oxide lattice. Dopants that form on interstitial sites, such as Ag, Cu, and Ni, also tend to form cation filaments and thus CBRAM devices. Conversely, dopants that strongly prefer to substitute onto the Hf lattice site are unlikely to produce competing cation filaments, leaving the VO-based switching mechanism intact. The relative favorability is also independent of dopant chemical potential, allowing a direct one-to-one comparison of different dopant species in HfOx. In this work, the quantity of relative stability is formulated and quantified over all 42 cation dopants. The results are shown in Fig. 1. The impact of these dopants on VO formation energy is also analyzed, as shown in Fig. 2. The effect of dopants on device properties is strongly dependent upon dopant species, and Fig.’s 2 and 3 clearly show that these effects follow predictable trends. Finally, the trends underlying dopant behavior are explored, and a model is developed to predict dopant behavior in the oxide system. The behavior of the dopant was found to vary predictably as a function of six key quantities: (1) the valence difference between the dopant and Hf, (2) the dopant atomic radius, (3) the dopant oxide enthalpy of formation, (4) the change in dopant coordination from native oxide to HfOx, (5) the magnetization of the doped system, and (6) the perturbation of neighboring ion charge states by the dopant. The derived composite function predicts the relative dopant site favorability with 95% accuracy. Since these quantities can all be calculated for different oxides, it is expected that these trends can be used as guidelines for doping in any transition metal-oxide for both RRAM and CBRAM applications. [1] D. Duncan, B. Magyari-Köpe, and Y. Nishi, Appl. Phys. Lett. 108, 043501, 2016. [2] L. Zhao, S. W. Ryu, A. Hazeghi, D. Duncan, B. Magyari-Köpe, and Y. Nishi, Tech. Digest VLSI Tech. Symp., T106-T107, 2013. [3] D. Ning, P. Hua, and W. Wei, Chin. Phys. B 23(10), 107306, 2014. [4] D. Duncan, B. Magyari-Köpe, and Y. Nishi, Elec. Dev. Lett. 37 (4), 400-403, 2016. [5] B. Magyari-Köpe, S.G. Park, H.-D. Lee, and Y. Nishi, J. Mater. Sci. 47, pp. 7498-7514, 2012. [6] S.G. Park, B. Magyari-Köpe, and Y. Nishi, Phys. Rev. B, 82, 115109, 2010. [7] H.-D. Lee, B. Magyari-Köpe, and Y. Nishi, Phys. Rev. B 81, 193202, 2010. [8] S.G. Park, B. Magyari-Köpe, and Y. Nishi, Tech. Digest VLSI Tech. Symp., 2011. [9] B. Magyari-Köpe, M. Tendulkar, S.G. Park, H.-D. Lee, and Y. Nishi, Nanotechnology 22, 254029, 2011. Figure 1

Here, we present the synaptic characteristics of AlN-based conductive bridge random access memory (CBRAM) as a synaptic device for neuromorphic systems. Both non-volatile and volatile memory are observed by simply controlling the strength of the Cu filament inside the AlN film. For non-volatile switching induced by high compliance current (CC), good retention with a strong Cu metallic filament is verified. Low-resistance state (LRS) and high-resistance state (HRS) conduction follow metallic Ohmic and trap-assisted tunneling (TAT), respectively, which are supported by I–V fitting and temperature dependence. The transition from long-term plasticity (LTP) to short-term plasticity (STP) is demonstrated by increasing the pulse interval time for synaptic device application. Also, paired-pulse facilitation (PPF) in the nervous system is mimicked by sending two identical pulses to the CBRAM device to induce STP. Finally, potentiation and depression are achieved by gradually increasing the set and reset voltage in pulse transient mode.

Conductive-bridging random access memory (CBRAM) using a cobalt (Co) electrode has recently featured a CMOS-compatible process, excellent data retention, and a sub-μA operating current level, which are difficult to achieve by conventional CBRAM. However, the resistive switching (RS) mechanism of Co CBRAM has not been extensively explored compared to that of the conventional CBRAM cells using Ag, Cu, or Ni as active metal electrodes. Because only implicit inferences based on electrical measurements have been made, the formation of Co filaments is not yet clearly understood. This study presents evidence of Co filament formation in Co/10 nm SiOx/TiN resistive random access memory (RRAM) using direct transmission electron microscopy (TEM) observations. Co protrusions larger than 5 nm are observed, and their exact atomic composition is investigated by spectroscopy. We explain the RS operation of Co/SiOx/TiN cells based on the Co electromigration-mediated filament development process through Co electrode deformation, protrusion, and subsequent Co conductive bridge formation. An asymmetric RS operation and negative temperature correlation of the resistance are found in low resistance state (LRS) cells, which further supports the Co-involved RS operation in Co/SiOx/TiN devices.

In this paper, we investigate the quantized conduction behavior of conductive bridge random access memory (CBRAM) with varied materials and ramping rates. We report stable and reproducible quantized conductance states with integer multiples of fundamental conductance obtained by optimizing the voltage ramping rate and the Ti-diffusion barrier (DB) at the Cu/HfO2 interface. Owing to controlled diffusion of Cu ions by the Ti-DB and the optimized ramping rate, through which it was possible to control the time delay of Cu ion reduction, more than seven levels of discrete conductance states were clearly observed. Analytical modeling was performed to determine the rate-limiting step in filament growth based on an electrochemical redox reaction. Our understanding of the fundamental mechanisms of quantized conductance behaviors provide a promising future for the multi-bit CBRAM device.

Large parameter variability is identified as one of the main problems in conductive bridging random access memory (CBRAM) at the device level impeding large-scale commercial applications. Here, we ...

In this letter, we demonstrate a self-limited conductive-bridging random access memory (CBRAM) that removes the necessity for external current compliance in a one selector–one resistor (1S1R) architecture. The standard Ge2Sb2Te5 (GST) is used as a CBRAM switching layer. In addition, Te-rich GST is also considered. Their performance is then compared. Both samples exhibit self-limited on-current characteristics, and the on-currents of the standard GST and Te-rich GST are ~300 and $\sim 20~\mu \text{A}$ , respectively. The observed self-limited characteristics are caused by the Te in the GST layer because in the presence of Te, Cu tends to form a more stable CuTe phase that restrict Cu filament growth. Furthermore, to confirm the feasibility of crossbar array applications, the 1S1R device is evaluated using a Ag/TiO2-based threshold selector device reported in our previous work. Hence, we confirm leakage current reduction, a uniform resistance distribution, and stable retention characteristics in the 1S1R configuration with no external current compliance.

Filamentary-type resistive switching devices, such as conductive bridge random-access memory and valence change memory, have diverse applications in memory and neuromorphic computing. However, the randomness in filament formation poses challenges to device reliability and uniformity. To overcome this issue, various defect engineering methods have been explored, including doping, metal nanoparticle embedding, and extended defect utilization. In this study, we present a simple and effective approach using self-assembled uniform Au nanoelectrodes to controll filament formation in HfO2 resistive switching devices. By concentrating the electric field near the Au nanoelectrodes within the BaTiO3 matrix, we significantly enhanced the device stability and reduced the threshold voltage by up to 45% in HfO2-based artificial neurons compared to the control devices. The threshold voltage reduction is attributed to the uniformly distributed Au nanoelectrodes in the insulating matrix, as confirmed by COMSOL simulation. Our findings highlight the potential of nanostructure design for precise control of filamentary-type resistive switching devices.

Resistive random access memory (RRAM) is a prime candidate to replace Flash memory. Of the two classes of RRAM, conductive bridge RAM (CBRAM) is favored over that based on filaments of oxygen vacancies because of its larger on/off resistance ratio. The nature of the filament in Cu/Al2O3-based CBRAM is analyzed using density functional theory. The defect and binding energies of Cu interstitials and clusters in Al2O3 are calculated. The binding energy per Cu interstitial is shown to significantly increase with increasing Cu coordination, whereas the binding per oxygen vacancy only slightly increases with vacancy concentration. This explains why metal filaments in CBRAM devices tend to be denser than oxygen vacancy filaments. Using three different filament models, we discover that the strong binding between Cu interstitials drives filament formation, resulting in Al ions being driven out of the Cu-rich environment. This leads to the formation of densely packed metallic Cu filaments with bonding similar to Cu...

The threshold switching effect is considered of outmost importance for a variety of applications ranging from the reliable operation of crossbar architectures to emulating neuromorphic properties with artificial neural networks. This property is strongly believed to be associated with the rich inherit dynamics of a metallic conductive filament (CF) formation and its respective relaxation processes. Understanding the origin of these dynamics is very important in order to control the degree of volatility and design novel electronic devices. Here, we present a synergistic numerical and experimental approach in order to deal with that issue. The distribution of relaxation time is addressed through time-resolved pulse measurements whereas the entire switching behavior is modeled through a 2D dynamical model by taking into account the destructive interference of the drift/diffusion transport mechanisms and the Soret diffusion flux due to the intense local Joule heating. The proposed mechanism interprets successfully both the threshold to bipolar switching transition as well as the self-rectifying effects in SiO2-based memories. The model incorporates the effect of electrode materials on the switching pattern and provides a different perception of the ionic transport processes, shading light into the ultra-small lifetimes of the CF and explaining the different behavior of the silver or copper active materials in a conductive bridge random access memory architecture.

The variation of resistive switching (RS) parameters is one of the major challenges to push resistive random access memory (RRAM) into application. In this paper, with a statistical methodology, for the first time we study the statistical distributions of set parameters of the Cu/HfO <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf> /Pt based conductive bridge random access memory (CBRAM). The distributions of the set parameters in each off resistance range are shown to nicely fit a Weibull model. The Weibull slope and scale factor of the set voltage increase logarithmically with the geometric mean of off resistance in different ranges, which can be well explained by the cell-based percolation model of set statistics. And also for the first time, we revel that the Weibull slope of the set current decreases logarithmically with the off resistance. Our study is of significance to reduce the fluctuation of the resistive switching parameters.

The impact of TiW barrier layer thickness dependent transition from electro-chemical metallization memory to valence change memory in ZrO2-based resistive switching random access memory devices

Control of filament growth within conductive bridge random access memories (CBRAM) is of crucial interest in order to ensure the reliability of such emerging devices. Here, we demonstrate that Scanning Joule Expansion Microscopy (SJEM) can be used to detect and precisely localize conductive filaments within operating crossbar CBRAM devices. Flexible memory devices based on Pd/Al2O3/Ag stacks are first elaborated at low temperature on polyimide substrate. These devices show set and reset operations at low voltage ( $4.3~\mu \text{m}$ is extracted at 50kHz and it is also demonstrated that the thermal expansion signal is proportional to the dissipated Joule power. We believe that the proposed procedure opens the way to reliability studies that can be applied to any family of memory device based on filamentary conduction.

Resistively switching Conductive Bridge Random Access Memories (CBRAMs) rely on the controlled formation and dissolution of metallic filaments within a solid insulator, and are emerging building blocks for beyond von Neumann computing architectures and neuromorphic computing. A lack of understanding of the underlying switching mechanisms currently prevents further device utilisation and optimisation. We present a study of lateral and vertical CBRAM model devices that allow us to systematically relate important switching properties and their statistics to a direct characterisation of their critical switching region by scanning and transmission electron microscopy, i.e. to the physical nature of metal filaments and inclusions on all relevant length scales. We find that filaments are composed of metallic clusters and show how filament dynamics link to migration effects of embedded nanoparticles under voltage bias stress. The formation of metal clusters is promoted by a dynamic interplay of cation mobility and redox rate during switching. These clusters are not completely removed from the switching material matrix upon RESET and appear to grow by consumption of smaller clusters. We discuss in detail the interfacial stress of the nanoparticles in the context of their interaction with the switching material and ambient atmosphere. This allows us to consistently interpret previous literature and to suggest future device improvements.

The development of artificial nerve systems is considered of great significance for the fabrication of artificial appendages or intelligent robotics with environmental communication and awareness. Along these lines, in this work, the synaptic properties of a SiO <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$_{\text{2}}$</tex-math> </inline-formula> -based conductive bridge random access memory (CBRAM) were thoroughly investigated. Initially, a detailed study was conducted to understand and catalog the short-and long-term properties of the device, and their underlying physical mechanisms. To that end, a numerical model was introduced for interpreting the synaptic pattern within the device and its results were validated with the acquired experimental data. In addition, an artificial afferent nerve concept was demonstrated, comprising a piezoelectric sensory receptor, to generate electrical signals in response to either impulsive (“touch and go”) or continuous external stimuli (“touch and stay”), a microcontroller-emulated leaky-integrate-and-fire neural node converting the signals into spiking action potentials, and the CBRAM element as a terminal synapse, responsible for processing/attenuating the action potentials Throughout these procedures, the synaptic adaptation and the memory consolidation processes were recorded and analyzed.