A chiral molecular sensing platform based on chiral nanohorn array and intelligent modeling and a detection method

Abstract

This invention discloses a chiral molecule sensing platform and detection method based on a chiral nanocone array and intelligent modeling. The chiral nanocone array sensing substrate includes a substrate and an Ag-SiO2-Au three-layer composite structure disposed on the substrate: an Ag layer, a SiO2 layer, and an Au layer; the Au layer is located outside the SiO2 layer, and the absolute value of the difference in the relative azimuth angle between the Au layer and the Ag layer is 90°. The chiral molecule sensing platform includes a chiral nanocone array sensing substrate, a spectral acquisition module, a signal processing module, and an intelligent modeling module. This invention solves the technical bottlenecks of traditional chiral sensing, such as weak response, poor analyte accessibility, low signal-to-noise ratio, and limited quantitative accuracy, achieving label-free identification of the chiral molecule thalidomide with 99% accuracy and 10... 2 -10 9 Precise quantification with a wide dynamic range (pg / mL) 2 With a value of ≥0.99, it is suitable for drug development, quality control and other scenarios, and has a wide range of application value.

Description

Technical Field

[0001] This invention relates to the field of chiral sensing technology, specifically to a chiral molecular sensing platform and detection method based on chiral nanocone arrays and intelligent modeling. Background Technology

[0002] The identification and quantitative analysis of chiral molecules are of paramount importance in the fields of chemistry, biology, and drug development. Statistics show that over 56% of approved small molecule drugs are chiral, and more than 88% are marketed as single enantiomers. Different enantiomers can exhibit drastically different pharmacological effects; for example, the R-enantiomer of thalidomide (Thd) has a sedative effect, while the S-enantiomer is teratogenic. Therefore, accurate identification and quantification of chiral molecules are crucial for drug safety evaluation.

[0003] Traditional chirality detection methods include circular dichroism (CD), vibrational circular dichroism (VCD), Raman optical activity (ROA), and nuclear magnetic resonance (NMR), but these methods have inherent drawbacks: the characteristic chiral signal is usually 10 times weaker than the non-chiral signal. 3 -10 6 The limitations of the structured chiral field and the analyte are significant; the chiral near-field is easily buried by the substrate, making it difficult for the analyte to effectively contact the enhancement region; the signal-to-noise ratio is low and easily interfered with, resulting in poor detection repeatability; the interaction between the structured chiral field and the analyte is nonlinear and system-dependent, making quantitative concentration regression difficult.

[0004] Chiral metamaterials, as artificially designed subwavelength structures, can achieve customized optical chirality, providing an alternative to traditional methods. However, existing chiral metamaterials still face challenges such as difficulty in achieving strong response and high integration, insufficient effective interaction between analytes and the chiral near field, and lagging signal processing and modeling methods, which restrict their practical applications. Therefore, developing chiral sensing technologies that combine high response intensity, high analyte accessibility, high signal-to-noise ratio, and high quantitative accuracy is of significant practical importance. Summary of the Invention

[0005] To address the aforementioned limitations of existing technologies, the present invention aims to provide a chiral molecular sensing platform and detection method based on chiral nanocone arrays and intelligent modeling. The invention first prepares a chiral nanocone array sensing substrate possessing multiple characteristics including sharp tips, subwavelength gaps, nanopores, and resonant cavities. By constructing the chiral nanocone array sensing substrate, a spectral acquisition module, a signal processing module, and an intelligent modeling module, the technical bottlenecks of traditional chiral sensing, such as weak response, poor analyte accessibility, low signal-to-noise ratio, and limited quantitative accuracy, are overcome.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides a chiral nanocone array sensing substrate, the chiral nanocone array sensing substrate comprising a substrate and a three-dimensional asymmetric structure disposed on the substrate; the three-dimensional asymmetric structure is an Ag-SiO2-Au three-layer composite structure; the Ag-SiO2-Au three-layer composite structure comprises an Ag layer, a SiO2 layer and an Au layer; the Au layer is located outside the SiO2 layer, and the absolute value of the difference in the relative azimuth angle between the Au layer and the Ag layer is 90°.

[0007] Preferably, the substrate is a glass substrate.

[0008] Preferably, the chiral nanocone array sensing substrate is prepared by the following method: (1) Monolayer self-assembly of PS microspheres: Monolayer self-assembly of PS microspheres was performed on a glass substrate coated with photoresist using a liquid-gas interface method; (2) Reactive ion etching: The photoresist film and PS microspheres are subjected to reactive ion etching to completely remove the PS microspheres and form a non-close-packed nanocone array; (3) Tilted deposition: The Ag layer, SiO2 layer and Au layer are deposited sequentially on the photoresist at an angle, so that the Au layer is located outside the SiO2 layer and the absolute value of the difference between the relative azimuth angles of the Ag layer and the Au layer is 90°. (4) Structure formation: Ethanol immersion to remove photoresist, and rotating the substrate by 90° to prepare left-handed and right-handed chiral structures, which are chiral nanocone array sensing substrates.

[0009] Preferably, the diameter of the PS microspheres is 500 nm; the tilting deposition angle is 40°; the thickness of the Ag layer is 20 nm, the thickness of the SiO2 layer is 40 nm, and the thickness of the Au layer is 20 nm.

[0010] A second aspect of the present invention provides the application of a chiral nanocone array sensing substrate in improving the sensitivity of chiral detection.

[0011] A third aspect of the present invention provides a chiral molecular sensing platform based on chiral nanocone arrays and intelligent modeling, characterized in that it includes a chiral nanocone array sensing substrate, a spectral acquisition module, a signal processing module, and an intelligent modeling module; the spectral acquisition module is a circular dichroism spectroscopy instrument connected to the chiral nanocone array sensing substrate; the signal processing module is connected to the spectral acquisition module, preprocesses the detected raw spectrum, constructs a differential spectrum through differential operations to retain full-band feature information, and transforms the one-dimensional ΔΔSpectra signal into three high-dimensional features—GASF, GADF, and CGAF—through Gram angle field technology to mine latent signal correlations; the intelligent module is connected to the signal processing module to construct a fully connected neural network dual-task model based on Bayesian optimization.

[0012] Preferably, the CD value of the chiral nanocone array sensing substrate is 1.86°. g Factor 0.39; The preprocessing involves reconstructing the detected raw spectrum in the 800-1000 nm band, followed by smoothing and normalization.

[0013] Preferably, the difference operation is ∆CD RH ( l i ) = CD RHwThd ( l i ) - CD RHw / oThd ( l i ) and ∆CD LH ( l i )= CD LHwThd ( l i ) -CD LHw / oThd ( l i The differential spectrum is ΔΔSpectra ( l i ) = ΔCD RH ( l i -ΔCD LH ( l i ).

[0014] A fourth aspect of the present invention provides the application of a chiral molecule sensing platform in improving the detection responsiveness of chiral molecules.

[0015] A fifth aspect of the present invention provides a method for detecting chiral molecules using a chiral molecule sensing platform, the method comprising the following steps: (1) Pretreatment of sensing substrate: The chiral nanocone array sensing substrate was ultrasonically cleaned in anhydrous ethanol and dried with nitrogen before use. (2) Sample preparation: The solution of the chiral molecule to be detected is dropped onto the surface of the sensing substrate and incubated. (3) Spectral acquisition: acquire circular dichroism chromatographic signals and background signals from the blank substrate; (4) Signal preprocessing: The original spectral signal is reconstructed in the 800-1000 nm band, and smoothing algorithm and normalization are applied in sequence to eliminate noise and amplitude scale effect interference. (5) Feature extraction: Calculate ΔCD RH ΔCD LH And ΔΔSpectra differential spectra, which are then converted into GASF, GADF and CGAF high-dimensional features through Gram angular field technology; (6) Intelligent detection: Input the high-dimensional features of CGAF into the intelligent modeling module, identify the enantiomer type through the classification sub-model, and predict the molecular concentration through the regression sub-model.

[0016] Preferably, the molecular weight of the chiral molecules in the solution to be detected is <500; the incubation time is 5-10 min; the smoothing algorithm is the Savitzky-Golay algorithm; the enantiomeric recognition accuracy is ≥99%, and the concentration prediction accuracy is ≥99%. R 2 ≥0.99, MSE<0.01.

[0017] Preferably, the chiral molecule contained in the solution of the chiral molecule to be detected is thalidomide.

[0018] The beneficial effects of this invention are: (1) Excellent sensing substrate performance: The Ag-SiO2-Au three-layer composite structure integrates multiple plasmon characteristics, achieving a broadband strong chiral response (CD value 1.86°). g With a factor of 0.39, the three-dimensional asymmetric design extends the superchiral field into the air cavity region, significantly improving analyte accessibility and solving the problems of weak response and difficulty in molecular contact in traditional substrates. (2) High efficiency in signal processing: By using differential operations and high-dimensional representation techniques, the hidden information in the spectrum is fully explored, noise interference is effectively suppressed, the signal-to-noise ratio is significantly improved, and the technical bottleneck of feature extraction under low signal-to-noise ratio is broken through. (3) High detection accuracy: Based on a Bayesian-optimized fully connected neural network dual-task model, it achieves 99% accuracy in label-free enantiomer recognition of thalidomide molecules and 10 2 -10 9 Precise quantification with a wide dynamic range (pg / mL) R 2(≥0.99), which solves the problem of difficult quantification under nonlinear spectral conditions; (4) Convenient preparation and application: The sensing substrate is prepared by colloidal etching technology, which does not require complex photolithography process, and is low in cost and can be mass-produced; the detection method is easy to operate and is suitable for the detection of a variety of chiral substances such as chiral drugs and biomolecules, and has wide application value in drug development, quality control, environmental monitoring and other fields. Attached Figure Description

[0019] Figure 1 Schematic diagram of the fabrication process of Ag-SiO2-Au chiral nanocone array; Figure 2 (a) Characterization results of Ag-SiO2-Au right-handed (RH-) and left-handed (LH-) nanocone arrays; (b) Reflectance and transmission spectra of Ag-SiO2-Au right-handed nanocone arrays; (c) CD spectra of Ag-SiO2-Au and (Ag-Au) right-handed nanocone arrays; (d) Influence of SiO2 layer thickness on optical properties; Figure 3 A complete spectral data processing workflow; Figure 4 Architecture and training strategy of chiral thalidomide identification and concentration prediction model based on fully connected neural network; Figure 5 The model training performance is demonstrated from multiple dimensions, including classification performance, regression accuracy, training process, and prediction stability. Among them, (a) uses ΔΔ l ΔΔSpectra l i Five data representation formats, including GADF, GASF, and CGAF, were used. The classification accuracy and regression average coefficient of determination were obtained after training with a fully connected neural network (FCN). R 2 (a) Comparison results; (b) Targeting R -Thd and S -Thd sample, after being trained by a fully connected neural network for classification, the confusion matrix between the true label and the predicted label; (c) for S -The loss function curves of fully connected neural network regression training when Thd samples are processed using different advanced signal processing methods; (dh) sequentially passes through ΔΔ l ΔΔSpectra ( l i After processing the data using five methods—GADF, GASF, and CGAF—a fully connected neural network regression model was used to obtain... R -Thd and S- Linear fitting plot of the true concentration and predicted concentration of the Thd sample; (i) After training a fully connected neural network regression model using the CGAF dataset, the model was analyzed on 50 samples. S -The concentration of the Thd sample was predicted, and the predicted concentration was compared with the actual concentration. Detailed Implementation

[0020] It should be noted that the following detailed descriptions are illustrative and intended to provide further explanation of this application. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0021] As described in the background section, the identification and quantitative analysis of molecular chirality has long been limited by the problem of weak and noisy chiral optical response signals.

[0022] This invention proposes a multi-scale chiral sensing platform that specifically addresses core bottlenecks such as insufficient response intensity, poor analyte accessibility, low signal-to-noise ratio, and limited quantitative accuracy. The platform integrates a wafer-scale Ag-SiO2-Au chiral nanocone array fabricated using colloidal etching technology, combined with spectral coding and machine learning analysis methods. This asymmetric three-dimensional nanostructure possesses multiple characteristics, including sharp tips, subwavelength gaps, nanopores, and resonant cavities, enabling the generation of a broadband superchiral field, which molecules can directly contact. By optimizing a 40 nm thick SiO2 spacer layer aligned with the Au layer orientation, this structure achieves four strong resonance modes, a CD value of 1.86°, and an asymmetry factor of 0.39—a 7.4-fold performance improvement compared to a bilayer structure. To address the issues of low signal-to-noise ratio and non-monotonic spectral response, the original circular dichroism signal is converted into multi-dimensional data formats such as spectral differentiation and Gram angle field, and then analyzed using a Bayesian optimization algorithm and neural network. Using thalidomide enantiomers as the detection target, this platform achieved label-free identification with 99% accuracy, and within 10... 2 -10 9 Accurate quantitative regression was performed within the concentration range of pg / mL. R 2 ≥0.99).

[0023] To enable those skilled in the art to better understand the technical solution of this application, the technical solution of this application will be described in detail below with reference to specific embodiments.

[0024] The test materials used in the embodiments of this invention are all conventional test materials in the art and can be purchased through commercial channels.

[0025] Example 1: Fabrication of a chiral nanocone array sensing substrate (1) The preparation process is as follows Figure 1 As shown, specifically: Substrate pretreatment: A clean glass substrate was selected and ultrasonically cleaned sequentially with deionized water and anhydrous ethanol for 15 minutes each. After drying with nitrogen, photoresist (BP212-37S, purchased from Beijing Kempur Microelectronics Co., Ltd.) was uniformly spin-coated onto the glass substrate surface at a speed of 3000 rpm for 30 seconds. The coated substrate was then placed in an oven and cured at 88°C for 2 hours, ultimately forming a 2 μm thick photoresist layer on the substrate.

[0026] PS microsphere self-assembly: A close-packed PS microsphere array (diameter = 500 nm, purchased from TSSphere Co., Ltd., China) was assembled on a photoresist-coated substrate using a liquid-gas interface method. A 5 wt% PS microsphere dispersion (1 mL ethanol and 1 mL deionized water) was added dropwise to the surface of deionized water. After the PS microspheres self-assembled to form a monolayer film, it was transferred to a photoresist / substrate and allowed to dry naturally to obtain a PS microsphere / photoresist / glass substrate.

[0027] Reactive ion etching: PS microspheres / photoresist / glass substrate are placed in a reactive ion etching instrument. A CF4 / O2 mixed gas (flow rates of 10 sccm and 20 sccm, respectively) is used as the etching gas. The RF source frequency is set to 13.56 MHz, the RF power to 8 W, and the initial process pressure to 10 W. -2 Pa, etching time 300 seconds, finally forming a non-closely packed photoresist nanocone array, with a unit cone diameter of 230 nm and a height of 260 nm.

[0028] Tilted deposition: A photoresist nanocone array based on a glass substrate is placed in a vacuum evaporator, and Ag layers (20 nm) are deposited sequentially at an incident angle of 40°. The central axis of the Ag layer is defined as the azimuth angle. f = 0°), SiO2 layer (40 nm, f = 90°) and Au layer (20 nm, f = 90°), and the deposition rate was controlled at 0.1 nm / s; thus, a composite layer / photoresist nanocone / substrate was obtained.

[0029] Structure formation: The composite layer / photoresist nanocone / substrate is immersed in anhydrous ethanol for 10 minutes, the photoresist is removed, and the substrate is dried with nitrogen. The substrate is rotated 90° to obtain the left-handed and right-handed chiral structures, which are the Ag-SiO2-Au chiral nanocone array sensing substrate.

[0030] Example 2: Preparation of a chiral molecule sensing platform (1) Sensing substrate: The chiral nanocone array sensing substrate prepared in Example 1 integrates multiple characteristics such as sharp tip, subwavelength gap, nanopore and resonant cavity. The SiO2 layer thickness is optimized to 40 nm and aligned with the Au layer azimuth angle to achieve four strong resonance modes, with a CD value of 1.86° and an asymmetry factor of 0.39.

[0031] (2) Spectral acquisition module: A circular dichroism chromatograph (MOS-450 / AF-CD, Biologic, France) was used to detect wavelengths covering the visible-near infrared region (400-1100 nm). The temperature was controlled at 25±0.5℃. The CD spectral signal after the interaction between the chiral molecules and the sensing substrate was acquired, and the background signal of the blank substrate was recorded for subsequent correction.

[0032] (3) Signal processing module: To address the noise interference problem of the original signal, a multi-level processing flow is designed: ① Preprocessing: The original spectrum is reconstructed in the 800-1000 nm band, and noise and amplitude interference are reduced through smoothing and normalization. ② Difference operation: Calculate ∆CD RH ( l i ) = CD RHwThd ( l i ) - CD RHw / oThd ( l i ) and ∆CD LH ( l i ) =CD LHwThd ( l i ) -CD LHw / oThd ( l i Construct ΔΔSpectra ( l i ) = ΔCD RH ( l i -ΔCD LH ( l i Differential spectroscopy preserves characteristic information across the entire spectral band; ③ High-dimensional representation: The one-dimensional ΔΔSpectra signal is transformed into three high-dimensional features, namely GASF, GADF and CGAF, through Gram corner field technology to explore latent signal correlations.

[0033] (4) Intelligent modeling module: Construct a dual-task model based on a fully connected neural network with Bayesian optimization: ① Classification sub-model: a two-layer network structure, with the input layer matching the feature dimension, the hidden layer containing 128 neurons, configured with ReLU activation function and Dropout regularization, and the output layer corresponding to 2 types of enantiomers to achieve enantiomer recognition; ② Regression sub-model: a three-layer network structure with 128 and 64 neurons in the middle hidden layers, configured with ReLU activation function and Dropout regularization, and a single neuron in the output layer to achieve concentration quantification; ③ Hyperparameter optimization: The learning rate, Dropout probability, and number of iterations are optimized using the Bayesian optimization algorithm to improve the model's generalization ability and prediction accuracy.

[0034] Example 3: Detection of thalidomide (Thd) enantiomers Pretreatment of sensing substrate: The chiral nanocone array sensing substrate prepared in Example 1 was ultrasonically cleaned in anhydrous ethanol for 10 minutes and dried with nitrogen gas for later use.

[0035] Sample preparation: Prepare a solution with a concentration of 10... 2 pg / mL, 10 5 pg / mL, 109 pg / mL R -Thd and S -Thd standard solution, and at the same time prepare blank control solution (deionized water).

[0036] Sample application: Take 5 μL of the above standard solution and add it to the surface of the sensing substrate. Incubate for 8 minutes to allow the molecules to be fully adsorbed and interact with the superchiral field.

[0037] Spectral acquisition: A circular dichroism chromatograph was used to acquire CD spectral signals in the 400-1100 nm band at 25±0.5℃. Each sample was acquired three times and the average value was taken.

[0038] Signal processing: The original spectrum is reconstructed in the 800-1000 nm band, smoothed using the Savitzky-Golay algorithm, and then normalized and corrected. The ΔΔSpectra differential spectrum is calculated and converted into CGAF high-dimensional features using Gram angular field technology.

[0039] Intelligent detection: Input CGAF features into a Bayesian-optimized fully connected neural network model, the classification sub-model identifies the enantiomer type, and the regression sub-model predicts the concentration.

[0040] like Figure 3As shown, a complete spectral data processing workflow is presented: First, the CD spectral dataset, which is affected by noise, is limited to the wavelength range of 800-1000 nm for recalibration. Then, the calibrated spectra are smoothed and normalized sequentially. During processing, the spectral differences of the right-handed assembly sample with and without thalidomide, and the corresponding spectral differences of the left-handed assembly sample are calculated to obtain the ΔΔ spectra. Finally, feature extraction is performed on the obtained spectra using dimensionality expansion methods (including the gradient magnitude direction histogram GADF and the gradient magnitude similarity histogram GASF).

[0041] Figure 4 This paper presents the architecture and training strategy of a chiral thalidomide identification and concentration prediction model based on a fully connected neural network. ΔΔλ and ΔΔSpectra are selected. l i Five data representation formats are used, including GADF, GASF, and CGAF, based on... R -Thd and S - Fully connected neural networks (FCNs) were trained on two datasets: one for the test samples and the other for the blank Ag-SiO2-Au chiral nanocone array circular dichroism chromatogram. Five representations were used as model inputs, respectively connected to classification / regression models. The training and test sets were split in a 9:1 ratio. The chiral classification model employed a two-layer FCN architecture, with the input layer dimension matching the feature dimension. The hidden layer was a 128-neuron fully connected layer, followed by a ReLU activation function and a Dropout regularization module (to prevent overfitting). The number of neurons in the output layer corresponded to the classification category. R -Thd is marked as 0, S -Thd is marked as 1); the concentration regression model adopts a three-layer FCN architecture, with the input layer matching the feature dimension; two hidden layers (128 and 64 neurons) are set, and each hidden layer is configured with a ReLU activation function and a Dropout regularization module; the output layer is a single neuron, which is paired with a Sigmoid activation function to realize the quantitative prediction of 12 concentration gradients; the model optimization introduces a Bayesian optimization algorithm to automatically optimize hyperparameters, thereby improving the model's generalization ability and the accuracy and stability of concentration prediction.

[0042] Figure 5 The model training performance is demonstrated from multiple dimensions, including classification performance, regression accuracy, training process, and prediction stability. (a) The model training effect is shown using ΔΔ l ΔΔSpectra l i Five data representation formats, including GADF, GASF, and CGAF, were used. The classification accuracy and regression average coefficient of determination were obtained after training with a fully connected neural network (FCN). R2 (a) Comparison results; (b) Targeting R -Thd and S -Thd sample, after being trained by a fully connected neural network for classification, the confusion matrix between the true label and the predicted label; (c) for S -The loss function curves of fully connected neural network regression training when Thd samples are processed using different advanced signal processing methods; (dh) sequentially passes through ΔΔ l ΔΔSpectra ( l i After processing the data using five methods—GADF, GASF, and CGAF—a fully connected neural network regression model was used to obtain... R -Thd and S - Linear fitting plot of the true concentration and predicted concentration of the Thd sample; (i) After training a fully connected neural network regression model using the CGAF dataset, the model was analyzed on 50 samples. S -The concentration of the Thd sample was predicted, and the predicted concentration was compared with the actual concentration.

[0043] Results verification: The accuracy rate of enantiomer recognition was 99.2%. R -Thd and S -Thd concentration prediction R 2 The values ​​were 0.993 and 0.991 respectively, and the MSE values ​​were 0.0006 and 0.0008 respectively. The test results were highly consistent with the standard values.

[0044] Comparative Example 1: Preparation of Ag-Au chiral nanocone arrays The difference from Example 1 is that no SiO2 layer is deposited. The resulting chiral nanocone array is denoted as Ag-Au chiral nanocone array.

[0045] Comparative Example 2: Fabrication of Ag-SiO2-Au chiral nanocone array sensing substrates with different SiO2 layer thicknesses The difference from Example 1 is that a SiO2 spacer layer was added to the structure, with deposition thicknesses of 30 nm, 50 nm, 60 nm, and 70 nm, respectively.

[0046] Example 4: The Ag-SiO2-Au right-handed (RH-) and left-handed (LH-) nanocone arrays prepared in Example 1 were characterized. Scanning electron microscopy (SEM) EDS mapping confirmed the structural morphology and compositional uniformity; transmission spectroscopy revealed the material's properties in... l Strong resonance peaks are observed at 500 nm (Au intrinsic resonance) and 900 nm (EOT anomalous optical transmission). lThe 680 nm shoulder peak is due to the coupling between the tip and the nanopore, and the transmission spectrum shows a significant difference under circularly polarized light illumination. CD spectroscopy indicates that the chiral nanocone array... l A quadruple strong resonance peak is formed at 525, 648, 825, and 1067 nm. The right-handed structure has a maximum CD value of 1.86° and a g-factor as high as 0.39, while the corresponding values ​​for the left-handed structure are 1.46° and 0.34, respectively. This strongly chiral responsive material can be fabricated through an efficient, low-cost, and scalable route, offering a significant cost advantage compared to traditional electron beam lithography or focused ion beam etching. The experiment also clarified the role of the SiO2 spacer layer. The results show that the chiral response of the Ag-Au structure without the SiO2 layer (Comparative Example 1) is significantly weakened; the material exhibits the strongest chiral optical response when the SiO2 layer thickness is 40 nm; and the performance is optimal when the SiO2 layer and Au layer are aligned. Experiments have shown that the presence, thickness, and orientation of the dielectric spacer layer are key factors affecting the chiral properties of the material. Compared with the thickness of the SiO2 spacer layer in Comparative Example 2, the Ag-SiO2-Au chiral nanoarray with a SiO2 layer thickness of 40 nm and an orientation consistent with the Au layer in Example 1 is the optimal structure.

[0047] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A chiral nanocone array sensing substrate, characterized in that, The chiral nanocone array sensing substrate includes a substrate and a three-dimensional asymmetric structure disposed on the substrate; the three-dimensional asymmetric structure is an Ag-SiO2-Au three-layer composite structure; the Ag-SiO2-Au three-layer composite structure includes an Ag layer, a SiO2 layer and an Au layer; the Au layer is located outside the SiO2 layer, and the absolute value of the difference in the relative azimuth angle between the Au layer and the Ag layer is 90°.

2. The chiral nanocone array sensing substrate according to claim 1, characterized in that, The chiral nanocone array sensing substrate is prepared by the following method: (1) Monolayer self-assembly of PS microspheres: Monolayer self-assembly of PS microspheres was performed on a glass substrate coated with photoresist using a liquid-gas interface method; (2) Reactive ion etching: The photoresist film and PS microspheres are subjected to reactive ion etching to completely remove the PS microspheres and form a non-close-packed nanocone array; (3) Tilted deposition: The Ag layer, SiO2 layer and Au layer are deposited sequentially on the photoresist at an angle, so that the Au layer is located outside the SiO2 layer and the absolute value of the difference between the relative azimuth angles of the Ag layer and the Au layer is 90°. (4) Structure formation: Ethanol immersion to remove photoresist, and rotating the substrate by 90° to prepare left-handed and right-handed chiral structures, which are chiral nanocone array sensing substrates.

3. The chiral nanocone array sensing substrate according to claim 1, characterized in that, The PS microspheres have a diameter of 500 nm; the tilted deposition angle is 40°; the Ag layer has a thickness of 20 nm, the SiO2 layer has a thickness of 40 nm, and the Au layer has a thickness of 20 nm.

4. The application of the chiral nanocone array sensing substrate according to any one of claims 1 to 3 in improving the sensitivity of chiral detection.

5. A chiral molecular sensing platform based on chiral nanocone arrays and intelligent modeling, characterized in that, The system includes a chiral nanocone array sensing substrate, a spectral acquisition module, a signal processing module, and an intelligent modeling module as described in any one of claims 1 to 3; the spectral acquisition module is a circular dichroism spectroscopy instrument connected to the chiral nanocone array sensing substrate; the signal processing module is connected to the spectral acquisition module, preprocesses the detected raw spectrum, constructs a differential spectrum through differential operations to retain full-band feature information, and transforms the one-dimensional ΔΔSpectra signal into three high-dimensional features—GASF, GADF, and CGAF—through Gram angle field technology to mine latent signal correlations; the intelligent module is connected to the signal processing module to construct a fully connected neural network dual-task model based on Bayesian optimization.

6. The chiral molecule sensing platform according to claim 5, characterized in that, The CD value of the chiral nanocone array sensing substrate is 1.86°. g Factor 0.39; The preprocessing involves reconstructing the detected raw spectrum in the 800-1000 nm band, followed by smoothing and normalization.

7. The chiral molecule sensing platform according to claim 5, characterized in that, The difference operation is ∆CD RH ( λ i )= CD RHwThd ( λ i ) - CD RHw / oThd ( λ i ) and ∆CD LH ( λ i ) = CD LHwThd ( λ i ) -CD LHw / oThd ( λ i The differential spectrum is ΔΔSpectra ( λ i ) = ΔCD RH ( λ i ) -ΔCD LH ( λ i ).

8. The application of the chiral molecule sensing platform according to any one of claims 5 to 7 in improving the response of chiral molecule detection.

9. The method for detecting chiral molecules using the chiral molecule sensing platform according to any one of claims 5 to 7, characterized in that, The method includes the following steps: (1) Pretreatment of sensing substrate: The chiral nanocone array sensing substrate according to any one of claims 1 to 3 is ultrasonically cleaned in anhydrous ethanol and dried with nitrogen gas for later use. (2) Sample preparation: The solution of the chiral molecule to be detected is dropped onto the surface of the sensing substrate and incubated. (3) Spectral acquisition: acquire circular dichroism chromatographic signals and background signals from the blank substrate; (4) Signal preprocessing: The original spectral signal is reconstructed in the 800-1000 nm band, and smoothing algorithm and normalization are applied in sequence to eliminate noise and amplitude scale effect interference. (5) Feature extraction: Calculate ΔCD RH ΔCD LH And ΔΔSpectra differential spectra, which are then converted into GASF, GADF and CGAF high-dimensional features through Gram angular field technology; (6) Intelligent detection: Input the high-dimensional features of CGAF into the intelligent modeling module, identify the enantiomer type through the classification sub-model, and predict the molecular concentration through the regression sub-model.

10. The method according to claim 9, characterized in that, The chiral molecules in the solution to be detected have a molecular weight < 500; the incubation time is 5-10 min; the smoothing algorithm is the Savitzky-Golay algorithm; the enantiomeric recognition accuracy is ≥ 99%, and the concentration prediction accuracy is [missing information]. R 2 ≥0.99, MSE<0.01.

Images