Ionic Driven Waveguide Integrated Memristor

Ionic Driven Waveguide Integrated Memristor | 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 Ionic Driven Waveguide Integrated Memristor Hongtao Lin, Kunhao Lei, Zijia Wang, Kai Xu, Shuo Lin, Kangjian Bao, and 10 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7782397/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted You are reading this latest preprint version Abstract Photonic memristors are anticipated to emerge as a novel hardware platform for neural network computing systems owing to their broad bandwidth communication capabilities and potential compatibility with non-von Neumann neuromorphic computing architectures. However, current photonic memristors are limited to a monotonic modulation mechanism and most demonstrate only a non-volatile temporal response scale, which restricts their applicability across more diverse neural network computing architectures. Here, we present a novel strategy that employs ion-doped chalcogenide glass, combined with a multi-dimensional modulation mechanism of optical and electrical fields, to realize the monolithic integration of non-volatile waveguide-integrated memristors featuring multi-level storage capacity and compact footprint, along with volatile reconfigurable memristors exhibiting high extinction ratios and excellent short-term plasticity. Furthermore, leveraging the powerful nonlinear dynamic response of the reconfigurable memristor, we developed an on-chip photonic reservoir computing system that operates without a feedback loop. This work provides a novel developmental approach for the development of neuron devices with varying time response scales and offers substantial support for neural networks in more accurately simulating brain functions. Physical sciences/Optics and photonics/Applied optics/Optoelectronic devices and components Physical sciences/Optics and photonics/Optical materials and structures Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Introduction The rapid advancement of artificial intelligence (AI) has recently resulted in the growing complexity of artificial neural network (ANN) architectures, accompanied by increased training costs and a continuous rise in the demand for enhanced hardware data processing capabilities. However, the von Neumann computing paradigm is increasingly struggling to meet the demands for efficient and low-latency data processing due to the significant time and energy overhead associated with frequent data transfers between computing and storage units. Recent advancements in electronic neuromorphic memristors have underscored the benefits of the neuro-bionic paradigm and have been utilized in efficient ANN architectures, such as the lightweight reservoir computing (RC) architecture based on time-division multiplexing, not only showcasing their exceptional capacity for processing temporal data, but also demonstrating their effectiveness in substantially reducing training costs and minimizing the overall system footprint 1 – 3 . Owing to the substantial advantages of optical interconnection in communication networks—such as large bandwidth, high parallelism, and low power consumption, and the synergistic potential with the non-von Neumann neuromorphic computing paradigm, photonic artificial memristor devices offer a novel hardware platform for ANNs 4 – 6 . Currently, a diverse range of emerging memristors based on photonic integrated circuits (PICs) has been widely applied for optical memories, post-fabrication trimming of PICs, and in-memory computing, such as Ⅲ-Ⅴ/Si hybrid integration memristor 7 , 8 , ferroelectric materials 9 , 10 , chalcogenide phase change materials (PCMs) 11 – 13 , and microelectromechanical systems (MEMS) 14 , 15 . Most of the aforementioned memristors demonstrate non-volatile behavior with a monotonous temporal response scale, rendering them suitable for storing trained weights or executing linear optical signal processing in the output layer of artificial neural networks. Meanwhile, owing to distinct functional requirements, the reservoir neurons in time-division multiplexed photonic RC systems without feedback loops necessitate memristors that possess dynamic and nonlinear response characteristics 16 —features that are challenging to achieve with purely non-volatile devices. Chalcogenide glass (ChGs), an amorphous soft glass material composed of elements such as sulfur (S), selenium (Se), and/or tellurium (Te), has demonstrated outstanding performance in chalcogenide solid state batteries, memristors, and the development of associated large-scale device arrays, owing to its superior ionic conductivity and excellent processability 17 – 21 . Modification through the ion doping mechanism and effect has currently emerged as a mainstream and significant strategy for chalcogenide solid-state batteries and metal-insulator-metal (MIM) memristors. Firstly, doping elements can enhance the ionic conductivity of the electrolyte as they not only generate point defects but also widen the ionic channels. Secondly, doping can also regulate the composition of the electrolyte, decrease interface resistance, and enhance the chemical stability of the solid electrolyte 22 , 23 . The MIM structure, composed of metal-ChGs-metal, not only demonstrates electrical responses but also realizes neuromorphic functionalities for optoelectronic multi-signal interactions via the photovoltaic effect induced by the asymmetric Schottky barrier formed between the top and bottom electrodes and the chalcogenide medium 24 . Ion-doped ChGs are capable of not only modifying their electrical conductivity but also tuning their band gap, thereby enabling control over the variation of optical constants within the ChGs material. This ionic behavior offers a promising pathway for the development of reconfigurable photonic devices 25 – 27 . In this work, we employed the ion-doping mechanism and effect of chalcogenide solid-state batteries to propose an ion-driven waveguide-integrated memristor platform for the first time, and demonstrate the application potential of the RC system based on these neuronal devices in the domain of underwater acoustic signal recognition, as illustrated in Fig. 1 . The proposed process flow eliminates the need for high-cost semiconductor ion implantation equipment and associated procedures. By modifying the Ge 28 Sb 12 Se 60 (GSSe) waveguide through electrochemically active metal doping and integrating the mechanism of optical- and electric field-induced ion migration, the approach enables multi-dimensional modulation of both the optical phase and attenuation within the GSSe waveguide. The stable non-volatile tuning memristor can be achieved by modifying the band gap of the chalcogenide glass through photo-induced sputtered metal Ag dissolution, while the dynamic attenuation memristor was achieved by controlling the formation of silver nanocluster in the Ag + solution-doped GSSe waveguide using an electric field, thereby enhancing the absorption and scattering losses within the waveguide. Furthermore, the intrinsic dynamic properties of the GSSe memristors, coupled with the variation characteristics of nonlinear analog transmission signals, render them highly suitable for implementing photonics RC systems. Within this architecture, the dynamic memristor can be mathematically modeled as a reservoir that nonlinearly maps the input encoded time-domain signal into a high-dimensional feature space. The ion-driven waveguide-integrated memristor platform, based on the multi-dimensional modulation mechanism, achieves monolithic integration of non-volatile and volatile dynamic memristors. It enables the development of diverse neuron devices with varying time response scales for the optical neural networks, offering a novel pathway for expanding the emerging memristor technologies and providing significant support for enhancing brain-like functionalities in neural network systems. Results Integrated Memristor-based RC system for Acoustic Target Recognition Underwater acoustic target recognition, as a pivotal technology, holds extensive application potential in areas such as marine resource exploration, environmental monitoring, and vessel identification. Currently, propelled by the ongoing advancements in various marine observation technologies, including satellite remote sensing, ocean buoys, and research vessels, the volume of marine data is increasing exponentially. This exponential growth poses significant challenges to existing maritime communication satellites, which struggle to effectively manage the surge in observational data and meet the demand for real-time monitoring. Consequently, the development of temporal data processing and identification technologies for facilitating communication among observation vessels, satellites, and shore stations has emerged as a critical focus. Here, we propose an ion-driven waveguide-integrated memristor-based RC system for processing temporal signals. In contrast to the traditional photonic RC architecture, which necessitates the establishment of a delay feedback loop to enhance the complexity of the reservoir node, this system leverages the time-domain analog signal response of the photonic memristor as the reservoir node. It directly reads the transmission changes of the optical signal following the application of a write pulse, eliminating the need for additional read pulses during data retrieval. This approach not only conserves spatial footprint but also simplifies the overall system architecture. The concept of the ion-driven waveguide-integrated memristor RC system for underwater acoustic signal recognition is illustrated in Fig. 1. When the sensors on the observation vessel detect the target signal, the encoder compresses and encodes the signal before transmitting it to the maritime satellite, thereby accommodating the satellite's limited computing resources and transmission bandwidth. Further, the satellite transmits the compressed data to the shore station, where the acoustic signal is decoded and recognized by the decoder utilizing the photonic RC system. For the photonic RC system, we propose an architecture that integrates on-chip wavelength division multiplexing (WDM) and time division multiplexing techniques with parallel memristor arrays to enhance the parallelism and signal processing capabilities. The architecture comprises a WDM and demultiplexing layer, a reservoir layer leveraging the dynamic response of memristors, and a fully connected output layer that can potentially exploit the non-volatile characteristics of memristors in future implementations, as illustrated in Fig. 1b. Fig. 1c presents the two schemes that can be monolithically integrated for doping and multi-physical field modulation of GSSe waveguides. The optical field modulation approach enables bandgap and refractive index tuning through a photochemical reaction induced by ultraviolet light, which drives the interaction between sputtered Ag and GSSe waveguide. The electric field modulation strategy facilitates ion doping by immersing the waveguides in an Ag + rich solution, followed by the aggregation of metal nanoclusters within the waveguide under an applied transverse electric field. By employing precise micro-nano fabrication techniques, we have successfully achieved monolithic integration of two distinct types of waveguide-integrated memristors featuring different doping profiles and modulation mechanisms, as shown in Fig. 1d. The photonic chip integrated hundreds of memristors has an overall footprint smaller than that of a one-yuan CNY coin. Among them, the dynamic memristor utilized for the reservoir layer forms a capacitive structure by placing electrodes on either side of the Ag + doped GSSe waveguide. Under the influence of an electric field, it governs the migration of silver nanoclusters within the waveguide. When these nanoclusters coalesce into larger Ag particles, they induce additional absorption and scattering losses, as depicted in Fig. 1f. Upon the low electrical voltage, owing to thermal effects and the driving force of interfacial energy, the nanoclusters undergo deformation and decay, causing the propagation loss of the waveguide to revert to its initial state 28 . The dynamic response of the electro-reconfigurable memristor is illustrated in Fig. 1g. The optical power decay curve can be further fitted to an exponential function. Fabrication and Principle Fig. 2a illustrates the fabrication processes of two types of memristors, which are described in detail in the methods section. To investigate the impact of Ag-doped chalcogenide glass on its molecular network, the doped chalcogenide thin films were characterized by Raman spectroscopy. As illustrated in Fig. 2d, the thermal evaporated GSSe chalcogenide film primarily exhibits three fundamental Raman scattering peaks. The most significant peak, located at 200 cm -1 , corresponds to the symmetric stretching mode associated with the corner-sharing [GeSe 4/2 ] tetrahedral units. The characterization results indicate that the scattering peak intensity at 170 cm⁻¹ and 100 cm⁻¹ increase with increasing silver doping concentration, whereas the intensity of the peak at 270 cm⁻¹ decreases to some extent, which can be attributed to the vibration of Ge–Ge bond, the umbrella vibration of [SbSe 3/2 ] pyramids unit and the symmetric stretching vibrations of Se–Se bond, respectively. Based on this phenomenon, we speculate that Ag doped and modified chalcogenide films create Se-deficient conditions (forming Ag 2 Se through bonding with Se), thereby promoting the rearrangement of Ge-Ge and Sb-Se bonds. This finding is consistent with similar experimental results reported in the literature 29-31 . Fig. 2b depicts the molecular structure of GSSe before and after doping with metallic Ag. Based on Raman spectroscopy characterization, it is found that the doping of GSSe films with Ag ions primarily occurs through the breaking of Se-Se homopolar bonds. Meanwhile, in Ge-rich chalcogenide films, the coordination bonds between two neighboring chains are relatively weak and are more likely to be broken, ionized, or form bonds with Ag ions, thus offering a natural transport channel for the migration of Ag ions. To verify the reconfigurable mechanism of the dynamic memristors, we carried out an in-situ transmission electron microscopy (TEM) experiment on the area of the device doped with an Ag ion solution, and the sample was prepared using focused ion beam (FIB) technology. As shown in the initial state of the device and the energy dispersive spectroscopy (EDS) analysis in Fig. 4e, after the memristor has been switched several times, a small amount of Ag species on both sides of the waveguide has already been unable to dissolve back into the GSSe waveguide spontaneously, and the distribution of silver ions within the waveguide was non-uniform (the initial uniformly doped distribution state prior to the switching operation was presented in supplementary information SI.2). This indicates that the device has approached a fatigue failure state. Subsequently, a voltage was applied from right to left at a step size of 5 mV per second. When the voltage reached 0.29 V, as depicted in Fig. 4f-i, it was observed that the Ag species on the right side of the waveguide dissolved in comparison to the initial state. Under the influence of the electric field, the Ag ions within the waveguide migrated to the lower left side, where they nucleated and aggregated. Furthermore, we restored the voltage to 0 and applied the voltage from left to right in the same step. When the voltage reached 0.21 V, we observed a similar phenomenon as shown in Fig. 4f-ii. Under the influence of the electric field, the Ag species on the left re-oxidized and dissolved back into the waveguide, and the Ag ions inside the waveguide migrated to the lower right of the waveguide and nucleated, realizing a reversible process regulated by the electric field. After observing the nucleation phenomenon twice, we immediately restored the voltage to zero and carried out EDS and high-resolution diffraction characterization on the species at the lower left corner of Fig. 4f-iii and the lower right corner of Fig. 4f-iv. Through the EDS and diffraction analysis presented in Fig .4f-iii, iv, v, it has been found that Ag ions migrate to the surface of the GSSe waveguide under the influence of the electric field and undergo a reduction reaction to form elemental Ag. Consequently, this leads to an increase in the absorption and scattering losses of the waveguide and the formation of an attenuation modulation. Meanwhile, according to the TEM and EDS diagram in Fig. 4f-iv, it is evident that a portion of the metallic Ag has spontaneously dissolved, which also suggests that this mechanism is a dynamic and volatile process. The high-resolution TEM image presented in Fig. 4f-v shows the regular lattice of the nucleated Ag species. The lattice spacings were measured to be 0.236 nm, corresponding to the (111) crystal planes of the cubic crystal system of elemental Ag. The mechanism of elemental Ag formation is speculated to be that Ag ions migrate to the surface of the waveguide under the influence of an electric field, where they undergo a reduction reaction with the defect states possessing lone pairs of electrons in the doped waveguide and the tunneling electrons under the influence of the electric field 32-34 . Photo-doping Response of the Non-volatile GSSe/Ag Memristors For the process of the non-volatile photo-induced diffusion of metallic Ag in chalcogenide glass, when light with energy exceeding the optical bandgap of the chalcogenide compound is incident on the waveguide, photogenerated holes and electrons are created. Electrons tend to be captured over short distances, forming negative charge centers, while holes migrate over longer distances, accumulate at the interface between metallic silver and the chalcogenide waveguide, and induce ionization. This process results in the formation of silver ions and free electrons, where the silver ions further diffuse and react with sulfur-based nanoclusters 35 . As the silver concentration increases, the absorption edge of the chalcogenide glass undergoes a redshift, accompanied by a simultaneous increase in the refractive index 27 . To further investigate the photoinduced dissolution process of Ag-doped GSSe waveguide, we conducted quantitative optical measurements and analysis of the photodoped micro-ring resonator (MRR) devices. Fig. 3a and b present the optical microscope image of the device and the scanning electron microscope (SEM) image of the doped region with the sputtered Ag cell. The transmission spectrum depicted in Fig. 3c illustrates an MRR with a doped length of 10 μm. The results indicate that the deposition of Ag on the waveguide leads to a significant decrease in the extinction ration and Q factor of the microring, suggesting a substantial increase in absorption loss due to the metal deposited. However, following the photoinduced dissolution of the Ag by using UV light, the extinction ration and Q factor nearly returns to its initial state. Simultaneously, we characterize the variations in the wavelength shift and quality factor (Q) of the MRRs as they transition from Ag deposition to photodoping under various doped lengths. As illustrated in Fig. 3d, following the photoinduced doping process, the redshift of the MRR resonance peak can be enhanced from 0.63 nm to 1.80 nm by varying the doped length within the range of 5 μm to 50 μm, which enables the realization of a compact phase shifter with an L 𝜋 of approximately 26 μm (L 𝜋 =𝜆/2Δneff, free spectral range: 2.3 ~ 2.4 nm). Next, by comparing the quality factor Q of the MRR, we analyze the changes in optical loss of the MRR during the process of photo doing. In Fig. 3e, the loaded Q factor of the single-mode waveguide GSSe MRR is about 2.25~3.39×10 4 . After Ag cells were deposited on the GSSe waveguide via sputtering, the loaded Q factor of the MRR decreased to a range of 0.34×10 3 to 0.72×10 3 . This reduction was attributed to increased metal absorption, which exhibited a rising trend as the length of the sputtered Ag cell increased. The loaded Q factor of the MRR is improved and nearly restored to its initial level (1.82~2.59×10 4 ) after implementing the photodoping GSSe waveguide process by using UV light-induced silver dissolution. Furthermore, the kinetic process of photoinduced Ag dissolution doping in GSSe MRRs was examined. As shown in Fig. 3f, we exposed the hybrid MRR to UV light for durations ranging from 15 to 150 minutes and recorded the transmission response. It can be observed from the results that, upon the initiation of UV illumination, the growth rate of redshift for the resonant wavelength decreases exponentially. This phenomenon can be attributed to the initial stage of photoinduced dissolution of chalcogenide films doped with Ag, where the doping rate is determined by the photochemical reaction occurring at the interface between the metal Ag and the chalcogenide film. When the thickness of the doped film reaches a critical value, the rate of the doping process shifts from being dominated by the photochemical reaction at the interface to being determined by the diffusion of silver ions and other charged particles within the doped film. Consequently, the process rate becomes diffusion-limited 36,37 . Further, the multilevel characteristics of the hybrid integrated GSSe/Ag waveguide attenuator were experimentally verified, the structural schematic diagram is presented in Fig. 3g. As illustrated in Fig. 3h, the insertion loss of the waveguide attenuator with 100 μm metal Ag deposited at a wavelength of 1630 nm was measured to be -23.8 dB. By progressively increasing the UV light dose to induce the dissolution of the metal Ag layer, the insertion loss of the waveguide attenuator can be gradually reduced, simultaneously enabling the realization of 32-level (~5 bit) intensity modulation. This characteristic positions it as a promising candidate for weight storage in the fully connected output layer of RC systems. To verify the non-volatile properties of the device, the retention time of both the deposited metal state (Ag deposited) and the photoinduced metal doping state (photodoped) was tested under the same wavelength. Neither state exhibited significant degradation over a period of 0 to 10 3 seconds, as shown in Fig. 3i. Chalcogenide Dynamic Waveguide Integrated Memristors Based on the Ag + solution doping process, we propose a novel electrically reconfigurable chalcogenide dynamic waveguide integrated memristor to simulate the reservoir layer of the photonics RC systems. Based on the aforementioned modulation mechanism, an MRR device featuring a doping length of 90 μm and achieving an intensity tuning exceeding 3 dB was developed, as illustrated in Fig. 4b. The set/reset switching cycle was found to exceed 20 cycles, and the corresponding transmission spectrum is presented in Fig. 4c. Based on the TEM characterization presented in the previous section, the observed switching failure is hypothesized to result from the presence of residual Ag nanocluster after the reset process 38 , which leads to increased insertion loss of the device. Future improvements could focus on optimizing the doping process and switching parameters to enhance device performance and reliability. After this, a switch characterization measurement was conducted on a waveguide attenuator device to further investigate the modulation mechanism. As shown in Fig. 4e, a reconfigurable attenuator device featuring a switch ratio of approximately 25 dB can be achieved with a doping length of 100 μm, while maintaining an insertion loss of -3 dB. Under the influence of the electric field, the doped waveguide demonstrates a relatively high absorption loss, which further corroborates our hypothesis that the electric field induces the formation of Ag nanoclusters within the waveguide, thereby causing substantial absorption and scattering losses. As illustrated in Fig. 4f, by progressively applying voltages ranging from 50 to 130 volts, a quasi-continuous 28-level (˃ 4 bit) multilevel intensity tuning was achieved. To simulate brain-like neuron functions, we systematically characterized the response of the waveguide attenuation dynamic memristor under different pulsed voltage conditions. Similar to electronic dynamic memristors that regulate the excitation or inhibition of postsynaptic currents via presynaptic electrical pulse mechanisms, ion-driven waveguide-integrated memristors can also modulate the inhibition of postsynaptic light intensity responses at specific wavelengths by leveraging the light absorption mechanism induced through electrically driven Ag nanoclusters. Among these, the key parameters of the pulse voltage primarily include the pulse amplitude, pulse duration, and pulse delay (i.e., the time interval between consecutive pulses). Firstly, we maintained the pulse duration constant (3.39 seconds) while systematically varying the pulse amplitude from 110 to 130V. The test results are presented in Fig. 4g. It can be observed that as the pulse amplitude increases, the attenuation absorption of the waveguide attenuator at a fixed wavelength gradually rises, with the modulation depths exceeding 25 dB. This indicates that under the influence of a stronger electric field, a larger number of Ag ions are involved in diffusion motion, leading to the formation of nanoclusters. Consequently, these not only enhance the absorption and scattering losses of the attenuator but also result in a significant increase in the relaxation time of the device. Meanwhile, by fixing a pulse amplitude at 110V and increasing the pulse duration (1.70 to 7.35s), we observed the same phenomenon, as illustrated in Fig. 4h. However, the results indicate that the scheme with variable pulse duration has a more restricted influence on the diffusion of Ag nanoclusters compared to the variable pulse amplitude. This is attributed to the fact that an increase in pulse duration primarily affects the coarsening process following the nucleation of the Ag nanoclusters 39 , thereby limiting the change in absorption of the intrinsic optical mode. As illustrated in Fig. 4i, the dynamic attenuator's transmission can be modulated in a wide range by adjusting the pulse amplitude and duration to simulate various synaptic activities. Specifically, paired-pulse facilitation (PPF) and paired-pulse depression (PPD) are two important mechanisms of short-term plasticity, which describe the relative facilitation or depression of postsynaptic currents evoked by the second pulse compared to the first pulse after an inter-pulse interval 40,41 . Based on the characteristic that the optical absorption of the waveguide attenuator increases under electrical pulse stimulation, we measured the PPD index of the device by keeping the pulse amplitude and duration constant while increasing the pulse interval time, and the results are presented in SI.5. Memristor-based RC system for Underwater Acoustic Recognition The RC system is a computational framework that transforms input temporal signals into high-dimensional representations through the reservoir layer. The switching behavior exhibited by the dynamical waveguide integrated memristor we developed closely resembles that of a dynamic reservoir: the future optical transmission state is determined by both the applied input voltage signal and the current transmission condition of the device. To showcase the neuromorphic computing capabilities of the doped GSSe memristor, we propose a reservoir computing architecture leveraging WDM technology and apply it to an underwater acoustic target recognition task. We utilized the DeepShip dataset 42 , comprising four primary categories of vessels: cargo ships, tanker ships, passenger ships, and tug ships, to evaluate the temporal data processing capability of the RC system (as shown in Figs. 5a and b). Underwater acoustic target recognition serves as a critical technology for marine monitoring and has traditionally relied on deep learning methods for target identification. However, in practical deployment scenarios, these approaches face a significant data transmission bottleneck. Specifically, after underwater acoustic sensors collect raw signals and extract features using the constant-Q transform (CQT) (as illustrated in Fig. 5a), these feature data must be transmitted to shore stations via satellite communication. Considering that modern advanced satellite communication systems are capable of operating with data transmission rates of 50 Mbps 43,44 , the transmission of 100 feature matrices, each of dimension 96×188, requires approximately 1.15 seconds. When thousands of sensors within the coverage area of a single satellite are transmitting data concurrently, the resulting queuing delay can accumulate to hundreds of seconds. This latency significantly limits the system's ability to perform real-time target recognition. To address this issue, we propose a hybrid architecture that integrates data binarization compression with the photonic RC decoding system. First, the time-frequency features to be transmitted were subjected to binarization, which reduces the data volume from 72 Kb to 2.256 Kb and shortens the transmission time to 36 milliseconds, thereby improving satellite data transmission efficiency by 3200%. Subsequently, the binarized data received on the ground was reconstructed using an ion-driven photonic memristor-based RC framework incorporating WDM technology. The core advantage of the memristor-based RC system lies in its capacity to emulate the time-dependent characteristics of neurons by utilizing the exponential decay response of the GSSe dynamic memristors (as illustrated in the inset of Fig. 5b), thereby effectively recovering the temporal information that may be lost during the binarization process, and subsequently feeding it into convolutional neural network (CNN) for classification and recognition (as illustrated in Fig. 5b). The performance comparison analysis present in Fig. 5d indicates that the hybrid system achieves a recognition accuracy of 73.0% on the DeepShip dataset, which is less than 5% lower than the original data accuracy of 77.7% classified directly by the CNN model, while significantly increasing the transmission efficiency by a factor of 32 within the same time unit. More importantly, compared with the accuracy of 57.4% achieved by directly processing binary data, the hybrid RC module contributes to a notable improvement of 15.6%. Furthermore, as evidenced by the confusion matrices presented in Fig. 5e, the hybrid system exhibits a strong capacity for classification balance when applied to imbalanced datasets, thereby underscoring the robustness of the proposed approach. Discussion In conclusion, we successfully demonstrated the functionality of dynamic and non-volatile memristors on the ChGs integrated photonic platform by the ion-doping mechanism and effect for the first time. Meanwhile, we have disclosed the underlying ion migration reconfigurable working mechanism of the dynamic memristor via systematic Raman spectroscopy analysis and in-situ TEM characterization. This finding opens up a new developmental avenue for the development of diverse neuronal devices capable of operating across various temporal response scales within photonic neural networks. Non-volatile memristor is achieved via photoinduced Ag dissolution doping process, which modified the molecular chain structure and bandgap of chalcogenide glass. This approach enabled a compact MRR phase shifter with an L 𝜋 of 26 µm, as well as a multi-level tunable attenuator featuring 5-bit. By applying voltage to precisely control the nucleation process of silver nanoclusters in the GSSe waveguide, a reconfigurable dynamic memristor with a switching ratio of approximately 25 dB was realized. Meanwhile, its short-term memory and nonlinear characteristics were validated through dynamic response measurement, demonstrating its capability to generate well-separable responses to distinct temporal inputs. Furthermore, we proposed a photonic RC architecture based on a ion-doped GSSe waveguide-integrated memristor platform and applied it to underwater acoustic recognition tasks with temporal characteristics. Through data binarization, the data transmission rate per unit time was significantly enhanced. Subsequently, the compressed data was decoded and reconstructed using the GSSe dynamic memristor, ultimately achieving an accuracy of 73.0%, with a reduction in recognition accuracy of less than 5% compared to the original data, which provides substantial support for the application of waveguide-integrated memristors in broader and lightweight ANN architectures, thereby enhancing the brain-like functionalities of neural network systems. Methods Device fabrication The 500 nm Ge 28 Sb 12 Se 60 film was deposited on a Si substrate with a 2 μm SiO 2 layer via thermal evaporation under a background vacuum of 5 × 10 -4 Pa. The wafer surface is maintained in pristine condition through oxygen plasma cleaning before deposition, and the deposition rate is controlled at 15 Å/s. The hybrid integrated GSSe/Ag MRRs were patterned using 50 keV electron beam lithography (EBL) equipment (Raith Voyager). The photoresist used was the AllResist AR-P 6200.13 series, spun at 4000 rpm to achieve a thickness of approximately 400 nm. Subsequently, the pattern is transferred to the GSSe layer using inductively coupled plasma (ICP) dry etching with an Ar/CHF 3 gas chemistry, and the residual photoresist on top of the waveguide is removed via oxygen plasma etching. For the fabrication of hybrid integrated GSSe/Ag waveguide attenuator and dynamic reconfigurable devices, we utilized the stepper lithography machine (Canon 3030EX6) equipped with a 248 nm deep ultraviolet (DUV) light source to achieve wafer-level patterning of the waveguide, and the subsequent dry etching process followed the same procedure as described previously. After this, the photodoped regions were patterned via EBL, and a 30 nm layer of Ag was deposited on the GSSe waveguide via magnetron sputtering, followed by a lift-off process. The lift-off process utilized AR-P 6200.13 photoresist, spun-coated at 2000 rpm to achieve a thickness of approximately 600-700 nm. For the fabrication of the electrically reconfigurable devices, the negative photoresist NR9-1500PY and the Aligner photolithography equipment (Karl Suss MA6-BSA) were employed to define the electrode pad patterns. Subsequently, a metal layer of Ti (5 nm)/Au (100 nm) was deposited on the chip via magnetron sputtering. Last, the metal outside the defined pattern was removed through a lift-off process. The doping window of the electrically reconfigurable device was also determined by NR9-1500PY. Through exposure and development processes, the waveguide regions that require doping are exposed to air, while the regions that do not require doping are protected by a layer of photoresist. Doing process In this study, the photoinduced Ag dissolution doping process utilized a 45W UV lamp with a wavelength of 365 nm as the light source. The immersion doping process utilized a standardized 0.1 mol/L AgNO 3 solution, achieving thermal diffusion doping through water bath treatment at 40 °C. The immersion doping process has lower lithographic resolution requirements and higher reliability compared to depositing metal Ag on the waveguide surface followed by a lift-off process, making it well-suited for large-scale on-chip waveguide doping modification. Declarations Data Availability All the data supporting this study are available in the paper and Supplementary Information. Additional data related to this paper are available from the corresponding authors upon request. Acknowledgments This work was supported by the National Key Research and Development Program of China (2024YFB2808700), "Pioneer" R&D Program of Zhejiang (2025C01002), and the Key Project of Westlake Institute for Optoelectronics (2024GD002). The authors would like to acknowledge the fabrication support from the ZJU Nano-Fabrication Center at Zhejiang University, the Westlake Center for Micro/Nano Fabrication and Instrumentation at Westlake University, and the Nano-Fabrication Center at Zhejiang Lab. The authors would like to acknowledge Dr. Yangjian Lin and Dr. Pei Sheng from the Instrumentation and Service Center for Physical Sciences at Westlake University for their support and assistance in the in-situ TEM experiments. Author contributions H.L. conceived the idea. K.L. carried out the fabrication, measurement setup construction, and device testing. Z.W. contributed to the implementation and training of the neural networks. K.X., S.L., K.B., and B.S. assisted in the deposition of the GSSe thin films and device testing. R.L., H.M., J.W., and C.S. assisted in the design and layout of the device. J.J. and Q.D. performed the preparation of the metal electrodes. W.Z., C.L., and S.D. provided the raw material chalcogenide glass and offered support during the waveguide fabrication process. H.L. and L.L. supervised the research. All the authors contributed to the technical discussions and writing of the paper. Competing interests The authors declare no competing financial interests. References Zhong, Y. et al. Dynamic memristor-based reservoir computing for high-efficiency temporal signal processing. Nature Communications 12 , doi:10.1038/s41467-020-20692-1 (2021). Moon, J. et al. Temporal data classification and forecasting using a memristor-based reservoir computing system. Nature Electronics 2 , 480-487, doi:10.1038/s41928-019-0313-3 (2019). Chen, Z. W. et al. All-ferroelectric implementation of reservoir computing. Nature Communications 14 , doi:10.1038/s41467-023-39371-y (2023). Youngblood, N., Ríos Ocampo, C. A., Pernice, W. H. P. & Bhaskaran, H. Integrated optical memristors. Nature Photonics 17 , 561-572, doi:10.1038/s41566-023-01217-w (2023). Farmakidis, N., Dong, B. & Bhaskaran, H. Integrated photonic neuromorphic computing: opportunities and challenges. Nature Reviews Electrical Engineering 1 , 358-373, doi:10.1038/s44287-024-00050-9 (2024). Shastri, B. J. et al. Photonics for artificial intelligence and neuromorphic computing. Nature Photonics 15 , 102-114, doi:10.1038/s41566-020-00754-y (2021). Tossoun, B. et al. High-speed and energy-efficient non-volatile silicon photonic memory based on heterogeneously integrated memresonator. Nature Communications 15 , doi:10.1038/s41467-024-44773-7 (2024). Cheung, S. et al. Ultra‐Power‐Efficient, Electrically Programmable, Multi‐State Photonic Flash Memory on a Heterogeneous III‐V/Si Platform. Laser & Photonics Reviews 18 , doi:10.1002/lpor.202400001 (2024). Zhang, G. et al. Thin film ferroelectric photonic-electronic memory. Light: Science & Applications 13 , doi:10.1038/s41377-024-01555-6 (2024). Taki, K. et al. Nonvolatile optical phase shift in ferroelectric hafnium zirconium oxide. Nature Communications 15 , doi:10.1038/s41467-024-47893-2 (2024). Wei, M. et al. Monolithic back-end-of-line integration of phase change materials into foundry-manufactured silicon photonics. Nature Communications 15 , 2786 (2024). Wei, M. et al. Electrically programmable phase-change photonic memory for optical neural networks with nanoseconds in situ training capability. Advanced Photonics 5 , doi:10.1117/1.Ap.5.4.046004 (2023). Lei, K. et al. Magnetron-sputtered and thermal-evaporated low-loss Sb-Se phase-change films in non-volatile integrated photonics. Optical Materials Express 12 , doi:10.1364/ome.462426 (2022). Seok, T. J., Quack, N., Han, S., Muller, R. S. & Wu, M. C. Large-scale broadband digital silicon photonic switches with vertical adiabatic couplers. Optica 3 , doi:10.1364/optica.3.000064 (2016). Edinger, P. et al. Silicon photonic microelectromechanical phase shifters for scalable programmable photonics. Opt. Lett. 46 , doi:10.1364/ol.436288 (2021). Nie, Y. K. et al. Integrated laser graded neuron enabling high-speed reservoir computing without a feedback loop. Optica 11 , 1690-1699, doi:10.1364/OPTICA.537231 (2024). Waser, R., Dittmann, R., Staikov, G. & Szot, K. Redox‐Based Resistive Switching Memories – Nanoionic Mechanisms, Prospects, and Challenges. Advanced Materials 21 , 2632-2663, doi:10.1002/adma.200900375 (2009). Zhang, G. et al. Functional Materials for Memristor‐Based Reservoir Computing: Dynamics and Applications. Advanced Functional Materials 33 , doi:10.1002/adfm.202302929 (2023). Wang, R. et al. Recent Advances of Volatile Memristors: Devices, Mechanisms, and Applications. Advanced Intelligent Systems 2 , doi:10.1002/aisy.202000055 (2020). Kim, H. J. et al. Super-Linear-Threshold-Switching Selector with Multiple Jar-Shaped Cu-Filaments in the Amorphous Ge3Se7 Resistive Switching Layer in a Cross-Point Synaptic Memristor Array. Advanced Materials 34 , doi:10.1002/adma.202203643 (2022). Chen, Z. & Zhi, C. Chalcogens for high-energy batteries. Nature Reviews Materials 10 , 268-284, doi:10.1038/s41578-025-00773-7 (2025). Wang, Z. et al. Doping effects of metal cation on sulfide solid electrolyte/lithium metal interface. Nano Energy 84 , doi:10.1016/j.nanoen.2021.105906 (2021). Wang, Z. et al. Threshold Switching of Ag or Cu in Dielectrics: Materials, Mechanism, and Applications. Advanced Functional Materials 28 , doi:10.1002/adfm.201704862 (2017). Sarwat, S. G., Moraitis, T., Wright, C. D. & Bhaskaran, H. Chalcogenide optomemristors for multi-factor neuromorphic computation. Nat Commun 13 , 2247, doi:10.1038/s41467-022-29870-9 (2022). Elfarash, A. H. & Gholipour, B. Reconfigurable nanoionic and photoionic material and device platforms. Advances in Physics: X 9 , doi:10.1080/23746149.2024.2338285 (2024). McRae, L., Xie, Y. & Gholipour, B. Photoionic Driven Movement of Metallic Ions as a Nonvolatile Reconfiguration Mechanism in Amorphous Chalcogenide Metasurfaces. Advanced Optical Materials 9 , doi:10.1002/adom.202101046 (2021). Frumar, M. & Wagner, T. Ag doped chalcogenide glasses and their applications. Current Opinion in Solid State and Materials Science 7 , 117-126, doi:10.1016/s1359-0286(03)00044-5 (2003). Wang, Z. et al. Memristors with diffusive dynamics as synaptic emulators for neuromorphic computing. Nature Materials 16 , 101-108, doi:10.1038/nmat4756 (2016). Deepak, P. et al. Influence of Ag Doping on Physico-Chemical Properties of the Ge28Sb12Se60 Chalcogenide Glassy Matrix. Journal of Materials Science and Engineering A 6 , doi:10.17265/2161-6213/2016.5-6.002 (2016). Svoboda, R., Málek, J. & Liska, M. Correlation between the structure and structural relaxation data for (GeSe2)y(Sb2Se3)1-y glasses. Journal of Non-Crystalline Solids 505 , 162-169, doi:10.1016/j.jnoncrysol.2018.11.013 (2019). Kochubei, H. K. et al. Raman scattering and X-ray diffraction studies of Ge16Sb24Se60 alloys. Semiconductor Physics Quantum Electronics & Optoelectronics 28 , doi:10.15407/spqeo28.03.300 (2025). Tsuchiya, T., Nakayama, T. & Ariga, K. Nanoarchitectonics Intelligence with atomic switch and neuromorphic network system. Applied Physics Express 15 , 100101, doi:10.35848/1882-0786/ac926b (2022). Terabe, K., Hasegawa, T., Nakayama, T. & Aono, M. Quantized conductance atomic switch. Nature 433 , 47-50, doi:10.1038/nature03190 (2005). Viswanathan, A., Udayan, S., Musfir, P. N., Nampoori, V. P. N. & Thomas, S. Enhancement of defect states assisted thermal diffusivity in solution-processed GeSeSb chalcogenide glass matrix on silver incorporation. Journal of Non-Crystalline Solids 503-504 , 151-157, doi:https://doi.org/10.1016/j.jnoncrysol.2018.09.039 (2019). Kolobov, A. V. Photo-induced metastability in amorphous semiconductors . (John Wiley & Sons, 2006). Wagner, T., Dale, G., Ewen, P., Owen, A. & Perina, V. J. J. o. A. P. Kinetics of the thermally and photoinduced solid state reaction of Ag with As 33 S 67 films. 87 , 7758-7767 (2000). Wagner, T. et al. The kinetics of the photo-induced solid-state chemical reaction in Ag/As33S67 bilayers and its reaction products. 79 , 223-237 (1999). Xiong, Y. et al. Observation of the failure mechanism in Ag10Ge15Te75-based memristor induced by ion transport. Nano Research 17 , 8431-8437, doi:10.1007/s12274-024-6791-2 (2024). Chekol, S. A., Menzel, S., Ahmad, R. W., Waser, R. & Hoffmann‐Eifert, S. Effect of the Threshold Kinetics on the Filament Relaxation Behavior of Ag‐Based Diffusive Memristors. Advanced Functional Materials 32 , doi:10.1002/adfm.202111242 (2021). Yang, J. T. et al. Artificial Synapses Emulated by an Electrolyte‐Gated Tungsten‐Oxide Transistor. Advanced Materials 30 , doi:10.1002/adma.201801548 (2018). Nie, F. et al. An Adaptive Solid‐State Synapse with Bi‐Directional Relaxation for Multimodal Recognition and Spatio‐Temporal Learning. Advanced Materials 37 , doi:10.1002/adma.202412006 (2025). Irfan, M. et al. DeepShip: An underwater acoustic benchmark dataset and a separable convolution based autoencoder for classification. Expert Systems with Applications 183 , doi:10.1016/j.eswa.2021.115270 (2021). Alqurashi, F. S., Trichili, A., Saeed, N., Ooi, B. S. & Alouini, M.-S. Maritime Communications: A Survey on Enabling Technologies, Opportunities, and Challenges. IEEE Internet of Things Journal 10 , 3525-3547, doi:10.1109/jiot.2022.3219674 (2023). Wei, T. et al. Hybrid Satellite-Terrestrial Communication Networks for the Maritime Internet of Things: Key Technologies, Opportunities, and Challenges. IEEE Internet of Things Journal 8 , 8910-8934, doi:10.1109/jiot.2021.3056091 (2021). Additional Declarations There is NO Competing Interest. Supplementary Files 202512Supplementaryfinal.docx Supplementary information for Ionic Driven Waveguide Integrated Memristor Cite Share Download PDF Status: Under Review 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-7782397","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":575463297,"identity":"9dd4fd8b-bd69-4902-aeb8-621c069953e0","order_by":0,"name":"Hongtao Lin","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA0klEQVRIiWNgGAWjYBACAxiDn4EhAUgxE6nlABBLNpCsxeAAmE+EFnP27sTHHyoO5xmfP/BMgqHCOrGB/ewBvFose85uNjhw5nCx2YEDaRIMZ9ITG3jyEvA77EbuNomDbbcTtx1sSJNgbDuc2CDBY0BIy/YfIC2bmxmAWv4Rp2UbA0jLBjaQlgZitJw5u1nizJn/iTPOMCRbJBxLN27jySGg5Xjvxg8VFWmJ/f1nEm98qLGW7Wc/g18LEuBJAEcmG7HqgYD9AAmKR8EoGAWjYCQBALP/TSdWrW4oAAAAAElFTkSuQmCC","orcid":"https://orcid.org/0000-0001-7432-3644","institution":"Zhejiang University","correspondingAuthor":true,"prefix":"","firstName":"Hongtao","middleName":"","lastName":"Lin","suffix":""},{"id":575463298,"identity":"34ca030e-37ba-4acd-a588-78a57402e4b7","order_by":1,"name":"Kunhao Lei","email":"","orcid":"","institution":"Zhejiang University","correspondingAuthor":false,"prefix":"","firstName":"Kunhao","middleName":"","lastName":"Lei","suffix":""},{"id":575463301,"identity":"1be10806-85c7-4203-97ec-2f146a0a71a3","order_by":2,"name":"Zijia Wang","email":"","orcid":"","institution":"Zhejiang University","correspondingAuthor":false,"prefix":"","firstName":"Zijia","middleName":"","lastName":"Wang","suffix":""},{"id":575463305,"identity":"beaaa686-8d5c-4ab5-b09d-a6afbe5a97ef","order_by":3,"name":"Kai Xu","email":"","orcid":"","institution":"Zhejiang University","correspondingAuthor":false,"prefix":"","firstName":"Kai","middleName":"","lastName":"Xu","suffix":""},{"id":575463310,"identity":"c9fb732d-638d-442d-898d-1f6a0d32ff5f","order_by":4,"name":"Shuo Lin","email":"","orcid":"","institution":"Zhejiang University","correspondingAuthor":false,"prefix":"","firstName":"Shuo","middleName":"","lastName":"Lin","suffix":""},{"id":575463313,"identity":"2cf1dc64-48c4-4ca2-82fd-62194031c796","order_by":5,"name":"Kangjian Bao","email":"","orcid":"","institution":"Westlake University","correspondingAuthor":false,"prefix":"","firstName":"Kangjian","middleName":"","lastName":"Bao","suffix":""},{"id":575463318,"identity":"6eb09722-a304-4154-b467-f92a91814502","order_by":6,"name":"Boshu Sun","email":"","orcid":"","institution":"Westlake University","correspondingAuthor":false,"prefix":"","firstName":"Boshu","middleName":"","lastName":"Sun","suffix":""},{"id":575463319,"identity":"f151e2f4-93b8-48fc-850d-02546d6effb7","order_by":7,"name":"Ruizhe Liu","email":"","orcid":"","institution":"Zhejiang University","correspondingAuthor":false,"prefix":"","firstName":"Ruizhe","middleName":"","lastName":"Liu","suffix":""},{"id":575463321,"identity":"51f78605-f039-4923-b577-a143cf7f2cc7","order_by":8,"name":"Hui Ma","email":"","orcid":"","institution":"Zhejiang University","correspondingAuthor":false,"prefix":"","firstName":"Hui","middleName":"","lastName":"Ma","suffix":""},{"id":575463324,"identity":"acb7e27f-f733-4e29-a36e-ffdeb17f6edd","order_by":9,"name":"Jiawei Weng","email":"","orcid":"","institution":"Zhejiang University","correspondingAuthor":false,"prefix":"","firstName":"Jiawei","middleName":"","lastName":"Weng","suffix":""},{"id":575463327,"identity":"212513c3-adcd-4df4-a4da-1cc4eacf3222","order_by":10,"name":"Jialing Jian","email":"","orcid":"https://orcid.org/0000-0001-6809-6588","institution":"Westlake University","correspondingAuthor":false,"prefix":"","firstName":"Jialing","middleName":"","lastName":"Jian","suffix":""},{"id":575463329,"identity":"37aaa559-4939-4ab9-ab29-188a2f918264","order_by":11,"name":"Qingyan Deng","email":"","orcid":"","institution":"Westlake University","correspondingAuthor":false,"prefix":"","firstName":"Qingyan","middleName":"","lastName":"Deng","suffix":""},{"id":575463331,"identity":"d7041b17-c1d8-4fa7-91bc-60c8cfad3d53","order_by":12,"name":"Wei Zhang","email":"","orcid":"","institution":"Ningbo university","correspondingAuthor":false,"prefix":"","firstName":"Wei","middleName":"","lastName":"Zhang","suffix":""},{"id":575463334,"identity":"a0134c15-98a0-4956-9ff0-4236762b8383","order_by":13,"name":"Changgui Lin","email":"","orcid":"https://orcid.org/0000-0002-0574-6296","institution":"Ningbo University","correspondingAuthor":false,"prefix":"","firstName":"Changgui","middleName":"","lastName":"Lin","suffix":""},{"id":575463336,"identity":"5c05fbd2-0697-4bae-bd17-67e76549be0c","order_by":14,"name":"Shixun Dai","email":"","orcid":"","institution":"Ningbo University","correspondingAuthor":false,"prefix":"","firstName":"Shixun","middleName":"","lastName":"Dai","suffix":""},{"id":575463337,"identity":"e0969ec4-e55f-43df-816b-563afd0c0533","order_by":15,"name":"Lan Li","email":"","orcid":"https://orcid.org/0000-0002-9097-9157","institution":"Westlake University","correspondingAuthor":false,"prefix":"","firstName":"Lan","middleName":"","lastName":"Li","suffix":""}],"badges":[],"createdAt":"2025-10-05 02:40:08","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7782397/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7782397/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":100758914,"identity":"b8ccd9a4-9d9a-4094-9421-655445f09b24","added_by":"auto","created_at":"2026-01-21 07:07:38","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":11548616,"visible":true,"origin":"","legend":"","description":"","filename":"202512Manuscriptfinal.docx","url":"https://assets-eu.researchsquare.com/files/rs-7782397/v1/b1f677b50e2a7c2316ceb1b8.docx"},{"id":100758782,"identity":"d1182d1a-1623-4728-aad9-eb0eb7e979c0","added_by":"auto","created_at":"2026-01-21 07:06:04","extension":"json","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":15199,"visible":true,"origin":"","legend":"","description":"","filename":"NCOMMS2578361AZ.json","url":"https://assets-eu.researchsquare.com/files/rs-7782397/v1/43e734ac4b1f242b6240ba32.json"},{"id":100758812,"identity":"d27b1137-08e7-40c8-8f55-30701aae4e2d","added_by":"auto","created_at":"2026-01-21 07:06:23","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":3079420,"visible":true,"origin":"","legend":"","description":"","filename":"202512Supplementaryfinal.docx","url":"https://assets-eu.researchsquare.com/files/rs-7782397/v1/27ee3953edbea528cc8e0bb5.docx"},{"id":100758807,"identity":"389da136-aab3-45e7-999e-7583ff62eae7","added_by":"auto","created_at":"2026-01-21 07:06:14","extension":"xml","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":107277,"visible":true,"origin":"","legend":"","description":"","filename":"NCOMMS2578361AZ1enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-7782397/v1/b909bf51a0762567a7b7f226.xml"},{"id":100758806,"identity":"a58b4370-4e58-4f8b-b919-88e516278096","added_by":"auto","created_at":"2026-01-21 07:06:11","extension":"jpeg","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1397971,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage1.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7782397/v1/7a1189ee2d13c9a3284dfe2c.jpeg"},{"id":100758799,"identity":"81ea4588-b323-4af9-983f-a0ff5fbc4158","added_by":"auto","created_at":"2026-01-21 07:06:05","extension":"jpeg","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":3515675,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage2.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7782397/v1/e11116c081812477098ce2dd.jpeg"},{"id":100758809,"identity":"ede2551c-7ea3-4acd-9262-4686228aa6f7","added_by":"auto","created_at":"2026-01-21 07:06:16","extension":"jpeg","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1450166,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage3.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7782397/v1/34c31d7e62822b89645d4183.jpeg"},{"id":100758805,"identity":"06d30267-de77-49bd-9d1c-62747a95dd47","added_by":"auto","created_at":"2026-01-21 07:06:08","extension":"jpeg","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1524829,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage4.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7782397/v1/6b1c7eda2792e625ee3593c3.jpeg"},{"id":100758921,"identity":"a7bc2175-1eac-48d8-bd85-466ae4d339a3","added_by":"auto","created_at":"2026-01-21 07:08:01","extension":"jpeg","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":3560231,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage5.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7782397/v1/01efb771cca888a9e75988f9.jpeg"},{"id":100758913,"identity":"aec3146e-7e7e-4222-92ea-9896602c225b","added_by":"auto","created_at":"2026-01-21 07:07:37","extension":"png","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":867311,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7782397/v1/9c70677eb3afded291eb5012.png"},{"id":100758903,"identity":"33bf2558-38a0-4e19-83ea-1fb06872a6f1","added_by":"auto","created_at":"2026-01-21 07:07:29","extension":"png","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":2742619,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7782397/v1/8003adce373de6442d671c58.png"},{"id":100758923,"identity":"3b1802fd-38ce-4ea0-bd3d-cc67a6464b0a","added_by":"auto","created_at":"2026-01-21 07:08:05","extension":"png","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":801265,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7782397/v1/2dc7a7f3b1c33f62746904a7.png"},{"id":100758875,"identity":"43d750fa-116d-4973-8ab6-5d7537a31d13","added_by":"auto","created_at":"2026-01-21 07:07:15","extension":"png","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1010666,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-7782397/v1/0d5158f77d33aa154e407032.png"},{"id":100758822,"identity":"5562726d-a810-4808-bfc4-ccf6a2005c75","added_by":"auto","created_at":"2026-01-21 07:06:55","extension":"png","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":2186288,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-7782397/v1/f88a8d7ceaff84629b44b004.png"},{"id":100758922,"identity":"bd892490-840c-45d0-aa4f-a4d4364f653a","added_by":"auto","created_at":"2026-01-21 07:08:04","extension":"xml","order_by":14,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":106202,"visible":true,"origin":"","legend":"","description":"","filename":"NCOMMS2578361AZ1structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-7782397/v1/b4cb9734745d309c23fd9586.xml"},{"id":100758815,"identity":"7a1a7a11-1f74-41db-95b1-6a26f248ffe8","added_by":"auto","created_at":"2026-01-21 07:06:32","extension":"html","order_by":15,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":114334,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-7782397/v1/91ca3191cfbdcbeb1d397b94.html"},{"id":100758898,"identity":"b47add9f-6450-43d2-864d-dc4359ac3e36","added_by":"auto","created_at":"2026-01-21 07:07:21","extension":"jpeg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":1397971,"visible":true,"origin":"","legend":"\u003cp\u003eConcept of ionic-driven photonics RC system. a) Schematic diagram of the underwater acoustic signal recognition process. b) Schematic of the ionic-driven memristor-based RC system. c) Schematic of ionic-driven waveguide integrated memristor platform. d) Photograph of the photonic memristor chip. e) The structure of neurons and synapses. Inset: Schematic illustration of the synaptic junction. f) Schematic cross-section of the doped region of the dynamic memristor. g) Response curve of the electrically reconfigurable dynamic memristor.\u003c/p\u003e","description":"","filename":"floatimage1.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7782397/v1/e3290695d531781269cd8b20.jpeg"},{"id":100758810,"identity":"008390e2-7602-4e25-b1aa-7ec69feef118","added_by":"auto","created_at":"2026-01-21 07:06:19","extension":"jpeg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":3515675,"visible":true,"origin":"","legend":"\u003cp\u003eFabrication and principle of the metal-doped GSSe. a) Flowcharts of the memristor fabrication and doping processes. b) Molecular structure of GSSe before and after doping with Ag. c) Schematic diagram of the migration of Ag ions driven by the electric field. d) Raman spectroscopic characterization of GSSe films prepared via two doping processes. e) TEM images and EDS characterization of the dynamic memristor. f) In-situ TEM characterization of the reconfigurable and attenuation response mechanisms of the dynamic memristors.\u003c/p\u003e","description":"","filename":"floatimage2.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7782397/v1/63959e1b53fb92215df3075d.jpeg"},{"id":100758872,"identity":"ad9bc8fe-386c-40c6-b479-897d42ab9178","added_by":"auto","created_at":"2026-01-21 07:07:04","extension":"jpeg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":1450166,"visible":true,"origin":"","legend":"\u003cp\u003ea, b) Optical microscope image of the MRRs and the scanning electron microscope (SEM) image of the doped region. c) Transmission spectra of microring response with 10 μm Ag cell length deposited by magnetron sputtering. d) Resonant peak shift of the MRRs as they transition from Ag deposition to photodoping under various Ag cell lengths. e) Quality factor analysis of the MRRs under various Ag cell lengths. f) Kinetics analysis of resonant peak shift and propagation loss in photo-doped MRRs with 10 μm Ag cell length. g) Optical microscope image of the waveguide attenuator. h) Multilevel photodoping process of the waveguide attenuator. i) Retention characteristics of the waveguide attenuator after the deposited metal process (Ag deposited) or the photoinduced metal doping state process (photodoped).\u003c/p\u003e","description":"","filename":"floatimage3.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7782397/v1/39e267a39b7a616992cd8cfb.jpeg"},{"id":100758915,"identity":"f5bdd0e8-b05f-48d6-8855-906f724f9c5d","added_by":"auto","created_at":"2026-01-21 07:07:38","extension":"jpeg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":1524829,"visible":true,"origin":"","legend":"\u003cp\u003ea) Optical microscope image of the doped MRR. b) Transmission spectra of the reconfigurable switching events of the doped MRR. c) Twenty cycles of reconfigurable switching events of the doped MRR (top: set, bottom: reset). d) Optical microscope image of the doped waveguide attenuator. e) Transmission spectra of the reconfigurable switching events of the doped waveguide attenuator. f) Multilevel operation of the doped waveguide attenuator with voltage ranging from 50 to 130 V. g) Dynamic characteristics of the doped waveguide attenuator upon varying pulse amplitude. h) Dynamic characteristics of the doped waveguide attenuator upon varying pulse duration. i) Modulation of the transmission by adjusting the pulse amplitude and duration.\u003c/p\u003e","description":"","filename":"floatimage4.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7782397/v1/8e04cb0529189f4305594f97.jpeg"},{"id":100758902,"identity":"49968cae-7fcd-40d5-8be0-436dbacb1e3d","added_by":"auto","created_at":"2026-01-21 07:07:27","extension":"jpeg","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":3560231,"visible":true,"origin":"","legend":"\u003cp\u003ea, b) Data preprocessing and hybrid reservoir-CNN architecture for underwater acoustic signal recognition. c, d) The comparative analysis of the training performance and classification accuracy achieved using original data, binarized data, and hybrid reservoir-CNN model. e) The confusion matrix comparing the classification results obtained via the memristor RC system with the ground truth, demonstrating an overall recognition accuracy of 73.0%.\u003c/p\u003e","description":"","filename":"floatimage5.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7782397/v1/155db061504e553047ae7b09.jpeg"},{"id":100759227,"identity":"0a1ebe65-4f92-42a3-bd54-d3feb1806e3d","added_by":"auto","created_at":"2026-01-21 07:14:32","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":12107682,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7782397/v1/884fb5a8-4d0b-43d2-901e-e0489044bdbf.pdf"},{"id":100758781,"identity":"b22ff77f-d74b-4c0f-bb86-7e50b2c77beb","added_by":"auto","created_at":"2026-01-21 07:06:02","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":3079420,"visible":true,"origin":"","legend":"Supplementary information for Ionic Driven Waveguide Integrated Memristor","description":"","filename":"202512Supplementaryfinal.docx","url":"https://assets-eu.researchsquare.com/files/rs-7782397/v1/cc85125eea407fe1afefac02.docx"}],"financialInterests":"There is \u003cb\u003eNO\u003c/b\u003e Competing Interest.","formattedTitle":"Ionic Driven Waveguide Integrated Memristor","fulltext":[{"header":"Introduction","content":"\u003cp\u003eThe rapid advancement of artificial intelligence (AI) has recently resulted in the growing complexity of artificial neural network (ANN) architectures, accompanied by increased training costs and a continuous rise in the demand for enhanced hardware data processing capabilities. However, the von Neumann computing paradigm is increasingly struggling to meet the demands for efficient and low-latency data processing due to the significant time and energy overhead associated with frequent data transfers between computing and storage units. Recent advancements in electronic neuromorphic memristors have underscored the benefits of the neuro-bionic paradigm and have been utilized in efficient ANN architectures, such as the lightweight reservoir computing (RC) architecture based on time-division multiplexing, not only showcasing their exceptional capacity for processing temporal data, but also demonstrating their effectiveness in substantially reducing training costs and minimizing the overall system footprint\u003csup\u003e\u003cspan additionalcitationids=\"CR2\" citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u003c/sup\u003e. Owing to the substantial advantages of optical interconnection in communication networks\u0026mdash;such as large bandwidth, high parallelism, and low power consumption, and the synergistic potential with the non-von Neumann neuromorphic computing paradigm, photonic artificial memristor devices offer a novel hardware platform for ANNs\u003csup\u003e\u003cspan additionalcitationids=\"CR5\" citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e\u003c/sup\u003e. Currently, a diverse range of emerging memristors based on photonic integrated circuits (PICs) has been widely applied for optical memories, post-fabrication trimming of PICs, and in-memory computing, such as Ⅲ-Ⅴ/Si hybrid integration memristor\u003csup\u003e\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e,\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u003c/sup\u003e, ferroelectric materials\u003csup\u003e\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e,\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u003c/sup\u003e, chalcogenide phase change materials (PCMs)\u003csup\u003e\u003cspan additionalcitationids=\"CR12\" citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e\u003c/sup\u003e, and microelectromechanical systems (MEMS)\u003csup\u003e\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e,\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u003c/sup\u003e. Most of the aforementioned memristors demonstrate non-volatile behavior with a monotonous temporal response scale, rendering them suitable for storing trained weights or executing linear optical signal processing in the output layer of artificial neural networks. Meanwhile, owing to distinct functional requirements, the reservoir neurons in time-division multiplexed photonic RC systems without feedback loops necessitate memristors that possess dynamic and nonlinear response characteristics\u003csup\u003e\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e\u003c/sup\u003e\u0026mdash;features that are challenging to achieve with purely non-volatile devices.\u003c/p\u003e \u003cp\u003eChalcogenide glass (ChGs), an amorphous soft glass material composed of elements such as sulfur (S), selenium (Se), and/or tellurium (Te), has demonstrated outstanding performance in chalcogenide solid state batteries, memristors, and the development of associated large-scale device arrays, owing to its superior ionic conductivity and excellent processability\u003csup\u003e\u003cspan additionalcitationids=\"CR18 CR19 CR20\" citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e\u003c/sup\u003e. Modification through the ion doping mechanism and effect has currently emerged as a mainstream and significant strategy for chalcogenide solid-state batteries and metal-insulator-metal (MIM) memristors. Firstly, doping elements can enhance the ionic conductivity of the electrolyte as they not only generate point defects but also widen the ionic channels. Secondly, doping can also regulate the composition of the electrolyte, decrease interface resistance, and enhance the chemical stability of the solid electrolyte\u003csup\u003e\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e,\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e\u003c/sup\u003e. The MIM structure, composed of metal-ChGs-metal, not only demonstrates electrical responses but also realizes neuromorphic functionalities for optoelectronic multi-signal interactions via the photovoltaic effect induced by the asymmetric Schottky barrier formed between the top and bottom electrodes and the chalcogenide medium\u003csup\u003e\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e\u003c/sup\u003e. Ion-doped ChGs are capable of not only modifying their electrical conductivity but also tuning their band gap, thereby enabling control over the variation of optical constants within the ChGs material. This ionic behavior offers a promising pathway for the development of reconfigurable photonic devices\u003csup\u003e\u003cspan additionalcitationids=\"CR26\" citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eIn this work, we employed the ion-doping mechanism and effect of chalcogenide solid-state batteries to propose an ion-driven waveguide-integrated memristor platform for the first time, and demonstrate the application potential of the RC system based on these neuronal devices in the domain of underwater acoustic signal recognition, as illustrated in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e. The proposed process flow eliminates the need for high-cost semiconductor ion implantation equipment and associated procedures. By modifying the Ge\u003csub\u003e28\u003c/sub\u003eSb\u003csub\u003e12\u003c/sub\u003eSe\u003csub\u003e60\u003c/sub\u003e (GSSe) waveguide through electrochemically active metal doping and integrating the mechanism of optical- and electric field-induced ion migration, the approach enables multi-dimensional modulation of both the optical phase and attenuation within the GSSe waveguide. The stable non-volatile tuning memristor can be achieved by modifying the band gap of the chalcogenide glass through photo-induced sputtered metal Ag dissolution, while the dynamic attenuation memristor was achieved by controlling the formation of silver nanocluster in the Ag\u003csup\u003e+\u003c/sup\u003e solution-doped GSSe waveguide using an electric field, thereby enhancing the absorption and scattering losses within the waveguide. Furthermore, the intrinsic dynamic properties of the GSSe memristors, coupled with the variation characteristics of nonlinear analog transmission signals, render them highly suitable for implementing photonics RC systems. Within this architecture, the dynamic memristor can be mathematically modeled as a reservoir that nonlinearly maps the input encoded time-domain signal into a high-dimensional feature space. The ion-driven waveguide-integrated memristor platform, based on the multi-dimensional modulation mechanism, achieves monolithic integration of non-volatile and volatile dynamic memristors. It enables the development of diverse neuron devices with varying time response scales for the optical neural networks, offering a novel pathway for expanding the emerging memristor technologies and providing significant support for enhancing brain-like functionalities in neural network systems.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003e\u003cstrong\u003eIntegrated Memristor-based RC system\u003c/strong\u003e\u003cstrong\u003e\u0026nbsp;for Acoustic Target Recognition\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eUnderwater acoustic target recognition, as a pivotal technology, holds extensive application potential in areas such as marine resource exploration, environmental monitoring, and vessel identification. Currently, propelled by the ongoing advancements in various marine observation technologies, including satellite remote sensing, ocean buoys, and research vessels, the volume of marine data is increasing exponentially. This exponential growth poses significant challenges to existing maritime communication satellites, which struggle to effectively manage the surge in observational data and meet the demand for real-time monitoring. Consequently, the development of temporal data processing and identification technologies for facilitating communication among observation vessels, satellites, and shore stations has emerged as a critical focus. Here, we propose an ion-driven waveguide-integrated memristor-based RC system for processing temporal signals. In contrast to the traditional photonic RC architecture, which necessitates the establishment of a delay feedback loop to enhance the complexity of the reservoir node, this system leverages the time-domain analog signal response of the photonic memristor as the reservoir node. It directly reads the transmission changes of the optical signal following the application of a write pulse, eliminating the need for additional read pulses during data retrieval. This approach not only conserves spatial footprint but also simplifies the overall system architecture.\u003c/p\u003e\n\u003cp\u003eThe concept of the ion-driven waveguide-integrated memristor RC system for underwater acoustic signal recognition is illustrated in Fig. 1. When the sensors on the observation vessel detect the target signal, the encoder compresses and encodes the signal before transmitting it to the maritime satellite, thereby accommodating the satellite\u0026apos;s limited computing resources and transmission bandwidth. Further, the satellite transmits the compressed data to the shore station, where the acoustic signal is decoded and recognized by the decoder utilizing the photonic RC system. \u0026nbsp;For the photonic RC system, we propose an architecture that integrates on-chip wavelength division multiplexing (WDM) and time division multiplexing techniques with parallel memristor arrays to enhance the parallelism and signal processing capabilities. The architecture comprises a WDM and demultiplexing layer, a reservoir layer leveraging the dynamic response of memristors, and a fully connected output layer that can potentially exploit the non-volatile characteristics of memristors in future implementations, as illustrated in Fig. 1b. Fig. 1c presents the two schemes that can be monolithically integrated for doping and multi-physical field modulation of GSSe waveguides. The optical field modulation approach enables bandgap and refractive index tuning through a photochemical reaction induced by ultraviolet light, which drives the interaction between sputtered Ag and GSSe waveguide. The electric field modulation strategy facilitates ion doping by immersing the waveguides in an Ag\u003csup\u003e+\u003c/sup\u003e rich solution, followed by the aggregation of metal nanoclusters within the waveguide under an applied transverse electric field. By employing precise micro-nano fabrication techniques, we have successfully achieved monolithic integration of two distinct types of waveguide-integrated memristors featuring different doping profiles and modulation mechanisms, as shown in Fig. 1d. The photonic chip integrated hundreds of memristors has an overall footprint smaller than that of a one-yuan CNY coin. Among them, the dynamic memristor utilized for the reservoir layer forms a capacitive structure by placing electrodes on either side of the Ag\u003csup\u003e+\u003c/sup\u003e doped GSSe waveguide. Under the influence of an electric field, it governs the migration of silver nanoclusters within the waveguide. When these nanoclusters coalesce into larger Ag particles, they induce additional absorption and scattering losses, as depicted in Fig. 1f. Upon the low electrical voltage, owing to thermal effects and the driving force of interfacial energy, the nanoclusters undergo deformation and decay, causing the propagation loss of the waveguide to revert to its initial state\u003csup\u003e28\u003c/sup\u003e. The dynamic response of the electro-reconfigurable memristor is illustrated in Fig. 1g. The optical power decay curve can be further fitted to an exponential function.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFabrication\u003c/strong\u003e\u003cstrong\u003e\u0026nbsp;and Principle\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFig. 2a illustrates the fabrication processes of two types of memristors, which are described in detail in the methods section. To investigate the impact of Ag-doped chalcogenide glass on its molecular network, the doped chalcogenide thin films were characterized by Raman spectroscopy. As illustrated in Fig. 2d, the thermal evaporated GSSe chalcogenide film primarily exhibits three fundamental Raman scattering peaks. The most significant peak, located at 200 cm\u003csup\u003e-1\u003c/sup\u003e, corresponds to the symmetric stretching mode associated with the corner-sharing [GeSe\u003csub\u003e4/2\u003c/sub\u003e] tetrahedral units. The characterization results indicate that the scattering peak intensity at 170 cm⁻\u0026sup1; and 100 cm⁻\u0026sup1; increase with increasing silver doping concentration, whereas the intensity of the peak at 270 cm⁻\u0026sup1; decreases to some extent, which can be attributed to the vibration of Ge\u0026ndash;Ge bond, the umbrella vibration of [SbSe\u003csub\u003e3/2\u003c/sub\u003e] pyramids unit and the symmetric stretching vibrations of Se\u0026ndash;Se bond, respectively. Based on this phenomenon, we speculate that Ag doped and modified chalcogenide films create Se-deficient conditions (forming Ag\u003csub\u003e2\u003c/sub\u003eSe through bonding with Se), thereby promoting the rearrangement of Ge-Ge and Sb-Se bonds. This finding is consistent with similar experimental results reported in the literature\u003csup\u003e29-31\u003c/sup\u003e. Fig. 2b depicts the molecular structure of GSSe before and after doping with metallic Ag. Based on Raman spectroscopy characterization, it is found that the doping of GSSe films with Ag ions primarily occurs through the breaking of Se-Se homopolar bonds. Meanwhile, in Ge-rich chalcogenide films, the coordination bonds between two neighboring chains are relatively weak and are more likely to be broken, ionized, or form bonds with Ag ions, thus offering a natural transport channel for the migration of Ag ions.\u003c/p\u003e\n\u003cp\u003eTo verify the reconfigurable mechanism of the dynamic memristors, we carried out an in-situ transmission electron microscopy (TEM) experiment on the area of the device doped with an Ag ion solution, and the sample was prepared using focused ion beam (FIB) technology. As shown in the initial state of the device and the energy dispersive spectroscopy (EDS) analysis in Fig. 4e, after the memristor has been switched several times, a small amount of Ag species on both sides of the waveguide has already been unable to dissolve back into the GSSe waveguide spontaneously, and the distribution of silver ions within the waveguide was non-uniform (the initial uniformly doped distribution state prior to the switching operation was presented in supplementary information SI.2). This indicates that the device has approached a fatigue failure state. Subsequently, a voltage was applied from right to left at a step size of 5 mV per second. When the voltage reached 0.29 V, as depicted in Fig. 4f-i, it was observed that the Ag species on the right side of the waveguide dissolved in comparison to the initial state. Under the influence of the electric field, the Ag ions within the waveguide migrated to the lower left side, where they nucleated and aggregated. Furthermore, we restored the voltage to 0 and applied the voltage from left to right in the same step. When the voltage reached 0.21 V, we observed a similar phenomenon as shown in Fig. 4f-ii. Under the influence of the electric field, the Ag species on the left re-oxidized and dissolved back into the waveguide, and the Ag ions inside the waveguide migrated to the lower right of the waveguide and nucleated, realizing a reversible process regulated by the electric field. After observing the nucleation phenomenon twice, we immediately restored the voltage to zero and carried out EDS and high-resolution diffraction characterization on the species at the lower left corner of Fig. 4f-iii and the lower right corner of Fig. 4f-iv. Through the EDS and diffraction analysis presented in Fig .4f-iii, iv, v, it has been found that Ag ions migrate to the surface of the GSSe waveguide under the influence of the electric field and undergo a reduction reaction to form elemental Ag. Consequently, this leads to an increase in the absorption and scattering losses of the waveguide and the formation of an attenuation modulation. Meanwhile, according to the TEM and EDS diagram in Fig. 4f-iv, it is evident that a portion of the metallic Ag has spontaneously dissolved, which also suggests that this mechanism is a dynamic and volatile process. The high-resolution TEM image presented in Fig. 4f-v shows the regular lattice of the nucleated Ag species. The lattice spacings were measured to be 0.236 nm, corresponding to the (111) crystal planes of the cubic crystal system of elemental Ag. The mechanism of elemental Ag formation is speculated to be that Ag ions migrate to the surface of the waveguide under the influence of an electric field, where they undergo a reduction reaction with the defect states possessing lone pairs of electrons in the doped waveguide and the tunneling electrons under the influence of the electric field\u003csup\u003e32-34\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePhoto-doping Response of the Non-volatile GSSe/Ag Memristors\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFor the process of the non-volatile photo-induced diffusion of metallic Ag in chalcogenide glass, when light with energy exceeding the optical bandgap of the chalcogenide compound is incident on the waveguide, photogenerated holes and electrons are created. Electrons tend to be captured over short distances, forming negative charge centers, while holes migrate over longer distances, accumulate at the interface between metallic silver and the chalcogenide waveguide, and induce ionization. This process results in the formation of silver ions and free electrons, where the silver ions further diffuse and react with sulfur-based nanoclusters\u003csup\u003e35\u003c/sup\u003e. As the silver concentration increases, the absorption edge of the chalcogenide glass undergoes a redshift, accompanied by a simultaneous increase in the refractive index\u003csup\u003e27\u003c/sup\u003e. To further investigate the photoinduced dissolution process of Ag-doped GSSe waveguide, we conducted quantitative optical measurements and analysis of the photodoped micro-ring resonator (MRR) devices. Fig. 3a and b present the optical microscope image of the device and the scanning electron microscope (SEM) image of the doped region with the sputtered Ag cell. The transmission spectrum depicted in Fig. 3c illustrates an MRR with a doped length of 10 \u0026mu;m. The results indicate that the deposition of Ag on the waveguide leads to a significant decrease in the extinction ration and Q factor of the microring, suggesting a substantial increase in absorption loss due to the metal deposited. However, following the photoinduced dissolution of the Ag by using UV light, the extinction ration and Q factor nearly returns to its initial state.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eSimultaneously, we characterize the variations in the wavelength shift and quality factor (Q) of the MRRs as they transition from Ag deposition to photodoping under various doped lengths. As illustrated in Fig. 3d, following the photoinduced doping process, the redshift of the MRR resonance peak can be enhanced from 0.63 nm to 1.80 nm by varying the doped length within the range of 5 \u0026mu;m to 50 \u0026mu;m, which enables the realization of a compact phase shifter with an L\u003csub\u003e𝜋\u003c/sub\u003e of approximately 26 \u0026mu;m (L\u003csub\u003e𝜋\u003c/sub\u003e=𝜆/2\u0026Delta;neff, free spectral range: 2.3 ~ 2.4 nm). Next, by comparing the quality factor Q of the MRR, we analyze the changes in optical loss of the MRR during the process of photo doing. In Fig. 3e, the loaded Q factor of the single-mode waveguide GSSe MRR is about 2.25~3.39\u0026times;10\u003csup\u003e4\u003c/sup\u003e. After Ag cells were deposited on the GSSe waveguide via sputtering, the loaded Q factor of the MRR decreased to a range of 0.34\u0026times;10\u003csup\u003e3\u003c/sup\u003e to 0.72\u0026times;10\u003csup\u003e3\u003c/sup\u003e. This reduction was attributed to increased metal absorption, which exhibited a rising trend as the length of the sputtered Ag cell increased. The loaded Q factor of the MRR is improved and nearly restored to its initial level (1.82~2.59\u0026times;10\u003csup\u003e4\u003c/sup\u003e) after implementing the photodoping GSSe waveguide process by using UV light-induced silver dissolution. Furthermore, the kinetic process of photoinduced Ag dissolution doping in GSSe MRRs was examined. As shown in Fig. 3f, we exposed the hybrid MRR to UV light for durations ranging from 15 to 150 minutes and recorded the transmission response. It can be observed from the results that, upon the initiation of UV illumination, the growth rate of redshift for the resonant wavelength decreases exponentially. This phenomenon can be attributed to the initial stage of photoinduced dissolution of chalcogenide films doped with Ag, where the doping rate is determined by the photochemical reaction occurring at the interface between the metal Ag and the chalcogenide film. When the thickness of the doped film reaches a critical value, the rate of the doping process shifts from being dominated by the photochemical reaction at the interface to being determined by the diffusion of silver ions and other charged particles within the doped film. Consequently, the process rate becomes diffusion-limited\u003csup\u003e36,37\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eFurther, the multilevel characteristics of the hybrid integrated GSSe/Ag waveguide attenuator were experimentally verified, the structural schematic diagram is presented in Fig. 3g. As illustrated in Fig. 3h, the insertion loss of the waveguide attenuator with 100 \u0026mu;m metal Ag deposited at a wavelength of 1630 nm was measured to be -23.8 dB. By progressively increasing the UV light dose to induce the dissolution of the metal Ag layer, the insertion loss of the waveguide attenuator can be gradually reduced, simultaneously enabling the realization of 32-level (~5 bit) intensity modulation. This characteristic positions it as a promising candidate for weight storage in the fully connected output layer of RC systems. To verify the non-volatile properties of the device, the retention time of both the deposited metal state (Ag deposited) and the photoinduced metal doping state (photodoped) was tested under the same wavelength. Neither state exhibited significant degradation over a period of 0 to 10\u003csup\u003e3\u003c/sup\u003e seconds, as shown in Fig. 3i.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eChalcogenide\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003eDynamic Waveguide Integrated Memristors\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eBased on the Ag\u003csup\u003e+\u003c/sup\u003e solution doping process, we propose a novel electrically reconfigurable chalcogenide dynamic waveguide integrated memristor to simulate the reservoir layer of the photonics RC systems. Based on the aforementioned modulation mechanism, an MRR device featuring a doping length of 90 \u0026mu;m and achieving an intensity tuning exceeding 3 dB was developed, as illustrated in Fig. 4b. The set/reset switching cycle was found to exceed 20 cycles, and the corresponding transmission spectrum is presented in Fig. 4c. Based on the TEM characterization presented in the previous section, the observed switching failure is hypothesized to result from the presence of residual Ag nanocluster after the reset process\u003csup\u003e38\u003c/sup\u003e, which leads to increased insertion loss of the device. Future improvements could focus on optimizing the doping process and switching parameters to enhance device performance and reliability. After this, a switch characterization measurement was conducted on a waveguide attenuator device to further investigate the modulation mechanism. As shown in Fig. 4e, a reconfigurable attenuator device featuring a switch ratio of approximately 25 dB can be achieved with a doping length of 100 \u0026mu;m, while maintaining an insertion loss of -3 dB. Under the influence of the electric field, the doped waveguide demonstrates a relatively high absorption loss, which further corroborates our hypothesis that the electric field induces the formation of Ag nanoclusters within the waveguide, thereby causing substantial absorption and scattering losses. As illustrated in Fig. 4f, by progressively applying voltages ranging from 50 to 130 volts, a quasi-continuous 28-level (˃ 4 bit) multilevel intensity tuning was achieved.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTo simulate brain-like neuron functions, we systematically characterized the response of the waveguide attenuation dynamic memristor under different pulsed voltage conditions. Similar to electronic dynamic memristors that regulate the excitation or inhibition of postsynaptic currents via presynaptic electrical pulse mechanisms, ion-driven waveguide-integrated memristors can also modulate the inhibition of postsynaptic light intensity responses at specific wavelengths by leveraging the light absorption mechanism induced through electrically driven Ag nanoclusters. Among these, the key parameters of the pulse voltage primarily include the pulse amplitude, pulse duration, and pulse delay (i.e., the time interval between consecutive pulses). Firstly, we maintained the pulse duration constant (3.39 seconds) while systematically varying the pulse amplitude from 110 to 130V. The test results are presented in Fig. 4g. It can be observed that as the pulse amplitude increases, the attenuation absorption of the waveguide attenuator at a fixed wavelength gradually rises, with the modulation depths exceeding 25 dB. This indicates that under the influence of a stronger electric field, a larger number of Ag ions are involved in diffusion motion, leading to the formation of nanoclusters. Consequently, these not only enhance the absorption and scattering losses of the attenuator but also result in a significant increase in the relaxation time of the device. Meanwhile, by fixing a pulse amplitude at 110V and increasing the pulse duration (1.70 to 7.35s), we observed the same phenomenon, as illustrated in Fig. 4h. However, the results indicate that the scheme with variable pulse duration has a more restricted influence on the diffusion of Ag nanoclusters compared to the variable pulse amplitude. This is attributed to the fact that an increase in pulse duration primarily affects the coarsening process following the nucleation of the Ag nanoclusters\u003csup\u003e39\u003c/sup\u003e, thereby limiting the change in absorption of the intrinsic optical mode. As illustrated in Fig. 4i, the dynamic attenuator\u0026apos;s transmission can be modulated in a wide range by adjusting the pulse amplitude and duration to simulate various synaptic activities. Specifically, paired-pulse facilitation (PPF) and paired-pulse depression (PPD) are two important mechanisms of short-term plasticity, which describe the relative facilitation or depression of postsynaptic currents evoked by the second pulse compared to the first pulse after an inter-pulse interval\u003csup\u003e40,41\u003c/sup\u003e. Based on the characteristic that the optical absorption of the waveguide attenuator increases under electrical pulse stimulation, we measured the PPD index of the device by keeping the pulse amplitude and duration constant while increasing the pulse interval time, and the results are presented in SI.5.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMemristor-based RC system for Underwater Acoustic Recognition\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe RC system is a computational framework that transforms input temporal signals into high-dimensional representations through the reservoir layer. The switching behavior exhibited by the dynamical waveguide integrated memristor we developed closely resembles that of a dynamic reservoir: the future optical transmission state is determined by both the applied input voltage signal and the current transmission condition of the device. To showcase the neuromorphic computing capabilities of the doped GSSe memristor, we propose a reservoir computing architecture leveraging WDM technology and apply it to an underwater acoustic target recognition task. We utilized the DeepShip dataset\u003csup\u003e42\u003c/sup\u003e, comprising four primary categories of vessels: cargo ships, tanker ships, passenger ships, and tug ships, to evaluate the temporal data processing capability of the RC system (as shown in Figs. 5a and b). Underwater acoustic target recognition serves as a critical technology for marine monitoring and has traditionally relied on deep learning methods for target identification. However, in practical deployment scenarios, these approaches face a significant data transmission bottleneck. Specifically, after underwater acoustic sensors collect raw signals and extract features using the constant-Q transform (CQT) (as illustrated in Fig. 5a), these feature data must be transmitted to shore stations via satellite communication. Considering that modern advanced satellite communication systems are capable of operating with data transmission rates of 50 Mbps\u003csup\u003e43,44\u003c/sup\u003e, the transmission of 100 feature matrices, each of dimension 96\u0026times;188, requires approximately 1.15 seconds. When thousands of sensors within the coverage area of a single satellite are transmitting data concurrently, the resulting queuing delay can accumulate to hundreds of seconds. This latency significantly limits the system\u0026apos;s ability to perform real-time target recognition.\u003c/p\u003e\n\u003cp\u003eTo address this issue, we propose a hybrid architecture that integrates data binarization compression with the photonic RC decoding system. First, the time-frequency features to be transmitted were subjected to binarization, which reduces the data volume from 72 Kb to 2.256 Kb and shortens the transmission time to 36 milliseconds, thereby improving satellite data transmission efficiency by 3200%. Subsequently, the binarized data received on the ground was reconstructed using an ion-driven photonic memristor-based RC framework incorporating WDM technology. The core advantage of the memristor-based RC system lies in its capacity to emulate the time-dependent characteristics of neurons by utilizing the exponential decay response of the GSSe dynamic memristors (as illustrated in the inset of Fig. 5b), thereby effectively recovering the temporal information that may be lost during the binarization process, and subsequently feeding it into convolutional neural network (CNN) for classification and recognition (as illustrated in Fig. 5b). The performance comparison analysis present in Fig. 5d indicates that the hybrid system achieves a recognition accuracy of 73.0% on the DeepShip dataset, which is less than 5% lower than the original data accuracy of 77.7% classified directly by the CNN model, while significantly increasing the transmission efficiency by a factor of 32 within the same time unit. More importantly, compared with the accuracy of 57.4% achieved by directly processing binary data, the hybrid RC module contributes to a notable improvement of 15.6%. Furthermore, as evidenced by the confusion matrices presented in Fig. 5e, the hybrid system exhibits a strong capacity for classification balance when applied to imbalanced datasets, thereby underscoring the robustness of the proposed approach.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eIn conclusion, we successfully demonstrated the functionality of dynamic and non-volatile memristors on the ChGs integrated photonic platform by the ion-doping mechanism and effect for the first time. Meanwhile, we have disclosed the underlying ion migration reconfigurable working mechanism of the dynamic memristor via systematic Raman spectroscopy analysis and in-situ TEM characterization. This finding opens up a new developmental avenue for the development of diverse neuronal devices capable of operating across various temporal response scales within photonic neural networks. Non-volatile memristor is achieved via photoinduced Ag dissolution doping process, which modified the molecular chain structure and bandgap of chalcogenide glass. This approach enabled a compact MRR phase shifter with an L\u003csub\u003e\u0026#120587;\u003c/sub\u003e of 26 \u0026micro;m, as well as a multi-level tunable attenuator featuring 5-bit. By applying voltage to precisely control the nucleation process of silver nanoclusters in the GSSe waveguide, a reconfigurable dynamic memristor with a switching ratio of approximately 25 dB was realized. Meanwhile, its short-term memory and nonlinear characteristics were validated through dynamic response measurement, demonstrating its capability to generate well-separable responses to distinct temporal inputs. Furthermore, we proposed a photonic RC architecture based on a ion-doped GSSe waveguide-integrated memristor platform and applied it to underwater acoustic recognition tasks with temporal characteristics. Through data binarization, the data transmission rate per unit time was significantly enhanced. Subsequently, the compressed data was decoded and reconstructed using the GSSe dynamic memristor, ultimately achieving an accuracy of 73.0%, with a reduction in recognition accuracy of less than 5% compared to the original data, which provides substantial support for the application of waveguide-integrated memristors in broader and lightweight ANN architectures, thereby enhancing the brain-like functionalities of neural network systems.\u003c/p\u003e"},{"header":"Methods","content":"\u003cp\u003e\u003cstrong\u003eDevice fabrication\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe 500 nm Ge\u003csub\u003e28\u003c/sub\u003eSb\u003csub\u003e12\u003c/sub\u003eSe\u003csub\u003e60\u003c/sub\u003e film was deposited on a Si substrate with a 2 \u0026mu;m SiO\u003csub\u003e2\u003c/sub\u003e layer via thermal evaporation under a background vacuum of 5 \u0026times; 10\u003csup\u003e-4\u003c/sup\u003e Pa. The wafer surface is maintained in pristine condition through oxygen plasma cleaning before deposition, and the deposition rate is controlled at 15 \u0026Aring;/s. The hybrid integrated GSSe/Ag MRRs were patterned using 50 keV electron beam lithography (EBL) equipment (Raith Voyager). The photoresist used was the AllResist AR-P 6200.13 series, spun at 4000 rpm to achieve a thickness of approximately 400 nm. Subsequently, the pattern is transferred to the GSSe layer using inductively coupled plasma (ICP) dry etching with an Ar/CHF\u003csub\u003e3\u003c/sub\u003e gas chemistry, and the residual photoresist on top of the waveguide is removed via oxygen plasma etching. For the fabrication of hybrid integrated GSSe/Ag waveguide attenuator and dynamic reconfigurable devices, we utilized the stepper lithography machine (Canon 3030EX6) equipped with a 248 nm deep ultraviolet (DUV) light source to achieve wafer-level patterning of the waveguide, and the subsequent dry etching process followed the same procedure as described previously. After this, the photodoped regions were patterned via EBL, and a 30 nm layer of Ag was deposited on the GSSe waveguide via magnetron sputtering, followed by a lift-off process. The lift-off process utilized AR-P 6200.13 photoresist, spun-coated at 2000 rpm to achieve a thickness of approximately 600-700 nm. For the fabrication of the electrically reconfigurable devices, the negative photoresist NR9-1500PY and the Aligner photolithography equipment (Karl Suss MA6-BSA) were employed to define the electrode pad patterns. Subsequently, a metal layer of Ti (5 nm)/Au (100 nm) was deposited on the chip via magnetron sputtering. Last, the metal outside the defined pattern was removed through a lift-off process. The doping window of the electrically reconfigurable device was also determined by NR9-1500PY. Through exposure and development processes, the waveguide regions that require doping are exposed to air, while the regions that do not require doping are protected by a layer of photoresist.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDoing process\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eIn this study, the photoinduced Ag dissolution doping process utilized a 45W UV lamp with a wavelength of 365 nm as the light source. The immersion doping process utilized a standardized 0.1 mol/L AgNO\u003csub\u003e3\u003c/sub\u003e solution, achieving thermal diffusion doping through water bath treatment at 40 \u0026deg;C. The immersion doping process has lower lithographic resolution requirements and higher reliability compared to depositing metal Ag on the waveguide surface followed by a lift-off process, making it well-suited for large-scale on-chip waveguide doping modification.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eData Availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll the data supporting this study are available in the paper and Supplementary Information. Additional data related to this paper are available from the corresponding authors upon request.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgments\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis work was supported by the National Key Research and Development Program of China (2024YFB2808700), \u0026quot;Pioneer\u0026quot; R\u0026amp;D Program of Zhejiang (2025C01002), and the Key Project of Westlake Institute for Optoelectronics (2024GD002). The authors would like to acknowledge the fabrication support from the ZJU Nano-Fabrication Center at Zhejiang University, the Westlake Center for Micro/Nano Fabrication and Instrumentation at Westlake University, and the Nano-Fabrication Center at Zhejiang Lab. The authors would like to acknowledge Dr. Yangjian Lin and Dr. Pei Sheng from the Instrumentation and Service Center for Physical Sciences at Westlake University for their support and assistance in the in-situ TEM experiments.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eH.L. conceived the idea. K.L. carried out the fabrication, measurement setup construction, and device testing. Z.W. contributed to the implementation and training of the neural networks. K.X., S.L., K.B., and B.S. assisted in the deposition of the GSSe thin films and device testing. R.L., H.M., J.W., and C.S. assisted in the design and layout of the device. J.J. and Q.D. performed the preparation of the metal electrodes. W.Z., C.L., and S.D. provided the raw material chalcogenide glass and offered support during the waveguide fabrication process. H.L. and L.L. supervised the research. All the authors contributed to the technical discussions and writing of the paper.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare no competing financial interests.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eZhong, Y.\u003cem\u003e et al.\u003c/em\u003e Dynamic memristor-based reservoir computing for high-efficiency temporal signal processing. \u003cem\u003eNature Communications\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, doi:10.1038/s41467-020-20692-1 (2021).\u003c/li\u003e\n\u003cli\u003eMoon, J.\u003cem\u003e et al.\u003c/em\u003e Temporal data classification and forecasting using a memristor-based reservoir computing system. \u003cem\u003eNature Electronics\u003c/em\u003e \u003cstrong\u003e2\u003c/strong\u003e, 480-487, doi:10.1038/s41928-019-0313-3 (2019).\u003c/li\u003e\n\u003cli\u003eChen, Z. W.\u003cem\u003e et al.\u003c/em\u003e All-ferroelectric implementation of reservoir computing. \u003cem\u003eNature Communications\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, doi:10.1038/s41467-023-39371-y (2023).\u003c/li\u003e\n\u003cli\u003eYoungblood, N., R\u0026iacute;os Ocampo, C. A., Pernice, W. H. P. \u0026amp; Bhaskaran, H. Integrated optical memristors. \u003cem\u003eNature Photonics\u003c/em\u003e \u003cstrong\u003e17\u003c/strong\u003e, 561-572, doi:10.1038/s41566-023-01217-w (2023).\u003c/li\u003e\n\u003cli\u003eFarmakidis, N., Dong, B. \u0026amp; Bhaskaran, H. Integrated photonic neuromorphic computing: opportunities and challenges. \u003cem\u003eNature Reviews Electrical Engineering\u003c/em\u003e \u003cstrong\u003e1\u003c/strong\u003e, 358-373, doi:10.1038/s44287-024-00050-9 (2024).\u003c/li\u003e\n\u003cli\u003eShastri, B. J.\u003cem\u003e et al.\u003c/em\u003e Photonics for artificial intelligence and neuromorphic computing. \u003cem\u003eNature Photonics\u003c/em\u003e \u003cstrong\u003e15\u003c/strong\u003e, 102-114, doi:10.1038/s41566-020-00754-y (2021).\u003c/li\u003e\n\u003cli\u003eTossoun, B.\u003cem\u003e et al.\u003c/em\u003e High-speed and energy-efficient non-volatile silicon photonic memory based on heterogeneously integrated memresonator. \u003cem\u003eNature Communications\u003c/em\u003e \u003cstrong\u003e15\u003c/strong\u003e, doi:10.1038/s41467-024-44773-7 (2024).\u003c/li\u003e\n\u003cli\u003eCheung, S.\u003cem\u003e et al.\u003c/em\u003e Ultra‐Power‐Efficient, Electrically Programmable, Multi‐State Photonic Flash Memory on a Heterogeneous III‐V/Si Platform. \u003cem\u003eLaser \u0026amp; Photonics Reviews\u003c/em\u003e \u003cstrong\u003e18\u003c/strong\u003e, doi:10.1002/lpor.202400001 (2024).\u003c/li\u003e\n\u003cli\u003eZhang, G.\u003cem\u003e et al.\u003c/em\u003e Thin film ferroelectric photonic-electronic memory. \u003cem\u003eLight: Science \u0026amp; Applications\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, doi:10.1038/s41377-024-01555-6 (2024).\u003c/li\u003e\n\u003cli\u003eTaki, K.\u003cem\u003e et al.\u003c/em\u003e Nonvolatile optical phase shift in ferroelectric hafnium zirconium oxide. \u003cem\u003eNature Communications\u003c/em\u003e \u003cstrong\u003e15\u003c/strong\u003e, doi:10.1038/s41467-024-47893-2 (2024).\u003c/li\u003e\n\u003cli\u003eWei, M.\u003cem\u003e et al.\u003c/em\u003e Monolithic back-end-of-line integration of phase change materials into foundry-manufactured silicon photonics. \u003cem\u003eNature Communications\u003c/em\u003e \u003cstrong\u003e15\u003c/strong\u003e, 2786 (2024).\u003c/li\u003e\n\u003cli\u003eWei, M.\u003cem\u003e et al.\u003c/em\u003e Electrically programmable phase-change photonic memory for optical neural networks with nanoseconds in situ training capability. \u003cem\u003eAdvanced Photonics\u003c/em\u003e \u003cstrong\u003e5\u003c/strong\u003e, doi:10.1117/1.Ap.5.4.046004 (2023).\u003c/li\u003e\n\u003cli\u003eLei, K.\u003cem\u003e et al.\u003c/em\u003e Magnetron-sputtered and thermal-evaporated low-loss Sb-Se phase-change films in non-volatile integrated photonics. \u003cem\u003eOptical Materials Express\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, doi:10.1364/ome.462426 (2022).\u003c/li\u003e\n\u003cli\u003eSeok, T. J., Quack, N., Han, S., Muller, R. S. \u0026amp; Wu, M. C. Large-scale broadband digital silicon photonic switches with vertical adiabatic couplers. \u003cem\u003eOptica\u003c/em\u003e \u003cstrong\u003e3\u003c/strong\u003e, doi:10.1364/optica.3.000064 (2016).\u003c/li\u003e\n\u003cli\u003eEdinger, P.\u003cem\u003e et al.\u003c/em\u003e Silicon photonic microelectromechanical phase shifters for scalable programmable photonics. \u003cem\u003eOpt. Lett.\u003c/em\u003e \u003cstrong\u003e46\u003c/strong\u003e, doi:10.1364/ol.436288 (2021).\u003c/li\u003e\n\u003cli\u003eNie, Y. K.\u003cem\u003e et al.\u003c/em\u003e Integrated laser graded neuron enabling high-speed reservoir computing without a feedback loop. \u003cem\u003eOptica\u003c/em\u003e \u003cstrong\u003e11\u003c/strong\u003e, 1690-1699, doi:10.1364/OPTICA.537231 (2024).\u003c/li\u003e\n\u003cli\u003eWaser, R., Dittmann, R., Staikov, G. \u0026amp; Szot, K. Redox‐Based Resistive Switching Memories \u0026ndash; Nanoionic Mechanisms, Prospects, and Challenges. \u003cem\u003eAdvanced Materials\u003c/em\u003e \u003cstrong\u003e21\u003c/strong\u003e, 2632-2663, doi:10.1002/adma.200900375 (2009).\u003c/li\u003e\n\u003cli\u003eZhang, G.\u003cem\u003e et al.\u003c/em\u003e Functional Materials for Memristor‐Based Reservoir Computing: Dynamics and Applications. \u003cem\u003eAdvanced Functional Materials\u003c/em\u003e \u003cstrong\u003e33\u003c/strong\u003e, doi:10.1002/adfm.202302929 (2023).\u003c/li\u003e\n\u003cli\u003eWang, R.\u003cem\u003e et al.\u003c/em\u003e Recent Advances of Volatile Memristors: Devices, Mechanisms, and Applications. \u003cem\u003eAdvanced Intelligent Systems\u003c/em\u003e \u003cstrong\u003e2\u003c/strong\u003e, doi:10.1002/aisy.202000055 (2020).\u003c/li\u003e\n\u003cli\u003eKim, H. J.\u003cem\u003e et al.\u003c/em\u003e Super-Linear-Threshold-Switching Selector with Multiple Jar-Shaped Cu-Filaments in the Amorphous Ge3Se7 Resistive Switching Layer in a Cross-Point Synaptic Memristor Array. \u003cem\u003eAdvanced Materials\u003c/em\u003e \u003cstrong\u003e34\u003c/strong\u003e, doi:10.1002/adma.202203643 (2022).\u003c/li\u003e\n\u003cli\u003eChen, Z. \u0026amp; Zhi, C. Chalcogens for high-energy batteries. \u003cem\u003eNature Reviews Materials\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, 268-284, doi:10.1038/s41578-025-00773-7 (2025).\u003c/li\u003e\n\u003cli\u003eWang, Z.\u003cem\u003e et al.\u003c/em\u003e Doping effects of metal cation on sulfide solid electrolyte/lithium metal interface. \u003cem\u003eNano Energy\u003c/em\u003e \u003cstrong\u003e84\u003c/strong\u003e, doi:10.1016/j.nanoen.2021.105906 (2021).\u003c/li\u003e\n\u003cli\u003eWang, Z.\u003cem\u003e et al.\u003c/em\u003e Threshold Switching of Ag or Cu in Dielectrics: Materials, Mechanism, and Applications. \u003cem\u003eAdvanced Functional Materials\u003c/em\u003e \u003cstrong\u003e28\u003c/strong\u003e, doi:10.1002/adfm.201704862 (2017).\u003c/li\u003e\n\u003cli\u003eSarwat, S. G., Moraitis, T., Wright, C. D. \u0026amp; Bhaskaran, H. Chalcogenide optomemristors for multi-factor neuromorphic computation. \u003cem\u003eNat Commun\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 2247, doi:10.1038/s41467-022-29870-9 (2022).\u003c/li\u003e\n\u003cli\u003eElfarash, A. H. \u0026amp; Gholipour, B. Reconfigurable nanoionic and photoionic material and device platforms. \u003cem\u003eAdvances in Physics: X\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, doi:10.1080/23746149.2024.2338285 (2024).\u003c/li\u003e\n\u003cli\u003eMcRae, L., Xie, Y. \u0026amp; Gholipour, B. Photoionic Driven Movement of Metallic Ions as a Nonvolatile Reconfiguration Mechanism in Amorphous Chalcogenide Metasurfaces. \u003cem\u003eAdvanced Optical Materials\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, doi:10.1002/adom.202101046 (2021).\u003c/li\u003e\n\u003cli\u003eFrumar, M. \u0026amp; Wagner, T. Ag doped chalcogenide glasses and their applications. \u003cem\u003eCurrent Opinion in Solid State and Materials Science\u003c/em\u003e \u003cstrong\u003e7\u003c/strong\u003e, 117-126, doi:10.1016/s1359-0286(03)00044-5 (2003).\u003c/li\u003e\n\u003cli\u003eWang, Z.\u003cem\u003e et al.\u003c/em\u003e Memristors with diffusive dynamics as synaptic emulators for neuromorphic computing. \u003cem\u003eNature Materials\u003c/em\u003e \u003cstrong\u003e16\u003c/strong\u003e, 101-108, doi:10.1038/nmat4756 (2016).\u003c/li\u003e\n\u003cli\u003eDeepak, P.\u003cem\u003e et al.\u003c/em\u003e Influence of Ag Doping on Physico-Chemical Properties of the Ge28Sb12Se60 Chalcogenide Glassy Matrix. \u003cem\u003eJournal of Materials Science and Engineering A\u003c/em\u003e \u003cstrong\u003e6\u003c/strong\u003e, doi:10.17265/2161-6213/2016.5-6.002 (2016).\u003c/li\u003e\n\u003cli\u003eSvoboda, R., M\u0026aacute;lek, J. \u0026amp; Liska, M. Correlation between the structure and structural relaxation data for (GeSe2)y(Sb2Se3)1-y glasses. \u003cem\u003eJournal of Non-Crystalline Solids\u003c/em\u003e \u003cstrong\u003e505\u003c/strong\u003e, 162-169, doi:10.1016/j.jnoncrysol.2018.11.013 (2019).\u003c/li\u003e\n\u003cli\u003eKochubei, H. K.\u003cem\u003e et al.\u003c/em\u003e Raman scattering and X-ray diffraction studies of Ge16Sb24Se60 alloys. \u003cem\u003eSemiconductor Physics Quantum Electronics \u0026amp; Optoelectronics\u003c/em\u003e \u003cstrong\u003e28\u003c/strong\u003e, doi:10.15407/spqeo28.03.300 (2025).\u003c/li\u003e\n\u003cli\u003eTsuchiya, T., Nakayama, T. \u0026amp; Ariga, K. Nanoarchitectonics Intelligence with atomic switch and neuromorphic network system. \u003cem\u003eApplied Physics Express\u003c/em\u003e \u003cstrong\u003e15\u003c/strong\u003e, 100101, doi:10.35848/1882-0786/ac926b (2022).\u003c/li\u003e\n\u003cli\u003eTerabe, K., Hasegawa, T., Nakayama, T. \u0026amp; Aono, M. Quantized conductance atomic switch. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e433\u003c/strong\u003e, 47-50, doi:10.1038/nature03190 (2005).\u003c/li\u003e\n\u003cli\u003eViswanathan, A., Udayan, S., Musfir, P. N., Nampoori, V. P. N. \u0026amp; Thomas, S. Enhancement of defect states assisted thermal diffusivity in solution-processed GeSeSb chalcogenide glass matrix on silver incorporation. \u003cem\u003eJournal of Non-Crystalline Solids\u003c/em\u003e \u003cstrong\u003e503-504\u003c/strong\u003e, 151-157, doi:https://doi.org/10.1016/j.jnoncrysol.2018.09.039 (2019).\u003c/li\u003e\n\u003cli\u003eKolobov, A. V. \u003cem\u003ePhoto-induced metastability in amorphous semiconductors\u003c/em\u003e. (John Wiley \u0026amp; Sons, 2006).\u003c/li\u003e\n\u003cli\u003eWagner, T., Dale, G., Ewen, P., Owen, A. \u0026amp; Perina, V. J. J. o. A. P. Kinetics of the thermally and photoinduced solid state reaction of Ag with As 33 S 67 films. \u003cstrong\u003e87\u003c/strong\u003e, 7758-7767 (2000).\u003c/li\u003e\n\u003cli\u003eWagner, T.\u003cem\u003e et al.\u003c/em\u003e The kinetics of the photo-induced solid-state chemical reaction in Ag/As33S67 bilayers and its reaction products. \u003cstrong\u003e79\u003c/strong\u003e, 223-237 (1999).\u003c/li\u003e\n\u003cli\u003eXiong, Y.\u003cem\u003e et al.\u003c/em\u003e Observation of the failure mechanism in Ag10Ge15Te75-based memristor induced by ion transport. \u003cem\u003eNano Research\u003c/em\u003e \u003cstrong\u003e17\u003c/strong\u003e, 8431-8437, doi:10.1007/s12274-024-6791-2 (2024).\u003c/li\u003e\n\u003cli\u003eChekol, S. A., Menzel, S., Ahmad, R. W., Waser, R. \u0026amp; Hoffmann‐Eifert, S. Effect of the Threshold Kinetics on the Filament Relaxation Behavior of Ag‐Based Diffusive Memristors. \u003cem\u003eAdvanced Functional Materials\u003c/em\u003e \u003cstrong\u003e32\u003c/strong\u003e, doi:10.1002/adfm.202111242 (2021).\u003c/li\u003e\n\u003cli\u003eYang, J. T.\u003cem\u003e et al.\u003c/em\u003e Artificial Synapses Emulated by an Electrolyte‐Gated Tungsten‐Oxide Transistor. \u003cem\u003eAdvanced Materials\u003c/em\u003e \u003cstrong\u003e30\u003c/strong\u003e, doi:10.1002/adma.201801548 (2018).\u003c/li\u003e\n\u003cli\u003eNie, F.\u003cem\u003e et al.\u003c/em\u003e An Adaptive Solid‐State Synapse with Bi‐Directional Relaxation for Multimodal Recognition and Spatio‐Temporal Learning. \u003cem\u003eAdvanced Materials\u003c/em\u003e \u003cstrong\u003e37\u003c/strong\u003e, doi:10.1002/adma.202412006 (2025).\u003c/li\u003e\n\u003cli\u003eIrfan, M.\u003cem\u003e et al.\u003c/em\u003e DeepShip: An underwater acoustic benchmark dataset and a separable convolution based autoencoder for classification. \u003cem\u003eExpert Systems with Applications\u003c/em\u003e \u003cstrong\u003e183\u003c/strong\u003e, doi:10.1016/j.eswa.2021.115270 (2021).\u003c/li\u003e\n\u003cli\u003eAlqurashi, F. S., Trichili, A., Saeed, N., Ooi, B. S. \u0026amp; Alouini, M.-S. Maritime Communications: A Survey on Enabling Technologies, Opportunities, and Challenges. \u003cem\u003eIEEE Internet of Things Journal\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, 3525-3547, doi:10.1109/jiot.2022.3219674 (2023).\u003c/li\u003e\n\u003cli\u003eWei, T.\u003cem\u003e et al.\u003c/em\u003e Hybrid Satellite-Terrestrial Communication Networks for the Maritime Internet of Things: Key Technologies, Opportunities, and Challenges. \u003cem\u003eIEEE Internet of Things Journal\u003c/em\u003e \u003cstrong\u003e8\u003c/strong\u003e, 8910-8934, doi:10.1109/jiot.2021.3056091 (2021).\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":false,"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":"nature-portfolio","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"","title":"Nature Portfolio","twitterHandle":"","acdcEnabled":false,"dfaEnabled":false,"editorialSystem":"ejp","reportingPortfolio":"","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-7782397/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7782397/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003ePhotonic memristors are anticipated to emerge as a novel hardware platform for neural network computing systems owing to their broad bandwidth communication capabilities and potential compatibility with non-von Neumann neuromorphic computing architectures. However, current photonic memristors are limited to a monotonic modulation mechanism and most demonstrate only a non-volatile temporal response scale, which restricts their applicability across more diverse neural network computing architectures. Here, we present a novel strategy that employs ion-doped chalcogenide glass, combined with a multi-dimensional modulation mechanism of optical and electrical fields, to realize the monolithic integration of non-volatile waveguide-integrated memristors featuring multi-level storage capacity and compact footprint, along with volatile reconfigurable memristors exhibiting high extinction ratios and excellent short-term plasticity. Furthermore, leveraging the powerful nonlinear dynamic response of the reconfigurable memristor, we developed an on-chip photonic reservoir computing system that operates without a feedback loop. This work provides a novel developmental approach for the development of neuron devices with varying time response scales and offers substantial support for neural networks in more accurately simulating brain functions.\u003c/p\u003e","manuscriptTitle":"Ionic Driven Waveguide Integrated Memristor","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-01-21 06:53:32","doi":"10.21203/rs.3.rs-7782397/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":"2e677c6c-eda6-4822-8bfe-6699dab95fc5","owner":[],"postedDate":"January 21st, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[{"id":61228311,"name":"Physical sciences/Optics and photonics/Applied optics/Optoelectronic devices and components"},{"id":61228312,"name":"Physical sciences/Optics and photonics/Optical materials and structures"}],"tags":[],"updatedAt":"2026-03-11T13:20:14+00:00","versionOfRecord":[],"versionCreatedAt":"2026-01-21 06:53:32","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7782397","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7782397","identity":"rs-7782397","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}