Search

CN-117630265-B - Method for rapidly and automatically distinguishing components in mixture solution based on TLC-SERS technology

CN117630265BCN 117630265 BCN117630265 BCN 117630265BCN-117630265-B

Abstract

The invention discloses a rapid and automatic method for distinguishing components in a mixture solution based on a TLC-SERS technology, relates to the field of spectrum detection, and aims to solve the problems that TLC-SERS cannot accurately analyze substances with similar structures and is too dependent on subjective interpretation and capability of operators. The invention assembles silver sol nano particles on the surface of the TLC plate to prepare the large-area TLC-SERS substrate with high sensitivity and repeated detection. And (3) spectrum collection, and generating TLC-SERS spectrums of all target substances. The TLC-SERS spectral data was then analyzed using two machine learning algorithms. Compared with the traditional artificial SERS peak analysis method, the proposed machine learning algorithm relies on the overall analysis of spatial information and spectral information of TLC-SERS spectral data, rather than the spectral information determined by the main characteristic SERS peaks, and thus greatly improves the accuracy of SERS peak identification of different substances.

Inventors

  • HASULIGI
  • FANG GUOQIANG
  • CHEN ZHIJUN
  • BAI LINA
  • HU KAI

Assignees

  • 哈尔滨工业大学

Dates

Publication Date
20260512
Application Date
20231128

Claims (10)

  1. 1. A rapid, automated method for distinguishing components in a mixture solution based on TLC-SERS technology, characterized in that it is performed according to the following steps: step 1, preparing silver sol nano particles Ag NPs by using a sol method; step 2, preparing TLC-SERS substrate by using three-phase interface self-assembly method Mixing the prepared silver sol nano particles Ag NPs, methylene dichloride and deionized water according to the volume ratio of (0.01-0.1): 0.1-1): 1 to obtain a solution B, mixing the solution B and n-hexane according to the volume ratio of (2-100): 1, standing for 1-30 min, and transferring the silver sol nano particles Ag NPs onto a TLC plate by adopting a lifting method to obtain a TLC-SERS substrate; Step 3, separating the mixture by TLC-SERS substrate and collecting spectrum Dripping the mixture to be analyzed at the bottom end of the prepared TLC-SERS substrate, drying the mixture to be analyzed to 3-10 min, placing the mixture into a developing agent, developing the mixture for 10-40min, drying the mixture for 10-30min, and then carrying out one-dimensional scanning by adopting a Raman spectrometer along the developing path of the mixture on the TLC-SERS substrate to acquire the SERS spectrum, wherein the scanning step length is 50-200 nm, the scanning time is 5-30 seconds, and the laser power is 10-100 mW; Step 4, automatically analyzing the spectrum acquired on the TLC-SERS substrate by adopting a machine learning algorithm Based on a convolution neural network algorithm and a spectrum angle algorithm in a python program, respectively writing a spectrum classification model and a spectrum similarity model, and inputting the acquired SERS spectrum into the spectrum classification model and the spectrum similarity model for automatic analysis: The spectrum classification model is used for predicting substance type information corresponding to the SERS spectrum, the spectrum classification model adopts a convolutional neural network, the input of the convolutional neural network is SERS spectrum two-dimensional data, and the output of the convolutional neural network is a classification curve corresponding to the substance type information corresponding to the SERS spectrum; The spectrum similarity model is used for calculating the similarity between the standard spectrum and the collected SERS spectrum, and calculating the similarity by utilizing the spectrum angles of the standard spectrum and the collected SERS spectrum to obtain a similarity curve; And meanwhile, accurately identifying the range of different components in the mixture to be analyzed according to the positions of the classification model curve and the similarity model curve on the TLC-SERS substrate.
  2. 2. A method for rapid, automated differentiation of components in a mixture solution based on TLC-SERS technology according to claim 1, wherein the silver sol nanoparticles in step one are prepared as follows: dispersing ascorbic acid and sodium citrate in ultrapure water at room temperature, stirring and dissolving to obtain a solution A, transferring the solution A into a water bath, adding silver nitrate solution into the solution A, boiling for reaction, continuously reacting for 1-3 hours to obtain silver sol nano particles, finally cooling to room temperature after the reaction is completed, collecting Ag NPs, centrifugally washing twice with an aqueous solution, and dispersing the collected centrifugal product in water to obtain the silver sol nano particles.
  3. 3. The method for rapidly and automatically distinguishing components in a mixture solution based on TLC-SERS technology according to claim 2, wherein the mass-volume ratio of the ascorbic acid, sodium citrate and the ultrapure water is (1-100) mg (100-500) mg/1 mL.
  4. 4. The method for automatically distinguishing components in a mixture solution based on the TLC-SERS technology according to claim 2, wherein the mass-volume ratio of the silver nitrate solution to the ultrapure water is 1:1-20.
  5. 5. A rapid, automated method for distinguishing components in a mixture solution based on TLC-SERS according to claim 2, wherein the centrifugation speed is from 5000 to 10000rpm and the centrifugation time is from 10 to 40 to min.
  6. 6. A rapid, automated method for distinguishing between components in a mixture solution based on TLC-SERS technology according to claim 1, wherein the mixture to be analyzed is added in an amount of 1-10 μl.
  7. 7. The method for rapid, automated separation of components in a mixture solution according to claim 1, wherein the automated analysis in step 4 is automated determination of species information in the experimental spectrum by a classification model, and a similarity model compares the similarity of the experimental spectrum to a standard spectrum.
  8. 8. A method for rapid, automated differentiation of components in a mixture solution based on TLC-SERS technology according to claim 1 or 7, wherein said automated analysis of the input of said collected SERS spectra into a spectroscopic classification model is as follows: Predicting the SERS spectrum in a convolutional neural network classification algorithm, wherein the prediction result is substance type information corresponding to the SERS spectrum, and the convolutional neural network comprises a first convolutional layer, a first pooling layer, a second pooling layer, a third convolutional layer, a first full-connection layer, a second full-connection layer and a third full-connection layer; the first convolution layer comprises a convolution kernel of 1*3, the number of the convolution kernels is 10, and the step length is 2, and after Relu activation, a characteristic matrix of 650 x 10 is obtained; The first pooling layer is formed by selecting the pooling mode as the maximum pooling mode, wherein the pooling core size is 1*3, the step length is 2, and the size of an output matrix of the characteristic matrix output by the convolution layer 1 is 325 x 10 after pooling operation; the second convolution layer is characterized in that the convolution kernel size is 1*3, the number of the convolution kernels is 20, and the step length is 2, and after Relu activation, a feature matrix of 163 x 20 is obtained; The second pooling layer is a pooling mode selected as the maximum pooling, the pooling core size is 1*3, the step length is 2, and the output matrix size is 82 x 20; The third convolution layer is characterized in that the convolution kernel size is 1*3, the number of the convolution kernels is 40, and the step length is 1, and after Relu activation, a feature matrix of 82 x 40 is obtained; the first full connection layer is used for fully connecting the extracted and flattened 3280 features into 1000 features; a second full connection layer for connecting 1000 features to 300 features; And the third full connection layer is used for connecting 300 features into the required category number, and the final full connection number in the spectrum classification model is the identification category number.
  9. 9. The method for automatically distinguishing components in a mixture solution based on the TLC-SERS technology according to claim 1, wherein the standard spectrum and the collected SERS spectrum are input into a spectrum similarity algorithm for secondary distinguishing, a numerical value between 0 and 1 is obtained by calculating a spectrum angle between the standard spectrum and the collected SERS spectrum and then performing cosine operation, and if the output numerical value is close to 1, the spectrum similarity is high, namely the result of a convolutional neural network classification model is correct, so that the similarity degree between the standard spectrum and the collected SERS spectrum is determined, and the secondary distinguishing is completed.
  10. 10. A method for rapid, automated differentiation of components in a mixture solution based on TLC-SERS technique according to claim 9, wherein said calculation of the spectral angle is based on the following formula: Wherein, the Is the spectral angle, arccos is the inverse cosine, Is the value at the ith raman shift of the standard spectrum, To test the value at the ith raman shift of the spectrum.

Description

Method for rapidly and automatically distinguishing components in mixture solution based on TLC-SERS technology Technical Field The invention belongs to the field of chromatographic detection, and particularly relates to a method for rapidly and automatically distinguishing components in a mixture by combining SERS technology and thin-layer chromatography technology with a machine learning algorithm. Background Surface enhanced raman scattering thin layer chromatography (TLC-SERS) has shown great potential as a new analytical chemistry tool and has gained increasing research interest due to its high sensitivity and ease of implementation. TLC technology integrates multiple functions cooperatively on a latex plate, achieving separation of the mixture by combining different eluents (mobile phase) and adsorption layers (stationary phase). The SERS technology amplifies molecular vibration spectra based on the plasmon resonance effect between nanostructures, thereby enabling trace-level chemical detection. TLC-SERS can be very successful in detecting single target substances from complex chemical and biological samples, as long as the mobile and stationary phases are collocated. Therefore, TLC-SERS has been widely used in the fields of food safety, agriculture, medical diagnosis, industrial detection, and the like. Conventional SERS detection is typically performed in a wet or dry film forming state. The dynamic SERS method is performed during the transition of the nano-solution from wet to dry to find the best base state that ensures the maximum enhancement effect. Currently, TLC-SERS faces several key challenges (1) TLC-SERS techniques typically involve the loading of gold nanoparticles and silver Nanoparticles (NPs) onto a TLC plate containing a sample for dynamic SERS detection. The concentration and volume of gold sol and silver sol and the like can have great influence on the SERS detection result. Furthermore, this can also create adverse interference with the continuous SERS measurement. (2) The imperfection of TLC technology separation has led to TLC-SERS technology facing greater challenges in detecting a variety of target species. When multiple substances are identified, the fluctuation of SERS signal intensity of different substances is large, and the signal-to-noise ratio is low. Manual analysis SERS spectra can only analyze one or several characteristic peaks of the target substance, making it difficult to perform accurate analysis of the mixture. In particular, facing structurally similar substances, are difficult to separate by TLC and have the same type of chemical bonds. (3) Conventional TLC-SERS techniques act as a manual analysis technique that relies heavily on subjective interpretation and ability of the operator. This approach would be difficult to follow as the number of spectra increases. Disclosure of Invention The invention provides a method for rapidly and automatically distinguishing components in a mixture solution based on TLC-SERS technology aiming at the problems that the TLC-SERS can not accurately analyze substances with similar structures and is too dependent on subjective interpretation and capability of operators. The invention relates to a rapid and automatic method for distinguishing components in a mixture solution based on TLC-SERS technology, which comprises the following steps: step 1, preparing silver sol nano particles Ag NPs by using a sol method; step 2, preparing TLC-SERS substrate by using three-phase interface self-assembly method Mixing the prepared silver sol nano particles Ag NPs, methylene dichloride and deionized water according to the volume ratio of (0.01-0.1): 0.1-1): 1 to obtain a solution B, mixing the solution B and n-hexane according to the volume ratio of (2-100): 1, standing for 1-30min, and transferring the silver sol nano particles Ag NPs onto a TLC plate by adopting a lifting method to obtain a TLC-SERS substrate; Step 3, separating the mixture by TLC-SERS substrate and collecting spectrum Dripping a mixture to be analyzed at the bottom end of the prepared TLC-SERS substrate, drying for 3-10min, placing the mixture into a developing agent, developing for 10-40min, drying for 10-30min, and then carrying out one-dimensional scanning by adopting a Raman spectrometer along a path of developing the mixture on the TLC-SERS substrate to acquire an SERS spectrum, wherein the scanning step length is 50-200nm, the scanning time is 5-30 seconds, and the laser power is 10-100mW; Step 4, automatically analyzing the spectrum acquired on the TLC-SERS substrate by adopting a machine learning algorithm Based on a convolution neural network algorithm and a spectrum angle algorithm in a python program, respectively writing a spectrum classification model and a spectrum similarity model, inputting the acquired SERS spectrum into the spectrum classification model and the spectrum similarity model for automatic analysis, wherein the spectrum classification model is u