A surface-enhanced raman scattering flexible substrate material and a preparation method and application thereof

Abstract

The application belongs to the technical field of gas analysis, and relates to a surface-enhanced Raman scattering flexible substrate material and a preparation method and application thereof. The surface-enhanced Raman scattering flexible substrate material comprises a three-dimensional network structure composed of hydroxypropyl methyl cellulose nanofibers and silver nanoparticles wrapped in the hydroxypropyl methyl cellulose nanofibers, wherein the silver nanoparticles are in a polyhedral structure. The surface-enhanced Raman scattering flexible substrate material can detect mercaptans in a solution, and the detection limit can reach 10 ‑19 M level; after the silver nanoparticle surface is modified with a probe, the material can detect trace substances in exhaled air. In addition, after the material is stored at room temperature for 6 months, the SERS signal strength attenuation is less than or equal to 4%, and the relative standard deviation of 20 random sites is only 4.04%.

Description

Technical Field

[0001] This invention belongs to the field of surface-enhanced Raman scattering (SERS) technology, and relates to a surface-enhanced Raman scattering flexible substrate material, its preparation method and application. Background Technology

[0002] Surface-enhanced Raman scattering (SERS) has become a powerful tool for detecting trace chemical and biomolecules due to its unique fingerprinting capabilities and extremely high sensitivity. Theoretically, SERS can achieve single-molecule detection. However, in practical solution-phase detection applications, especially for ultrasensitive detection of molecules with key functional groups such as thiols (-SH), its performance still faces significant bottlenecks. Currently, most SERS substrates, even the high-performance cellulose-based or noble metal-based flexible substrates, typically only achieve a detection limit of 10 for thiol-based probe molecules in solution. -12 M to 10 -15 The M-level (M-level) is insufficient to meet the growing demand for the detection of extremely low concentration biomarkers. This limitation stems primarily from the following: First, silver nanoparticles (AgNPs), the active core of SERS, are highly susceptible to oxidation, aggregation, or contamination in high-humidity environments or complex solution systems, leading to the attenuation of their local surface plasmon resonance (LSPR) characteristics, insufficient stability of SERS "hot spots," and severe signal decay over time. Second, many substrates exhibit poor signal uniformity and large relative standard deviations (RSDs), affecting the reliability of quantitative analysis. Furthermore, traditional high-performance SERS substrates are mostly rigid structures, limiting their application in emerging scenarios such as flexible sensing and wearable devices.

[0003] Further applying SERS technology to the detection of trace volatile organic compounds (VOCs) in gaseous samples such as human exhaled breath is an important direction for non-invasive disease diagnosis. However, this also introduces more severe challenges, which can be seen as an extension of the bottlenecks in solution-phase detection in more complex scenarios. The concentration of target biomarkers in exhaled breath is extremely low, often at the nanomolar to picomolar level, and exists in a complex matrix composed of a large amount of water vapor, carbon dioxide, and hundreds of VOCs. This not only requires the substrate to have higher sensitivity, but also requires it to effectively capture and enrich gas molecules. However, traditional SERS strategies optimized for solution detection have extremely low enrichment efficiency for gaseous molecules. In addition, direct exposure to warm, humid exhaled breath will drastically accelerate the oxidative inactivation process of silver nanoparticles. Therefore, developing a SERS substrate with ultra-high sensitivity, excellent antioxidant stability, good signal reproducibility, and flexibility to first break through the limits of ultra-trace detection in solution phases, and then lay the foundation for overcoming the challenges of efficient and stable detection of complex gaseous samples such as exhaled breath, is of urgent need and great significance. Summary of the Invention

[0004] The present invention aims to provide a surface-enhanced Raman scattering flexible substrate material, its preparation method and application, to solve the primary problem of insufficient sensitivity of existing SERS technology for detecting thiol compounds in solution, and at the same time achieve efficient and stable detection of trace marker molecules in exhaled breath.

[0005] This invention utilizes a three-dimensional network structure with a high specific surface area and rich in hydroxyl groups to increase the specific surface area for active adsorption and enrichment of thiol molecules. When the pre-enriched thiol molecules diffuse to the vicinity of AgNPs, their thiol groups form strong Ag-S covalent bonds with the silver surface. The polyhedral structure of AgNPs provides extremely strong electromagnetic field enhancement, enabling the surface-enhanced Raman scattering flexible substrate material to achieve a detection limit of 10 for thiol compounds such as 4-ATP in solution. -19 M level.

[0006] This invention utilizes probe molecules attached to a silver surface to adsorb polar target molecules in exhaled gas, thereby achieving efficient detection of trace marker molecules in exhaled gas.

[0007] Furthermore, this invention encapsulates AgNPs within hydroxypropyl methylcellulose (HPMC) nanofibers to protect them from oxidation, and utilizes the hydroxyl groups on the HPMC surface to reduce silver oxide, thereby ensuring the stability of the detection results; moreover, this surface-enhanced Raman scattering flexible substrate material is flexible, adaptable to the curved surface of the human body, and can be integrated into wearable devices.

[0008] The technical solution provided by this invention is as follows: In a first aspect, the present invention provides a surface-enhanced Raman scattering flexible substrate material, comprising: A three-dimensional network structure composed of hydroxypropyl methylcellulose nanofibers; And silver nanoparticles encapsulated within the hydroxypropyl methylcellulose nanofibers; The silver nanoparticles have a polyhedral structure.

[0009] In the surface-enhanced Raman scattering (SERS) flexible substrate material of this invention, hydroxypropyl methylcellulose (HPMC) nanofibers encapsulate silver nanoparticles within them, isolating the silver nanoparticles from the external environment. This effectively mitigates the oxidation of silver nanoparticles in humid air, thus ensuring the long-term stability of the SERS hotspot and achieving a signal attenuation of ≤4% after 6 months of storage. The hydroxyl groups abundant in the HPMC molecular chain act as a mild reducing agent during preparation or storage, neutralizing Ag generated on the surface of the silver nanoparticles due to slight oxidation. + Restored to Ag 0 ( Figure 3XRD and XPS evidence from a~b in the figure shows that the silver nanoparticles maintain their metallic state and SERS activity through self-healing. The hydroxyl groups of HPMC can form hydrogen bonds with many polar molecules (such as benzaldehyde and 4-ATP), actively and selectively capturing and pre-enriching these molecules on and near the fiber surface. Subsequently, the enriched molecules diffuse to the adjacent AgNPs region. Furthermore, the three-dimensional network possesses high porosity and hydrophilicity, providing channels for rapid diffusion and deep penetration of gaseous or liquid analyte molecules. Target molecules can easily enter the interior of the three-dimensional network, greatly increasing the probability of contact between the target molecules and the internal AgNPs. Compared to ordinary spherical AgNPs, polyhedral AgNPs have more exposed crystal faces and edge sites, generating stronger local electromagnetic field enhancement, which is crucial for achieving ultra-high sensitivity (10⁻⁶). -19 The physical basis of M).

[0010] Preferably, the mass ratio of silver nanoparticles to hydroxypropyl methylcellulose is (0.10~0.50):1. Both excessively large and small mass ratios will adversely affect the performance of the surface-enhanced Raman scattering (SERS) flexible substrate material. Specifically: when the mass ratio is >0.50:1, there are too many silver nanoparticles, which will damage the morphology and structural integrity of the hydroxypropyl methylcellulose nanofibers, causing oxidation of the silver nanoparticles exposed to humid air and reducing their stability; when the mass ratio is <0.10:1, there are too few silver nanoparticles, resulting in insufficient effective SERS active sites. Preferably, the mass ratio of silver nanoparticles to hydroxypropyl methylcellulose is (0.35~0.50):1; the optimal mass ratio is 0.45:1.

[0011] Preferably, the thickness of the surface-enhanced Raman scattering (SERS) flexible substrate material is 10–100 μm. Both excessively thick and thin materials will adversely affect the performance of the SERS flexible substrate material. Specifically: when the thickness is too thin (<10 μm), the total amount of silver nanoparticles loaded is insufficient, resulting in a low SERS hotspot density and insufficient diffusion paths and residence times for analyte molecules; when the thickness is too large (>100 μm), the material's rigidity increases while its flexibility decreases, potentially causing cracks during bending or folding, and even leading to the destruction of the internal nanofiber network or AgNPs encapsulation structure, affecting SERS performance and mechanical stability.

[0012] Preferably, the pore size of the surface-enhanced Raman scattering flexible substrate material is 1~5μm; this size allows gas molecules to quickly penetrate into the entire three-dimensional network of the material through free diffusion, while maintaining the overall structural strength, flexibility and permeability of the material.

[0013] Preferably, the porosity of the surface-enhanced Raman scattering flexible substrate material is 60% to 90%. This porosity imparts an extremely high specific surface area to the material, which is 20 to 100 m² / g. 2 The porosity of the material allows silver nanoparticles to disperse throughout the entire three-dimensional space, providing a large exposed area for hydroxyl groups, enabling them to efficiently capture polar target molecules in exhaled gas. Simultaneously, it allows excitation light to penetrate the material and excite the silver nanoparticles, while maintaining a low solid-state filling ratio, reducing light absorption and scattering by the solid. Furthermore, this porosity maintains the necessary mechanical support of the material, preventing collapse or damage.

[0014] Preferably, the silver nanoparticles have a particle size of 100-250 nm, and the hydroxypropyl methylcellulose nanofibers have a diameter of 50-500 nm. The particle size of the silver nanoparticles can be larger or smaller than the diameter of the pure hydroxypropyl methylcellulose nanofibers.

[0015] Preferably, the surface-enhanced Raman scattering flexible substrate material has a roughness Ra value > 500 nm.

[0016] The above-mentioned surface-enhanced Raman scattering flexible substrate material can fix thiol target molecules on the surface of silver nanoparticles by the reaction between silver nanoparticles and thiol groups, and generate extremely strong SERS signals under laser excitation.

[0017] Furthermore, the surface of the silver nanoparticles is also modified with probe molecules; wherein the probe molecules contain a first functional group for anchoring to the surface of the silver nanoparticles and a second functional group for capturing target molecules. The probe molecules pre-modified on the surface of the silver nanoparticles can undergo a specific chemical reaction with the target molecules through the second functional group, thereby stably and specifically fixing the target molecules to the SERS "hot spot," generating an extremely strong SERS signal under laser excitation. For example, if the first functional group of the probe molecule is a thiol group and the second functional group is an amino group, it can be used to capture aldehyde target molecules; if the second functional group is a carboxyl group, it can be used to capture alcohol target molecules.

[0018] Secondly, the present invention provides a method for preparing a surface-enhanced Raman scattering flexible substrate material, comprising the following steps: dissolving hydroxypropyl methylcellulose in a water-ethanol mixed solvent, adding silver nanoparticles to form a spinning solution; spinning the spinning solution into a nanofiber membrane by electrospinning; and drying the nanofiber membrane to obtain the surface-enhanced Raman scattering flexible substrate material.

[0019] This preparation method disperses silver nanoparticles in a spinning solution and encapsulates them within hydroxypropyl methylcellulose nanofibers via electrospinning, thus isolating the silver nanoparticles from the external environment and solving the problems of easy oxidation and instability of silver nanoparticles. Simultaneously, the electrospinning process forms a three-dimensional network structure that allows for rapid diffusion of gas molecules and possesses a large adsorption area, achieving highly efficient enrichment of trace polar target molecules and improving the detection limit. This preparation method uses a water-ethanol mixed solvent, which conforms to the principles of green chemistry and avoids interference from residual solvents in subsequent detection.

[0020] Preferably, the water-ethanol mixed solvent is composed of water and ethanol in a volume ratio of 1:(1~3); this volume ratio helps to form continuous and uniform nanofibers.

[0021] Preferably, the mass ratio of the hydroxypropyl methylcellulose to the water-ethanol mixed solvent is (1~3):100; this mass ratio determines the concentration of the polymer in the spinning solution. If the concentration is too low, it is difficult to form continuous fibers; if the concentration is too high, the solution viscosity is too high, making it difficult to form a stable jet, and may result in excessively large fiber diameters, or even make spinning impossible.

[0022] Preferably, the mass ratio of the silver nanoparticles to hydroxypropyl methylcellulose is (0.10~0.50):1.

[0023] Preferably, the drying temperature is 50~70℃, and the time is 2~5 hours. Drying at this temperature efficiently removes residual solvents, gently solidifies the fiber structure, enhances the material's mechanical properties, and protects the material's porous structure.

[0024] Furthermore, the preparation method further includes a probe molecule modification step, which includes: immersing the dried nanofiber membrane in a solution of probe molecules; after sufficient reaction, washing away probe molecules that have not bound to the silver nanoparticles; wherein the probe molecule contains a first functional group for anchoring to the surface of the silver nanoparticles and a second functional group for capturing target molecules. Preferably, the first functional group is one or more of thiol, carboxyl, phosphonic acid, and cyano groups, and the second functional group is one or more of carboxyl, aldehyde, ketone, biotin, antibody, and aptamer. More preferably, the probe molecule is 4-aminothiophenol.

[0025] Thirdly, the present invention provides a face mask comprising the surface-enhanced Raman scattering flexible substrate material described in the first aspect. Integrating the surface-enhanced Raman scattering flexible substrate material into the inner layer or interlayer of the face mask allows it to directly and imperceptibly contact and collect polar target molecules in exhaled air during wear.

[0026] Fourthly, the present invention provides an exhaled gas detection device, comprising: A face mask comprising the surface-enhanced Raman scattering flexible substrate material as described in the first aspect, for capturing and pre-enriching target molecules in exhaled gas; A Raman spectroscopy detection device is used to scan the surface-enhanced Raman scattering flexible substrate material in a used mask to obtain the Raman spectral signal of the target molecules. The data processing unit is configured to use machine learning algorithms to analyze the Raman spectral signal, identify the characteristic peaks of the target molecule, and output the detection results.

[0027] In this exhaled gas detection device, a mask is used for non-intrusive, in-situ sampling and signal pre-enhancement; a Raman spectroscopy detection device is used to read the signal; and a data processing unit is used to analyze the signal and output the results.

[0028] In some embodiments, the data processing unit includes: The data preprocessing module is used to smooth, denoise, and correct the baseline of the Raman spectral signal; The feature extraction module is used to extract the position and intensity information of the characteristic peaks of the target molecules from the corrected spectral signal; The intelligent recognition module is used to identify the types and concentrations of target molecules in exhaled gas based on the characteristic peak information.

[0029] In this data processing unit, the data preprocessing module reduces noise in the raw signal acquired by the Raman spectroscopy detection device to prevent it from masking or distorting the weak characteristic peaks of the target molecules, ensuring that the input to the subsequent analysis process is spectral data with a high signal-to-noise ratio and a clean background; the feature extraction module transforms the complex spectrum into a set of feature vectors that can characterize the presence and concentration of the target substance, greatly improving the efficiency and accuracy of subsequent identification; the intelligent identification module transforms the technically complex spectral data into intuitive detection results, making it easy for non-professionals to use.

[0030] Compared with the prior art, the present invention has at least the following beneficial effects: 1. The surface-enhanced Raman scattering flexible substrate material of this invention can be used to detect thiol compounds in solution, with a detection limit of up to 10. -19 M-level sensitivity, which is higher than that of cellulose-based SERS substrates (typical detection limit 10) recently reported in top journals. -12 ~10 -9 M) and noble metal-based flexible substrates (typical detection limit 10). -15 ~ 10 -12 The M) was improved by 3 to 10 orders of magnitude. After the probe was modified with silver nanoparticles, the material was able to detect trace amounts of substances in exhaled breath.

[0031] 2. This invention encapsulates AgNPs uniformly within hydroxypropyl methylcellulose nanofibers, forming a biomimetic "pearl shell" structure, providing AgNPs with dual protection through physical isolation and chemical anchoring. This results in the SERS signal intensity of the surface-enhanced Raman scattering (SERS) flexible substrate material decreasing by ≤4% after 6 months of storage at room temperature, effectively solving the industry problem of signal attenuation caused by the easy oxidation and aggregation of precious metal nanoparticles. Simultaneously, the substrate surface exhibits uniform signal, with a relative standard deviation (RSD) of only 4.04% for 20 random sites, demonstrating excellent signal reproducibility.

[0032] 3. Traditional strategies rely on hydrophobic surface evaporation and concentration to enrich gaseous polar molecules. This invention utilizes the natural hydrophilicity and abundant hydroxyl groups (-OH) of the HPMC nanofiber network to actively capture and pre-enrich polar VOCs markers (such as acetone and benzaldehyde) in exhaled gas through hydrogen bonding. This synergistic enrichment mechanism of "three-dimensional porous structure + hydrogen bonding" (…) Figure 5 The y in the middle solves the fundamental defect that traditional hydrophobic surfaces have extremely low enrichment efficiency for gaseous molecules.

[0033] 4. The entire preparation process of this invention is based on a water-ethanol mixed solvent and employs electrospinning technology, completely avoiding the use of toxic chemical reagents and organic solvents. The raw material, hydroxypropyl methylcellulose, is derived from natural cellulose and possesses excellent biocompatibility and biodegradability, conforming to green chemistry principles and avoiding environmental burden and biosafety risks. 5. This invention integrates a surface-enhanced Raman scattering flexible substrate material into a smart mask, and utilizes a data processing unit based on machine learning algorithms to perform rapid and automatic analysis of complex spectra, accurately identify specific disease biomarkers, and achieve the integration of sampling enrichment, signal detection, and intelligent analysis. Attached Figure Description

[0034] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0035] Figure 1 Conceptual diagram of the core innovation of the flexible SERS platform.

[0036] Figure 2 : A biomimetic "pearl shell" structure with embedded anti-oxidation silver hotspots; wherein: (a) Schematic diagram of the preparation of biomimetic “pearl shell” SERS substrate by green electrospinning; (b) Scanning electron microscope (SEM) images of AgNWs; (c) SEM images of AgNPs; (d) TEM images of AgNPs; (e) High-resolution TEM images of AgNPs; (f) SEM image of pure HPMC membrane; (g) SEM image of 15 wt.% AH-NFM loading; (h) SEM image of AH-NFM with a loading of 25 wt.%; (i) SEM image of 35 wt.% AH-NFM loading; (j) SEM image of AH-NFM with a loading of 45 wt.%; (k) TEM image of 45 wt.% AgNPs@HPMC nanofibers; (l) Figure 2 The distribution diagram of C elements corresponding to k in the diagram; (m) Figure 2 The distribution diagram of O elements corresponding to k in the diagram; (n) Figure 2 The distribution diagram of Ag elements corresponding to k in the diagram; (o) EDS energy spectrum of 45 wt.% AgNPs@HPMC nanofibers; (p) Size distribution of pure AgNPs; (q) Size distribution of pure HPMC nanofibers; (r) Size distribution of 45wt.% AH-NFM with a loading of 45wt.%.

[0037] Figure 3 Elucidation of the hydroxyl-mediated silver reduction and stabilization mechanism; among which: (a) XRD patterns of pristine AgNPs, pure HPMCs and AH-NFMs; (b) High-resolution Ag 3d XPS spectra of pristine AgNPs and AH-NFMs; (c) C1s XPS spectra of pure HPMCs and AH-NFMs; (d) O1s XPS spectra of pure HPMCs and AH-NFMs; (e) Raman spectra of pure HPMCs and AH-NFMs; (f) FT-IR spectra of pure HPMCs and AH-NFMs; (g) UV-Vis spectra of pristine AgNPs, pure HPMCs and AH-NFMs; (h) Schematic diagram of the reduction and anchoring mechanism of silver by HPMC.

[0038] Figure 4 10 -19 M-level SERS detection limit; where: (a) Schematic diagram of AH-NFM SERS substrate detection of 4-MBN / 4-ATP; (b) SERS spectrum of 4-MBN on AH-NFM with a loading of 15 wt.%; (c) SERS spectrum of 4-MBN on AH-NFM with a loading of 25 wt.%; (d) SERS spectrum of 4-MBN on AH-NFM with a loading of 35 wt.%; (e) SERS spectrum of 4-MBN on AH-NFM with a loading of 45 wt.%; (f) Figure 4 A magnified image of the low-concentration peak of e in the image; the signal-to-noise ratio (S / N) is calculated and determined using the following formula: S / N ≥ 3; where S represents the signal intensity (au) of the target characteristic Raman peak. The standard deviation (au) of baseline noise is expressed by the following formula:

[0039] In the formula: Baseline noise region (800~1000 cm) -1 The i-th measured intensity value (au) within ) ; The average value (au) of all intensity data within the baseline noise region; This represents the total number of intensity data points within the baseline noise region; n The degrees of freedom for calculating the sample standard deviation are 1, and the result is greater than 3. (g) Different concentrations (10) on AH-NFM with a loading of 45 wt.% -3 ~10 -19 SERS spectrum of M)4-ATP; (h) Figure 4 A magnified view of the low concentration peak of g in the image; (i) Normal Raman spectrum of solid 4-ATP, 10 on HPMC -3 SERS spectrum of M4-ATP and 10 on AH-NFM -3 Comparison of SERS spectra of M4-ATP; (j) 4-ATP concentration (10 -4 ~10 -9 M) and 1079cm-1 The linear relationship of SERS intensity at the concentration level (R²=0.901), and detailed data on the full concentration range and low concentration fluctuations can be found in (…). Figure 11 ); (k) 4-ATP on AH-NFM 1079cm -1 Raman mapping of characteristic peaks, and the corresponding optical images of the mapping regions are shown in ( ). Figure 13 ); (l) SERS intensity map of 20 random sites; (m) 20 random loci, 1079 cm -1 Comparison of SERS intensity of peaks (RSD=4.04%). (n) SERS spectra of 4-ATP before and after 6 months of storage in AH-NFM. Note: The detection limit of this surface-enhanced Raman scattering flexible substrate material for 4-ATP molecules can be achieved to 1.0 × 10⁻⁶ molecules through three independent experiments. -19 M; At the specified concentration, the signal-to-noise ratio (S / N) calculated from its Raman spectrum is greater than 3, indicating that the signal can be clearly identified and the results have good repeatability; (o) Performance comparison of AH-NFM with reported SERS substrates (detection limit).

[0040] Figure 5 The excellent molecular enrichment properties of hydrophilic three-dimensional nanofiber networks; among which: (a) SEM image of 2D-HPMC; (b) SEM images of 2D-Ag / HPMC; (c) SEM image of 3D-HPMC; (d) SEM images of 3D-AH-NFM; (e) SEM images of 3D-PVDF, with corresponding photographs of the samples shown in ( Figure 14 ); (f) Two-dimensional AFM image of 2D-HPMC; (g) Two-dimensional AFM images of 2D-Ag / HPMC; (h) Two-dimensional AFM images of 3D-HPMC; (i) Two-dimensional AFM images of 3D-AH-NFM; (j) Two-dimensional AFM image of 3D-PVDF, (k) Three-dimensional AFM images of 2D-HPMC; (l) Three-dimensional AFM images of 2D-Ag / HPMC; (m) 3D-HPMC three-dimensional AFM image; (n) Three-dimensional AFM images of 3D-AH-NFM; (o) Three-dimensional AFM image of 3D-PVDF; (p) Fluorescence microscopy image of 2D-HPMC after soaking in Rhodamine B (RhB) solution; (q) Fluorescence microscopy image of 2D-Ag / HPMC after soaking in Rhodamine B solution; (r) Fluorescence microscopy image of 3D-HPMC after immersion in Rhodamine B solution; (s) Fluorescence microscopy image of 3D-AH-NFM after immersion in Rhodamine B solution; (t) Fluorescence microscopy image of 3D-PVDF after soaking in Rhodamine B solution; (u) Statistical histogram of surface roughness (Ra) of different samples; (v) Water contact angle images of 3D-HPMC, 3D-AH-NFM and 3D-PVDF; (w) High-resolution XPS spectra of 3D-PVDF, 3D-HPMC and 3D-AH-NFM after soaking in 4-ATP solution (S 2p). (x) High-resolution XPS spectra (N 1s) of 3D-PVDF, 3D-HPMC and 3D-AH-NFM after soaking in 4-ATP solution. (y) Schematic diagram of the synergistic enrichment mechanism of AH-NFM (three-dimensional porous structure + hydrogen bonding + Ag-S chemisorption).

[0041] Figure 6 : A smart mask for respiratory diagnostics of multiple volatile organic compounds assisted by machine learning; among which: (a) Schematic diagram of SERS substrate integrated into smart mask to achieve exhaled gas detection; (b) Schematic diagram of the pre-enrichment of benzaldehyde molecules mediated by 3D-AH-NFM through physical confinement and hydrogen bonding; (c) Schematic diagram of chemical anchoring of 4-ATP with benzaldehyde via Schiff base reaction; (d) SERS spectrum of benzaldehyde; (e) SERS spectrum of methanol; (f) SERS spectrum of acetone; (g) A schematic diagram of the machine learning workflow for a random split strategy (70% training set and 30% test set); (h) Score map of spectral dimensionality reduction and key feature extraction achieved through PCA; (i) Confusion matrix of the SVM model; (j) Confusion matrix of the RF model; (k) Confusion matrix of CNN model.

[0042] Figure 7 Characterization of silver nanowires (AgNWs); where: (a) Scanning electron microscope (SEM) image of AgNWs at 10,000x magnification; (b) SEM image of AgNWs at 30,000x magnification; (c) SEM image of AgNWs at 50,000x magnification; (d) Figure 7 The silver element X-ray energy dispersive spectroscopy (EDS) mapping corresponding to the SEM image in a; (e) X-ray diffraction (XRD) pattern of AgNWs; (f) Detection of surface-enhanced Raman scattering (SERS) spectra of 4-mercaptobenzonitrile on AgNWs-hydroxypropyl methylcellulose (HPMC) surface-enhanced Raman scattering flexible substrate (NFM); (g) Figure 7 f in 10 -14 Locally magnified SERS spectrum of M concentration sample.

[0043] Figure 8 Preparation process of AgNPs-HPMC surface-enhanced Raman scattering flexible substrate material (AH-NFM); wherein: (a) AgNPs were synthesized by hydrothermal method and the synthesized AgNPs were dispersed in anhydrous ethanol; (b) Prepare spinning solution by mixing AgNPs dispersion with HPMC; (c) Electrospinning was performed using the prepared AgNPs-HPMC spinning solution; (d) Collect the prepared AH-NFM on aluminum foil.

[0044] Figure 9 : SEM images of AH-NFM at high magnification (50,000x); where: (a) SEM image of 15 wt.% AH-NFM loading; (b) SEM image of 25 wt.% AH-NFM loading; (c) SEM image of 35 wt.% AH-NFM loading; (d) SEM image of 45wt.% AH-NFM loading.

[0045] Figure 10X-ray photoelectron spectroscopy (XPS) full spectra of different samples; where: (a) HPMC; (b) AgNPs; (c) 15 wt.% AH-NFM loading; (d) 25 wt.% AH-NFM loading; (e) 35 wt.% AH-NFM loading; (f) 45 wt.% AH-NFM loading.

[0046] Figure 11 HPMC Surface-Enhanced Raman Scattering Flexible Substrate Material Testing 10 -12 The original normal Raman spectrum of M4-ATP.

[0047] Figure 12 : 4-Aminothiophenol (4-ATP) at different concentrations at 1079 cm⁻¹ -1 The relationship between SERS peak intensity and concentration at a given location; where: (a) SERS peak intensity and 4-ATP concentration (10) -2 M to 10 -19 The correlation of M); (b) Linear interval (10 -4 M to 10 -9 M); (c) Low concentration range (10 -9 M to 10 -19 Intensity fluctuations of M).

[0048] Figure 13 : Optical image corresponding to Raman mapping (scanning area: 30μm×30μm).

[0049] Figure 14 Photographs of membrane materials prepared by drop casting or electrospinning (size: 5×5cm) 2 );in: (a) HPMC two-dimensional membrane; (b) Ag / HPMC two-dimensional membrane; (c) HPMC three-dimensional membrane; (d) AH-NFM three-dimensional membrane; (e) Polyvinylidene fluoride (PVDF) three-dimensional membrane.

[0050] Figure 15 : Original surface roughness data of different substrate membrane materials; among which: (a) HPMC two-dimensional membrane; (b) Ag / HPMC two-dimensional membrane; (c) HPMC three-dimensional membrane; (d) AH-NFM three-dimensional membrane; (e) PVDF three-dimensional membrane.

[0051] Figure 16 Data preprocessing; where: (Category 0: blank group, Category 1: low concentration group, Category 2: medium concentration group, Category 3: high concentration group). (a) Raw SERS spectra of all samples; (b) Average raw SERS spectra of each group; (c) Baseline-corrected SERS spectra of all samples; (d) Average baseline-corrected SERS spectra for each group; (e) Standardized SERS spectra of each group; (f) Principal Component Analysis (PCA) Loading Plot; The data were derived from baseline-corrected SERS spectra through Z-score normalization. Principal component 1 (PC1) and principal component 2 (PC2) explained 32.6% and 16.4% of the variance, respectively, cumulatively covering 49.0% of the original spectral information variation and characterizing the core spectral features.

[0052] Figure 17 Performance of different machine learning models; where: (a) Feature importance of Support Vector Machine (SVM); (b) Learning curve of SVM (accuracy of 5-fold cross-validation); (c) Receiver Operating Characteristic (ROC) curve of SVM; (d) Feature importance of Random Forest (RF); (e) Learning curve of RF (200 decision trees); (f) ROC curve of RF; (g) Feature importance of Convolutional Neural Networks (CNNs); (h) The learning curve of CNN (30 training epochs); (i) ROC curve of CNN. Detailed Implementation

[0053] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below in conjunction with the embodiments of this invention. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0054] The technical solution of the present invention will be described in detail below. All reagents used in the following examples are of analytical grade and can be used without further purification. Among them: hydroxypropyl methylcellulose (HPMC, molecular weight ≈120000), polyvinylpyrrolidone (PVP, molecular weight ≈58000), rhodamine B (C 28 H 31ClN2O3 was purchased from Acros Organics; ethylene glycol (C2H6O2, purity ≥99.5%), anhydrous ethanol (C2H6O, purity ≥99.7%), silver nitrate (AgNO3, purity ≥99.8%), and acetone (C3H6O, purity ≥99.5%) were purchased from Sinopharm Chemical Reagent Co., Ltd.; 4-mercaptobenzonitrile (C7H5NS, purity 95%) was purchased from Bide Pharmaceutical Technology Co., Ltd.; and 4-aminothiophenol (C6H7NS, purity 97%) was purchased from Bailingwei Technology Co., Ltd. Deionized water prepared using a Milliporedirect-Q system was used throughout the experiment.

[0055] The characterization methods and instruments involved in the following examples are as follows: morphology and structure of samples were characterized using scanning electron microscopy (SEM, Hitachi SU5000) and transmission electron microscopy (TEM, JEOL JEM-2100P); crystal structure was analyzed using X-ray diffraction (XRD, Panaco Empyrean); chemical states were determined using X-ray photoelectron spectroscopy (XPS, Shimadzu AXISSUPRA); infrared absorption spectra were recorded using Fourier transform infrared spectroscopy (FTIR, Thermo Fisher Scientific Nicoletti S50); ultraviolet-visible-near-infrared absorption spectra were obtained using ultraviolet-visible-near-infrared spectrophotometer (Shimadzu SolidSpec-3700); fluorescence images were captured using laser scanning confocal microscopy (LSCM, Nikon AX); and two-dimensional and three-dimensional surface roughness and topology of samples were observed using atomic force microscopy (AFM, Shimadzu SPM-9700).

[0056] In the following embodiments, a confocal micro-Raman imaging system (Horiba, LabRAM Odyssey) was used for SERS testing, with an excitation wavelength of 532 nm. All SERS spectra were acquired using a 50x objective lens, with laser power attenuated to 1%. The spectral acquisition parameters for single-point tests were set as follows: integration time 5 seconds, cumulative 3 times; SERS mapping images were obtained by monitoring the intensity of characteristic Raman peaks, with a step size of 1 μm, an integration time of 1 second per pixel, cumulative 3 times. All experiments were independently repeated 3 times, and the obtained spectral data were baseline corrected to eliminate background interference, and the average value was presented.

[0057] Example 1: Determining the optimal morphology of SERS active units The size, shape, and spacing of metallic nanostructures are key parameters determining their local surface plasmon resonance (SERS) effect; different morphologies produce drastically different electromagnetic field enhancement (SERS) effects. To determine the optimal morphology of the active SERS unit, this embodiment prepared silver nanoparticles (AgNPs), silver nanowires (AgNWs), and silver nanoparticle-silver nanowire composites (AgNP-AgNWs). (1) Hydrothermal synthesis of AgNPs: S1. Prepare 3 mL of silver nitrate ethylene glycol solution (0.67 M) and 3 mL of polyvinylpyrrolidone ethylene glycol solution (2 M) respectively, and stir vigorously until the solid is completely dissolved.

[0058] S2. Mix the two solutions prepared in S1, then add 1 mL of ethylene glycol and 2 mL of glycerol, and stir continuously at 60°C until the solution is evenly mixed.

[0059] S3. Transfer the well-mixed solution from S2 to a stainless steel reactor with a polytetrafluoroethylene liner, and heat it in an oven at 160°C for 2 hours. After heating, allow the reactor to cool naturally to room temperature to obtain an AgNPs suspension.

[0060] S4. Centrifuge the AgNPs suspension obtained in S3 at 10,000 rpm for 10 minutes, and wash twice each with deionized water and ethanol to remove excess PVP and unreacted reagents. Finally, redisperse the AgNPs in ethanol and store at 4°C for later use.

[0061] (2) Hydrothermal synthesis of AgNWs: The steps and parameters are basically the same as in (1), except that in S2, 1 mL of NaCl ethylene glycol solution (0.17 M) and 2 mL of glycerol are added, and the concentration of polyvinylpyrrolidone ethylene glycol solution is 1.5 M.

[0062] (3) Hydrothermal synthesis of AgNP-AgNW: The steps and parameters are basically the same as in (2), the only difference being that the concentration of the polyvinylpyrrolidone glycol solution is 2M.

[0063] like Figure 2 As shown in b, (2) AgNWs with a length of 5~10μm and a diameter of 50~80nm were obtained; Figure 2 As shown in c~e, (1) AgNPs with uniform dispersion and polyhedral structure were obtained. The above results show that in the above method for synthesizing silver nanomaterials, the concentration of PVP and whether or not NaCl is added are the key morphology regulating factors. Figure 2 (a) in the middle.

[0064] like Figure 7 As shown, the isotropic geometry of AgNPs is conducive to the formation of dense and uniform electromagnetic "hot spots", while the hot spots of AgNWs are mainly concentrated at random intersections, resulting in poor signal reproducibility. It is speculated that the abundant exposed crystal planes and edge sites of the polyhedral AgNPs are the key to generating a strong LSPR effect. Therefore, AgNPs with polyhedral structures were finally selected for the preparation of subsequent materials.

[0065] Example 2: 1. Preparation of spinning membrane (1) Preparation of AH-NFM with an AgNPs loading of 45 wt.%: S1. Mix 5 g of water and 5 g of ethanol, add 0.2 g of HPMC, and magnetically stir at 25°C until the HPMC is completely dissolved. Then add 0.09 g of AgNPs prepared in Example 1 to the solution and continue stirring at 25°C for 3 hours to obtain the spinning solution. Note: The AgNPs loading of 45 wt.% here is calculated based on the ratio of the mass of AgNPs to the mass of solid HPMC, i.e., (0.09 g / 0.2 g) × 100% = 45 wt.%. All AgNPs loadings in this article are calculated in this way.

[0066] S2. The spinning solution obtained in S1 was loaded into a 10 mL syringe equipped with an 18-gauge needle for electrospinning. A voltage of 25 kV was applied, the solution flow rate was controlled at 1 mL / h, the rotation speed of the rotating collector wrapped with aluminum foil was set to 200 rpm, and the distance between the needle and the collector was fixed at 13 cm. After 8 hours of electrospinning, the nanofibers were carefully removed from the collector and vacuum dried at 60 °C for 4 hours to obtain a spun membrane, named AH-NFM.

[0067] (2) Preparation of pure HPMC membrane The preparation steps and parameters are basically the same as those in (1), the only difference being that AgNPs were not added to the spinning solution.

[0068] (3) Preparation of AH-NFM with different AgNP loadings (15 wt.%, 25 wt.%, 35 wt.%): The preparation steps and parameters are basically the same as in (1), the only difference being the amount of AgNPs added to the spinning solution.

[0069] 2. Characterization: All prepared spun films were cut to the required size and characterized by scanning electron microscopy (SEM), transmission electron microscopy (TEM), X-ray diffraction (XRD), X-ray photoelectron spectroscopy (XPS), Fourier transform infrared spectroscopy (FTIR), Raman spectroscopy, ultraviolet-visible spectroscopy (UV-Vis), atomic force microscopy (AFM), and contact angle measurement. like Figure 2 As shown in p~q, the average diameter of HPMC fibers (pure HPMC membrane fiber diameter: 132 nm) is... Figure 2 The q in Example 1 and the AgNPs prepared in Example 1 (average size: 176 nm) Figure 2The p-type ions in HPMC have similar nanoscale dimensions. Under the anchoring effect of the abundant hydroxyl groups in HPMC, most AgNPs are uniformly embedded within the fiber, forming a stable biomimetic "pearl shell" structure. For example... Figure 2 As shown in f~j, the gaps between AgNPs gradually decrease with increasing AgNP loading. The average fiber diameter of AH-HPMC with an AgNP loading of 45 wt.% increases to 248 nm. Figure 2 The average fiber diameter (132 nm) of the membrane was significantly increased compared to that of the pure HPMC membrane, directly proving the successful loading of AgNPs. Notably, even at a high AgNP loading of 45 wt.%, the self-assembled three-dimensional fiber network, with its inherent porous structure, effectively limited AgNP aggregation and promoted the formation of high-density, uniformly dispersed SERS active "hot spots." Figure 9 ).

[0070] TEM image ( Figure 2 The k) visually confirms that AgNPs are uniformly distributed inside HPMC fibers, indicating that AgNPs are successfully embedded, ensuring their long-term stability during storage and use.

[0071] Figure 2 The presence of l~o further validated the successful encapsulation and uniform dispersion of Ag in the fibrous membrane, with no impurity elements detected. Simultaneously, this three-dimensional fibrous network, through its high specific surface area and abundant hydrogen bonding sites (-OH), enhanced the adsorption of trace analytes. These structural and chemical properties collectively underpin the high sensitivity and stability of the resulting SERS substrate.

[0072] To elucidate the dual role of -OH in silver redox and AgNPs stabilization in HPMC, this invention presents a systematic structural characterization. Figure 3 In the figure, 'a' represents the XRD pattern of the synthesized AgNPs. Obvious diffraction peaks appear at 2θ = 38.3°, 44.5°, 64.6°, and 77.5°, corresponding to the face-centered cubic (FCC) structure of metallic silver (Ag). 0 The (111), (200), (220), and (311) crystal planes of Ag (JCPDS No. 04-0783). This is consistent with the 0.236 nm lattice fringe spacing observed by high-resolution transmission electron microscopy (HRTEM) (corresponding to Ag). 0 The (111) crystal plane; Figure 2 The consistency of 'e' in the data confirms the successful synthesis of crystalline metallic Ag via a hydrothermal method. 0Nanoparticles. Notably, the pristine AgNPs exhibit a weak but identifiable diffraction peak at 2θ = 32.5°, attributed to the (111) plane of silver oxide (Ag₂O) (JCPDS No. 41-1041), indicating slight surface oxidation of the AgNPs upon exposure to air—a common phenomenon resulting from the high reactivity of surface atoms. The low peak intensity suggests that the oxidation is confined to a thin surface layer and does not form a bulk oxide.

[0073] Pure HPMC films exhibit a broad diffraction peak at 2θ = 23°, characterizing their predominantly amorphous structure with a small amount of microcrystals, which originates from the random entanglement of HPMC molecular chains. After recombination with AgNPs, AH-NFMs show sharp Ag peaks at 2θ = 38.1°, 44.3°, 64.4°, and 77.3°. 0 Diffraction peaks ( Figure 3 (a) confirms that AgNPs were successfully embedded in HPMC fibers while maintaining their crystallinity. Notably, the intensity of the (111) crystal plane peak (38.08°) increases linearly with the increase of AgNPs loading and remains the strongest diffraction peak, indicating that AgNPs maintain a (111) preferred orientation in the composite material, consistent with the uniform dispersion of AgNPs within the fibers observed by SEM. Crucially, the characteristic peak of Ag2O (32.5°) present in the original AgNPs completely disappears in the composite material, directly demonstrating that the -OH group of HPMC reduces surface Ag2O to Ag. 0 This achieves "chemical repair" of silver oxide.

[0074] The elemental composition and chemical interactions of the samples were analyzed by XPS characterization. Figure 10 High-resolution XPS spectra ( Figure 3 The analysis of Ag3d core energy levels (b~d) elucidates the interaction between AgNPs and HPMC and the changes in silver chemical states. Figure 3 b) shows that the original AgNPs, except at 368.5 eV (3d 5 / 2 ) and 374.5eV (3d 3 / 2 The dominant Ag appears at ) 0 Outside the peak, at 367.1 eV (3d 5 / 2 ) and 373.1eV (3d 3 / 2 A weaker peak also appears at this location, which is attributed to Ag in Ag2O. + The species are consistent with the surface oxidation shown by XRD. In contrast, the Ag3d spectra of AH-NFMs only show symmetrical Ag. 0 Bimodal, Ag + The signal completely disappeared, further confirming that the -OH group of HPMC will block Ag. + Restored to Ag 0In HPMC, the -OH group, which is adjacent to the ether bond and has higher reactivity, can act as an electron donor during the preparation process, promoting the reduction reaction (Ag). + +e - →Ag 0 ).

[0075] C1s and O1s region analysis ( Figure 3 Further evidence is provided in sections c and d). The C1s spectrum of the pure HPMC film can be unconvoluted to three peaks: 284.8 eV (CC), 286.4 eV (CO, from the -OH / ether bond), and 287.6 eV (OCO). The O1s spectrum shows a main peak at 532.9 eV, attributed to oxygen in the COC / C-OH group. Compared to the pure HPMC film, the CO peak in the C1s spectrum of AH-NFMs is slightly shifted to a higher binding energy (≈286.5 eV, Δ=+0.1 eV), while the COC / C-OH peak in the O1s spectrum is shifted to a lower binding energy (≈532.8 eV, Δ=-0.1 eV). This bidirectional shift originates from electron transfer: the lone pair electrons of the O atom (in the -OH group) are transferred to Ag. 0 On the surface, the electron density of O decreases (leading to a decrease in the O 1s binding energy) and the electron density of adjacent C atoms increases (leading to an increase in the C 1s binding energy of CO). This directly demonstrates electron transfer between HPMC and AgNPs at the electronic level, strongly supporting the "hydroxyl-mediated reduction" mechanism.

[0076] Fourier transform infrared (FTIR) spectroscopy Figure 3 f) reveals the chemical interaction between AgNPs and HPMC. Pure HPMC films exhibit a characteristic peak: 3401 cm⁻¹. -1 (Broad peak, -OH stretching vibration), 2912 cm⁻¹ -1 (CH stretching vibration) and 1057cm -1 (Asymmetric COC stretching vibration). OH stretching bands of AH-NFMs (3200~3600cm) -1 A significant redshift (shift to lower wavenumbers) and decreased intensity were observed. This redshift is a hallmark of O-Ag coordination: electron transfer from O to Ag reduces the polarity and strength of the OH bond, leading to a decrease in vibrational frequency; the decreased peak intensity indicates that some free -OH groups participate in coordination and no longer exist in isolated form. This directly proves the interaction between -OH and Ag, revealing the coordination anchoring mechanism of HPMC on AgNPs and inhibiting nanoparticle aggregation.

[0077] Raman spectroscopy provides more in-depth structural information. Figure 3 (e). Characteristic peaks of pure HPMC membranes include: 2896 cm⁻¹ -1 and 2935cm -1(Split peak, attributed to CH stretching vibrations of -CH3 and -CH2- caused by molecular chain orientation driven by electrospinning), 1123 cm⁻¹ -1 (COC stretching vibration of HPMC main chain), 950cm -1 and 895cm -1 (Glucose ring vibration, the latter being a characteristic peak of β-glycosidic bonds), 1455 cm⁻¹ -1 (G-band) and 1369cm -1 (D band, attributed to the bending vibrations of -CH2- and -CH3 respectively), 299cm -1 (Intermolecular hydrogen bonds and lattice vibrations reflect the close packing of polymer networks). In AH-NFMs, 299 cm⁻¹ -1 The complete disappearance of the hydrogen bond-related peak indicates that the introduction of AgNPs disrupted the original intermolecular hydrogen bonds in HPMCs, with -OH preferentially coordinating with Ag. Simultaneously, a new 241 cm⁻¹ peak appeared in AH-NFMs. -1 The peak is attributed to the Ag-O coordination vibration. This further confirms at the molecular vibrational level that the -OH group of HPMC forms a coordination bond with AgNPs, strengthening the "anchoring stability" mechanism. Furthermore, the intensity ratio of the D band to the G band of AH-NFMs (ID / IG=I) 1369 / I 1455 The ratio is significantly lower than that of pure HPMC films. This ratio is usually related to the crystallite size and packing density of such materials, and its reduction indicates that although the introduction of AgNPs slightly disrupts the ordered packing of HPMC chains, it ultimately forms a more stable composite structure through coordination bonds.

[0078] Figure 3 The image shows the UV-Vis spectra of pure HPMC films and AgNPs / HPMC. Pure HPMC films show no obvious absorption peaks (consistent with their amorphous nature), while pristine AgNPs exhibit an absorption peak near 300 nm, corresponding to interband electronic transitions in AgNPs. AH-NFMs display two characteristic peaks: the peak at 300 nm is still related to the electronic transitions of AgNPs, and the peak at 395 nm corresponds to the LSPR effect of AgNPs. Notably, the intensity of the LSPR peak at 395 nm increases linearly with the mass loading of AgNPs, and the peak shape is symmetrical without broadening. The increased intensity confirms that higher concentrations of AgNPs produce a stronger LSPR effect. These optical properties verify the effective stabilization of AgNPs by the anchoring effect of HPMC, laying the structural foundation for the SERS performance enhancement discussed in subsequent chapters.

[0079] In summary, through synergistic analysis using XRD, XPS, FTIR, Raman spectroscopy, and UV-Vis, this invention reveals that HPMC possesses a dual function of reduction and anchoring for AgNPs. HPMC not only reduces surface silver oxide to metallic silver but also effectively anchors AgNPs through coordination (mechanism diagram shown in Figure 1). Figure 3 (as shown in h in the figure), thus solving the long-standing oxidation and aggregation problems of AgNPs.

[0080] Example 3: To evaluate the SERS activity of AH-NFMs, this invention designed 4-mercaptobenzonitrile (4-MBN) and 4-aminothiophenol (4-ATP) as probe molecules and investigated their detection sensitivity. Figure 4 (a) in the middle.

[0081] (1) Sample processing AH-NFM was immersed in 1 mL of 4-MBN ethanol solution of different concentrations and incubated at 4°C for 2 hours to obtain AH-NFM@4-MBN, which was then stored at 4°C for use in SERS detection.

[0082] AH-NFM was immersed in 1 mL of 4-ATP ethanol solution of different concentrations and incubated at 4°C for 2 hours to obtain AH-NFM@4-ATP, which was then stored at 4°C for use in SERS detection.

[0083] Figure 4 Figures b to e in the spectrum show the SERS response of AH-NFMs with different AgNP loadings to 4-MBN. Four strong characteristic peaks appear in the spectrum: 2228 cm⁻¹ -1 (C≡N stretching vibration, νC≡N), 1585cm -1 (Benzene ring C=C stretching vibration, νC=C), 1180cm -1 (Symmetric CN stretching vibration, νsC-N) and 1077cm -1 (In-plane C-C shear vibration of the benzene ring, δC-C). The limit of detection (LOD) of AH-NFM with 15 wt.% AgNPs loading for 4-MBN is 10. -9 The detection limits of AH-NFM for M, AgNPs loading of 25 wt.% and AgNPs loading of 35 wt.% both reached 10. -14 M ( Figure 4 (b~d); 45wt.%AH-NFM performed best, with the detection limit further reduced to 10. -15 M ( Figure 4(e~f in the original text). These results confirm that all four types of AH-NFM possess SERS activity. The detection limit significantly increased in the AgNP loading range of 15 wt.% to 25 wt.%, because a rich and effective three-dimensional "hot spot" network had not yet formed before the AgNP loading reached 25 wt.%. Although the SERS activity gradually increased with increasing AgNP loading, the increase was limited. Considering that excessive AgNPs not only increases the preparation cost but may also promote nanoparticle aggregation and disrupt fiber formation during electrospinning, AH-NFM with an AgNP loading of 45 wt.% was ultimately selected as the optimal substrate for further in-depth research.

[0084] Using 4-ATP as the probe molecule, the maximum SERS sensitivity of AH-NFM was investigated, and its detection limit and enhancement mechanism (electromagnetic enhancement (EM) / chemical enhancement (CM)) were systematically analyzed. The key structural difference between 4-ATP and 4-MBN lies in the terminal functional groups: amino (-NH2) and cyano (-CN), which form hydrogen bonds with the hydroxyl-rich cellulose matrix with different strengths. Figure 4 The g~h in the figure shows 4-ATP at different concentrations (10 -3 Up to 10 -19 The SERS spectrum on AH-NFM under M) shows five strong characteristic peaks: 1578 cm⁻¹ -1 1435cm -1 1391cm -1 1144cm -1 and 1079cm -1 They are respectively attributed to the C=C stretching vibration of the benzene ring (ν8a), the coupled vibration of C=C stretching and CH in-plane bending of the benzene ring (ν19b), the coupled vibration of CH in-plane bending and CC stretching of the benzene ring (ν3), the CH in-plane bending vibration of the benzene ring (δ9b), and the CS stretching vibration of the benzene ring side chain (ν7a).

[0085] Comparison of the normal Raman spectrum of solid 4-ATP with that of AH-NFM soaked in 10 -3 SERS spectrum after M 4-ATP ( Figure 4 The i) can provide deeper insights into adsorption behavior and enhancement mechanisms. The b2 vibrational mode peak in the SERS spectrum (1435 cm⁻¹) -1 1391cm -1 1144cm -1 The enhancement was significant. According to the surface selection rule, the b2 mode only enhances when charge transfer (CT) occurs between the substrate and the probe, confirming that the energy levels of AgNPs (Fermi level ≈ -4.8 eV) and 4-ATP (highest occupied molecular orbital (HOMO) ≈ -5.0 eV) are matched, enabling efficient charge transfer. Furthermore, peak shift phenomena (e.g., 1078 cm⁻¹) were observed. -1The solid 4-ATP is 9 cm away from the solid. -1 This indicates that the -SH group of 4-ATP forms a direct Ag-S bond with AgNPs, further promoting charge transfer. The a1 mode peak (1578 cm⁻¹) -1 1079 cm -1 ) dominates in SERS spectra, especially at low concentrations (e.g., 10). -15 ~10 -19 The effect is more pronounced under M). Since the a1 mode is closely related to the electromagnetic enhancement-induced local electromagnetic field enhancement, it confirms that EM is the dominant enhancement mechanism of AH-NFM—attributed to the three-dimensional "hot spots" formed by uniformly dispersed AgNPs in HPMC fibers. Therefore, 1079 cm⁻¹ was selected. -1 The enhancement factor (AEF) of the strongly a1 mode peak is calculated using the following formula: AEF=(I SERS / I Raman )×(C Raman / C SERS ) Among them, I SERS and I Raman The 1079 cm⁻¹ values ​​in the SERS and normal Raman spectra are shown in order. -1 Peak intensity at C SERS and C Raman These represent the molar concentrations of the corresponding target molecules.

[0086] like Figure 12 As shown in b, the concentration of 4-ATP is related to the SERS signal intensity at 10 -4 ~10 -9 A good linear relationship is observed within the range of M (R 2 =0.90), the signal intensity is proportional to the concentration within the linear interval, which can accurately reflect the enhancement ability of the substrate, therefore C is determined. SERS =10 -9 M is the detection limit; when C SERS =10 -9 M, C Raman =10 -2 When M is reached, AEF is calculated to be approximately 1.46 × 10⁻⁶. 8 This is two orders of magnitude higher than the detection limit reported recently, confirming that AH-NFM has a strong SERS enhancement effect.

[0087] Significant signal fluctuations at low concentrations are a typical characteristic of single-molecule SERS (SM-SERS) studies. Although the AH-NFM substrate possesses a rich three-dimensional "hotspot" network, the adsorption of analytes at ultra-low concentrations is extremely low. Furthermore, the randomness and dynamic orientation changes of molecules occupying hotspot regions lead to spatiotemporal fluctuations in SERS intensity. This causes the signal intensity distribution to shift from a symmetric Gaussian distribution at high concentrations to an asymmetric long-tailed non-Gaussian distribution at low concentrations. Figure 11 This manifests as a mixture of low-intensity signals (molecules located outside strong hot spots) and high-intensity signal events (molecules accidentally located inside strong hot spots).

[0088] Reliable SERS quantitative analysis requires a linear relationship between signal intensity and analyte concentration. Based on the Langmuir adsorption isotherm, the relationship between surface coverage (Γ, a physical quantity directly detected by SERS) and solution concentration (C) can be described as follows: Where Γ0 is the maximum surface coverage, K a This is the adsorption equilibrium constant. At K... a In the low concentration range where C << 1, the formula simplifies to Γ ≈ Γ0K. a C indicates a linear relationship between Γ and C, corresponding to the theoretical linear response range of SERS. Analysis at 1079 cm⁻¹ -1 Peak intensity was used to establish a quantitative relationship, and it was found that at 10 -4 Up to 10 -9 It exhibits a good linear response within the M concentration range (R² = 0.90). Figure 4 The value of j in the equation indicates that the signal enhancement of hotspot adsorbed molecules within this range is stable, with no obvious saturation or fluctuation effects. It is noteworthy that even at 10... -19 At ultra-low concentrations of M, 1079 cm⁻¹ -1 The characteristic peaks are still clearly discernible, directly confirming that AH-NFM has extremely high SERS activity, setting a new record of 10. -19 The detection limit was recorded at the M level. The detection limit of 4-ATP was significantly lower than that of 4-MBN, mainly due to the stronger hydrogen bonding between the amino group (-NH2) of 4-ATP and the -OH group of HPMC, which promoted the efficient enrichment of the molecule on the fiber matrix. This mechanism will be further verified in the next section.

[0089] Based on 1079 cm -1 Raman mapping of peak intensity ( Figure 4 The k-values ​​in the image show that the SERS active regions on the substrate surface are spatially uniform, confirming the uniformity of the interactions and the consistency of the membrane performance. Comparison of the SERS spectra at 1079 cm⁻¹ from 20 randomly selected sites further confirms this. -1 Peak intensity ( Figure 4The reproducibility was verified using l and m, and the relative standard deviation (RSD) was calculated to be 4.037%. This RSD is far below the 20% threshold of the quantitative SERS substrate, indicating excellent spatial reproducibility and providing assurance for practical applications.

[0090] AgNPs are easily oxidized, leading to a gradual decline in SERS-enhancing activity. In this invention, the HPMC shell of AH-NFM provides dual protection: (1) a physical barrier to isolate oxygen / moisture; (2) the abundant -OH groups on the HPMC protect trace amounts of Ag. + Chemical reduction was performed. To verify the antioxidant stability of AH-NFM, the same substrate was stored at room temperature (≈25℃) and humidity for 6 months, and the SERS of 4-ATP was measured. The results showed that the detection limit remained at 10. -19 M, 1079cm -1 The characteristic peak intensity decreased by only 3.93%, confirming that AH-NFM has excellent long-term stability. Figure 4 (n in the text).

[0091] Comparing AH-NFM with cellulose-based SERS substrates (typical detection limit 10) reported in recent top journals (such as *Advanced Materials* and *Nature Communications*) -12 ~10 -9 M) and noble metal-based flexible substrates (typical detection limit 10). -15 ~10 -12 Comparative analysis of M) Figure 4 (o in the text). This comparison quantifies the 3-10 order of magnitude improvement in sensitivity achieved in this study, establishing AH-NFM's international leading position in the field of trace SERS detection. These findings indicate that HPMC-encapsulated AgNPs form a surface-enhanced Raman scattering flexible substrate material, which is a highly efficient SERS substrate with broad application prospects in fields such as biosensing.

[0092] Thiophenol functionalization of AH-NFM and SERS testing of volatile organic compounds (VOCs): AH-NFM was functionalized with 4-aminothiophenol (4-ATP) to construct a SERS substrate for VOCs detection: AH-NFM was immersed in 1 mL of 4-ATP ethanol solution (0.01 M) and incubated at 4 °C for 6 hours to obtain 4-ATP modified SERS substrate.

[0093] The 4-ATP-modified SERS substrate was placed in a 1L sealed glass vial along with 0.1 μL, 1 μL, or 10 μL of the target analyte (benzaldehyde, methanol, or acetone). After incubation at 35°C for 6 hours, SERS analysis was performed. The equilibrium vapor concentration was calculated as follows: First, according to the formula m=V liq Calculate the mass m of the target liquid (benzaldehyde) using the formula ×ρ (where V is the mass of the target liquid). liq (where ρ is the liquid volume and ρ is the density of benzaldehyde at the experimental temperature); then, the amount of substance n is calculated using the formula n=m / M (where M is the molar mass of benzaldehyde); the equilibrium pressure P generated by benzaldehyde vapor in the sealed container is calculated using the ideal gas law PV=nRT (where P is the equilibrium pressure in the sealed container, in Pa; T is the absolute temperature, in K; V...). vap Steam volume, unit: m 3 n is the amount of benzaldehyde, in mol; R is the ideal gas constant, with a value of 8.314 J·mol. -1 ·K -1 Given the experimental temperature T and equilibrium pressure P, the steam volume V can be further derived using the above equation. vap Finally, by comparing the steam volume V vap Total volume V of the sealed container total The volume fraction of steam φ (φ=V) is obtained. vap / V total ), then multiply by 10 6 Convert to ppm concentration (1ppm=10) 6 Volume fraction).

[0094] Based on the above steps, the simplified formula for calculating the ppm concentration of benzaldehyde vapor is: .

[0095] Example 4: Excellent molecular enrichment properties of hydrophilic three-dimensional nanofiber networks 1. To clarify the influence of three-dimensional structure and wettability on enrichment behavior, a series of control experiments were designed in this embodiment: (1) Preparation of pure HPMC two-dimensional membranes (abbreviated as 2D-HPMC) by drop casting: S1. Take 0.2 g HPMC, add a mixed solvent consisting of 5 g water and 5 g ethanol, and stir magnetically at 25°C for 2 hours until HPMC is completely dissolved to form a homogeneous HPMC solution.

[0096] S2. Use tape to mark a 5×5 cm rectangular area on the surface of a clean glass plate. Slowly pour the HPMC solution into the rectangular area, ensuring that the HPMC solution evenly covers the surface of the glass plate.

[0097] S3. Place the glass plate in a vacuum drying oven at 60°C and dry for 6 hours to remove the solvent. Carefully peel off the tape to obtain 2D-HPMC, and seal it for later use.

[0098] (2) Preparation of Ag / HPMC two-dimensional membranes (abbreviated as 2D-Ag / HPMC) by drop casting: S1. Take 0.2 g of HPMC and 0.09 g of AgNPs (particle size 100~250 nm) prepared in Example 1, add them together to a mixed solvent consisting of 5 g of water and 5 g of ethanol, and stir magnetically at 25 °C for 2 hours to form a uniformly dispersed mixed solution.

[0099] S2. Using the same method as for preparing 2D-HPMC, a 5×5cm rectangular area is marked on the surface of a clean glass plate with tape. The above mixed solution is slowly poured into the rectangular area and spread evenly.

[0100] After vacuum drying at 3.60℃ for 6 hours, the tape was peeled off to obtain 2D-Ag / HPMC, which was then sealed and stored for later use.

[0101] (3) Preparation of PVDF-spun three-dimensional membranes (abbreviated as 3D-PVDF) by electrospinning (ES): S1. Take 1g of polyvinylidene fluoride (PVDF) powder, add it to 9g of N,N-dimethylformamide (DMF) solvent, and stir magnetically at 30℃ for 12 hours until the PVDF is completely dissolved to form a uniform PVDF solution.

[0102] S2. Let the PVDF solution stand for 6 hours to remove the bubbles generated during stirring, ensuring that it is free of impurities and bubbles.

[0103] S3. Load the PVDF solution into a 10 mL syringe equipped with an 18-gauge needle and perform electrospinning: apply a voltage of 17 kV, control the flow rate of the PVDF solution at 1 mL / h, rotate the receiver (wrapped in aluminum foil) at 200 rpm, keep the distance between the needle and the receiver at 14 cm, and maintain an ambient temperature of 30℃ and a humidity of 40%.

[0104] S4. After electrospinning, the fiber membrane is removed from the receiver and dried under vacuum at 60°C for 4 hours to obtain 3D-PVDF, which is then sealed and stored for later use.

[0105] like Figure 5 The SEM images shown in a~e in the figure show that the two two-dimensional membranes prepared by the drop casting method have a dense and non-porous structure, while all three-dimensional membranes prepared by the electrospinning method have a three-dimensional interconnected fiber network with micron-level pores (1~5μm), which can promote the permeation of analytes through capillary action.

[0106] Corresponding two-dimensional and three-dimensional atomic force microscopy (AFM) images ( Figure 5 f~j, k~o) and surface roughness statistics (Ra, Figure 5 The study revealed three key trends: the Ra values ​​of 3D-HPMC (165.98 nm) and 3D-AH-NFM (181.48 nm) were approximately 4 to 13 times higher than those of 2D-HPMC (12.69 nm) and 2D-Ag / HPMC (39.46 nm), attributed to the high specific surface area of ​​the electrospun network—maximizing physical adsorption and analyte capture through high specific surface area and capillary-driven permeation, resulting in superior performance compared to two-dimensional membranes; AgNPs doping further increased roughness by introducing micro / nano protrusions (2D-Ag / HPMC > 2D-HPMC, 3D-AH-NFM > 3D-HPMC); and 3D-PVDF, due to its larger fiber diameter, had the highest Ra value (548.58 nm). More importantly, the water contact angle test ( Figure 5 (v) Clearly distinguishes the wettability differences: 3D-PVDF exhibits significant hydrophobicity (142.5°), while all HPMC-based samples have significant hydrophilicity (≈70°). This difference plays a decisive role in molecular enrichment.

[0107] 2. Rhodamine B (RhB, a fluorescent polar probe) and 4-ATP (a polar SERS probe) were selected as model analytes, as their amino groups can interact with the -OH group of HPMC through hydrogen bonds.

[0108] (1) Rhodamine B (RhB) fluorescence assay (to assess the enrichment efficiency of polar molecules): S1. The pure HPMC membrane (hereinafter referred to as 3D-HPMC), the AH-NFM with AgNPs loading of 45wt.% (hereinafter referred to as 3D-AH-NFM), and the five samples (2D-HPMC, 2D-Ag / HPMC, and 3D-PVDF) prepared in Example 2 were cut into 5×5mm square pieces to ensure that the sample size was uniform.

[0109] S2. Immerse each of the cut samples separately in 1 mL of a 10% concentration solution. -6 In an ethanol solution of Rhodamine B (RhB) of M, the sample was incubated at 25°C for 6 hours to allow RhB molecules to be fully adsorbed onto the sample surface and inside.

[0110] S3. Take out the sample, quickly rinse the surface with anhydrous ethanol to remove unadsorbed free RhB molecules, and then place it in a vacuum drying oven at 60℃ for 2 hours to remove residual solvent.

[0111] S4. Fluorescence tests were performed on each sample using a laser scanning confocal microscope (LSCM, Nikon AX). The excitation wavelength was set to 532 nm. Fluorescence intensity and distribution images of the samples were acquired, and the enrichment capacity of different samples for RhB was compared.

[0112] (2) XPS test of 4-ATP adsorption (quantitative verification of enrichment mechanism): S1. Cut the three types of samples, 3D-HPMC, 3D-AH-NFM, and 3D-PVDF, into 5×5mm square pieces for later use.

[0113] S2. Immerse each sample separately in 1 mL of a 10% concentration solution. -2 In a 4-ATP ethanol solution of M, the sample was incubated at 25°C for 6 hours to allow 4-ATP molecules to bind to the sample via hydrogen bonds or chemical interactions.

[0114] S3. Remove the sample, rinse the surface with anhydrous ethanol to remove unbound 4-ATP, and vacuum dry at 60°C for 2 hours to remove the solvent.

[0115] S4. X-ray photoelectron spectroscopy (Shimadzu AXISSUPRA) was used to test each sample, focusing on acquiring high-resolution spectra of the S2p and N1s orbitals. The adsorption amount of 4-ATP by the sample was quantitatively analyzed by the intensity of characteristic peaks to verify the enrichment difference between hydrophilic and hydrophobic interfaces.

[0116] Fluorescence microscope images ( Figure 5 The enrichment efficiency was directly visualized using the p~t method: the electrospun three-dimensional samples (3D-HPMC, 3D-AH-NFM) exhibited strong and uniform fluorescence, significantly better than the drop-cast two-dimensional samples, confirming the physical confinement effect of the three-dimensional porous structure; in contrast, the hydrophobic 3D-PVDF showed almost no fluorescence signal, verifying the key finding—for the enrichment of polar analytes, hydrophilicity is more decisive than roughness. Although 3D-PVDF has high roughness, it cannot effectively capture polar molecules; while the hydrogen bonding between the -OH group of HPMC and the probe molecules achieves active "pre-enrichment," overcoming the core limitation of traditional superhydrophobic substrates.

[0117] Quantitative analysis of 4-ATP adsorption by XPS ( Figure 5The presence of w~x in the data further validates the enrichment advantage of the hydrophilic interface. The S2p and N1s signal intensities of 3D-PVDF are negligible, confirming that hydrophobic repulsion leads to extremely low adsorption. In contrast, 3D-HPMC exhibits distinct characteristic peaks, indicating effective enrichment driven by hydrogen bonds. Notably, 3D-AH-NFM exhibits the strongest signal response, attributed to a synergistic mechanism of "hydrogen bond-mediated pre-enrichment and Ag-S chemical anchoring": the HPMC framework enriches polar molecules at the interface through hydrogen bonds, subsequently forming Ag-S bonds with 4-ATPase, achieving molecular stabilization (preventing desorption).

[0118] The above results indicate that the synergistic enrichment mechanism of 3D-AH-NFM comprises three key components ( Figure 5 (y) (1) The three-dimensional network achieves physical adsorption and restriction through high specific surface area and capillary action; (2) The hydrophilic HPMC interface achieves pre-enrichment of target molecules through hydrogen bonds; (3) AgNPs firmly anchor the analyte to the hot spot region through Ag-S bonds. This multi-mode synergistic effect makes the detection limit of the AH-NFM substrate for 4-ATP as low as 10. -19 M overcomes the limitations of traditional hydrophobic substrates that rely solely on physical constraints.

[0119] This invention provides a new paradigm for the design of high-performance SERS substrates, utilizing hydrophilic interactions to achieve efficient molecular enrichment, laying a solid foundation for the high sensitivity performance of AH-NFM in advanced biosensing applications such as VOC detection in smart masks.

[0120] AH-NFM (AgNPs loading 45wt.%) functionalized with 4-ATP molecules is used as the core detection layer, replacing the middle filter layer of disposable masks and integrated into the main body of the mask. The substrate is cut into a sheet (5cm×5cm) that matches the middle filter layer of the mask. Utilizing its flexibility and breathability, it ensures wearing comfort while achieving efficient adsorption and detection of VOCs in exhaled air.

[0121] Simulated SERS detection process while wearing a mask: (1) Substrate pretreatment: AH-NFM was functionalized with 4-ATP and then immersed in 0.01M 4-ATP ethanol solution and incubated at 4°C for 6 hours. After that, it was taken out and air-dried for later use. (2) Simulated Adsorption Experiment: The functionalized AH-NFM was fixed at the mouth of a 1L sealed glass bottle. Then, 100μL, 10μL, and 1μL of benzaldehyde (one of the VOCs that are lung cancer biomarkers) were added to the bottle, respectively. After sealing, the entire device was placed in a constant temperature environment of 35℃ and left to stand for 2 hours to simulate the complete adsorption process of VOCs in exhaled air with the substrate when wearing a mask. Studies have shown that the concentration of benzaldehyde in human exhaled air has diagnostic reference value: in healthy individuals, it is approximately 0.01 ~ 0.1 mg / m³. 3 Within this range, the concentration in the exhaled breath of lung cancer patients can increase to 0.5 ~ 5 mg / m³. 3 The benzaldehyde addition amounts used in this experiment (1–100 μL, generating corresponding concentration gradients in a 1L container) were designed to cover and far exceed this physiological concentration range to comprehensively test the material's adsorption performance. Based on this experiment, this method can achieve highly sensitive detection of benzaldehyde, with a detection limit as low as 0.01 mg / m³. 3 It is fully capable of distinguishing between background concentrations in healthy individuals and abnormally elevated concentrations in lung cancer patients.

[0122] (3) Raman detection: The AH-NFM substrate, after being incubated at a constant temperature, was removed and placed directly on the sample stage of the Raman spectrometer without additional pretreatment (simulating the convenience of on-site detection). SERS spectra were acquired using a confocal micro Raman imaging system (excitation wavelength 532 nm, laser power 1%, integration time 5 seconds, 3 times in total), with a focus on analyzing the 1621 cm⁻¹ region. -1 (C=N stretching vibration, characteristic peak of Schiff base reaction) and 1079 cm⁻¹ -1 Peak intensity and peak shape of (CS stretching vibration).

[0123] Anti-interference test procedure: Using the same experimental conditions as the simulated wearing test, the specificity of AH-NFM for identifying interfering VOCs was verified.

[0124] (1) Selection of interfering substances: Methanol and acetone, common interfering substances in exhaled breath, were selected as test subjects. The anti-interference results are as follows: Figure 6 As shown in e and f, the spectral detection results are basically consistent with those of the pure probe molecule; (2) Adsorption experiment: Functionalized AH-NFM was fixed at the mouth of a 1L sealed glass bottle, and 100μL, 10μL, and 1μL of methanol or acetone were added respectively. The bottle was then sealed and incubated at 35℃ for 2 hours. (3) Spectral acquisition and analysis: SERS spectra were acquired using the same Raman detection parameters, and the presence of 1621 cm⁻¹ was observed. -1 Characteristic peaks are used to determine the impact of interfering substances on benzaldehyde detection.

[0125] The extracted SERS data is directly input into the machine learning model of Example 5 for identification and prediction.

[0126] Example 5: Machine Learning Models and Statistical Analysis (1) Data preparation process (SERS spectrum → model input vector): SERS spectral data needs to undergo multiple standardization steps to ensure the consistency and validity of the model input. The specific steps are as follows: S1. Raw Spectrum Preprocessing: Preprocessing the acquired raw SERS spectra (wavelength range 400~2000 cm⁻¹) - ¹) Smoothing and denoising are performed using the Savitzky-Golay filtering algorithm (window size 11 points, polynomial order 2) to eliminate instrument noise interference; S2. Baseline Correction: Baseline background is removed by adaptive iterative reweighted penalized least squares (airPLS) algorithm, preserving the characteristic peak signal of the analyte and avoiding the impact of baseline drift on feature extraction; S3. Spectral Standardization: The Z-score standardization method is used to normalize the calibrated spectrum, eliminating intensity differences between different detection batches and base batches, so that the data meets the normal distribution requirements for model training. S4. Characteristic Peak Extraction: Automatically selects the key characteristic peak positions of the target analyte (4-ATP characteristic peak: 1079 cm⁻¹). -1 1391cm -1 1435cm -1 1578cm -1 1144cm -1 Benzaldehyde characteristic peak: 1621 cm⁻¹ -1 1168cm -1 ), extract the peak intensity value of each feature peak, and form a 7-dimensional feature vector; S5. Feature Dimensionality Reduction: Principal Component Analysis (PCA) is used to reduce the dimensionality of the 7-dimensional feature vector, retaining the top 3 principal components that explain more than 97% of the cumulative variance, resulting in simplified 3-dimensional input features, reducing the computational complexity of the model and avoiding overfitting.

[0127] (2) Detailed description of the parameters of the optimal model (SVM) The Support Vector Machine (SVM) model is the core analysis model for the detection equipment. It uses a radial basis function (RBF) as the kernel function and optimizes the hyperparameters through a grid search method (search range: C∈[0.1,1,10,20,100], gamma∈[0.001,0.01,0.05,0.1,1]) combined with 5-fold cross-validation. The optimal parameter combination is determined as follows: The penalty parameter C=20.0 is used to balance the model training error and generalization ability, and to avoid overfitting; The kernel coefficient gamma = 0.05 controls the width of the RBF kernel and adjusts the distribution density of samples in the feature space; The loss function parameter epsilon=0.1 defines the ε-insensitive region to reduce the impact of noise on model predictions; Other parameters: Double-precision floating-point arithmetic is used, the maximum number of iterations is set to 1000, and the probability output mode is enabled.

[0128] (3) Application process: All data were processed using Python code: Support Vector Machine (SVM) and Random Forest (RF) models were built using the scikit-learn library, and Convolutional Neural Network (CNN) models were built using the TensorFlow framework and Keras API. Spectral data preprocessing (including smoothing and denoising, baseline correction, and standardization), principal component analysis (PCA) for feature extraction and dimensionality reduction, and multivariate statistical evaluation (including receiver operating characteristic (ROC) curve plotting, area under the curve (AUC) calculation, and precision-recall curve analysis) were all implemented using the scikit-learn library; statistical tests such as t-tests were performed using the scipy library; all data visualizations (including PCA loading plots, score plots, confusion matrices, model learning curves, and ROC curves) were implemented using Python.

[0129] The model training employed a data splitting strategy: 70% of the data was used as the training set, and 30% as an independent test set; 5-fold cross-validation was used to evaluate the model's generalization stability. A combination of grid search and manual tuning was employed to optimize key hyperparameters: for the SVM model, the penalty parameter C and kernel coefficient gamma of the radial basis function (RBF) kernel were optimized; for the RF model, the number of decision trees was optimized; for the CNN model, the convolutional kernel size, number of training epochs, and learning rate were optimized to avoid overfitting. All spectral data in this invention were derived from multiple independent experiments to ensure the reliability and reproducibility of the results.

[0130] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features claimed herein.

Claims

1. A surface-enhanced Raman scattering flexible substrate material, characterized in that, include: A three-dimensional network structure composed of hydroxypropyl methylcellulose nanofibers; And silver nanoparticles encapsulated within the hydroxypropyl methylcellulose nanofibers; The silver nanoparticles have a polyhedral structure.

2. The surface-enhanced Raman scattering flexible substrate material according to claim 1, characterized in that: The silver nanoparticles have a particle size of 100~250 nm; and / or, The mass ratio of the silver nanoparticles to hydroxypropyl methylcellulose is (0.10~0.50):

1.

3. The surface-enhanced Raman scattering flexible substrate material according to claim 1, characterized in that: The surface-enhanced Raman scattering flexible substrate material has one or more of the following characteristics: Thickness ranges from 10 to 100 μm; The pore size is 1~5μm; Porosity is 60%~90%; Specific surface area is 20~100 m² 2 / g; The diameter of hydroxypropyl methylcellulose nanofibers is 50~500 nm; The roughness Ra value of the surface-enhanced Raman scattering flexible substrate material is >500 nm.

4. The surface-enhanced Raman scattering flexible substrate material according to claim 1, characterized in that: The surface of the silver nanoparticles is further modified with probe molecules; wherein: The probe molecule contains a first functional group for anchoring to the surface of silver nanoparticles and a second functional group for capturing target molecules.

5. A method for preparing a surface-enhanced Raman scattering flexible substrate material, characterized in that, Includes the following steps: Hydroxypropyl methylcellulose was dissolved in a water-ethanol mixture, and silver nanoparticles were added to form a spinning solution. The spinning solution is spun into a nanofiber membrane using an electrospinning process. The nanofiber membrane is dried to obtain the surface-enhanced Raman scattering flexible substrate material.

6. The method for preparing the surface-enhanced Raman scattering flexible substrate material according to claim 5, characterized in that: The water-ethanol mixed solvent is composed of water and ethanol in a volume ratio of 1:(1~3); and / or, The mass ratio of the hydroxypropyl methylcellulose to the water-ethanol mixed solvent is (1~3):100; and / or, The mass ratio of the silver nanoparticles to hydroxypropyl methylcellulose is (0.10~0.50):1; and / or, The drying temperature is 50~70℃, and the time is 2~5 hours.

7. The method for preparing the surface-enhanced Raman scattering flexible substrate material according to claim 5, characterized in that: The preparation method further includes a probe molecule modification step, which includes: The dried nanofiber membrane is placed in a solution of probe molecules and immersed; after sufficient reaction, the probe molecules that have not bound to the silver nanoparticles are washed away; wherein the probe molecules contain a first functional group for anchoring to the surface of the silver nanoparticles and a second functional group for capturing target molecules.

8. A face mask, characterized in that: The mask comprises the surface-enhanced Raman scattering flexible substrate material as described in claim 4.

9. An exhaled gas detection device, characterized in that, include: The face mask includes the surface-enhanced Raman scattering flexible substrate material as described in claim 4, for capturing and pre-enriching target molecules in exhaled gas; A Raman spectroscopy detection device is used to scan the surface-enhanced Raman scattering flexible substrate material in a used mask to obtain the Raman spectral signal of the target molecules. The data processing unit is configured to use machine learning algorithms to analyze the Raman spectral signal, identify the characteristic peaks of the target molecule, and output the detection results.

10. The exhaled gas detection device according to claim 9, characterized in that, The data processing unit includes: The data preprocessing module is used to smooth, denoise, and correct the baseline of the Raman spectral signal; The feature extraction module is used to extract the position and intensity information of the characteristic peaks of the target molecules from the corrected spectral signal; The intelligent recognition module is used to identify the types and concentrations of target molecules in exhaled gas based on the characteristic peak information.

Images