Raman spectrum is essentially the inelastic scattering of light at the molecular level. The vibrational energy level of the molecule is affected by the incident light (frequency ω₀）Excitation, that is, from the ground state to the excited state (extremely unstable), and then by emitting scattered photons (frequency ω_R）Return to a lower energy state. ω₀And ω_RThe difference between them corresponds to the vibrational energy levels of molecules. Scattered light relative to incident light (ω₀ ± ω_R）The frequency shift of is represented by the frequency of the vibration peak displayed in the Raman spectrum. Therefore, Raman scattering has great advantages in the study of molecular vibration and rotation^[22]。

As a non-destructive, rapid and highly sensitive detection method of non-polar functional groups, Raman spectroscopy can accurately analyze the micro nano plastics mainly composed of non-polar covalent bonds qualitatively and quantitatively. Raman spectroscopy can identify the chemical composition of micro/nano plastics, so as to understand its source and degradation process. Different types of plastics, such as polyethylene (PE), polypropylene (PP), polyethylene terephthalate (PET), have unique spectral characteristics in Raman spectra^[23]Through the comparison and analysis of these characteristic spectra, we can track the pollution path of micro nano plastics and evaluate its environmental impact.

In addition, Raman spectroscopy can detect the interaction between micro/nano plastics and other pollutants. For example, the surface of micro nano plastic is easy to absorb organic pollutants and heavy metals, which may enter the organism through the food chain, further aggravating the ecological toxicity^[24-25]Through Raman spectroscopy, the existing forms and concentrations of these composite pollutants can be studied in detail, so as to provide a scientific basis for environmental risk assessment.

Plastic samples exposed to the environment will undergo long-term aging and crushing, and the number of small-size micro plastics will increase sharply^[26-27]However, in many studies, small-size micro plastics are often ignored, which is due to the large gap of the interception net used in the sampling process. As Enders et al^[28]In a study, microplastics were collected in the Atlantic Ocean and identified by µ -raman technology. Among them, 64% of the microplastics with sizes less than 40 µ m followed the exponential distribution, and the scale index λ was 1.96. Erni cassola et al^[29]Microplastics less than 40 µ m were identified in the surface water of Plymouth Bay (UK) using µ -raman technology. Schymanski et al^[8]In the study of 38 brands of bottled water, the µ -raman technology was used to identify 80% of plastic particles with particle sizes between 5 and 20 µ M.

The applicability of classical Raman scattering techniques (such as µ -raman) for the analysis of micro nano plastic samples in the environment is narrow, which is due to the low environmental concentration of micro nano plastic and the low Raman scattering efficiency (only 10% under favorable conditions^-8Incident photons converted to Raman photons)^[22]As a result, the target signal is easily masked by noise. The solution is to add the concentration step in the sample pretreatment step to increase the concentration of micro nano plastics in the test sample. Usually, the solution to the operation of the instrument is to extend the laser exposure time and increase the laser intensity, but this will prolong the test time, and even cause damage to the plastic sample by the laser. Therefore, during the actual operation of the instrument, the experimenter should choose the exposure time and laser intensity according to the actual situation of the sample. Another effective way is to replace the ordinary charge coupled device (CCD) with the electron multiplying charge coupled device (EM-CCD). Compared with ordinary CCD, EM-CCD is equipped with a multiplication register, which can amplify the gain by 1000 times^[30]This can greatly save sampling time and suppress the noise of the instrument itself. Dieing et al^[31]The Raman images of polymethylmethacrylate (PMMA) on glass were obtained by ordinary CCD, and the collection time was 36 MS/piece of spectrum. The signal-to-noise ratio of the collected images is very low, and it is almost impossible to distinguish glass and PMMA. Using EM-CCD detector, images with the same signal-to-noise ratio can be obtained in one tenth of the time, and glass and PMMA are very easy to distinguish under the same acquisition time. With the expansion of Raman theory, the generation of surface enhanced Raman spectroscopy and coherent Raman spectroscopy has effectively improved the Raman signal of the analysis target^[32-33]There are many researches on the application of micro nano plastics in the analysis of environmental samples.

Generally, the content of micro/nano plastics in environmental media is low. In order to facilitate detection, scientists often include concentration and enrichment in the process of sample pretreatment, and use the technology of enhancing Raman to improve the Raman signal response value of micro/nano plastics^[32,34]Surface enhanced Raman spectroscopy (SERS) is a commonly used method to concentrate electromagnetic energy through metal nanostructures to enhance Raman spectral signals. The enhancement principle is divided into electromagnetic enhancement and chemical enhancement. First, the incident light drives the conduction band electrons on the metal surface. When the light frequency matches the oscillation frequency of the electrons, local surface plasmon resonance (LSPR) will be generated, resulting in the formation of photoelectric field at the metal nanostructure interface. The intensity of the photoelectric field is much stronger than that of the incident photoelectric field, which can provide about 10⁶∼10⁸Enhancement factor (EF). Secondly, the spatial position of target molecules directly affects the intensity of Raman scattering signal. The gap where plasma coupling occurs on the surface of nanoparticles is usually referred to as "hot spot", where the adsorbed molecules will have a completely different SERS enhancement effect from other places. As the distance from the hot spot gradually increases, the enhancement effect will decay rapidly. Chemical enhancement refers to the interaction between the target material and the metal substrate, in which the adsorption of molecules on the substrate surface undergoes chemical complexation or charge transfer, resulting in changes in the Raman polarization characteristics of molecules, and the EF provided is 10 ∼ 100. When SERS technology is used to detect organic molecules, the enhancement coefficient EF is defined as formula (1).

Where:I_SERSandI_NRSAre the peak strengths of specific peaks in SERS and conventional Raman spectra respectively,N_SERSandN_NRSIs the average number of signal molecules in the focus area.

For the determination of single particle plastics, theN_SERSandN_NRSAll fixed asN =1. It is worth noting that although this facilitates the calculation of EF value, the collected Raman spectrum needs to use the same sample and the same test conditions, especially the distance from the objective lens to the sample, to ensure that the polymer is exposed to the same intensity of laser spot^[35]。

SERS can be realized in two forms: metal nanoarray substrate and metal nanoparticle colloid (e.gTable 1andTable 2As shown in). The metal nano array substrate can be divided into filter membrane type and non filter membrane type. The filter membrane type substrate has the characteristics of filtration and concentration, which can simplify the sample pretreatment steps and detect micro nano plastics in situ on the SERS filter membrane. Klarite reinforced substrate is a common commercial non filtration membrane metal nano array substrate, which can be used as a reference for the development of other reinforced substrates. Xu et al^[35]The inverted pyramid klarite reinforced substrate prepared with gold recognized 360 nm standard PS and PMMA microspheres at the lowest, which was a diffractive size difficult to reach by ordinary Raman. The main reason was the enhancement of the electric field caused by particles entering the bottom of the pyramid pit structure, rather than the enhancement of the electric field between particles and gold nanoparticles. However, the size of plastic particles in the environment is different, and not all particles can just enter the bottom of the pyramid pit, so the reinforcement effect of different particles is different. At the same time, the manufacturing process and cost of this substrate are complex and expensive. Repeated cleaning and reuse will cause the gold film to fall off and damage the structure, and the cost performance is not high. Park et al^[52]Taking advantage of the low melting point of plastic, heating the SERS substrate made of Ag nanoparticles to melt the plastic particles and enter the "hot spot" region of SERS can overcome the difficulty of Raman signal weakness caused by the difficulty of micro nano plastic particles with different shapes entering the "hot spot". Chang et al^[36]In order to solve the problem that different sizes of nano plastics have different reinforcement effects on the metal nano array substrate, a new type of SiO 2 was designed₂@AG substrate, using the coffee ring effect and the gravity of the nano plastic itself to trap a single nano plastic in three adjacent SIOs₂@In the nanowells formed by Ag nanospheres, higher spatial resolution and better consistency are achieved. SiO 2 was treated with 100 µ mol/l ethanol solution of 4-Aminothiophenol₂@The coefficient of variation CV of Ag substrate was 23.4%, which was better than 39.05% of klarite substrate. The highly consistent signal provides favorable conditions for quantitative analysis of micro plastics. In the range of 0.00005% -0.1%, the Raman response intensity of PS nano plastics is highly linear with the concentration（R²>97). Yang et al^[46]It is proposed to use one-dimensional Ag nanowires self-assembly to form a filter membrane to preconcentrate nano plastics. The high SERS activity of Ag nanostructures can enhance the Raman signal of nano plastics in situ. The detection limit of PS plastic microspheres with a diameter of 50 nm is as low as 0.10 µ g/L.

The colloids of metal nanoparticles used in the analysis of micro and nano plastics are mainly composed of Ag, Au, Cu and other metals that are easily excited by light to produce surface plasma. In the process of using metal nanoparticles colloid, metal nanoparticles are first required to be mixed with micro nano plastics. The effect of signal enhancement is mainly related to the shape and size of metal nanoparticles and the degree of combination between metal particles and micro nano plastics. Lee et al^[51]The Raman signal of PS plastic at 600 nm was significantly enhanced by the use of Au nanourchin like particles at about 50 nm. Kleinman et al^[53]It is found that Au nanoparticles with more vertices will produce more "hot spots" (number of hot spots: spherical<triangular<star), which further improves the Raman response. LV, etc^[34]And Hu et al^[48]Ag nanoparticles were used to detect micro and nano plastics in water samples. They used the polymerization agent NaCl or ki to destroy the double electron layer of Ag nanoparticles, so that Ag particles could gather and wrap on the surface of PS nanoparticles. The detection limit of PS microspheres at 100 nm could reach 6.25 µ g/ml. Although studies have shown that competitive adsorption of ions from polymeric agents can significantly reduce the SERS sensitivity of the target^[54⇓-56]The researchers still recommend using ki as a polymerization agent to increase the aggregation of metal nanoparticles and micro nano plastics, which can not only remove the impurities on the surface of metal nanoparticles^[57]Can also increase the number of "hot spots"^[58]。

In general, the metal nano array substrate has a more efficient and stable "hot spot" structure and good reproducibility, which is conducive to the realization of single particle imaging and quantitative research of micro nano plastics. However, its manufacturing process is complex and requires more expensive processing equipment. In the future, we can find a simpler manufacturing process and cheap preparation materials to achieve the goal of efficient, rapid and convenient identification of micro nano plastics in the environment.

Coherent Raman spectroscopy is a nonlinear optical method, which mainly uses two synchronous pulse lasers (pump light and Stokes light). When the frequency difference between the two beams corresponds to the vibration frequency of the target chemical bond, coherent anti Stokes scattering (cars) and stimulated Raman scattering (SRS) will occur, and a stronger signal will be obtained at the expense of some spectral information^[33]. Rhee et al^[59]Cars method was used to directly and rapidly detect and identify PE, PS, PMMA and pa-12 micro nano plastics in soil samples. Laptenok et al^[60]Using SRS technology, plastic fibers were found in drinking water, surface seawater, coastal sediments, deep-sea sediments and fish gastrointestinal tract and were distinguished from natural fibers. Qian et al^[61]SRS technology was used to directly detect the micro nano plastics in bottled water, and the quantification showed that there were more than 10% of micro nano plastics per liter of bottled water⁵Plastic particles, most of which are nano plastic particles. In addition, as nonlinear optical means, cars and SRS also have the characteristics of real-time fast imaging, fluorescence suppression, small damage to samples and high spatial resolution. Huber et al^[62]The PS, PE and PMMA plastic particles with the diameter of 100~5000 nm were rapidly detected in the flow field by SRS technology. Thanks to the high time resolution of 60.5 µ s, this technology can directly detect the single signal of each particle and calculate the particle diameter through the average peak width and intensity of the signal. Cars technology and SRS technology can also be used for label free detection in biological samples^{[63⇓⇓⇓-67]}For example, Choi et al^[66]Cars was used to track the movement of 2 µ m PS microspheres in human living cells in real time, and the evidence of its internalization into cells was successfully obtained from the dynamic characteristics of PS microspheres. Fueser et al^[68]Cars technology was used to detect the intake of 1 µ m PS microspheres by Caenorhabditis elegans. It was found that the lipid distribution area of nematodes exposed to PS microspheres was 79% higher than that of non exposed nematodes. Xin et al^[69]SRS technology was used to quantitatively monitor the bioaccumulation and metabolic toxicity caused by the pollution of micro/nano plastics in the early development stage of zebrafish. It was found that the gut and liver were the main target organs of micro/nano plastics. It was revealed that micro/nano plastics disturbed the development and lipid metabolism of zebrafish larvae, resulting in growth retardation. It is worth noting that plastics are composed of C, h, O and other elements. In cells, proteins, nucleic acids, glycans and other molecules are also polymers composed of C, h, O and other elements. This similarity makes their coherent Raman signals very similar and difficult to distinguish. One solution is to use isotope labeling method. For example, molecules labeled with isotope deuterium will produce C-D bonds with different vibration frequencies from C-H bonds, so as to distinguish the background signal^[70]。

At present, the concentration level of micro nano plastic in most environmental sample detection reports basically depends on the microscope to determine the total number and morphology of plastic particles in the sample. But some studies have pointed out that this method is prone to false positive and false negative^[71]An effective way is to add a chemical composition detector such as Raman and microscope in series. However, these methods rely on manual preselection, which is a huge waste of manpower and time. Especially, nano plastics that exceed the diffraction limit of visible light are often ignored, resulting in a large difference between the statistical results and the actual concentration^[72]. Park et al^[52]Using a dark field microscope, particles and substrates can be clearly distinguished. A more feasible method is to identify and visualize micro nano plastics through Raman spectral imaging.

Compared with single point analysis, Raman spectral imaging scans the sample surface by changing the laser position, and collects the signals in the preset area in the form of matrix to generate a hyperspectral matrix. Subsequent analysis can convert this matrix into a mapping image, and visualize the scanning area from the perspective of spectrum or through the chemical information channel. The spatial resolution of Raman imaging can usually reach sub micron level, which can analyze the shape, distribution, behavior and other complex details of micro nano plastics in complex environmental samples^[73]. Qian et al^[61]Using SRS technology, standard PS particles can be quickly identified in tens of microseconds. Although the best spatial resolution is 365 nm, 100 nm PS single particles can still be easily identified according to the diffraction limit mode and intensity distribution. Sobhani et al^[74]The micro nano plastics with a diameter as low as 100 nm were successfully mapped, and it was proved that billions of micro nano plastics with a diameter of 200 nm~7 µ m would be produced in this process by collecting the polishing powder in the process of polishing automobile paint. Due to the use of conventional micro Raman technology, scanning 4 µ M²The signal-to-noise ratio is low, and some particles cannot be identified. Ruan et al^[43]The SRS technology was used to test 1248 µ m on the SERS substrate for several edible sea salt samples²The imaging time was significantly reduced to about 2 min. They assume that adults consume 5 g of sea salt per day. In some areas, an adult can consume up to 6 × 10% of sea salt per year⁶Nano plastic. It is worth noting that this value may be far underestimated, because this method does not count nano plastics with diameters below 200 nm. The correlation between particle morphology and chemical composition has an important impact on toxicology. Studies have shown that the toxicity induced by micro nano particles is not only related to the dose, but also related to the physicochemical properties of particles, their interactions with cells and their effects on uptake^[75-76]Therefore, whether in water samples or biological tissues, visual analysis of single micro nano plastic particles is helpful to understand their aggregation behavior, transport mechanism and potential interactions with organisms and ecosystems. Real time data acquisition is another attractive function of Raman imaging. This ability is helpful to dynamically study the behavior of micro/nano plastics and track their characteristics over time^[77]This is particularly important for observing how micro/nano plastics change in response to environmental conditions, chemical processes or biological interactions, and for further understanding their persistence and toxicity.

In order to efficiently convert hyperspectral matrix signals into images, different algorithms (such asTable 3As shown in). Advanced algorithms combine machine learning technology, pattern recognition, image merging and signal deconvolution methods to significantly improve the accuracy and efficiency of plastic recognition^[78-79]For example, su et al^[80]The machine learning algorithm is used to train on large-scale data sets containing various plastics and environmental conditions, so as to distinguish the subtle differences of the spectrum and improve its ability to distinguish plastics and background signals. In addition, the emerging signal processing technology can extract meaningful information from the complex Raman spectrum in the presence of interfering substances^[81]The progress of these algorithms not only improves the accuracy of micro/nano plastics recognition, but also helps to automate the analysis process, thus speeding up the research on the distribution, transformation and potential environmental impact of micro/nano plastics. With the continuous development of technology, the synergy of Raman imaging and spectral analysis algorithm provides important support for the study of the behavior of micro nano plastics in various environmental and biological backgrounds.

In the Raman spectrum analysis of micro nano plastics, the sources of fluorescence signals are very wide, including the inherent fluorescence of plastics, colorants, stabilizers, degradation products and the fluorescence of various impurities in the environment. The presence of fluorescence causes the baseline to rise, and even completely masks the Raman signal. There are many methods to suppress the fluorescence signal, and pre clean the material to remove the interference of fluorescence impurities. Photobleaching and changing the excitation wavelength are alternative methods.

At present, the common method to reduce the fluorescence effect caused by organic matter is to use the cleaning step to remove pollutants^[89-90]However, chemical cleaning will cause damage to micro nano plastics, making it difficult to accurately identify plastics. Usually, the collected environmental samples will be pretreated with acid, alkali, oxidant and enzyme to reduce the interference of surface adsorbed pollutants and facilitate accurate analysis. Acids (such as sulfuric acid, nitric acid or hydrochloric acid) and bases (such as sodium hydroxide and potassium hydroxide) can quickly digest organic matter in the sample, and the digestion efficiency of organic matter can be greater than 90%^[91]However, in addition to the function of digesting organics, strong acids and bases can also lead to the oxidation of the plastic surface^[92-93]During the oxidation process, the plastic will also be discolored and slightly degraded^[94]Because the Raman spectra of degraded plastics and standard samples do not match, the degradation of plastics induced by digestion or oxidation may affect the accuracy of Raman spectra identification of micro plastics. Enzymatic digestion is an alternative method of chemical digestion, which can minimize the degradation of plastics. However, the enzyme digestion efficiency is relatively low, and it is sensitive to the substrate and digestion conditions^[95]Therefore, appropriate digestion methods should be selected and optimized by combining several digestion processes.

Despite the thorough digestion process, some samples will still show fluorescence of impurities, because there are often colorants in plastics that are difficult to remove. Two common methods are switching excitation light with different wavelengths and photobleaching samples^[96-97]For exampleTable 1andTable 2As shown, the most widely used excitation wavelength is 785 nm near infrared laser. This laser wavelength achieves a high balance in signal intensity, fluorescence suppression, cost and overall performance. Compared with expensive laser sources, photobleaching is more widely used, that is, the samples are placed under the laser to degrade the fluorescent substances continuously. However, in addition to increasing the detection time, this method may also lead to photodegradation or pyrolysis of micro/nano plastics, so photobleaching is not always effective^[96-97]A more radical approach is to use nonlinear Raman technology^[22]This technology can provide high signal-to-noise ratio without fluorescence interference, but it needs more expensive equipment.

Ghosal et al^[98]A rapid detection scheme that can correctly identify fluorescent samples is proposed. They use automatic polynomial fitting algorithm to eliminate the wide background, so as to remove the fluorescent background and enhance the polymer spectrum. When analyzing plastic samples covered by biofilm, the spectrum obtained directly is saturated by fluorescence signal, which almost covers the characteristic peak of plastic. After using the automatic algorithm to process the spectrum, the polymer characteristic peak is clearly visible, so the chemical composition of plastics can be identified by Raman database software. Xie et al^[99]The random forest algorithm is used to overcome the interference of fluorescence generated by complex environmental media on plastic recognition, and the recognition accuracy of more than 97% is achieved in the labeled tap water. The verification experiment shows that PS and PVC nano plastics are successfully detected in the rainwater.

Another way to remove fluorescence interference is to use coherent Raman technology. Because the signal only comes from molecular vibration, this technology can effectively avoid the fluorescence interference and obtain Raman spectra with high signal-to-noise ratio. Rhee et al^[46]The cars method was used to identify PE, PS, PMMA and pa-12 micro nano plastics in undigested soil samples, which effectively reduced the pretreatment steps. C-H bond specific cars imaging and spectral analysis help to quickly search for microplastics particles, and can realize rapid chemical recognition even if it is interfered by residual particles and highly fluorescent substances in soil. Zada et al^[100]After density separation of sediment samples from the Rhine estuary, 1 cm²88 kinds of micro nano plastics, including nylon, pet, PS, PP and PE, were successfully identified on the surface of the membrane. Wang et al^[67]Using SRS technology, PE, PP, PVC, pet, PS and PMMA particles were successfully detected in protozoan cells, which effectively reduced the effects of other fluorescent substances in water samples and cells on plastic Raman signals.

The automatic Raman spectrum program uses the spectrum library matching software to compare the sample spectrum with the customized or commercial spectrum library, and automatically identify the specific chemical composition of the micro plastic. The probability of successful matching largely depends on the comprehensiveness of the spectrum library. However, the customized spectrum library is usually based on the spectrum obtained from the original polymer particles, which may be different from the spectrum of micro nano plastics collected from the environment. Micro/nano plastics in the environment are mainly composed of fragments of commercial plastics, including foam, sheet and fiber, and often contain various additives, fillers and colorants. In some cases, these components may mask the spectrum of the polymer, so the use of a more comprehensive commercial spectrum library will undoubtedly improve the accuracy of matching^[71]。

Under the influence of various environmental action sources (such as ultraviolet, heat and biodegradation), plastics will age^[26,71]For example, polyvinyl chloride (PVC) is prone to photodegradation in aqueous media. Under high humidity conditions, additives such as light stabilizers may accelerate leakage. After UV exposure, the spectral fingerprints of PVC changed significantly, showing a double peak of 693 cm representing the C-Cl bond^-1And 637 cm^-1The adjacent peaks at^[71]In the spectrum after strong UV irradiation, there are no such double peaks at all, but 1139 cm due to the vibration of carbon carbon double bond (c=c) will appear^-1And 1540 cm^-1There are two strong peaks. Because the spectrum library only contains the spectra of non degraded PVC, it is impossible to search and match the spectra successfully through the spectrum library. Considering this problem, it is very important to include the spectra of polymers at different aging and degradation stages into the spectrum library, which can increase the opportunity for matching software to correctly identify polymer components. It is worth noting that, compared with FTIR technology, Raman spectroscopy can better identify photooxidized micro nano plastics without fluorescence interference. For the most common micro nano plastics (such as PE and PP), photooxidation leads to the formation of oxygen-containing groups, mainly c=o and - Oh, which show strong strength in the infrared spectrum^{[26,101⇓-103]}In sharp contrast to its weak Raman signal. For example, Cai et al^[101]It is found that there is a significant difference between the infrared spectra of the original and degraded PE, PP and PS, while the corresponding Raman spectral peak intensities are only slightly different, which is conducive to the software to automatically match the standard Raman spectra. Therefore, Raman spectroscopy has a significant advantage in simply identifying the polymer type of aging plastics.

In addition, some scholars suggest that it is also valuable to include the spectra of some non plastic materials that are often confused with micro nano plastics in the spectrum library. These non plastic materials include cellulose, keratin, inorganic particles, and the most important synthetic fibers (i.e., viscose fibers), which are common in micro nano plastic samples^[71,104]The main limitation of the popularization of this method is that each laboratory needs to establish a comprehensive spectrum library, which is a time-consuming work. The field of micro nano plastics is highly interdisciplinary, and scholars in this field have a variety of academic backgrounds and experiences. Therefore, it will be very beneficial to establish an open-source and well-designed spectral database, which can avoid the repeated establishment of spectral databases, minimize the energy and material resource consumption of each researcher, and realize the complexity and comprehensiveness that a single commercial spectral database does not have. Such a spectral library can also contain spectra from real environmental samples, so that its components can be confirmed by other identification technologies in case of doubt, thus greatly improving the accuracy of matching. In addition, the free spectrum library will encourage research groups with limited funds to carry out more research on the identification of microplastics, thereby increasing researchers' awareness of global microplastics pollution.

After having a complete Raman spectrum library, the difficulty is to quickly compare the Raman spectrum of the sample with the spectrum library. Machine learning has attracted extensive attention because of its ability to automatically analyze complex spectral data. The combination of machine learning and Raman spectroscopy can improve the recognition efficiency and accuracy of micro nano plastics^[105⇓-107]Before machine learning, it is necessary to properly process the spectrum. The first is the smoothing of the spectral curve. The common smoothing method is savitzky Golay smoothing. Its principle is to dynamically fit polynomials to continuous data point windows (local least squares polynomial approximation) to track the shape of the spectrum, so as to reduce the impact of random noise signals. However, in some cases, this smoothing method may have a negative impact on the spectral features. When applied to the highly smooth high noise spectrum, it may seriously affect the sharp local features and lead to errors in subsequent machine learning. Barton et al^[108]The maximum likelihood estimation is used to enhance the savitzky Golay smoothing, which effectively prevents the obvious deviation from the real Raman signal, and retains the strong smoothing characteristics of the savitzky Golay method, so that the signal-to-noise ratio is enhanced by at least 50%, providing a good Raman spectrum for subsequent machine learning.

Machine learning can dynamically and quickly process complex Raman spectral data. It uses the algorithm to automatically extract features from the spectrum and train the classification model, so as to explore the potential relationship between researchers' multiple concerns, and finally realize the accurate classification of micro nano plastic types^[109]. Xie et al^[99]Five common micro nano plastics (PE, PTFE, PS, PMMA and PVC) were identified by random forest model. The average accuracy and sensitivity were 98.8% and 98.5%, respectively. Nanoscale PS and PVC were identified in the rainwater detection, which proved the potential of the model for analyzing actual environmental samples. Lim et al^[110]The convolutional neural network is used to realize the rapid recognition of 1-10 µ m single micro plastic. The exposure time is only 0.4 s, and the confidence level is about 85.47%, which greatly speeds up the qualitative analysis ability of micro plastic in different environmental scenes. Feng et al^[106]It is considered that the single machine learning model is not efficient for the recognition of micro nano Plastics Exposed to the environment, so a multi model algorithm with three-layer discrimination is proposed. For the unknown spectrum, savitzky Golay smoothing method was used, then principal component analysis linear discriminant analysis (pca-lda) was used to identify PP, PS, pet and PVC plastics, and then principal component analysis-k nearest neighbor analysis (pca-knn) was used to identify chlorinated polyethylene plastics (CPE), and finally 980~1385 cm^-1Multi layer perceptron model (MLP) is used to classify HDPE and LDPE in the Raman spectral band of, and the accuracy is more than 97%. The real micro plastic samples in daily life are successfully identified.

It is an interesting method to quickly identify micro plastics with complex components by Raman barcode method^[111]This method converts the Raman spectra of various plastics into bar codes, where each line represents a Raman spectral peak, and then identifies the bar codes of samples by binary comparison with the reference bar codes, so as to speed up the matching and eliminate the standardization of spectral intensity. Although this bar code method has not been used for the identification of micro/nano plastics in actual environmental samples, its reference bar code can include the bar codes of polymers, colorants and common plastic additives. By ignoring the spectral intensity, the recognition rate of the spectrum of micro/nano plastics covered by the signals of colorants or additives can be improved.

In the quantitative determination of micro/nano plastics, the mass concentration is an important parameter to formulate its pollution control countermeasures, and the difference of particle size and density of micro plastics is an important factor affecting the quantitative accuracy. Chaisrikhwun et al^[112]An interesting experiment is proposed. The micro nano plastic (PS particles) in water are dissolved in organic solvent after drying to avoid the influence of particle size on Raman signal response. The organic solution containing plastic is dripped onto the gold film substrate for drying, and the plastic molecules are concentrated at the edge of the droplet under the effect of "coffee ring". There are three points worth paying attention to. First, the method effectively avoids the dependence of signal and particle size; Second, according to the principle of "similarity dissolves", the solvent that dissolves plastic molecules is generally a non-polar solvent, so polar interfering substances will be separated in the process of dissolution; Third, the "coffee ring" effect not only concentrates plastic molecules, but also separates other interferents from plastic molecules. Based on the above three points, the experimental method can eliminate the interference of salt, sugar, detergent and protein, with the detection limit as low as 0.10 µ g/l and the quantitative range between 10 and 40 µ g/l.

The "coffee ring" effect refers to that when a droplet is on a hydrophilic surface, the boundary of the droplet is fixed due to capillarity, and the liquid evaporating near the boundary must be supplemented by the internal liquid, thus forming a liquid flow from the interior to the boundary, which can bring the vast majority of solutes to the boundary^[113-114]Although the "coffee ring" effect helps to concentrate the micro/nano plastics to the gas-liquid solid interface, in most cases, the micro/nano plastics will not be completely distributed in the ring, which will lead to large differences in Raman signal peaks. The hydrophobicity detection substrate can effectively concentrate the micro nano plastic in the small contact area between the droplet and the hydrophobic substrate. Li et al^[115]A hydrophobic substrate with a contact angle of 143.71 ° was constructed by immersing the clean silicon wafer in n-hexane solution of methyltrichlorosilane, which effectively concentrated the target material evenly and tightly in a small area. At the same time, Ag nanoparticles were mixed with plastic particles to further enhance the Raman signal. The detection limit of this method is as low as 0.5 mg/L (500 nm PS) and 1 mg/L (100 nm PS), and the relative standard deviation (RSD) can be as low as 1.94%. The reproducibility is the best at present, which provides a potential direction for the quantitative detection of micro plastics. However, the strength of SERS depends on the distribution of "hot spots", which may be affected by the preparation method of Ag nanoparticles, the combination mode of Ag particles and micro nano plastics, and the experimental operation, so it is very difficult to test the operating ability of the experimenters.

The membrane filtration method based on physical closure has the advantages of simple operation, high enrichment efficiency, maintaining the original shape of particles and retaining particles of different sizes, and can effectively separate and enrich micro and nano plastics^[116-117]Because the background signal of plastic filter membrane will mask the signal of micro nano plastic sample, glass fiber membrane, inorganic membrane (alumina) and mixed cellulose membrane are often used in the membrane filtration of micro nano plastic. In addition to using optical instruments to directly analyze the micro and nano plastics on the membrane, when membrane filtration is used as a preconcentration method before other analysis technologies, sample transfer is also required. However, small plastic particles tend to be adsorbed on the membrane, resulting in a low transfer rate (only 53.1% of micro plastics)^[91]The direct preparation of nano reinforced structure on the filter membrane is an effective method to improve the detection sensitivity of Raman spectrum by combining the filtration and detection of microplastics. Qin et al^[45]The spinel Au nanocrystals were synthesized and evenly deposited on the glass fiber filter membrane for in-situ enrichment and high sensitivity SERS detection of micro/nano plastics. The SERS membrane was successfully used to detect PS microspheres with a detection limit of 0.1 mg/L, ranging from 20 nm to 10 µ M. The detection limit of 100 nm PS microspheres dispersed in tap water and rainwater was 0.1 mg/L. Yang et al^[35]It is proposed to use one-dimensional Ag nanowires self-assembly to form a filter membrane to preconcentrate nano plastics. The retention rate of PS plastic microspheres with a diameter of 50 nm is as high as 86.7%, and the high SERS activity of Ag nanostructures can enhance the Raman signal of nano plastics in situ. The detection limit is as low as 0.10 µ g/L. In the analysis of water samples from the actual seafood market, considering the low concentration of nano plastics in the water sample, the concentration of nano plastics on the membrane was increased by increasing the filtering capacity of the actual water sample within the maximum retention capacity of the membrane. The method detected 500 nm PS plastics at the level of µ g/L.

Whether it is concentrated through the "coffee ring" effect or through membrane filtration, we recommend that we first choose the particle size of micro nano plastics that need to be focused on. The ability of particles with larger particle size to form "coffee ring" is often weak^[118]The mismatch between particle size and membrane pore size will seriously affect the concentration efficiency of the sample^[119]As a result, the quantitative accuracy is low.

In addition to the detection of the mass concentration of micro/nano plastics, the analysis of the particle size distribution and number concentration of micro/nano plastics in environmental samples is also very important because there may be many different ways for plastic particles with different particle sizes to enter organisms and interact with biological target organs. Zhang et al^[38]The gold triangle cavity array substrate was constructed by electron beam evaporation technology. When particles fall into or close to the edge or apex of the triangle cavity, significant electric field enhancement can be obtained, and the concentration of 1.5 × 10¹¹PS standard microspheres with diameter of 50 nm were prepared. Using the Golden Triangle cavity array substrate and nano particle tracking analyzer, the concentration of 10⁸Particles/ml PET plastic particles with an average diameter of 88.2 nm. However, this is the qualitative analysis of particles using SERS technology and the quantification of the number concentration of micro and nano plastics using nano particle tracking analyzer, which is not the real quantitative research using Raman spectrometer. Huber et al^[62]Standard PS microspheres with diameters of 300 nm and 600 nm were selected as representatives. The relationship between SRS signal value and number concentration was explored in the flow field using SRS technology. It was found that this technology can only be applied to the quantitative study of plastic particles with high number concentration, and the signal intensity depends on the particle diameter. This is because most particles cannot appear near the very small SRS focus in the experiment, so no signal is generated. It is necessary to further optimize the flow unit to improve the probability of particles appearing at the focus. Another scheme is to use Raman spectral imaging within the spatial resolution of the instrument to visualize single particle micro nano plastics, so that researchers can easily count the number of particles in the scanning range. Qian et al^[61]The SRS technology is used to scan and filter the micro nano plastics on the alumina membrane with a diameter of 13 mm. Five or more 0.2 mm × 0.2 mm imaging areas are selected on each filter membrane to calculate the total number of micro nano plastics on the whole filter membrane. Finally, the total number of micro nano plastics per liter of bottled water is more than 10⁵Particles, and most of them are plastic. Qian et al. Carried out quantitative research on transforming Raman spectral imaging from qualitative identification to numerical concentration, which is helpful to understand the impact of micro nano plastics on organisms from the perspective of numerical concentration. Raman spectral imaging is also helpful to analyze the number concentration of micro nano plastics with different shapes and sizes, especially nano plastics which are easy to be ignored in the mass concentration test. Although these nanoplastics contribute little to the mass concentration, they have the ability to cross the biological barrier and play a leading role in toxicity assessment^[120-121]Therefore, it is still of profound significance to evaluate the number concentration of nano plastics.