A high-sensitivity detection method for carbendazim based on terahertz metamaterial resonance enhancement

Abstract

This invention discloses a highly sensitive detection method for carbendazim based on terahertz metamaterial resonance enhancement. The invention proposes a terahertz metamaterial sensor with a "+" shaped composite double-peak structure. Transmission spectra were acquired using terahertz spectroscopy, and the differences in terahertz spectra at different concentrations were investigated. Analysis revealed that the amplitude of the transmission peak in the terahertz transmission spectrum of this structure decreases with increasing carbendazim solution concentration. Data preprocessing was performed using multivariate scattering correction (MSC), feature extraction was conducted using the iterative information-preserving variable (IRIV) method, and a quantitative detection model for carbendazim terahertz transmission spectra was established based on LS-SVM. This provides a new approach for the quantitative detection of carbendazim. Experimental results show that the model has a high R-value. P The assay yielded a value of 0.9825, an RMSEP of 0.2001, and a detection limit (LOD) of 0.672 μg / ml. This invention provides a novel method for the rapid, non-destructive, and highly sensitive detection of carbendazim in food.

Description

Technical Field

[0001] This invention relates to the field of chemical detection technology, and in particular to a highly sensitive detection method for carbendazim based on terahertz metamaterial resonance enhancement. Background Technology

[0002] Carbendazim (CBZ) is a widely used fungicide belonging to the benzimidazole class of compounds. It is primarily used to control fungal diseases in various crops, such as apple ring rot, anthracnose, and fruit rust. As a systemic fungicide, carbendazim achieves its control effect by inhibiting fungal cell wall synthesis and mitosis. However, with the increasing use of pesticides in agricultural production, carbendazim residues have gradually attracted attention in the field of food safety. Studies have shown that carbendazim residues on crops may pose potential health hazards to consumers; long-term consumption of food containing carbendazim residues may lead to endocrine disorders or other chronic diseases. To ensure that carbendazim residue levels in food are within legal and safe limits, many countries and regions have established corresponding regulatory standards. For example, the Chinese national standard GB 2763-2021, "Maximum Residue Limits for Pesticides in Food," stipulates that the maximum permissible residue concentration of carbendazim in apples is 5 μg / ml. Therefore, researching an accurate, rapid, and non-destructive method for detecting carbendazim residues in agricultural products has significant practical importance and application value.

[0003] Currently, commonly used methods for detecting pesticide residues both domestically and internationally include high-performance liquid chromatography (HPLC), enzyme-linked immunosorbent assay (ELISA), electrochemical methods, near-infrared spectroscopy (NIRS), and Raman spectroscopy. While existing methods offer advantages in sensitivity, they still suffer from several drawbacks, such as high detection costs, long processing times, poor result stability, cumbersome and complex experimental procedures, and high equipment costs that may damage samples. Therefore, there is an urgent need to develop a detection technology that is both highly sensitive and rapid, and non-destructive.

[0004] Terahertz (THz) waves typically refer to electromagnetic waves with frequencies ranging from 0.1 to 10 THz. Due to their high stability, high resolution, and rich optical properties, they have become ideal tools for biological and chemical molecular detection. In recent years, an increasing number of researchers have begun to explore the application of terahertz spectroscopy in areas such as food safety. However, traditional terahertz spectroscopy techniques often face problems such as insufficient sensitivity and inability to meet detection standards when detecting trace toxic substances. This may be because the low concentration of the substance leads to a weaker response of the terahertz wave, thus limiting the resolution and detection accuracy of the terahertz spectrum. The emergence of metamaterial sensors has provided new ideas for improving the detection limit of terahertz detection technology.

[0005] Metamaterial sensors possess unique electromagnetic response characteristics, and through the principle of resonance, they can enhance the sensitivity of terahertz spectroscopy to trace components. By analyzing the frequency shifts and amplitude changes caused by the sample, the accuracy and sensitivity of detection are significantly improved. Summary of the Invention

[0006] In view of the above, the purpose of this invention is to provide a highly sensitive detection method for carbendazim based on terahertz metamaterial resonance enhancement, so as to develop a new metamaterial structure and use the metamaterial to realize the quantitative detection of carbendazim, thus providing a new approach for the quantitative detection of carbendazim.

[0007] This invention provides a highly sensitive detection method for carbendazim based on terahertz metamaterial resonance enhancement. The terahertz metamaterial sensor has the following parameters: a period P1 of 70 μm, a gold coating, and a silicon substrate. The terahertz metamaterial absorber consists of multiple periodic structural units, each of which is a square structure with a width of 70 μm. Each periodic structural unit contains four cross-shaped structures, and each periodic structural unit is surrounded by a strip structure. The width L1 of each cross-shaped structure is 30 μm, and each cross-shaped structure is composed of two mutually perpendicular "I"-shaped structures. The width L3 at the end of each "I"-shaped structure is 4 μm, and the width W2 in the middle of each "I"-shaped structure is 2 μm. The length L2 of each strip structure is 60 μm, and the width W1 of each strip structure is 2 μm. The silicon substrate has a refractive index of 3.335 and a thickness of 500 μm.

[0008] The detection method includes the following steps:

[0009] (1) Prepare multiple carbendazim standard solutions of different concentrations, and drop the carbendazim standard solutions onto the terahertz metamaterial absorber and dry them in a drying oven;

[0010] (2) Place the dried terahertz metamaterial absorber into the terahertz system and use the transmission mode to measure and collect the original terahertz spectral information of the corresponding sample.

[0011] (3) Use Mahalanobis distance to remove outliers from the original terahertz spectral information to obtain the remaining spectral data;

[0012] (4) Use multivariate scattering correction to preprocess the remaining spectral data, and then use the iterative information-preserving variable method to extract features from the preprocessed data;

[0013] (5) A quantitative detection model for carbendazim terahertz transmission spectrum was established based on the extracted features and LS-SVM;

[0014] (6) Obtain the terahertz spectral information of the sample containing carbendazim to be tested and input it into the quantitative detection model to obtain the content of carbendazim in the sample to be tested.

[0015] Compared with existing technologies, this invention provides a highly sensitive detection method for carbendazim based on terahertz metamaterial resonance enhancement. It proposes a terahertz metamaterial sensor with a "+"-shaped composite double-peak structure, and uses terahertz spectroscopy to acquire transmission spectra, exploring the differences in terahertz spectra among different concentrations. Analysis reveals that the transmission peak amplitude of this structure decreases with increasing carbendazim solution concentration. Data preprocessing is performed using multivariate scattering correction (MSC), feature extraction is conducted using iterative information-preserving variable (IRIV) method, and a quantitative detection model for carbendazim terahertz transmission spectra is established based on LS-SVM. This provides a new approach for the quantitative detection of carbendazim. Experimental results show that the model has a high R-value. P The assay yielded a value of 0.9825, an RMSEP of 0.2001, and a detection limit (LOD) of 0.672 μg / ml. This invention provides a novel method for the rapid, non-destructive, and highly sensitive detection of carbendazim in food.

[0016] In addition, the highly sensitive detection method for carbendazim based on terahertz metamaterial resonance enhancement provided by the present invention also has the following technical features:

[0017] The terahertz metamaterial absorber is prepared by the following method:

[0018] First, a silicon substrate is selected. The substrate surface is cleaned to remove impurities and then oxidized to enhance the adhesion of the photoresist. Next, a chromium layer is deposited on the silicon substrate using thin-film deposition technology, followed by a metal layer to enhance the adhesion between the metal layer and the silicon substrate. Then, a layer of photoresist with a thickness of 1-2 μm is coated onto the silicon wafer using a spin-coating process. After coating, pre-baking is performed at 100°C for 120 seconds to remove the solvent from the photoresist, ensuring it hardens and is ready for exposure. After exposure, the development process begins, with a development time controlled at 30 seconds, removing the unexposed areas of photoresist and leaving the pattern. After development, solvent stripping is used to remove residual photoresist. Then, reactive ion etching or ion beam etching is performed. After etching, the photoresist mask is removed by ultrasonic cleaning with acetone to ensure no residue on the chip surface. Finally, surface oxidation or metal deposition post-treatment is performed to enhance conductivity and corrosion resistance.

[0019] Specifically, step (1) includes:

[0020] The concentration of the carbendazim standard solution was selected as 100 μg / ml. The carbendazim standard solution was diluted with ultrapure deionized water using a pipette to prepare 21 carbendazim solutions with different concentration gradients. The prepared sample solutions were placed on a vortex mixer and shaken for 3 min to ensure that the carbendazim standard solution was fully diluted in deionized water, and then stored in volumetric flasks.

[0021] Specifically, step (2) includes:

[0022] Before the experiment, the system was preheated using an air compressor equipped with an air dryer, and dry air was continuously filled into the optical cavity. The humidity was kept below 10% and the temperature was kept constant at 25±0.5℃ by monitoring with a thermometer and hygrometer. Using a pipette, 20 μl of carbendazim sample solution was dropped onto the terahertz metamaterial absorber in order of increasing sample concentration. Before acquiring the sample spectrum, the terahertz metamaterial absorber with the sample was placed in a drying oven at 50℃ for 30 min. Finally, the terahertz metamaterial absorber with the sample was placed into the terahertz system and measured using transmission mode. After each sample was placed, a 2-minute wait was made for the system to stabilize before starting the measurement. Five points were taken for each sample, and 10 measurements were performed at each point. A total of 1050 spectra were obtained from 21 concentration samples.

[0023] Specifically, step (3) includes:

[0024] The variables are normalized by the covariance matrix, the correlation between the features is identified, the Mahalanobis distance threshold is set to 0.99, and then outliers in the original terahertz spectral information are removed to obtain the remaining spectral data.

[0025] In step (5), F-test and t-test are used to verify the predicted and actual values ​​of the established quantitative detection model of carbendazim terahertz transmission spectrum. Attached Figure Description

[0026] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which:

[0027] Figure 1 This is a schematic diagram of a terahertz metamaterial absorber.

[0028] Figure 2 This is a simulation diagram of the terahertz transmission spectrum;

[0029] Figure 3 To simulate the electromagnetic field distribution and terahertz transmission spectrum of the metamaterial structure using FDTD: (a) spatial distribution of electric field intensity at 1.22 THz; (b) spatial distribution of electric field intensity at 2.30 THz;

[0030] Figure 4 Transmission spectra of carbendazim solutions at different concentrations on the surface of metamaterial sensors;

[0031] Figure 5 Mahalanobis distance distribution for different spectral data;

[0032] Figure 6 1 st The results of variable selection using different feature extraction methods on the data after D processing;

[0033] Figure 7 The variable selection results for MSC-processed data after different feature extraction methods;

[0034] Figure 8 This is a fitting graph showing the THz spectrum of carbendazim solution predicted by the LS-SVM model and the actual values. Detailed Implementation

[0035] To make the objectives, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Several embodiments of the present invention are shown in the drawings. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided so that the disclosure of the present invention will be thorough and complete.

[0036] Unless otherwise defined, 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 invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0037] Please see Figures 1 to 3 This invention provides a terahertz metamaterial absorber. In this embodiment, the structural parameters of the metamaterial absorber are designed using FDTD Solutions software parameter simulation. Since the metamaterial is composed of periodic structural units, its overall transmission or reflection characteristics can be represented by the performance of a single periodic unit. Therefore, in simulation analysis, simulating a single periodic unit of the metamaterial effectively improves simulation efficiency.

[0038] Specifically, the terahertz metamaterial sensor is a cross-shaped composite bimodal terahertz metamaterial sensor with the following parameters: period P1 is 70 μm, the coating metal is gold, and a silicon substrate is used; the terahertz metamaterial absorber is composed of multiple periodic structural units, each periodic structural unit is a square structure with a width of 70 μm, and each periodic structural unit has four cross-shaped structures 10 inside, and a strip structure 20 is provided around each of the periodic structural units. The width L1 of each cross-shaped structure 10 is 30 μm, and each cross-shaped structure 10 is composed of two mutually perpendicular I-shaped structures 11. The width L3 of the end of each I-shaped structure 11 is 4 μm, and the width W2 of the middle of each I-shaped structure 11 is 2 μm. The length L2 of each strip structure 20 is 60 μm, and the width W1 of each strip structure 20 is 2 μm; the refractive index of the silicon substrate is 3.335, and the thickness of the silicon substrate is 500 μm.

[0039] The fabrication process of metamaterial sensors is a precise operation. First, a high-quality silicon substrate is selected, ensuring its smoothness and flatness. The substrate surface is cleaned to remove impurities and then oxidized to enhance the adhesion of the photoresist. Subsequently, a chromium layer of approximately 5 nm is deposited on the silicon substrate using thin-film deposition technology, followed by a metal layer of approximately 100 nm to enhance the adhesion between the metal layer and the silicon substrate. Then, a layer of photoresist is coated onto the silicon wafer using a spin-coating process, with the spin speed controlled at approximately 3000 rpm to ensure uniform distribution of the photoresist, typically with a thickness of 1-2 μm. After coating, a pre-baking (soft baking) process is performed at 100°C for 120 seconds to remove the solvent from the photoresist, ensuring it hardens and is ready for exposure. After exposure, the development process begins, with a development time controlled at 30 seconds, removing the unexposed areas of photoresist and leaving the pattern. After development, a solvent (such as acetone) is used for stripping to remove residual photoresist. Finally, reactive ion etching (RIE) or ion beam etching (IBE) is used for etching. After etching, the photoresist mask needs to be removed by ultrasonic cleaning with acetone to ensure no residue remains on the chip surface. Finally, to improve the performance of the metamaterial chip, post-processing such as surface oxidation or metal deposition is usually performed. The surface oxidation temperature is typically controlled at 1000℃, and metal deposition uses evaporation or sputtering techniques to enhance conductivity and corrosion resistance, further improving chip performance.

[0040] After the metamaterial chip was fabricated, FDTD Solutions was used to perform full-band terahertz simulation of the "+" composite bimodal terahertz metamaterial sensor. Figure 2 The image shows a simulated terahertz transmission spectrum. The simulation results indicate that the metamaterial sensor has two characteristic peaks in the terahertz band, at 1.22 THz and 2.30 THz.

[0041] like Figure 3 This represents the spatial distribution of the electric field intensity around the "+"-shaped composite terahertz metamaterial sensor, calculated using FDTD. Figure 3 (a) shows the spatial distribution of the electric field intensity at 1.22 THz. Figure 3 (b) shows the spatial distribution of the electric field intensity at 2.30 THz. The electric field intensity gradually changes from blue to red, with the red region representing the highest value and the dark red region corresponding to the maximum value. The region with the strongest electric field intensity is considered a terahertz hotspot, which typically enhances the terahertz wave signal. In-depth analysis of the simulation results can provide a theoretical basis for selecting a suitable resonant frequency, further optimizing the performance of the metamaterial sensor.

[0042] After simulation using FDTD software, the metamaterial structure theoretically possesses the ability to detect trace substances. To verify the detection capability of the metamaterial sensor chip fabricated based on this metamaterial structure for trace carbendazim solutions, subsequent work will proceed from a practical perspective, configuring carbendazim solutions with different concentration gradients. Through THz spectral acquisition, algorithm model construction and optimization, and other methods, the practical application effects and significance of this metamaterial structure will be demonstrated.

[0043] Another embodiment of the present invention provides a highly sensitive detection method for carbendazim based on terahertz metamaterial resonance enhancement, wherein the terahertz metamaterial absorber is the aforementioned terahertz metamaterial absorber, and the detection method includes the following steps (1) to (6):

[0044] (1) Prepare multiple carbendazim standard solutions of different concentrations, and drop the carbendazim standard solutions onto the terahertz metamaterial absorber and dry them in a drying oven.

[0045] The concentration of the carbendazim standard solution was 100 μg / ml. To detect trace amounts of carbendazim residue, the carbendazim standard solution was diluted with ultrapure deionized water using a pipette, and 21 different concentration gradients of carbendazim solutions were prepared according to national standards. The prepared sample solutions were vortexed for 3 minutes to ensure sufficient dilution in deionized water and then stored in volumetric flasks. The specific concentration gradients of the carbendazim solutions are shown in Table 1.

[0046] Table 1. Concentration gradient of carbendazim solution

[0047]

[0048] (2) The dried terahertz metamaterial absorber was placed in the terahertz system and measured using the transmission mode to collect the original terahertz spectral information of the corresponding sample.

[0049] The terahertz experimental equipment used for the detection of trace carbendazim solution residues was the TAS7500 terahertz time-domain spectrometer from Advantest, Japan. This system consists of three main parts: a femtosecond laser, a terahertz emitter, and a terahertz detector. The spectral measurement range is 0.1–5.0 THz, and the instrument resolution is 7.6 GHz. To avoid interference from the surrounding environment and to obtain stable sample information, the system was preheated using an air compressor with an air dryer before the experiment, and dry air was continuously filled into the optical cavity. The humidity was maintained below 10%, and the temperature was kept constant at 25 ± 0.5℃, monitored by a hygrometer. To reduce the influence of concentration residues on the metamaterial sheet, 20 μl of carbendazim sample solution was pipetted onto the metamaterial sheet in ascending order of sample concentration. Since water strongly absorbs terahertz waves, the metamaterial sheet with the sample was placed in a drying oven at 50℃ for 30 minutes before sample spectrum acquisition to ensure the solution on the metamaterial sheet was thoroughly dried. Finally, the metamaterial sheet containing the sample was placed into the terahertz system and measured using transmission mode. After each sample was placed, a 2-minute wait was allowed for the system to stabilize before measurement began. To reduce the impact of random errors on the experimental results, 5 points were taken for each sample, and 10 measurements were performed at each point. A total of 1050 spectra were obtained from the 21 concentration samples.

[0050] First, terahertz time-domain spectra of 21 different concentrations of carbendazim were collected, and the frequency domain signals were obtained by fast Fourier transform (FFT), expressed as formula (1):

[0051]

[0052] Where A(ω) represents the electric field amplitude, Let E(t) be the phase difference between the reference signal and the sample signal, and E(t) be the terahertz time-domain signal. The refractive index and absorption coefficient of the sample are obtained by formulas (2) and (3).

[0053]

[0054] Where n(ω) is the refractive index, α(ω) is the absorption coefficient, ω is the frequency, k(ω) is the extinction coefficient, ρ(ω) is the amplitude ratio function, d is the sample thickness, and c is the speed of light in vacuum.

[0055] Subsequently, the terahertz transmission spectrum of carbendazim was extracted by comparing the sample spectrum and the reference spectrum.

[0056] T = (A S / A R ) 2 (4)

[0057] Where A S and AR These are the amplitudes of the sample signal and the reference signal, respectively.

[0058] The detection model for carbendazim solution content is based on the correlation coefficient of the prediction set (Ri). P ), Root mean square error of prediction set (RMSEP), Related coefficient of correction set (R²) C The model is evaluated using the root mean square error of correction set (RMSEC), where R0... P and R C R is between 0 and 1 P and R C The closer the value is to 1, the higher the accuracy of the model. The closer the values ​​of RMSEP and RMSEC are and the smaller they are, the more stable the model is.

[0059]

[0060]

[0061] The limit of detection (LOD) refers to the minimum concentration of a substance that can be detected. The LOD with a confidence interval of 99.86% can be calculated from the slope of the fitting curve between the true and predicted values ​​established using THz transmission spectroscopy and the variance of the prediction error.

[0062]

[0063] Where σ is the standard error of the predicted concentration, m is the slope of the fitted curve, and RMSEP in the model is equal to the value of σ.

[0064] (3) Use Mahalanobis distance to remove outliers from the original terahertz spectral information to obtain the remaining spectral data.

[0065] Among these methods, a cross-shaped composite metamaterial combined with THz-TDS technology was used to detect and analyze carbendazim solution. Figure 4The transmission spectra of a partial concentration of carbendazim solution on the surface of the metamaterial sensor are shown. Considering the significant noise interference in the frequency bands below 1.0 THz and above 3.0 THz, to ensure accuracy, the transmission spectra in the 1.0-3.0 THz band were selected, containing 263 spectral data points for subsequent modeling. Although the carbendazim content on the surface of the "+" composite metamaterial is extremely low, a clear amplitude response pattern was observed in the transmission peaks after signal enhancement sensing. Specifically, the transmission peak amplitude decreased with increasing carbendazim solution concentration. This phenomenon may be due to the increased absorption of terahertz waves by more carbendazim molecules as the concentration increases, coupled with a change in the refractive index of the metamaterial surface, leading to a weakening of the transmitted wave energy and a decrease in the transmission peak amplitude.

[0066] During spectral data acquisition, changes in the measurement environment and the stability of the spectrometer may cause some samples to exhibit abnormal spectra. These abnormal data can interfere with model training and affect the final prediction accuracy; therefore, they need to be screened out before modeling.

[0067] The Mahalanobis Distance (MD) method used in this invention to remove outliers normalizes variables using the covariance matrix, enabling the identification of correlations between features. In spectral data processing, absorbance values ​​at different wavelengths often exhibit correlations, and Mahalanobis distance can effectively detect outliers that deviate significantly from the overall distribution. Furthermore, spectral data is typically high-dimensional, and Mahalanobis distance can measure the degree of deviation of a sample from the overall distribution in multidimensional space. Therefore, it is more commonly used for spectral outlier detection than one-dimensional statistical methods (such as Z-scores or mean ± standard deviation).

[0068] The Mahalanobis distance threshold is set to 0.99; points exceeding this value are considered outliers. Figure 5 This is the result of the Mahalanobis distance algorithm for filtering. A total of 115 outlier spectral data points were removed, leaving 935 spectral data points.

[0069] The original 1050 data points and the remaining 935 spectral data points after Mahalanobis distance removal were used to build PLS (Partial Least Squares) and LS-SVM (Least Squares Support Vector Machine) models, and the model performance before and after outlier removal was compared. The results are shown in Table 2. The results show that for both models, RP and RMSEP were improved after outlier removal. Therefore, using Mahalanobis distance to remove outliers from spectral data is an effective method to improve the prediction accuracy and robustness of the model, reducing the interference of outliers on the modeling process and thus improving the reliability of the analysis results.

[0070] Table 2 Validation results of the two models before and after outlier removal.

[0071]

[0072] (4) Use multivariate scattering correction to preprocess the remaining spectral data, and then use the iterative information-preserving variable method to extract features from the preprocessed data.

[0073] In spectral analysis, raw data is often affected by various external factors, such as ambient light interference, baseline drift, light scattering effects, and instrument noise. These factors can distort spectral signals, reduce data reliability, and thus affect the accuracy of modeling and analysis. Therefore, spectral data must be preprocessed before data modeling to improve data quality and make it more suitable for subsequent analysis.

[0074] In this application, the Discrete Cosine Transform (DCT) and the First Derivative (1) are employed. st Four commonly used preprocessing methods are D), multiplicative scattering correction (MSC), wavelet transform (WT), etc. They can effectively reduce and remove interference factors such as random noise, background interference, baseline drift, and light scattering effects, thereby improving the stability and accuracy of modeling.

[0075] The preprocessed data was divided into a calibration set and a prediction set at a ratio of approximately 3:1 using the Kennard-Stone (KS) algorithm, and then input into the PLS model. The results are shown in Table 3. The results show that the PLS model performance was reduced after preprocessing with DCT and WT algorithms. This may be because the preprocessing methods, in removing noise, smoothing data, and correcting the baseline, may have lost high-frequency detail information, weakening or distorting the key features of the original spectral data, making the model less adaptable and reducing its accuracy and stability. In contrast, after 1... st The data preprocessed by the D and MSC algorithms can better preserve the characteristics of the original data, while reducing noise and other interference from the original spectrum, thus improving the model performance of PLS. After 1 st The data preprocessed by D showed the best results, with a predictive correlation coefficient R0. P The value reached 0.8961, and the root mean square error (RMSEP) reached 0.4435.

[0076] Table 3 Validation results of PLS ​​models with different preprocessing methods

[0077]

[0078] Similarly, the calibration and prediction sets partitioned by the KS algorithm were input into the LS-SVM model, and the results are shown in Table 4. The results show that the performance of the LS-SVM model improved and decreased after preprocessing. The best preprocessing method was MSC, with a prediction correlation coefficient R0. P The value reaches 0.9766, and the root mean square error (RMSEP) reaches 0.2349. At this point, the parameter combination is γ = 1192.861, σ... 2 =73.2734. In this embodiment, MSC was ultimately selected as the preprocessing method.

[0079] Table 4. Validation results of LS-SVM models with different preprocessing methods

[0080]

[0081] After preprocessing the original data, the model achieved good results. However, the spectral data has a high dimensionality, containing 263 variables, only a few of which are meaningful for the target analysis. High-dimensional data not only increases computational cost but also easily leads to the "curse of dimensionality," meaning the model is susceptible to data noise and overfitting. Feature extraction can reduce the dimensionality of high-dimensional data, remove redundant information, and retain the most representative features, thereby reducing computational complexity and improving analytical efficiency. Furthermore, by removing irrelevant or redundant bands, feature extraction helps the model focus on learning the parts meaningful to the target variable, improving the model's generalization ability, avoiding overfitting, and improving its performance on new data. Therefore, to remove redundancy, noise, and irrelevant information, this application will employ feature extraction methods to help extract key information from the data, improve analytical accuracy, and enhance model stability.

[0082] This application will use five commonly used feature extraction algorithms on the preprocessed data: Iteratively Retaining Informative Variables (IRIV), Principal Component Analysis (PCA), Successive Projection Algorithm (SPA), Competitive Adaptive Reweighting Sampling (CARS), and Genetic Algorithm (GA) to extract features, and compare the model performance of each feature extraction algorithm. Figure 6 1 st The variables selected after feature extraction from the data processed by D.

[0083] The spectral data after feature extraction were input into the PLS model, and the model results are shown in Table 5. The results show that the best performing model is CARS-PLS, with a prediction correlation coefficient (RP) of 0.9484 and a root mean square error (RMSEP) of 0.3175. As can be seen from the table, the model performance slightly decreased after PCA and GA feature extraction. This may be because feature extraction and dimensionality reduction failed to preserve the class distribution or nonlinear relationships between features, resulting in the loss of key information and preventing the model from effectively learning the discriminative features of the original data. Furthermore, if the number of features after dimensionality reduction is too small, the model's expressive power may be limited, potentially introducing noise or losing data discriminability.

[0084] Table 5. Validation results of PLS ​​models using different feature extraction methods.

[0085]

[0086]

[0087] Similarly, Figure 7 The variables selected after feature extraction of the data processed by MSC.

[0088] The extracted spectral data were input into the LS-SVM model, and the model results are shown in Table 6. The results show that the LS-SVM model generally outperforms the PLS model, with the MSC-IRIV-LS-SVM model based on the RBF kernel function showing the best performance and the highest prediction correlation coefficient R0. P The prediction root mean square error (RMSEP) reached 0.9825, and the prediction root mean square error (RMSEP) reached 0.2001. At this point, the parameter combination was γ = 10.9260, σ... 2 =2.3702.

[0089] Table 6. Validation results of LS-SVM models using different feature extraction methods.

[0090]

[0091] (5) A quantitative detection model for carbendazim terahertz transmission spectrum was established based on the extracted features and LS-SVM.

[0092] This invention ultimately selects to use multivariate scattering correction to preprocess the remaining spectral data, and then uses the iterative information-preserving variable method to extract features from the preprocessed data. Based on the extracted features and LS-SVM, a quantitative detection model for carbendazim terahertz transmission spectrum is established (this invention is referred to as the MSC-IRIV-LS-SVM model).

[0093] The MSC-IRIV-LS-SVM model exhibits the best overall performance because: 1) Different concentrations of solutions dropped onto the surface of terahertz metamaterials and dried result in varying residue thicknesses, leading to scattering effects. The MSC preprocessing algorithm is commonly used to correct the impact of scattering on the spectrum. Therefore, MSC enhances the spectral performance in the model; 2) IRIV iteratively optimizes feature subsets, while LS-SVM iteratively optimizes model parameters. The two algorithms have a hierarchical relationship in their optimization logic, making the entire modeling process more consistent with data characteristics. Therefore, IRIV-LS-SVM effectively improves the model's prediction accuracy and generalization ability; 3) Compared to other feature extraction algorithms such as CARS and SPA, IRIV not only considers the importance ranking of variables during feature selection but also avoids local optima through an inverse variable selection strategy, making the selected feature subset more stable and robust. Combined with LS-SVM, the model can fit within a more optimal feature space, further improving the modeling performance.

[0094] Figure 8 The figure shows the fitted curve of the optimal model prediction and the actual value of the transmission spectrum of terahertz metamaterial carbendazim solution using LS-SVM. The slope of the fitted curve reaches 0.8933, and the LOD is 0.672 μg / ml according to formula (9). This verifies the high accuracy and feasibility of the "+" composite metamaterial structure and the MSC-IRIV-LS-SVM model in the analysis of trace carbendazim residues.

[0095] To explore the statistical significance of the optimal MSC-IRIV-LS-SVM model, this invention used F-test and t-test to validate the predicted and actual values. The validation results are shown in Tables 7 and 8. The results show that in the F-test, "F = 0.994" < "one-tailed critical value = 1.213", and "(F ≤ f) = 0.52" > 0.05. In the t-test, "t = 0.071" < "two-tailed critical value = 1.964", and "(T ≤ t) double-tail = 0.944" > 0.05. There was no significant difference between the predicted and actual values ​​of carbendazim content; therefore, the MSC-IRIV-LS-SVM model has good statistical robustness and predictive effectiveness.

[0096] Table 7 F-test Results

[0097]

[0098] Table 8 t-test results

[0099]

[0100] (6) Obtain the terahertz spectral information of the sample containing carbendazim to be tested and input it into the quantitative detection model to obtain the content of carbendazim in the sample to be tested.

[0101] In summary, this invention provides a highly sensitive detection method for carbendazim based on terahertz metamaterial resonance enhancement. It proposes a terahertz metamaterial sensor with a "+"-shaped composite bimodal structure. Transmission spectra were acquired using terahertz spectroscopy, and the differences in terahertz spectra at different concentrations were investigated. Analysis revealed that the amplitude of the transmission peak in the terahertz transmission spectrum of this structure decreases with increasing carbendazim solution concentration. Data preprocessing was performed using multivariate scattering correction (MSC), feature extraction was conducted using iterative information-preserving variable (IRIV) method, and a quantitative detection model for carbendazim terahertz transmission spectra was established based on LS-SVM. This provides a new approach for the quantitative detection of carbendazim. Experimental results show that the model has a high R-value. P The assay yielded a value of 0.9825, an RMSEP of 0.2001, and a detection limit (LOD) of 0.672 μg / ml. This invention provides a novel method for the rapid, non-destructive, and highly sensitive detection of carbendazim in food.

[0102] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

Claims

1. A highly sensitive detection method for carbendazim based on terahertz metamaterial resonance enhancement, characterized in that, The terahertz metamaterial sensor has the following parameters: a period P1 of 70 μm, a gold coating, and a silicon substrate; the terahertz metamaterial absorber is composed of multiple periodic structural units, each periodic structural unit being a square structure with a width of 70 μm. Each periodic structural unit contains four cross-shaped structures, and each periodic structural unit has a strip-shaped structure around its perimeter. The width L1 of each cross-shaped structure is 30 μm, and each cross-shaped structure is composed of two mutually perpendicular "I"-shaped structures. The width L3 at the end of each "I"-shaped structure is 4 μm, and the width W2 in the middle of each "I"-shaped structure is 2 μm. The length L2 of each strip-shaped structure is 60 μm, and the width W1 of each strip-shaped structure is 2 μm; the refractive index of the silicon substrate is 3.335, and the thickness of the silicon substrate is 500 μm. The detection method includes the following steps: (1) Prepare multiple carbendazim standard solutions of different concentrations, and drop the carbendazim standard solutions onto the terahertz metamaterial absorber and dry them in a drying oven; (2) Place the dried terahertz metamaterial absorber into the terahertz system and use the transmission mode to measure and collect the original terahertz spectral information of the corresponding sample. (3) Use Mahalanobis distance to remove outliers from the original terahertz spectral information to obtain the remaining spectral data; (4) Use multivariate scattering correction to preprocess the remaining spectral data, and then use the iterative information-preserving variable method to extract features from the preprocessed data; (5) A quantitative detection model for carbendazim terahertz transmission spectrum was established based on the extracted features and LS-SVM; (6) Obtain the terahertz spectral information of the sample containing carbendazim to be tested and input it into the quantitative detection model to obtain the content of carbendazim in the sample to be tested.

2. The method for high-sensitivity detection of carbendazim based on terahertz metamaterial resonance enhancement according to claim 1, characterized in that, The terahertz metamaterial absorber was prepared by the following method: First, a silicon substrate is selected. The substrate surface is cleaned to remove impurities and then oxidized to enhance the adhesion of the photoresist. Subsequently, a chromium layer is deposited on the silicon substrate using thin film deposition technology, followed by the deposition of a metal layer to enhance the adhesion between the metal layer and the silicon substrate. Then, a layer of photoresist with a thickness of 1-2 μm is coated onto the silicon wafer using a spin coating process; After coating, pre-baking is performed at 100°C for 120 seconds to remove the solvent from the photoresist, ensuring it hardens and is ready for exposure; After exposure, the development process begins, with the development time controlled at 30 seconds, to remove the photoresist from the unexposed areas and leave the pattern. After development, a solvent is used for photoresist coating to remove residual photoresist. Then, reactive ion etching or ion beam etching is used for etching. After etching, the photoresist mask is removed by ultrasonic cleaning with acetone to ensure that there are no residues on the chip surface. Finally, surface oxidation or metal deposition post-treatment is performed to enhance conductivity and corrosion resistance.

3. The terahertz metamaterial resonance enhanced method for high-sensitivity detection of carbendazim according to claim 1, characterized in that, Step (1) specifically includes: The concentration of the carbendazim standard solution was selected as 100 μg / ml. The carbendazim standard solution was diluted with ultrapure deionized water using a pipette to prepare 21 carbendazim solutions with different concentration gradients. The prepared sample solutions were placed on a vortex mixer and shaken for 3 min to ensure that the carbendazim standard solution was fully diluted in deionized water, and then stored in volumetric flasks.

4. The terahertz metamaterial resonance enhanced method for high-sensitivity detection of carbendazim according to claim 1, characterized in that, Step (2) specifically includes: Before the experiment, the system was preheated using an air compressor equipped with an air dryer, and dry air was continuously filled into the optical cavity. The humidity was kept below 10% and the temperature was kept constant at 25±0.5℃ by monitoring with a thermometer and hygrometer. Using a pipette, 20 μl of carbendazim sample solution was dropped onto the terahertz metamaterial absorber in order of increasing sample concentration. Before acquiring the sample spectrum, the terahertz metamaterial absorber with the sample was placed in a drying oven at 50℃ for 30 min. Finally, the terahertz metamaterial absorber with the sample was placed into the terahertz system and measured using transmission mode. After each sample was placed, a 2-minute wait was made for the system to stabilize before starting the measurement. Five points were taken for each sample, and 10 measurements were performed at each point. A total of 1050 spectra were obtained from 21 concentration samples.

5. The highly sensitive detection method for carbendazim based on terahertz metamaterial resonance enhancement according to claim 1, characterized in that, Step (3) specifically includes: The variables are normalized by the covariance matrix, the correlation between the features is identified, the Mahalanobis distance threshold is set to 0.99, and then outliers in the original terahertz spectral information are removed to obtain the remaining spectral data.

6. The terahertz metamaterial resonance enhanced method for high-sensitivity detection of carbendazim according to claim 1, characterized in that, In step (5), F-test and t-test are used to verify the predicted and actual values ​​of the quantitative detection model of carbendazim terahertz transmission spectrum.

Images