1. Introduction

Surface-enhanced Raman scattering (SERS) technology is an ultra-sensitive micro-trace molecular fingerprint spectroscopy analytical tool [1]. Due to the combined action of the chemical enhancement mechanism (CM) and the electromagnetic enhancement mechanism (EM) of SERS sensors, the weak Raman scattering signal intensity of the molecule can be enhanced to a higher level and can be more easily collected by the spectrometer [2–5]. However, SERS sensors with rigid substrates require complex analyte extraction steps, which are inconvenient for on-site detection of complex surfaces such as fruits and vegetables [6–8]. In recent years, a variety of novel flexible SERS sensor structures suitable for complex surface detection have been reported [9–13]. Carbon fiber (CF) has excellent electrical conductivity and stable physical-chemical properties. Simultaneously, the low relative density and low coefficient of linear expansion of CF make it an excellent substrate material for flexible SERS sensors [14]. Flexible carbon fiber cloth (CFC) can bend and fold under external stress conditions. Therefore, the flexible CFC based SERS sensor can be attached to the surface of the test object with different flatness to on-site sampling and detection [15,16]. In 2018, Bian et al. [17] combined three methods (water bath heating, high temperature calcination, and ultraviolet reduction) to prepare a CFC composite SERS structure. The Raman enhancement factor (EF) of rhodamine 6 G (R6G) molecule can reach 1.2 × 10⁸. In 2020, Liu et al. [18] used a combination of three methods (immersion, chemical vapor deposition and displacement reaction) to fabricate a CFC composite SERS structure. The lowest concentration of thiram that can be detected is 0.1 nmol/L. In the same year, Lu et al. [19] used the electrochemical deposition method to connect the CFC to the cathode of the electrochemical workstation, and used a mixed solution of citric acid and silver nitrate as the electrolyte to prepare a carbon fiber cloth composite SERS structure. The lowest concentration of crystal violet (CV) that can be detected is 0.1 nmol/L. In recent years, there are many fabrication methods to prepare cotton and Ag composites as SERS substrates [20–25], such as dipping and drying method, in-suit growth, seed-mediated in-situ growth, bio-template, etc. There is still a potential necessity to optimize methods to prepare large-area SERS substrates.

The photoreduction of metal ions can realize the photochemical synthesis of metal nanoparticles [26,27]. In 2010, Mildred et al. [28] synthesized gold nanoparticles on single-walled carbon nanotubes by UV light irradiation. It is demonstrated that the oxidized groups on carbon nanotubes can provide sites for the nucleation and growth of metallic structures. In the same year, Moon et al. [29] reported the chemical-free growth of metal nanoparticles on graphene oxide sheets with light irradiation. It is demonstrated that the photoreduction technique of nanoparticles could be applied to carbon material substrates with oxygen-containing functional groups.

In this paper, we further improved Raman enhancement performances of carbon fiber cloths decorated with UV photo reduced silver nanoparticles (UV-AgNPs@CFC), and main studies are as follows. (1) The structures were characterized and their formation caused by irradiation time and AgNO₃ concentration were analyzed in detail. (2) SERS properties were simultaneous measured using R6G and CV as standard analytes. Raman signals of low-concentration CV and thiram can be detected on the curved surface of an apple. (3) Finally, the experimental observations are further explained by numerical simulations based on the finite-difference time-domain (FDTD) software and COMSOL Multiphysics software. Furthermore, we replaced the hydrophilized CFC with TiO₂ nanorods-modified carbon cloth (TiO₂-CFC) to fabricate a reusable self-cleaning UV-AgNPs@TiO₂-CFC substrate.

2. Experimental

2.1 Materials

Materials and equipment to fabricate the substrate and the equipment used to characterize substrate are shown in the Supplement 1. The excitation light wavelength of the Raman spectrometer is 532 nm, with laser power of 5 mW, an integration time of 10 s, and the microscope objective of 50×. LabSpec software is used to deal with the baseline of Raman spectroscopy data. In order to reduce the random error, each spectrum line is an averaged result.

2.2 Fabrication of UV-AgNPs@CFC

The structure and fabrication steps of UV-AgNPs@CFC substrate are shown in Fig. 1. First step (Fig. 1(a)): Use an ultrasonic cleaner to clean the carbon fiber cloth. Orderly use acetone, absolute ethanol, and deionized (DI) water sequentially as cleaning agents for the CFC for half an hour, respectively. Then use a nitrogen gun to remove impurities on the surface of the CFC.

 

Fig. 1. Schematic diagram of photochemical modification of CFC with AgNPs: (a) ultrasonically cleaning for the CFC; (b) hydrophilic oxidation treatment for CFC; (c) UV irradiation for the hydrophilized CFC; (d) rinse and clean the prepared SERS substrate.

Download Full Size | PDF

Second step (Fig. 1(b)): The cleaned CFC was put into an oxidizing hydrophilic agent (made by mixing concentrated sulfuric acid and nitric acid in a volume ratio of 1:1), and boiled in a constant temperature water bath for 1 hour. Then the CFC was washed with deionized water to be neutrality. Finally, hydrophilized CFC was placed in a vacuum drying oven at 120°C for 16 hours.

Third step (Fig. 1(c)): The hydrophilized carbon fiber cloth was immersed in a glass petri dish filled with AgNO₃ solution. After that, the petri dish was placed on the rotating sample platform of the KW-4AC UV curer. Then turn on the light source to fabricate UV-AgNPs@CFC SERS substrate with the ultraviolet irradiation.

Fourth step (Fig. 1(d)): Rinse in deionized water to remove residual AgNO₃ on the surface of UV-AgNPs@CFC substrate. The substrate was placed in a fume hood to dry naturally and then the substrate was sealed and stored for further use.

3. Results and discussion

3.1 Structural characteristics of UV-AgNPs@CFC SERS substrate

We analyzed the influence of concentration parameters and time parameters on the structure of the UV-AgNPs@CFC substrate. 0.5 mol/L(M), 1 mol/L, 1.5 mol/L and 2 mol/L AgNO₃ solutions were selected as the concentration variation. Simultaneously, UV irradiation duration of 10 minutes (min), 20 min, 30 min, 40 min, 50 min and 60 min was selected as the time variation.

Figures 2(a1∼a6) are SEM images of SERS substrates prepared by immersing CFC in 0.5 mol/L(M) AgNO₃ solution, named 0.5 M-UV-AgNPs@CFC). Figures 2(b1∼b6) are SEM images of samples in AgNO₃ solution of 1 M (1 M-UV-AgNPs@CFC). Figures 2(c1∼c6) are SEM images of samples in AgNO₃ solution of 1.5 M (1.5 M-UV-AgNPs@CFC). Figures 2(d1∼d6) are SEM images of samples in AgNO₃ solution of 2 M (2 M-UV-AgNPs@CFC). The UV irradiation duration at orderly column is 10 min to 60 min, respectively.

 

Fig. 2. (a1)∼(a6) SEM images of 0.5 M-UV-AgNPs@CFC. (b1)∼(b6) SEM images of 1 M-UV-AgNPs@CFC. (c1)∼(c6) SEM images of 1.5 M-UV-AgNPs@CFC. (d1)∼(d6) SEM images of 2 M-UV-AgNPs@CFC. The UV irradiation duration at orderly column is 10 min to 60 min, respectively. Each SEM image scale is 10 μm.

Download Full Size | PDF

In order to select the optimum samples, we analyzed as follows.

3.2 Formation of UV-AgNPs@CFC

In order to further understand the forming mechanism of AgNPs on carbon fibers, it is necessary to analyze the preparation principle of the substrate. Conventional photochemical reduction methods for the preparation of metal nanoparticles are presented in the Supplement 1. The X-ray photoelectron spectroscopy(XPS) spectrum of the 1.5 M-UV-40 min-AgNPs@CFC sample (Figs. 3(a), (b)) shows that the intensities of Ag 3d_5/2(368.22 eV) and Ag 3d_3/2(374.32 eV) on the sample are higher than other peaks. It is proved that the majority of Ag on the substrate exists in the form of silver element Ag⁰ [30–32]. During the hydrophilization of the CFC with nitric acid and concentrated sulfuric acid, the surface of the CFC (Fig. 3(c)) is gradually covered by oxygen defects such as hydroxyl and carboxyl groups (Fig. 3(d)). Oxygen defects such as carboxyl groups on the surface of CFC will provide growth sites for the formation of silver seeds [26]. According to the Oswald ripening theory, AgNPs after nucleation can continue to absorb the tiny AgNPs in AgNO₃ solution and grow continuously. During the irradiation process, oxygen-containing functional groups become an aggregation center of silver particles [33], which will continue to attract the silver element in the solution to converge (Fig. 3(e)). Finally, silver nanoparticles are grown on the carbon fiber cloth (Fig. 3(f)). Figure 3(g) is the SEM image of the hydrophilic carbon fiber. Figure 3(h) is the EDS image of the rectangular area in (g).

 

Fig. 3. (a) and (b) the XPS spectra of 1.5 M-UV-40 min-AgNPs@CFC samples; (c) clean carbon fiber; (d) carbon fiber with oxygen-containing functional groups; (e) adsorption of oxygen-containing functional groups silver element; (f) nucleation of AgNPs on oxygen-containing functional groups. (g) is the SEM of the hydrophilic carbon fiber. (h) EDS of carbon fiber after hydrophilic. (i1) Hollow hexagonal prism AgNPs on UV-AgNPs@CFC sample. (i2) Semi-hollow hexagonal prism AgNPs. (i3) Hexagonal prism AgNPs; (j1) EDS image of (i1). (j2) EDS image of (i3); (k1) AgNPs in nucleation stage. (k2) AgNPs in forming growth stage. (k3) AgNPs after forming growth. Schematic figure of (l1) hollow hexagonal prism AgNPs, (l2) semi-hollow hexagonal prism AgNPs, (l3) hexagonal prism AgNPs.

Download Full Size | PDF

Nanoparticles on carbon fibers are mainly grown from multiple twinned nanocrystals [34]. Wang et al. demonstrated that as the proportion of <111 > crystal planes increase, hexagonal prism metal nanocrystals appear [35]. SEM images of Figs. 3(i1-i3), show that the hollow hexagonal AgNPs are first formed on CFC (Fig. 3(i1)), and then the interior is gradually filled (Fig. 3(i2)). Finally, the regular hexagonal prism is formed, shown in Fig. 3(i3). The calculated edge lengths are shown in Table S1 (in the Supplement 1). The EDS images also show that the silver content of the hexagonal prism AgNPs (Fig. 3(j2)) is significantly higher than hollow hexagonal prism AgNPs (Fig. 3(j1)).

In the hollow hexagonal prism stage, silver atoms are simultaneously adsorbed by the inner and outer sides of the hollow hexagonal prism AgNPs. After the prism length increases to ∼437 nm, it no longer continues to increase, and the inner side of the prism is gradually filled to form a hexagonal prism AgNPs. SEM images of Figs. 3(k1-k3) show that the length of the hexagonal prism AgNPs also continues to grow during the process from nucleation to full formation. In addition, silver particles with a diameter of ∼40 nm exist on the sidewalls of the hexagonal AgNPs, and their gap is ∼14 nm.

The growth process of AgNPs from hollow hexagonal prism to fully formed can be analyzed according to the Ostwald ripening theory. When UV light is irradiated into the solution, the aggregation of particles also occurs in the AgNO₃ solution to form small-sized AgNPs. Due to the high Gibbs free energy of the small-sized AgNPs in the reaction system, they are relatively unstable. However, the twin interface is the position of the highest energy of Ag multiple twins on the carbon fiber, which is helpful for the adsorption of silver atoms in the reaction system [36]. The silver element of the small-sized AgNPs will be absorbed for the continuous growth of the AgNPs on the carbon fibers. As shown in Figs. 3(l1-l3), the large-grained AgNPs on the carbon fibers absorb the silver atoms of the solution and the small-sized AgNPs under the continuous irradiation of UV light. However, when the irradiation time exceeds 30 min, the AgNO₃ solution becomes turbidity obviously (black brown colloid). This is especially evident in the 2 M AgNO₃. The AgNPs in the solution also increase gradually with irradiation. This may lead to two results. (a) A large number of silver atoms are adsorbed by AgNPs in colloid, delaying the growth of AgNPs on the surface of carbon fiber cloth. (b) Due to the presence of colloid on the surface of the solution. The energy reaching the surface of the carbon fiber cloth at the bottom of the solution is attenuated, which causes the AgNPs on the sample surface to grow slowly.

3.3 Raman enhancement of UV-AgNPs@CFC

We characterized the limit of detection and stability of 1.5 M-UV-40 min-AgNPs@CFC using CV and R6G solutions as probe molecules. The Raman characteristic peak information of CV and R6G is presented in the Supplement 1. Raman detection preprocessing: Sample 1.5 M-UV-40 min-AgNPs@CFC was immersed in the solution for 18 h. Then the sample was taken out to dry naturally and measured.

Figure 4(a) shows Raman spectra of 10⁻¹⁰mol/L CV solution on 1.5 M-UV-40 min-AgNPs@CFC sample and 10⁻²mol/L CV solution on CFC sample. The Raman peak at ∼913 cm⁻¹ is used to calculate the analytical enhancement factor (AEF) with a maximal value of 1.22 × 10⁹. The calculation method of AEF is shown in the Supplement 1. Figure 4(b) shows Raman spectra of 10⁻¹⁰mol/L R6G solution on 1.5 M-UV-40 min-AgNPs@CFC sample and 10⁻²mol/L R6G solution on CFC sample. The Raman peak at ∼611 cm⁻¹ is used to calculate the AEF with a maximal value of 1.69 × 10⁸. In order to reduce random error, all Raman spectra are averaged spectra from 9 Raman measurements. We measured 10⁻⁶mol/L CV solution on 1.5 M-UV-40 min-AgNPs@CFC samples to analyze the stability of the substrate by Raman spectroscopy (Fig. 4(c)). Samples are stored and sealed as follows. Firstly, secure the substrate in the substrate box. Then, the substrate box is then vacuumed using a plastic vacuum bag and sealed. Finally, store the sample in the sample cabinet. After 83 days of sealed storage, the Raman signal at ∼913 cm⁻¹ position dropped to 70.59%, which shows that 1.5 M-UV-40 min-AgNPs@CFC sample is relatively stable. Table S3 (in the Supplement 1) shows the fabrication methods and properties of other SERS substrates. The method reported in this article has certain advantages in terms of high EF and low cost. Actually, we checked all SERS signal from all the substrates, but we found it is difficult to see obvious Raman signals of R6G with lower concentration (∼10⁻¹⁰mol/L) from most of samples 0.5/1/2M-UV-AgNPs@CFC under different irradiation light.

 

Fig. 4. Raman intensity of (a) 10⁻¹⁰mol/L CV and (b) 10⁻¹⁰mol/L R6G with our SERS sample and 10⁻²mol/L CV and R6G on pure CFC as substrate. (c) Raman intensity of 10⁻⁶mol/L CV solution on 1.5 M-UV-40 min sample stored in air.

Download Full Size | PDF

The development of SERS technology for on-site detection puts forward new requirements for multi-molecular detection. Figure 5(a) shows Raman intensities of 10⁻²mol/L crystal violet, rhodamine 6 G and the mixed solution of crystal violet and rhodamine 6 G on CFC. Figure 5(b) shows Raman intensities of 10⁻⁵mol/L R6G + CV mixed solution on 1.5 M-UV-40 min-AgNPs@CFC sample. In order to reduce the random error, multi-points measurements within an area of 2 cm × 2 cm were carried on. The averaged intensity is shown in red line. Figure 5(c) shows the intensity distribution and relative standard deviation at ∼611 cm⁻¹ in each spectrum (Fig. 5(b)). The calculated relative standard deviation (RSD) value is 17.2% at 611 cm⁻¹. The calculation method of RSD is shown in the Supplement 1. Figure 5(d) shows the averaged Raman spectrum within a line mapping of 10⁻⁵mol/L R6G + CV mixed solution on 1.5 M-UV-40 min-AgNPs@CFC sample. The test method is to probe the Raman spectrum point by point along a CFC using the line mapping. Figure 5(e) is the intensity distribution and RSD at ∼611 cm⁻¹ in each spectrum (Fig. 5(d)). The calculated RSD value is 14.4%.

 

Fig. 5. (a) Raman spectra of 10⁻²mol/L CV, R6G and their mixed solution on CFC. (b) Averaged Raman area mapping intensity of 10⁻⁵mol/L R6G and CV mixed solution. (c) RSD at the ∼611 cm⁻¹ peak in (b). (d) Raman line mapping intensity of 10⁻⁵mol/L R6G and CV mixed solution. (e) RSD at the ∼611 cm⁻¹ peak in (d).

Download Full Size | PDF

3.4 On-site detection of AgNPs@CFC substrates

In order to improve the on-site detection performance, some questions must be taken into consideration. (i) How to extract surface molecules without damaging the surface of the analyte? (ii) How to avoid damaging the surface of the SERS substrate? (iii) How to simplify the molecular extraction process? Carbon fiber cloth is a flexible fabric material that can elastically deform under external stress (Fig. 6(a)). It can be used for SERS testing of various complex surfaces, such as fruits with curved surfaces. We design an on-site detection method for flexible Raman measurement based on the characteristics of UV-AgNPs@CFC substrates. Firstly, 10⁻⁵mol/L CV and 10⁻³mol/L thiram solution was sprayed evenly on an apple surface and waited for it to dry naturally, respectively. Then, solvent was sprayed on the apple surface. Finally, the apple surface was fricated by our SERS substrate and then the SERS substrate (1.5 M-UV-40 min-AgNPs@CFC) is used to carry on Raman measurement, shown in Figs. 6 (b) and 6(c). It is demonstrated that our flexible SERS substrate can perform an efficient molecular extraction and Raman measurement on the curved surface.

 

Fig. 6. (a) Flexibility of sample. Raman spectrum of (b) 10⁻⁵mol/L CV and (c) 10⁻³mol/L thiram from an apple surface using our SERS sample.

Download Full Size | PDF

3.5 Electric field simulation analysis

In order to investigate the sample’s Raman enhancement mechanism, we performed 3D modeling in Lumerical finite difference time domain (FDTD) solutions software, and the geometric sizes of the model are based on SEM images and Table S1, shown in Figs. 7(a1-a4)). The simulation environment is set as a vacuum environment. The center wavelength of the incident laser is 532 nm, the incident direction is along the negative Z-axis, and the electric field strength is 1 V/m. The complex refractive index of Ag was set as 0.054007 + 3.4290·i [37], and the refractive index of carbon fiber was set as 2.6913 + 1.4557·i [38]. According to the SEM images, the edge length of the hexagonal prism AgNPs was set as 437 nm, the length was set as 5 μm, the prism center coordinates were (x = 0, y = 0, z = 0). The diameter of the carbon fiber filaments was 7 μm. In order to reduce the computational complexity, the length of the carbon fiber was set to 10 μm.

 

Fig. 7. Schematic diagram of the simulation model: (a1) XY plane, (a2) 3D view, (a3) YZ plane, (a4) XZ plane. The electric field (E-field) distribution (b1) on the XY plane (Z = 0, polarized along the X axis), (b2) on the YZ plane (X = 0, polarized along the X axis), (b3) on the XZ plane (Y = 0, polarized along the X axis). E-field distribution (c1) on the XY plane (Z = 0, polarized along the Y axis), (c2) on the YZ plane (X = 0, polarized along the Y axis), (c3) on the XZ plane (Y = 0, polarized along the Y axis).

Download Full Size | PDF

When the polarization direction of the incident electric field is along the X-axis direction, the simulation results shown in Figs. 7(b1-b3) show that the strongest electric field region of hexagonal AgNPs is at the tip edge. This is because of the surface plasmon resonance (SPR) effect caused by the rearrangement of free electrons along the direction of the electric field.

When the polarization direction is along the Y-axis direction, the simulation results show that (Figs. 7(c1-c3) hexagonal AgNPs will change the distribution pattern of free electrons, and eventually lead to the change of the surface electric field distribution. Figure 7(c1) shows that when the electric field polarization direction is perpendicular to the long axis of the hexagonal prism AgNPs, the SPPs mode that propagates along the prism surface appears on the particle surface.

According to the SEM image in Fig. 2, we selected the sample with the highest density on each concentration sample for simulation. The local electric field maximal value (25.7 V/m) of the 1.5 M-UV-40 min-AgNPs@CFC sample is much higher than that of other samples (Fig. S2 and S3) (in the Supplement 1).

3.6 Self-cleaning properties

Most SERS sensors are not reusable. Because the substrate can be contaminated with target analyte molecules. The application scenarios of SERS sensors are severely limited. Therefore, photocatalytic self-cleaning methods are employed to improve the reusability of substrates. UV light irradiation generates electrons on the surface of the TiO₂ nanorods, which degrades the molecules on the surface [39].

UV-AgNPs@TiO₂-CFC substrate was fabricated using UV irradiation to grow AgNPs on the surface of TiO₂ nanorods-modified carbon fiber cloth (TiO₂-CFC) structure, which has reusable, high hydrophilicity, high sensitivity and self-cleaning properties. We fabricated −2M-UV-10min-AgNPs@TiO₂-CFC substrates using UV irradiation of TiO₂-CFCs soaked in 10⁻²mol/L (−2 M) AgNO₃ solution for 10 min.

The first (1st) spectrum (red solid line) of Fig. 8(a) is the Raman spectrum of −2M-UV-10 min-AgNPs@TiO₂-CFC on the substrate with 10⁻⁶mol/L (−6 M) R6G dropwise added. The typical Raman characteristic peaks of R6G can be observed in the spectrum. However, the UV-1st spectrum (black dot line) is the Raman spectrum obtained by irradiating the substrate with an LED UV lamp for 20 min, and the Raman characteristic peak of R6G cannot be observed. Then, 20 μl of R6G was added dropwise on the substrate again, and the Raman characteristic peak of R6G could be detected, as shown in the second (2nd) spectrum (orange solid line). This is regarded as a cycle of the substrate self-cleaning ability test. The difference in Raman spectra indicates that the −2M-UV-10 min-AgNPs@TiO₂-CFC substrate exhibits self-cleaning ability and can be used for repeated Raman detection. Four cycles of experiments were performed sequentially on the same substrate. The intensity of the Raman characteristic peak ∼773 cm⁻¹ of R6G was selected as the observation object, and the intensity contrast map was drawn (Fig. 8(b)). Therefore, the −2M-UV-10 min-AgNPs@TiO₂-CFC substrate exhibits good self-cleaning performance and reusability. Table S4 shows the fabrication methods and molecular detection information of other SERS substrates. The method reported in this article has certain advantages in terms of low cost, good self-cleaning properties and reusability.

 

Fig. 8. Raman spectra of −2M-UV-10 min-AgNPs@TiO₂-CFC: (a) Raman spectra of 5 times of dropwise addition of R6G and 4 times of self-cleaning at R6G concentration of 10⁻⁶mol/L; (b) Raman intensity at 773 cm⁻¹ during the self-cleaning cycle.

Download Full Size | PDF

Because the Fermi level of silver is lower compared to the energy level of TiO₂, photogenerated electrons can be transferred from the surface of TiO₂ to the surface of metal nanoparticles, greatly reducing the chance of electron and hole recombination and improving the photocatalytic activity of TiO₂ [40]. Since the substrate is a three-dimensional structure, dense “hot spots” can be formed to achieve Raman enhancement. The high hydrophilicity of titanium dioxide-carbon fiber cloth can promote the adsorption of the molecules to be tested on the surface of the substrate. However, the detection limit of R6G on the −2M-UV-10 min-AgNPs@TiO₂-CFC substrate did not lower than 10⁻¹⁰mol/L. Therefore, the −1M-UV-40 min-AgNPs@TiO₂-CFC substrate was fabricated. When the −1M-UV-40 min-AgNPs@TiO₂-CFC substrate was immersed in 10⁻¹³mol/L (−13 M) R6G solution for 18 h, the obvious Raman characteristic peaks of R6G could be detected (in the Supplement 1).

4. Conclusions

A flexible SERS substrate based on CFC-AgNPs with a UV light irradiation method was completely prepared, and an optimized irradiation of 40 min was obtained. Raman measurements with high EF (1.22 × 10⁹), a good uniformity (RSD, 14.4%) were experimentally carried on. The sample can also accomplish two different molecules measurement simultaneously. Additionally, this flexible SERS substrate enables efficient molecular extraction and Raman measurements on the curved surface of apples. Furthermore, the −2M-UV-10 min-AgNPs@TiO₂-CFC substrate exhibits good self-cleaning properties and reusability. This method can be combined with a fluidic chip to fabricate a low-cost SERS lab-on-chip. It would be possible to modify the internal standard on the substrate to make a SERS substrate for on-site quantitative detection, using the Raman peaks of CFC.