Reverse solvent spectrum stripping quantitative analysis method

Abstract

The invention belongs to the technical field of spectral analysis and chemical quantitative detection, and particularly relates to a reverse solvent spectral stripping quantitative analysis method, which comprises the following steps of: acquiring a Raman spectrum by selecting reference solutions with different concentrations, performing spectral subtraction by taking a high-concentration reference spectrum as a reference spectrum and a low-concentration spectrum as a to-be-processed spectrum, and eliminating a solvent background, so as to obtain a Raman spectrum; the method comprises the following steps: obtaining a difference spectrum only containing solute characteristic peaks, carrying out curve fitting on the difference spectrum, extracting a solute Raman characteristic peak intensity value, establishing a linear relation model of concentration and characteristic peak intensity, taking a spectrum of a solution to be detected as a spectrum to be processed, and subtracting the spectrum from a high-concentration reference spectrum to obtain a reverse difference spectrum; and calculating the solute concentration in the solution to be detected according to the established model. The problems of poor detection precision, high detection limit and the like of the low-concentration solute in the prior art are solved, thorough deduction and signal enhancement of a solvent background can be realized under the condition of not depending on pure solvent reference, and high-precision quantitative analysis of the low-concentration solute is realized.

Description

Technical Field

[0001] This application belongs to the field of spectroscopic analysis and chemical quantitative detection technology, and specifically relates to a reverse solvent spectral stripping quantitative analysis method. Background Technology

[0002] Currently, Raman spectroscopy is widely used for qualitative and quantitative analysis of solute components in solution systems due to its advantages of being non-destructive, rapid, and highly sensitive. Commonly used quantitative analysis methods include the internal standard method and the external standard method. A solution system consists of two parts: a solvent and a solute. The solvent is a liquid medium capable of dissolving other substances to form a homogeneous system, such as water, methanol, ethanol, and acetone, and is the main component of the system. The solute is the substance dissolved in the solvent, such as sulfates, nitrates, phosphates, or other ions or organic molecules, and is the target of spectroscopic detection. When a Raman laser irradiates a solution, the solute and solvent molecules each produce characteristic Raman scattering spectra. In the resulting spectra, different chemical bonds or molecular vibrations correspond to different wavenumber positions, forming several prominent signal peaks, which are called Raman characteristic peaks. Among them, the solute characteristic peak is the signal peak generated by the vibration of the target analyte (i.e., solute) molecules. It is the main basis for quantitative and qualitative analysis, and its peak intensity is directly related to the solute concentration. The solvent peak is the Raman scattering peak generated by the solvent molecules. In the internal standard method, it can be used as a reference signal for normalization. It usually has a large peak intensity and will overlap with the solute peak, causing interference. The solvent background signal refers to the continuous or broadband scattering signal formed by the solvent molecules in the entire wavelength range. It will be superimposed on the solute characteristic peak, forming background interference, making it difficult to accurately identify the weak solute peak.

[0003] Traditional internal standard methods involve simultaneously measuring the Raman characteristic peak intensities of the solute and solvent, using the solvent peak as a reference for normalization, and establishing a linear relationship between intensity and concentration. However, at low solute concentrations, the solute characteristic peaks are often obscured by the solvent background signal. Furthermore, the signal-to-noise ratio decreases and the curve fitting accuracy is reduced due to factors such as inconsistent pixel responses of CCD (Charge-Coupled Device) sensors, dark noise (i.e., electrical signal noise generated by the CCD detector itself in the absence of light illumination) and interference from solvent scattering signals, making it difficult to achieve accurate detection of trace concentrations.

[0004] The external standard method uses the spectrum of the pure solvent as a reference spectrum and subtracts the spectrum containing the solute solution to obtain a Raman spectrum containing only the characteristic peaks of the solute. This method can eliminate solvent interference to some extent and improve measurement accuracy. However, when the sample concentration is low, the signal-to-noise ratio decreases significantly, and dark noise fluctuations can mask the characteristic signal of the solute, resulting in insufficient stability and accuracy of curve fitting. The detection limit is the lowest concentration or minimum content of a substance that can be reliably detected, and this detection limit is usually difficult to be lower than 1.6 mg / L. For example, if the detection limit of a Raman method is 1.6 mg / L, it means that when the solute concentration is below 1.6 mg / L, the instrument cannot distinguish between signal and noise.

[0005] Therefore, existing methods suffer from inaccuracies, low signal-to-noise ratios, and high detection limits in the low concentration range, failing to meet the demand for highly sensitive quantitative analysis of trace solutes. There is an urgent need for a Raman spectroscopy quantitative analysis method that can effectively eliminate background interference, improve the signal-to-noise ratio, and raise the detection limit. Summary of the Invention

[0006] To address the technical problems of poor detection accuracy and high detection limits of low-concentration solutes in existing technologies, this application provides a reverse solvent spectral stripping quantitative analysis method. This method uses the spectrum of a high-concentration solution as a reference spectrum and the spectrum of a low-concentration solution as the spectrum to be processed. By performing reverse differential processing on the spectra of high and low concentration solutions, the signal of low-concentration solutes is enhanced and background interference is effectively eliminated, making it suitable for the accurate quantitative determination of low-concentration solutes.

[0007] This application provides a reverse solvent spectral stripping quantitative analysis method, the method comprising:

[0008] Step 1: Select multiple sets of reference solutions with the same solute type and known solute concentration, collect Raman spectra of each reference solution, obtain the corresponding reference Raman spectra, and sort them from high to low solute concentration;

[0009] Step 2: Select the reference Raman spectrum corresponding to the highest concentration reference solution as the reference spectrum, and use the Raman spectra of other low concentration reference solutions as the spectra to be processed. Perform inter-spectral difference calculations on the reference spectrum and each spectrum to be processed on the same wavenumber axis to obtain differential Raman spectra without solvent background and containing only solute Raman characteristic peaks.

[0010] Step 3: Fit the peak shape of the differential Raman spectrum within the preset wavenumber range, extract the peak intensity parameters of the Raman characteristic peaks of the target solute, and construct a linear calibration model between solute concentration and peak intensity parameters based on the peak intensity parameters corresponding to different known concentration reference solutions.

[0011] Step 4: Obtain the Raman spectrum of the test solution, and use the Raman spectrum of the test solution as the spectrum to be processed and the reference Raman spectrum corresponding to the highest concentration reference solution as the reference spectrum to perform interspectral difference operation to obtain the reverse difference Raman spectrum of the test solution. Based on the reverse difference Raman spectrum of the test solution, extract the peak intensity parameters of the corresponding target solute Raman characteristic peaks, and substitute them into the linear calibration model in Step 3 to obtain the concentration of solute in the test solution.

[0012] In a preferred implementation, step 1 further includes:

[0013] Step 1.1: Select the target solute and solvent, and prepare no less than three sets of reference solutions with the same solute species and known solute concentrations, and different from each other, according to the preset concentration gradient;

[0014] Step 1.2: Set the Raman spectroscopy measurement parameters based on the wavenumber position of the Raman characteristic peak of the target solute, and select the reference solution with the highest concentration in the reference solution as the reference reference solution R1, and control the signal-to-noise ratio of the target Raman characteristic peak of the reference reference solution R1 to be within the preset range;

[0015] Step 1.3: Number the reference solutions sequentially from high to low concentration as R1, R2, R3... and corresponding to concentrations N1, N2, N3... respectively. Perform Raman spectroscopy measurements on each reference solution under the same acquisition parameters to obtain the corresponding reference Raman spectra G1, G2, G3... Normalize the spectra of all concentrations of reference solutions based on the height of the solvent characteristic peak to obtain the normalized reference Raman spectra, and sort the reference Raman spectra from high to low concentration.

[0016] In a preferred implementation, further, in step 2, the interspectral difference operation satisfies the following relationship:

[0017]

[0018] In the formula: The difference spectrum is obtained by performing a difference operation between the high-concentration reference spectrum and the low-concentration reference spectrum; The reference Raman spectrum is the corresponding to the low-concentration reference solution numbered i and i>1; The reference Raman spectrum is the one corresponding to the highest concentration reference solution.

[0019] In a preferred implementation, step 3 further includes:

[0020] Step 3.1: Detect the Raman characteristic peak signal of the target solute in the differential spectrum, and determine whether the Raman characteristic peak appears within the preset target wavenumber range and whether its peak shape polarity is consistent with the differential operation direction. At the same time, compare the peak intensity variation trend of the Raman characteristic peak in the differential spectrum corresponding to different concentration samples, determine whether it is monotonically increasing or monotonically decreasing with the change of solute concentration, and calculate the signal-to-noise ratio of the differential spectrum. When the peak amplitude of the Raman characteristic peak is greater than 3 times the noise amplitude and the baseline noise amplitude is less than 5% of the total signal amplitude of the differential spectrum, the differential spectrum is determined to be an effective differential spectrum.

[0021] Step 3.2: After determining the effective difference spectrum, smooth and denoise the effective difference spectrum, and then align the wavenumber axis and correct the peak position to ensure that the position deviation of the characteristic peaks corresponding to adjacent samples does not exceed 0.5. ;

[0022] Step 3.3: After determining the wavenumber range of the Raman characteristic peak of the target solute, perform peak shape curve fitting on the differential Raman spectrum processed in step 3.2. Use the absolute value of the difference between the ordinate of the fitted peak and the baseline, or the absolute value of the integral area between the characteristic peak and the baseline, as the Raman characteristic peak intensity value of the solute. Mark the characteristic peak intensity values ​​corresponding to different concentration samples with each differential spectrum.

[0023] Step 3.4: Using the known solute concentration of each reference solution as the dependent variable and the corresponding standardized characteristic peak intensity as the independent variable, perform linear regression fitting using the least squares method to establish a linear calibration model between solute concentration and standardized characteristic peak intensity, and output the fitting coefficients.

[0024] Step 3.5: Calculate the correlation coefficient R between the known concentration and the standardized characteristic peak intensity based on the regression fitting results of the linear calibration model established in Step 3.4. When R ≥ 0.99, the linear model established in Step 3.4 is deemed to meet the linear correlation requirement. Calculate the mean square error of the fitting residuals of the linear calibration model and compare it with the total variance. When the mean square error of the residuals is less than 5% of the total variance, the linear calibration model is deemed to meet the quantitative accuracy requirement. When both the correlation coefficient R and the mean square error of the residuals are met, the linear calibration model is solidified into a quantitative standard equation for solute concentration inversion calculation.

[0025] In a preferred implementation, further, in step 3.4, the linear relationship model is:

[0026] y = a·x + b

[0027] In the formula: y is the concentration of the solute; x is the standardized characteristic peak intensity I; and a and b are regression coefficients, respectively.

[0028] In a preferred implementation, step 4 further includes:

[0029] Step 4.1: Use a spectroscopic detection device to acquire the Raman spectrum of the test solution, obtain the Raman spectrum of the test solution, and call the reference Raman spectrum corresponding to the highest concentration reference solution stored in advance as the reference spectrum;

[0030] Step 4.2: Using the highest concentration reference spectrum as the reference spectrum, perform interspectral difference operation on the same wavenumber axis for the Raman spectrum to be measured to obtain the reverse difference Raman spectrum to be measured;

[0031] Step 4.3: Select the preset wavenumber interval where the Raman characteristic peak of the target solute is located in the reverse differential Raman spectrum to be measured. Within the preset wavenumber interval, use the Gaussian peak shape function to perform curve fitting on the Raman characteristic peak of the target solute, and use the least squares method to solve the fitting parameters. When the sum of squares of the fitting residuals meets the convergence and the ratio of the fitting residuals to the total variance is less than the preset threshold, the fitting is determined to be effective. When the fitting is effective, the absolute value of the difference between the ordinate of the fitting peak and the fitting baseline is used as the peak intensity parameter of the Raman characteristic peak of the target solute.

[0032] Step 4.4: Substitute the peak intensity parameter into the linear relationship model between solute concentration and standardized characteristic peak intensity established in Step 3.4 to calculate the concentration of the target solute in the test solution;

[0033] Step 4.5: Determine the validity of the concentration results obtained in Step 4.4. When the peak intensity parameter falls within the effective range of the linear relationship model and the proportion of the root mean square error of the fitting residual of the linear relationship model does not exceed the preset threshold, the concentration result is determined to be valid and the solute concentration value of the test solution is output. If the above conditions are not met, the Raman spectrum to be tested is re-acquired and Steps 4.1 to 4.4 are repeated.

[0034] In a preferred implementation, further, in step 1, the concentration difference between each reference solution is in the range of 2-5 times.

[0035] In a preferred implementation, further, in step 1, the signal-to-noise ratio of the Raman characteristic peak of the solute measured in the reference solution R1 is between 20 and 120.

[0036] In a preferred embodiment, further, in step 1, the solvent of the solution includes one of water, alcohol, ketone, acid, and lipid, wherein the water includes one of water, hydrogen peroxide, deuterium water, and tritium water, the alcohol includes one of methanol and ethanol, the ketone includes acetone, the acid includes one of formic acid and acetic acid, and the lipid includes one of ethyl acetate and ethyl butyrate.

[0037] In a preferred embodiment, further, in step 1, the solute in the solution includes one of sulfate, nitrate, carbonate, bicarbonate, borate, phosphate, perchlorate, chlorophyll, water, ethanol, methanol, acetone, acetic acid, and ethyl acetate.

[0038] The beneficial effects of this application are:

[0039] First, the reverse solvent spectral stripping quantitative analysis method of this application, compared with existing technologies, can significantly improve the problems of low signal-to-noise ratio, strong background interference, and high detection limit in low-concentration detection of traditional internal and external standard methods. This method selects multiple sets of reference solutions with known concentrations and uses spectral subtraction between high-concentration and low-concentration reference spectra to effectively strip away systematic errors caused by solvent background and CCD pixel response inconsistencies, thereby obtaining a differential spectrum containing only solute Raman characteristic peaks. Furthermore, by curve fitting of the differential spectrum to extract characteristic peak intensities and establishing a linear relationship model between concentration and peak intensity, the influence of dark noise can be effectively suppressed, improving the recognition of characteristic signals and fitting accuracy. In specific implementation, this invention uses a "reverse" subtraction method, i.e., using a high-concentration reference solution as the reference spectrum and the spectrum of the test solution as the spectrum to be processed, to obtain a reverse differential spectrum, which can enhance the low-concentration solute signal while eliminating the superposition effects of solvent background and systematic noise. Compared to the traditional external standard method, this invention achieves background subtraction without the need for additional measurement of the pure solvent spectrum, avoiding error accumulation caused by differences in solvent signal intensity. Compared to the internal standard method, this invention does not rely on the solvent peak as a reference signal, thus avoiding errors caused by overlap between solvent and solute peaks. Therefore, the method of this invention has the advantages of thorough background subtraction, improved signal-to-noise ratio, high quantitative accuracy, and lower detection limit, achieving quantitative analysis results with high sensitivity, high accuracy, and high stability.

[0040] Secondly, in the preferred implementation, this application achieves systematic configuration and standardized measurement of the reference solution in the initial stage of the experiment. By selecting the target solute and solvent and preparing multiple sets of reference solutions with different concentrations, it can ensure that the concentration gradient covers the target detection range. By setting the spectral measurement parameters according to the position of the solute characteristic peak and controlling the signal-to-noise ratio of the high-concentration reference solution within a reasonable range, it can effectively avoid oversaturation or signal distortion. The Raman spectra of all concentrations of reference solutions are normalized based on the peak height of the solvent characteristic peak, reducing the overall intensity inconsistency caused by factors such as laser power fluctuations, so that the reference spectra of different concentrations are comparable in intensity scale. The spectra are numbered from high to low concentration and measured and archived in sequence, so that the subsequent spectral subtraction operation has strict comparability and data consistency.

[0041] Third, in the preferred implementation, this application ensures the authenticity and effectiveness of the extracted peak signals by detecting characteristic peak signals and verifying their wavenumber range, signal direction, and signal-to-noise ratio, thereby reducing the interference of noise errors. Furthermore, it performs smoothing and noise reduction, wavenumber alignment, and peak position correction on the differential spectrum to control the position deviation of characteristic peaks for samples of different concentrations within 0.5. Within this range, a linear relationship model between concentration and characteristic peak intensity was established. Combined with the dual criteria of correlation coefficient R≥0.99 and residual mean squared error ≤5% of total variance, the model was ensured to have excellent linear correlation and quantitative accuracy.

[0042] Fourth, in the preferred implementation, this application achieves high-precision inversion calculation of the solute concentration of the test solution through step 4. The spectrum of the high-concentration reference solution is used as the reference spectrum and the spectrum of the test sample is used as the spectrum to be processed. Inverse spectral subtraction is performed to effectively remove solvent background and system noise, and obtain an inverse difference spectrum containing only solute characteristic peaks, thereby improving signal purity. Secondly, Gaussian function fitting and least squares optimization are used in the characteristic peak extraction process. Combined with the judgment criterion that the residual / total variance ratio is less than 5%, high fitting accuracy and accurate peak shape identification are ensured. Furthermore, by substituting the obtained characteristic peak intensity into the previously established linear relationship model, the solute concentration is calculated quickly and accurately.

[0043] Fifth, in the preferred implementation, this application controls the concentration difference between each reference solution within the range of 2-5 times, and limits the signal-to-noise ratio of the Raman characteristic peak of the high-concentration reference solution R1 to between 20 and 120. This ensures sufficient spectral signal intensity while avoiding signal saturation caused by excessively high concentrations, ensuring that high-concentration samples have sufficient signal contrast, and keeping low-concentration samples within the detectable range. This effectively improves the sensitivity and stability of spectral subtraction operations. Attached Figure Description

[0044] Figure 1 This is a flowchart of the reverse solvent spectral stripping quantitative analysis method of the present invention;

[0045] Figure 2 The original Raman spectra obtained by using the internal standard method on aqueous sulfate solutions of different concentrations in Example 1 of the present invention are shown below.

[0046] Figure 3 for Figure 2 China 970-985 A magnified view of a portion of the band;

[0047] Figure 4 This is the Raman spectrum of Example 1 of the present invention after removing the background of the pure solvent using the external standard method;

[0048] Figure 5 for Figure 4970-995 A magnified view of the spectrum in the band;

[0049] Figure 6 The reverse differential Raman spectrum obtained by reverse solvent spectral stripping quantitative analysis method in Example 1 of the present invention;

[0050] Figure 7 According to Figure 6 The concentration-peak intensity quantitative relationship curve established by the reverse differential spectroscopy is shown below.

[0051] Figure 8 The quantitative relationship curve in Example 1 of the present invention is obtained by using a solution with a high signal-to-noise ratio as the background reference spectrum.

[0052] Figure 9 This is the quantitative relationship curve when a solution with a low signal-to-noise ratio is used as the background reference spectrum in Example 1 of the present invention;

[0053] Figure 10 In Embodiment 2 of the present invention, noise is introduced and in Reference Raman spectra of simulated Raman characteristic peaks within the wavenumber range;

[0054] Figure 11 In Embodiment 2 of the present invention Figure 10 Noisy Raman spectra with random noise superimposed on noiseless simulated peak shapes;

[0055] Figure 12 This is the difference spectrum after interspectral difference processing in Embodiment 2 of the present invention;

[0056] Figure 13 In Embodiment 2 of the present invention Figure 12 A magnified view of the bottom of the differential spectrum. Detailed Implementation

[0057] To enable those skilled in the art to better understand the technical solutions of this application, the following will provide a more detailed description of this application in conjunction with the accompanying drawings and embodiments.

[0058] The directional terms such as above, below, left, right, front, and back used in this application are based on the positional relationships shown in the attached drawings. Different attached drawings may result in different positional relationships, therefore they should not be interpreted as limitations on the scope of protection.

[0059] In this application, the terms "installation," "connection," "interlocking," "linking," and "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, an integral connection, a mechanical connection, an electrical connection, or a connection that allows communication between components. They can also refer to a direct connection or an indirect connection through an intermediate medium. They can refer to the internal connection of two components or the interaction between two components. For those skilled in the art, the specific meaning of the above terms in this application can be understood according to the specific circumstances.

[0060] The accurate quantitative determination of solute concentration in solution systems is a fundamental requirement in fields such as chemical analysis, environmental monitoring, and biological detection. Raman spectroscopy, as a non-contact, rapid, and highly selective spectroscopic analysis method, can simultaneously identify solute molecular structure and concentration information without the need for chemical colorimetry and complex pretreatment. However, the acquisition of Raman signals is limited by multiple factors, including solute concentration, solvent background, and photodetector performance. The signal-to-noise ratio (SNR) has a decisive impact on detection accuracy and linear fitting results; a high SNR means that the useful signal is much stronger than the noise background, resulting in clear characteristic peaks, stable fitting, small errors, and a lower detection limit. For low-concentration solute systems, the inventors found that even with a high-sensitivity CCD detector and multiple integrations, it is still difficult to accurately distinguish solute characteristic peaks from a strong background spectrum, leading to significant fitting errors and a high detection limit.

[0061] It should be noted that the signal-to-noise ratio (S / N) is calculated as the signal intensity of the solute Raman characteristic peak divided by the spectral background, which includes the intensity of dark noise, scattering, baseline fluctuations, etc. A lower detection limit indicates that weaker signals can be identified.

[0062] Through systematic spectroscopic experiments and data fitting analysis, the inventors confirmed that the root cause of this phenomenon lies in the fact that both traditional internal and external standard methods rely on the original Raman spectral signal for characteristic peak extraction, and their processing includes the superimposed information of solute and solvent. When the solute concentration is low, its characteristic peak intensity is much lower than that of the solvent peak or noise fluctuations. At this time, not only does the background scattering of the solvent cause baseline drift, but the inconsistent response of each pixel in the CCD array also introduces nonlinear artifacts into the spectrum, causing the low-signal solute peak to be covered by noise, thus distorting the linear relationship between Raman characteristic peak intensity and concentration. In other words, the detectable limit of the solute is limited by the superimposed threshold of solvent interference and pixel noise, rather than the instrument sensitivity itself. This is also the fundamental reason why the traditional external standard method is unstable in fitting and cannot further reduce the detection limit when the concentration is below 1.6 mg / L.

[0063] To address the aforementioned bottlenecks, the inventors proposed a novel signal construction approach: a reverse solvent spectral stripping quantitative analysis method. This method no longer uses the pure solvent spectrum as the subtraction benchmark, but instead utilizes the spectrum containing a higher concentration of solute as the reference spectrum. The spectrum of the low-concentration sample is subtracted in reverse, and the net signal of the solute characteristic peaks is amplified through the concentration gradient difference, thus obtaining a high signal-to-noise ratio difference spectrum containing only the solute Raman characteristic peaks. During this process, the solvent background and CCD pixel differences are simultaneously canceled out, and the dark noise effect is significantly suppressed. Subsequently, by extracting the peak intensity of the solute characteristic peaks in the difference spectrum and establishing a concentration linear model, high-precision quantitative analysis can be achieved under low-energy, low-noise conditions. Practical results show that this reverse spectral stripping method can still obtain stable and linear quantitative results even at solute concentrations as low as 0.5 mg / L. The detection limit is significantly lower than that of the traditional external standard method, and the correlation coefficient of the spectral fitting is significantly improved. This method achieves enhanced extraction of low-concentration solute signals and simultaneous noise subtraction, providing an efficient and scalable technical path for the quantitative application of Raman spectroscopy in trace analysis.

[0064] As per the instruction manual Figure 1 This invention provides a reverse solvent spectral stripping quantitative analysis method, comprising:

[0065] Step 1: Select multiple sets of reference solutions with the same solute type and known solute concentration, collect Raman spectra of each reference solution, obtain the corresponding reference Raman spectra, and sort them from high to low solute concentration.

[0066] The purpose of Step 1 is to establish a basic Raman spectral database as a reference standard for subsequent analysis. By selecting multiple solutions with the same solute but different known concentrations and measuring their Raman spectra, we can obtain the variation patterns of Raman spectral signals at different concentrations, observe the changing trends of solute characteristic peaks with concentration, provide high signal-to-noise ratio raw data for difference calculations and linear fitting, and eliminate the influence of systematic errors between solutions, such as CCD response differences and background light. This step lays the foundation for the entire quantitative model and is a prerequisite for spectral subtraction and modeling.

[0067] Specifically, step 1 includes:

[0068] Step 1.1: Select the target solute and solvent, and prepare no less than three sets of reference solutions with the same solute type and known solute concentration, and different from each other, according to the preset concentration gradient.

[0069] Based on the research or detection objectives, determine the types of solutes: Solvents in the solution include, but are not limited to, aqueous solutions such as water, hydrogen peroxide, deuterium water, and tritium water; alcohols such as methanol and ethanol; ketones such as acetone; acids such as formic acid and acetic acid; and lipids such as ethyl acetate and ethyl butyrate. Solutes in the solution include, but are not limited to, sulfates, nitrates, carbonates, bicarbonates, borates, phosphates, perchlorates, chlorophyll, water, ethanol, methanol, acetone, acetic acid, and ethyl acetate.

[0070] Based on the solubility range of the solute and the instrument's detection sensitivity, determine a suitable concentration range (e.g., 0.1-50 mg / L). Weigh the solute using a high-precision electronic balance and prepare at least three sets of solutions with different concentrations, named the reference solution set. To ensure the effectiveness of subsequent linear calibration, set the concentration interval between adjacent reference solutions to be in the range of 2-5 times, ensuring that the Raman spectral signals at different concentrations have distinguishability and linear variation range, for example: 1 mg / L, 3 mg / L, 9 mg / L, 27 mg / L. Prepare at least 10 mL of each solution to meet the requirements of repeatable measurements and background correction.

[0071] After preparation, solute should be fully dissolved by ultrasonic vibration for 5 minutes. Store in an inert material (such as PTFE or quartz bottle) to avoid reaction with the container. Label the concentration (N1, N2, N3, etc.) and allow to stand in a constant temperature environment for at least 30 minutes to stabilize the system.

[0072] Step 1.2: Set the Raman spectroscopy measurement parameters based on the wavenumber position of the Raman characteristic peak of the target solute, and select the reference solution with the highest concentration in the reference solution as the reference reference solution R1, and control the signal-to-noise ratio of the target Raman characteristic peak of the reference reference solution R1 to be within the preset range.

[0073] Raman spectroscopy measurement parameters include preset parameters for dark noise correction, wavenumber calibration, laser power, and integration time. Dark noise correction involves measuring a signal with no light input and setting it as a baseline. Wavenumber calibration is performed using a silicon wafer such as a 520nm. The Raman shift was corrected using standard materials. The laser power and integration time were adjusted to ensure the signal remained within the instrument's linear response range. Additionally, the spot size, incident angle, and depth of focus were kept consistent to reduce system bias.

[0074] Based on literature or preliminary experimental results, determine the position of the Raman characteristic peak of the target solute, for example, 980. Corresponding to sulfate ions. It should be noted that the Raman characteristic peak position refers to the wavenumber position in a Raman spectrum corresponding to the maximum intensity of scattered light produced by the vibrational or rotational modes of specific chemical bonds in the sample molecules; the unit is usually 1 / 2 voltammetric unit (VQU). For example, each chemical bond, such as CH, OH, C=O, and SO, has a different vibrational frequency. These vibrational frequencies are reflected as different wavenumber shifts by Raman scattering. The positions of the characteristic peaks reflect the type of chemical bond and information about the molecular structure. Table 1 shows the correspondence between typical chemical bonds and Raman shifts.

[0075] Table 1

[0076]

[0077] High-concentration samples with a signal-to-noise ratio (SNR) of 20 to 120 and clear peak shapes were selected as reference solutions R1 to ensure the quality of the spectral signal and the accuracy of subsequent differential analysis.

[0078] Step 1.3: Number the reference solutions sequentially from high to low concentration as R1, R2, R3... and corresponding to concentrations N1, N2, N3... respectively. Perform Raman spectroscopy measurements on each reference solution under the same acquisition parameters to obtain the corresponding reference Raman spectra G1, G2, G3... Normalize the spectra of all concentrations of reference solutions based on the height of the solvent characteristic peak to obtain the normalized reference Raman spectra, and sort the reference Raman spectra from high to low concentration.

[0079] The samples are numbered sequentially from highest to lowest concentration, with R1 representing the highest concentration, and R2, R3, and so on, ensuring a correspondence between the numbering and concentration. R1 corresponds to N1, R2 to N2, and R3 to N3. Each group of samples is measured at least three times, with time intervals of approximately 30 seconds to eliminate instantaneous fluctuations. Identical measurement conditions are used, including laser power, exposure time, spot size, and number of integrations. After acquisition, the results from multiple measurements are averaged to generate the final spectral data.

[0080] To reduce overall intensity differences caused by different acquisitions, such as slight variations in laser power or focusing differences, peak height normalization is performed using the stable characteristic peak heights of the solvent in each spectrum. This involves selecting the same solvent characteristic peak in each spectrum, reading its peak height relative to the baseline (e.g., using the characteristic peak of the spectral line of the highest concentration solution as a reference or the average of multiple spectral lines' characteristic peaks as a reference), and then scaling the overall spectral intensity of all other spectral lines proportionally to ensure that the peak height of the solvent characteristic peak is consistent across all spectra, thus obtaining a normalized reference Raman spectrum. All spectra are stored in descending order of concentration: G1, G2, G3… Furthermore, it should be noted that this invention requires selecting a solvent or characteristic peak such that the solute does not interfere with the solvent for that characteristic peak; that is, the solute does not contain solvent characteristic peak spectral lines.

[0081] Step 2: Select the reference Raman spectrum corresponding to the highest concentration reference solution as the reference spectrum, and use the Raman spectra of other low concentration reference solutions as the spectra to be processed. Perform inter-spectral difference operations on the reference spectrum and each spectrum to be processed on the same wavenumber axis to obtain differential Raman spectra without solvent background and containing only solute Raman characteristic peaks.

[0082] The purpose of step 2 is to eliminate solvent background signals and extract a pure signal containing only the solute Raman characteristic peaks. Subtracting the spectrum from the high-concentration reference solution spectrum, using the low-concentration spectrum as the target spectrum, effectively removes the solvent Raman background, eliminates instrument system errors, obtains a differential spectrum containing only solute characteristic peaks, and improves the signal-to-noise ratio of the solute signal in the low-concentration solution, thereby enhancing the accuracy of subsequent fitting. The core function of this step is to achieve solvent removal and signal-to-noise ratio improvement.

[0083] Specifically, the highest concentration reference spectrum is set as the reference spectrum, and other low concentration reference spectra are used as the spectra to be processed. The difference operation between the spectra satisfies the following relationship:

[0084] (1)

[0085] In the formula: The difference spectrum is obtained by performing a difference operation between the high-concentration reference spectrum and the low-concentration reference spectrum; The reference Raman spectrum is the corresponding to the low-concentration reference solution numbered i and i>1; The reference Raman spectrum is the one corresponding to the highest concentration reference solution.

[0086] For example, the difference spectrum is obtained according to formula (1). , At this point, the characteristic peak of the solute is negative. Because the larger the solute concentration difference, the more prominent the characteristic peak in the difference result, thus improving the signal-to-noise ratio and outputting the difference spectrum file. Used for subsequent peak fitting analysis.

[0087] Step 3: Perform peak shape fitting on the differential Raman spectrum within the preset wavenumber range, extract the peak intensity parameters of the Raman characteristic peaks of the target solute, and construct a linear calibration model between solute concentration and peak intensity parameters based on the peak intensity parameters corresponding to different known concentration reference solutions.

[0088] The purpose of step 3 is to establish the mathematical relationship for quantitative analysis, achieving a precise correspondence between solute concentration and Raman intensity. By curve fitting the differential spectrum, the intensity values ​​of the solute characteristic peaks are extracted, and a linear relationship model is constructed using the peak intensities of samples with different concentrations. This yields a linear function of concentration and Raman characteristic peak intensity, corrects experimental systematic errors, and provides a reliable mathematical model for quantitative calculations of unknown samples. This step constructs the core model of the entire quantitative analysis: y = a·x + b.

[0089] Specifically, step 3 includes:

[0090] Step 3.1: Detect the Raman characteristic peak signal of the target solute in the differential spectrum, and determine whether the Raman characteristic peak appears within the preset target wavenumber range and whether its peak shape polarity is consistent with the differential operation direction. At the same time, compare the peak intensity variation trend of the Raman characteristic peak in the differential spectrum corresponding to different concentration samples, and determine whether it is monotonically increasing or monotonically decreasing with the change of solute concentration. Calculate the signal-to-noise ratio of the differential spectrum. When the peak amplitude of the Raman characteristic peak is greater than 3 times the noise amplitude and the baseline noise amplitude is less than 5% of the total signal amplitude of the differential spectrum, the differential spectrum is determined to be an effective differential spectrum.

[0091] It should be noted that peak polarity refers to whether a characteristic peak is a positive peak that bulges upward relative to the baseline or zero line, or a negative peak that dips downward.

[0092] First, within the preset target wavenumber range, such as 950-1000 The differential spectral signal is extracted, and it is determined whether the direction of the characteristic peak is consistent with the subtraction setting of the spectrum. For example, when subtracting a high concentration from a low concentration, it should be a negative peak. If the peak signal direction is opposite or the peak position deviates from the preset range by more than ±1, the signal is not considered. If necessary, readjust the direction of spectral subtraction or calibrate the wavenumber.

[0093] Then, the intensity of characteristic peaks is calculated for the differential spectra of samples at different concentrations, and the trend of their intensity changes with concentration is compared. If the intensity of the characteristic peaks shows a monotonically increasing or monotonically decreasing relationship, the concentration response is considered reasonable. If anomalies are found, such as non-monotonic changes, the sample is re-measured or the abnormal data is removed.

[0094] Select a region without characteristic peaks, such as a region 30 meters away from the main peak. For bands other than the baseband noise amplitude, calculate the signal-to-noise ratio S / N = signal intensity of solute Raman characteristic peak / spectral background. When S / N ≥ 20 and the baseband noise amplitude is less than 5% of the total signal amplitude of the differential spectrum, the spectral signal quality is considered to meet the fitting requirements.

[0095] Step 3.2: After determining the effective difference spectrum, smooth and denoise the effective difference spectrum, and then align the wavenumber axis and correct the peak position to ensure that the position deviation of the characteristic peaks corresponding to adjacent samples does not exceed 0.5. .

[0096] Step 3.2 Improve the accuracy of subsequent fitting by optimizing the differential spectrum.

[0097] The Savitzky-Golay filtering algorithm is used to smooth and reduce noise in the differential spectrum. The optimal window width is 5-11 data points, and the polynomial order is 2-3 to balance smoothness and peak shape preservation. If the noise is high, moving average or repeated integral averaging can be used for further noise reduction. The positions of known solute characteristic peaks in the differential spectrum are calibrated to ensure that the peak position deviation between adjacent samples does not exceed 0.5. The correction method can be the conventional cross-correlation function peak alignment method or the quadratic polynomial interpolation method.

[0098] Step 3.3: After determining the wavenumber range of the Raman characteristic peak of the target solute, perform peak shape curve fitting on the differential Raman spectrum processed in step 3.2. Use the absolute value of the difference between the ordinate of the fitted peak and the baseline, or the absolute value of the integral area between the characteristic peak and the baseline, as the Raman characteristic peak intensity value of the solute. Mark the characteristic peak intensity values ​​corresponding to different concentration samples with each differential spectrum.

[0099] Select the target peak wavenumber range, for example, 970-990. The fitting function, such as the Gaussian function, Lorentz function, or Voigt mixture function, is selected based on the spectral peak shape, and the fitting parameters are calculated using the least squares method. Taking the Gaussian function as an example, the process of curve fitting for the characteristic peaks of the solute in the difference spectrum includes: selecting the wavenumber range containing the characteristic peak as the fitting interval based on the known Raman characteristic wavenumbers of the target solute; preferably, the fitting interval covers ±10 of the center wavenumber of the characteristic peak. Within the range, establish a Gaussian fitting model:

[0100] (2)

[0101] In the formula: A is the characteristic peak amplitude; The wavenumber at the center of the characteristic peak; denoted as the characteristic peak half-width; B is the baseline constant term.

[0102] The Raman light intensity f(x) at wavenumber x in formula (2) is given by a... The Gaussian peak centered on the solute and a constant background term B together constitute the signal. The Gaussian peak represents the characteristic signal of the solute, and the background term B represents the baseline or noise background.

[0103] Initial parameters are set based on measured differential spectroscopy data, where the peak wavenumber corresponds to... The initial value is taken as the location of the maximum signal within the fitting interval, and the initial value of amplitude A is taken as the intensity value at that point. The half-width of the characteristic peak is... The initial value is set to 3-10. The initial value of baseline term B is taken as the baseline average intensity.

[0104] Furthermore, the least squares algorithm is used to solve for the parameters A in the Gaussian function mentioned above. , And B, to minimize the sum of squared residuals between the Gaussian fitting function f(x) and the measured differential spectrum G(x). After obtaining the optimal parameters, the peak intensity of the characteristic peak is calculated, which is defined as the absolute value of the difference between the peak value and the baseline value, and the intensity value is used as the quantitative intensity of the solute Raman characteristic peak.

[0105] The process of calculating the fitting parameters using the least squares method includes: using the measured difference spectrum G(x) and the Gaussian fitting function. The sum of squared differences is taken as the objective function, which is defined as follows:

[0106] (3)

[0107] In the formula: The sum of squared residuals is the sum of squares of the differences between the measured values ​​and the fitted values ​​at all sampling points, used to reflect the overall error of the fitting; n is the total number of data points, i.e. the number of wavenumber points in the measured spectrum, used to calculate the spectral resolution; i is the data point index, i=1,2,3……n, used to traverse each spectral sampling point. Let be the wavenumber of the i-th sampling point, and be the abscissa of the spectrum, used to measure the wavenumber position of the spectrum; The value is the calculated value of the Gaussian fitting function, which is the theoretical fitting strength value calculated by formula (2).

[0108] By analyzing the objective function Regarding each parameter A, , Taking the partial derivatives of B and setting them to zero, we establish a system of four nonlinear equations:

[0109] (4)

[0110] An iterative algorithm is used to solve the system of equations, with gradient descent or the Levenberg-Marquardt algorithm being preferred for iterative optimization. When the number of iterations reaches a preset condition or the objective function converges to its minimum value, the current A is determined. , B and are the optimal fitting parameters. When the percentage of the sum of squared residuals is less than 5%, the fitting accuracy is considered satisfactory. Finally, the optimal parameters A and B are output. , And B, and calculate the intensity of the Raman characteristic peak of the solute. As input data for establishing subsequent linear models, for example, the Raman characteristic peak intensity of the solute in the difference spectrum G2-1 of the solution is Q2-1, and the Raman characteristic peak intensity of the solute in the difference spectrum G3-1 is Q3-1.

[0111] Step 3.4: Using the known solute concentration of each reference solution as the dependent variable and the corresponding standardized characteristic peak intensity as the independent variable, perform linear regression fitting using the least squares method to establish a linear calibration model between solute concentration and standardized characteristic peak intensity, and output the fitting coefficients.

[0112] Using the known concentrations of the reference solution N1, N2, N3… as the ordinate (y) and the corresponding peak intensities Q2-1, Q3-1… as the abscissa (x), a linear regression model is used to establish the linear relationship between concentration and characteristic peak intensity.

[0113] y = a·x + b (5)

[0114] In the formula: y is the concentration of the solute; x is the intensity of the standardized characteristic peak Q; and a and b are the regression coefficients.

[0115] When the number of samples is greater than 3 groups, the least squares method is used for linear regression calculation to minimize the objective function and obtain the best-fit coefficients a and b, thus obtaining a linear relationship model between concentration and characteristic peak intensity. The minimized objective function is:

[0116] (6)

[0117] In the formula: is the sum of squared residuals; i is the i-th sample, i=1,2,…,n; n is the number of samples; For the measured solute concentration values ​​N1, N2, N3…, The standardized characteristic peak intensities are Q2-1, Q3-1, ...; a is the linear regression slope, a fitting parameter that characterizes the sensitivity of concentration to changes in peak intensity; b is the linear regression intercept, a fitting parameter that represents the theoretical intensity shift when the concentration is zero. The fitting residual for a single sample represents the difference between the actual measured value and the model prediction. These are the model's predicted values.

[0118] Take the partial derivatives with respect to a and b respectively and set them to 0:

[0119] (7)

[0120] The optimal regression coefficients are obtained:

[0121] , (8)

[0122] When S(a,b) is minimized, the resulting straight line y=a·x+b is the best linear model between concentration and Raman peak intensity, where a and b are the best fitting coefficients, and the model can most accurately reflect the relationship between concentration and peak intensity.

[0123] Step 3.5: Calculate the correlation coefficient R between the known concentration and the standardized characteristic peak intensity based on the regression fitting results of the linear calibration model established in Step 3.4. When R ≥ 0.99, the linear model established in Step 3.4 is deemed to meet the linear correlation requirement. Calculate the mean square error of the fitting residuals of the linear calibration model and compare it with the total variance. When the mean square error of the residuals is less than 5% of the total variance, the linear calibration model is deemed to meet the quantitative accuracy requirement. When both the correlation coefficient R and the mean square error of the residuals are met, the linear calibration model is solidified into a quantitative standard equation for solute concentration inversion calculation.

[0124] Step 3.5 is used to verify the correlation and accuracy of the linear model of solute concentration and characteristic peak intensity established in step 3.4.

[0125] First, using n sets of sample data with known concentrations ( , ) is the input, where The intensity of the solute Raman characteristic peak or the normalized peak intensity I. For the corresponding solute concentration values, a and b are the linear regression coefficients obtained by the least squares method in step 3.4.

[0126] Then calculate the predicted value. And calculate the linear correlation coefficient R, the formula is:

[0127] (9)

[0128] In the formula: , These are the average values ​​of x and y, respectively.

[0129] According to formula (9), when R≥0.99, it indicates that there is a high linear correlation between the characteristic peak intensity and the solute concentration, and the model is judged to meet the correlation requirement.

[0130] Furthermore, the sum of squared residuals SSE of the model fit is calculated:

[0131] (10)

[0132] Calculate the total sample variance SST:

[0133] (11)

[0134] Calculate the mean square error (MSE):

[0135] (12)

[0136] The degrees of freedom n-2 are derived from the fact that linear regression includes two fitting parameters a and b.

[0137] Define the residual mean squared error proportion E as:

[0138] (13)

[0139] When E < 5%, it indicates that the model error is small and the fitting accuracy meets the requirements for quantitative analysis.

[0140] When both of the following conditions are met simultaneously: correlation coefficient R ≥ 0.99 and residual mean square error E < 5%, the linear relationship model established in step 3.4 is deemed to meet the quantitative analysis accuracy requirements, and this model y = a·x + b is determined as the quantitative standard equation for solute concentration inversion calculation. If either condition is not met, the fitting wavenumber interval can be reselected, or differential spectral optimization can be performed again, i.e., returning to step 3.2 and fitting again, to improve the fitting accuracy.

[0141] Step 4: Obtain the Raman spectrum of the test solution, and use the Raman spectrum of the test solution as the spectrum to be processed and the reference Raman spectrum corresponding to the highest concentration reference solution as the reference spectrum to perform interspectral difference operation to obtain the reverse difference Raman spectrum of the test solution. Based on the reverse difference Raman spectrum of the test solution, extract the peak intensity parameters of the corresponding target solute Raman characteristic peaks, and substitute them into the linear calibration model in Step 3 to obtain the concentration of solute in the test solution.

[0142] The purpose of step 4 is to eliminate the influence of solvent through inverse difference analysis, thereby achieving quantitative calculation of the solute concentration in the unknown solution. By subtracting the Raman spectrum of the test solution as the spectrum to be processed and the high-concentration reference spectrum as the reference spectrum through inverse spectral subtraction, a "reverse difference spectrum" containing only the characteristic peaks of the solute can be obtained. This further reduces noise interference, allows for the inverse estimation of the solute concentration using a linear model, and improves the detection sensitivity and quantitative accuracy of low-concentration solutes. This step is the application stage of the entire method, achieving high-precision inversion of the unknown solution concentration.

[0143] Specifically, step 4 includes:

[0144] Step 4.1: Analyze the test solution using a spectroscopic detection device. Raman spectroscopy was performed to obtain the Raman spectrum of the solution to be tested. It also retrieves the reference Raman spectrum corresponding to the highest concentration reference solution stored in advance. As a reference spectrum.

[0145] Step 4.2: Use the highest concentration reference spectrum The reference spectrum is used for the Raman spectrum to be measured. Perform interspectral difference operations on the same wavenumber axis to obtain the reverse difference Raman spectrum of the target.

[0146] The formula for calculating spectral subtraction is:

[0147] (14)

[0148] The obtained spectrum is a reverse difference spectrum. This spectrum is used to remove the solvent background signal and retain only the solute Raman characteristic peak signal.

[0149] Step 4.3: Analyze the reverse differential Raman spectrum to be measured. A preset wavenumber interval containing the Raman characteristic peak of the target solute is selected. Within this interval, a Gaussian peak shape function is used to fit the Raman characteristic peak of the target solute, and the least squares method is used to solve for the fitting parameters. When the sum of squares of the fitting residuals converges and the ratio of the fitting residuals to the total variance is less than a preset threshold, the fitting is considered valid. When the fitting is valid, the absolute value of the difference between the ordinate of the fitted peak and the fitted baseline is used as the peak intensity parameter of the Raman characteristic peak of the target solute. .

[0150] In this application, a preset threshold is set for the sum of squared fitted residuals to be convergent and for the ratio of fitted residuals to total variance to be less than 5%.

[0151] Fitting using a Gaussian function:

[0152] (15)

[0153] Solving for parameters A using the least squares method , And B.

[0154] Step 4.4: Set peak intensity parameters Substituting the linear relationship model between solute concentration and standardized characteristic peak intensity established in step 3.4, the solution to be tested is calculated. The concentration of the target solute.

[0155] The solute concentration of the solution to be tested is the y value.

[0156] Step 4.5: Determine the validity of the concentration results obtained in Step 4.4, when the peak intensity parameter... If the concentration falls within the effective range of the linear relationship model, and the percentage of the mean square error of the fitting residuals of the linear relationship model does not exceed a preset threshold, the concentration result is deemed valid and the solution to be tested is output. The solute concentration value is determined, and if the above conditions are not met, the Raman spectrum to be measured is re-acquired. Then repeat steps 4.1 to 4.4.

[0157] In this application, the percentage of the mean squared error of the fitting residuals of the linear relationship model does not exceed a preset threshold of 5%.

[0158] Example 1

[0159] To verify the effectiveness and superiority of the reverse solvent spectral stripping quantitative analysis method proposed in this application in Raman quantitative detection, sulfate ions in aqueous solution were selected as the target solute, and the detection performance of the method in this application and the internal standard method and external standard method in the prior art at different concentrations was compared and analyzed.

[0160] The samples tested were aqueous solutions of sulfate at concentrations ranging from 0.52 to 31.25 mg / L. Raman spectroscopy was used for detection, with a wavenumber range of 900-1000. The main focus is on the sulfate characteristic peak at approximately 980°. The comparison methods include the traditional internal standard method, the external standard method (such as patent number 202111655091.9), and the reverse solvent spectral stripping quantitative analysis method of this application, i.e., the reverse external standard method.

[0161] As per the instruction manual Figure 2-3 Raman spectroscopy was performed on sulfate aqueous solutions of different concentrations (31.25 mg / L-0.52 mg / L) using the internal standard method. Figure 2 The images show the raw Raman spectra of sulfate aqueous solutions at different concentrations, including 31.25, 15.625, 7.8125, 3.90625, 2.19726, 1.64795, 0.92697, 0.69522, and 0.52142 mg / L, as well as a pure water sample. The x-axis represents the Raman Shift (σ). The vertical axis represents Raman Intensity (Counts).

[0162] In approximately 980 Sulfate ions are visible at this location. The characteristic peaks of the pure water sample were observed, but as the solution concentration decreased, the overall spectrum gradually converged towards the pure water curve. The pure water sample showed peaks around 1640... A distinct OH bending vibration peak is present at 980°C; this peak is broad and strong, forming a high background signal. Medium and low concentration samples show this peak at 980°C. The characteristic peaks in the region almost overlap with the water sample curve, and the peak shapes are not obvious. Furthermore, due to the combined effects of solvent Raman background, inconsistent CCD pixel response, and dark noise, the sulfate characteristic peaks in low-concentration samples are difficult to distinguish visually across the entire spectrum. Therefore, in the original spectrum of the internal standard method, as the concentration decreases, the intensity of the characteristic peak signal gradually falls below the background fluctuation level, making it difficult to use as a reliable quantitative basis.

[0163] Figure 3 For internal standard method 970-980 The locally magnified spectrum of the band reveals details of the sulfate characteristic peak region. Figure 3As can be seen, the signal-to-noise ratio is low, mainly due to the following factors: baseline unevenness caused by differences in the photoelectric response of CCD pixels; residual solvent background signal; and fluctuations in dark noise in the low-signal region. Although the sample curve is slightly higher than the pure water baseline at approximately 0.5 mg / L, and a slight difference can be visually identified, this difference is insufficient to support stable quantitative analysis. The noise level is on the same order of magnitude as the signal amplitude, resulting in large uncertainties in peak height and integral area extraction, and a significant reduction in the linear correlation of quantitative fitting.

[0164] In summary, in the internal standard method, the original spectrum is affected by both solvent background and differences in instrument response, approximately 980 The characteristic peak signal of sulfate in low-concentration samples is easily drowned out by background noise. Although the characteristic peak can be clearly distinguished in high-concentration samples, the peak height gradually approaches the baseline noise level as the concentration decreases, making it difficult to maintain linearity. Therefore, the internal standard method is insufficient to meet the quantitative accuracy requirements of low-concentration samples.

[0165] As per the instruction manual Figure 4-5 Raman spectroscopy was performed on sulfate aqueous solutions of different concentrations ranging from 31.25 mg / L to 0.52 mg / L using the external standard method. The external standard method aims to eliminate systematic errors caused by differences in CCD pixel response and solvent background by subtracting the background spectrum of the pure solvent, thereby highlighting the characteristic peak signals of the solute. Figure 4 This shows the spectral results after background subtraction using the external standard method. The horizontal axis represents the Raman Shift (...). The vertical axis represents Raman Intensity (Counts). From Figure 4 It can be seen that at 980 Nearby sulfate ions ( The characteristic peaks of the sample were enhanced to some extent, and the high-concentration samples >1.6 mg / L showed a clear trend of peak height change proportional to concentration. Since the external standard method eliminated the inconsistency of instrument pixel response and solvent background signal, the overall baseline was relatively smooth. For example, the high-concentration samples of 31.25 mg / L and 15.625 mg / L showed obvious peak shape differences, which basically conformed to the rule that the peak intensity increases with concentration.

[0166] However, this method still has limitations. For example... Figure 5 As shown, when dealing with low concentration samples (<1.6 mg / L) in the range of 970-995... When the region is magnified locally, obvious spectral fluctuations are visible, with dark noise and characteristic peak signal intensities on the same order of magnitude. As the concentration further decreases, the random fluctuation amplitude of dark noise increases, leading to instability in peak position and intensity. The characteristic peak and noise signal mix, making accurate identification difficult. In samples below 1.6 mg / L, the spectrum exhibits irregular oscillations, with blurred or even disappeared peak shapes, making it impossible to obtain reliable quantitative data through peak height or integrated area.

[0167] In summary, the external standard method significantly improves the errors caused by solvent interference and instrument response differences through background subtraction, achieving relatively stable quantitative analysis in the medium to high concentration range. However, in the low concentration region, especially <1.6 mg / L, dark noise becomes the main interfering factor, making it difficult to distinguish characteristic peaks, reducing signal linearity, and limiting the detection limit. Therefore, although the external standard method has better baseline flatness and signal significance than the internal standard method, it is still insufficient to meet the high-precision quantitative requirements of low-concentration samples.

[0168] As per the instruction manual Figure 6-7 The Raman spectral analysis method proposed in this application was used to detect sulfate aqueous solutions of different concentrations (31.25 mg / L - 0.52 mg / L). This method uses the spectrum of a high-concentration reference solution as the reference spectrum and the spectrum of a low-concentration or analyte sample as the spectrum to be processed. By subtracting the spectra in reverse, the solvent background signal and instrument system error are simultaneously removed, preserving the pure characteristic peak signals of the solute.

[0169] Figure 6 The Raman spectra after reverse spectral stripping are shown. The horizontal axis represents the Raman Shift (...). The vertical axis represents Raman Intensity (Counts). At approximately 980 cm⁻¹, the sulfate ion concentration (...) The characteristic peaks of the sample transformed from positive peaks to a distinct inverted negative peak structure, with symmetrical peak shapes and clear signal contours. Compared with traditional internal and external standard methods, the background signal in the inverted differential spectrum was almost completely eliminated, the baseline was flat and without significant drift, indicating that the difference between the solvent background and the CCD response was effectively separated. The inverted spectral peak intensity of samples at various concentrations showed a stable monotonic variation with increasing concentration, and even low-concentration samples ≤0.5 mg / L maintained clear characteristic peak signals. These results demonstrate that this method can significantly suppress dark noise in the low-signal region, making the characteristic peaks stand out above the background, thereby improving detection sensitivity and quantitative reliability.

[0170] Figure 7 Quantitative relationship curves between peak intensity and concentration obtained from inverse differential spectroscopy are presented. To ensure repeatability, each sample was tested three times, and the results were labeled with different symbols. It can be seen that the data in each group exhibit a good linear distribution on the concentration-peak intensity coordinate, with a linear fitting correlation coefficient R ≥ 0.99. Even at the lowest concentration of 0.5 mg / L, identifiable and trend-compliant data points can still be obtained, indicating that the detection limit of this method is significantly lower than that of traditional internal standard and external standard methods. According to the quantitative analysis criteria of this application, the model is considered to meet the quantitative accuracy requirements when the sum of squared residuals is less than 5% of the total variance. Figure 7As can be seen, the data points obtained by this method have extremely low dispersion, the fitting curve is stable, and it fully meets the accuracy standard.

[0171] As per the instruction manual Figure 8-9 This is a comparative example of the "quantitative relationship curve between peak intensity and concentration obtained based on reverse differential spectroscopy" in the method of this application. This experiment aims to verify the impact of the selection of the signal-to-noise ratio of the reference spectrum on the quantitative accuracy and linear correlation when performing reverse solvent spectral stripping.

[0172] Figure 8 The quantitative curves obtained when using a solution with a high signal-to-noise ratio (SNR) as the reference spectrum for background subtraction are shown. The horizontal axis represents solute concentration (mg / L), and the vertical axis represents Raman peak intensity (Counts). Different symbols represent three sets of repeated test data. It can be seen that when using a reference spectrum with an excessively high SNR for subtraction, the overall background signal remains strong, especially in the low-concentration region, where the curve points almost overlap and are abnormally densely distributed. In this case, the signal changes of the low-concentration sample are masked by the high background uplift effect, resulting in insensitivity to changes in the relative intensity of characteristic peaks, exhibiting a clear compression trend. Therefore, although the overall linear relationship is still visible, the signal accuracy in the low-concentration range decreases, the fitting residual increases, and quantitative inaccuracies occur. An excessively high reference SNR introduces excessive background residue, submerging the effective signal of the low-concentration sample and affecting detection sensitivity and linear stability.

[0173] Figure 9 The results are obtained using a solution with a low signal-to-noise ratio (SNR) as the reference spectrum for background subtraction. In this case, due to the inherently high noise in the subtracted spectrum, the noise is amplified and superimposed on the difference spectrum after inverse subtraction, resulting in a significant decrease in the overall SNR. As seen in the concentration-peak intensity graph, the dispersion of data points increases significantly, especially in the low concentration region (<2 mg / L), where fluctuations are severe, and the curve loses its monotonicity. The noise fluctuations have amplified the peak intensity disturbance to near the signal amplitude, making the linear regression results unstable and difficult to maintain a reliable correlation coefficient R. In summary, an excessively low reference SNR amplifies the influence of random noise, weakening the smoothness and quantitative repeatability of the inverse spectral difference.

[0174] pass Figure 7-9 As can be seen from the comparison, the reverse solvent spectral stripping quantitative analysis method of this application has an optimal range requirement for the signal-to-noise ratio (SNR) of the reference spectrum. When the SNR of the reference spectrum is too high, background residue is enhanced; when the SNR is too low, the noise amplification effect is significant. Comprehensive experimental results show that when the SNR of the reference spectrum is controlled within the range of 20-120, the optimal balance can be achieved between effectively stripping the background and suppressing noise, resulting in a stable linear quantitative relationship. This result verifies the accuracy of the reference solution method described in this application. The method for limiting the signal-to-noise ratio of Raman characteristic peaks to between 20 and 120 is described, and the rationality of the parameter selection in the method of this application is further explained.

[0175] By comparing this application, the internal standard method, and the external standard method, it can be seen that the reverse solvent spectral stripping quantitative analysis method of this application has outstanding advantages in the following aspects, as shown in Table 2:

[0176] Table 2

[0177]

[0178] In summary, the reverse solvent spectral stripping quantitative analysis method of this application can simultaneously eliminate influencing factors such as CCD pixel response inconsistency, solvent background signal, and dark noise interference in a single processing step. Through reverse differential operation, the characteristic peak signals of low-concentration samples are enhanced, achieving a higher signal-to-noise ratio and a lower detection limit. Experimental results show that when the signal-to-noise ratio of the Raman characteristic peaks of the reference solution is controlled within the range of 20-120, an optimal balance can be achieved between effective background stripping and noise suppression. Within this range, the method achieves optimal quantitative linearity and reproducibility, with a fitting residual less than 5% of the total variance and a linear correlation coefficient R ≥ 0.99. This method can still achieve stable quantification at sulfate ion concentrations as low as 0.5 mg / L, demonstrating excellent linear consistency and signal-to-noise robustness, thus improving the sensitivity and reliability of Raman spectroscopy-based quantitative analysis.

[0179] Example 2

[0180] This embodiment uses numerical simulation to verify the effect of interspectral difference operation in step 2 of the reverse solvent spectral stripping quantitative analysis method on improving the detection capability of weak Raman characteristic peaks.

[0181] As per the instruction manual Figure 10 ,exist A set of simulated Raman characteristic peak signals was constructed within the wavenumber range, and the simulated single Raman characteristic peak was located at approximately At this point, different peak intensities were set to represent different concentration levels, with peak intensities of 100, 200, 300, 400, 500, 1000, 2000, and 3000 respectively. No noise was introduced, and the characteristic peaks corresponding to each concentration were [values ​​missing]. Figure 10 Both exhibit good separation and linear intensity variation, and can be regarded as the benchmark Raman spectrum under ideal conditions.

[0182] As per the instruction manual Figure 11 Based on the above noise-free simulated peak shape, random noise with an amplitude of approximately 900 Counts is introduced to simulate instrument noise and baseline fluctuations present in actual Raman measurements, resulting in the following... Figure 11 The Raman spectrum shown is noisy. From... Figure 11It can be seen that when the characteristic peak intensity is 2000 or 3000, the signal-to-noise ratio of the corresponding Raman characteristic peak is greater than approximately 3:1, and the peak intensity parameter can still be stably extracted. However, when the characteristic peak intensity is in the range of 100-1000, the peak shape is severely submerged by noise, making it impossible to accurately obtain its peak intensity, which is actually below the detection limit. This comparative simulation reflects that in the traditional detection mode without inter-spectral differential processing, weak Raman signals are significantly limited by noise, resulting in a high detection limit.

[0183] As per the instruction manual Figure 12 To simulate the process of inter-spectral difference using the highest concentration reference spectrum as the reference spectrum in this invention, a high-intensity characteristic peak with a peak intensity of 10000 Counts is further set in the simulation model above, serving as the characteristic peak signal corresponding to the reference sample. Under noisy conditions, the noisy spectra corresponding to each low-intensity characteristic peak are subjected to inter-spectral difference calculation on the same wavenumber axis as the spectrum of the high-intensity reference characteristic peak, i.e., the following is performed for each wavenumber point:

[0184]

[0185] in, The reference characteristic peak spectrum is for an intensity of 10000. The spectra to be processed are those with different peak intensities ranging from 100 to 3000. This is the inverse subtracted spectrum after differentiation.

[0186] The result after differential processing is as follows Figure 12 As shown, the differential spectrum exhibits a distinct negative characteristic peak at the target peak position. The peak depth varies monotonically with the intensity of the original characteristic peak from 100 to 3000. Even with added noise, the spacing between the differential curves remains clearly discernible, allowing for accurate determination of the characteristic peak intensity changes corresponding to each concentration.

[0187] As per the instruction manual Figure 13 By locally magnifying the bottom of the differential spectrum, the weak signal characteristic peaks with original intensities of 100-1000 still maintain good distinguishability after local magnification. This demonstrates that the detectability of weak signals is significantly improved after the inter-spectral differential processing of the reverse solvent spectrum according to the present invention.

[0188] In summary, from Figure 11 and Figure 12 , Figure 13The comparison shows that without reverse differential processing, characteristic peaks with intensities below approximately 2000 Counts are difficult to reliably identify from the baseline under noise, resulting in a high detection limit. Using the reverse solvent spectral stripping method of this invention, by performing inter-spectral difference between the low-intensity characteristic peak spectrum and a reference characteristic peak with an intensity of 10000 Counts, a differential spectrum with a significant negative characteristic peak is obtained. Further amplification and analysis of the bottom of the differential spectrum allows for stable identification of weak peaks with original intensities of 100-1000 Counts. Therefore, this method can extract more weak signal information, reduce the effective detection limit, and provide a reliable data foundation for peak intensity extraction in step 3 and linear calibration and concentration inversion in step 4.

[0189] The above descriptions are merely embodiments of this application, and common knowledge regarding specific structures and characteristics in the solutions is not described in detail here. It will be apparent to those skilled in the art that this application is not limited to the details of the above exemplary embodiments, and that this application can be implemented in other specific forms without departing from the spirit or essential characteristics of this application. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of this application is defined by the appended claims rather than the foregoing description. Therefore, it is intended that all variations falling within the meaning and scope of equivalents of the claims be included within this application. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A reverse solvent spectral stripping quantitative analysis method, characterized in that, The method includes: Step 1: Select multiple sets of reference solutions with the same solute type and known solute concentration, collect Raman spectra of each reference solution, obtain the corresponding reference Raman spectra, and sort them from high to low solute concentration; Step 2: Select the reference Raman spectrum corresponding to the highest concentration reference solution as the reference spectrum, and use the Raman spectra of other low concentration reference solutions as the spectra to be processed. Perform inter-spectral difference calculations on the reference spectrum and each spectrum to be processed on the same wavenumber axis to obtain differential Raman spectra without solvent background and containing only solute Raman characteristic peaks. Step 3: Fit the peak shape of the differential Raman spectrum within the preset wavenumber range, extract the peak intensity parameters of the Raman characteristic peaks of the target solute, and construct a linear calibration model between solute concentration and peak intensity parameters based on the peak intensity parameters corresponding to different known concentration reference solutions. Step 4: Obtain the Raman spectrum of the test solution, and use the Raman spectrum of the test solution as the spectrum to be processed and the reference Raman spectrum corresponding to the highest concentration reference solution as the reference spectrum to perform interspectral difference operation to obtain the reverse difference Raman spectrum of the test solution. Based on the reverse difference Raman spectrum of the test solution, extract the peak intensity parameters of the corresponding target solute Raman characteristic peaks, and substitute them into the linear calibration model in Step 3 to obtain the concentration of solute in the test solution.

2. The reverse solvent spectral stripping quantitative analysis method according to claim 1, characterized in that, Step 1 includes: Step 1.1: Select the target solute and solvent, and prepare no less than three sets of reference solutions with the same solute species and known solute concentrations, and different from each other, according to the preset concentration gradient; Step 1.2: Set the Raman spectroscopy measurement parameters based on the wavenumber position of the Raman characteristic peak of the target solute, and select the reference solution with the highest concentration in the reference solution as the reference reference solution R1, and control the signal-to-noise ratio of the target Raman characteristic peak of the reference reference solution R1 to be within the preset range; Step 1.3: Number the reference solutions sequentially from high to low concentration as R1, R2, R3... and corresponding to concentrations N1, N2, N3... respectively. Perform Raman spectroscopy measurements on each reference solution under the same acquisition parameters to obtain the corresponding reference Raman spectra G1, G2, G3... Normalize the spectra of all concentrations of reference solutions based on the height of the solvent characteristic peak to obtain the normalized reference Raman spectra, and sort the reference Raman spectra from high to low concentration.

3. The reverse solvent spectral stripping quantitative analysis method according to claim 1, characterized in that, In step 2, the inter-spectral difference operation satisfies the following relationship: ； In the formula: The difference spectrum is obtained by performing a difference operation between the high-concentration reference spectrum and the low-concentration reference spectrum; The reference Raman spectrum is the one corresponding to the low-concentration reference solution numbered i and i>1; The reference Raman spectrum is the one corresponding to the highest concentration reference solution.

4. The reverse solvent spectral stripping quantitative analysis method according to claim 1, characterized in that, Step 3 includes: Step 3.1: Detect the Raman characteristic peak signal of the target solute in the differential spectrum, and determine whether the Raman characteristic peak appears within the preset target wavenumber range and whether its peak shape polarity is consistent with the differential operation direction. At the same time, compare the peak intensity variation trend of the Raman characteristic peak in the differential spectrum corresponding to different concentration samples, determine whether it is monotonically increasing or monotonically decreasing with the change of solute concentration, and calculate the signal-to-noise ratio of the differential spectrum. When the peak amplitude of the Raman characteristic peak is greater than 3 times the noise amplitude and the baseline noise amplitude is less than 5% of the total signal amplitude of the differential spectrum, the differential spectrum is determined to be an effective differential spectrum. Step 3.2: After determining the effective difference spectrum, smooth and denoise the effective difference spectrum, and then align the wavenumber axis and correct the peak position to ensure that the position deviation of the characteristic peaks corresponding to adjacent samples does not exceed 0.

5. ; Step 3.3: After determining the wavenumber range of the Raman characteristic peak of the target solute, perform peak shape curve fitting on the differential Raman spectrum processed in step 3.

2. Use the absolute value of the difference between the ordinate of the fitted peak and the baseline, or the absolute value of the integral area between the characteristic peak and the baseline, as the Raman characteristic peak intensity value of the solute. Mark the characteristic peak intensity values ​​corresponding to different concentration samples with each differential spectrum. Step 3.4: Using the known solute concentration of each reference solution as the dependent variable and the corresponding standardized characteristic peak intensity as the independent variable, perform linear regression fitting using the least squares method to establish a linear calibration model between solute concentration and standardized characteristic peak intensity, and output the fitting coefficients. Step 3.5: Calculate the correlation coefficient R between the known concentration and the standardized characteristic peak intensity based on the regression fitting results of the linear calibration model established in Step 3.

4. When R ≥ 0.99, the linear model established in Step 3.4 is deemed to meet the linear correlation requirement. Calculate the mean square error of the fitting residuals of the linear calibration model and compare it with the total variance. When the mean square error of the residuals is less than 5% of the total variance, the linear calibration model is deemed to meet the quantitative accuracy requirement. When both the correlation coefficient R and the mean square error of the residuals are met, the linear calibration model is solidified into a quantitative standard equation for solute concentration inversion calculation.

5. The reverse solvent spectral stripping quantitative analysis method according to claim 4, characterized in that, In step 3.4, the linear relationship model is as follows: y = a·x + b; In the formula: y is the concentration of the solute; x is the standardized characteristic peak intensity I; and a and b are regression coefficients, respectively.

6. The reverse solvent spectral stripping quantitative analysis method according to claim 5, characterized in that, Step 4 includes: Step 4.1: Use a spectroscopic detection device to acquire the Raman spectrum of the test solution, obtain the Raman spectrum of the test solution, and call the reference Raman spectrum corresponding to the highest concentration reference solution stored in advance as the reference spectrum; Step 4.2: Using the highest concentration reference spectrum as the reference spectrum, perform interspectral difference operation on the same wavenumber axis for the Raman spectrum to be measured to obtain the reverse difference Raman spectrum to be measured; Step 4.3: Select the preset wavenumber interval where the Raman characteristic peak of the target solute is located in the reverse differential Raman spectrum to be measured. Within the preset wavenumber interval, use the Gaussian peak shape function to perform curve fitting on the Raman characteristic peak of the target solute, and use the least squares method to solve the fitting parameters. When the sum of squares of the fitting residuals meets the convergence and the ratio of the fitting residuals to the total variance is less than the preset threshold, the fitting is determined to be effective. When the fitting is effective, the absolute value of the difference between the ordinate of the fitting peak and the fitting baseline is used as the peak intensity parameter of the Raman characteristic peak of the target solute. Step 4.4: Substitute the peak intensity parameter into the linear relationship model between solute concentration and standardized characteristic peak intensity established in Step 3.4 to calculate the concentration of the target solute in the test solution; Step 4.5: Determine the validity of the concentration results obtained in Step 4.

4. When the peak intensity parameter falls within the effective range of the linear relationship model and the proportion of the root mean square error of the fitting residual of the linear relationship model does not exceed the preset threshold, the concentration result is determined to be valid and the solute concentration value of the test solution is output. If the above conditions are not met, the Raman spectrum to be tested is re-acquired and Steps 4.1 to 4.4 are repeated.

7. The reverse solvent spectral stripping quantitative analysis method according to claim 1, characterized in that, In step 1, the concentration difference between the reference solutions is in the range of 2-5 times.

8. The reverse solvent spectral stripping quantitative analysis method according to claim 2, characterized in that, In step 1, the signal-to-noise ratio of the Raman characteristic peak of the solute measured in the reference solution R1 is between 20 and 120.

9. The reverse solvent spectral stripping quantitative analysis method according to claim 1, characterized in that, In step 1, the solvent of the solution includes one of water, alcohol, ketone, acid, and lipid. The water includes one of water, hydrogen peroxide, deuterium water, and tritium water. The alcohol includes one of methanol and ethanol. The ketone includes acetone. The acid includes one of formic acid and acetic acid. The lipid includes one of ethyl acetate and ethyl butyrate.

10. The reverse solvent spectral stripping quantitative analysis method according to claim 1, characterized in that, In step 1, the solute in the solution includes one of the following: sulfate, nitrate, carbonate, bicarbonate, borate, phosphate, perchlorate, chlorophyll, water, ethanol, methanol, acetone, acetic acid, and ethyl acetate.

Images