Gradual Switching Research Articles

(Invited) Engineering the Switching Kinetics of Valence Change-Based Memristive Devices for Neuromorphic Computing

Memristive devices based on the valence mechanism are highly interesting candidates for the use as hardware representation of synapses in neuromorphic computing. The memristive model system SrTiO3 exhibit gradual switching and can be tuned between short-term and long-term plasticity. We will discuss the impact of the switching kinetics on the gradual switching mode and provide general guidelines for the design of gradual switching systems. Moreover, we present an approach to accelerate the switching kinetics of SrTiO3 by up to 103 times, or reduce the operating voltage by ≈ 30% to maintain the switching speed.Our approach is to introduce a low thermal conductivity layer inside the active electrode of the active electrode of the devices, which blocks the heat dissipation caused by Joule heating during switching. Our method leaves the switching layer and its interfaces with the electrodes intact, while the use of HfO2 and TaOx as the heat blocking layers ensures ease of fabrication and CMOS compatibility. We will demonstrate that this approach is transferable to other more common material systems.

Enhanced Synaptic Characteristics Under Applied Magnetic Field in V2O5/NiMnIn Based Switching Device for Neuromorphic Computing

The present study reports a memory structure Al/V2O5/NiMnIn on a flexible stainless-steel (SS) substrate for neuromorphic applications. The fabricated device exhibits gradual SET and RESET switching characteristics with an OFF/ON resistance ratio of ~100, good consistency of 4500, and excellent data retention capability up to 3000 s. The current-voltage (I-V) study supports an Ohmic conduction mechanism in the low resistance state (LRS). In contrast, the trap-controlled modified space charge conduction mechanism demonstrated the high resistance state (HRS). The resistance versus temperature measurement (R-T) in the LRS and HRS of the device signifies that oxygen vacancies form the conduction filament. We further analyze the synaptic functioning by applying identical consecutive voltage pulses, and the device's conductance change has been observed. These characteristics show a good representation of the biological memory synapse in terms of the artificial memory device. Long-term potentiation (LTP) and long-term depression (LTD) show nonlinear and asymmetery behavior, which is substantial for neuromorphic applications. A considerable shift in LTP and LTD was detected by applying external temperature and magnetic field. This is explained via temperature and magnetic field strain in the functional NiMnIn bottom electrode of the fabricated device. The mechanical flexibility of the memory structure was tested by exploring the switching characteristics with various bending angles and bending cycles. Therefore, the present study offers new avenues for flexible devices with high data storage capability for futuristic neuromorphic applications.

Engineering of TiN/ZnO/SnO2/ZnO/Pt multilayer memristor with advanced electronic synapses and analog switching for neuromorphic computing

The two-terminal memristor is a promising neuromorphic artificial electronic device, mirroring biological synapses in structure and replicating various synaptic functions. Despite its advantages, challenges in achieving high reliability, gradual switching, and low energy consumption hinder progress in neuromorphic devices. This study explores electronic synapses and simulates analog switching in a Pt/TiN/ZnO/SnO2/ZnO/Pt multilayer (ML) configuration, featuring a ∼3 nm SnO2 layer between ZnO layers. Results show enhanced cycling endurance (more than 250 cycles), resistance window (102), tunable synaptic plasticity, and multilevel switching. ML memristors exhibit low coefficient of variation (4.5 %) in set voltage, low energy consumption (set = ∼0.12 nj, reset = ∼0.1 nj), and fast switching speeds (set = 300 ns, reset = 200 ns), suitable for high-density memory and neuromorphic systems. They successfully emulate synaptic functions, including paired-pulse facilitation (PPF), spike voltage-dependent plasticity (SVDP), spike width-dependent plasticity (SWDP), spike frequency-dependent plasticity (SFDP), and post-tetanic potentiation (PTP). Modulating pulse amplitude and width achieves multilevel conductance in long-term potentiation (LTP) and long-term depression (LTD). Using nonlinear conductance data, a 96.5 % image pattern recognition accuracy is achieved in a deconvolution neural network (DNN) simulation. These results highlight the ML memristor's potential in efficient neuromorphic computing systems.

Implementation of 8-bit reservoir computing through volatile ZrOx-based memristor as a physical reservoir

In this study, we employed a sputtering process to construct a memristive device within the ITO/ZrOx/TaN structure for implementing neuromorphic computing. Initially, we scanned the basic electrical properties of the ITO/ZrOx/TaN device using a DC voltage sweep on the top ITO electrode. A highly uniform gradual resistive switching phenomenon was observed over 100 cycles. The current decay in the low-resistance state was effectively controlled by the volatile memory properties. Gradual conductance changes for potentiation and depression were achieved by applying electrical pulses, enabling the establishment of multi-level conductance states. In addition, the emulation of various synaptic functions was achieved by following the learning rules of SRDP, EPSC, STDP, ADSP, Pavlovian associative learning, and PPF. Finally, 8-bit reservoir computing was demonstrated in cost-effective pattern generation and recognition, highlighting the ITO/ZrOx/TaN device’s advantageous memory storage properties for synaptic characteristics.

Evaluation of Imprint and Multi‐Level Dynamics in Ferroelectric Capacitors

AbstractFluorite‐structured ferroelectrics are one of the most promising material systems for emerging memory technologies. However, when integrated into electronic devices, these materials exhibit strong imprint effects that can lead to a failure during writing or retention operations. To improve the performance and reliability of these devices, it is cardinal to understand the physical mechanisms underlying the imprint during operation. In this work, the comparison of First‐Order Reversal Curves measurements with a new gradual switching experimental approach named “Unipolar Reversal Curves” is used to analyze both the fluid imprint and the time‐dependent imprint effects within a 10 nm‐thick Hf0.5Zr0.5O2 capacitor. Interestingly, the application of delay times (ranging from 100 µs up to 10 s) between the partial switching pulses of a Unipolar Reversal Curve sequence enables analysis of the connection between the two aforementioned imprint types. Based on these results, the study finally reports a unified physical interpretation of imprint in the context of a charge injection model, which explains both types of imprint and sheds light on the dynamics of multi‐level polarization switching in ferroelectrics.

Effect of RF power on analog synaptic behavior of sputter-deposited InGaZnO films for neuromorphic computing applications

We demonstrate that an analog synaptic weight update can be achieved using identical pulses via gradual resistance switching of memristors based on sputter-deposited InGaZnO (IGZO) films. When a positive voltage is applied to the IGZO memristor, oxygen vacancies (Vo) are ionized and electrons that serve as dopants are generated. This lowers the Schottky barrier formed at the interface and decreases the resistance. However, a negative voltage provides electrons from the Pd electrode, annihilating the ionized vacancies. As the barrier is reconstructed, the IGZO memristor exhibits a high-resistance state (HRS). As Vo plays a crucial role in switching, the effect of the radio frequency (RF) power on the memristor during the IGZO deposition is examined. A physical analysis reveals that the Vo concentration in the IGZO is proportional to the RF power, inducing a leaky HRS. However, at the expense of a reduced memory window in the IGZO deposited at a higher power of 200 W, the resultant low-resistance state is reliable over time. The strengthened retention allows the synaptic weights to be linearly decreased even by consecutive negative pulses. Hence, when the IGZO memristors are used for synaptic elements to construct neural networks, a higher recognition accuracy for the MNIST dataset is achieved.

Artificial Synapse Based on a δ-FAPbI3/Atomic-Layer-Deposited SnO2 Bilayer Memristor.

Halide perovskite-based resistive switching memory (memristor) has potential in an artificial synapse. However, an abrupt switch behavior observed for a formamidinium lead triiodide (FAPbI3)-based memristor is undesirable for an artificial synapse. Here, we report on the δ-FAPbI3/atomic-layer-deposited (ALD)-SnO2 bilayer memristor for gradual analogue resistive switching. In comparison to a single-layer δ-FAPbI3 memristor, the heterojunction δ-FAPbI3/ALD-SnO2 bilayer effectively reduces the current level in the high-resistance state. The analog resistive switching characteristics of δ-FAPbI3/ALD-SnO2 demonstrate exceptional linearity and potentiation/depression performance, resembling an artificial synapse for neuromorphic computing. The nonlinearity of long-term potentiation and long-term depression is notably decreased from 12.26 to 0.60 and from -8.79 to -3.47, respectively. Moreover, the δ-FAPbI3/ALD-SnO2 bilayer achieves a recognition rate of ≤94.04% based on the modified National Institute of Standards and Technology database (MNIST), establishing its potential in an efficient artificial synapse.

A Diffusive Artificial Synapse Based on Charged Metal Nanoparticles.

We show that a diffusive memristor with analogue switching characteristics can be achieved in a layer of gold nanoparticles (AuNPs) functionalized with charged self-assembled monolayers (deprotonated 11-mercaptoundecanoic acid). The nanoparticle core and the anchored stationary charges are jammed within the layer while the mobile counterions [N(CH3)4+] can respond to the electric field and spontaneously diffuse back to the initial positions upon removal of the field. This metal nanoparticle device is set-step free, energy consumption efficient, mechanically flexible, and analogous to bio-Ca2+ dynamics and has tunable conductance modulation capabilities at the counterion concentrations. The gradual resistive switching behavior enables us to implement several important synaptic functions such as potentiation/depression, spike voltage-dependent plasticity, spike duration-dependent plasticity, spike frequency-dependent plasticity, and paired-pulse facilitation. Finally, on the basis of the paired-pulse facilitation characteristics, the metal nanoparticle diffusive artificial synapse is used for edge extraction with exhibits excellent performance.

Unraveling the origins of the coexisting localized-interfacial mechanism in oxide-based memristors in CMOS-integrated synaptic device implementations.

The forefront of neuromorphic research strives to develop devices with specific properties, i.e., linear and symmetrical conductance changes under external stimuli. This is paramount for neural network accuracy when emulating a biological synapse. A parallel exploration of resistive memory as a replacement for conventional computing memory ensues. In search of a holistic solution, the proposed memristive device in this work is uniquely poised to address this elusive gap as a unified memory solution. Opposite biasing operations are leveraged to achieve stable abrupt and gradual switching characteristics within a single device, addressing the demands for lower latency and energy consumption for binary switching applications, and graduality for neuromorphic computing applications. We evaluated the underlying principles of both switching modes, attributing the anomalous gradual switching to the modulation of oxygen-deficient layers formed between the active electrode and oxide switching layer. The memristive cell (1R) was integrated with 40 nm transistor technology (1T) to form a 1T-1R memory cell, demonstrating a switching speed of 50 ns with a pulse amplitude of ±2.5 V in its forward-biased mode. Applying pulse trains of 20 ns to 490 ns in the reverse-biased mode exhibited synaptic weight properties, obtaining a nonlinearity (NL) factor of <0.5 for both potentiation and depression. The devices in both modes also demonstrated an endurance of >106 cycles, and their conductance states were also stable under temperature stress at 85 °C for 104 s. With the duality of the two switching modes, our device can be used for both memory and synaptic weight-storing applications.

On-receptor computing with classical associative learning in semiconductor oxide memristors.

The increasing demand for energy-efficient data processing leads to a growing interest in neuromorphic computing that aims to emulate cerebral functions. This approach offers cost-effective and rapid parallel data processing, surpassing the limitations of the conventional von Neumann architecture. Key to this emulation is the development of memristors that mimic biological synapses. Recently, research efforts have focused on the incorporation of nociceptors-sensory neurons capable of detecting external stimuli-into memristors for applications in robotics and artificial intelligence. This integration enables memristors to adapt to various circumstances while remaining cost-effective. A nonfilamentary gradual resistive switching memristor is utilized to implement artificial nociceptor and synaptic behaviors. The fabricated Pt/indium gallium zinc oxide (IGZO)/SnOx/TiN device exhibits essential properties of biological nociceptors, including threshold response, no-adaptation, relaxation, sensitization, and recovery. Furthermore, the device leverages short-term memory principles to emulate learning behaviors observed in the brain by showcasing "forgetting" paradigms. Additionally, control of the input spikes yields different synaptic plasticity responses, thus emulating the key functions of our synapse. Computational simulations demonstrate the device's ability to perform both computing and sensing tasks effectively, thus enabling on-receptor computing with associative learning capabilities.

Experimental demonstration and analysis of crossbar array memristor for brain-inspired computing

The human brain accomplishes intricate computational tasks like learning, recognition, and cognition with minimal power usage, and showcases synaptic behavior well-suited for neuromorphic computing systems. The latest advancements in memristive devices are pivotal for designing memory and synapses, as well as for non-volatile data storage and computing. However, a directive on realizing memristive analog behavior is still a major concern. Here we propose forming free digital and analog resistive switching (ARS) dynamics of 4 × 4 crossbar array MgZnO-based memristive devices for neural activity (Potentiation/Depression). Moreover, our device demonstrates both abrupt and gradual resistive switching (RS) behaviors, with acceptable endurance (10 K cycles at 85 °C), data retention (107 s), a read voltage range of 0.1 V to 0.5 V, and pulse widths ranging from 50 μs to 250 μs (with a pulse interval of 50 μs) for synaptic behavior for neuromorphic computing. Our approach deals with multifunctional memory and synaptic behavior with low voltage linearity and excellent switching characteristics. Our results highlight the potential of memristor for energy-efficient edge computing and brain-inspired computing systems.

Formation of Cluster-Structured Metallic Filaments in Organic Memristors for Wearable Neuromorphic Systems with Bio-Mimetic Synaptic Weight Distributions.

With increasing demand for wearable electronics capable of computing huge data, flexible neuromorphic systems mimicking brain functions have been receiving much attention. Despite considerable efforts in developing practical neural networks utilizing several types of flexible artificial synapses, it is still challenging to develop wearable systems for complex computations due to the difficulties in emulating continuous memory states in a synaptic component. In this study, polymer conductivity is analyzed as a crucial factor in determining the growth dynamics of metallic filaments in organic memristors. Moreover, flexible memristors with bio-mimetic synaptic functions such as linearly tunable weights are demonstrated by engineering the polymer conductivity. In the organic memristor, the cluster-structured filaments are grown within the polymer medium in response to electric stimuli, resulting in gradual resistive switching and stable synaptic plasticity. Additionally, the device exhibits the continuous and numerous non-volatile memory states due to its low leakage current. Furthermore, complex hardware neural networks including ternary logic operators and a noisy image recognitions system are successfully implemented utilizing the developed memristor arrays. This promising concept of creating flexible neural networks with bio-mimetic weight distributions will contribute to the development of a new computing architecture for energy-efficient wearable smart electronics.

Eightwise Switching Mechanism in Memristive SrTiO3Devices and Its Implications on the Device Performance

Filamentary‐type resistive switching devices based on the valence change mechanism (VCM) have been studied extensively during the past decades, due to their great performance as nonvolatile memories and their promising nature for novel computing architectures. For many materials exhibiting the VCM, two different polarities can be observed, namely, counter‐eightwise and eightwise switching. The more technologically relevant materials, such as TaOxand HfOx, usually exhibit the counter‐eightwise resistive switching mode, however, the nature of its abrupt switching limits the potential for their implementation in more analog systems. On the contrary, the eightwise resistive switching mode exhibits a more gradual switching profile, which makes it a more suitable candidate for neuromorphic and in‐memory computing applications. Contrary to its counterpart, the eightwise switching mode has not been fully understood for several years, hampering the improvement of memristive devices based on this mode. This review presents the major advancements that have been made to understand the mechanism behind the eightwise switching mode and its implications on the properties of the memristive devices. The works presented here are based on the model system of SrTiO3, however, the obtained knowledge and the strategies to tailor their properties can be transferred to other technologically relevant systems.

Uniform multilevel switching and synaptic properties in RF-sputtered InGaZnO-based memristor treated with oxygen plasma.

Bipolar gradual resistive switching was investigated in ITO/InGaZnO/ITO resistive switching devices. Controlled intrinsic oxygen vacancy formation inside the switching layer enabled the establishment of a stable multilevel memory state, allowing for RESET voltage control and non-degradable data endurance. The ITO/InGaZnO interface governs the migration of oxygen ions and redox reactions within the switching layer. Voltage-stress-induced electron trapping and oxygen vacancy formation were observed before conductive filament electroforming. This device mimicked biological synapses, demonstrating short- and long-term potentiation and depression through electrical pulse sequences. Modulation of post-synaptic currents and pulse frequency-dependent short-term potentiation were successfully emulated in the InGaZnO-based artificial synapse. The ITO/InGaZnO/ITO memristor exhibited spike-amplitude-dependent plasticity, spike-rate-dependent plasticity, and potentiation-depression synaptic learning with low energy consumption, making it a promising candidate for large-scale integration.

Investigation of resistive switching dynamics in e-beam evaporated P-type tin-oxide based cross-cell memristor for synaptic and memory application

In this letter, we have examined the gradual and abrupt resistive switching (RS) dynamics in E-beam evaporated P-type Tin-oxide based cross-cell memristor by adjusting the DC voltage sweep, and it has been used for synaptic and memory application, respectively. We have observed gradual RS dynamics at low DC sweeping voltage (+0.5 V/-0.5 V) and we have also studied its synaptic behavior (Potentiation/Depression) at different read voltage and pulse widths to optimize its synaptic performance. As well as we have also observed abrupt RS dynamics at high DC sweeping voltage (+1.5 V/-1.5 V) and utilized it as a memory element. The proposed device shows a good Switching window (∼102) up to 20 K times read/write operation without any considerable degradation at an optimized read voltage (0.5 V) and we have also studied data retention property for 105 sec at 85 °C.

Effectiveness and safety of lubiprostone after switching from stimulant laxatives in elderly patients with chronic constipation.

Stimulant laxatives may cause electrolyte abnormalities, dehydration, and abdominal pain; their long-term use can lead to tolerance and subsequent refractory constipation. We investigated the effectiveness, safety, and quality of life after switching from stimulant laxatives to lubiprostone in elderly patients with chronic constipation (CC). This multicenter, interventional, open-label, single-arm, before-and-after comparison study enrolled 99 Japanese patients aged 65-90 years with CC who took stimulant laxatives for ≥2 weeks prior to switching to lubiprostone monotherapy. The mean ± SD spontaneous defecations at Week 1 of 7.8 ± 6.2 times/week was not significantly different from that at baseline (8.3 ± 4.7). Spontaneous defecations were significantly reduced at Weeks 2 (-1.5 ± 4.0, P < 0.001) and 4 (-1.5 ± 3.7, P < 0.001). The Bristol Stool Form Scale score did not change from baseline (4.7 ± 0.9) at Weeks 1 (4.5 ± 1.3) or 4 (4.3 ± 1.3), but it did at Week 2 (4.3 ± 1.5, P < 0.05). The Patient Assessment of Constipation Quality of Life questionnaire score increased (0.36 ± 0.07, P < 0.001) after 28 days. Nausea was the only symptom that worsened from baseline and was the most frequently reported adverse drug reaction (15.2%). Switching to lubiprostone monotherapy for CC was not associated with significant concerns in short-term spontaneous defecation frequency and safety, but it might affect the efficacy and patient quality of life over 2 weeks. Careful treatment strategies facilitating gradual switching to lubiprostone monotherapy may be needed in patients using stimulant laxatives.

Enhanced Synaptic Characteristics under Applied Magnetic Field in V2O5/NiMnIn-Based Switching Device for Neuromorphic Computing

The present study reports a memory structure Al/V2O5/NiMnIn on a flexible stainless steel (SS) substrate for neuromorphic applications. The fabricated device exhibits gradual SET and RESET switching characteristics with an OFF/ON resistance ratio of ∼100, good consistency of 4500, and excellent data retention capability up to 3000 s. The current–voltage (I–V) study supports an Ohmic conduction mechanism in the low-resistance state (LRS). In contrast, the trap-controlled modified space charge conduction mechanism demonstrated the high-resistance state (HRS). The resistance vs temperature measurement (R–T) in the LRS and HRS of the device signifies that oxygen vacancies form the conduction filament. We further analyze the synaptic functioning by applying identical consecutive voltage pulses, and the device’s conductance change has been observed. These characteristics show a good representation of the biological memory synapse in terms of the artificial memory device. Long-term potentiation (LTP) and long-term depression (LTD) show nonlinear and asymmetry behavior, which is substantial for neuromorphic applications. A considerable shift in LTP and LTD was detected by applying external temperature and magnetic field. This is explained via temperature and magnetic field strain in the functional NiMnIn bottom electrode of the fabricated device. The mechanical flexibility of the memory structure was tested by exploring the switching characteristics with various bending angles and bending cycles. Therefore, the present study offers new avenues for flexible devices with high data storage capability for futuristic neuromorphic applications.

Hidden phases in homovalent and heterovalent substituted BaTiO3

Ferroelectric materials can exhibit metastable phases when exposed to THz pulses, characterized by a polarization integration capability related to the amplitude and frequency of the pulses. These so-called ``hidden'' phases enable gradual switching of polarization that can be utilized in artificial synapses for nonconventional (neuromorphic) computing machines. In this work, we employ large-scale molecular-dynamics simulation based on an effective Hamiltonian approach, and we report on the discovery of hidden phases in Zr- and Nb-doped barium titanate ($\mathrm{BaTi}{\mathrm{O}}_{3}$). We investigate the formation and the stability of those phases at different stimuli and temperatures (20 and 200 K). Our results shed light on the compositional dependence of the properties of these phases, demonstrating the potential of lead-free relaxor ferroelectrics for near-room-temperature neuromorphic computing.

Collective Control of Potential-Constrained Oxygen Vacancies in Oxide Heterostructures for Gradual Resistive Switching.

Filamentary resistive switching in oxides is one of the key strategies for developing next-generation non-volatile memory devices. However, despite numerous advantages, their practical applications in neuromorphic computing are still limited due to non-uniform and indeterministic switching behavior. Given the inherent stochasticity of point defect migration, the pursuit of reliable switching likely demands an innovative approach. Herein, a collective control of oxygen vacancies is introduced in LaAlO3 /SrTiO3 (LAO/STO) heterostructures to achieve reliable and gradual resistive switching. By exploiting an electrostatic potential constraint in ultrathin LAO/STO heterostructures, the formation of conducting filaments is suppressed, but instead precisely control the concentration of oxygen vacancies. Since the conductance of the LAO/STO device is governed by the ensemble concentration of oxygen vacancies, not their individual probabilistic migrations, the resistive switching is more uniform and deterministic compared to conventional filamentary devices. It provides direct evidence for the collective control of oxygen vacancies by spectral noise analysis and modeling by Monte-Carlo simulation. As a proof of concept, the significantly-improved analog switching performance of the filament-free LAO/STO devices is demonstrated, revealing potential for neuromorphic applications. The results establish an approach to store information by point defect concentration, akin to biological ionic channels, for enhancing switching characteristics of oxide materials.

Dual-pulse disturb-free programming scheme for FeFET based neuromorphic computing

With the compatibility and scalability of Hf0.5Zr0.5O2 (HZO) film, the HZO-based Ferroelectric field-effect transistors (FeFETs) have attracted extensive attention as artificial synapses to break the Von Neumann bottleneck. For neuromorphic computing, improved synaptic features and suppressed sneak paths are of great importance in a large-scale FeFET array. In this work, a dual-pulse programming scheme is proposed with an identical VG pulse and an incremental VD pulse. The gradual domain switching is achieved corresponding to the potentiation and depression process. Moreover, both write-disturb and read-disturb are investigated under 50 successive pulses. With the proposed dual-pulse programming scheme, 93.8% accuracy of image recognition is achieved in artificial tasks by using the MNIST dataset. Our findings may provide potential for developing HZO-FeFET-based neuromorphic computing.