Laser-Guided Ion Dynamics in a Dual-Mode Memristor for Bioinspired Neuronal and Synaptic Integration

Laser-Guided Ion Dynamics in a Dual-Mode Memristor for Bioinspired Neuronal and Synaptic Integration | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Laser-Guided Ion Dynamics in a Dual-Mode Memristor for Bioinspired Neuronal and Synaptic Integration Keon Jae Lee, Yu Jin Jeong, Kyunghwan Kim, Young Bin Kim, Hyera Shin, and 2 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6140700/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Neuromorphic computing aims to replicate the parallel, adaptive nature of biological intelligence in electronic systems. Despite considerable advances in memristor technology, material-encoded neurosynaptic bifunctionality has not been demonstrated. We introduce a laser-guided dual-mode memristor that integrates both volatility for neuronal spiking and nonvolatility for synaptic plasticity within a single-phase material. By precisely modulating silver ion dynamics through XeCl excimer laser irradiation, we achieve local and dynamic control of the dual-mode memristive behavior without requiring a heterogeneous device array or stacking. The neurosynaptic tunability with optimal computational efficiency demonstrates reconfigurable reservoir computing and a positive feedback loop for adaptive learning. Physical sciences/Materials science/Materials for devices/Electronic devices Physical sciences/Materials science/Materials for optics/Lasers, LEDs and light sources/Semiconductor lasers Laser-guided ion dynamics Single coplanar memristor Bioinspired neuronal and synaptic integration Volatile and nonvolatile switching Neuromorphic reservoir computing Figures Figure 1 Figure 2 Figure 3 Figure 4 Main The human brain achieves remarkable efficiency by integrating neuronal and synaptic functions for real-time adaptation and long-term memory. Emulating this dual functionality is critical for energy-efficient, scalable neuromorphic hardware capable of dynamic learning and cognitive tasks 1 – 5 . However, despite progress in CMOS-based neuromorphic circuits, current hardware rigidly separates neuronal and synaptic processes, even though transient neuronal activity and stable synaptic memory must function in tandem. Furthermore, the conventional electrical stimulus lacks the spatial selectivity required to control volatile and nonvolatile states independently 6 – 8 . A strategy that provides localized and reconfigurable neurosynaptic connections by mimicking the 3D functionality of the brain is a potential means of overcoming these limitations. While memristors are promising candidates for integrating neuronal and synaptic functions, most implementations rely on external circuit-based links, rather than a unified material for bifunctionality 9 – 12 . Direct neuron-synapse integration using a stacked architecture to support brain-like computation has attracted considerable attention for memristor-based neurosynaptic systems 13 , 14 . However, previous neurosynaptic integrations, including stacked or circuit-based neuron-synapse connections and material-constrained non-tunable volatility, have faced fundamental challenges such as insufficient spatial selectivity, lack of direct neurosynaptic interconnection, and limited freedom of ion transport control 15 , 16 . Laser technologies, widely adopted in semiconductor patterning 17 , 18 and surface modification 19 – 21 , offer precise control of material properties and ionic conductivity 22 . Unlike conventional microfabrication with limited spatial selectivity, laser energy transfer can drive localized structural and electronic property changes and thereby support tunable neurosynaptic behaviors 23 . However, its application to selective modulation of volatile-nonvolatile transitions within a homogeneous composition has not been explored. Here, we report a laser-guided dual-mode memristor that enables direct neurosynaptic interconnection by selectively controlling ion dynamics within a single-phase material. A pristine Ag-doped SiO₂ threshold switch (TS) exhibits frequency-dependent neuronal firing, while laser-irradiated regions transition into synapse-like conductive states via Ag clustering. The direct connectivity between the laser-induced synapse and TS neuron demonstrates neuromorphic reservoir computing (RC) through a reconfigurable crossbar architecture. The dual-mode neurosynaptic memristor enhances RC efficiency by integrating short-term neuronal dynamics and long-term synaptic plasticity. This tunability facilitates in-situ adaptive learning and a positive feedback in synaptic strength. Results Laser-guided dual-mode memristor with neuron-synapse pair The laser-guided dual-mode memristor integrates neuron- and synapse-like functionalities within a single layer. We developed an integrated neurosynaptic architecture, shown in Fig. 1 (a), in which a pristine Ag-doped SiO₂ TS exhibits volatile switching and selectively transitions to a nonvolatile state upon XeCl excimer laser irradiation. To fabricate the volatile TS layer, an Au bottom electrode was deposited, followed by a 1 nm Ag reservoir layer, a 15 nm Ag-doped SiO₂ switching layer, and a Pt top electrode. Within the crossbar array, localized laser exposure on the Ag:SiO₂ matrix stabilizes the filaments and converts specific regions to nonvolatile operation (see Methods and Supplementary Fig. 1 for fabrication details). The excimer laser was optimized at 308 nm, 150–175 mJ/cm², with a 25 ns pulse width (see Supplementary Fig. 2 for details). Figure 1 (b) presents the device's behavior before and after laser irradiation of the Au/Ag/SiO₂:Ag/Pt structure. Before laser exposure, dispersed Ag nanoparticles form transient filaments, resulting in volatile switching that mimics rapid signaling dynamics of biological neurons 24 – 26 . After laser irradiation, plasmonic resonance and electromagnetic gradients drive Ag filament aggregation. The aggregated Ag filaments stabilize the conduction path and transition the TS region into a nonvolatile state 27 – 31 . This stable reconfiguration affords synaptic plasticity for adaptive learning in neuromorphic systems. Figure 1 (c) depicts the volatile mode that provides transient signal processing and action potential firing in neuronal plasticity. The nonvolatile mode mirrors neurotransmitter release in synaptic plasticity for long-term memory consolidation 32 . These complementary mechanisms reflect the hierarchical structure of neural networks, where short-term adaptability and long-term retention coexist 3 . Laser-guided ion dynamics provide the dual-mode memristor with the capability of adaptive neurosynaptic reconfiguration. Figure 1 (d) illustrates a bioinspired reservoir computing scenario based on the dual-mode operation of the reconfigurable memristor. The volatile mode functions as a physical reservoir for processing temporal signals whereas the nonvolatile mode serves as a readout layer for storing stable synaptic weights 33 – 35 . Laser-guided Ag filament clustering provides precise spatial control over volatility and allows direct neurosynaptic interactions in a single-phase material. Dynamically controlled ion transport supports adaptive connectivity and in turn reconfigurable neuromorphic architectures can be realized 30 , 31 . Laser-induced ion dynamics and resistive switching in a dual-mode memristor Laser irradiation induces localized energy absorption within the Ag-doped SiO₂ matrix, which triggers a temperature rise that enhances Ag-ion mobility. As the excitation energy propagates, localized surface plasmon resonance (LSPR) amplifies the local electric field, optical absorption, and structural transformations 27 – 29 . Figure 2 (a) illustrates the 308 nm excimer laser-driven material transition into Ag nanoparticles (Ag NPs) growth for selective volatility control in dual-mode memristor. The LSPR effect on Ag NP clustering is depicted in Fig. 2 (b), as electron cloud oscillations generate strong local electric fields 27 , 28 . Initially, Ag NPs are uniformly distributed within the SiO₂:Ag switching layer, with the reservoir layer serving as an ion source 13 , 24 . Upon laser irradiation, LSPR generates strong local electric fields and electron cloud oscillations that lower the energy barrier for Ag-ion transport 27 , 28 . To validate this structural modification, cross-sectional Transmission Electron Microscopy (TEM) images were obtained at different fluences, as shown in Fig. 2 (c). At 125 mJ/cm², Ag NPs remain dispersed with diameters ranging from 1.5 to 3 nm. This indicates that the energy is insufficient to induce significant clustering and filament formation. As the fluence increased to 150–175 mJ/cm², Ag NPs coalesced into larger clusters (~ 27 nm) consistent with the LSPR-enhanced aggregation. The resonance effect is attributed to LSPR rather than conventional SPR due to the discrete nanoparticle structure of Ag, which contrasts with the continuous film required for SPR. Fluences over 175 mJ/cm², however, result in excessive filament formation due to structural instabilities and degrade electrical performance (Supplementary Fig. 3). UV-Vis absorption spectra were analyzed before and after laser treatment, as shown in Supplementary Fig. 4. A noticeable redshift in the absorption peak was observed and indicated an increase in Ag NP size. The observed redshift is attributed to the formation of larger Ag NP clusters, which modify the local dielectric environment and shift the LSPR resonance condition. Further analysis using energy-dispersive X-ray spectroscopy (EDS) mapping confirms the redistribution of Ag NPs under laser irradiation (Supplementary Fig. 5). Fast Fourier Transform (FFT) analysis of the TEM images reveals crystalline domains with d-spacing values, matching theoretical Ag lattice parameters (Supplementary Figs. 6, 7). X-ray photoelectron spectroscopy (XPS) and Time-of-Flight Secondary Ion Mass Spectrometry (ToF-SIMS) profiling confirm the change of Ag chemical states and support LSPR-mediated structural modifications (Supplementary Figs. 8–10). Figure 2 (d) presents COMSOL Multiphysics simulations of the electric field distribution between Ag NPs (radius = 2 nm, gap = 1 nm) performed at a frequency of 9.73×10 14 Hz, corresponding to 308 nm wavelength of laser. The strong local electric fields generated by LSPR lower the clustering barrier by enhancing energy absorption and inducing localized heating 27 – 29 . The simulation results confirm that localized plasmonic enhancement strengthens Ag-ion migration and establishes the necessary conditions for filament formation. Supplementary Figs. 11–13 present the dependence of the nanoparticle size and gap on field enhancement and its role in structural transformations within the switching layer. Figure 2 (e) shows an optical microscopy image of the fabricated 4×4 crossbar array of memristors, each initially designed as a volatile device. A single laser shot selectively induces nonvolatile transitions in targeted cells while preserving volatility in non-irradiated areas. The blue-box region retains original threshold switching characteristics, while the irradiated red-box region is transformed into stable nonvolatile pathways. This partial conversion yields both volatile and nonvolatile regions that coexist within the same crossbar and offers flexible neuron-synapse functionality with short-term and long-term memory in one architecture. In the volatile regime in Fig. 2 (f), a threshold-switching mechanism emerges near 0.24 V. The mechanism creates transient Ag filaments under bias, which revert to a high-resistance state upon bias removal. Conduction returns to a high-resistance state within 0.6 ms (Supplementary Fig. 14), mirroring transient neuronal signals 15 . Repeated switching over 100 cycles confirms stable threshold behavior with minimal voltage variation (Supplementary Fig. 15). Schottky emission governs the conduction mechanism, as shown in the inset of Fig. 2 (f), which indicates that electrons traverse a potential barrier at the metal-dielectric interface rather than forming a fully metallic path 36 . Figure 2 (g) presents the nonvolatile bipolar switching behavior, where the memristor transitions from a high-resistance state (HRS) to a low-resistance state (LRS) at an average SET voltage of approximately 0.41 V; conduction persists after bias removal, verifying stable retention. The conduction mechanism shifts from Schottky to Ohmic behavior, and continuous metallic pathways replace barrier-limited electron transport 36 . Extended retention measurements confirm stable conduction beyond 8,250 seconds 33 , while repeated switching over 100 cycles exhibits robust threshold behavior with minimal voltage variation. Multi-shot laser processing achieves tunable retention from milliseconds to 2,000 s by incrementally accumulating energy at a fluence of 102 mJ/cm², surpassing single-shot or thermal methods (Supplementary Fig. 16). To compare the effects of laser irradiation with conventional thermal processing, Rapid Thermal Annealing (RTA) was performed under different conditions (Supplementary Fig. 17). Annealing at 400°C maintained volatility, while treatment at 550°C led to severe degradation of memory functionality. These results highlight the advantages of laser processing in achieving controlled structural modifications without inducing significant thermal damage. Integrated emulation of neuronal spiking and synaptic plasticity The laser-guided dual-mode memristor emulates neuron-like spiking and synaptic plasticity within a single-phase material. This capability provides seamless transitions between short-term memory (STM) and long-term memory (LTM) and effectively replicates the spatiotemporal dynamics of biological neural networks. Figure 3 (a) presents the neuronal membrane structure and its circuit, where the memristor mimics a voltage-gated ion channel and the parallel capacitor represents the lipid bilayer. In biological neurons, AP generation occurs through the movement of ions (Na⁺, K⁺, Cl⁻) across the lipid bilayer 37 . The capacitor separates charge and generates a potential difference, while the memristor modulates current flow through Ag filament formation, mimicking the gating dynamics of ion channels (see Supplementary Figs. 18,19, Supplementary note 1). Signal transmission occurs when neurotransmitters are released from presynaptic vesicles and bind to postsynaptic receptors, generating an excitatory postsynaptic potential (EPSP) 38 , as shown in Fig. 3 (b). The memristor’s volatile mode reproduces fundamental neuronal spiking patterns, essential for sensory processing, motor control, and higher cognitive functions 25 , 26 . Figures 3 (c)–(e) show representative tonic spiking, phasic spiking, and threshold variation, while additional neuron spikes are presented in Supplementary Figs. 20–22. In a parallel RC circuit, the repetitive charging and discharging of the capacitor drive the voltage spiking behavior. The n th charging ( Vₙ ch ) and discharging ( Vₙ dis ) functions govern the capacitor’s voltage evolution in the circuit, where I in , tₙ ch , tₙ dis , τ ch , and τ dis denote the input current, charging/discharging time, and time constants. $$\:{V}_{n}^{ch}\left(t\right)={I}_{in}{R}_{H}\left\{1-{exp}\left(-\frac{t-{t}_{n}^{ch}}{{\tau\:}_{ch}}\right)\right\}$$ 1 $$\:{V}_{n}^{dis}\left(t\right)=I{R}_{L}{exp}\left(-\frac{t-{t}_{n}^{dis}-{\tau\:}_{dis}{ln}\left({V}_{th}/I{R}_{L}\right)}{{\tau\:}_{dis}}\right)$$ 2 When \(\:{I}_{in}{R}_{H}\:\) exceeds V th (and V th > 0), the capacitor voltage rises to V th according to the charging equation (Eq. ( 1 )). At that threshold, the memristor abruptly switches from \(\:{R}_{H}\:\) to \(\:{R}_{L}\) , causing the capacitor to discharge from V th to V min (Eq. ( 2 )). When a 0.8 µA input current is applied, the memristor with the 2.2 nF parallel capacitor generates stable tonic spiking, as seen in Fig. 3 (c). Increasing the capacitance extends the average spike period by slowing the charging and discharging. The voltage spike period ( T ) for the time-dependent charging and discharging equation is defined by the following equation (see Supplementary note 1 for details). $$\:T=C\left[{R}_{H}{ln}\left(\frac{I{R}_{H}-{V}_{min}}{I{R}_{H}-{V}_{th}}\right)+{R}_{L}{ln}\left(\frac{{V}_{th}}{{V}_{min}}\right)\right]$$ 3 Because the time constant τ scales with the capacitance C, the spike period T also depends on C. The average spike period increases from 9.2 ms to 17 ms as the capacitance rises from 2.2 nF to 10 nF in Supplementary Fig. 20. Raising the input current to 2 µA shifts the behavior from tonic spiking to tonic bursting, characterized by periodic clusters of spikes with quiescent periods exceeding 300 ms. This indicated a higher excitability threshold and enhanced burst firing, akin to cortical neuron activity. The memristor RC circuit replicates inhibition-induced spiking/bursting, where a negative current pulse induces rebound firing. After prolonged inhibitory input, the memristor transitions to rhythmic burst firing and mirrors T-type calcium channel-mediated spikes 37 . Phasic spiking is a bursting pattern with silent intervals and resembles thalamic relay cell activity, which regulates attention gating and network synchronization 39 (Fig. 3 (d)). At a higher current level of 5 µA, the electric field exceeds the Ag-ion diffusion rates, causing intermittent filament formation and dissolution. The volatile region shows threshold variation when alternating ± 1.2 V pulses are applied, revealing the device’s sensitivity to small voltage changes (Fig. 3 (e)). This adaptive threshold mechanism mirrors the influence of ionic dynamics and synaptic conditions on firing thresholds in biological neurons 37 , 40 . The laser-guided dual-mode memristor experimentally validates the leaky integrate-and-fire (LIF) model, where abrupt firing occurs as repeated input signals are accumulated to reach V th (see Supplementary Figs. 22). The all-or-nothing principle, which controls action potential generation in excitable cells, is exhibited in the device. It ensures binary signal transmission, essential for cortical and motor neurons 41 , 42 . The nonvolatile characteristics of laser-guided dual-mode memristor replicate synaptic plasticity. Figure 3 (f) demonstrates paired-pulse facilitation (PPF), a short-term synaptic plasticity mechanism where two closely spaced presynaptic spikes cause an increased postsynaptic response. This behavior results from the finite retention time of Ag conductive filaments 30 , which enlarge under repeated input and temporarily increase conductance 32 . Figure 3 (g) presents long-term potentiation (LTP) and depression (LTD), fundamental for learning and memory. Potentiation was induced by applying 0.4–1.0 V, 2 ms pulses in 0.01 V increments, while a -0.2 V, 2 ms pulses effected depression (V read at 0.1 V). To validate LTP/LTD nonlinearity, the open-source MATLAB tool Neurosim V3.0 was used, confirming the device’s ability to modulate synaptic weights accurately. The analysis yielded a nonlinearity of 2.03 for LTP and 4.88 for LTD, a symmetricity value of 3.09 × 10⁵, and a dynamic range (G max / G min ) of 144.1. The energy consumption at 0.1 V was 274 pJ/µm² for LTP and 52 pJ/µm² for LTD, indicating low-power potential for neuromorphic systems 43 – 45 (see Supplementary Fig. 23–25 and note 2–4). Figure 3 (h) verifies the symmetric Hebbian learning rule, a core principle of associative learning. Identical pre- and post-spike pulse conditions (0.6 V, 1 ms) were applied. The device’s conductance update follows an exponential decay, which confirms adaptive synaptic strength based on input timing 46 . Neuromorphic melody recognition and feedback-driven adaptation The RC system enables melody recognition by processing binary-encoded musical sequences using a dual-mode memristor 33 – 35 . Figure 4 (a) illustrates the RC simulation, where the memristor-based reservoir processes input signals and the nonvolatile readout layer stores synaptic weights for classification. The memristor’s threshold-switching behavior in volatile mode generates 80 virtual nodes that transform time-series inputs into distinct current spikes. Encoded spike outputs are then processed by a nonvolatile readout network of a 80 × 5 crossbar for weight storage and computation. Five representative melodies (Mary Had a Little Lamb, Twinkle Twinkle Little Star, Jingle Bells, Ode to Joy, and London Bridge Is Falling Down) were classified by the RC system. Figure 4 (b) illustrates the binary-to-pulse encoding scheme, where each musical pitch (C, D, E, F, F♯, G, A, and B) is mapped to binary codes (000–111). Binary ‘0’ corresponds to a single 0.7 V, 2 ms pulse that induces a transient threshold-switching event in volatile mode, while binary ‘1’ is represented by two consecutive 0.7 V pulses separated by 1 ms for an extended volatile response. To maintain echo-state properties, a 4 ms interval was introduced between consecutive symbols (e.g., ‘0’–‘0’, ‘0’–‘1’, and ‘1’–‘1’) and a 7 ms gap between distinct notes (e.g., ‘C’-‘D’, ‘F’-‘A’, and ‘F#’-‘A’). The interval allows clear differentiation between new input signals and previously processed data. With each note lasting 30 ms, a melody composed of eight notes had a total duration of 240 ms per sequence. Post-synaptic current (PSC) values were sampled every 3 ms and 80 virtual nodes that preserved the spatiotemporal input structure were generated. Gaussian noise was applied to create 5,000 variations per melody, yielding a dataset of 25,000 samples (20,000 for training, 5,000 for testing). The memristor’s threshold-switching converted input pulse streams into distinct current spikes at each virtual node. Classification thus could be performed by the intrinsic switching characteristics rather than by predefined computational rules 47 . Following the reservoir stage, the 80 virtual node outputs were utilized in an 80 × 5 readout layer, which stores and updates weights based on LTP/LTD. The weight adaptation uses in-situ backpropagation by the device-pair configurations (G⁺ and G⁻) 48 , 49 . The nonlinearity of 2.03 for LTP and 4.88 for LTD, a symmetricity value of 3.09 × 10⁵, and a dynamic range (G max /G min ) of 144.1 reveal excellent synaptic programmability. Moreover, the energy consumption at 0.1 V was 274 pJ /µm² for LTP and 52 pJ /µm² for LTD. Unlike conventional reservoir computing systems that require individual components for dynamic processing and memory, this approach employs a single memristor, making it feasible for low-power neuromorphic applications 43 – 45 . Figure 4 (c) presents a confusion matrix comparing actual and predicted outputs across five encoded melodies for high classification accuracy. After five training epochs, the RC system reached 94.34% accuracy, as shown in Fig. 4 (d), highlighting the effectiveness of in-situ weight adaptation. To validate the simulation-driven RC approach, we conducted experimental measurements on the dual-mode memristor (see Supplementary Fig. 26 for the current response under varied pulse amplitudes/widths in volatile mode). We verified robust echo-state properties at different pulse intervals, as presented in Supplementary Fig. 27, while Supplementary Fig. 28 provides details of the separability of the final conductance states for 16 different “0/1” input sequences. Furthermore, the pitch configuration (000–111) and melody-specific data for C–B notes, depicted in Supplementary Figs. 29–30, illustrate how the device encodes musical sequences as transient spikes. In addition to classification, the RC system exhibits biologically inspired feedback adaptation for dynamic neuromorphic learning. Figure 4 (e) presents a neuron-synapse coupled scheme to experimentally verify the feedback effect. The integrated array of directly connected neuro-synaptic memristors allows on-chip transient spiking and stable weight retention. The dual-mode synergy supports functional feedback loops, signal processing, and adaptive learning within a unified hardware framework. Feedback-driven spiking adaptation in a coupled volatile–nonvolatile memristor, where repeated pulse stimulation progressively increased spiking frequency and amplitude. A pulse train with 20 ms intervals progressively increased spiking frequency and amplitude over five repetitions. This mirrors biological synaptic potentiation in which repeated stimulation enhances neurotransmitter release 14 . First-spike latency decreased from 36.2 ms to 0.32 ms, while the peak current reached 0.24 mA. The memristor’s threshold-switching physics facilitates charge accumulation and lowers the activation energy required for filament formation, resulting in faster or stronger spiking responses 24 – 26 , 50 . The experimental results confirm that a single dual-mode memristor in an RC-based neuromorphic system can incorporate both transient neuronal dynamics and stable synaptic plasticity. This capability provides a hardware-efficient path toward real-time pattern recognition and adaptive learning. Discussion The laser-guided dual-mode memristor applies LSPR to achieve dynamic control over transient and stable conduction states within a single Ag-doped SiO₂ matrix. Excimer laser fluences of 150–175 mJ/cm² induce Ag-ion migration and clustering, forming 5–27 nm Ag nanoclusters, as observed by TEM. These clusters establish robust conduction pathways, enabling tunable volatility control and extended retention, a capability that surpasses conventional memristive devices reliant on static material properties. In the volatile regime, the device operates at a 0.24 V threshold with a conduction decaying within 0.6 ms, approximating rapid neuronal activity. Transition to the nonvolatile mode achieves a 3.4×10⁶ ON/OFF ratio, retention beyond 8,250 s, and a dynamic range of 144.1, with energy consumption as low as 274 pJ/µm 2 for potentiation and 52 pJ/µm 2 for depression. The ability to selectively modulate volatility within the same material system introduces a reconfigurable approach to neuromorphic computing, reducing energy overhead associated with separate volatile and nonvolatile components. Integrated into a reservoir computing framework, the dual-mode device yields 94.34% accuracy for melody recognition and exhibits positive feedback adaptation under repeated stimulation. Unlike conventional reservoir computing systems that require separate components for dynamic processing and memory storage, this approach consolidates both functionalities within a single memristor, significantly reducing hardware complexity and power consumption. The synergy between transient neuronal dynamics and stable synaptic weight retention addresses a critical gap in neuromorphic hardware, providing a direct link between computational neuroscience principles and practical AI implementations. A pathway emerges for large-scale, energy-efficient architectures that require real-time adaptability and self-directed learning in advanced AI hardware. Methods Fabrication of dual-mode neuromorphic memristor. A 4-inch silicon wafer was prepared for the fabrication of dual-mode neuromorphic memristors. A 5-µm-diameter cross-point Au/Ag/SiO₂:Ag/Pt memristor was then integrated on the substrate using the following steps. First, a 5-nm-thick Cr adhesion layer and a 30-nm-thick Au bottom electrode were deposited by e-beam evaporation (SNTEK). Next, a bottom contact pad was patterned via a conventional photolithography process (Midas MDA-600S), followed by wet etching. The switching medium of the volatile TS layer was deposited by sputtering a 1 nm thick Ag layer and a 15 nm thick Ag:SiO 2 layer. To deposit Ag:SiO₂, Ag and SiO₂ targets were co-sputtered in an Ar atmosphere. Subsequently, an XeCl excimer laser was directed onto a moving stage that scanned selective areas of the device to modulate the volatility of the TS layer. A 5 µm-diameter via was defined by photolithography, followed by RF sputtering of an SiO₂ insulating layer in an Ar atmosphere. After lift-off, the SiO₂ layer remained intact except in the via. Furthermore, the lift-off step exposed the switching layer while maintaining electrical connectivity with the bottom electrode. Finally, a top contact pad was defined by a conventional lithography process, followed by deposition of a 30 nm thick Pt (Ateck) layer and lift-off. Excimer Laser Annealing (ELA). A XeCl excimer laser (Coherent, COMPex Pro 205, λ = 308 nm, pulse duration 25 ns) was used for selective laser irradiation on top of the Ag:SiO 2 layer in the crossbar array. The ELA system was composed of a beam homogenizer, attenuator, and delivery optics designed to produce a flat-top, square beam profile. It also has an XY linear stage (Dukin, SLS-200) that provides full-area substrate scanning. It controls the irradiating fluence with a flat-top square beam on the sample. The laser shot was scanned along a “zig-zag” shaped pattern with an optimized energy density level. Electrical characterization. All electrical measurements were conducted using a Keithley 4200-SCS semiconductor parameter analyzer. A Keithley 4225-PMU with a remote amplifier/switch (Keithley 4225-RPM) was utilized for the voltage pulse measurements. In the spiking emulations, voltage spikes were measured by a Tektronix DPO 3054 digital phosphor oscilloscope with a P6139B voltage probe (10 MΩ input resistance). A conventional ceramic capacitor with 2.2 nF or 10 nF capacitance was connected in parallel to the memristor RC circuit. Electric Field Simulation. A finite-element approach in COMSOL Multiphysics (version 6.2) was employed to calculate the spatial distribution of the electric field in the electromagnetic wave, frequency domain (emw) module. The governing equation was solved under time‐harmonic conditions, satisfying Maxwell’s equations: $$\:\nabla\:\times\:{{\mu\:}_{r}}^{-1}(\nabla\:\times\:\text{E})-{k}_{0}^{2}\left({ϵ}_{r}-j\frac{\sigma\:}{\omega\:{ϵ}_{0}}\right)E=0$$ where \(\:{\mu\:}_{r}\) ​ denotes the relative permeability, \(\:{ϵ}_{r}\) ​ is the relative permittivity, \(\:\sigma\:\) is the electrical conductivity, \(\:\omega\:\) is the angular frequency, and \(\:{ϵ}_{0}\) ​ is the permittivity of free space. All material parameters were obtained from the COMSOL material library or reported data for SiO₂ and Ag. The electric field simulation was performed at a frequency of 9.7335×10¹⁴ Hz (corresponding to a wavelength of approximately 308 nm). A linearly polarized plane wave was applied along the z-direction with an initial field amplitude of 1 V/m. A Gaussian beam approximation was applied to model the incident field, addressing wavefront curvature and beam divergence. The applied electric field at lateral position 𝑥 and axial position 𝑦 was given by: $$\:{E}_{b}\left(x,\:y,z\right)={E}_{b0}\sqrt{\frac{{\omega\:}_{0}}{\omega\:\left(y\right)}}exp\left[\frac{-{x}^{2}}{{\omega\:}^{2}\left(y\right)}-jky-jk\frac{{x}^{2}}{2R\left(y\right)}+\frac{j\eta\:\left(y\right)}{2}\right]$$ where \(\:\omega\:\left(y\right)\:\) defines the beam waist as a function of propagation distance, \(\:\:R\left(y\right)\:\) is the wavefront radius of curvature, and \(\:\eta\:\left(y\right)\:\) is the Gouy phase shift. The Gaussian beam was linearly polarized along the 𝑥-direction and normally incident along the z-axis, with an initial field amplitude of \(\:{E}_{0}\) =1 V/m, a beam waist of \(\:{\omega\:}_{0}\) =0.308 µm, and a focal position at \(\:{y}_{0}\) =9.676 µm. To prevent unwanted reflections, multi-layered perfectly matched layers (PMLs) were applied along the x- and y-boundaries, ensuring efficient absorption of outgoing waves. The PML thickness was set to 2λ to minimize numerical artifacts while preserving computational efficiency. The bottom boundary was defined a Perfect Electric Conductor (PEC) condition to simulate a grounded electrode, while all other boundaries remained open with PMLs. A physics-controlled mesh was employed to optimize element distribution based on electromagnetic field variations, ensuring accurate resolution of rapid field gradients. The finest mesh setting was used, with refinement concentrated near Ag–SiO₂ interfaces and nanoparticles to capture localized plasmonic effects and strong permittivity contrasts. The smallest element size was set to resolve subwavelength field variations, preserving numerical accuracy. To solve the frequency-domain Maxwell’s equations efficiently, a direct solver was utilized for small-scale problems, while an iterative solver (GMRES with preconditioning) was applied for large-scale simulations to optimize memory usage. A relative tolerance of 10 − 6 was applied, and convergence was monitored by tracking the residual norm, ensuring it remained below 10 − 6 . The physics-controlled adaptive mesh ensured high accuracy, with automatic refinement applied in regions exhibiting strong field gradients. To analyze the simulated electric field distribution, postprocessing involved extracting the field amplitude ∣𝐸∣, phase distributions, and energy-flux profiles. These parameters were used to examine localized plasmonic enhancement in laser-irradiated regions, which played a crucial role in determining field intensities necessary for Ag-ion migration and filament formation in the switching layer. Declarations Data availability All data that support the findings of this study are reported in the Article and its supplementary information. Acknowledgements This work was supported by Samsung Electronics Co., Ltd (No.IO201214-08153-01). It was also supported by the Convergent Technology R&D Program for Human Augmentation through the National Research Foundation of Korea (NRF) funded by the Ministry of Science and ICT (No. NRF- 2020M3C1B8081519). This work was additionally supported by a National Research Foundation of Korea (NRF) grant funded by the Korean Government (MSIP) (No. NRF- 2020M3F3A2A02082445). Author contributions Y.J.J. and K.J.L. conceived the idea of the laser-guided dual-mode memristor, designed the experiments, and analyzed corresponding data. K.K., Y.B.K., H.S., J.W.O., and S.H.S. helped with experiments and data analysis. Y.J.J. and K.J.L. wrote the manuscript. K.J.L. supervised the research and contributed to the discussion of the overall methodology and results. All authors discussed the results and commented on the manuscript. Competing interests The authors declare no conflict of interest. References Ham, D., Park, H., Hwang, S. & Kim, K. Neuromorphic electronics based on copying and pasting the brain. Nat. Electron. 4, 635-644 (2021). Liu, Z. et al. A memristor-based adaptive neuromorphic decoder for brain–computer interfaces. Nat. Electron. Preprint at https://doi.org/10.1038/s41928-025-01340-2 (2025). Sung, S. H. et al. Memory-centric neuromorphic computing for unstructured data processing. Nano Research 14, 3126-3142 (2021). Prezioso, M. et al. Training and operation of an integrated neuromorphic network based on metal-oxide memristors. Nature 521, 61-64 (2015). Chen, S. et al. Wafer-scale integration of two-dimensional materials in high-density memristive crossbar arrays for artificial neural networks. Nat. Electron. 3, 638-645 (2020). Fu, T. et al. Bioinspired bio-voltage memristors. Nat Commun. 11, 1861 (2020). John, R. A. et al. Reconfigurable halide perovskite nanocrystal memristors for neuromorphic computing. Nat. Commun. 13, 2074 (2022). Wang, T. et al. Reconfigurable neuromorphic memristor network for ultralow-power smart textile electronics. Nat. Commun. 13, 7432 (2022). Xia, Q. & Yang, J. J. Memristive crossbar arrays for brain-inspired computing. Nat. Mater. 18, 309-323 (2019). Wan, W. et al. A compute-in-memory chip based on resistive random-access memory. Nature 608, 504-512 (2022). Kim, H., Mahmoodi, M. R., Nili, H. & Strukov, D. B. 4K-memristor analog-grade passive crossbar circuit. Nat. Commun. 12, 5198 (2021). Zhang, X. et al. An artificial spiking afferent nerve based on Mott memristors for neurorobotics. Nat. Commun. 11, 51 (2020). Sung, S. H., Kim, T. J., Shin, H., Im, T. H. & Lee, K. J. Simultaneous emulation of synaptic and intrinsic plasticity using a memristive synapse. Nat. Commun. 13, 2811 (2022). Sung, S. H. et al. Bio-plausible memristive neural components towards hardware implementation of brain-like intelligence. Mater. Today 62, 251-270 (2023). Woo, K. S. et al. Tunable stochastic memristors for energy-efficient encryption and computing. Nat. Commun. 15, 3245 (2024). Han, J.-K. et al. Cointegration of single-transistor neurons and synapses by nanoscale CMOS fabrication for highly scalable neuromorphic hardware. Sci. Adv. 7, eabg8836 (2021). Yang, Q. et al. High-speed, scanned laser structuring of multi-layered eco/bioresorbable materials for advanced electronic systems. Nat. Commun. 13, 6518 (2022). Chen, L. et al. Direct laser patterning of organic semiconductors for high performance OFET-based gas sensors. J. Mater. Chem. C 11, 7088-7097 (2023). Rho, Y. et al. A laser-assisted chlorination process for reversible writing of doping patterns in graphene. Nat. Electron. 5, 505-510 (2022). Meier, T. et al. Obtaining auxetic and isotropic metamaterials in counterintuitive design spaces: an automated optimization approach and experimental characterization. npj Comput. Mater. 10, 3 (2024). Lee, G. et al. Implantable, bioresorbable radio frequency resonant circuits for magnetic resonance imaging. Adv. Sci. 11, 2301232 (2024). Park, H. et al. Laser‐Based Selective Material Processing for Next‐Generation Additive Manufacturing. Adv. Mater. 36, e2307586 (2024). Nath, S. K. et al. Optically tunable electrical oscillations in oxide‐based memristors for neuromorphic computing. Adv. Mater. 36, e2400904 (2024). Wang, Z. et al. Memristors with diffusive dynamics as synaptic emulators for neuromorphic computing. Nat. Mater. 16, 101-108 (2017). Milozzi, A., Ricci, S. & Ielmini, D. Memristive tonotopic mapping with volatile resistive switching memory devices. Nat. Commun. 15, 2812 (2024). Kim, G. et al. Mott neurons with dual thermal dynamics for spatiotemporal computing. Nat. Mater. 23, 1237-1244 (2024). Koirala, K. P., Sandireddy, V. P., Garcia, H., Duscher, G. & Kalyanaraman, R. Nanosecond switchable localized surface plasmons through resettable contact angle behavior in silver nanoparticles. Nanotechnol. 31, 355503 (2020). Sandireddy, V. P., Koirala, K. P. & Kalyanaraman, R. In situ contact angle tuning of silver nanostructures by laser heating in different fluid ambient. Langmuir 35, 10744-10751 (2019). Yu, H., Peng, Y., Yang, Y. & Li, Z.-Y. Plasmon-enhanced light–matter interactions and applications. npj Comput. Mater. 5, 45 (2019). Wang, W. et al. Surface diffusion-limited lifetime of silver and copper nanofilaments in resistive switching devices. Nat. Commun. 10, 81 (2019). Valov, I., Waser, R., Jameson, J. R. & Kozicki, M. N. Electrochemical metallization memories—fundamentals, applications, prospects. Nanotechnol. 22, 254003 (2011). Shen, Z. et al. Advances of RRAM devices: Resistive switching mechanisms, materials and bionic synaptic application. Nanomaterials 10, 1437 (2020). Moon, J. et al. Temporal data classification and forecasting using a memristor-based reservoir computing system. Nat. Electron. 2, 480-487 (2019). Choi, S. et al. 3D-integrated multilayered physical reservoir array for learning and forecasting time-series information. Nat. Commun. 15, 2044 (2024). Liang, X. et al. Physical reservoir computing with emerging electronics. Nat. Electron. 7, 193-206 (2024). Li, W. et al. Dual‐Functional Nonvolatile and Volatile Memory in Resistively Switching Indium Tin Oxide/HfOx Devices. phys. status solidi (a) 216, 1900555 (2019). Yi, W. et al. Biological plausibility and stochasticity in scalable VO2 active memristor neurons. Nat. Commun. 9, 4661 (2018). Li, X. et al. Design of artificial neurons of memristive neuromorphic networks based on biological neural dynamics and structures. IEEE Trans.Circuits Syst. I 71, 2320-2333 (2024). Xiao, Y. et al. Bio-plausible reconfigurable spiking neuron for neuromorphic computing. Sci. Adv. 11, eadr6733 (2025). Kim, G. et al. Threshold modulative artificial GABAergic nociceptor. Adv. Mater. 35, 2304148 (2023). Yang, J.-Q. et al. Leaky integrate-and-fire neurons based on perovskite memristor for spiking neural networks. Nano Energy 74, 104828 (2020). Kim, S. J., Kim, S. & Jang, H. W. Competing memristors for brain-inspired computing. Iscience 24 101889 (2021). Chen, P.-Y., Peng, X. & Yu, S. NeuroSim+: An Integrated Device¬to¬Algorithm Framework for Benchmarking Synaptic Devices and Array Architectures. In 2017 IEEE International Electron Devices Meeting (IEDM) 6.1.1-6.1.4 (IEEE, 2017) Seo, S. et al. Artificial van der Waals hybrid synapse and its application to acoustic pattern recognition. Nat. Commun. 11, 3936 (2020). Wu, C., Kim, T. W., Choi, H. Y., Strukov, D. B. & Yang, J. J. Flexible three-dimensional artificial synapse networks with correlated learning and trainable memory capability. Nat. Commun. 8, 752 (2017). Prezioso, M. et al. Spike-timing-dependent plasticity learning of coincidence detection with passively integrated memristive circuits. Nat. Commun. 9, 5311 (2018). Milano, G. et al. In materia reservoir computing with a fully memristive architecture based on self-organizing nanowire networks. Nat. Mater. 21, 195-202 (2022). Lillicrap, T. P., Cownden, D., Tweed, D. B. & Akerman, C. J. Random synaptic feedback weights support error backpropagation for deep learning. Nat. Commun. 7, 13276 (2016). Zhang, T. et al. Self-backpropagation of synaptic modifications elevates the efficiency of spiking and artificial neural networks. Sci. Adv. 7, eabh0146 (2021). Pattnaik, D. et al. Stress-induced artificial neuron spiking in diffusive memristors. Commun. Eng. 3, 163 (2024). Additional Declarations There is NO Competing Interest. Supplementary Files SupplementaryInformation.docx Supplementary Information Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-6140700","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":433469403,"identity":"9ec9ee2d-5693-4886-8a6e-fd814350fac9","order_by":0,"name":"Keon Jae Lee","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA3ElEQVRIie3QsQqCQBjA8S+CpgvXTwR9BSVoCp/lRNCldqEgQTiXHkAJehZBOBejV1BaG4T26qyt4bSt4f7Dx3Hwg+8OQKX6w7AfDYCJMIk/V2QMoQCLNyl+IV7WH0YRPU14Q6NVmKdlcu/AtYCcGykxCA9tWgebI/EYFuA78Ty1pcTE9RI9Vm5O4DGx2JSCNpMvZlo3QR7P0NTapCtgP0wMJILEBTXEFIuVFOZMTvRDECDlvpNnLcParhxGuJxgVXLsdq6Fl/DaRdHW0kggJ1+Jvxp4iUqlUqnG9AKKuT+8wT0f5wAAAABJRU5ErkJggg==","orcid":"https://orcid.org/0000-0001-7500-3079","institution":"Korea Advanced Institute of Science and Technology (KAIST)","correspondingAuthor":true,"prefix":"","firstName":"Keon","middleName":"Jae","lastName":"Lee","suffix":""},{"id":433469404,"identity":"99614f3e-3c66-42a3-bd45-f3cc9bef6fcf","order_by":1,"name":"Yu Jin Jeong","email":"","orcid":"","institution":"Korea Advanced Institute of Science and Technology (KAIST)","correspondingAuthor":false,"prefix":"","firstName":"Yu","middleName":"Jin","lastName":"Jeong","suffix":""},{"id":433469405,"identity":"b2b349ef-7219-46d2-95ab-5a4046e8e0ca","order_by":2,"name":"Kyunghwan Kim","email":"","orcid":"","institution":"Korea Advanced Institute of Science and Technology (KAIST)","correspondingAuthor":false,"prefix":"","firstName":"Kyunghwan","middleName":"","lastName":"Kim","suffix":""},{"id":433469406,"identity":"3f8fac1c-0e18-41ff-9baf-0eff865b9625","order_by":3,"name":"Young Bin Kim","email":"","orcid":"","institution":"Korea Advanced Institute of Science and Technology (KAIST)","correspondingAuthor":false,"prefix":"","firstName":"Young","middleName":"Bin","lastName":"Kim","suffix":""},{"id":433469407,"identity":"388d93b4-ca28-4470-b910-219c3fba6dd2","order_by":4,"name":"Hyera Shin","email":"","orcid":"","institution":"Korea Advanced Institute of Science and Technology (KAIST)","correspondingAuthor":false,"prefix":"","firstName":"Hyera","middleName":"","lastName":"Shin","suffix":""},{"id":433469408,"identity":"b9410530-c29b-4d09-b125-0bfc90d55481","order_by":5,"name":"Jung Won Oh","email":"","orcid":"","institution":"Samsung Electronics Co.Ltd.","correspondingAuthor":false,"prefix":"","firstName":"Jung","middleName":"Won","lastName":"Oh","suffix":""},{"id":433469409,"identity":"108e0ce2-3d97-4bfb-b54e-7406eb78c0e9","order_by":6,"name":"Sang Hyeon Sung","email":"","orcid":"","institution":"Samsung Electronics Co.Ltd.","correspondingAuthor":false,"prefix":"","firstName":"Sang","middleName":"Hyeon","lastName":"Sung","suffix":""}],"badges":[],"createdAt":"2025-03-02 17:25:10","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6140700/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6140700/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":80198537,"identity":"931ce9ee-09f7-496c-9871-3bdf26ddfa05","added_by":"auto","created_at":"2025-04-09 06:18:49","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":451626,"visible":true,"origin":"","legend":"\u003cp\u003eStructure and dual-mode operation of the laser-driven memristor. aSchematic illustration of the dual-mode memristor array, where laser guidance induces a transition to nonvolatile states in selected regions, while adjacent pristine areas retain volatile behavior. Both volatile and nonvolatile modes share the same electrode plane and Ag reservoir layer. b Structural transition of Ag filaments in an Au/Ag/SiO₂:Ag/Pt memristor. Before laser exposure, dispersed Ag nanoparticles lead to volatile (threshold-switching) behavior. After laser irradiation, clustered Ag filaments yield stable nonvolatile states. c Biological analogy illustrating neuronal (left) and synaptic (right) functions. The device mimics neuronal signal integration via volatile switching and synaptic weight modulation via nonvolatile retention. d Conceptual diagram of bioinspired reservoir computing using a reconfigurable memristor. The volatile mode enables short-term temporal processing, while the nonvolatile mode supports long-term memory storage, integrating neuron- and synapse-like functionalities into a unified platform.\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-6140700/v1/f5c557213422e870c64d7c7a.png"},{"id":80198538,"identity":"d6e08144-182d-460e-82ce-f905a794a736","added_by":"auto","created_at":"2025-04-09 06:18:49","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":814119,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eLSPR-driven Ag nanoparticle clustering and electrical characteristics. a\u003c/strong\u003e Schematic of laser-induced Ag NP growth within the SiO₂:Ag matrix. Excimer laser irradiation at 308 nm enhances Ag-ion migration, promoting NP aggregation. \u003cstrong\u003eb \u003c/strong\u003eConceptual diagram of the LSPR effect on Ag NP clustering. In the initial state, Ag NPs remain dispersed within the SiO₂:Ag switching layer. Laser irradiation excites LSPR at Ag NP sites, inducing electron cloud oscillations and generating strong localized electric fields, which lower the migration barrier for Ag ions. \u003cstrong\u003ec\u003c/strong\u003e Cross-sectional TEM images showing Ag clustering under different laser fluence conditions. At 125 mJ/cm², Ag NPs remain dispersed, while fluences of 150–175 mJ/cm² induce larger Ag clusters. Scale bar, 5 nm. \u003cstrong\u003ed\u003c/strong\u003e COMSOL simulations of the near-field electric field distribution between two adjacent Ag NPs (radius = 2 nm, gap = 1 nm) under 308 nm laser irradiation. The strongest local electric field enhancement occurs at the inter-particle gap. Scale bar, 2 nm. \u003cstrong\u003ee \u003c/strong\u003eOptical microscopy image of a 4×4 memristor crossbar array. The pristine region (blue box) retains volatile switching, whereas the laser-irradiated region (red box) exhibits nonvolatile operation. Scale bar, 100 µm. \u003cstrong\u003ef\u003c/strong\u003e I–V characteristics of the volatile mode showing transient conduction, where RESET restores the high-resistance state upon bias removal. The inset depicts the Schottky emission mechanism. \u003cstrong\u003eg\u003c/strong\u003e I–V characteristics of the nonvolatile mode exhibit stable resistive switching. The device maintains a low-resistance state after SET operation, with Ohmic conduction dominating nonvolatile transport.\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-6140700/v1/b2c051ed2501727f466eee22.png"},{"id":80198540,"identity":"7e82c0dd-b1c6-4ddc-ad75-d880d399e3c5","added_by":"auto","created_at":"2025-04-09 06:18:49","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":16630185,"visible":true,"origin":"","legend":"\u003cp\u003eDual-mode neuromorphic properties in a unified memristor. a A neuronal membrane structure and circuit model. The memristor functions as an ion channel, while the capacitor represents the lipid bilayer, mimicking neuronal membrane behavior. b Synaptic weight modulation schematic, where neurotransmitter release strengthens or weakens synaptic connections, emulating biological plasticity. c Tonic spiking behavior under repeated pulses, demonstrating sustained firing in the volatile mode. d Phasic spiking behavior, characterized by burst discharges followed by quiescent intervals, attributed to repeated filament formation and dissolution. e Threshold variation demonstrating an all-or-nothing switching effect, where small voltage changes induce abrupt current transitions. f PPF observed in the nonvolatile mode. The transient increase in conductance under closely spaced pulses emulates synaptic plasticity and short-term memory effects. g LTP/LTD demonstrate synaptic weight modulation. The inset shows the input voltage waveform used for LTP/LTD induction. h Emulation of symmetric Hebbian learning rule. a laser-guided dual-mode memristor. The device follows a symmetric Hebbian learning rule, where conductance updates depend on the timing between pre- and post-spikes.\u003c/p\u003e","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-6140700/v1/0aa9526f3d4864df3b7b798c.png"},{"id":80198539,"identity":"407748e0-9c2d-4bb0-acb3-79e431a38bfe","added_by":"auto","created_at":"2025-04-09 06:18:49","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":607472,"visible":true,"origin":"","legend":"\u003cp\u003eNeuromorphic melody recognition and feedback-driven adaptation a Schematic of the reservoir computing (RC) system employing a unified memristor device. The volatile (neuron-like) reservoir dynamically processes input signals, while the nonvolatile (synapse-like) readout layer stabilizes learned patterns, enabling the system to identify target melodies such as \u003cem\u003e\"Jingle Bells\"\u003c/em\u003e. b Encoding scheme for melody recognition. Eight musical notes (C, D, E, F, F♯, G, A, B) are mapped to binary codes (000–111) and converted into pulse sequences. For example, “D” is encoded as 100 (a single pulse followed by a consecutive pulse). c Confusion matrix comparing actual and predicted outputs across five encoded melodies, demonstrating high classification accuracy. d Recognition accuracy over five training epochs, showing continuous improvement as the RC system optimizes connection weights in the readout layer. e Feedback-induced spiking adaptation in a volatile-nonvolatile coupled memristor device. The system processes voltage inputs and current readouts to analyze feedback-driven modulation in response to external stimuli. Spiking frequency and amplitude increase progressively with repeated pulse stimulation, demonstrating the synergistic interaction between volatile and nonvolatile modes, analogous to biological synaptic potentiation.\u003c/p\u003e","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-6140700/v1/12322a18f7ab5ae0a7c2447a.png"},{"id":80199827,"identity":"cfa45977-cdff-4963-9481-0ec8b62e4734","added_by":"auto","created_at":"2025-04-09 06:34:55","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":19271156,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6140700/v1/7a1eeb1f-46c7-4a28-b67b-967ae72cea22.pdf"},{"id":80199395,"identity":"d793f1df-1aee-4027-a5fe-55ecce98197f","added_by":"auto","created_at":"2025-04-09 06:26:49","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":4167781,"visible":true,"origin":"","legend":"Supplementary Information","description":"","filename":"SupplementaryInformation.docx","url":"https://assets-eu.researchsquare.com/files/rs-6140700/v1/722575ff19eaa743c9fea23f.docx"}],"financialInterests":"There is \u003cb\u003eNO\u003c/b\u003e Competing Interest.","formattedTitle":"Laser-Guided Ion Dynamics in a Dual-Mode Memristor for Bioinspired Neuronal and Synaptic Integration","fulltext":[{"header":"Main","content":"\u003cp\u003eThe human brain achieves remarkable efficiency by integrating neuronal and synaptic functions for real-time adaptation and long-term memory. Emulating this dual functionality is critical for energy-efficient, scalable neuromorphic hardware capable of dynamic learning and cognitive tasks\u003csup\u003e\u003cspan additionalcitationids=\"CR2 CR3 CR4\" citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e\u003c/sup\u003e. However, despite progress in CMOS-based neuromorphic circuits, current hardware rigidly separates neuronal and synaptic processes, even though transient neuronal activity and stable synaptic memory must function in tandem. Furthermore, the conventional electrical stimulus lacks the spatial selectivity required to control volatile and nonvolatile states independently\u003csup\u003e\u003cspan additionalcitationids=\"CR7\" citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u003c/sup\u003e. A strategy that provides localized and reconfigurable neurosynaptic connections by mimicking the 3D functionality of the brain is a potential means of overcoming these limitations.\u003c/p\u003e \u003cp\u003eWhile memristors are promising candidates for integrating neuronal and synaptic functions, most implementations rely on external circuit-based links, rather than a unified material for bifunctionality\u003csup\u003e\u003cspan additionalcitationids=\"CR10 CR11\" citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u003c/sup\u003e. Direct neuron-synapse integration using a stacked architecture to support brain-like computation has attracted considerable attention for memristor-based neurosynaptic systems\u003csup\u003e\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e,\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e\u003c/sup\u003e. However, previous neurosynaptic integrations, including stacked or circuit-based neuron-synapse connections and material-constrained non-tunable volatility, have faced fundamental challenges such as insufficient spatial selectivity, lack of direct neurosynaptic interconnection, and limited freedom of ion transport control\u003csup\u003e\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e,\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eLaser technologies, widely adopted in semiconductor patterning\u003csup\u003e\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e,\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e\u003c/sup\u003e and surface modification\u003csup\u003e\u003cspan additionalcitationids=\"CR20\" citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e\u003c/sup\u003e, offer precise control of material properties and ionic conductivity\u003csup\u003e\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e\u003c/sup\u003e. Unlike conventional microfabrication with limited spatial selectivity, laser energy transfer can drive localized structural and electronic property changes and thereby support tunable neurosynaptic behaviors\u003csup\u003e\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e\u003c/sup\u003e. However, its application to selective modulation of volatile-nonvolatile transitions within a homogeneous composition has not been explored. Here, we report a laser-guided dual-mode memristor that enables direct neurosynaptic interconnection by selectively controlling ion dynamics within a single-phase material. A pristine Ag-doped SiO₂ threshold switch (TS) exhibits frequency-dependent neuronal firing, while laser-irradiated regions transition into synapse-like conductive states via Ag clustering. The direct connectivity between the laser-induced synapse and TS neuron demonstrates neuromorphic reservoir computing (RC) through a reconfigurable crossbar architecture. The dual-mode neurosynaptic memristor enhances RC efficiency by integrating short-term neuronal dynamics and long-term synaptic plasticity. This tunability facilitates \u003cem\u003ein-situ\u003c/em\u003e adaptive learning and a positive feedback in synaptic strength.\u003c/p\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eLaser-guided dual-mode memristor with neuron-synapse pair\u003c/h2\u003e \u003cp\u003eThe laser-guided dual-mode memristor integrates neuron- and synapse-like functionalities within a single layer. We developed an integrated neurosynaptic architecture, shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e(a), in which a pristine Ag-doped SiO₂ TS exhibits volatile switching and selectively transitions to a nonvolatile state upon XeCl excimer laser irradiation. To fabricate the volatile TS layer, an Au bottom electrode was deposited, followed by a 1 nm Ag reservoir layer, a 15 nm Ag-doped SiO₂ switching layer, and a Pt top electrode. Within the crossbar array, localized laser exposure on the Ag:SiO₂ matrix stabilizes the filaments and converts specific regions to nonvolatile operation (see Methods and Supplementary Fig.\u0026nbsp;1 for fabrication details). The excimer laser was optimized at 308 nm, 150\u0026ndash;175 mJ/cm\u0026sup2;, with a 25 ns pulse width (see Supplementary Fig.\u0026nbsp;2 for details).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eFigure\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e(b) presents the device's behavior before and after laser irradiation of the Au/Ag/SiO₂:Ag/Pt structure. Before laser exposure, dispersed Ag nanoparticles form transient filaments, resulting in volatile switching that mimics rapid signaling dynamics of biological neurons\u003csup\u003e\u003cspan additionalcitationids=\"CR25\" citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e\u003c/sup\u003e. After laser irradiation, plasmonic resonance and electromagnetic gradients drive Ag filament aggregation. The aggregated Ag filaments stabilize the conduction path and transition the TS region into a nonvolatile state\u003csup\u003e\u003cspan additionalcitationids=\"CR28 CR29 CR30\" citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e\u003c/sup\u003e. This stable reconfiguration affords synaptic plasticity for adaptive learning in neuromorphic systems. Figure\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e(c) depicts the volatile mode that provides transient signal processing and action potential firing in neuronal plasticity. The nonvolatile mode mirrors neurotransmitter release in synaptic plasticity for long-term memory consolidation\u003csup\u003e\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e\u003c/sup\u003e. These complementary mechanisms reflect the hierarchical structure of neural networks, where short-term adaptability and long-term retention coexist\u003csup\u003e\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eLaser-guided ion dynamics provide the dual-mode memristor with the capability of adaptive neurosynaptic reconfiguration. Figure\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e(d) illustrates a bioinspired reservoir computing scenario based on the dual-mode operation of the reconfigurable memristor. The volatile mode functions as a physical reservoir for processing temporal signals whereas the nonvolatile mode serves as a readout layer for storing stable synaptic weights\u003csup\u003e\u003cspan additionalcitationids=\"CR34\" citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e\u003c/sup\u003e. Laser-guided Ag filament clustering provides precise spatial control over volatility and allows direct neurosynaptic interactions in a single-phase material. Dynamically controlled ion transport supports adaptive connectivity and in turn reconfigurable neuromorphic architectures can be realized\u003csup\u003e\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e,\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eLaser-induced ion dynamics and resistive switching in a dual-mode memristor\u003c/h3\u003e\n\u003cp\u003eLaser irradiation induces localized energy absorption within the Ag-doped SiO₂ matrix, which triggers a temperature rise that enhances Ag-ion mobility. As the excitation energy propagates, localized surface plasmon resonance (LSPR) amplifies the local electric field, optical absorption, and structural transformations\u003csup\u003e\u003cspan additionalcitationids=\"CR28\" citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e\u003c/sup\u003e. Figure\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e(a) illustrates the 308 nm excimer laser-driven material transition into Ag nanoparticles (Ag NPs) growth for selective volatility control in dual-mode memristor. The LSPR effect on Ag NP clustering is depicted in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e(b), as electron cloud oscillations generate strong local electric fields\u003csup\u003e\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e,\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e\u003c/sup\u003e. Initially, Ag NPs are uniformly distributed within the SiO₂:Ag switching layer, with the reservoir layer serving as an ion source\u003csup\u003e\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e,\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e\u003c/sup\u003e. Upon laser irradiation, LSPR generates strong local electric fields and electron cloud oscillations that lower the energy barrier for Ag-ion transport\u003csup\u003e\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e,\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eTo validate this structural modification, cross-sectional Transmission Electron Microscopy (TEM) images were obtained at different fluences, as shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e(c). At 125 mJ/cm\u0026sup2;, Ag NPs remain dispersed with diameters ranging from 1.5 to 3 nm. This indicates that the energy is insufficient to induce significant clustering and filament formation. As the fluence increased to 150\u0026ndash;175 mJ/cm\u0026sup2;, Ag NPs coalesced into larger clusters (~\u0026thinsp;27 nm) consistent with the LSPR-enhanced aggregation. The resonance effect is attributed to LSPR rather than conventional SPR due to the discrete nanoparticle structure of Ag, which contrasts with the continuous film required for SPR. Fluences over 175 mJ/cm\u0026sup2;, however, result in excessive filament formation due to structural instabilities and degrade electrical performance (Supplementary Fig.\u0026nbsp;3).\u003c/p\u003e \u003cp\u003eUV-Vis absorption spectra were analyzed before and after laser treatment, as shown in Supplementary Fig.\u0026nbsp;4. A noticeable redshift in the absorption peak was observed and indicated an increase in Ag NP size. The observed redshift is attributed to the formation of larger Ag NP clusters, which modify the local dielectric environment and shift the LSPR resonance condition. Further analysis using energy-dispersive X-ray spectroscopy (EDS) mapping confirms the redistribution of Ag NPs under laser irradiation (Supplementary Fig.\u0026nbsp;5). Fast Fourier Transform (FFT) analysis of the TEM images reveals crystalline domains with d-spacing values, matching theoretical Ag lattice parameters (Supplementary Figs.\u0026nbsp;6, 7). X-ray photoelectron spectroscopy (XPS) and Time-of-Flight Secondary Ion Mass Spectrometry (ToF-SIMS) profiling confirm the change of Ag chemical states and support LSPR-mediated structural modifications (Supplementary Figs.\u0026nbsp;8\u0026ndash;10).\u003c/p\u003e \u003cp\u003eFigure\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e(d) presents COMSOL Multiphysics simulations of the electric field distribution between Ag NPs (radius\u0026thinsp;=\u0026thinsp;2 nm, gap\u0026thinsp;=\u0026thinsp;1 nm) performed at a frequency of 9.73\u0026times;10\u003csup\u003e14\u003c/sup\u003e Hz, corresponding to 308 nm wavelength of laser. The strong local electric fields generated by LSPR lower the clustering barrier by enhancing energy absorption and inducing localized heating\u003csup\u003e\u003cspan additionalcitationids=\"CR28\" citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e\u003c/sup\u003e. The simulation results confirm that localized plasmonic enhancement strengthens Ag-ion migration and establishes the necessary conditions for filament formation. Supplementary Figs.\u0026nbsp;11\u0026ndash;13 present the dependence of the nanoparticle size and gap on field enhancement and its role in structural transformations within the switching layer.\u003c/p\u003e \u003cp\u003eFigure\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e(e) shows an optical microscopy image of the fabricated 4\u0026times;4 crossbar array of memristors, each initially designed as a volatile device. A single laser shot selectively induces nonvolatile transitions in targeted cells while preserving volatility in non-irradiated areas. The blue-box region retains original threshold switching characteristics, while the irradiated red-box region is transformed into stable nonvolatile pathways. This partial conversion yields both volatile and nonvolatile regions that coexist within the same crossbar and offers flexible neuron-synapse functionality with short-term and long-term memory in one architecture.\u003c/p\u003e \u003cp\u003eIn the volatile regime in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e(f), a threshold-switching mechanism emerges near 0.24 V. The mechanism creates transient Ag filaments under bias, which revert to a high-resistance state upon bias removal. Conduction returns to a high-resistance state within 0.6 ms (Supplementary Fig.\u0026nbsp;14), mirroring transient neuronal signals\u003csup\u003e\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u003c/sup\u003e. Repeated switching over 100 cycles confirms stable threshold behavior with minimal voltage variation (Supplementary Fig.\u0026nbsp;15). Schottky emission governs the conduction mechanism, as shown in the inset of Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e(f), which indicates that electrons traverse a potential barrier at the metal-dielectric interface rather than forming a fully metallic path\u003csup\u003e\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e\u003c/sup\u003e. Figure\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e(g) presents the nonvolatile bipolar switching behavior, where the memristor transitions from a high-resistance state (HRS) to a low-resistance state (LRS) at an average SET voltage of approximately 0.41 V; conduction persists after bias removal, verifying stable retention. The conduction mechanism shifts from Schottky to Ohmic behavior, and continuous metallic pathways replace barrier-limited electron transport\u003csup\u003e\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e\u003c/sup\u003e. Extended retention measurements confirm stable conduction beyond 8,250 seconds\u003csup\u003e\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e\u003c/sup\u003e, while repeated switching over 100 cycles exhibits robust threshold behavior with minimal voltage variation.\u003c/p\u003e \u003cp\u003eMulti-shot laser processing achieves tunable retention from milliseconds to 2,000 s by incrementally accumulating energy at a fluence of 102 mJ/cm\u0026sup2;, surpassing single-shot or thermal methods (Supplementary Fig.\u0026nbsp;16). To compare the effects of laser irradiation with conventional thermal processing, Rapid Thermal Annealing (RTA) was performed under different conditions (Supplementary Fig.\u0026nbsp;17). Annealing at 400\u0026deg;C maintained volatility, while treatment at 550\u0026deg;C led to severe degradation of memory functionality. These results highlight the advantages of laser processing in achieving controlled structural modifications without inducing significant thermal damage.\u003c/p\u003e\n\u003ch3\u003eIntegrated emulation of neuronal spiking and synaptic plasticity\u003c/h3\u003e\n\u003cp\u003eThe laser-guided dual-mode memristor emulates neuron-like spiking and synaptic plasticity within a single-phase material. This capability provides seamless transitions between short-term memory (STM) and long-term memory (LTM) and effectively replicates the spatiotemporal dynamics of biological neural networks.\u003c/p\u003e \u003cp\u003eFigure\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e(a) presents the neuronal membrane structure and its circuit, where the memristor mimics a voltage-gated ion channel and the parallel capacitor represents the lipid bilayer. In biological neurons, AP generation occurs through the movement of ions (Na⁺, K⁺, Cl⁻) across the lipid bilayer\u003csup\u003e\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e\u003c/sup\u003e. The capacitor separates charge and generates a potential difference, while the memristor modulates current flow through Ag filament formation, mimicking the gating dynamics of ion channels (see Supplementary Figs.\u0026nbsp;18,19, Supplementary note 1). Signal transmission occurs when neurotransmitters are released from presynaptic vesicles and bind to postsynaptic receptors, generating an excitatory postsynaptic potential (EPSP)\u003csup\u003e\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e\u003c/sup\u003e, as shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e(b).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eThe memristor\u0026rsquo;s volatile mode reproduces fundamental neuronal spiking patterns, essential for sensory processing, motor control, and higher cognitive functions\u003csup\u003e\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e,\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e\u003c/sup\u003e. Figures\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e(c)\u0026ndash;(e) show representative tonic spiking, phasic spiking, and threshold variation, while additional neuron spikes are presented in Supplementary Figs.\u0026nbsp;20\u0026ndash;22. In a parallel RC circuit, the repetitive charging and discharging of the capacitor drive the voltage spiking behavior. The \u003cem\u003en\u003c/em\u003eth charging (\u003cem\u003eVₙ\u003c/em\u003e\u003csup\u003e\u003cem\u003ech\u003c/em\u003e\u003c/sup\u003e) and discharging (\u003cem\u003eVₙ\u003c/em\u003e\u003csup\u003e\u003cem\u003edis\u003c/em\u003e\u003c/sup\u003e) functions govern the capacitor\u0026rsquo;s voltage evolution in the circuit, where \u003cem\u003eI\u003c/em\u003e\u003csub\u003e\u003cem\u003ein\u003c/em\u003e\u003c/sub\u003e, \u003cem\u003etₙ\u003c/em\u003e\u003csup\u003e\u003cem\u003ech\u003c/em\u003e\u003c/sup\u003e, \u003cem\u003etₙ\u003c/em\u003e\u003csup\u003e\u003cem\u003edis\u003c/em\u003e\u003c/sup\u003e, \u003cem\u003eτ\u003c/em\u003e\u003csub\u003e\u003cem\u003ech\u003c/em\u003e\u003c/sub\u003e, and \u003cem\u003eτ\u003c/em\u003e\u003csub\u003e\u003cem\u003edis\u003c/em\u003e\u003c/sub\u003e denote the input current, charging/discharging time, and time constants.\u003cdiv id=\"Equ1\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ1\" name=\"EquationSource\"\u003e\n$$\\:{V}_{n}^{ch}\\left(t\\right)={I}_{in}{R}_{H}\\left\\{1-{exp}\\left(-\\frac{t-{t}_{n}^{ch}}{{\\tau\\:}_{ch}}\\right)\\right\\}$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e1\u003c/div\u003e\u003c/div\u003e\u003cdiv id=\"Equ2\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ2\" name=\"EquationSource\"\u003e\n$$\\:{V}_{n}^{dis}\\left(t\\right)=I{R}_{L}{exp}\\left(-\\frac{t-{t}_{n}^{dis}-{\\tau\\:}_{dis}{ln}\\left({V}_{th}/I{R}_{L}\\right)}{{\\tau\\:}_{dis}}\\right)$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e2\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003eWhen \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{I}_{in}{R}_{H}\\:\\)\u003c/span\u003e\u003c/span\u003eexceeds \u003cem\u003eV\u003c/em\u003e\u003csub\u003e\u003cem\u003eth\u003c/em\u003e\u003c/sub\u003e (and \u003cem\u003eV\u003c/em\u003e\u003csub\u003e\u003cem\u003eth\u003c/em\u003e\u003c/sub\u003e \u0026gt; 0), the capacitor voltage rises to \u003cem\u003eV\u003c/em\u003e\u003csub\u003e\u003cem\u003eth\u003c/em\u003e\u003c/sub\u003e according to the charging equation (Eq.\u0026nbsp;(\u003cspan refid=\"Equ1\" class=\"InternalRef\"\u003e1\u003c/span\u003e)). At that threshold, the memristor abruptly switches from \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{R}_{H}\\:\\)\u003c/span\u003e\u003c/span\u003eto \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{R}_{L}\\)\u003c/span\u003e\u003c/span\u003e, causing the capacitor to discharge from \u003cem\u003eV\u003c/em\u003e\u003csub\u003e\u003cem\u003eth\u003c/em\u003e\u003c/sub\u003e to \u003cem\u003eV\u003c/em\u003e\u003csub\u003e\u003cem\u003emin\u003c/em\u003e\u003c/sub\u003e (Eq.\u0026nbsp;(\u003cspan refid=\"Equ2\" class=\"InternalRef\"\u003e2\u003c/span\u003e)). When a 0.8 \u0026micro;A input current is applied, the memristor with the 2.2 nF parallel capacitor generates stable tonic spiking, as seen in Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e(c). Increasing the capacitance extends the average spike period by slowing the charging and discharging. The voltage spike period (\u003cem\u003eT\u003c/em\u003e) for the time-dependent charging and discharging equation is defined by the following equation (see Supplementary note 1 for details).\u003cdiv id=\"Equ3\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ3\" name=\"EquationSource\"\u003e\n$$\\:T=C\\left[{R}_{H}{ln}\\left(\\frac{I{R}_{H}-{V}_{min}}{I{R}_{H}-{V}_{th}}\\right)+{R}_{L}{ln}\\left(\\frac{{V}_{th}}{{V}_{min}}\\right)\\right]$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e3\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003eBecause the time constant τ scales with the capacitance C, the spike period T also depends on C. The average spike period increases from 9.2 ms to 17 ms as the capacitance rises from 2.2 nF to 10 nF in Supplementary Fig.\u0026nbsp;20. Raising the input current to 2 \u0026micro;A shifts the behavior from tonic spiking to tonic bursting, characterized by periodic clusters of spikes with quiescent periods exceeding 300 ms. This indicated a higher excitability threshold and enhanced burst firing, akin to cortical neuron activity. The memristor RC circuit replicates inhibition-induced spiking/bursting, where a negative current pulse induces rebound firing. After prolonged inhibitory input, the memristor transitions to rhythmic burst firing and mirrors T-type calcium channel-mediated spikes\u003csup\u003e\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003ePhasic spiking is a bursting pattern with silent intervals and resembles thalamic relay cell activity, which regulates attention gating and network synchronization\u003csup\u003e\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e\u003c/sup\u003e (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e(d)). At a higher current level of 5 \u0026micro;A, the electric field exceeds the Ag-ion diffusion rates, causing intermittent filament formation and dissolution. The volatile region shows threshold variation when alternating\u0026thinsp;\u0026plusmn;\u0026thinsp;1.2 V pulses are applied, revealing the device\u0026rsquo;s sensitivity to small voltage changes (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e(e)). This adaptive threshold mechanism mirrors the influence of ionic dynamics and synaptic conditions on firing thresholds in biological neurons\u003csup\u003e\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e,\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e\u003c/sup\u003e. The laser-guided dual-mode memristor experimentally validates the leaky integrate-and-fire (LIF) model, where abrupt firing occurs as repeated input signals are accumulated to reach \u003cem\u003eV\u003c/em\u003e\u003csub\u003e\u003cem\u003eth\u003c/em\u003e\u003c/sub\u003e (see Supplementary Figs.\u0026nbsp;22). The all-or-nothing principle, which controls action potential generation in excitable cells, is exhibited in the device. It ensures binary signal transmission, essential for cortical and motor neurons\u003csup\u003e\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e,\u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eThe nonvolatile characteristics of laser-guided dual-mode memristor replicate synaptic plasticity. Figure\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e(f) demonstrates paired-pulse facilitation (PPF), a short-term synaptic plasticity mechanism where two closely spaced presynaptic spikes cause an increased postsynaptic response. This behavior results from the finite retention time of Ag conductive filaments\u003csup\u003e\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e\u003c/sup\u003e, which enlarge under repeated input and temporarily increase conductance\u003csup\u003e\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eFigure\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e(g) presents long-term potentiation (LTP) and depression (LTD), fundamental for learning and memory. Potentiation was induced by applying 0.4\u0026ndash;1.0 V, 2 ms pulses in 0.01 V increments, while a -0.2 V, 2 ms pulses effected depression (V\u003csub\u003eread\u003c/sub\u003e at 0.1 V). To validate LTP/LTD nonlinearity, the open-source MATLAB tool Neurosim V3.0 was used, confirming the device\u0026rsquo;s ability to modulate synaptic weights accurately. The analysis yielded a nonlinearity of 2.03 for LTP and 4.88 for LTD, a symmetricity value of 3.09 \u0026times; 10⁵, and a dynamic range (G\u003csub\u003emax\u003c/sub\u003e / G\u003csub\u003emin\u003c/sub\u003e) of 144.1. The energy consumption at 0.1 V was 274 pJ/\u0026micro;m\u0026sup2; for LTP and 52 pJ/\u0026micro;m\u0026sup2; for LTD, indicating low-power potential for neuromorphic systems\u003csup\u003e\u003cspan additionalcitationids=\"CR44\" citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e\u003c/sup\u003e (see Supplementary Fig.\u0026nbsp;23\u0026ndash;25 and note 2\u0026ndash;4).\u003c/p\u003e \u003cp\u003eFigure\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e(h) verifies the symmetric Hebbian learning rule, a core principle of associative learning. Identical pre- and post-spike pulse conditions (0.6 V, 1 ms) were applied. The device\u0026rsquo;s conductance update follows an exponential decay, which confirms adaptive synaptic strength based on input timing\u003csup\u003e\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\n\u003ch3\u003eNeuromorphic melody recognition and feedback-driven adaptation\u003c/h3\u003e\n\u003cp\u003eThe RC system enables melody recognition by processing binary-encoded musical sequences using a dual-mode memristor\u003csup\u003e\u003cspan additionalcitationids=\"CR34\" citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e\u003c/sup\u003e. Figure\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e(a) illustrates the RC simulation, where the memristor-based reservoir processes input signals and the nonvolatile readout layer stores synaptic weights for classification. The memristor\u0026rsquo;s threshold-switching behavior in volatile mode generates 80 virtual nodes that transform time-series inputs into distinct current spikes. Encoded spike outputs are then processed by a nonvolatile readout network of a 80 \u0026times; 5 crossbar for weight storage and computation.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eFive representative melodies (Mary Had a Little Lamb, Twinkle Twinkle Little Star, Jingle Bells, Ode to Joy, and London Bridge Is Falling Down) were classified by the RC system. Figure\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e(b) illustrates the binary-to-pulse encoding scheme, where each musical pitch (C, D, E, F, F♯, G, A, and B) is mapped to binary codes (000\u0026ndash;111). Binary \u0026lsquo;0\u0026rsquo; corresponds to a single 0.7 V, 2 ms pulse that induces a transient threshold-switching event in volatile mode, while binary \u0026lsquo;1\u0026rsquo; is represented by two consecutive 0.7 V pulses separated by 1 ms for an extended volatile response. To maintain echo-state properties, a 4 ms interval was introduced between consecutive symbols (e.g., \u0026lsquo;0\u0026rsquo;\u0026ndash;\u0026lsquo;0\u0026rsquo;, \u0026lsquo;0\u0026rsquo;\u0026ndash;\u0026lsquo;1\u0026rsquo;, and \u0026lsquo;1\u0026rsquo;\u0026ndash;\u0026lsquo;1\u0026rsquo;) and a 7 ms gap between distinct notes (e.g., \u0026lsquo;C\u0026rsquo;-\u0026lsquo;D\u0026rsquo;, \u0026lsquo;F\u0026rsquo;-\u0026lsquo;A\u0026rsquo;, and \u0026lsquo;F#\u0026rsquo;-\u0026lsquo;A\u0026rsquo;). The interval allows clear differentiation between new input signals and previously processed data. With each note lasting 30 ms, a melody composed of eight notes had a total duration of 240 ms per sequence.\u003c/p\u003e \u003cp\u003ePost-synaptic current (PSC) values were sampled every 3 ms and 80 virtual nodes that preserved the spatiotemporal input structure were generated. Gaussian noise was applied to create 5,000 variations per melody, yielding a dataset of 25,000 samples (20,000 for training, 5,000 for testing). The memristor\u0026rsquo;s threshold-switching converted input pulse streams into distinct current spikes at each virtual node. Classification thus could be performed by the intrinsic switching characteristics rather than by predefined computational rules\u003csup\u003e\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e\u003c/sup\u003e. Following the reservoir stage, the 80 virtual node outputs were utilized in an 80 \u0026times; 5 readout layer, which stores and updates weights based on LTP/LTD. The weight adaptation uses \u003cem\u003ein-situ\u003c/em\u003e backpropagation by the device-pair configurations (G⁺ and G⁻)\u003csup\u003e\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e,\u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e49\u003c/span\u003e\u003c/sup\u003e. The nonlinearity of 2.03 for LTP and 4.88 for LTD, a symmetricity value of 3.09 \u0026times; 10⁵, and a dynamic range (G\u003csub\u003emax\u003c/sub\u003e/G\u003csub\u003emin\u003c/sub\u003e) of 144.1 reveal excellent synaptic programmability. Moreover, the energy consumption at 0.1 V was 274 pJ /\u0026micro;m\u0026sup2; for LTP and 52 pJ /\u0026micro;m\u0026sup2; for LTD. Unlike conventional reservoir computing systems that require individual components for dynamic processing and memory, this approach employs a single memristor, making it feasible for low-power neuromorphic applications\u003csup\u003e\u003cspan additionalcitationids=\"CR44\" citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eFigure\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e(c) presents a confusion matrix comparing actual and predicted outputs across five encoded melodies for high classification accuracy. After five training epochs, the RC system reached 94.34% accuracy, as shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e(d), highlighting the effectiveness of \u003cem\u003ein-situ\u003c/em\u003e weight adaptation. To validate the simulation-driven RC approach, we conducted experimental measurements on the dual-mode memristor (see Supplementary Fig.\u0026nbsp;26 for the current response under varied pulse amplitudes/widths in volatile mode). We verified robust echo-state properties at different pulse intervals, as presented in Supplementary Fig.\u0026nbsp;27, while Supplementary Fig.\u0026nbsp;28 provides details of the separability of the final conductance states for 16 different \u0026ldquo;0/1\u0026rdquo; input sequences. Furthermore, the pitch configuration (000\u0026ndash;111) and melody-specific data for C\u0026ndash;B notes, depicted in Supplementary Figs.\u0026nbsp;29\u0026ndash;30, illustrate how the device encodes musical sequences as transient spikes.\u003c/p\u003e \u003cp\u003eIn addition to classification, the RC system exhibits biologically inspired feedback adaptation for dynamic neuromorphic learning. Figure\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e(e) presents a neuron-synapse coupled scheme to experimentally verify the feedback effect. The integrated array of directly connected neuro-synaptic memristors allows on-chip transient spiking and stable weight retention. The dual-mode synergy supports functional feedback loops, signal processing, and adaptive learning within a unified hardware framework. Feedback-driven spiking adaptation in a coupled volatile\u0026ndash;nonvolatile memristor, where repeated pulse stimulation progressively increased spiking frequency and amplitude. A pulse train with 20 ms intervals progressively increased spiking frequency and amplitude over five repetitions. This mirrors biological synaptic potentiation in which repeated stimulation enhances neurotransmitter release\u003csup\u003e\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e\u003c/sup\u003e. First-spike latency decreased from 36.2 ms to 0.32 ms, while the peak current reached 0.24 mA. The memristor\u0026rsquo;s threshold-switching physics facilitates charge accumulation and lowers the activation energy required for filament formation, resulting in faster or stronger spiking responses\u003csup\u003e\u003cspan additionalcitationids=\"CR25\" citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e,\u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e50\u003c/span\u003e\u003c/sup\u003e. The experimental results confirm that a single dual-mode memristor in an RC-based neuromorphic system can incorporate both transient neuronal dynamics and stable synaptic plasticity. This capability provides a hardware-efficient path toward real-time pattern recognition and adaptive learning.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eThe laser-guided dual-mode memristor applies LSPR to achieve dynamic control over transient and stable conduction states within a single Ag-doped SiO₂ matrix. Excimer laser fluences of 150–175 mJ/cm² induce Ag-ion migration and clustering, forming 5–27 nm Ag nanoclusters, as observed by TEM. These clusters establish robust conduction pathways, enabling tunable volatility control and extended retention, a capability that surpasses conventional memristive devices reliant on static material properties. In the volatile regime, the device operates at a 0.24 V threshold with a conduction decaying within 0.6 ms, approximating rapid neuronal activity. Transition to the nonvolatile mode achieves a 3.4×10⁶ ON/OFF ratio, retention beyond 8,250 s, and a dynamic range of 144.1, with energy consumption as low as 274 pJ/µm\u003csup\u003e2\u003c/sup\u003e for potentiation and 52 pJ/µm\u003csup\u003e2\u003c/sup\u003e for depression. The ability to selectively modulate volatility within the same material system introduces a reconfigurable approach to neuromorphic computing, reducing energy overhead associated with separate volatile and nonvolatile components. Integrated into a reservoir computing framework, the dual-mode device yields 94.34% accuracy for melody recognition and exhibits positive feedback adaptation under repeated stimulation. Unlike conventional reservoir computing systems that require separate components for dynamic processing and memory storage, this approach consolidates both functionalities within a single memristor, significantly reducing hardware complexity and power consumption. The synergy between transient neuronal dynamics and stable synaptic weight retention addresses a critical gap in neuromorphic hardware, providing a direct link between computational neuroscience principles and practical AI implementations. A pathway emerges for large-scale, energy-efficient architectures that require real-time adaptability and self-directed learning in advanced AI hardware.\u003c/p\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003cp\u003e\u003c/p\u003e \u003cp\u003e\u003c/p\u003e \u003c/div\u003e"},{"header":"Methods","content":"\u003cp\u003e \u003cb\u003eFabrication of dual-mode neuromorphic memristor.\u003c/b\u003e A 4-inch silicon wafer was prepared for the fabrication of dual-mode neuromorphic memristors. A 5-µm-diameter cross-point Au/Ag/SiO₂:Ag/Pt memristor was then integrated on the substrate using the following steps. First, a 5-nm-thick Cr adhesion layer and a 30-nm-thick Au bottom electrode were deposited by e-beam evaporation (SNTEK). Next, a bottom contact pad was patterned via a conventional photolithography process (Midas MDA-600S), followed by wet etching. The switching medium of the volatile TS layer was deposited by sputtering a 1 nm thick Ag layer and a 15 nm thick Ag:SiO\u003csub\u003e2\u003c/sub\u003e layer. To deposit Ag:SiO₂, Ag and SiO₂ targets were co-sputtered in an Ar atmosphere. Subsequently, an XeCl excimer laser was directed onto a moving stage that scanned selective areas of the device to modulate the volatility of the TS layer. A 5 µm-diameter via was defined by photolithography, followed by RF sputtering of an SiO₂ insulating layer in an Ar atmosphere. After lift-off, the SiO₂ layer remained intact except in the via. Furthermore, the lift-off step exposed the switching layer while maintaining electrical connectivity with the bottom electrode. Finally, a top contact pad was defined by a conventional lithography process, followed by deposition of a 30 nm thick Pt (Ateck) layer and lift-off.\u003c/p\u003e\u003cp\u003e \u003cb\u003eExcimer Laser Annealing (ELA).\u003c/b\u003e A XeCl excimer laser (Coherent, COMPex Pro 205, λ = 308 nm, pulse duration 25 ns) was used for selective laser irradiation on top of the Ag:SiO\u003csub\u003e2\u003c/sub\u003e layer in the crossbar array. The ELA system was composed of a beam homogenizer, attenuator, and delivery optics designed to produce a flat-top, square beam profile. It also has an XY linear stage (Dukin, SLS-200) that provides full-area substrate scanning. It controls the irradiating fluence with a flat-top square beam on the sample. The laser shot was scanned along a “zig-zag” shaped pattern with an optimized energy density level.\u003c/p\u003e\u003cp\u003e \u003cb\u003eElectrical characterization.\u003c/b\u003e All electrical measurements were conducted using a Keithley 4200-SCS semiconductor parameter analyzer. A Keithley 4225-PMU with a remote amplifier/switch (Keithley 4225-RPM) was utilized for the voltage pulse measurements. In the spiking emulations, voltage spikes were measured by a Tektronix DPO 3054 digital phosphor oscilloscope with a P6139B voltage probe (10 MΩ input resistance). A conventional ceramic capacitor with 2.2 nF or 10 nF capacitance was connected in parallel to the memristor RC circuit.\u003c/p\u003e\u003cp\u003e \u003cb\u003eElectric Field Simulation.\u003c/b\u003e A finite-element approach in COMSOL Multiphysics (version 6.2) was employed to calculate the spatial distribution of the electric field in the electromagnetic wave, frequency domain (emw) module. The governing equation was solved under time‐harmonic conditions, satisfying Maxwell’s equations:\u003c/p\u003e\u003cdiv id=\"Equa\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equa\" name=\"EquationSource\"\u003e\n$$\\:\\nabla\\:\\times\\:{{\\mu\\:}_{r}}^{-1}(\\nabla\\:\\times\\:\\text{E})-{k}_{0}^{2}\\left({ϵ}_{r}-j\\frac{\\sigma\\:}{\\omega\\:{ϵ}_{0}}\\right)E=0$$\u003c/div\u003e\u003c/div\u003e\u003cp\u003ewhere \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\mu\\:}_{r}\\)\u003c/span\u003e\u003c/span\u003e​ denotes the relative permeability, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{ϵ}_{r}\\)\u003c/span\u003e\u003c/span\u003e​ is the relative permittivity,\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\sigma\\:\\)\u003c/span\u003e\u003c/span\u003eis the electrical conductivity, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\omega\\:\\)\u003c/span\u003e\u003c/span\u003e is the angular frequency, and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{ϵ}_{0}\\)\u003c/span\u003e\u003c/span\u003e​ is the permittivity of free space. All material parameters were obtained from the COMSOL material library or reported data for SiO₂ and Ag.\u003c/p\u003e\u003cp\u003eThe electric field simulation was performed at a frequency of 9.7335×10¹⁴ Hz (corresponding to a wavelength of approximately 308 nm). A linearly polarized plane wave was applied along the z-direction with an initial field amplitude of 1 V/m. A Gaussian beam approximation was applied to model the incident field, addressing wavefront curvature and beam divergence. The applied electric field at lateral position 𝑥 and axial position 𝑦 was given by:\u003c/p\u003e\u003cdiv id=\"Equb\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equb\" name=\"EquationSource\"\u003e\n$$\\:{E}_{b}\\left(x,\\:y,z\\right)={E}_{b0}\\sqrt{\\frac{{\\omega\\:}_{0}}{\\omega\\:\\left(y\\right)}}exp\\left[\\frac{-{x}^{2}}{{\\omega\\:}^{2}\\left(y\\right)}-jky-jk\\frac{{x}^{2}}{2R\\left(y\\right)}+\\frac{j\\eta\\:\\left(y\\right)}{2}\\right]$$\u003c/div\u003e\u003c/div\u003e\u003cp\u003ewhere \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\omega\\:\\left(y\\right)\\:\\)\u003c/span\u003e\u003c/span\u003edefines the beam waist as a function of propagation distance,\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\:R\\left(y\\right)\\:\\)\u003c/span\u003e\u003c/span\u003eis the wavefront radius of curvature, and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\eta\\:\\left(y\\right)\\:\\)\u003c/span\u003e\u003c/span\u003eis the Gouy phase shift. The Gaussian beam was linearly polarized along the 𝑥-direction and normally incident along the z-axis, with an initial field amplitude of \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{E}_{0}\\)\u003c/span\u003e\u003c/span\u003e=1 V/m, a beam waist of \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\omega\\:}_{0}\\)\u003c/span\u003e\u003c/span\u003e=0.308 µm, and a focal position at \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{y}_{0}\\)\u003c/span\u003e\u003c/span\u003e=9.676 µm.\u003c/p\u003e\u003cp\u003eTo prevent unwanted reflections, multi-layered perfectly matched layers (PMLs) were applied along the x- and y-boundaries, ensuring efficient absorption of outgoing waves. The PML thickness was set to 2λ to minimize numerical artifacts while preserving computational efficiency. The bottom boundary was defined a Perfect Electric Conductor (PEC) condition to simulate a grounded electrode, while all other boundaries remained open with PMLs.\u003c/p\u003e\u003cp\u003eA physics-controlled mesh was employed to optimize element distribution based on electromagnetic field variations, ensuring accurate resolution of rapid field gradients. The finest mesh setting was used, with refinement concentrated near Ag–SiO₂ interfaces and nanoparticles to capture localized plasmonic effects and strong permittivity contrasts. The smallest element size was set to resolve subwavelength field variations, preserving numerical accuracy.\u003c/p\u003e\u003cp\u003eTo solve the frequency-domain Maxwell’s equations efficiently, a direct solver was utilized for small-scale problems, while an iterative solver (GMRES with preconditioning) was applied for large-scale simulations to optimize memory usage. A relative tolerance of 10\u003csup\u003e− 6\u003c/sup\u003e was applied, and convergence was monitored by tracking the residual norm, ensuring it remained below 10\u003csup\u003e− 6\u003c/sup\u003e. The physics-controlled adaptive mesh ensured high accuracy, with automatic refinement applied in regions exhibiting strong field gradients.\u003c/p\u003e\u003cp\u003eTo analyze the simulated electric field distribution, postprocessing involved extracting the field amplitude ∣𝐸∣, phase distributions, and energy-flux profiles. These parameters were used to examine localized plasmonic enhancement in laser-irradiated regions, which played a crucial role in determining field intensities necessary for Ag-ion migration and filament formation in the switching layer.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eData availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll data that support the findings of this study are reported in the Article and its supplementary information.\u003c/p\u003e\n\n\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis work was supported by Samsung Electronics Co., Ltd (No.IO201214-08153-01). It was also supported by the Convergent Technology R\u0026amp;D Program for Human Augmentation through the National Research Foundation of Korea (NRF) funded by the Ministry of Science and ICT (No. NRF- 2020M3C1B8081519). This work was additionally supported by a National Research Foundation of Korea (NRF) grant funded by the Korean Government (MSIP) (No. NRF- 2020M3F3A2A02082445). \u003c/p\u003e\n\n\u003cp\u003e\u003cstrong\u003eAuthor contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eY.J.J. and K.J.L. conceived the idea of the laser-guided dual-mode memristor, designed the experiments, and analyzed corresponding data. K.K., Y.B.K., H.S., J.W.O., and S.H.S. helped with experiments and data analysis. Y.J.J. and K.J.L. wrote the manuscript. K.J.L. supervised the research and contributed to the discussion of the overall methodology and results. All authors discussed the results and commented on the manuscript.\u003c/p\u003e\n\n\u003cp\u003e\u003cstrong\u003eCompeting interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare no conflict of interest. \u003c/p\u003e\n"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eHam, D., Park, H., Hwang, S. \u0026amp; Kim, K. Neuromorphic electronics based on copying and pasting the brain. \u003cem\u003eNat. Electron.\u003c/em\u003e 4, 635-644 (2021). \u003c/li\u003e\n\u003cli\u003eLiu, Z. et al. A memristor-based adaptive neuromorphic decoder for brain\u0026ndash;computer interfaces. \u003cem\u003eNat. Electron. \u003c/em\u003ePreprint at https://doi.org/10.1038/s41928-025-01340-2 (2025).\u003c/li\u003e\n\u003cli\u003eSung, S. H. et al. Memory-centric neuromorphic computing for unstructured data processing. \u003cem\u003eNano Research \u003c/em\u003e14, 3126-3142 (2021).\u003c/li\u003e\n\u003cli\u003ePrezioso, M. et al. Training and operation of an integrated neuromorphic network based on metal-oxide memristors. \u003cem\u003eNature\u003c/em\u003e 521, 61-64 (2015).\u003c/li\u003e\n\u003cli\u003eChen, S. et al. Wafer-scale integration of two-dimensional materials in high-density memristive crossbar arrays for artificial neural networks. \u003cem\u003eNat. Electron.\u003c/em\u003e 3, 638-645 (2020).\u003c/li\u003e\n\u003cli\u003eFu, T. et al. Bioinspired bio-voltage memristors. \u003cem\u003eNat Commun.\u003c/em\u003e 11, 1861 (2020). \u003c/li\u003e\n\u003cli\u003eJohn, R. A. et al. Reconfigurable halide perovskite nanocrystal memristors for neuromorphic computing. \u003cem\u003eNat. Commun.\u003c/em\u003e 13, 2074 (2022). \u003c/li\u003e\n\u003cli\u003eWang, T. et al. Reconfigurable neuromorphic memristor network for ultralow-power smart textile electronics. \u003cem\u003eNat. Commun.\u003c/em\u003e 13, 7432 (2022). \u003c/li\u003e\n\u003cli\u003eXia, Q. \u0026amp; Yang, J. J. Memristive crossbar arrays for brain-inspired computing. \u003cem\u003eNat. Mater.\u003c/em\u003e 18, 309-323 (2019). \u003c/li\u003e\n\u003cli\u003eWan, W. et al. A compute-in-memory chip based on resistive random-access memory. \u003cem\u003eNature \u003c/em\u003e608, 504-512 (2022). \u003c/li\u003e\n\u003cli\u003eKim, H., Mahmoodi, M. R., Nili, H. \u0026amp; Strukov, D. B. 4K-memristor analog-grade passive crossbar circuit. \u003cem\u003eNat. Commun.\u003c/em\u003e 12, 5198 (2021). \u003c/li\u003e\n\u003cli\u003eZhang, X. et al. An artificial spiking afferent nerve based on Mott memristors for neurorobotics. \u003cem\u003eNat. Commun.\u003c/em\u003e 11, 51 (2020). \u003c/li\u003e\n\u003cli\u003eSung, S. H., Kim, T. J., Shin, H., Im, T. H. \u0026amp; Lee, K. J. Simultaneous emulation of synaptic and intrinsic plasticity using a memristive synapse. \u003cem\u003eNat. Commun.\u003c/em\u003e 13, 2811 (2022). \u003c/li\u003e\n\u003cli\u003eSung, S. H. et al. Bio-plausible memristive neural components towards hardware implementation of brain-like intelligence. \u003cem\u003eMater. Today\u003c/em\u003e 62, 251-270 (2023). \u003c/li\u003e\n\u003cli\u003eWoo, K. S. et al. Tunable stochastic memristors for energy-efficient encryption and computing. \u003cem\u003eNat. Commun. \u003c/em\u003e15, 3245 (2024). \u003c/li\u003e\n\u003cli\u003eHan, J.-K. et al. Cointegration of single-transistor neurons and synapses by nanoscale CMOS fabrication for highly scalable neuromorphic hardware. \u003cem\u003eSci. Adv.\u003c/em\u003e 7, eabg8836 (2021). \u003c/li\u003e\n\u003cli\u003eYang, Q. et al. High-speed, scanned laser structuring of multi-layered eco/bioresorbable materials for advanced electronic systems. \u003cem\u003eNat. Commun.\u003c/em\u003e 13, 6518 (2022). \u003c/li\u003e\n\u003cli\u003eChen, L. et al. Direct laser patterning of organic semiconductors for high performance OFET-based gas sensors. \u003cem\u003eJ. Mater. Chem. C\u003c/em\u003e 11, 7088-7097 (2023). \u003c/li\u003e\n\u003cli\u003eRho, Y. et al. A laser-assisted chlorination process for reversible writing of doping patterns in graphene. \u003cem\u003eNat. Electron.\u003c/em\u003e 5, 505-510 (2022). \u003c/li\u003e\n\u003cli\u003eMeier, T. et al. Obtaining auxetic and isotropic metamaterials in counterintuitive design spaces: an automated optimization approach and experimental characterization. \u003cem\u003enpj Comput. Mater.\u003c/em\u003e 10, 3 (2024). \u003c/li\u003e\n\u003cli\u003eLee, G. et al. Implantable, bioresorbable radio frequency resonant circuits for magnetic resonance imaging. \u003cem\u003eAdv. Sci.\u003c/em\u003e 11, 2301232 (2024). \u003c/li\u003e\n\u003cli\u003ePark, H. et al. Laser‐Based Selective Material Processing for Next‐Generation Additive Manufacturing. \u003cem\u003eAdv. Mater.\u003c/em\u003e 36, e2307586 (2024).\u003c/li\u003e\n\u003cli\u003eNath, S. K. et al. Optically tunable electrical oscillations in oxide‐based memristors for neuromorphic computing. \u003cem\u003eAdv. Mater.\u003c/em\u003e 36, e2400904 (2024). \u003c/li\u003e\n\u003cli\u003eWang, Z. et al. Memristors with diffusive dynamics as synaptic emulators for neuromorphic computing. \u003cem\u003eNat. Mater.\u003c/em\u003e 16, 101-108 (2017). \u003c/li\u003e\n\u003cli\u003eMilozzi, A., Ricci, S. \u0026amp; Ielmini, D. Memristive tonotopic mapping with volatile resistive switching memory devices. \u003cem\u003eNat. Commun.\u003c/em\u003e 15, 2812 (2024). \u003c/li\u003e\n\u003cli\u003eKim, G. et al. Mott neurons with dual thermal dynamics for spatiotemporal computing. \u003cem\u003eNat. Mater.\u003c/em\u003e 23, 1237-1244 (2024). \u003c/li\u003e\n\u003cli\u003eKoirala, K. P., Sandireddy, V. P., Garcia, H., Duscher, G. \u0026amp; Kalyanaraman, R. Nanosecond switchable localized surface plasmons through resettable contact angle behavior in silver nanoparticles. \u003cem\u003eNanotechnol.\u003c/em\u003e 31, 355503 (2020). \u003c/li\u003e\n\u003cli\u003eSandireddy, V. P., Koirala, K. P. \u0026amp; Kalyanaraman, R. In situ contact angle tuning of silver nanostructures by laser heating in different fluid ambient. \u003cem\u003eLangmuir\u003c/em\u003e 35, 10744-10751 (2019). \u003c/li\u003e\n\u003cli\u003eYu, H., Peng, Y., Yang, Y. \u0026amp; Li, Z.-Y. Plasmon-enhanced light\u0026ndash;matter interactions and applications. \u003cem\u003enpj Comput. Mater.\u003c/em\u003e 5, 45 (2019). \u003c/li\u003e\n\u003cli\u003eWang, W. et al. Surface diffusion-limited lifetime of silver and copper nanofilaments in resistive switching devices. \u003cem\u003eNat. Commun.\u003c/em\u003e 10, 81 (2019). \u003c/li\u003e\n\u003cli\u003eValov, I., Waser, R., Jameson, J. R. \u0026amp; Kozicki, M. N. Electrochemical metallization memories\u0026mdash;fundamentals, applications, prospects. \u003cem\u003eNanotechnol.\u003c/em\u003e 22, 254003 (2011). \u003c/li\u003e\n\u003cli\u003eShen, Z. et al. Advances of RRAM devices: Resistive switching mechanisms, materials and bionic synaptic application. \u003cem\u003eNanomaterials \u003c/em\u003e10, 1437 (2020). \u003c/li\u003e\n\u003cli\u003eMoon, J. et al. Temporal data classification and forecasting using a memristor-based reservoir computing system. \u003cem\u003eNat. Electron.\u003c/em\u003e 2, 480-487 (2019). \u003c/li\u003e\n\u003cli\u003eChoi, S. et al. 3D-integrated multilayered physical reservoir array for learning and forecasting time-series information. \u003cem\u003eNat. Commun.\u003c/em\u003e 15, 2044 (2024). \u003c/li\u003e\n\u003cli\u003eLiang, X. et al. Physical reservoir computing with emerging electronics. \u003cem\u003eNat. Electron. \u003c/em\u003e7, 193-206 (2024). \u003c/li\u003e\n\u003cli\u003eLi, W. et al. Dual‐Functional Nonvolatile and Volatile Memory in Resistively Switching Indium Tin Oxide/HfOx Devices. \u003cem\u003ephys. status solidi (a)\u003c/em\u003e 216, 1900555 (2019). \u003c/li\u003e\n\u003cli\u003eYi, W. et al. Biological plausibility and stochasticity in scalable VO2 active memristor neurons. \u003cem\u003eNat. Commun.\u003c/em\u003e 9, 4661 (2018). \u003c/li\u003e\n\u003cli\u003eLi, X. et al. Design of artificial neurons of memristive neuromorphic networks based on biological neural dynamics and structures. \u003cem\u003eIEEE Trans.Circuits Syst.\u003c/em\u003e I 71, 2320-2333 (2024).\u003c/li\u003e\n\u003cli\u003eXiao, Y. et al. Bio-plausible reconfigurable spiking neuron for neuromorphic computing. \u003cem\u003eSci. Adv.\u003c/em\u003e 11, eadr6733 (2025). \u003c/li\u003e\n\u003cli\u003eKim, G. et al. Threshold modulative artificial GABAergic nociceptor. \u003cem\u003eAdv. Mater.\u003c/em\u003e 35, 2304148 (2023). \u003c/li\u003e\n\u003cli\u003eYang, J.-Q. et al. Leaky integrate-and-fire neurons based on perovskite memristor for spiking neural networks. \u003cem\u003eNano Energy\u003c/em\u003e 74, 104828 (2020). \u003c/li\u003e\n\u003cli\u003eKim, S. J., Kim, S. \u0026amp; Jang, H. W. Competing memristors for brain-inspired computing. \u003cem\u003eIscience\u003c/em\u003e 24 101889 (2021).\u003c/li\u003e\n\u003cli\u003eChen, P.-Y., Peng, X. \u0026amp; Yu, S. NeuroSim+: An Integrated Device\u0026not;to\u0026not;Algorithm Framework for Benchmarking Synaptic Devices and Array Architectures. In 2017 IEEE International Electron Devices Meeting (IEDM) 6.1.1-6.1.4 (IEEE, 2017)\u003c/li\u003e\n\u003cli\u003eSeo, S. et al. Artificial van der Waals hybrid synapse and its application to acoustic pattern recognition. \u003cem\u003eNat. Commun.\u003c/em\u003e 11, 3936 (2020). \u003c/li\u003e\n\u003cli\u003eWu, C., Kim, T. W., Choi, H. Y., Strukov, D. B. \u0026amp; Yang, J. J. Flexible three-dimensional artificial synapse networks with correlated learning and trainable memory capability. \u003cem\u003eNat. Commun.\u003c/em\u003e 8, 752 (2017). \u003c/li\u003e\n\u003cli\u003ePrezioso, M. et al. Spike-timing-dependent plasticity learning of coincidence detection with passively integrated memristive circuits. \u003cem\u003eNat. Commun.\u003c/em\u003e 9, 5311 (2018). \u003c/li\u003e\n\u003cli\u003eMilano, G. et al. In materia reservoir computing with a fully memristive architecture based on self-organizing nanowire networks. \u003cem\u003eNat. Mater.\u003c/em\u003e 21, 195-202 (2022). \u003c/li\u003e\n\u003cli\u003eLillicrap, T. P., Cownden, D., Tweed, D. B. \u0026amp; Akerman, C. J. Random synaptic feedback weights support error backpropagation for deep learning. \u003cem\u003eNat. Commun.\u003c/em\u003e 7, 13276 (2016). \u003c/li\u003e\n\u003cli\u003eZhang, T. et al. Self-backpropagation of synaptic modifications elevates the efficiency of spiking and artificial neural networks. \u003cem\u003eSci. Adv.\u003c/em\u003e 7, eabh0146 (2021).\u003c/li\u003e\n\u003cli\u003ePattnaik, D. et al. Stress-induced artificial neuron spiking in diffusive memristors. \u003cem\u003eCommun. Eng.\u003c/em\u003e 3, 163 (2024).\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":false,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Laser-guided ion dynamics, Single coplanar memristor, Bioinspired neuronal and synaptic integration, Volatile and nonvolatile switching, Neuromorphic reservoir computing","lastPublishedDoi":"10.21203/rs.3.rs-6140700/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6140700/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eNeuromorphic computing aims to replicate the parallel, adaptive nature of biological intelligence in electronic systems. Despite considerable advances in memristor technology, material-encoded neurosynaptic bifunctionality has not been demonstrated. We introduce a laser-guided dual-mode memristor that integrates both volatility for neuronal spiking and nonvolatility for synaptic plasticity within a single-phase material. By precisely modulating silver ion dynamics through XeCl excimer laser irradiation, we achieve local and dynamic control of the dual-mode memristive behavior without requiring a heterogeneous device array or stacking. The neurosynaptic tunability with optimal computational efficiency demonstrates reconfigurable reservoir computing and a positive feedback loop for adaptive learning.\u003c/p\u003e","manuscriptTitle":"Laser-Guided Ion Dynamics in a Dual-Mode Memristor for Bioinspired Neuronal and Synaptic Integration","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-04-09 06:10:44","doi":"10.21203/rs.3.rs-6140700/v1","editorialEvents":[],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"nature-communications","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"NCOMMS","sideBox":"Learn more about [Nature Communications](http://www.nature.com/ncomms/)","snPcode":"","submissionUrl":"https://mts-ncomms.nature.com/","title":"Nature Communications","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"ejp","reportingPortfolio":"Nature Communications","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"765ac5cc-2e73-4641-be5a-7e7cbd6da87a","owner":[],"postedDate":"April 9th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":46155023,"name":"Physical sciences/Materials science/Materials for devices/Electronic devices"},{"id":46155024,"name":"Physical sciences/Materials science/Materials for optics/Lasers, LEDs and light sources/Semiconductor lasers"}],"tags":[],"updatedAt":"2025-09-26T06:11:45+00:00","versionOfRecord":[],"versionCreatedAt":"2025-04-09 06:10:44","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-6140700","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6140700","identity":"rs-6140700","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}