Zuyu Xu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Zuyu Xu has a total of 139 co-authors. Here are the top co-authors by number of shared publications:
- Zuheng Wu
Anhui University
25 shared publications - Yunlai Zhu
Anhui University
23 shared publications Zhe Feng
Anhui University
16 shared publicationsJianxun Zou
Anhui University
14 shared publicationsYuehua Dai
Anhui University
14 shared publications- Yuehua Dai
Anhui University
12 shared publications
Latest Publications
- 17 October 2026
- Published by American Chemical Society (ACS) in ACS Applied Materials & Interfaces
- Vol. 18 (42) , 57820-57832
Memcapacitive-Dominated Reservoir Computing with Self-Organizing Nanowire Networks
Nanowire networks have emerged as a promising platform for physical reservoir computing. However, networks based on memristive switching exhibit inherent limitations in dynamical richness and short-term memory capacity. Memcapacitors are expected to overcome these dynamical constraints through their topology-dependent dynamics. In this work, inspired by the distributed processing of neural systems, we endow Ag nanowire networks with memcapacitive properties via a temperature-modulated fabrication process, realizing a multi-path channel-dependent reservoir system with richer reservoir dynamics and establishing its circuit model. Benefiting from its memcapacitive nature, the system exhibits channel-specific responses and approximate linear superposition, enabling parallel and spontaneous fusion of multi-input signals with controllable and predictable behavior. This allows input states to expand exponentially with channel count. Leveraging this capability, we designed a multi-threshold encoding scheme that generates 256 highly separable states within four time steps while consuming only 18.98 pJ/pulse. This approach reduces the information loss inherent in conventional binarization and delivers over 94% accuracy on a compressed MNIST recognition task, demonstrating a potent strategy for high-performance neuromorphic edge computing.
- 17 October 2026
- Published by American Chemical Society (ACS) in ACS Applied Materials & Interfaces
- Vol. 18 (42) , 57820-57832
Memcapacitive-Dominated Reservoir Computing with Self-Organizing Nanowire Networks
Nanowire networks have emerged as a promising platform for physical reservoir computing. However, networks based on memristive switching exhibit inherent limitations in dynamical richness and short-term memory capacity. Memcapacitors are expected to overcome these dynamical constraints through their topology-dependent dynamics. In this work, inspired by the distributed processing of neural systems, we endow Ag nanowire networks with memcapacitive properties via a temperature-modulated fabrication process, realizing a multi-path channel-dependent reservoir system with richer reservoir dynamics and establishing its circuit model. Benefiting from its memcapacitive nature, the system exhibits channel-specific responses and approximate linear superposition, enabling parallel and spontaneous fusion of multi-input signals with controllable and predictable behavior. This allows input states to expand exponentially with channel count. Leveraging this capability, we designed a multi-threshold encoding scheme that generates 256 highly separable states within four time steps while consuming only 18.98 pJ/pulse. This approach reduces the information loss inherent in conventional binarization and delivers over 94% accuracy on a compressed MNIST recognition task, demonstrating a potent strategy for high-performance neuromorphic edge computing.
- 1 October 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Electron Devices
- Vol. 73 (10) , 6282-6288
Suppressing Poole--Frenkel Instability in NbO x Memristors Through an Ultrathin SiO x Tunneling Layer
- 1 October 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Electron Devices
- Vol. 73 (10) , 6282-6288
Suppressing Poole--Frenkel Instability in NbO x Memristors Through an Ultrathin SiO x Tunneling Layer
- Open Access
- 20 September 2026
- Published by Wiley in Advanced Materials
- Vol. 38 (62) , e75096
A Memristor‐Based Saturation‐Constrained Information Capacity Framework for Temporal Information Processing
Physical reservoir computing (PRC) has emerged as a highly energy‐efficient paradigm for processing high‐dimensional spatiotemporal data in edge‐intelligent systems. However, the deployment of PRC is often limited by the quantization mismatch between continuous input signals and the finite dynamic range of hardware substrates. Conventional binary time‐multiplexing ensures high noise immunity but necessitates protracted pulse sequences that induce high latency and premature physical saturation. Conversely, multi‐valued analog encoding enhances information density but frequently induces state overlap, operating devoid of physical quantization boundaries. To resolve this physical‐information trade‐off, we propose a saturation‐constrained information capacity (SCIC) framework. This framework considers task‐specific encoding variables and substrate‐intrinsic constraints, enabling device‐informed optimization of input mapping. We experimentally demonstrate this framework on a highly uniform, selector‐free 32 × 32 crossbar array, utilizing the volatile dynamics of memristors. Under the experimentally selected encoding conditions, the SCIC‐guided RC achieves exceptional performance in temporal prediction (NRMSE of 0.019), speech recognition (free‐spoken digit dataset (FSDD) subset accuracy of ∼94.5%), and real‐time obstacle avoidance. Our approach links encoding constraints to solid‐state device physics, providing a device‐informed method for PRC parameterization at the edge.
- 20 September 2026
- Published by Wiley in Advanced Materials
- Vol. 38 (62) , e75096
Open AccessA Memristor‐Based Saturation‐Constrained Information Capacity Framework for Temporal Information Processing
Physical reservoir computing (PRC) has emerged as a highly energy‐efficient paradigm for processing high‐dimensional spatiotemporal data in edge‐intelligent systems. However, the deployment of PRC is often limited by the quantization mismatch between continuous input signals and the finite dynamic range of hardware substrates. Conventional binary time‐multiplexing ensures high noise immunity but necessitates protracted pulse sequences that induce high latency and premature physical saturation. Conversely, multi‐valued analog encoding enhances information density but frequently induces state overlap, operating devoid of physical quantization boundaries. To resolve this physical‐information trade‐off, we propose a saturation‐constrained information capacity (SCIC) framework. This framework considers task‐specific encoding variables and substrate‐intrinsic constraints, enabling device‐informed optimization of input mapping. We experimentally demonstrate this framework on a highly uniform, selector‐free 32 × 32 crossbar array, utilizing the volatile dynamics of memristors. Under the experimentally selected encoding conditions, the SCIC‐guided RC achieves exceptional performance in temporal prediction (NRMSE of 0.019), speech recognition (free‐spoken digit dataset (FSDD) subset accuracy of ∼94.5%), and real‐time obstacle avoidance. Our approach links encoding constraints to solid‐state device physics, providing a device‐informed method for PRC parameterization at the edge.
- Open Access
- 23 July 2026
- Published by Wiley in Advanced Functional Materials
- Vol. 36 (67) , e77093
High‐Throughput True Random Number Generator Enabled by NbO x Memristor With Dual‐Clock‐Edge Sampling for Dual‐Image Encryption
As a fundamental hardware security primitive, true random number generators (TRNGs) offer superior unpredictability over pseudo‐random number generators (PRNGs). However, limited throughput hinders their widespread adoption. Here, leveraging the abundant intrinsic entropy of memristors, we co‐optimize the device switching speed and the sampling method of the entropy extraction circuit. For the device, a bilayer Ti/NbO x /NbO y /Pt (T2NP) threshold switching (TS) memristor, driven by an insulator‐metal transition (IMT) mechanism, delivers the relatively short delay and relaxation time, facilitating both high‐speed and high‐quality random entropy generation among its peers. For the circuit, the conventional nonlinear feedback shift register (NFSR) is augmented with a resistor‐capacitor (RC) delay unit. This forms a dual‐input module comprising the native memristor response and its RC‐delayed counterpart, driving dual‐edge clock sampling to enable dual‐channel output from a single memristor. The optimized TRNG achieves a 4.2 Mb/s throughput with an energy consumption of 1.47 nJ/bit, a 1.89× enhancement in throughput over prior memristor‐based TRNGs. Dual‐image encryption and comprehensive tests confirm the high randomness, reliability, and robustness of the generated bitstreams. This co‐optimized framework underscores significant potential for high‐speed, scalable hardware security in future secure data processing.
- 23 July 2026
- Published by Wiley in Advanced Functional Materials
- Vol. 36 (67) , e77093
Open AccessHigh‐Throughput True Random Number Generator Enabled by NbO x Memristor With Dual‐Clock‐Edge Sampling for Dual‐Image Encryption
As a fundamental hardware security primitive, true random number generators (TRNGs) offer superior unpredictability over pseudo‐random number generators (PRNGs). However, limited throughput hinders their widespread adoption. Here, leveraging the abundant intrinsic entropy of memristors, we co‐optimize the device switching speed and the sampling method of the entropy extraction circuit. For the device, a bilayer Ti/NbO x /NbO y /Pt (T2NP) threshold switching (TS) memristor, driven by an insulator‐metal transition (IMT) mechanism, delivers the relatively short delay and relaxation time, facilitating both high‐speed and high‐quality random entropy generation among its peers. For the circuit, the conventional nonlinear feedback shift register (NFSR) is augmented with a resistor‐capacitor (RC) delay unit. This forms a dual‐input module comprising the native memristor response and its RC‐delayed counterpart, driving dual‐edge clock sampling to enable dual‐channel output from a single memristor. The optimized TRNG achieves a 4.2 Mb/s throughput with an energy consumption of 1.47 nJ/bit, a 1.89× enhancement in throughput over prior memristor‐based TRNGs. Dual‐image encryption and comprehensive tests confirm the high randomness, reliability, and robustness of the generated bitstreams. This co‐optimized framework underscores significant potential for high‐speed, scalable hardware security in future secure data processing.
- 18 February 2026
- Published by Elsevier in Physica D: Nonlinear Phenomena
- Vol. 489
Enhancing physics-informed neural networks with a hybrid parallel Kolmogorov-Arnold and MLP architecture
- 18 February 2026
- Published by Elsevier in Physica D: Nonlinear Phenomena
- Vol. 489
Enhancing physics-informed neural networks with a hybrid parallel Kolmogorov-Arnold and MLP architecture
- 3 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal on Emerging and Selected Topics in Circuits and Systems
- Vol. 16 (2) , 302-311
Hierarchical Memristor-Based IMPLY Comparator Architecture for In-Memory Computing
To address the inherent bottlenecks of comparison latency and energy efficiency caused by the separation of computation and storage in the von Neumann architecture, this study proposes a memristor-based modular hierarchical tree comparator that employs implication (IMPLY) logic. The proposed architecture performs parallel bit-wise comparisons, generating ternary outputs (">", "<", "=") that are aggregated through a balanced binary tree structure. This approach reduces comparison latency from linear to logarithmic scaling with bit width. A dedicated sign-bit processing module enables flexible support for both signed and unsigned data formats. Circuit simulations demonstrate that, compared with existing memristor-based IMPLY logic unsigned comparators, the proposed unsigned comparator achieves a 7.8%–21.9% energy reduction for unsigned comparisons, while the proposed signed comparator not only extends the functionality but also achieves a 6.7%–15.1% reduction in energy consumption compared with unsigned comparators of the same bit-width. The architecture also achieves significant latency and Figure of Merit (FoM) improvements over memristor-aided logic and IMPLY-based designs, outperforming existing approximate computing approaches at large bit widths. Validation in image processing applications confirms the architecture's potential for high-throughput, low-power tasks, establishing a promising design paradigm for efficient in-memory comparison.
- 3 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal on Emerging and Selected Topics in Circuits and Systems
- Vol. 16 (2) , 302-311
Hierarchical Memristor-Based IMPLY Comparator Architecture for In-Memory Computing
To address the inherent bottlenecks of comparison latency and energy efficiency caused by the separation of computation and storage in the von Neumann architecture, this study proposes a memristor-based modular hierarchical tree comparator that employs implication (IMPLY) logic. The proposed architecture performs parallel bit-wise comparisons, generating ternary outputs (">", "<", "=") that are aggregated through a balanced binary tree structure. This approach reduces comparison latency from linear to logarithmic scaling with bit width. A dedicated sign-bit processing module enables flexible support for both signed and unsigned data formats. Circuit simulations demonstrate that, compared with existing memristor-based IMPLY logic unsigned comparators, the proposed unsigned comparator achieves a 7.8%–21.9% energy reduction for unsigned comparisons, while the proposed signed comparator not only extends the functionality but also achieves a 6.7%–15.1% reduction in energy consumption compared with unsigned comparators of the same bit-width. The architecture also achieves significant latency and Figure of Merit (FoM) improvements over memristor-aided logic and IMPLY-based designs, outperforming existing approximate computing approaches at large bit widths. Validation in image processing applications confirms the architecture's potential for high-throughput, low-power tasks, establishing a promising design paradigm for efficient in-memory comparison.
- 26 January 2026
- Published by IOP Publishing in Nanotechnology
- Vol. 37 (5)
Optimizing the analog synaptic characteristics of memristors by regulating the distribution of barrier layers
Memristors are promising candidates for artificial synapses in neuromorphic computing systems, yet their performance is often limited by nonlinear conductance modulation in oxide-based memristors. This work systematically investigates the modulation mechanism governing conductive filament (CF) rupture behavior utilizing distributed barrier layers based on finite element simulations. Our initial electro-thermal simulations of a HfO 2 -based memristor with a single-layer Al 2 O 3 barrier (SLB) thickness ( h = 0, 3, 6, 12 nm) showed only limited improvement in synaptic linearity. In contrast, the introduction of a (HfO 2 /Al 2 O 3 ) n multilayer barrier (MLB) structure fundamentally alters the switching dynamics. Simulations reveal that appropriately increasing the number of layers ( n ) promotes a transition from continuous to spatially discrete oxygen vacancy migration pathways. This engineered disorder expands the CF rupture region from a localized position to multiple interfaces, thereby reducing the electric field and temperature peaks and driving the set and reset process from abrupt to gradual switching. The optimized MLB device ( n = 4) exhibits significantly enhanced synaptic linearity and analog switching characteristics, closely emulating biological synapse behavior. Furthermore, system-level validation using this device model achieved an accuracy of 94.34% in handwritten digit recognition. This work elucidates the physical mechanism by which multilayer barriers enhance conductance linearity, providing a novel design strategy for high-performance memristive synapses.
- 26 January 2026
- Published by IOP Publishing in Nanotechnology
- Vol. 37 (5)
Optimizing the analog synaptic characteristics of memristors by regulating the distribution of barrier layers
Memristors are promising candidates for artificial synapses in neuromorphic computing systems, yet their performance is often limited by nonlinear conductance modulation in oxide-based memristors. This work systematically investigates the modulation mechanism governing conductive filament (CF) rupture behavior utilizing distributed barrier layers based on finite element simulations. Our initial electro-thermal simulations of a HfO 2 -based memristor with a single-layer Al 2 O 3 barrier (SLB) thickness ( h = 0, 3, 6, 12 nm) showed only limited improvement in synaptic linearity. In contrast, the introduction of a (HfO 2 /Al 2 O 3 ) n multilayer barrier (MLB) structure fundamentally alters the switching dynamics. Simulations reveal that appropriately increasing the number of layers ( n ) promotes a transition from continuous to spatially discrete oxygen vacancy migration pathways. This engineered disorder expands the CF rupture region from a localized position to multiple interfaces, thereby reducing the electric field and temperature peaks and driving the set and reset process from abrupt to gradual switching. The optimized MLB device ( n = 4) exhibits significantly enhanced synaptic linearity and analog switching characteristics, closely emulating biological synapse behavior. Furthermore, system-level validation using this device model achieved an accuracy of 94.34% in handwritten digit recognition. This work elucidates the physical mechanism by which multilayer barriers enhance conductance linearity, providing a novel design strategy for high-performance memristive synapses.
- 20 November 2025
- Published by IOP Publishing in Nanotechnology
- Vol. 36 (47) , 475202
Artificial nociceptor with Pt/NbO x /Graphene memristive device
Nociceptors, the body’s essential alarm system, detect harmful stimuli (mechanical, thermal, chemical) and trigger protective neural responses. Memristive devices offer a powerful platform for emulating such biological functions due to their ion-migration dynamics, which naturally replicate neural threshold behaviors and plasticity. While prior memristive artificial nociceptors have successfully mimicked key characteristics of A-MH I-type nociceptors (threshold activation, relaxation, allodynia, hyperalgesia), the implementation of A-MH II-type responses, such as subthreshold response, rapid active, adaption and fatigue behaviors, remains scarce despite its critical role in realistic biomimetics. Here, we develop a Pt/NbO x /Graphene memristive device-based nociceptor leveraging electron trapping/de-trapping dynamics. This device comprehensively emulates both A-MH I-type features and previously overlooked A-MH II-type properties. The successful dual-type nociceptor emulation demonstrates exceptional potential for constructing high-fidelity biomimetic systems in advanced robotics, prosthetics, and neuromorphic sensing.
- 20 November 2025
- Published by IOP Publishing in Nanotechnology
- Vol. 36 (47) , 475202
Artificial nociceptor with Pt/NbO x /Graphene memristive device
Nociceptors, the body’s essential alarm system, detect harmful stimuli (mechanical, thermal, chemical) and trigger protective neural responses. Memristive devices offer a powerful platform for emulating such biological functions due to their ion-migration dynamics, which naturally replicate neural threshold behaviors and plasticity. While prior memristive artificial nociceptors have successfully mimicked key characteristics of A-MH I-type nociceptors (threshold activation, relaxation, allodynia, hyperalgesia), the implementation of A-MH II-type responses, such as subthreshold response, rapid active, adaption and fatigue behaviors, remains scarce despite its critical role in realistic biomimetics. Here, we develop a Pt/NbO x /Graphene memristive device-based nociceptor leveraging electron trapping/de-trapping dynamics. This device comprehensively emulates both A-MH I-type features and previously overlooked A-MH II-type properties. The successful dual-type nociceptor emulation demonstrates exceptional potential for constructing high-fidelity biomimetic systems in advanced robotics, prosthetics, and neuromorphic sensing.