Raman spectroscopic analysis method for various halogenated phenols based on hydrogen bond enhanced halogen bond recognition

Abstract

The application discloses a method for analyzing multiple halogenated phenols based on hydrogen bond enhanced halogen bond recognition, which mainly comprises the following steps: constructing a selective recognition surface enhanced Raman substrate, selectively capturing target molecules, collecting signals and extracting chemical characteristics, embedding physical / chemical prior knowledge into machine learning multi-component decoupling, and the like. The application overcomes the interference of complex matrix in water, realizes selective capture, breaks through the black box limitation of traditional pure data driven model, constructs a physical information guided machine learning model, and realizes the detection of multiple halogenated phenols. The application can be used for selectively capturing halogenated phenol persistent organic pollutants (such as trichlorophenol, tribromophenol, tetrabromobisphenol A and the like) in a water phase environment containing background inorganic salt or non-halogenated aromatic organic matter interference, and realizes the qualitative screening of target pollutants in combination with a surface enhanced Raman spectroscopy technology.

Description

Technical Field

[0001] This invention belongs to the field of analytical detection technology and provides a variety of Raman spectroscopy analysis methods for halogenated phenols based on hydrogen bond-enhanced halogen bond recognition. Background Technology

[0002] With the acceleration of industrialization, a large number of new persistent organic pollutants (such as various halogenated phenol homologues) with diverse types and highly similar structures have emerged in water bodies. Traditional targeted analysis methods usually rely on a "one-to-one" detection mode. Although this method has high quantitative accuracy, it is time-consuming and costly to establish standard curves and perform quantitative detection one by one for the ever-increasing number of new pollutants. It is difficult to meet the needs of rapid, high-throughput screening and real-time monitoring of multi-component pollutants in complex water bodies. Halogenated phenols (such as trichlorophenol, tribromophenol, tetrabromobisphenol A, etc.) are widely used industrial raw materials and are a class of persistent organic pollutants with high toxicity and bioaccumulation. In-situ, multi-component online analysis and monitoring of these pollutants in complex water bodies is of great significance.

[0003] Surface-enhanced Raman spectroscopy (SERS) has great potential for application in the screening of multi-component environmental pollutants due to its high sensitivity and molecular fingerprint spectral characteristics. However, when applying SERS to the online monitoring of multi-component environmental water bodies, the following technical bottlenecks still exist: First, real-world water bodies (such as industrial wastewater and surface water) have complex matrices, typically containing inorganic background salt ions and large organic molecules like humic acid. These complex matrices easily cause salt shielding and non-specific interference to conventional electrostatic adsorption probes. Furthermore, water molecules in water act as strong hydrogen bond donors and acceptors, interfering with the weak supramolecular interactions between the probe and the target. Although current techniques attempt to repel water molecule interference by introducing a physically hydrophobic layer, this often increases the spatial distance between the target molecules and the metal substrate, leading to an exponential decay of the surface-enhanced Raman plasmon resonance (SMR) signal. This technical contradiction—the difficulty in simultaneously repelling hydration interference and maintaining short-range, high-sensitivity signals—weakens the target capture capability and detection specificity of conventional sensors in real-world water phases.

[0004] Secondly, due to their highly similar molecular skeletons, different halogenated homologues exhibit dense and severe overlap of their Raman characteristic peaks in multi-component mixed systems. Existing pure data-driven machine learning analytical models, when processing such complex overlapping spectra, rely heavily on "black box" mapping mechanisms for blind feature optimization, easily extracting environmental baseline drift, complex background fluorescence, and random noise as classification features, thus leading to overfitting. Especially when it is difficult to obtain a large number of high-quality field water samples as training sets, the generalization ability of conventional algorithm models in real complex water bodies decreases, making it difficult to achieve accurate decoupling of multiple components of structurally similar pollutants.

[0005] Therefore, there is a need in this field for an online monitoring technology that can overcome the interference of complex water matrix, achieve selective capture, and improve the accuracy of multi-component decoupling by combining physical laws, in order to meet the current demand for multi-component screening of new pollutants with similar structures. Summary of the Invention

[0006] In view of this, the purpose of this invention is to provide a variety of halogenated phenol Raman spectroscopy analysis methods based on hydrogen bond-enhanced halogen bond recognition (HBeXB), aiming to solve the problems of insufficient selectivity, reduced probe binding ability due to water molecule competition, and overfitting of multi-component overlapping spectra in conventional data models in the detection of complex water bodies by existing surface-enhanced Raman technology.

[0007] To achieve the above objectives, the present invention provides the following technical solution: The specific steps of Raman spectroscopy analysis of various halophenols based on hydrogen bond-enhanced halogen bond recognition are as follows: S1. Construction of surface-enhanced Raman substrates: Long-chain alkane molecules and rigid nitrogen-containing heteroaromatic ring probe molecules are co-assembled on the surface of a metal substrate to obtain surface-enhanced Raman substrates; S2. Selective capture of target molecules: The surface-enhanced Raman substrate is placed in the water sample to be tested. The hydrophobic dehydration microenvironment constructed by long-chain alkane molecules and the rigid nitrogen-containing heteroaromatic ring probe molecules generate hydrogen bond-enhanced halogen bond (HBeXB) supramolecular three-dimensional key-locking recognition of halogenated phenols, thereby achieving selective capture of halogenated phenols. S3. Signal Acquisition and Chemical Feature Extraction: The surface-enhanced Raman substrate after capturing halogenated phenols was laser-excited using a Raman spectrometer, and the surface-enhanced Raman spectrum was acquired and chemical features were extracted. S4. Decoupling of Machine Learning with Embedded Physical / Chemical Prior Knowledge: Construct a Physical Information Guided Machine Learning (PIML) model, extract weighted features from surface-enhanced Raman spectra using a pre-defined feature mask matrix, and introduce a penalty term based on the monotonically changing polarizability of halogen atoms in the target molecule into the model's loss function as a mathematical constraint for data analysis.

[0008] Preferably, the halogen atom in the halophenol is chlorine (Cl), bromine (Br), or iodine (I).

[0009] More preferably, the halogenated phenol is selected from trichlorophenol, tribromophenol, tetrabromobisphenol A, and triiodophenol.

[0010] Preferably, in step S1, the metal substrate is made of gold or silver, and its surface has nano-gap and plasmon hot spots. It is prepared by conventional chemical deposition, hydrothermal synthesis or galvanic cell replacement methods in the art.

[0011] Preferably, in step S1, the specific method for co-assembly modification is as follows: first, dissolve the long-chain alkane molecule and the rigid nitrogen-containing heteroaromatic ring probe molecule in anhydrous ethanol to obtain a mixed solution; then, completely immerse the metal substrate in the mixed solution and incubate it for self-assembly at room temperature in the dark for 3 to 6 hours. After that, remove the metal substrate, rinse it with anhydrous ethanol, and then blow it dry.

[0012] Preferably, in step S1, the long-chain alkane molecule is selected from straight-chain alkane thiols with 8 to 18 carbon atoms; the rigid nitrogen-containing heteroaryl ring probe molecule is selected from monocyclic nitrogen-containing heteroaryl thiols or fused-ring nitrogen-containing heteroaryl thiols.

[0013] More preferably, the rigid nitrogen-containing heterocyclic aromatic ring probe molecule is selected from one or more of mercaptopyridine, mercaptopyrimidine, mercaptopyrazine, mercaptotriazine, mercaptoimidazolium, mercaptotriazole, mercaptotetrazole, mercaptothiadiazole, mercaptoquinoline, mercaptobenzimidazole, mercaptobenzothiazole, or mercaptopurine.

[0014] More preferably, the long-chain alkane molecule is n-dodecylthiol (1-DDT), and the rigid nitrogen-containing heteroaromatic ring probe molecule is 4-mercaptopyridine (4-MPy).

[0015] More preferably, in step S1, the specific method for co-assembly modification is as follows: prepare a mixed solution of 1-DDT and 4-MPy in ethanol; then completely immerse the metal substrate in the mixed solution, incubate for self-assembly at room temperature in the dark for 3-6 hours, remove the metal substrate, rinse with anhydrous ethanol, and then blow dry; wherein, the concentration of 1-DDT and 4-MPy in the mixed solution is 1×10⁻⁶. -3 mol / L.

[0016] For even better results, the incubation time is 4 hours.

[0017] Preferably, in step S3, the excitation wavelength of the Raman spectrometer is 785 nm.

[0018] Preferably, in step S3, the chemical feature extraction includes: the characteristic vibrational peak shifts and peak intensity variations of the probe molecule induced by the hydrogen bond-enhanced halogen bond supramolecular binding effect, as well as the characteristic peaks of the target molecule's own skeleton.

[0019] Preferably, the specific method of step S4 is as follows: (S4-1) Constructing a physical prior feature mask: Preset a feature mask matrix for the feature bands where polarizability-related physical frequency shifts are expected (e.g., 1060 cm⁻¹). -1 1220 cm -1 1580 cm -1The region near the target analyte, as well as the redistribution of low-frequency skeleton characteristic peaks and high-frequency double peaks caused by the steric hindrance of polycyclic aromatic rings, and the characteristic bands that generate the target analyte's own skeleton peaks (such as triiodophenol at 170 cm⁻¹). -1 Tribromophenol at 232 cm -1 The feature peaks appearing at the location are assigned higher feature extraction weights; at the same time, the extraction weights of irrelevant frequency bands such as water fluorescence and inorganic background salts are reduced to reduce data input interference from environmental noise. (S4-2) Construct a loss function with polarizability monotonicity constraints: Based on the trend of polarizability of halogen atoms in the target molecule exhibiting I > Br > Cl in quantum chemistry, the Raman frequency shift amplitude of the induced probe characteristic peak is positively correlated with this polarizability; during the model training phase, the total loss function L is set. total : Where L data The loss term is the data fitting loss, and λ is a hyperparameter that adjusts the strength of the physical constraints. A penalty term L is introduced to penalize violations of monotonicity for the displacement amplitude of the characteristic peak. physics Its mathematical expression is: in, , and These represent the Raman frequency shift amplitudes induced by the corresponding halogenated (I, Br, Cl) phenols extracted by the model, respectively, and ɛ is the set tolerance threshold. During the parameter optimization iteration process, if the displacement characteristics predicted by the model do not conform to the monotonicity law of the polarizability, the output value of the penalty term is increased to increase the total loss, thereby constraining the algorithm model to converge within the parameter space that conforms to the physical law. Finally, the qualitative classification and semi-quantitative prediction results of multiple halogenated phenol homologues in the mixed water sample are output.

[0020] The beneficial effects of this invention are: This invention provides a Raman spectroscopy analysis method for various halogenated phenols based on hydrogen bond-enhanced halogen bond recognition. The method mainly includes steps such as constructing a surface-enhanced Raman substrate, selectively capturing target molecules, signal acquisition and chemical feature extraction, and multi-group decoupling using machine learning embedded with physical / chemical prior knowledge. This invention overcomes the interference of complex aqueous matrices, achieves selective capture, and breaks through the black-box limitations of traditional pure data-driven models by constructing a physical information-guided machine learning (PIML) model, enabling the detection of various halogenated phenols. This invention can be used in aqueous environments containing background inorganic salts or non-halogenated aromatic organic compounds for the selective capture of persistent halogenated phenolic pollutants (such as trichlorophenol, tribromophenol, tetrabromobisphenol A, etc.), and combined with surface-enhanced Raman spectroscopy for the analysis and screening of target pollutants.

[0021] The specific analysis is as follows: 1. Overcome the drawbacks of the mutual repulsion between physical hydrophobic extraction and surface-enhanced Raman short-range enhancement, and achieve synergistic effect of microenvironment. This invention constructs a low-dielectric-constant hydrophobic dehydration microenvironment on a surface-enhanced Raman spectroscopy (SERS) substrate through the synergistic co-assembly of a long-chain alkane (e.g., 1-DDT) and a rigid nitrogen-containing heterocyclic probe (e.g., 4-MPy). This microenvironment effectively reduces hydrogen bond competition interference from bulk water molecules for supramolecular recognition, allowing the weak interaction between the probe and the target to remain stable in an aqueous environment. Simultaneously, the rigid probe molecules guide the polarization perturbation of the target molecules into the plasmon electromagnetic field enhancement region. This design overcomes the technical deficiency of relying solely on a long-chain physical hydrophobic layer, which leads to the attenuation of SERS enhancement with spatial distance, achieving a synergistic effect of anti-hydration interference and signal transduction.

[0022] A surface-enhanced Raman spectroscopy (SERS) substrate is placed in the water sample to be tested. Long-chain alkane molecules in the synergistic modification layer locally construct a low-dielectric-constant hydrophobic dehydration microenvironment on the metal surface, repelling water molecules. Within this microenvironment, the nitrogen-containing heterocycle of a rigid nitrogen-containing heterocyclic probe molecule acts as an acceptor, generating hydrogen-bonded halogen bond-enhanced supramolecular three-dimensional key-locked recognition with the hydroxyl and halogen atoms of halophenols diffused from the aqueous phase. This selectively anchors the halophenols within the plasmon hotspot range of the SERS substrate. Since supramolecular recognition depends on σ-holes with sufficient positive charge strength, the halogen atoms in the halophenols are limited to chlorine (Cl), bromine (Br), or iodine (I), which have high polarizability.

[0023] 2. Enhancing selective recognition based on hydrogen bond-enhanced halogen bond bonding This invention utilizes the structural characteristics of the target pollutant. When it simultaneously possesses a hydrogen bond donor (-OH) and a halogen bond donor with high polarizability (-Cl, -Br, -I), the highly polarizable halogen atoms form a hydrogen-bonded halogen-bonded supramolecular network with the nitrogen-containing acceptor of the probe molecule in a dehydrated microenvironment. This dual verification mechanism based on specific molecular spatial structure and electrical characteristics effectively overcomes the defect of traditional electrostatic adsorption probes being easily shielded by inorganic background salts in environmental water. More importantly, this mechanism eliminates interference from single supramolecular interactions. Experiments have shown that aromatic organic compounds lacking halogen atoms (such as phenol) or polyhalogenated aromatic hydrocarbons lacking hydrogen bond donors (such as bromopentafluorobenzene and dibromotetrafluorobenzene) cannot form effective supramolecular anchoring with the probe because they can only provide single hydrogen or halogen bond interactions. Only target molecules that simultaneously meet the conditions of hydrogen and halogen donation structures can trigger the synergistic effect of hydrogen-bonded halogen bonds and be selectively captured, thereby significantly improving the selectivity of the sensor in complex interference matrices.

[0024] 3. Multidimensional physical modulation of the spectrum using dual channels of polarizability and steric hindrance. Due to the differences in the polarizability of halogen substituents in different halogenated phenols, and the steric hindrance effect of polycyclic aromatic molecules (such as tetrabromobisphenol A), the target compound causes varying degrees of perturbation to the probe's electron cloud upon binding. This not only induces Raman shifts in the direction and amplitude of the probe's characteristic peaks, but also causes a regular redistribution of skeletal vibrations and characteristic peak intensities. This multidimensional physical modulation mechanism provides richer characteristic spectral information for the subsequent decoupling of highly similar homologues.

[0025] 4. Introduce Physical Prior Machine Learning (PIML) to improve the robustness of multi-group decoupling in complex water bodies. To address the problem of severe Raman spectral overlap and strong background interference in real-world water environments, this invention constructs a physically-guided machine learning model. On one hand, by introducing a physical prior feature mask to extract active bands and transforming the objective law of characteristic peak shifts caused by polarizability variations (I > Br > Cl) into a constraint term in the model's loss function, the optimization of the model in non-physical spaces is limited. This reduces the risk of overfitting by mistaking environmental stray fluorescence or random noise for classification features, thus improving the model's generalization ability in high-noise environments. On the other hand, the embedding of physical prior knowledge effectively reduces the algorithm's solution space, enabling the model to establish effective classification boundaries with a smaller training sample size, alleviating the difficulty of obtaining high-quality labeled samples in in-situ monitoring of water quality in real-world environments.

[0026] Other advantages, objectives, and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination, or may be learned from practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description

[0027] To make the objectives, technical solutions, and advantages of the present invention clearer, the preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings, wherein: Figure 1 4-MPy modified substrate for the detection of tribromophenol; Figure 2 1-DDT-modified substrates for the detection of tribromophenol; Figure 3 1-DDT / 4-MPy modified substrate for the detection of tribromophenol; Figure 4 1-DDT / 4-MPy modified substrate for phenol detection; Figure 5 1-DDT / 4-MPy modified substrate for the detection of bromopentafluorobenzene; Figure 61-DDT / 4-MPy modified substrate for the detection of dibromotetrafluorobenzene; Figure 7 1-DDT / 4-MPy modified substrate for triiodophenol detection; Figure 8 1-DDT / 4-MPy modified substrate for trichlorophenol detection; Figure 9 1-DDT / 4-MPy modified substrate for the detection of tetrabromobisphenol A; Figure 10 Surface-enhanced Raman spectroscopy (SERS) for TIP, TBP, and TCP is performed at 1060 cm⁻¹. -1 Nearby characteristic peaks shift; Figure 11 Surface-enhanced Raman spectroscopy (SERS) of TIP, TBP, and TCP was performed at 1220 cm⁻¹. -1 Nearby characteristic peaks shift; Figure 12 Surface-enhanced Raman spectroscopy (SERS) of TIP, TBP, and TCP was performed at 1580 cm⁻¹. -1 Nearby characteristic peak shifts and new TIP and TBP peaks; Figure 13 Signal response of tribromophenol and its mixed solutions with interfering substances on surface-enhanced Raman substrates; Figure 14 Detection of surface-enhanced Raman signal intensity of tribromophenol (TBP) and TBP mixed solutions with different interfering substances on modified surface-enhanced Raman substrates; Figure 15 Detection of different concentrations of triiodophenol on a 1-DDT / 4-MPy modified substrate; Figure 16 Detection of different concentrations of tribromophenol on a 1-DDT / 4-MPy modified substrate. Detailed Implementation

[0028] The present invention will be further described below with reference to the accompanying drawings and embodiments. It should be noted that the following description is only for explaining the present invention and does not limit its content.

[0029] Example 1: Preparation and testing process of surface-enhanced Raman spectroscopy substrate (1) Preparation of surface-enhanced Raman spectroscopy (SERS) active metal substrates: Surface-enhanced Raman spectroscopy (SERS) substrates of gold and silver noble metals with nano-interstic gaps and plasmon hot spots on the surface were prepared using conventional chemical deposition, hydrothermal synthesis, or galvanic cell replacement methods in the art. Taking copper foam@silver SERS substrate as an example, copper foam was cut into small pieces of 0.5 × 0.5 cm, ultrasonically cleaned with isopropanol and anhydrous ethanol for 15 min respectively, rinsed with deionized water, and then dried and stored. The replacement solution was a mixture of silver nitrate and polyvinylpyrrolidone (molecular weight ~40000) in ethanol, with a silver nitrate concentration of 1 × 10⁻⁶. -2 The concentration of silver nitrate was mol / L, and the molar ratio of silver nitrate to polyvinylpyrrolidone was 1:1. Before the reaction, the cleaned copper foam was soaked in dilute sulfuric acid for 2 min to remove the oxide layer on the copper surface. After cleaning, the copper foam was immersed in a displacement solution (30 mL solution) and reacted for 6 min. After the reaction, it was washed with anhydrous ethanol.

[0030] (2) Two-component synergistic co-assembly modification: 1×10⁻⁶ components were prepared respectively. -3 mol / L 4-mercaptopyridine (4-MPy) with 1×10 -3 A solution of 1 mol / L n-dodecylthiol (1-DDT) in anhydrous ethanol was prepared by mixing 4-MPy and 1-DDT solutions at a volume ratio of 1:9. The surface-enhanced Raman spectroscopy (SERS) substrate was immersed in this mixture and incubated for 4 hours at room temperature in the dark. A synergistic modification layer composed of 1-DDT and 4-MPy was spontaneously formed on the substrate surface through strong covalent bonds between the thiol groups and the silver surface. The substrate was then removed, rinsed with anhydrous ethanol, and dried for later use.

[0031] (3) Selective capture and surface-enhanced Raman spectroscopy acquisition: The above-mentioned surface-enhanced Raman substrate was placed in the water sample to be tested (which can be integrated into the flow cell of online water quality monitoring). The hydrophobic microenvironment constructed by 1-DDT at the interface reduces the interference of water molecules; 4-MPy uses the lone pair electrons of the nitrogen-containing heterocycle as an acceptor to form hydrogen bonds with the halophenol molecules, thereby enhancing the halogen bond and anchoring them. After incubation, surface-enhanced Raman spectroscopy was performed using a Raman spectrometer (excitation wavelength 785 nm).

[0032] Example 2: Experimental verification of the synergistic mechanism between hydrophobic microenvironment and receptor Experimental Design: Three different surface-enhanced Raman spectroscopy (SERS) substrates were prepared to detect tribromophenol (TBP) in a pure water system (1×10⁻⁶ substrates). -4 mol / L): Comparative Example 2A: 4-MPy probes were modified only on a metal substrate.

[0033] Comparative Example 2B: Long-chain 1-DDT modified only on a metal substrate.

[0034] Test group 2C of this invention: 1-DDT and 4-MPy were co-modified using the method of Example 1.

[0035] Experimental results and microscopic mechanism confirmation: Comparative Example 2A: Figure 1 As shown, no obvious tribromophenol characteristic peaks were detected in the surface-enhanced Raman spectrum; they were only detected at 233.6 and 855.5 cm⁻¹. -1 A weak characteristic peak of tribromophenol was observed at 1583.6 cm⁻¹. -1 The characteristic peak shifted to 1587.1 cm⁻¹. -1 The reason is that in a pure aqueous phase, the hydrogen bond network of water molecules competes with 4-MPy and the target substance, hindering effective supramolecular recognition.

[0036] Comparative Example 2B: such as Figure 2 As shown, surface-enhanced Raman spectroscopy also failed to reveal a valid target characteristic signal. This is because, although 1-DDT can physically extract hydrophobic tribromophenol through hydrophobic interactions, its long carbon chains form a thick insulating dielectric layer on the metal surface. This physically blocks the tribromophenol molecules outside the strong plasmon enhancement region, and due to the exponential distance attenuation of the surface-enhanced Raman effect, the Raman signal cannot be detected.

[0037] Test group 2C of this invention: as follows Figure 3 As shown, at 231.5 cm -1 853.8 cm -1 A distinct characteristic peak of tribromophenol was detected at 1033.1 cm⁻¹, and it showed a certain degree of shift compared to the characteristic peak of tribromophenol. The shift in Raman characteristic peaks was due to selective coupling of 4-MPy and TBP. -1 The characteristic peak shifted to 1035.4 cm⁻¹. -1 1063.6 cm -1 The characteristic peak shifted to 1061.9 cm⁻¹. -1 1221.8 cm -1 The characteristic peak shifted to 1218.4 cm⁻¹. -1 1592 cm -1 The characteristic peak shifted to 1588.5 cm⁻¹. -1 And a new peak of 1559.3 cm appeared. -1 The results validated the role of the two-component recognition layer: 1-DDT repelled water molecules, reduced the local dielectric constant, and decreased hydration interference; while the rigid short-chain 4-MPy probes interspersed within it provided specific binding sites, shortening the signal transduction distance. The combination of the two produced a synergistic enhancement effect of physical and chemical processes.

[0038] Example 3: Theoretical Verification of Probe Selectivity and Multidimensional Physical Spectral Modulation Experimental design: A 1-DDT / 4-MPy modified substrate was used to detect aqueous solutions containing common phenol, bromopentafluorobenzene, dibromotetrafluorobenzene, triiodophenol (TIP), tribromophenol (TBP), trichlorophenol (TCP), and tetrabromobisphenol A (TBBPA), all at a concentration of 10. -4 mol / L.

[0039] Comparative Example 3A: 1-DDT / 4-MPy modified substrate for phenol detection.

[0040] Comparative Example 3B: 1-DDT / 4-MPy modified substrate for the detection of bromopentafluorobenzene and dibromotetrafluorobenzene.

[0041] Test group 3C of this invention: The 1-DDT / 4-MPy synergistic modified substrate of Example 1 is used for the detection of triiodophenol, tribromophenol, trichlorophenol and tetrabromobisphenol A.

[0042] After immersion enrichment under the same conditions, surface-enhanced Raman spectroscopy was performed.

[0043] Experimental results and mechanism confirmation: Comparative Example 3A: such as Figure 4 As shown, no obvious phenol signal was observed in the spectrum except for the substrate background. This is because the hydrophobic microenvironment constructed by 1-DDT prevents phenol molecules from approaching the surface to enhance Raman spectroscopy, and the hydroxyl groups on phenol are unlikely to form hydrogen bonds with mercaptopyridine.

[0044] Comparative Example 3B: such as Figure 5 , Figure 6 As shown, no obvious signals of bromopentafluorobenzene and dibromotetrafluorobenzene were observed in the spectrum. Due to the structural limitation of the HBeXB mechanism of "hydrogen donor + halogen donor", ordinary halogenated aromatic hydrocarbons lacking hydroxyl groups cannot form supramolecular structures with the probe.

[0045] Test group 3C of this invention: modified substrate p-triiodophenol ( Figure 7 There was a clear response; 169.6 cm -1 Location, 851 cm -1 A distinct triiodophenol characteristic peak appears at [location missing]. The 4-MPy characteristic peak shifts to 1036.4 cm⁻¹. -1 1060.8 cm -1 1216.9cm -1 1588.4 cm -1 Location, 1538 cm -1 A new peak appears at this location. The substrate contains p-trichlorophenol ( Figure 8 There was also a response, 199 cm -1The characteristic peak of trichlorophenol appeared at 1035 cm⁻¹, and the characteristic peak of 4-MPy shifted to 1035 cm⁻¹. -1 1063 cm -1 1219.5 cm -1 And 1589.5cm -1 Overall, the modified substrate showed a stronger response to halogenated phenols, with triiodophenol being greater than tribromophenol, which in turn was greater than trichlorophenol, consistent with the halogen bond strength rule. For tetrabromobisphenol A (… Figure 9 ), 256 cm -1 Location, 853 cm -1 The characteristic peak of tetrabromobisphenol A appears at 1592.2 cm⁻¹, and the characteristic peak of 4-MPy increases from 1592.2 cm⁻¹. -1 Shifted to 1586.5 cm -1 .

[0046] When detecting TIP, TBP, and TCP, 4-MPy is located at 1060 cm⁻¹. -1 1220 cm -1 The characteristic in-plane bending peak of CH is located at 1580 cm⁻¹. -1 The characteristic peaks of the CC / CN asymmetric stretching vibrations in the nearby ring framework exhibit a regular Raman shift. Due to the physical law that the polarizability of halogen atoms follows I > Br > Cl, their polarization drag on the probe electron cloud decreases during the formation of HBeXB. The experimentally measured peak at 1060 cm⁻¹... -1 The nearby characteristic peaks shifted by 2.7 cm. -1 1.6 cm -1 0.5 cm -1 ( Figure 10 1220cm -1 The nearby characteristic peaks shifted by 4.8 cm. -1 3.3 cm -1 2.2 cm -1 ( Figure 11 1580 cm -1 The nearby characteristic peaks shifted by 3.8 cm. -1 3.7 cm -1 2.7 cm -1 In addition, TIP is at 1538 cm -1 TBP is 1559.3 cm. -1 A new peak appeared ( Figure 12 The displacement amplitude strictly follows the monotonicity of polarizability. This spectral variation caused by polarizability provides more dimensions of physical characteristics for the decoupling of homologous series.

[0047] Example 4: Designing functionalized substrates with interference immunity Experimental design: Prepare a mixed solution of tribromophenol and interfering agents. The concentration of tribromophenol is 1×10⁻⁶. -6 The concentration was mol / L, and interfering substances included total phosphorus standard solution (0.5 mg / L), ammonia nitrogen standard solution (1 mg / L), COD standard solution (20 mg / L), chloride ion (30 mg / L), sulfate ion (30 mg / L), sodium dodecyl sulfonate (1 mg / L), and bromobenzene (1×10 mol / L). -4 mol / L), hexafluorobenzene (1×10 -4 mol / L), sulfamethoxazole (1×10 -4 mol / L), and another 1×10 mol / L of tribromophenol was prepared. -6 mol / L) and triiodophenol (1×10 -6 A mixed solution (mol / L) was used. The mixed solution was analyzed using a 1-DDT / 4-MPy modified substrate.

[0048] Experimental results and mechanisms: such as Figure 13 , Figure 14 As shown, characteristic peaks of tribromophenol (approximately 232 cm⁻¹) were detected in the mixed solutions. -1 Characteristic peaks of two components were detected in a mixed solution of tribromophenol and triiodophenol (triiodophenol approximately 170 cm⁻¹). -1 The presence of interfering substances such as total phosphorus, ammonia nitrogen, high concentrations of COD, and inorganic salt ions did not mask the frequency or quench the intensity of the tribromophenol characteristic peak. This is attributed to the good anti-interference ability of the 1-DDT hydrophobic layer against polar and large organic molecules. Sodium dodecyl sulfonate, as a typical surfactant, usually disrupts the stability of the monolayer, but in this system, the sensing interface maintained good structural integrity due to the strong Ag-S bond formed between 1-DDT and the metal surface. In a mixed solution of tribromophenol and triiodophenol, the substrate was able to simultaneously detect the 232 cm⁻¹ peak. -1 and 170 cm -1 The independent characteristic peaks demonstrate the resolution potential of this method in multi-component mixed pollution events, with a peak at 232 cm⁻¹ in a mixed solution of tribromophenol with chloride ions and bromobenzene. -1 The relatively weak characteristic peaks are attributed to the strong nucleophilicity of chloride ions, an anion with high charge density, in the aqueous phase. Although 1-DDT constructs a hydrophobic microenvironment, trace amounts of Cl... -It is still possible for it to approach the nitrogen atom sites of 4-MPy through electrostatic attraction or localized permeation, thereby occupying some active sites. Secondly, although the bromobenzene molecule lacks a hydroxyl group and cannot trigger the HBeXB mechanism, its bromine atom still has a σ-hole, which can form a single halogen bond with pyridine nitrogen. Due to the simple structure and certain hydrophobicity of the bromobenzene molecule, and its concentration being much higher than that of the target molecule, it can enter the 1-DDT modification layer and compete with the target molecule for nitrogen atom acceptor sites.

[0049] Example 5: Detection of different concentrations of triiodophenol and tribromophenol Experimental design: Prepare concentrations of 10... -5 10 -6 10 -7 10 -8 mol / L solutions of triiodophenol and tribromophenol were used to detect gradient concentration solutions using a 1-DDT / 4-MPy modified substrate.

[0050] Experimental Results: To quantitatively evaluate the sensor's performance, 10 tests were conducted. -5 mol / L to 10 -8 Concentration gradient in mol / L. From Figure 15 , Figure 16 As can be seen from the data, as the concentration decreases, triiodophenol at 170 cm⁻¹... -1 The characteristic peak of triiodophenol is at 232 cm⁻¹. -1 The characteristic peak gradually decreases.

[0051] Example 6: Implementation Scheme for Data Processing Based on Physical Priors-Based Machine Learning (PIML) Algorithm Experimental objective: Based on the multidimensional physical spectral modulation features extracted in Example 3, this example provides a machine learning algorithm implementation architecture that integrates physical laws to reduce the model's dependence on a large amount of training data and improve classification robustness.

[0052] Implementation steps: 1. Data acquisition and preprocessing: Surface-enhanced Raman spectra of mixed water samples (such as those containing TCP, TBP, TBBPA, etc.) were collected, and baseline correction and smoothing were performed.

[0053] 2. Constructing a Physical Prior Feature Mask: In the feature extraction stage, a weight mask matrix is ​​configured based on the aforementioned physical mechanism. For bands dominated by the "polarizability electronic effect" (e.g., 1033 cm⁻¹), -1 1063 cm -1 (nearby) and bands dominated by spatial steric effects (such as the low-frequency skeleton peak of TBBPA and 1590 cm⁻¹). -1Intensity redistribution region), and characteristic bands that produce the target's own skeleton peaks (such as triiodophenol at 170 cm⁻¹). -1 Tribromophenol at 232 cm -1 Characteristic peaks appearing at certain locations are assigned higher extraction weight coefficients; while easily disturbed water fluorescence background regions are assigned lower weights.

[0054] 3. Model Training and Physical Consistency Penalty: Select a suitable machine learning classifier (such as a support vector machine or neural network). In the model's loss function L... total In addition to the classification fitting error term L, data In addition, a regularization penalty term L, designed based on the monotonically changing law of polarizability, is added. physics .

[0055] Where L data The loss term is the data fitting loss, and λ is a hyperparameter that adjusts the strength of the physical constraints. A penalty term L is introduced to penalize violations of monotonicity for the displacement amplitude of the characteristic peak. physics Its mathematical expression is: in, , and These represent the Raman frequency shift amplitudes induced by the corresponding halogenated phenols extracted by the model, and ɛ is the set tolerance threshold. During model training parameter optimization, if the input batch of samples does not simultaneously contain the three target molecule categories corresponding to I, Br, and Cl, the algorithm can use a masking mechanism or adaptive conditional branching to activate only the comparison terms containing currently existing homologues (e.g., when I is missing, only those containing...). , The penalty term is used to ensure the robustness of the model in samples with missing classes.

[0056] 4. Output prediction: Input the sample data to be tested into the trained model, and output the classification information and semi-quantitative concentration range of each halogenated phenol component according to the model classification boundary.

[0057] This method reduces the computational complexity of systematically classifying multihalogenated homologues by extracting specific bands and applying directional penalties.

[0058] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for Raman spectral analysis of various halogenated phenols based on hydrogen bond enhanced halogen bond recognition, characterized by, The specific steps are as follows: S1. Constructing a surface-enhanced Raman substrate: Long-chain alkane molecules containing thiol anchoring groups and rigid nitrogen-containing heteroaromatic ring probe molecules are co-assembled and modified on the surface of a metal substrate to obtain the surface-enhanced Raman substrate; S2. Selective capture of target molecules: The surface-enhanced Raman substrate is placed in the water sample to be tested. The hydrophobic dehydration microenvironment constructed by long-chain alkane molecules and the rigid nitrogen-containing heteroaromatic ring probe molecules generate hydrogen bond-enhanced halogen bond supramolecular three-dimensional key-locking recognition of halogen bonds in hydroxyl and halogen atoms of halogenated phenols, thereby achieving selective capture of halogenated phenols. S3. Signal Acquisition and Chemical Feature Extraction: The surface-enhanced Raman substrate after capturing halogenated phenols was laser-excited using a Raman spectrometer, and the surface-enhanced Raman spectrum was acquired and chemical features were extracted. S4. Machine learning multi-group decoupling with embedded physical / chemical prior knowledge: Construct a machine learning model guided by physical information, extract weighted features from surface-enhanced Raman spectra through a preset feature mask matrix, and introduce a penalty term based on the monotonically changing polarizability of target molecule halogen atoms into the model's loss function as a mathematical constraint for data analysis; The specific method for step S4 is as follows: (S4-1) Constructing a physical prior feature mask: Preset a feature mask matrix, assign high feature extraction weights to the feature bands that are expected to undergo polarizability-related physical frequency shifts, as well as the low-frequency skeleton feature peaks and high-frequency double-peak intensity redistribution intervals caused by the spatial steric hindrance of multi-aryl rings, and the feature bands that generate the skeleton peaks of the target object itself; at the same time, reduce the extraction weights of frequency bands unrelated to water fluorescence and inorganic background salts, so as to reduce the interference of environmental noise in the data input. (S4-2) Constructing loss function containing monotonicity constraint of polarizability: According to the characteristics that the polarizability of halogen atoms of the target molecule in quantum chemistry presents I>Br>Cl, the Raman frequency shift amplitude generated by the characteristic peak of the probe is positively correlated with the polarizability; in the model training stage, the total loss function L total is set as where L data is the data fitting loss, and λ is a hyperparameter that adjusts the strength of the physical constraint; a penalty term L physics is introduced for the feature peak displacement amplitude to penalize the violation of monotonicity, and its mathematical expression is: in, ∆v I , ∆v Br and ∆v Cl α represents the Raman frequency shift amplitude induced by the corresponding halogenated phenols extracted by the model, and α is the set tolerance threshold. During the parameter optimization iteration process, if the displacement characteristics predicted by the model do not conform to the monotonicity law of the polarizability, the output value of the penalty term is increased to increase the total loss, thereby constraining the algorithm model to converge in the parameter space that conforms to the physical law. Finally, the qualitative classification and semi-quantitative prediction results of multiple halogenated phenol homologues in the mixed water sample are output.

2. The method according to claim 1, wherein the method is a multi-halophenol Raman spectroscopic method based on hydrogen bond enhanced halogen bond recognition. The halogen atom in the halogenated phenol is chlorine, bromine, or iodine.

3. The method according to claim 1, wherein the method is a multi-halogen substituted phenol Raman spectroscopic method based on hydrogen bond enhanced halogen bond recognition. In step S1, the material of the metal substrate is selected from gold, silver or copper.

4. The method according to claim 1, wherein the method is a multi-halogen substituted phenol Raman spectroscopic method based on hydrogen bond enhanced halogen bond recognition. In step S1, the specific method for co-assembly modification is as follows: first, dissolve the long-chain alkane molecule and the rigid nitrogen-containing heteroaromatic ring probe molecule in anhydrous ethanol to obtain a mixed solution; then, completely immerse the metal substrate in the mixed solution and incubate it for self-assembly at room temperature in the dark for 3 to 6 hours. After that, remove the metal substrate, rinse it with anhydrous ethanol, and then blow it dry.

5. The method according to claim 1, wherein the method is a multi-halophenol Raman spectroscopic analysis method based on hydrogen bond enhanced halogen bond recognition. In step S1, the long-chain alkane molecule is selected from straight-chain alkathiols with 8 to 18 carbon atoms; the rigid nitrogen-containing heteroaromatic ring probe molecule is selected from mercaptopyridine, mercaptopyrimidine, or mercaptoimidazole.

6. The method according to claim 1, wherein the method is a multi-halophenol Raman spectroscopic analysis method based on hydrogen bond enhanced halogen bond recognition. In step S3, the excitation wavelength of the Raman spectrometer is 785 nm.

7. The method according to claim 1, wherein the method is a multi-halophenol Raman spectroscopic analysis method based on hydrogen bond enhanced halogen bond recognition. In step S3, chemical feature extraction includes: characteristic vibrational peak shifts and peak intensity variations of probe molecules induced by hydrogen-enhanced halogen supramolecular binding, as well as characteristic peaks of the target molecule's own skeletal structure.

Images