Wide-Field Digital Surface-Enhanced Raman Scattering: Quantitative Single-Molecule Detection with High Sensitivity and Throughput

Abstract

Due to its exceptionally high sensitivity and specificity, surface-enhanced Raman scattering (SERS) is widely employed in diverse fields, including biomedicine, environmental monitoring, and food safety. Nonetheless, the lack of reproducibility and substrate uniformity has seriously hindered its application in quantitative detection, particularly at low analyte concentrations. Recently, the concept of digitization has been integrated into SERS, enabling quantitative and sensitive detection with promising applications (Bi et al. Nature 2024, 628, 771−775 38632399 ). In this work, we further developed a wide-field digital SERS (WidiSERS) by employing wide-field microscopy for high throughput. Protein assembled gold nanorod dimers are used for largely enhancing the Raman signals. Reproducible quantification of a wide range of target molecules at extremely low concentrations is achievable through single-molecule measurements. Trace-level quantification of ciprofloxacin in a complex milk environment and phenylalanine in cell culture medium was also achieved, verifying the practicability and accuracy of this method. Meanwhile, the gold nanorod dimer substrate is both simple to prepare and reusable after UV/ozone cleaning. WidiSERS is expected to emerge as a preferred method for ultrafast and effective detection across various fields.

Introduction

Label-free detection technologies have revolutionized molecular analysis by preserving the native properties of target analytes, enabling real-time, noninvasive, and highly sensitive detection. − Commonly employed label-free techniques include surface plasmon resonance (SPR), , quartz crystal microbalance (QCM), , electrochemical impedance spectroscopy (EIS), , photothermal imaging, − and surface-enhanced Raman scattering (SERS). − SPR is highly effective for real-time, label-free molecular interaction analysis, while EIS offers notable sensitivity and adaptability for portable on-site applications. QCM excels in detecting minute mass changes, making it ideal for multiphase sample analysis. Among these methods, surface-enhanced Raman scattering (SERS) stands out due to its exceptional sensitivity, allowing single-molecule detection through plasmonic nanostructures. When a target molecule adsorbs onto the rough surface of a metallic nanostructures such as gold or silver, the Raman signal can be enhanced by 6–10 orders or magnitudes. The enhancement is achieved through two primary mechanisms: electromagnetic enhancement, driven by localized surface plasmon resonance (LSPR), and chemical enhancement, involving charge transfer between the analyte and metal surface. In addition, SERS also provides molecular-specific vibrational fingerprints, making it highly suitable for analyzing complex samples. ,

The exceptional sensitivity and selectivity of SERS have enabled its broad application in diverse fields. In biomedicine, SERS has been utilized for early cancer detection and noninvasive biomarker monitoring, including circulating tumor DNA (ctDNA), providing new opportunities for precision medicine. − In environmental monitoring, SERS enables the detection of hazardous substances, such as heavy metals and pesticides, at femtomolar concentrations, supporting regulatory compliance and environmental safety. , Furthermore, SERS is pivotal in catalysis, providing real-time insights into reaction intermediates and mechanisms, thereby supporting the development of more efficient catalytic processes. Despite its transformative potential, SERS also faces challenges in reproducibility, substrate uniformity, and quantitative detection, particularly at ultralow analyte concentrations.

To address challenges in achieving reliable and reproducible molecular quantification, digital surface-enhanced Raman scattering (dSERS) has been introduced as an innovative method. , Drawing inspiration from digital polymerase chain reaction (dPCR), dSERS transforms analog Raman signal intensities into binary outputs (“detected” or “not detected”), enabling the quantification of ultralow analyte concentrations through statistical models such as the Poisson distribution. , For instance, Bi et al. developed digital colloidal-enhanced Raman spectroscopy (dCERS) by integrating confocal Raman mapping and single-molecule counting. This approach demonstrated quantitative detection of diverse analytes, including dye molecules, metabolic small molecules, nucleic acids, and proteins, with a detection limit below 1 fM. Similarly, Wen et al. employed a novel dSERS chip combined with confocal Raman mapping for precise, rapid, and label-free quantification of viable bacteria, showcasing its potential for applications in complex biological systems. However, confocal Raman spectroscopy, a key component in many dSERS methods, typically sacrifices the acquisition speed to enhance the resolution of weak signal spectra. Conversely, ensuring high throughput demands larger imaging areas and faster acquisition ratesrequirements that are difficult to fulfill with confocal Raman techniques.

In this study, we present a wide-field digital SERS (WidiSERS) method that integrates the high throughput of wide-field single molecule microscopy with the extraordinary enhancement factors provided by gold nanorod (GNR) dimers. WidiSERS is achieved by wide-field imaging of a large area with diameter of 60 μm, which provides the ability to simultaneously capture the SERS kinetics of 30–50 nanorod dimers. By analyzing the blinking events of each individual GNR dimer, we can quantitatively count the diffusing molecules near the GNR dimers. Thus, WidiSERS enables the simultaneous and rapid detection of multiple analytes while achieving trace-level quantification through single-molecule counting. By overcoming the intrinsic challenges of traditional SERS, our approach demonstrates robust reproducibility and sensitivity, as validated in various scenarios, including ciprofloxacin detection in complex milk matrices and phenylalanine in cell culture medium.

Results and Discussion

To achieve large Raman enhancement, GNR dimers were prepared following a reported method. The detailed preparation and characterization of the GNR and dimers are shown in Section S1 in the Supporting Information. The GNRs exhibited a uniform morphology with an aspect ratio of 2.6, as shown in Figure S1. The GNR dimers are dispersed onto a 22 × 22 mm glass substrate followed by UV/ozone cleaning to remove the intermediate TBS linker molecules between the GNRs (Experimental Section and Figure S2). The distance between the two GNRs is about 5 nm, which generates a high electromagnetic field and provides an enhancement factor as high as 10⁶ based on the theoretical simulation, which ensures the extraction of effective signals (Figure S3). The steady state and dynamic Raman spectra of different molecules were measured using a confocal microscope to obtain a better signal-to-noise ratio; see Figures S4 and S5 in the Supporting Information. For WidiSERS, a wide-field microscopy system was employed to simultaneously capture the dynamics of multiple GNR dimers under 808 nm laser excitation (Figure a). Owing to the high throughput and real-time dynamic detection capabilities of wide-field microscopy, the SERS kinetics of 30–50 GNR dimers were collected simultaneously. Figure b illustrates the typical dynamics of the two dimers as an example. The effective number of blinking events for each dimer were counted by setting an appropriate threshold. The ratio between the blinking events and the number of GNR dimers is defined as the ratio of positive voxels (RPV) which can be used as a quantitative parameter to characterize the concentration of the analytes.

1.

Schematic of the WidiSERS method. (a) Principle of the WidiSERS method. GNR dimers were dispersed on a coverslip and imaged (left); sample solution was added to the reaction pool. SERS dynamics of 30–50 GNR dimers are simultaneously collected (middle). RPV can be calculated from the blinking events and number of GNR dimers (right). (b) Typical dynamics of two GNR dimers. The effective blinking events for each dimer were counted by applying an appropriate threshold.

Furthermore, signal consistency is often influenced by the uneven distribution of hot spots in the wide-field method. To address this, a digital quantitative approach was employed to statistically minimize individual variations. Videos of the Raman dynamics were recorded with 1500 frames with an exposure time of 20 ms. For each sample, 20 videos were recorded and analyzed, giving the results presented in Figure . The distribution of blinking positive number of both methylene blue (MB) and crystal violet (CV) reveals that the frequency of the Raman signal decreases with a gradual decrease of the sample concentration. This is because, when the concentration of the target molecules decreases, the chance that a molecule moving into the hot spot decreases, which is consistent with theoretical expectations. Notably, all dimers within the laser spot region have to be counted, including the ones that do not have any Raman signals, because the number of these dimers are closely associated with the concentration of solution and should be retained in the final calculation. To exclude the signal from other possible impurities, a control experiment was performed using pure water as the sample. Much less Raman signal was observed (Figure S6). To confirm the quantitative ability of WidiSERS, the RPV values were calculated for samples with different concentrations. A good linearity with concentration in log–log coordinates was obtained, consistent with the results of Ye et al. It is worth noting that the detection ranges are different for different analytes which is mainly due to three factors: (1) The intrinsic Raman scattering cross-section of the analyte, as stronger Raman scattering naturally yields more intense blinking signals under the same level of surface enhancement, providing higher RPV values; (2) The diffusion coefficient, which determines the probability of the molecule, moves into the gap between two GNRs; (3) The interaction between the analyte molecule with GNR, which not only contributes to signal amplification through chemical enhancement but also increases the duration time in the gap between GNRs.

2.

Quantitative detection of trace amounts of chemicals by WidiSERS. (a) The distribution of the positive number at different concentrations of CV. (b) Dependence of RPV values on the concentration of CV. The detection limit is estimated to be 9 pM based on the control RPV_water = 0.16. (c) The distribution of positive number at different concentrations of MB. (d) Dependence of RPV values on the concentration of MB. The detection limit is estimated to be 1 pM based on the control RPV_water = 0.16. These results highlight the capability of WidiSERS to achieve quantitative detection across a broad concentration range.

Most of the SERS detection is based on Stokes Raman shift, which provides strong signal for high sensitivity. However, for those molecules with fluorescence, the Raman detection can be seriously embedded by the strong fluorescence background, which is several orders of magnitude stronger and can also be enhanced by the GNR. Anti-Stokes signals can effectively eliminate the fluorescent background of molecules compared to Stokes signals. To expand the application of WidiSERS, we also test the detection CV molecules via anti-Stokes signals, as illustrated in Figure . Benefitting from the large enhancement from the GNR dimers, a clear burst of Raman signal can be observed from the dynamics of anti-Stokes Raman spectra recorded by a confocal Raman microscope under 785 nm laser excitation. After collecting the Raman dynamics of CV samples with different concentrations, the statistics of the Raman bursts and the dependence of RPV on concentrations are shown in Figure c,d, respectively. A very good linear relation was obtained in log–log coordinates, demonstrating the validity of WidiSERS in anti-Stokes mode.

3.

Quantitative detection of trace amounts of CV by WidiSERS in the anti-Stokes range. (a) The schematic of WidiSERS in anti-Stokes mode. (b) Single-molecule dynamics of CV on a GNR dimer by confocal Raman microscopy. (c) The distribution of positive number at different concentrations of CV. (d) Dependence of RPV values on the concentration of CV from (c) with an R ² value of 0.99. These results highlight the capability of WidiSERS to achieve quantitative detection across a broad concentration range in anti-Stokes Raman mode.

To further validate the effectiveness and practical applicability of WidiSERS, we tried to use it in real applications. Ciprofloxacin (CFX) is an antibiotic that is frequently overused in animal husbandry. Phenylalanine has been proven to be related to the metabolism of bone cancer cells. Thus, we applied WidiSERS for the detection of CFX and phenylalanine in milk and DMEM, respectively. Proteins in both cases are removed by centrifugation using a 10 kDa ultrafiltration tube. Raman spectra and spectral kinetics were also recorded using confocal Raman microscopy to confirm the molecules (Figures S7 and S10). Based on National food safety standard-Maximum residue limits for veterinary drugs in foods standard (GB 31650-2019), the maximum allowable concentration of CFX in milk is approximately 0.1 μg/mL. Therefore, samples with different concentrations of CFX (0.01–1 μg/mL) were measured by WidiSERS. As shown in Figure b and Figure S8, the number of Raman bursts increases with increasing concentration, and a good linear relation between RPV and concentration was obtained under log–log coordinates. A similar result was obtained for phenylalanine in DMEM (Figure ). The slightly reduced CFX values in milk may be attributed to residual carbohydrates in milk that compete with CFX, even after protein removal. Control experiments were performed without CFX or phenylalanine showing a few Raman burst signals (Figures S9 and S11).

4.

Quantitative detection of trace amounts of CFX in milk by WidiSERS. (a) Schematic of CFX detection. Typical wide-field Raman imaging (left) and dynamics (right) of two gold nanorod dimers in CFX solution. (b) The distribution of positive number at different concentrations of the CFX in milk. (c) Dependence of RPV values on the concentration of CFX from (b) with an R ² value of 0.99. The detection limit is estimated to be 8 pg/mL (25 fM) based on the control RPV_milk = 0.22. These results highlight the capability of WidiSERS to achieve quantitative detection in the actual complex environment.

5.

Quantitative detection of trace amounts of phenylalanine in cell culture medium DMEM by WidiSERS. (a) Schematic of phenylalanine detection. Typical wide-field Raman imaging (left) and dynamics (right) of two gold nanorod dimers in phenylalanine solution. (b) The distribution of positive number at different concentrations of the phenylalanine in DMEM. (c) Dependence of RPV values on the DMEM of phenylalanine from (b) with an R ² value of 0.98. The detection limit is estimated to be 3 nM based on the control RPV_DMEM = 0.29. These results highlight the capability of WidiSERS to achieve quantitative detection in the actual complex environment.

By combining the advantage of gold nanorod (GNR) dimers and wide-field microscopy, WidiSERS resolved the trade-off problem between sensitivity and throughput, achieved precise quantification over a broad concentration range, and demonstrated its effectiveness in complex scenarios, including ciprofloxacin detection in milk and phenylalanine in cell culture medium. Compared to confocal Raman-based digital SERS techniques, ,, WidiSERS significantly enhances detection speed and throughput. The digital nature of WidiSERS, combined with its ability to capture transient single-molecule events over a wide field of view, provides a statistical framework that could help distinguish true signals from background fluctuations. In addition, the anti-Stokes Raman detection can significantly offer an effective way to mitigate autofluorescence and improve its anti-interference capability, broadening its potential applications for complex analytes, such as whole blood, body fluids, or wastewater containing multiple coexisting contaminants. To further improve the sensitivity and specificity of WidiSERS for robust and scalable detection in complex and unprocessed samples, two complementary strategies are proposed: (1) the fabrication of high-performance SERS substrates using advanced techniques such as lithography and 3D printing to achieve improved signal uniformity and enhancement; (2) the integration of machine learning algorithms to extract molecular specificity from both blinking kinetics and spectral features.

Conclusion

In summary, we have developed a wide-field digital SERS (WidiSERS) method based on the dynamical measurement of the SERS signal of individual dimers of GNR using a wide-field microscope. WidiSERS integrates the advantages of wide-field microscopy (high throughput), the high enhancement factors provided by GNR dimers, and single-molecule counting to achieve rapid and effective quantitative molecular detection. We also demonstrated the validity of WidiSERS for different dye solutions, antibiotic detection in milk, and amino acids in cell culture medium. The combination of high throughput, sensitive detection, and robust statistical analysis makes WidiSERS a promising tool for the ultrafast and reliable detection of trace analytes in complex environments.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/cbmi.5c00014.

• Experimental methods, gold nanorod characterization, SERS substrate assembly, Raman spectra of analytes, electric field simulations, quantitative CFX detection in water, and control experiments (water/milk/DMEM) (PDF)

S.Y. carried out the experiments and wrote the manuscript. W.Z., L.L., W.X., and Z.X. helped with the optical setup. L.T. wrote the codes for data analysis. K.M. and J.M. helped with the theoretical simulation. Y.T. conceived the idea and supervised the project. All authors were involved in the writing and revising process of the manuscript.

The authors declare no competing financial interest.