ABSTRACT In‐sensor processing is a key enabler for edge computing because it allows sensor data to be processed locally, thereby reducing data movement and improving the efficiency of complex workloads. In particular, emerging computationally intensive applications such as explainable artificial intelligence (XAI) further increase the demand for dedicated in‐sensor hardware capable of efficient sensor‐level processing. Here, we report an optoelectronic memtransistor (OEMT) that integrates a photoresponsive indium zinc oxide (IZO) channel with a NbO x /AlO x charge‐trapping gate stack, enabling optical sensing, electrical masking, and non‐volatile memory within a single three‐terminal device architecture. We experimentally demonstrate sensor‐level vision explainable artificial intelligence (VXAI) operation using a 3 × 3 OEMT array and validate its scalability through hardware‐based simulations. Compared with a conventional sensing–processing pipeline, the OEMT‐based framework achieves a 21‐fold improvement in system‐level energy efficiency, highlighting its strong potential for efficient sensor‐level processing.

Probabilistic computing is an emerging approach for solving combinatorial optimization and probabilistic inference by leveraging stochastic dynamics to explore complex energy landscapes. However, most hardware implementations remain constrained to binary probabilistic bits (p-bits), limiting efficiency for inherently multi-class problems. Here, we propose and experimentally demonstrate a ternary Potts machine (TPM) enabled by a NbOx oscillator-based probabilistic ternary digit unit (i.e., p-trits). The p-trit exploits electro-thermal dynamics of a NbOx threshold-switching memristor that exhibits two distinct negative differential resistance regimes, producing stochastic transitions among three discrete states (off/osc/on). A compact p-trit circuit digitizes the oscillatory signals into two-bit outputs, yielding a practical p-trit with an operation rate of 2.5 µs/bit, while the NbOx oscillator, which is the primary source of energy consumption, consumes 2.4 nJ/bit. A microcontroller unit -integrated TPM prototype solves Max-3-Cut with ∼30% higher accuracy than p-bit-based computing while requiring ∼33% of its computational overhead. Simulations on the weighted ternary number partitioning problem confirm scalability and robustness, showing convergence of up to 99.96%, comparable to near-optimal solutions.

Memristors exhibit tunable resistance, which has been widely exploited in non-volatile memory, in-memory computing, and neuromorphic computing. They can also serve as an entropy source due to their inherent instability, making them attractive for security devices and probabilistic computing. When these two characteristics are coupled, memristors can act as tunable entropy sources; however, this direction remains largely unexplored. Here, we propose a spiking-rate-programmable probabilistic neuron that leverages the tunable noise characteristics of a Ru/TaOx/Pt memristor. In this memristor, the conduction mechanism varies across resistance states, leading to distinct noise behaviors and signal-to-noise ratios that depend on the programmed resistance. This noise can be harnessed to realize frequency-selective, frequency-domain probabilistic neural encoding. By integrating these probabilistic neurons, an identical network architecture can process input signals spanning a wide frequency range, achieving around 95% classification performance on both low-frequency human activity data (UCI HAR, 0.4-25 Hz) and high-frequency speech data (Audio MNIST, 20 Hz-8 kHz). These results highlight a new direction that leverages the intrinsic properties of memristors for compact, adaptive, and energy-efficient time-series encoding.

Memristors exhibit changes in internal resistance in response to external voltage, introducing new functionalities to electronic devices. This enables diverse applications in non-volatile memory, neuromorphic devices, sensors, and computing systems, highlighting their growing importance in electronics. These applications leverage various mechanisms underlying memristors. Therefore, understanding these mechanisms and discovering new memristive mechanisms are essential for overcoming implementation challenges and developing emerging applications. Here, a new type of memristor is introduced comprising an Ag/Ag:Cu-islands/HfO2/Pt structure, characterized by a hybrid mechanism and its potential for multifunctional applications. The memristor combines the metallic filament (Ag/Cu alloy) of the electrochemical metallization (ECM) mechanism with the oxygen vacancy filament of the valence change memory (VCM) mechanism, achieving both the high on/off ratio of ECM and the analog characteristics of VCM with enhanced reliability. Both resistive and threshold switching characteristics are shown by controlling the compliance current, making the device applicable to artificial synapses and neurons. Notably, this device exhibits heat-responsive nociceptor characteristics, positioning it as a promising candidate for next-generation neuromorphic devices.

The human brain's efficiency and adaptability in processing information is largely attributed to spatiotemporal spiking activities and intrinsic plasticity-the ability of neurons to autonomously modulate their excitability. Mott memristors, with their threshold switching characteristics, have been effectively utilized as artificial neurons, or neuristors, to generate spiking activities. However, the implementation of intrinsic plasticity and its significance in neuromorphic computing has yet to be systematically explored. Here, a frequency switching (FS) neuristor is presented that emulates neuron's intrinsic plasticity characteristics. By combining a volatile Mott memristor with a non-volatile valence change memory (VCM) memristor, the FS neuristor achieves programmable multi-level frequency-voltage (f-V) characteristics analogous to the transfer functions of neuronal intrinsic plasticity. Through device-based simulations of sparse neural networks, it is proposed that this intrinsic plasticity acts as memory and processor itself, enhancing network performance and reducing energy consumption. Additionally, intrinsic plasticity endows the network with structural plasticity, enabling full recovery of the network's performance after random neuron damage, suggesting a pathway toward more adaptive and resilient neuromorphic computing systems.

Optoelectronic devices using circularly polarized light (CPL) offer enhanced sensitivity and specificity for efficient data processing. There is a growing demand for CPL sensing mediums with strong optical activity, stability and sensitivity, multiple transition bands, and environmental compatibility. Here, defect-engineered chiroferromagnetic quantum dots (CFQDs) are used as a new type of CPL sensing material. By inducing amorphization defects through chiral molecules, CFQDs with high unpaired electron density, atomic structural chirality, amplified chiroptical activity, and multiple exciton transition bands are developed. CFQDs enable nonlinear, long-term plastic behavior with linear optical input, acting as in situ noise filters that reduce noise by over 20%. Additionally, CFQDs provide over nine times higher integration for photon polarization and wavelength distinctions, paving the way for next-generation processors with improved energy efficiency, integration, and reduced retention time.

Edge computing devices, which generate, collect, process, and analyze data near the source, enhance the data processing efficiency and improve the responsiveness in real-time applications or unstable network environments. To be utilized in wearable and skin-attached electronics, these edge devices must be compact, energy efficient for use in low-power environments, and fabricable on soft substrates. Here, we propose a flexible memristive dot product engine (f-MDPE) designed for edge use and demonstrate its feasibility in a real-time electrocardiogram (ECG) monitoring system. The f-MDPE comprises a 32×32 crossbar array embodying a low-temperature processed self-rectifying charge trap memristor on a flexible polyimide substrate and exhibits high uniformity and robust electrical and mechanical stability even under 5-mm bending conditions. Then, we design a neural network training algorithm through hardware-aware approaches and conduct real-time edge ECG diagnosis. This approach achieved an ECG classification accuracy of 93.5%, while consuming only 0.3% of the energy compared to digital approaches. Furthermore, our simulations indicated that the energy reduction could be further reduced to 0.001% through device scaling to a 100-nm-line width, highlighting the strong potential of this approach for emerging edge neuromorphic hardware.

Genuine stochastic information represented by true random numbers (TRNs) is essential for entropy-associated applications such as cryptography and energy-based computing. They demand specialized hardware called true random number generators (TRNGs) capable of rapid, energy-efficient TRN generation. In this study, we enhance Johnson-Nyquist noise to demonstrate the fastest, most energy-efficient memristor-based TRNG. The TRNG comprises an NbOx-based negative resistance oscillator, a T flip-flop for digitalization, and a heater as a noise source. The heater enhances Johnson-Nyquist noise, achieving a TRNG speed of 100 kbit/s, 2.5× faster than without the heater. Furthermore, we propose a TRNG array that utilizes heat across the array for energy-efficient, parallel TRN generation. The 128-sized TRNG array is expected to achieve 0.65 μJ/bit at 1.28 Mbit/s, demonstrating significant improvements in speed and efficiency. By focusing on noise engineering rather than conventional material- or circuit-based methods, our approach enables broader applications in entropy-based computing.

Over the past decade, dendrites of neurons, which were previously thought to perform only information pooling and networking, have now been shown to express complex temporal dynamics, Boolean-like logic, arithmetic, signal discrimination, and edge detection. Mimicking this rich functionality could offer a powerful primitive for neuromorphic computing. Here, using Ovonic threshold switching in Sb-Te-doped GeSe, we demonstrate a single two-terminal component capable of self-sustained dynamics and universal Boolean logic in addition to XOR operations (which is traditionally thought to require a network of active components). We then employed logic-driven dynamics to detect and estimate the gradients of edges in images. The Ovonic switch exhibits properties of a half adder and a full adder in addition to discriminative logic accommodating inhibitory and excitatory signals. We show that this simple computational primitive offers a highly improved energy efficiency. As such, this work paves the path for potentially emulating dendrites for efficient postdigital neuromorphic computing.

A memristive crossbar array can execute Boolean logic operations directly within the memory, which is highly noteworthy as it addresses the data bottleneck issue in traditional von Neumann computing. Although its potential has been widely demonstrated, achieving practical levels of operational reliability and computational efficiency remains a challenge. Here, we introduce a three-input majority logic gate supported by near-memory operations, serving as a universal gate and achieving both robust reliability and high efficiency in versatile logic operations. We fabricated a highly reliable HfOx-based memristive array, incorporating a series resistor to increase the reset voltage of the memristor, thereby increasing the operational voltage margin of the gate operation. This ensured reliable operation of the majority gate, resulting in successful experimental proof of combined 1-bit full adder and subtractor operations performed in 5 steps using 7 cells. Additionally, we propose that an N-bit parallel prefix adder (PPA) operation is possible in O(log2 N) steps, by taking advantage of the parallel operation capability of the majority gate. This achieves 8.5x higher spatiotemporal efficiency than the previously reported NOR-based logic system in 64-bit adder operation. Moreover, as N increases, the spatiotemporal efficiency further improves, which significantly enhances the applicability of memristive logic-in-memory. Reliability was ensured and 8.5 times higher efficiency in a 64-bit adder was achieved through the parallel operation of the MAJ logic.

Artificial intelligence (AI) is often considered a black box because it provides optimal answers without clear insight into its decision-making process. To address this black box problem, explainable artificial intelligence (XAI) has emerged, which provides an explanation and interpretation of its decisions, thereby promoting the trustworthiness of AI systems. Here, a memristive XAI hardware framework is presented. This framework incorporates three distinct types of memristors (Mott memristor, valence change memristor, and charge trap memristor), each responsible for performing three essential functions (perturbation, analog multiplication, and integration) required for the XAI hardware implementation. Three memristor arrays with high robustness are fabricated and the image recognition of 3 × 3 testing patterns and their explanation map generation are experimentally demonstrated. Then, a software-based extended system based on the characteristics of this hardware is built, simulating a large-scale image recognition task. The proposed system can perform the XAI operations with only 4.32% of the energy compared to conventional digital systems, enlightening its strong potential for the XAI accelerator.

Insects can efficiently perform object motion detection via a specialized neural circuit, called an elementary motion detector (EMD). In contrast, conventional machine vision systems require significant computational resources for dynamic motion processing. Here, a fully memristive EMD (M-EMD) is presented that implements the Hassenstein-Reichardt (HR) correlator, a biological model of the EMD. The M-EMD consists of a simple Wye (Y) configuration, including a static resistor, a dynamic memristor, and a Mott memristor. The resistor and dynamic memristor introduce different signal delays, enabling spatio-temporal signal integration in the subsequent Mott memristor, resulting in a direction-selective response. In addition, a neuromorphic system is developed employing the M-EMDs to predict a lane-changing maneuver by vehicles on the road. The system achieved a high accuracy (> 87%) in predicting future lane-changing maneuvers on the Next Generation Simulation (NGSIM) dataset while reducing the computational cost by 92.9% compared to the conventional neuromorphic system without the M-EMD, suggesting its strong potential for edge-level computing.

Heat dissipation is a natural consequence of operating any electronic system. In nearly all computing systems, such heat is usually minimized by design and cooling. Here, we show that the temporal dynamics of internally produced heat in electronic devices can be engineered to both encode information within a single device and process information across multiple devices. In our demonstration, electronic NbOx Mott neurons, integrated on a flexible organic substrate, exhibit 18 biomimetic neuronal behaviours and frequency-based nociception within a single component by exploiting both the thermal dynamics of the Mott transition and the dynamical thermal interactions with the organic substrate. Further, multiple interconnected Mott neurons spatiotemporally communicate purely via heat, which we use for graph optimization by consuming over 106 times less energy when compared with the best digital processors. Thus, exploiting natural thermal processes in computing can lead to functionally dense, energy-efficient and radically novel mixed-physics computing primitives. Heat in electronic devices is normally dissipated via cooling. Here the authors engineer the thermal dynamics of the Mott transition and dynamical thermal interactions with the substrate to enable neuromorphic computing in a NbOx-based device.

A one-selector-one-memristor crossbar array was developed, capable of driving Monte Carlo DropConnect network. This could be achieved through a hardware and algorithm co-design approach, involving mutual improvement of them.

Energy-based computing is a promising approach for addressing the rising demand for solving NP-hard problems across diverse domains, including logistics, artificial intelligence, cryptography, and optimization. Probabilistic computing utilizing pbits, which can be manufactured using the semiconductor process and seamlessly integrated with conventional processing units, stands out as an efficient candidate to meet these demands. Here, we propose a novel pbit unit using an NbOx volatile memristor-based oscillator capable of generating probabilistic bits in a self-clocking manner. The noise-induced metal-insulator transition causes the probabilistic behavior, which can be effectively modeled using a multi-noise-induced stochastic process around the metal-insulator transition temperature. We demonstrate a memristive Boltzmann machine based on our proposed pbit and validate its feasibility by solving NP-hard problems. Furthermore, we propose a streamlined operation methodology that considers the autocorrelation of individual bits, enabling energy-efficient and high-performance probabilistic computing.

Gamma-aminobutyric acid (GABA) is a crucial inhibitory neurotransmitter of the central nervous system. It modifies the signal threshold of the nociceptor, allowing it to react to external stimuli in various circumstances. Thus, GABAergic behaviors are critical characteristics of adaptive behavior in life. Here, a threshold-modulative artificial GABAergic nociceptor is reported for the first time at a Pt/Ti/Nb2 O5- x /Al2 O3- y /Pt/Ti (top to bottom) of the double charge trapping structure. The Al2 O3- y layer contains deep defect states that function similarly to the GABA neurotransmitter in modulating the signal threshold. Meanwhile, the Nb2 O5- x layer traps volatile charges and produces nociceptive behaviors. The combined dynamics of the two layers readily offer threshold-modulative GABAergic nociceptive behaviors. Based on these GABAergic behaviors, a method of implementing hot- and cold-sensitive thermoreceptors is demonstrated and shows its potential applications in advanced sensory devices.

Cu interconnects suffer from increased resistance and poor reliability at a sub-10 nm width. Ru and Mo have been highlighted recently as the next interconnection material candidate due to their various advantages over Cu; they have lower resistance than Cu at sub-10 nm, do not diffuse into SiO2, and are etchable. Here, we evaluated the electromigration (EM) reliability of Ru and Mo to confirm their feasibility for the next-generation interconnection. The activation energy for EM failure is calculated by measuring the mean time to failure (MTTF) of film and wire structures while factoring in temperature increases with thermal coefficient of resistance (TCR) measurements. In addition, we investigate the EM properties in terms of resistivity-increasing parameters that originate from geometry and additional fabrication processes. Furthermore, we evaluate the EM performance in terms of electrochemical potential. Our findings confirm the feasibility of Ru as a promising candidate for next-generation interconnection applications, providing enhanced reliability compared to conventional Cu interconnects.

Abstract Energy-based computing is a promising approach for solving NP-hard problems. Probabilistic computing using pbits, which can be fabricated through the semiconductor process and integrated with conventional processing units, can be an efficient candidate for fulfilling these demands. Here, we propose a novel pbit unit comprising a NbOx mott memristor-based oscillator, capable of generating probabilistic bits in a self-clocking manner. The noise-induced mott transition causes the probabilistic behavior, which can be effectively modeled using a multi-noise-induced stochastic process around the mott transition temperature. We demonstrate a memristive Boltzmann machine based on our proposed pbit and validate its feasibility by solving NP-hard problems. Furthermore, we propose a streamlined operation methodology that considers the autocorrelation of individual bits, enabling energy-efficient high-performance probabilistic computing.

NbOx-based Mott memristors exhibit fast threshold switching behaviors, making them suitable for spike generators in neuromorphic computing and stochastic clock generators in security devices. In these applications, a high output spike amplitude is necessary for threshold level control and accurate signal detection. Here, we propose a materialwise solution to obtain the high amplitude spikes by inserting Au nanodots into the NbOx device. The Au nanodots enable increasing the threshold voltage by modulating the oxygen contents at the electrode-oxide interface, providing a higher ON current compared to nanodot-free NbOx devices. Also, the reduction of the local switching region volume decreases the thermal capacitance of the system, allowing the maximum spike amplitude generation. Consequently, the Au nanodot incorporation increases the spike amplitude of the NbOx device by 6 times, without any additional external circuit elements. The results are systematically supported by both a numerical model and a finite-element-method-based multiphysics model.

Heat dissipation is a natural consequence of operating any electronic system. In nearly all of computing, such heat is minimized by design and discarded via cooling, while some post-digital electronics (such as phase-change memories) utilize only the static electrical outcome of such heat dissipation within a single device to represent information. Thus, neither the naturally produced heat energy nor its dynamics are directly used for computing. Here we demonstrate electronic NbO x Mott neurons integrated on a biocompatible flexible organic substrate, designed to exploit both the thermal dynamics of the Mott transition and the dynamics of thermal interactions with the organic substrate, to exhibit 18 bio-mimetic neuronal behaviors and frequency-based nociception (sensing of noxious stimuli) within a single component. Further, multiple Mott neurons can spatiotemporally communicate purely via their thermal dynamics, which we use for graph optimization by consuming over 10 7 × lower energy compared to the best digital processors. Thus, we show that naturally produced heat in electronic systems and its dynamics can be used directly to encode and process information, thereby enabling a higher-complexity energy efficient and radically novel mixed-physics computing primitive.