Two-dimensional platinum (II)-alkyne organic framework, preparation method thereof and application of two-dimensional platinum (II)-alkyne organic framework in memristor

Abstract

According to the two-dimensional platinum (II)-acetylene organic framework material and the preparation method and application thereof, due to the fact that the two-dimensional platinum (II)-acetylene organic framework material is provided with a molecular potential well and a double donor-acceptor structure at the same time, the two-dimensional platinum (II)-acetylene organic framework material has excellent photoelectric performance, and it is proved that the two-dimensional platinum (II)-acetylene organic framework material has huge application potential in the field of data storage and calculation. According to the invention, the material is applied to a photon memristor, and excellent light modulation multi-mode non-volatile resistance change characteristic is realized. The RC system in the sensor based on the device can adapt to multi-task processing, the accuracy rate of static number recognition exceeds 90.775%, the accuracy rate of dynamic letter recognition exceeds 89.6%, and the NRMSE of waveform classification is 0.1. According to the method, the theoretical basis of the photon memristor on the molecular level is supplemented, the bottleneck of separation of sensing, storage and calculation units in an existing calculation system is overcome, and a promising material-algorithm collaborative strategy is provided for designing RC in an efficient sensor for edge calculation.

Description

Technical Field

[0001] This invention belongs to the field of storage and computing materials and devices, specifically relating to a two-dimensional platinum(II)-yne organic framework, its preparation method, and its application in memristors. Background Technology

[0002] In recent years, the latest advancements in bio-inspired hardware design have focused on improving computational efficiency through in-sensor computing, integrating sensing and computational functions into artificial neural network (ANN) architectures. At the algorithmic level, reservoir computing (RC) has attracted significant attention due to its unique advantages in reducing the complexity of temporal data processing. By leveraging the nonlinear response characteristics of the "reservoir" to input signals, RC exhibits extremely high computational efficiency. However, current RC implementations are still limited by electrical signal transmission, requiring complex interface circuits to connect the sensor and the computing unit. This design not only leads to increased latency, power consumption, and architectural complexity but also diminishes the inherent advantages of in-sensor computing.

[0003] Memristors with photonic sensing capabilities can serve as RC elements within sensors in machine vision systems, eliminating redundant data conversion processes and fully leveraging the inherent advantages of the retina. However, significant technical bottlenecks remain at the device level in developing photonic memristors capable of simultaneously sensing, filtering, storing, and processing optical signals—these devices need to maintain high responsiveness to illumination and be resistant to rapid nonlinear decay. Therefore, there is an urgent need to explore a filter-free photonic memristor that can operate as a dynamic storage repository, possessing advantages such as high expressive power under ultraviolet light modulation, nonlinear response characteristics, short-term memory (STM), polymorphism, and a large dynamic range. Another challenge is that current research focuses primarily on device-level mechanism verification, while the molecular-level mechanism of the active medium in photonic memristors remains unclear, hindering its further development in machine vision systems.

[0004] A material-algorithm collaborative strategy holds promise for overcoming current limitations and unleashing the potential of neuromorphic computing in RC systems within biomimetic sensors. In terms of material design, to mimic photoreceptors and peripheral optic nerves in the retina, the active layer material of photonic memristors must simultaneously possess excellent photosensitivity and multi-level response characteristics to external light stimuli. This ensures that photonic devices can achieve real-time sensing of key feature information, instantaneous photoelectric conversion, and efficient data processing. Theoretically, the inherent electronic structure of molecules—including energy level distribution, redox sites, oxygen vacancies, and deep and shallow trapping sites—affects exciton separation efficiency and photoinduced carrier recombination behavior, and regulates electron transfer direction and trapping sites, thereby achieving multiple dynamic responses in photonic devices. However, traditional organic polymers are limited by anisotropy and inhomogeneity, and conventional inorganic materials are constrained by uncontrollable modification; both have shortcomings in exploring the underlying mechanisms. Summary of the Invention

[0005] To overcome the aforementioned existing problems, one objective of this invention is to provide a two-dimensional platinum(II)-yne organic framework. A second objective is to provide a method for preparing the aforementioned two-dimensional platinum(II)-yne organic framework. A third objective is to provide applications of the aforementioned two-dimensional platinum(II)-yne organic framework. A fourth objective is to provide a photonic memristor. A fifth objective is to provide a method for preparing the aforementioned memristor. A sixth objective is to provide an in-sensor storage computing system. A seventh objective is to provide applications of the aforementioned in-sensor storage computing system.

[0006] Metal-acetylene frameworks (MAFs) are a type of framework consisting of an organic ligand and a single metal atom linked by a metal di(acetyl) group. C≡C M C≡C A novel two-dimensional highly conjugated metal-organic framework (MOF) was developed. Research shows that by controlling the type of organic ligands, metal centers, heteroatoms, and defects, the distribution of electrons and holes within the two-dimensional plane can be precisely optimized. This provides a solid foundation for constructing tunable energy levels and multiple trapping sites at the molecular level. Furthermore, C≡C M C≡C The metal-carbon σ bonds in the structure allow for large-area overlap between the transition metal d orbitals and the alkyne unit p orbitals. This coupling effect not only significantly enhances the material's stability but also, through ingenious chemical bonding properties, modulates electronic delocalization characteristics, thereby improving carrier mobility and concentration. Notably, the two-dimensional topology complements the cross-structure of the memristor, with the layered structure of the two-dimensional active medium perfectly mimicking the layered tissue structure of the human retina. This material, possessing controllable properties at both the molecular and atomic levels, offers unique advantages for systematically studying the optical characteristics and operational mechanisms of photonic memristors in RC systems within sensors.

[0007] This invention proposes a flexible, two-terminal photonic memristor based on a novel two-dimensional platinum(II)-acetylene organic framework (Pt-AFs). Through a synergistic molecular design strategy involving the construction of a molecular potential well and a dual donor-acceptor (D–A) structure, the fabricated photonic memristor exhibits not only excellent multimodal non-volatile resistive properties but also good durability and retention. This characteristic can be attributed to the synergistic mechanism of space charge-restricted conduction (SCLC) and photoenhanced bonded charge transport (TBCT). This photonic memristor successfully simulates various biological synaptic behaviors, including excitatory postsynaptic currents (EPSC), inhibitory postsynaptic currents (IPSC), double-pulse enhancement (PPF), short-term plasticity (STP), and long-term plasticity (LTP). Its rich tunability, spatiotemporal correlation, and attenuation memory characteristics under illumination modulation enable this device to initially simulate the image memory, forgetting, and preprocessing functions of the human retina. The static sensor-based RC system using photonic memristors achieves a 90.775% accuracy rate in recognizing handwritten digits, while the dynamic sensor-based RC system achieves an 89.6% accuracy rate in recognizing dynamic letters. Through masking techniques, the RC system is progressively improved, enabling precise classification of sine waves and square waves into their corresponding categories. By selecting an appropriate mask length, the system achieves an excellent performance with a normalized root mean square error (NRMSE) of only 0.1. This research fully demonstrates the rich characteristics and application potential of photonic memristors in sensor-based RC systems, providing an efficient and economical solution for temporal and spatial signal processing in the field of edge computing.

[0008] To achieve the above objectives, the technical solution adopted by the present invention is as follows: The first aspect of this invention provides a two-dimensional platinum(II)-yne organic framework, the chemical structural formula of which is shown in formula (I): Equation (Ⅰ); Ar is either phenyl or aziridine.

[0009] Preferably, the azinyl group is a triazinyl group.

[0010] More preferably, the chemical structural formula of the two-dimensional platinum(II)-yne organic framework is selected from: ; .

[0011] A second aspect of the present invention provides a method for preparing the two-dimensional platinum(II)-yne organic framework described in the first aspect, comprising the following steps: In the presence of a catalyst, ligands and transThe two-dimensional platinum(II)-yne organic framework was prepared by reacting -[Pt(PBu3)2Cl2].

[0012] Preferably, the catalyst includes at least one of cuprous iodide, cuprous bromide, and cuprous chloride.

[0013] Preferably, the trans The structural formula of -[Pt(PBu3)2Cl2] is: .

[0014] More preferably, the trans The preparation method of -[Pt(PBu3)2Cl2] includes the following steps: reacting potassium tetrachloroplatinate with tributylphosphine to obtain the [Pt(PBu3)2Cl2] trans -[Pt(PBu3)2Cl2].

[0015] More preferably, the method further includes the following step: the product after the reaction contains trans -[Pt(PBu3)2Cl2] and cis -[Pt(PBu3)2Cl2], the product was heated at 170–220 °C for 1–3 hours to carry out a configuration inversion reaction, yielding trans -[Pt(PBu3)2Cl2].

[0016] More preferably, the molar ratio of potassium tetrachloroplatinate to tributylphosphine is 1:(2–5).

[0017] Preferably, the reaction is carried out in the presence of an alkaline agent.

[0018] More preferably, the alkali agent includes at least one of triethylamine, diisopropylamine, diisopropylethylamine, tetramethylethylenediamine, and pyridine.

[0019] Preferably, the reaction is carried out under an inert atmosphere.

[0020] More preferably, the inert atmosphere is nitrogen.

[0021] Preferably, the reaction is carried out in an organic solvent.

[0022] More preferably, the organic solvent includes at least one of tetrahydrofuran (THF), N,N-dimethylformamide (DMF), N,N-dimethylacetamide, dimethyl sulfoxide, and acetonitrile.

[0023] Preferably, the ligand and trans The molar ratio of -[Pt(PBu3)2Cl2] is 1:(1–2).

[0024] More preferably, the ligand and transThe molar ratio of -[Pt(PBu3)2Cl2] is 1:(1.3–1.7).

[0025] Preferably, the molar ratio of the ligand to the catalyst is 1:(0.1–0.5).

[0026] More preferably, the molar ratio of the ligand to the catalyst is 1:(0.2–0.4).

[0027] Preferably, the ligand is selected from 1,3,5-tris(4'-ethynylphenyl)benzene (TEPB) or 2,4,6-tris(4-ethynylphenyl)-1,3,5-triazine (TEPT).

[0028] Preferably, the reaction temperature is 30–60 °C.

[0029] More preferably, the reaction temperature is 35–50 °C.

[0030] Preferably, the reaction time is 60–85 h.

[0031] More preferably, the reaction time is 65–80 h.

[0032] Preferably, the process further includes the following steps: the crude product after the reaction is washed with dichloromethane, purified and dried to obtain the two-dimensional platinum(II)-yne organic framework.

[0033] The third aspect of the present invention provides the application of the two-dimensional platinum(II)-yne organic framework described in the first aspect in memristors.

[0034] A fourth aspect of the present invention provides a memristor, the structure of which includes a substrate, a bottom electrode layer, a dielectric layer and a top electrode layer arranged sequentially; the material of the dielectric layer includes the two-dimensional platinum(II)-yne organic framework described in the first aspect.

[0035] Preferably, the material of the dielectric layer further includes an organic polymer.

[0036] More preferably, the organic polymer includes polymethyl methacrylate (PMMA).

[0037] Preferably, the substrate is an insulating substrate, a semiconductor substrate, or a conductive substrate.

[0038] Preferably, the material of the bottom electrode layer is any one of metal, conductive oxide, conductive nitride, and conductive carbon material.

[0039] More preferably, the bottom electrode is made of tin oxide (ITO).

[0040] Preferably, the material of the top electrode layer includes at least one of gold, silver, copper, platinum, titanium, chromium, platinum, and indium.

[0041] Preferably, the thickness of the top electrode layer is 1–500 nm.

[0042] The fifth aspect of this invention provides a method for fabricating the memristor described in the fourth aspect, comprising the following steps: A bottom electrode layer, a dielectric layer, and a top electrode layer are sequentially formed on a substrate using a coating process; the coating process includes thermal evaporation, magnetron sputtering, sol-gel, chemical vapor deposition, or coating methods.

[0043] Preferably, the method for preparing the dielectric layer includes the following steps: dispersing a two-dimensional platinum(II)-yne organic framework in chlorobenzene, adding an organic polymer, coating the resulting dispersion onto the bottom electrode layer, and annealing to obtain the dielectric layer.

[0044] More preferably, the annealing temperature is 80–120 °C.

[0045] More preferably, the ratio of the two-dimensional platinum(II)-yne organic framework to chlorobenzene is (0.5–5) mg:1 mL.

[0046] More preferably, the ratio of the organic polymer to chlorobenzene is (30–50) mg:1 mL.

[0047] A sixth aspect of the present invention provides an in-sensor storage computing system, including the memristor described in the fifth aspect.

[0048] The seventh aspect of the present invention provides the application of the in-sensor storage computing system described in the sixth aspect in artificial synapses and neuromorphic computing.

[0049] The beneficial effects of this invention are: This invention proposes a two-dimensional platinum(II)-acetylene oxide framework, which can be used as an active layer to fabricate high-performance flexible photonic memristors. Multimodal non-volatile resistive properties are achieved by controlling the uniform distribution of molecular potential wells and the unique double D–A structure in the TEPB-Pt-AF thin film. Specifically, in the TEPB-Pt-AF material, the electron transfer rate of the D1→A transition in the double D–A sequence dominates after excitation. This reveals... dπ-pπAfter coupling, more electrons are distributed in the alkyne group, providing a pathway for multiple driving forces to be generated within the molecule, enabling efficient charge transport, rapid exciton dissociation, and slow charge recombination. These characteristics give TEPB-photonic memristors excellent photoelectric response performance, especially their analog switching and photoinduced properties, enabling them to simulate the real-time sensing, memory, and dynamic light signal preprocessing capabilities of the human retina. Particularly noteworthy are the time-dependent photoresponse, nonlinear decay memory effect, and stimulus accumulation effect, which allow for the construction of an in-sensor RC system that can map received spatiotemporal signals to a high-dimensional computational space without the need for additional sensors and processors. The TEPB-photonic memristor-based static sensing-storage-computing integrated in-sensor RC system achieves an accuracy of up to 90.8% in handwritten digit recognition tasks, fully demonstrating its powerful recognition capabilities.

[0050] This invention further applies the RC system within a dynamic sensor to a dynamic letter recognition task, verifying its ability to simulate human vision in processing dynamic spatiotemporal optical signals, achieving an accuracy of 89.6%. Furthermore, the parallel RC system within the sensor, incorporating a mask, achieves an NRMSE value of 0.1 in a waveform classification task. Based on the concept of designing rational memristor materials at the molecular level, this invention overcomes the bottlenecks of complex architecture and high energy consumption in traditional sensor computing systems, providing an innovative material-algorithm synergy strategy for designing efficient RC systems within sensors in the edge computing field. Attached Figure Description

[0051] Figure 1 The synthesis and structural characterization of two-dimensional platinum(II)-acetylene organic frameworks are presented; Figure a shows the synthetic schemes of TEPB-Pt-AF and TEPT-Pt-AF; Figure b shows the solid-state structure of TEPB-Pt-AF. 13 Figure c shows the high-resolution C1s XPS spectrum of TEPB-Pt-AF; Figure d shows the high-resolution Pt 4f XPS spectrum; Figure e shows the normalized Pt K-edge XANES spectrum of TEPB-Pt-AF; Figure f shows the corresponding spectrum obtained through k... 2 The weighted Fourier transform EXAFS spectrum is compared with the reference values ​​for Pt foil and PtO2; the g-plot represents the k-plot. 2 Figure h shows the Pt K-edge EXAFS (blue dots) and fitting curve (blue line) of TEPB-Pt-AF in weighted R space; Figure i shows the wavelet transform analysis of TEPB-Pt-AF EXAFS data; Figure j shows the HAADF STEM image of TEPB-Pt-AF with double spherical aberration correction, showing the single-atom Pt sites in TEPB-Pt-AF (marked with red circles); Figure j shows the HAADF-STEM image of TEPB-Pt-AF and the corresponding EDX mapping image (the EDX mappings of C, P, and Pt elements are represented by red, green, and blue, respectively). Figure 2 Solid state of TEPT-Pt-AF 13 C NMR spectrum; Figure 3 Figure 1 shows the FTIR spectra of two-dimensional platinum(II)-yne organic frameworks; Figure 2a shows the FTIR spectra of TEPB and TEPB-Pt-AF; Figure 3b shows the FTIR spectra of TEPT and TEPT-Pt-AF. Figure 4 For k 2 - Weighted EXAFS spectrum fitting analysis; where a is the Pt foil passing through k in R space. 2 Figure 1 shows the EXAFS spectrum after weighted Fourier transform; Figure 2 shows the oscillation spectrum of Pt foil in k-space after Fourier transform; Figures 3 and 4 show the correlation EXAFS spectra of PtO2; Figures 5 and 6 show the correlation EXAFS spectra of TEPB-Pt-AF. Figure 5 SEM images of TEPB-Pt-AF and TEPT-Pt-AF; where images a and b are SEM images of TEPB-Pt-AF at different magnifications; and images c and d are SEM images of TEPT-Pt-AF at different magnifications. Figure 6 The images are transmission electron microscopy (TEM) images of two-dimensional platinum(II)-yne organic frameworks; images a and b are low-resolution TEM images of TEPB-Pt-AF and TEPT-Pt-AF, respectively; images c and d are high-resolution TEM images of TEPB-Pt-AF and TEPT-Pt-AF, respectively. Figure 7 Thermogravimetric (TG) curves of TEPB-Pt-AF and TEPT-Pt-AF are shown. Figure 8 This study investigates the electrical properties of a photonic memristor based on a platinum(II)-yne organic framework (Pt-AFs). Figure a shows a schematic diagram of the memristor (ITO / Pt-AFs@PMMA / Pt) architecture and its corresponding SEM cross-sectional image, where the bottom electrode is ITO, the active layer is Pt-AFs@PMMA, and the top electrode is Pt. Figure b shows a schematic diagram of the flexible memristor structure. Figure c shows a schematic diagram of the mechanism by which the TEPB photonic memristor achieves ordered charge trapping through potential well formation during the SET process. Figure d shows the bipolar IV characteristic curves of the TEPB photonic memristor (voltage scan: 0→). 4 V→0→4 V→0, limiting current I CC = 1 mA); Figure e shows the low voltage (0→ (2 V→0→2 V→0) Multimode analog switching characteristics of the TEPT photonic memristor; Figure f shows the schematic diagram of the disordered charge trapping mechanism of the TEPT photonic memristor without a potential well; Figure g shows the bipolar IV characteristic curve of the TEPT photonic memristor (voltage scan: 0→-8 V→0→8 V→0); Figure h shows the non-analog switching characteristics of the TEPT photonic memristor (0→ Figure i shows the EPSC response of the TEPB photonic memristor to different voltage pulse amplitudes under a fixed pulse width (10 ms); Figure j shows the relationship between the electric field-induced PPF exponent and the two-voltage pulse interval (inset: experimental data); Figure k shows the long-term enhancement (LTP) and long-term suppression (LTD) characteristics of the TEPB photonic memristor induced by continuous voltage pulses (±2.5 V, 5 ms width, 7 ms interval); Figure 9 Atomic force microscopy morphology of the dielectric layer (TEPB-Pt-AF@PMMA) in the selected 5 μm × 5 μm test area; Figure 10 The current retention characteristics of the TEPB photonic memristor in the high-resistivity state (HRS) and low-resistivity state (LRS) at a constant read voltage of 0.1 V; Figure 11 The TEPB optical memristor exhibits the synaptic plasticity (EPSC) effect induced by an electrical pulse: the current response to different amplitudes with a fixed pulse width of 10 ms. Figure 12 The image shows the inhibitory postsynaptic current (IPSC) effect of the TEPB photonic memristor under electrical pulse induction; Figure a shows the current response as a function of pulse width at a constant voltage amplitude of 2 V; Figure b shows the current response as a function of voltage amplitude at a fixed pulse width of 10 ms. Figure 13 Figure 1 shows the optical characteristics of the TEPB photonic memristor; Figure 2a shows the IV characteristic curves of the TEPB photonic memristor under darkness and 375nm ultraviolet light, respectively; Figure 3b shows the TEPB photonic memristor under different incident light intensities (0 to 180 mW·cm). 2 Figure 1 shows the photoresponse current; Figure 2 shows the dynamic photoresponse of the TEPB photonic memristor to different light pulse widths (0 to 900 ms); Figure 3 shows the relationship between the photoinduced PPF exponent and the ultraviolet light pulse interval (inset: experimental data); Figure 4 shows the multi-pulse photoresponse of the TEPB photonic memristor at different frequencies; Figure 5 shows the photoresponsivity (R) and photodetectivity (D) at the same light intensity. (Figure g is a diagram illustrating the memory and forgetting of the letter "T" pattern;) Figure 14To visually demonstrate the light response characteristics of various devices during the training process of memorizing the letter "T", their ability to capture and retain visual information was intuitively shown. Figure 15 Figure a in the diagram is a schematic diagram of the image contrast enhancement effect based on the training process; Figure b is the current response curve of the device to different light intensities, corresponding to different gray levels in the input image. Figure 16 Figure a shows the current response characteristics of the TEPB photonic memristor to a 4-bit binary optical pulse sequence; Figure b shows the statistical distribution of the current response of the TEPB photonic memristor under 4-bit binary optical pulse excitation. Figure 17 This is a multi-task RC processing system within a sensor based on a TEPB photonic memristor. Figures a and b represent schematic diagrams of a human visual recognition system and an RC architecture within the sensor, respectively. Figure c shows a comparison of static handwritten digit recognition and dynamic handwritten letter recognition tasks. Figure d shows the confusion matrix for static handwritten digit array recognition. Figure e shows the confusion matrix for dynamic handwritten letter recognition. Figure f shows the high-dimensional mapping of words such as "FMEG", "RRAM", "BLUE", "STAR", and "LUCK" in the dynamic RC system within the sensor. Figure g is a schematic diagram of a waveform classification task. Figure h shows the input signal (top) composed of superimposed triangular and square waves and the classification result (bottom), demonstrating high accuracy even when the input noise level (NRMSE) is 0.1. Figure 18 This is a paradigm for handwritten digit recognition based on the TEPB photonic memristor RC system; Figure 19 In this context, 'a' represents the handwritten digit recognition accuracy, and 'b' represents the evolution of the loss function of the fully connected network as the number of training iterations increases. Figure 20 The evolution of dynamic word recognition accuracy in the TEPB photonic memristor network with the training iteration cycle; Figure 21 A schematic diagram illustrating the introduction of a mask in an RC system based on a dynamic memristor. Detailed Implementation

[0052] The present invention will be further described in detail below through specific embodiments. Unless otherwise specified, the raw materials used in the following embodiments can be obtained from conventional commercial channels or prepared and isolated through simple synthesis; unless otherwise specified, the processes employed are conventional processes in the art.

[0053] Example 1 This embodiment provides a two-dimensional platinum(II)-yne organic framework, the preparation method of which is as follows: 1. transThe preparation route of -[Pt(PBu3)2Cl2] is as follows:

[0054] The specific preparation method is as follows: K₂PtCl₄ (415.1 mg, 1 mmol) was dissolved in deionized water (7.5 mL) and placed in a round-bottom flask. The mixture was stirred until K₂PtCl₄ was completely dissolved, forming an orange-red solution. PBu₃ (0.6 mL, 2.5 mmol) was added dropwise, and stirring was continued for 3 h until a white precipitate formed. After the reaction was complete, the product was extracted with dichloromethane, the organic phase was collected, and the solvent was evaporated to obtain a pale yellow oil. The oil was heated to 190 °C for 1 h, and the resulting mixture was purified by rapid column chromatography (silica gel, eluent: V). 正己烷 / V 二氯甲烷 =5:1). After drying under reduced pressure, a yellow powdery product was obtained. trans -[Pt(PBu3)2Cl2].

[0055] The product was characterized as follows: 1 H NMR (400 MHz, Chloroform- d ): δ ppm 1.98–1.76 (m, 12H,CH2), 1.61–1.40 (m, 24H, CH2), 0.95 (t, J = 7.2 Hz, 18H, CH3). 31 P NMR (400 MHz, Chloroform- d ): δ ppm 4.44 ( 1 J Pt-P = H2).

[0056] 2. The preparation equation for the two-dimensional platinum(II)-yne organic framework is as follows: ; Two-dimensional platinum(II)-yne organic framework is achieved through trans Synthesized by a base-catalyzed dehydrohalogenation reaction between -[Pt(PBu3)2Cl2] (100.6 mg, 0.15 mmol) and 1,3,5-tris(4'-ethynylphenyl)benzene (TEPB) (37.9 mg, 0.1 mmol).

[0057] The specific preparation method is as follows: TEPB of the above-mentioned mass, trans[Pt(PBu3)2Cl2] and cuprous iodide (2 mg, 0.03 mmol, catalyst) were placed in a dry Schlenk flask. The system was evacuated and purged with nitrogen to ensure an inert atmosphere. Anhydrous THF (3 mL) was then added, followed by Et3N (3 mL). The mixture was stirred at 40 °C for 72 h to obtain a yellow powder. The crude product was washed with dichloromethane to remove unreacted ligands and residual cuprous iodide. It was then purified using a Soxhlet extractor with ACN and THF (24 h each) to remove the catalyst, oligomers, and residual platinum complexes. Finally, it was dried under vacuum to obtain a pale yellow two-dimensional platinum(II)-yne organic framework bulk material, denoted as TEPB-Pt-AF (90% yield).

[0058] Example 2 This embodiment provides a two-dimensional platinum(II)-yne organic framework, the preparation equation of which is as follows: ; Two-dimensional platinum(II)-yne organic framework is achieved through trans Synthesized by a base-catalyzed dehydrohalogenation reaction between -[Pt(PBu3)2Cl2] (100.6 mg, 0.15 mmol) and 2,4,6-tris(4-ethynylphenyl)-1,3,5-triazine (TEPT) (38.2 mg, 0.1 mmol).

[0059] The specific preparation method is as follows: TEPT of the above-mentioned mass, trans [Pt(PBu3)2Cl2] and cuprous iodide (2 mg, 0.03 mmol, catalyst) were placed in a dry Schlenk flask. The system was evacuated and purged with nitrogen to ensure an inert atmosphere. Anhydrous THF (3 mL) was then added, followed by Et3N (3 mL). The mixture was stirred at 40 °C for 72 h to obtain a yellow powder. The crude product was washed with dichloromethane to remove unreacted ligands and residual cuprous iodide. It was then purified using a Soxhlet extractor with ACN and THF (24 h each) to remove the catalyst, oligomers, and residual platinum complexes. Finally, it was dried under vacuum to obtain a yellow two-dimensional platinum(II)-yne organic framework bulk material, denoted as TEPT-Pt-AF (90% yield).

[0060] Material characterization pass 13 For the first time, C-cross polarized magic angle rotation (CP-MAS) solid-state NMR spectroscopy has resolved the chemical structures of TEPB-Pt-AF and TEPT-Pt-AF. For example... Figure 1As shown in b, the resonance peaks at δ ≈ 8 and 19 ppm correspond to the butyl group of the PBu3 unit, while the peaks at δ ≈ 80 and 110 ppm belong to the –C≡C–Pt–C≡C– structural unit. For TEPB-Pt-AF, the peaks in the 120–150 ppm range belong to the aromatic carbon of the benzene ring, while TEPT-Pt-AF also shows the characteristic peak of the C–N bond in the triazine core at δ ≈ 170 ppm. Figure 2 The Fourier Transform Infrared (FTIR) spectra of TEPB-Pt-AF and TEPT-Pt-AF are shown at 3200 cm⁻¹. 1 The sharp peaks nearby disappeared, corresponding to the C–H stretching vibration, which confirms that the dehydrohalogenation reaction has been completed. Figure 3 (Approximately 2950 cm) 1 2920 cm 1 and 2860 cm 1 Characteristic absorption peaks were detected at the locations, corresponding to the butyl saturated C–H stretching vibration modes, proving that... trans Successful coordination of -[Pt(PBu3)2Cl2] with organic ligands. X-ray photoelectron spectroscopy (XPS) analysis confirmed the elemental composition and bonding environment of TEPB-Pt-AF. X-ray photoelectron spectroscopy showed the presence of Pt, C, and P elements, and the high-resolution C 1s spectrum decomposed into six sub-peaks (283.19, 284.59, 285.02, 285.78, 286.63, and 288.98 eV), corresponding to C–Pt, C–C (Pt, C(PBu3)2Cl2) and organic ligands, respectively. sp 3 , sp 2 and sp C=O bond, C-P bond and C=O bond Figure 1 c). High-resolution Pt 4f spectrum at 71.54 eV (4f) 7 / 2 ) and 74.76 eV (f 5 / 2 The presence of a double peak at point () indicates that Pt is in a divalent state () Figure 1 (d in the text)

[0061] The electronic structure and coordination environment of Pt atoms were studied using X-ray absorption spectroscopy (XAS). Figure 1As shown in e, the adsorption edge of TEPB-Pt-AF lies between Pt foil (0) and PtO2 (+4), indicating that its oxidation state is between 0 and +4. Extended X-ray absorption fine structure (EXAFS) spectra show a dominant peak at approximately 1.83 Å, corresponding to the first coordination shell (Pt–C / P), while no Pt–Pt (~2.46 Å) or Pt–O (~1.63 Å) was observed, confirming the absence of Pt clusters or oxides. Figure 1 f in (f). For k 2 Fitting analysis of weighted EXAFS data yielded a coordination number of approximately 4.1, with Pt–P and Pt–C bond lengths of ~2.34 Å and ~2.01 Å, respectively, consistent with the expected molecular structure. Figure 1 g and Figure 4 Wavelet transform analysis further revealed a point at 5.69 Å corresponding to the Pt–C / P bond, further verifying that the metal ion Pt(II) in TEPB-Pt-AF is dispersed in the two-dimensional material in the form of single atoms. Figure 1 h). Inductively coupled plasma atomic emission spectroscopy (ICP-OES) quantitatively confirmed that the Pt content in TEPB-Pt-AF was 23.96 wt%, consistent with the theoretical value. These results confirm the successful synthesis of TEPB-Pt-AF and TEPT-Pt-AF.

[0062] The morphological characteristics of the Pt-AF framework were studied using scanning electron microscopy (SEM) and transmission electron microscopy (TEM), revealing a clear two-dimensional layered structure. Figure 5 and 6 Atomic-level structure analysis was performed using aberration-corrected high-angle annular dark-field scanning transmission electron microscopy (AC-HAADF-STEM), clearly demonstrating the atomic-level dispersion of platinum atoms in TEPB-Pt-AF. Figure 1 The 'i' in the diagram is marked with a red circle. Energy-dispersive X-ray spectroscopy (EDS) elemental distribution maps further confirm the uniform distribution of platinum, as well as carbon and phosphorus. Figure 1 Thermogravimetric analysis (TG) curves show that the initial decomposition temperatures of both TEPB-Pt-AF and TEPT-Pt-AF are approximately 280 ℃, indicating that the materials possess good thermal stability. Figure 7 ).

[0063] Example 3 This embodiment provides a photonic memristor based on Pt-AF, and its fabrication method is as follows: The glass substrate was first pre-cleaned with ethanol and deionized water, followed by nitrogen drying. Then, the substrate underwent 15 min of UV-ozone cleaning to ensure surface cleanliness. The bottom electrode was deposited using indium tin oxide (ITO) via magnetron sputtering. TEPB-Pt-AF or TEPT-Pt-AF was dispersed in chlorobenzene at a concentration of 2 mg / mL. 1 The mixture was ultrasonicated for 24 hours. Subsequently, this dispersion was mixed with a PMMA dispersion at a volume ratio of 1:1, wherein the PMMA was dissolved in chlorobenzene at a concentration of 40 mg / mL. 1 The obtained TEPB-Pt-AF@PMMA or TEPT-Pt-AF@PMMA dispersions were spin-coated onto an ITO-coated substrate at 3000 rpm for 40 s. The coated films were annealed at 100 °C for 2 hours in a nitrogen atmosphere to form a uniform TEPB-Pt-AF@PMMA or TEPT-Pt-AF@PMMA layer. Finally, a 50 nm thick platinum (Pt) layer was deposited as the top electrode using shadow mask-assisted magnetron sputtering.

[0064] In the fabrication of flexible devices, polyimide films are used as substrates, and all subsequent steps are completely consistent with those for rigid devices.

[0065] Device performance characterization 1. Resistive switching performance of Pt-AF-based photonic memristors Figure 8 Image a shows a 3D schematic of an 8 × 8 cross-array memristor, accompanied by a scanning electron microscope (SEM) cross-sectional image of the Pt / TEPB-Pt-AF@PMMA / ITO memristor (TEPB photonic memristor), visually illustrating the structure of the platinum electrode, resistive layer, and ITO electrode. Atomic force microscopy (AFM) reveals that the roughness (Ra) of the dielectric film (TEPB-Pt-AF@PMMA) of the TEPB photonic memristor is 1.45 nm. Figure 9 ).also, Figure 8 The dh model demonstrates the performance of flexible memristors, with the TEPB photonic memristor exhibiting typical bipolar non-volatile memory characteristics, its I... on / I off The ratio is 10 3 (At 0→4V→0→) Under a voltage scan from 4V to 0, limit the current I. CC =1 mA). like Figure 8 As shown in d, the IV characteristic curve indicates that the device is initially in a high-impedance state (HRS, off state, read voltage 0.1 V), and when the voltage drops to approximately At 2.8 V, the device will suddenly switch to a low-impedance state (LRS, on-state, read voltage 0.1 V). Subsequently, the LRS state can be maintained after the voltage is removed, but when a negative voltage of +2.8 V is applied during reset, the device will suddenly switch back to a high-impedance state (HRS). This can be verified by reverse voltage scanning (0→ At 4 V→0→4 V→0, near-mirror-symmetric non-volatile behavior was observed, indicating that both positive and negative voltages can control the switching state of the memristor. More importantly, the TEPB photonic memristor still exhibits analog memristor characteristics at low voltages (<2 V), making it an ideal choice for simulating the plasticity of biological synapses—which are key elements in RC calculations within sensors. Figure 8 The IV curve shown in Figure e indicates that during 20 negative voltage scans (curve 0 → curve 20), the current gradually increases from the initial state to a higher level; subsequently, during 20 consecutive positive voltage scans (curve 21 → curve 40), the current drops back to a lower level, corresponding to the conductance enhancement and inhibition processes in biological synapses, respectively. However, although the Pt / TEPT-Pt-AF / ITO memristor (i.e., TEPT photonic memristor) also exhibits bipolar storage characteristics, its high switching voltage prevents it from producing analog memristor characteristics. Figure 8 (g and h in the text).

[0066] Studies have shown that the potential well structure formed in TEPB-Pt-AF materials enables ordered electron trapping behavior between layers, thereby forming stable analog memristor characteristics, while TEPT-Pt-AF materials without potential wells do not possess this characteristic. Figure 8 (c and f in the original text). Furthermore, the TEPB photonic memristor exhibits remarkable retention, durability, and flexibility, providing strong evidence for its potential applications in flexible or wearable devices. Under both HRS and LRS test modes, its resistance state can be stably maintained for over 10,000 s without significant fluctuations or degradation; durability tests show that by applying pulse sequences to the flexible TEPB photonic memristor at different bending radii (14 mm, 11 mm, 7 mm, 5 mm), both states remain relatively stable (…). Figure 10 Synaptic weight regulation and synaptic plasticity are core elements for building intelligent systems that more closely resemble the information processing mechanisms of the human brain, which is particularly important for realizing neuromorphic circuits.

[0067] This study simulated the current enhancement (EPSC) and inhibition (IPSC) phenomena induced by electrical stimulation in artificial synaptic devices by applying voltages of different pulse amplitudes and widths to a TEPB photonic memristor. Both the EPSC and IPSC of the device exhibited a synchronous dependence on pulse amplitude and pulse width, indicating enhanced synaptic connection strength and increased synaptic weight. Figure 8 As shown in Figure i, when the pulse amplitude changes from... Increased from 1.5 V to At 3.0 V (pulse width = 10 ms), the peak intensity of EPSC under negative voltage continuously increases until saturation, and then the current gradually decays after the pulse disappears. On the other hand, IPSC under positive voltage stimulation decreases with different pulse widths and amplitudes, and gradually recovers over time. Figure 11 In addition, both EPSC and IPSC can be improved by increasing the pulse width (2–20 ms, pulse amplitude = 2 V) for modulation ( Figure 12 Meanwhile, the potential of the device to simulate short-term synaptic plasticity (PPF) was explored by applying paired voltage pulses with intervals ranging from 1 ms to 500 ms. The results showed that the response current intensity generated by the second pulse was greater than that of the first pulse, and the fitted PPF ratios were: response time (τ1) = 2.07 ms, synaptic duration (τ2) = 79.12 ms, comparable to the magnitude of biological synapses. Figure 8 j in the text). Furthermore, such as... Figure 8 As shown in k, by applying 50 consecutive electrical pulses, the characteristics of long-term enhancement (LTP) and long-term inhibition (LTD) were successfully simulated. Experimental results show that when a negative voltage is applied (each pulse amplitude is... When a positive voltage is applied (2.5 V, duration 5 ms, interval 7 ms), the response current of the memristor gradually increases, exhibiting LTP characteristics; while when a positive voltage is applied (each pulse amplitude 2.5 V, duration 5 ms, interval 7 ms), the response current gradually decreases, corresponding to LTD characteristics. These electrical performance characteristics demonstrate that the flexible TEPB photonic memristor possesses excellent artificial synaptic plasticity, offering great potential for short-range and long-range learning simulations in neuromorphic computing.

[0068] The resistive characteristics of TEPB photonic memristors are not only modulated by an electric field, but also exhibit a significant response to light stimulation. For example... Figure 13 As shown in the IV curve in (a), when the device is at a light intensity of 165.5 mW·cm 2 Under 375 nm ultraviolet light irradiation, the current increases approximately four times compared to the dark condition. Furthermore, the optical modulation effect of these devices exhibits a different order of magnitude in its spontaneous recovery characteristics compared to electrical modulation. Notably, the photoresponse current is highly dependent on the incident light intensity (30 mW·cm⁻¹). 2 52 mW·cm 2 85mW·cm 2 120 mW·cm 2165.5 mW·cm 2 The pulse width is fixed at 500 ms and the optical pulse widths are 100 ms, 200 ms, 300 ms, 500 ms, and 1000 ms, with a power density fixed at 165.5 mW·cm². 2 All showed a positive correlation (see) Figure 13 (b and c in the text). The study also found that the enhancement of photocurrent has a significant time-dependent characteristic, exhibiting a slow nonlinear decay process, which is more advantageous in simulating artificial vision systems compared to the rapid saturation and relaxation characteristics of traditional devices. Figure 13 As shown in d, the double-pulse light-induced photodynamic response (PPF) experiment demonstrates that the intensity of the photoresponse signal is not only related to the current pulse but also influenced by the previous pulse. The fitted curve shows that the PPF ratio gradually decreases as the pulse interval increases. This cumulative effect indicates that the device can store historical input information and respond to subsequent inputs accordingly, laying the foundation for exploring in-sensor storage computing systems. Equally important, at different pulse frequencies (1.00 Hz, 1.25 Hz, and 1.67 Hz), the multiple photoresponses exhibit a nonlinear positive correlation with the number of pulses, providing an important basis for subsequent applications in reservoir computing. Figure 13 (e). Two key parameters used to evaluate a device's response to weak light—photoresponsivity (R) and photodetectivity (D). * (The former quantifies optical response efficiency, while the latter reflects the device's ability to detect weak optical signals), and its definition is as follows: ; ; Among them I ph Photocurrent; P inc S represents the incident light intensity; S represents the channel area; A represents the effective area of ​​the device; q represents the fundamental charge; I d It is dark current. Figure 13 f in the figure shows R and D The dependence of both on light intensity indicates that they are negatively correlated with light intensity, suggesting that the device is sensitive to low-light environments. Particularly noteworthy is the effect observed when the device is under low pulse intensity (30 mW·cm⁻¹). 2 When ), the R value reaches its maximum value of 23.52 × 10. 2 AW 1 High pulse intensity (165.5 mW·cm) 2 The value at that time was increased by 3.46 times. Similarly, the device's D... The detection sensitivity under low-light conditions was also improved by 3.46 times compared to strong-light environments. Therefore, the TEPB photonic memristor exhibits dose-dependent output current and hold-time characteristics, and holds promise for multi-signal preprocessing functions such as image storage and contrast enhancement. To evaluate its image storage capability, this invention applied an intensity of 165.5 mW·cm² to a flexible TEPB photonic memristor employing a 3 × 3 array structure using a photomask with a "T"-shaped pattern. 2 The light pulses, each storage unit corresponds to one image pixel. For example... Figure 13 As shown in g, the photocurrent on the pixel excited by the ultraviolet light pulse is significantly greater than that on the pixel excited by the ultraviolet light pulse, and the longer the duration of the light pulse, the greater the increase in response current. Furthermore, even 60 minutes after the light pulse stops, the letter "T" can still be sensed and retained, indicating that the device has excellent sensing and memory capabilities. The output currents of five randomly selected memory cells in the array are almost identical, further demonstrating the good consistency in photoresponse of this type of device. Figure 14 The contrast enhancement characteristics of four-grayscale images were further verified using a 3 × 3 photonic memristor array. Most importantly, the output current increases significantly faster at high light intensities than at low light intensities. Conversely, the current decays faster in low-intensity pixels after the input process than in high-intensity pixels, resulting in a more vivid output image. Figure 15 (a) This is because, compared to the original image, the current differences between pixels in the output image are amplified, ultimately forming a reconstructed output image with enhanced contrast. Figure 15 (b) The nonlinear dynamic evolution of the channel current under light stimulation is crucial to the RC mechanism within the sensor. This dynamic evolution exhibits short-term memory characteristics, which can be visually demonstrated by applying a 4-bit binary optical pulse sequence to the photonic memristor array. Figure 16 Figure a shows that an increase in current was detected under the action of pulse "1", while pulse "0" caused a decrease in current (where "1" and "0" represent intensities of 80 mW·cm). 2 (0 mW·cm) 2 (500 ms pulse width, 300 ms interval optical pulses). The resulting sixteen sets of 4-bit pulse sequences (from 0000 to 1111) form sixteen clearly distinguishable states, indicating that this method can effectively map complex spatiotemporal signals to reservoir states. Figure 16 (b).

[0069] 2. In-sensor multi-task reservoir calculation based on TEPB photonic memristor devices TEPB photonic memristors hold great potential for in-sensor RC hardware implementation, with their core advantage being the ability to map spatiotemporal optical signals to a high-dimensional computational space without the need for additional sensors or processors. Inspired by the integrated perception and processing functions of the human visual system, this invention constructs an in-sensor RC system based on TEPB photonic memristors as the physical reservoir and their adapted neural network structure. This system applies the input optical signal to the physical reservoir and ultimately extracts the readout layer conductance (...) through the TEPB photonic memristor. Figure 17 a and Figure 18 To comprehensively evaluate its application value in the field of RC within sensors, this invention conducted static digit recognition and dynamic letter recognition tasks. For example... Figure 17 c. The system uses the MNIST digit dataset, which contains 28 × 28 pixel handwritten digits from 0 to 9, as input images (this dataset consists of 60,000 training images and 10,000 test images). Taking the digit "6" as an example, the original 28 × 28 grayscale image is first converted to binary and divided into seven 4-bit pulse sequences (each pixel is represented by "0" or "1", where "1" represents an input intensity of 80 mW·cm). 2 The ultraviolet light pulses ("0" indicates no light pulse input) are then converted into ultraviolet light pulses and input into the photonic memristor. The last photocurrent response of each 4-bit pulse sequence is selected as the storage state and then passed to the readout layer for subsequent processing and analysis. From the perspective of neural network architecture, the weights of the output layer network are first randomly initialized and used as the starting point for network learning. The network learning rate is set to 0.001 to control the step size of weight updates during training, thereby ensuring that the neural network can learn continuously and efficiently. To convert the network output into a probability distribution, the Softmax function is selected as the activation function, and the cross-entropy loss function is referenced to quantify the deviation between the predicted result and the actual label. Finally, by minimizing the loss function and combining it with gradient descent-based training rules, the weights of the readout layer are iteratively optimized to continuously improve network performance. After 2000 iterations, the recognition rate reached 90.775%. As the number of iterations increases, the recognition accuracy and loss function value of the fully connected neural network fluctuate rapidly and eventually tend to stabilize. Figure 19 The confusion matrix plotted based on the classification data further demonstrates that the system has excellent handwritten digit recognition capabilities. Figure 17 (d) In addition, this invention also performs a dynamic handwritten letter recognition task to simulate real-time capture of external dynamic information and perform preliminary preprocessing. Each word containing four letters ("FMEG", "RRAM", "BLUE", "STAR", "LUCK") is input into the RC system within the dynamic sensor for classification and recognition. Figure 17In the 'c' section, four frames of letter images "F", "M", "E", and "G" selected from the "EMNIST-letters" dataset are input sequentially in chronological order to simulate the real-time dynamic reception of letters by the human eye. Each letter image is 28 × 28 pixels in size, and the input of each pixel is converted into a 4-bit light pulse sequence and input into the RC system. After completing the high-dimensional mapping of the dynamic letter light information, these reservoir states are passed to a subsequent single-layer neural network for further classification and recognition processing. Figure 17 The high-dimensional mapping of f in the image shows the visualization of the current corresponding to the reservoir state after five different words are sequentially input into the TEPB photonic memristor RC system. Due to the nonlinear decay characteristics of the photoresponse, the current corresponding to the last two letters increases significantly, while the current corresponding to the two earlier input letters decreases accordingly. This characteristic is consistent with human memory characteristics—information acquired early gradually fades over time, while dynamic image information leaves a deeper impression. After training, the readout layer network eventually achieved a test set recognition accuracy of 89.6% with increasing iterations. Figure 20 ).from Figure 17 The confusion matrix diagram clearly shows that all five words can be accurately classified. It's worth noting that this system is not limited to recognizing four-letter words; through simple cropping, it can also recognize more complex dynamic text images.

[0070] To directly process the raw data, this application employs a dynamic sensor-internal RC system to further complete the waveform classification task. This system simulates the working principle of a visual neural network. To address the challenge of random connections between multiple neuromorphic devices required in traditional RC architectures, this application constructs a dynamic sensor-internal RC system by introducing the concept of time-division multiplexing—this system combines masking techniques to generate virtual nodes in the time domain. These virtual nodes utilize the dynamic nonlinear response characteristics of photonic memristors to achieve nonlinear coupling, generating a large number of reservoir states at the hardware level using a small number of photonic memristors. This optimizes hardware complexity while maintaining the reservoir system's ability to handle complex tasks. Figure 21 The architecture of the RC system within the dynamic sensor is demonstrated: the input optical signal is first preprocessed through a mask matrix, and then converted into an optical pulse sequence by a signal generation system. Each frame of the input signal generates a specific pulse sequence with a total length τ and a pulse width δ, which is then input into a photonic memristor. The photonic memristor samples the output of each virtual node and outputs a linear combination of reservoir states, which is then used to train weights through linear regression to finally output the result. In this study, multiple memories based on TEPB photonic memristors are connected in parallel to form a parallel RC system to improve performance, such as... Figure 17As shown in g in the diagram. By using six masks in parallel, a single input data can generate six sets of pulse sequences. Positive and negative pulses are input into two photonic memristors respectively, driving twelve parallel photonic memristors in the reservoir. Due to the use of 5-frame masks, the reservoir state, composed of the virtual node states of all photonic memristors, is expanded from 12 to 120 (12 × 10), and then fed back to a 120 × 1 readout network for classification. The input signal consists of random triangular and square waves. Each point in the input sequence is multiplied by a five-pixel one-dimensional vector mask composed of random binary bits 0 and 1. The mask is then converted into a 5-frame pulse sequence, extracting 2 virtual sampling points per frame. Optical pulses (τ = 2 s) are input to the photonic memristors using a pulsed laser. The readout network is trained via linear regression to output a binary sequence, with -1 and 1 representing the triangular and square waves respectively. Figure 17 As shown in Figure h (top: input waveform; bottom: classification result), the trained dynamic sensor-based RC system can accurately distinguish between sine waves and square waves with a classification accuracy (NRMSE) of 0.1. This NRMSE value is excellent, even surpassing the recent reported benchmark of 0.13 for all-ferroelectric RC systems. This breakthrough is mainly attributed to the TEPB photonic memristor's ability to generate sufficiently strong feedback and rich reservoir states.

[0071] The above embodiments are preferred embodiments of the present invention, but the embodiments of the present invention are not limited to the above embodiments. Any changes, modifications, substitutions, combinations, or simplifications made without departing from the spirit and principle of the present invention shall be considered equivalent substitutions and shall be included within the protection scope of the present invention.

Claims

1. A two-dimensional platinum(II)-yne organic framework, characterized in that, The chemical structural formula of the two-dimensional platinum(II)-yne organic framework is shown in formula (I): Equation (Ⅰ); Ar is either phenyl or triazine.

2. The two-dimensional platinum(II)-yne organic framework according to claim 1, characterized in that, The chemical structural formula of the two-dimensional platinum(II)-yne organic framework is selected from: ； 。 3. The method for preparing the two-dimensional platinum(II)-yne organic framework according to claim 1 or 2, characterized in that, Includes the following steps: In the presence of a catalyst, ligands and trans The two-dimensional platinum(II)-yne organic framework was prepared by reacting -[Pt(PBu3)2Cl2].

4. The method for preparing the two-dimensional platinum(II)-yne organic framework according to claim 3, characterized in that, The catalyst includes at least one of cuprous iodide, cuprous bromide, and cuprous chloride.

5. The application of the two-dimensional platinum(II)-yne organic framework according to claim 1 or 2 in memristors.

6. A memristor, characterized in that, Its structure includes a substrate, a bottom electrode layer, a dielectric layer and a top electrode layer arranged sequentially; the material of the dielectric layer includes the two-dimensional platinum(II)-yne organic framework as described in claim 1 or 2.

7. The memristor according to claim 6, characterized in that, The material of the dielectric layer also includes organic polymers.

8. The method for fabricating a memristor according to claim 6 or 7, characterized in that, Includes the following steps: A bottom electrode layer, a dielectric layer, and a top electrode layer are sequentially formed on a substrate using a coating process; the coating process includes thermal evaporation, magnetron sputtering, sol-gel, chemical vapor deposition, or coating methods.

9. A sensor-in-sense storage computing system, characterized in that, Includes the memristor as described in claim 6 or 7.

10. The application of the in-sensor storage computing system of claim 9 in artificial synapses and neuromorphic computing.

Images