Patch-Based Convolutional Encoder: A Deep Learning Algorithm for Spectral Classification Balancing the Local and Global Information.

Molecular vibrational spectroscopies, including infrared absorption and Raman scattering, provide molecular fingerprint information and are powerful tools for qualitative and quantitative analysis. They benefit from the recent development of deep-learning-based algorithms to improve the spectral, spatial, and temporal resolutions. Although a variety of deep-learning-based algorithms, including those to simultaneously extract the global and local spectral features, have been developed for spectral classification, the classification accuracy is still far from satisfactory when the difference becomes very subtle. Here, we developed a lightweight algorithm named patch-based convolutional encoder (PACE), which effectively improved the accuracy of spectral classification by extracting spectral features while balancing local and global information. The local information was captured well by segmenting the spectrum into patches with an appropriate patch size. The global information was extracted by constructing the correlation between different patches with depthwise separable convolutions. In the five open-source spectral data sets, PACE achieved a state-of-the-art performance. The more difficult the classification, the better the performance of PACE, compared with that of residual neural network (ResNet), vision transformer (ViT), and other commonly used deep learning algorithms. PACE helped improve the accuracy to 92.1% in Raman identification of pathogen-derived extracellular vesicles at different physiological states, which is much better than those of ResNet (85.1%) and ViT (86.0%). In general, the precise recognition and extraction of subtle differences offered by PACE are expected to facilitate vibrational spectroscopy to be a powerful tool toward revealing the relevant chemical reaction mechanisms in surface science or realizing the early diagnosis in life science.

Revealing the Denoising Principle of Zero-Shot N2N-Based Algorithm from 1D Spectrum to 2D Image.

Denoising is a necessary step in image analysis to extract weak signals, especially those hardly identified by the naked eye. Unlike the data-driven deep-learning denoising algorithms relying on a clean image as the reference, Noise2Noise (N2N) was able to denoise the noise image, providing sufficiently noise images with the same subject but randomly distributed noise. Further, by introducing data augmentation to create a big data set and regularization to prevent model overfitting, zero-shot N2N-based denoising was proposed in which only a single noisy image was needed. Although various N2N-based denoising algorithms have been developed with high performance, their complicated black box operation prevented the lightweight. Therefore, to reveal the working function of the zero-shot N2N-based algorithm, we proposed a lightweight Peak2Peak algorithm (P2P) and qualitatively and quantitatively analyzed its denoising behavior on the 1D spectrum and 2D image. We found that the high-performance denoising originates from the trade-off balance between the loss function and regularization in the denoising module, where regularization is the switch of denoising. Meanwhile, the signal extraction is mainly from the self-supervised characteristic learning in the data augmentation module. Further, the lightweight P2P improved the denoising speed by at least ten times but with little performance loss, compared with that of the current N2N-based algorithms. In general, the visualization of P2P provides a reference for revealing the working function of zero-shot N2N-based algorithms, which would pave the way for the application of these algorithms toward real-time (in situ, in vivo, and operando) research improving both temporal and spatial resolutions. The P2P is open-source at https://github.com/3331822w/Peak2Peakand will be accessible online access at https://ramancloud.xmu.edu.cn/tutorial.

Rapid Sample Pretreatment Facilitating SERS Detection of Trace Weak Organic Acids/Bases in Complex Matrices.

Fast and efficient sample pretreatment is the prerequisite for realizing surface-enhanced Raman spectroscopy (SERS) detection of trace targets in complex matrices, which is still a big issue for the practical application of SERS. Recently, we have proposed a highly performed liquid-liquid extraction (LLE)-back extraction (BE) for weak acids/bases extraction in drinking water and beverage samples. However, the performance efficiency decreased drastically on facing matrices like food and biological blood. Based on the total interaction energies among target, interferent, and extractant molecules, solid-phase extraction (SPE) with a higher selectivity was introduced in advance of LLE-BE, which enabled the sensitive (µg L-1 level) and rapid (within 10 min) SERS detection of both koumine (a weak base) and celastrol (a weak acid) in different food and biological samples. Further, the high SERS sensitivity was determined unmanned by Vis-CAD (a machine learning algorithm), instead of the highly demanded expert recognition. The generality of SPE-LLE-BE for various weak acids/bases (2 < pKa < 12), accompanied by the high efficiency, easy operation, and low cost, offers SERS as a powerful on-site and efficient inspection tool in food safety and forensics.

Signal2signal: Pushing the Spatiotemporal Resolution to the Limit by Single Chemical Hyperspectral Imaging.

There is growing interest in developing a high-performance self-supervised denoising algorithm for real-time chemical hyperspectral imaging. With a good understanding of the working function of the zero-shot Noise2Noise-based denoising algorithm, we developed a self-supervised Signal2Signal (S2S) algorithm for real-time denoising with a single chemical hyperspectral image. Owing to the accurate distinction and capture of the weak signal from the random fluctuating noise, S2S displays excellent denoising performance, even for the hyperspectral image with a spectral signal-to-noise ratio (SNR) as low as 1.12. Under this condition, both the image clarity and the spatial resolution could be significantly improved and present an almost identical pattern with a spectral SNR of 7.87. The feasibility of real-time denoising during imaging was well demonstrated, and S2S was applied to monitor the photoinduced exfoliation of transition metal dichalcogenide, which is hard to accomplish by confocal Raman spectroscopy. In general, the real-time denoising capability of S2S offers an easy way toward in situ/in vivo/operando research with much improved spatial and temporal resolution. S2S is open-source at https://github.com/3331822w/Signal2signal and will be accessible online at https://ramancloud.xmu.edu.cn/tutorial.

Deep Learning-Assisted Spectrum-Structure Correlation: State-of-the-Art and Perspectives.

Spectrum-structure correlation is playing an increasingly crucial role in spectral analysis and has undergone significant development in recent decades. With the advancement of spectrometers, the high-throughput detection triggers the explosive growth of spectral data, and the research extension from small molecules to biomolecules accompanies massive chemical space. Facing the evolving landscape of spectrum-structure correlation, conventional chemometrics becomes ill-equipped, and deep learning assisted chemometrics rapidly emerges as a flourishing approach with superior ability of extracting latent features and making precise predictions. In this review, the molecular and spectral representations and fundamental knowledge of deep learning are first introduced. We then summarize the development of how deep learning assist to establish the correlation between spectrum and molecular structure in the recent 5 years, by empowering spectral prediction (i.e., forward structure-spectrum correlation) and further enabling library matching and de novo molecular generation (i.e., inverse spectrum-structure correlation). Finally, we highlight the most important open issues persisted with corresponding potential solutions. With the fast development of deep learning, it is expected to see ultimate solution of establishing spectrum-structure correlation soon, which would trigger substantial development of various disciplines.

Seeing Is Not Necessarily Believing: Is the Surface-Enhanced Raman Spectroscopy Signal Really from the Target?

Reagent purity is crucial to experimental research, considering that the ignorance of ultratrace impurities may induce wrong conclusions in either revealing the reaction nature or qualifying the target. Specifically, in the field of surface science, the strong interaction between the impurity and the surface will bring a non-negligible negative effect. Surface-enhanced Raman spectroscopy (SERS) is a highly surface-sensitive technique, providing fingerprint identification and near-single molecule sensitivity. In the SERS analysis of trace chloromethyl diethyl phosphate (DECMP), we figured out that the SERS performance of DECMP is significantly distorted by the trace impurities from DECMP. With the aid of gas chromatography-based techniques, one strongly interfering impurity (2,2-dichloro-N,N-dimethylacetamide), the byproduct during the synthesis of DECMP, was confirmed. Furthermore, the nonignorable interference of impurities on the SERS measurement of NaBr, NaI, or sulfadiazine was also observed. The generality ignited us to refresh and consolidate the guideline for the reliable SERS qualitative analysis, by which the potential misleading brought by ultratrace impurities, especially those strongly adsorbed on Au or Ag surfaces, could be well excluded.

1D Gradient-Weighted Class Activation Mapping, Visualizing Decision Process of Convolutional Neural Network-Based Models in Spectroscopy Analysis.

Being characterized by the self-adaption and high accuracy, the deep learning-based models have been widely applied in the 1D spectroscopy-related field. However, the "black-box" operation and "end-to-end" working style of the deep learning normally bring the low interpretability, where a reliable visualization is highly demanded. Although there are some well-developed visualization methods, such as Class Activation Mapping (CAM) and Gradient-weighted Class Activation Mapping (Grad-CAM), for the 2D image data, they cannot correctly reflect the weights of the model when being applied to the 1D spectral data, where the importance of position information is not considered. Here, aiming at the visualization of Convolutional Neural Network-based models toward the qualitative and quantitative analysis of 1D spectroscopy, we developed a novel visualization algorithm (1D Grad-CAM) to more accurately display the decision-making process of the CNN-based models. Different from the classical Grad-CAM, with the removal of the gradient averaging (GAP) and the ReLU operations, a significantly improved correlation between the gradient and the spectral location and a more comprehensive spectral feature capture were realized for 1D Grad-CAM. Furthermore, the introduction of difference (purity or linearity) and feature contribute in the CNN output in 1D Grad-CAM achieved a reliable evaluation of the qualitative accuracy and quantitative precision of CNN-based models. Facing the qualitative and adulteration quantitative analysis of vegetable oils by the combination of Raman spectroscopy and ResNet, the visualization by 1D Grad-CAM well reflected the origin of the high accuracy and precision brought by ResNet. In general, 1D Grad-CAM provides a clear vision about the judgment criterion of CNN and paves the way for CNN to a broad application in the field of 1D spectroscopy.

Visualization of a Machine Learning Framework toward Highly Sensitive Qualitative Analysis by SERS.

Surface-enhanced Raman spectroscopy (SERS), providing near-single-molecule-level fingerprint information, is a powerful tool for the trace analysis of a target in a complicated matrix and is especially facilitated by the development of modern machine learning algorithms. However, both the high demand of mass data and the low interpretability of the mysterious black-box operation significantly limit the well-trained model to real systems in practical applications. Aiming at these two issues, we constructed a novel machine learning algorithm-based framework (Vis-CAD), integrating visual random forest, characteristic amplifier, and data augmentation. The introduction of data augmentation significantly reduced the requirement of mass data, and the visualization of the random forest clearly presented the captured features, by which one was able to determine the reliability of the algorithm. Taking the trace analysis of individual polycyclic aromatic hydrocarbons in a mixture as an example, a trustworthy accuracy no less than 99% was realized under the optimized condition. The visualization of the algorithm framework distinctly demonstrated that the captured feature was well correlated to the characteristic Raman peaks of each individual. Furthermore, the sensitivity toward the trace individual could be improved by least 1 order of magnitude as compared to that with the naked eye. The proposed algorithm distinguished by the lesser demand of mass data and the visualization of the operation process offers a new way for the indestructible application of machine learning algorithms, which would bring push-to-the-limit sensitivity toward the qualitative and quantitative analysis of trace targets, not only in the field of SERS, but also in the much wider spectroscopy world. It is implemented in the Python programming language and is open-source at https://github.com/3331822w/Vis-CAD.

Improving SERS Sensitivity toward Trace Sulfonamides: The Key Role of Trade-Off Interfacial Interactions among the Target Molecules, Anions, and Cations on the SERS Active Surface.

In recent years, ensuring the rational use and effective control of antibiotics has been a major focus in the eco-environment, which requires an effective monitoring method. However, on-site rapid detection of antibiotics in water environments remains a challenging issue. In this study, surface-enhanced Raman spectroscopy (SERS) was used to systematically achieve selective, rapid, and highly sensitive detection of sulfonamides, based on their fingerprint characteristics. The results show that the trade-off between the competitive and coadsorption behaviors of target molecules and agglomerates (inorganic salts) on the surface of the SERS substrate determines whether the molecules can be detected with high sensitivity. Based on this, the qualitative differentiation and quantitative detection of three structurally similar antibiotics, sulfadiazine, sulfamerazine, and sulfamethazine, were achieved, with the lowest detectable concentration being 1 µg/L for sulfadiazine and 50 µg/L for sulfamerazine and sulfamethazine.

Developing a Peak Extraction and Retention (PEER) Algorithm for Improving the Temporal Resolution of Raman Spectroscopy.

In spectroscopic analysis, push-to-the-limit sensitivity is one of the important topics, particularly when facing the qualitative and quantitative analyses of the trace target. Normally, the effective recognition and extraction of weak signals are the first key steps, for which there has been considerable effort in developing various denoising algorithms for decades. Nevertheless, the lower the signal-to-noise ratio (SNR), the greater the deviation of the peak height and shape during the denoising process. Therefore, we propose a denoising algorithm along with peak extraction and retention (PEER). First, both the first and second derivatives of the Raman spectrum are used to determine Raman peaks with a high SNR whose peak information is kept away from the denoising process. Second, an optimized window smoothing algorithm is applied to the left part of the Raman spectrum, which is combined with the untreated Raman peaks to obtain the denoised Raman spectrum. The PEER algorithm is demonstrated with much better signal extraction and retention and successfully improves the temporal resolution of Raman imaging of a living cell by at least 1 order of magnitude higher than those by traditional algorithms.

First Report of Phomopsis longicolla Causing Stem Canker of Eggplant in Fujian Province, China.

Eggplant (Solanum melongena L.) is one of the most popular vegetable in China. In July 2019, a serious stem canker disease of eggplant cv. Hangqieyiha has been found in commercial fields in Pingnan County, Fujian Province. The disease incidence ranged from 38% to 72%. The symptoms were found on stems but not on fruits. At first the lesions are small, more or less circular, later becoming elongated, blackish-brown lesions, eventually containing pycnidia. When stem girdling occurs, the shoot above the infected area wilts and dries up. The teleomorph of the fungus has not been encountered in sympotomatic stem. Single-conidial isolate has been obtained by using routine fungal-isolation methods and single-spore purification technique. The fungus was cultivated on potato dextrose agar (PDA), incubated under 12h/12h cycles of light and darkness until sporulation to determine. The fungus initially produced white fluffy aerial hyphae, forming relatively dense concentric pattern colony, which subsequently exhibited yellow-green pigmentation. Pycnidias had globose locules and prominent beaks, which immersed in medium, black, solitary, discoid or irregular. Conidiophores were colorless, separated, branched, 10.0 to 20.0 × 1.0 to 2.5 µm. Alpha-conidia were single-celled, ellipsoidal to fusiform, guttulate, 5.4 to 8.7 × 1.5 to 3.2 µm. Beta-conidia were found occasionally in older stock cultures, hyaline, filiform, hamate, and 17.0 to 26.9 × 0.86 to 1.23 µm. Based on these morphological characters, the fungus was identified as Phomopsis longicolla (Hobbs et al., 1985). The rDNA－ITS of the isolate FAFU01 was amplified with primers ITS1/ ITS4 (TCCGTAGGTGAACCTGCGG/ TCCTCCGCTTATTGATATGC) (White et al., 1990)，and A 578 bp sequence obtained (GenBank Accession No. MW380387 ) was 96% to 98.3% identical to the known sequence of P. longicolla or Diaporthe longicolla in GenBank. For further confirmation, P. longicolla specific primers Phom.I /Phom.II (GAGCTCGCCACTAGATTTCAGGG/GGCGGCCAACCAAACTCTTGT) (Zhang et al., 1997) were used and a 337-bp amplification product was obtained which was previously reported only for P. longicolla, whereas no product was amplified from control. Based on these morphological and molecular characters, the fungus was identified as P. longicolla. In greenhouse tests, each of 35-day-old plants of eggplant cv. Hangqieyihao was maintained in 30-cm-diameter pot. Healthy stem on the plants was wounded by pinpricking. Both wounded and non-wounded stems were inoculated respectively with mycelial plugs (4 mm in diameter) from a 7-day-old PDA culture or PDA medium plugs as controls, with six replicates. The plants were covered with plastic bags to maintain high relative humidity for two days. Four days after inoculation, the plugs were washed from the stems. Thirty-five days after inoculation, canker lesions and small, black pycnidias, which were similar to those in the field, were observed on the surface of non-wounded and wounded healthy stems inoculated with pathogen, whereas all the control stems remained healthy. The fungi was re-isolated from the infected stems of plants and was further confirmed with the species-specific primers. These results confirmed the fungus's pathogenicity. This is the first report of P. longicolla causing stem canker in eggplant in Fujian Province, China.

Single-Molecule Level Rare Events Revealed by Dynamic Surface-Enhanced Raman Spectroscopy.

Surface-enhanced Raman spectroscopy (SERS) is a powerful tool to monitor various interfacial behaviors providing molecular level information with high spatial and temporal resolutions. However, it is a challenge to obtain SERS spectra with high quality for analytes having a weak binding affinity with plasmonic nanostructures due to the short dwell time of the analyte on the surface. Here, we employed dynamic SERS, an acquisition method consisting of the rapid acquisition of a series of consecutive SERS spectra, to study the adsorption/desorption behavior of R6G on Ag surfaces. We demonstrated that the signal-noise ratio of SERS spectra of mobile molecules can be improved by dynamic SERS even when the acquisition time cannot catch up with the diffusion time of the molecule. More interestingly, we captured the neutral R6G0 state (spectroscopically different from the dominated positive R6G+ state) of R6G at the single-molecule level, which is a rare molecule event hardly detectable by traditional SERS. Dynamic SERS provides near real-time molecular vibrational information with an improved signal-noise ratio, which opens a new avenue to capture metastable or rare molecule events for the comprehensive understanding of interfacial processes related to catalysis and life science.

In Situ Raman Study of CO Electrooxidation on Pt(hkl) Single-Crystal Surfaces in Acidic Solution.

The adsorption and electrooxidation of CO molecules at well-defined Pt(hkl) single-crystal electrode surfaces is a key step towards addressing catalyst poisoning mechanisms in fuel cells. Herein, we employed in situ electrochemical shell-isolated nanoparticle-enhanced Raman spectroscopy (SHINERS) coupled with theoretical calculation to investigate CO electrooxidation on Pt(hkl) surfaces in acidic solution. We obtained the Raman signal of top- and bridge-site adsorbed CO* molecules on Pt(111) and Pt(100). In contrast, on Pt(110) surfaces only top-site adsorbed CO* was detected during the entire electrooxidation process. Direct spectroscopic evidence for OH* and COOH* species forming on Pt(100) and Pt(111) surfaces was afforded and confirmed subsequently via isotope substitution experiments and DFT calculations. In summary, the formation and adsorption of OH* and COOH* species plays a vital role in expediting the electrooxidation process, which relates with the pre-oxidation peak of CO electrooxidation. This work deepens knowledge of the CO electrooxidation process and provides new perspectives for the design of anti-poisoning and highly effective catalysts.

Experimental and Theoretical Study of Surface-Enhanced Raman Spectra of Sulfadiazine Adsorbed on Nanoscale Gold Colloids.

Sulfadiazine, as a class of antibiotics, has been widely used in the world for decades; however, its surface-enhanced Raman spectra (SERS) on gold colloids are obviously different from ordinary Raman spectra in the solid powder and liquid solution. To explore the reasons for such significant differences, we used density functional theory calculations and normal-mode analysis to investigate the effects of the configuration, conformation, protonation, hydrogen-bonding interaction, and adsorption configurations of sulfadiazine on gold clusters to check these different effects on the vibrational assignments. Our calculated results can be summarized as two points. First, the Raman spectra strongly depend on the configuration, conformation, protonation, and hydrogen bonding of sulfadiazine. Second, the wagging vibration displays a significant vibrational frequency shift and a very strong SERS peak responsible for the observed SERS signal when sulfadiazine is adsorbed on gold clusters through the terminal amino group. This is different from another adsorption configuration through two oxygen atoms of the -SO2NH- group on gold clusters. Finally, we further investigate the potential energy surfaces along the wagging vibration and the binding interaction of -NH2 adsorbed on different sites of gold surfaces.

Organic-inorganic interactions of single crystalline organolead halide perovskites studied by Raman spectroscopy.

Organolead halide perovskites exhibit superior photoelectric properties, which have given rise to the perovskite-based solar cells whose power conversion efficiency has rapidly reached above 20% in the past few years. However, perovskite-based solar cells have also encountered problems such as current-voltage hysteresis and degradation under practical working conditions. Yet investigations into the intrinsic chemical nature of the perovskite material and its role on the performance of the solar cells are relatively rare. In this work, Raman spectroscopy is employed together with CASTEP calculations to investigate the organic-inorganic interactions in CH3NH3PbI3 and CH3NH3PbBr3-xClx perovskite single crystals with comparison to those having ammonic acid as the cations. For Raman measurements of CH3NH3PbI3, a low energy line of 1030 nm is used to avoid excitation of strong photoluminescence of CH3NH3PbI3. Raman spectra covering a wide range of wavenumbers are obtained, and the restricted rotation modes of CH3-NH3(+) embedded in CH3NH3PbBr3 (325 cm(-1)) are overwhelmingly stronger over the other vibrational bands of the cations. However, the band intensity diminishes dramatically in CH3NH3PbBr3-xClx and most of the bands shift towards high frequency, indicating the interaction with the halides. The details of such an interaction are further revealed by inspecting the band shift of the restricted rotation mode as well as the C-N, NH3(+) and CH3 stretching of the CH3NH3(+) as a function of Cl composition and length of the cationic ammonic acids. The results show that the CH3NH3(+) interacts with the PbX3(-) octahedral framework via the NH3(+) end through N(+)-HX hydrogen bonding whose strength can be tuned by the composition of halides but is insensitive to the size of the organic cations. Moreover, an increase of the Cl content strengthens the hydrogen bonding and thus blueshifts the C-N stretching bands. This is due to the fact that Cl is more electronegative than Br and an increase of the Cl content decreases the lattice constant of the perovskite. The findings of the present work are valuable in understanding the role of cations and halides in the performance of MAPbX3-based perovskite solar cells.

[Instrumentation and Application of a Portable Pretreating System for a Speedy Plasmon-Enhanced Raman Spectroscopy Detection Towards Food Safety].

The present paper developed a portable and fast sample pretreatment apparatus. It has many virtues like being portable, low power consumption and convenient operation, short extraction time and sound repeatability. Therefore, it can meet the requirements of on-site rapid detection pretreatment. The apparatus consists of four functional modules: ultrasonic extraction unit, heating unit, exhaust gas evaporation and absorption unit and control system. In addition, LED control panel and alarm device were designed. The whole treatment process needs three steps: ultrasonic extraction, liquid-liquid extraction and solvent evaporation by heating and pumping. For test of this apparatus performance, three real samples (pepper powder, pepper oil, bean sauce) containing banned additive Rhodamine B were taken as experiment objects. Compared with conventional laboratory pretreatment method, the PERS spectra achieved by two methods were little changed, but the experiment time was halved. In addition, the test results showed relative standard deviation less than ±5%.

[SERS spectrum recognition for Malachite Green real-time detection in aquaculture used seawater].

Surface enhanced Raman spectroscopy (SERS) is a useful chemical analysis technique for its high sensitivity, which was used for Malachite Green qualitative analysis in real cases in the present article. Automatic recognition algorithms were put forward, which is a combination of three modules, including a robust Fourier transform for background rejection, a principal component analysis based character extraction method and artificial neural networks for classifying. Low-frequency background was rejected by iterative Fourier transform in order to eliminate the effect of variable background. The best principal component combination was obtained according to the Euclidean distances between-class and within-class in the sample space. And a three-layer back-propagating neural network was constructed for classifying. As it was shown, it would both minimize the network and reduce the classifying mistakes from variable baseline and Raman characters of other substances in seawater with best principal component combination. Malachite Green real-time detection in aquaculture used seawater was realized with a lower density limit of 0. 1 microg L-1. Moreover, the method proposed in this article could be extended for other sol analysis based on SERS technique.

Resolving nanostructure and chemistry of solid-electrolyte interphase on lithium anodes by depth-sensitive plasmon-enhanced Raman spectroscopy.

The solid-electrolyte interphase (SEI) plays crucial roles for the reversible operation of lithium metal batteries. However, fundamental understanding of the mechanisms of SEI formation and evolution is still limited. Herein, we develop a depth-sensitive plasmon-enhanced Raman spectroscopy (DS-PERS) method to enable in-situ and nondestructive characterization of the nanostructure and chemistry of SEI, based on synergistic enhancements of localized surface plasmons from nanostructured Cu, shell-isolated Au nanoparticles and Li deposits at different depths. We monitor the sequential formation of SEI in both ether-based and carbonate-based dual-salt electrolytes on a Cu current collector and then on freshly deposited Li, with dramatic chemical reconstruction. The molecular-level insights from the DS-PERS study unravel the profound influences of Li in modifying SEI formation and in turn the roles of SEI in regulating the Li-ion desolvation and the subsequent Li deposition at SEI-coupled interfaces. Last, we develop a cycling protocol that promotes a favorable direct SEI formation route, which significantly enhances the performance of anode-free Li metal batteries.

Surface-enhanced Raman spectroscopic study of p-aminothiophenol.

p-aminothiophenol (PATP) is an important molecule for surface-enhanced Raman spectroscopy (SERS). It can strongly interact with metallic SERS substrates and produce very strong SERS signals. It is a molecule that has often been used for mechanistic studies of the SERS mechanism as the photon-driven charge transfer (CT) mechanism is believed to be present for this molecule. Recently, a hot debate over the SERS behavior of PATP was triggered by our finding that PATP can be oxidatively transformed into 4,4'-dimercaptoazobenzene (DMAB), which gives a SERS spectra of so-called "b2 modes". In this perspective, we will give a general overview of the SERS mechanism and the current status of SERS studies on PATP. We will then demonstrate with our experimental and theoretical evidence that it is DMAB which contributes to the characteristic SERS behavior in the SERS spectra of PATP and analyze some important experimental phenomena in the framework of the surface reaction instead of the contribution "b2 modes". We will then point out the existing challenges of the present system. A clear understanding of the reaction mechanism for nitrobenzene or aromatic benzene will be important to not only understand the SERS mechanism but to also provide an economic way of producing azo dyes with a very high selectivity and conversion rate.