1. Introduction

The detection of analytes at ultralow concentrations is of utmost importance for sensing applications in several areas, especially in environmental monitoring [1], life sciences [2] and homeland security [3]. Among these applications, the early detection of diseases [4,5], explosives, chemical warfare agents [6,7] and pollutants [1,8–10] is of extreme value to tackle current world challenges. In this context, surface enhanced Raman spectroscopy (SERS) is a promising technique, with plasmonic metallic nanostructures of various shapes, sizes and materials used as substrates, enhancing the Raman cross-section by several orders of magnitude [11]. The large enhancement in sensitivity arises from a combination of the chemical (CM) and electromagnetic (EM) mechanisms of SERS. As the CM is based on charge transfer and changes in the molecular polarizability [12,13], it requires the analyte to be chemically bound to the substrate [14,15]. On the other hand, the EM occurs by amplification of the electromagnetic fields by excitation of localized surface plasmons, enhancing the Raman signal scattered by molecules bound to or in the close vicinity of the SERS substrate [16].

Recently, graphene-based materials, including graphene oxide (GO), have been proposed as SERS substrates with improved optical properties [17,18]. Unlike the case of metallic nanostructures, the CM is believed to be dominant when graphene-based materials are used as substrates [19–21], improving the SERS signal through charge transfer and fluorescence background quenching [22–25]. These materials have the additional advantage of possessing ultimately high surface-to-volume ratios, and can act as anchoring sites for specific molecules [26–28]. Furthermore, graphene-based materials/metallic nanostructure hybrids are easy to assemble [29–31] and have been shown to provide higher enhancement factors compared to SERS substrates composed of the individual components [21,32]. They can also yield signals with superior temporal stability [33] and reproducibility [34]. In fact, we have shown in a previous study [33] that the interaction dynamics between AuNRs and GO helps to improve the temporal stability of the obtained SERS spectrum, while providing an enhancement factor of up to 10¹⁰.

Despite all their benefits, planar SERS substrates are usually limited by the maximum number of adsorbate molecules that contributes to the signal from each sample [35]. On the other hand, in optofluidic geometries [36,37], consisting of microfluidic channels through which analyte and light simultaneously flow, the signal from several molecules along the channel accumulates, facilitating the detection of samples with very small volumes and concentrations. In this context, special optical fibers with a single [38] or multiple [39] microscopic capillaries running along their length, known as capillary fibers and microstructured optical fibers (MOFs), respectively, emerge as convenient and versatile sensing platforms. They provide simultaneous light and analyte confinement for long lengths, as well as straightforward integration with other optical fiber components [40,41].

Indeed, the inner walls of MOFs [41,42] and capillary fibers [3,9,43] have been demonstrated to serve as a substrate for the deposition of metallic nanostructures and the SERS detection of different molecules. Shanthil et al. [43], for example, showed the uniform coverage of 2-cm-long glass capillaries (0.05-mm inner diameter, 2-mm outer diameter) with silver-silica core-shell nanoparticles. SERS detection of different molecules was then successfully performed. However, while the silica shell helped to prevent silver oxidation and acted as a molecular trap, it also limited the achievable CM and EM enhancements. In general, coating microcapillaries requires pretreatment and functionalization of the inner walls and characterization is performed by scanning electron microscopy (SEM), which can only be carried out by cutting the fiber open. No direct, nondestructive, spectroscopic characterization has been proposed for this purpose.

So far, neither graphene-based materials nor nanocomposites have been used for SERS in microcapillaries, even though microcapillaries in fibers have been coated with GO for applications such as chromatography [44,45], electrophoresis [46] and optoelectronics/photonics [47,48]. Combining the highly stable and sensitive detection obtained by SERS using these materials with the compactness and the low volume/concentration requirements of the optofluidic platform promises a robust, simple and efficient platform for the detection of molecules.

Therefore, in this paper, we propose and demonstrate a silica microcapillary fiber coated with a graphene oxide/gold nanorod (GO/AuNR) nanocomposite and its application as an optofluidic SERS substrate. The inner capillary walls were coated with GO by a simple solvent evaporation method from a GO water dispersion, which allowed the continuous internal coating of fibers of tens of centimeters in length. The GO coating then provided anchoring sites for the AuNRs on the inner capillary wall, which were nondestructively characterized by Raman imaging. The nanocomposite-coated fiber was then successfully used for the detection of Rhodamine 6G (R6G) and Rhodamine 640 (RH640) at concentrations as low as 10⁻⁷ M and 10⁻⁹ M, respectively, with sub-second acquisition times. It also provided sufficient sensitivity for the SERS detection of R6G (10⁻⁵ M) and RH640 (10⁻⁶ M) with acquisition times as low as 0.05 s, showing that the combination of a highly efficient GO/AuNRs SERS substrate with the capillary fiber geometry provides an excellent platform for stable real-time optofluidic SERS.

2. Preparation and characterization of the nanocomposite-coated capillary fibers

Up to ~20-cm-long sections of silica capillary fibers with 80-µm inner diameter and 125-µm outer diameter were coated with the GO/AuNR nanocomposite, with ~3-cm-long sections typically used for SERS. The GO was synthesized by a modified Hummers method [49]; after synthesis, the GO aqueous suspension (1 mg mL⁻¹ concentration) was filtered using a syringe filter with 0.45-µm pore size to remove larger GO flakes, which avoided clogging and obstruction of the capillary. Gold nanorods were synthesized by a seed mediated method with CTAB (cetyltrimethylammonium bromide) as the stabilizing agent [33,50]. The bare AuNRs had average dimensions of 14.6 ± 1.8 nm in diameter and 48.7 ± 3.7 nm in length, which yielded localized surface plasmon resonance peaks at 517 nm and 636 nm. AuNRs extinction spectra were recorded on a Hewlett Packard 8453A diode-array spectrophotometer.

Figure 1 schematically shows the method for coating the capillary with the nanocomposite. In step 1, Fig. 1(a), the tip of the capillary was carefully immersed in the GO suspension through the perforation of a vial septum, made with a syringe needle. The GO suspension quickly filled the fiber hole by capillary force. The coating of the capillary’s inner wall was then obtained as reported elsewhere [48], by drying the fiber from the tips to its center, through local heat provided by a hot finger that approached the fiber from the side. Any remnant water trace was then evaporated by leaving the fiber in an oven at 65°C for 2 hours and, subsequently, in vacuum at room temperature for 12 hours.

 

Fig. 1 Schematic illustration of the coating of the capillary fiber with a GO/AuNR nanocomposite.

Download Full Size | PDF

In step 2, Fig. 1(b), the AuNR’s aqueous suspension (6.7 × 10⁻¹⁴ mol/L) was introduced into the capillary by using a similar procedure to that used for the GO suspension. In this case, however, by applying a continuous 0.8 bar pressure difference (obtained by pressurizing the vial with compressed air), approximately 80 µl of the suspension flowed through the fiber within 1 minute; the filled fiber was then allowed to rest for 12h. Increasing the flow time (while keeping the same flow rate) or the AuNR concentration in the suspension was found to yield lower or no SERS signal, which can be attributed to higher propagation losses. Indeed, it has been shown that optimum signal intensity in fiber-based SERS arises from a tradeoff between the Raman signal enhancement and scattering/absorption losses, both induced by the metal nanoparticles; with higher nanoparticle densities, losses tend to diminish the resulting Raman signal [42].

As previously shown [33], AuNRs synthesized by the seed mediated method present a net positive charge on their surface due to the formation of a CTAB bilayer [51]. Thus, the AuNRs are expected to electrostatically link to the highly oxygenated negative sites at the GO surface [52]. Subsequently, a gentle N₂ flow was introduced to remove the remaining water. The capillary fiber was, then, continuously rinsed with deionized water for 5 min to remove any weakly attached Au particles; no damage or removal of the GO layer due to this procedure was identified. Finally, N₂ flow was again induced, resulting in a dry GO/AuNR coated capillary fiber. Therefore, note that no functionalization of the capillary surface is required, and the preparation method is extremely simple.

For the characterization of the coated capillaries, confocal Raman spectroscopy (WITec Alpha 300R, 633-nm excitation wavelength) and SEM (JEOL JSM-7800F Schottky type field emission electron microscope) were used; the results are depicted in Fig. 2. Figure 2(a) shows an optical microscopy image of the side of the fiber, with the 150 µm × 150 µm red square indicating the selected area for Raman hyperspectral imaging. Figures 2(b) and 2(c) show the Raman intensity (color scale) images for graphene oxide’s G and D bands, respectively. The images were obtained with the laser beam focused slightly above the center of the fiber. The two horizontal high intensity (yellow) lines across the whole width of each image correspond to the intersection between the inner capillary surface and the focal plane, and indicate a continuous GO coverage along the capillary’s inner wall. The gradual reduction in intensity as one moves away from these lines occurs as a consequence of the capillary’s cylindrical geometry. The same continuity was obtained along the entire fiber length over tens of centimeters for several produced samples.

 

Fig. 2 Characterization of the GO/AuNR’s coating to the inner walls of a capillary fiber. (a) Optical microscope image of the side of the fiber; red square represents the area analyzed by Raman hyperspectral imaging. (b)-(d) Raman intensity images for graphene oxide’s G (b) and D (c) bands and for the AuNRs’ ν(Au-Br) mode (d); normalized intensities are represented by a color scale (color bar on the right of each image; dark blue represents the lowest intensity and yellow the highest intensity in each case). (e) Raman spectrum obtained at the position marked by the red cross in (d); the red asterisk indicates the ν(Au-Br) mode. Raman data obtained with 4.2-mW laser power at 633 nm and with 0.5 s integration times. (f) SEM image of a section of the capillary inner wall, exposed by angle cleaving the fiber; white spots correspond to AuNRs.

Download Full Size | PDF

As previously shown [33] the negatively charged GO interacts and attracts CTA⁺ towards its surface, leaving only the Br^– ion covalently bonded to the AuNRs surface. The presence of the ion then allows for an easy and nondestructive confirmation of the AuNRs presence via the identification of the Au-Br stretching mode, ν(Au-Br), at 179 cm⁻¹. It is important to mention that the ν(Au-Br) is only visible under 633 nm laser light excitation, due to its spectral proximity to the AuNR’s plasmonic resonance. Figure 2(d) shows the Raman intensity image of the ν(Au-Br), confirming the presence of AuNRs in several spots along the capillary’s longitudinal section. Note that at some places the signal is so strong that it appears with a high intensity even being significantly out of focus, which may be attributed to specific AuNRs aggregation configurations, responsible for the higher SERS intensities [33]. Figure 2(e) shows the spectrum taken in one of the points exhibiting the Au-Br Raman signal red cross in Fig. 2(d), with the bromide peak marked by an asterisk, allowing for the immediate characterization of the nanocomposite on the capillary’s inner walls. The presence of the AuNRs is also confirmed in the SEM image of Fig. 2(f). For this measurement, the capillary tip was carefully angle cleaved so that the electron beam could reach the nanocomposite. Higher SEM magnifications could not be obtained due to electron charging on the silica surface, which caused image distortion.

3. Nanocomposite-coated capillaries as optofluidic SERS substrates

The nanocomposite-coated fibers were tested as an optofluidic SERS platform, with R6G and RH640 as probe molecules. Virtually identical results were obtained with fibers that were used right after coating and fibers that were stored for a few months and then used, demonstrating the long-term stability of the platform. For SERS, the tip of the fibers was immersed in aqueous Rhodamine solutions with concentrations varying between 10⁻⁵ M and 10⁻⁹ M, so that the inner hole was filled by capillary force. 3-cm-long capillaries were typically used for the SERS measurements, accommodating volumes of 150 nanoliters, which stresses the method’s ability to probe extremely low volumes, as well as low analyte concentrations. For the measurements, the fibers were vertically held under the same confocal Raman microscope used before, so that light from the excitation laser source (at 633-nm wavelength) was launched along the fiber axis. As discussed later, it was found that optimum SERS signal was obtained by launching the laser light right at the capillary wall (i.e., at the interface between the silica region and the air hole), with a 50 × objective lens. The Raman signal was collected in the backscattering mode, and directed to the spectrometer.

Figure 3 presents the detection results for R6G and RH640. Figure 3(a) compares two R6G SERS spectra taken with different integration times (0.5 s – blue; 0.05 s – black), at an analyte concentration of 10⁻⁵ M. A laser power of 185 µW was used and the spectra were normalized by the highest significant peak. The shorter integration time corresponds to 20 spectra per second, approaching standard video frame rate (24 frames per second). Clear SERS signals are observed, thus allowing for real-time chemical analysis. Both spectra present clear R6G spectral signatures, in good agreement with the literature [53–55], with the most prominent Raman modes at 615 cm⁻¹ (C-C-C in plane ring deformation); 775 cm⁻¹ (out of plane C-H angle deformation); 1186 cm⁻¹ (in plane xanthene ring deformation, C-H bend and N-H bend); and 1315, 1365, 1513, 1576 and 1653 cm⁻¹ (all aromatic stretching vibrations of the xanthene moiety).

 

Fig. 3 SERS spectra for R6G (a)-(b) and RH640 (d)-(e) in the nanocomposite-coated capillary (vertically displaced for clarity). (a) R6G concentration of 10⁻⁵ M and integration times of 0.5 s (blue) and 0.05 s (black). Laser power of 185 µW. (b) R6G at concentrations of 10⁻⁵ M (blue), 10⁻⁶ M (red) and 10⁻⁷ M (black). Laser powers of 185 µW (blue) and 1.77 mW (red/black). Integration time of 0.5 s. (c) SERS spectra time series for R6G. Laser power of 185 µW and 0.5 s integration time. (d) RH640 concentration of 10⁻⁶ M and integration times of 0.5 s (blue) and 0.05 s (black). Laser power of 105 µW. (e) RH640 at concentrations of 10⁻⁶ M (blue), 10⁻⁸ M (red) and 10⁻⁹ M (black). Laser powers of 105 µW (blue) and 1.85 mW (red/black). Integration time of 0.5 s. (f) SERS spectra time series for RH640. Laser power of 105 µW and 0.5 s integration time.

Download Full Size | PDF

To infer the detection limit of the optofluidic SERS substrate, the R6G concentration was lowered. For each tested concentration, at least 3 different fiber samples were used, consistently presenting the same spectral features. In Fig. 3(b), the spectra for R6G at 10⁻⁵ M, 10⁻⁶ M and 10⁻⁷ M are shown as the blue, red and black curves, respectively. R6G shows a resonant Raman effect when excited within its absorption band, with a maximum at ≈528 nm. The 633-nm excitation is far enough from this peak to be considered non-resonant [53,56]. Therefore, for a non-resonant molecule, the SERS platform still provides a clear and distinguishable signal at 10⁻⁷ M, which was detected with a sub-microliter sample volume, leading to a R6G mass close to 10 picogram. Note that the capillary hole is easily filled with the probe molecule in solution and that the SERS detection is performed with a static SERS substrate. In contrast, the use of metal nanoparticle colloids [53,56,57] for SERS may require the addition of chemicals and constant optimization to prevent aggregation, preventing their use as readily-usable practical SERS substrates.

The nanocomposite-coated capillary was also tested as a SERS substrate for the detection of RH640, for which the 633-nm excitation leads to the resonant Raman effect. Figure 3(d) compares two RH640 SERS spectra with different integration times (0.5 s – blue; 0.05 s – black) at the analyte concentration of 10⁻⁶ M. A laser power as low as 105 µW was used in this case, to reduce molecular photobleaching. Despite the higher noise in the lower integration time case, both spectra present a clear RH640 spectral signature, agreeing with the Raman shifts that were previously reported [33].

Figure 3(e) shows RH640 SERS spectra for concentrations of 10⁻⁶ M, 10⁻⁸ M and 10⁻⁹ M. At 10⁻⁹ M, the platform still presents enough sensitivity to provide a clear and distinguishable RH640 signal, specially through the peaks at 1510 cm⁻¹ (symmetric C-C stretching mode combined with an in-plane C-O-C deformation of the Xanthene ring) and 1653 cm⁻¹ (C = O stretching mode) [33]. Different relative peak intensities can be observed in the curves of Fig. 3(e), which can be attributed to specific molecular interactions resulting from the different RH640 adsorption geometries in the nanocomposite. In addition, as the concentration is lowered, less RH640 molecules are available at the plasmonic hot spots, making detection more sensitive to environmental changes and inducing variations in the peak intensities and frequency positions [33,53]. The 2 orders of magnitude difference in the detection limit between RH640 and R6G can be attributed to the resonant Raman effect. Indeed, as estimated by DFT calculations in our previous work [33], RH640 presents a pre-resonance condition at 633 nm, involving π- π* electronic transitions between the Xanthene and Benzene π orbitals, which leads to an enhancement factor of 10² in the overall SERS signal.

For applications that require high spectral resolution and real-time assessment, as in the clinical field, integration times ranging from 0.5 – 5 s are desirable [58]. This requirement tends to somewhat limit the number of SERS platforms that are candidates for real-world applications. Here, real-time analyses were performed for R6G and RH640 with integration times of 0.5 s and laser powers of 185 µW and 105 µW, respectively, and are shown as spectral time series in Fig. 3(c) and 3(f). As reported in our previous work [33], graphene oxide-surfactant interaction prevents blinking in GO/AuNRs samples, which is also verified in Figs. 3(c) and 3(f) by the temporally-stable peaks associated to the analyte molecules (black asterisks). In Fig. 3(c), the diffuse background fluctuations between 45 – 75 s and 182 – 230 s can be attributed to a continuum emission of the substrate or fluorescence from non-adsorbed R6G [59] and do not prevent precise molecular assignment. In the case of RH640 [Fig. 3(f)] after ≈45 s an intensity decrease can be observed due to photobleaching reactions caused by the resonance condition. Still, the molecule’s main peaks are observed along the entire time series.

In the proposed optofluidic SERS substrate, the high sensitivity is attributed to two main factors: (i) the GO/AuNR nanocomposite’s SERS properties and (ii) the light and analyte confinement offered by the capillary geometry. In the case of RH640 the resonant Raman effect corresponds to a third mechanism, as already mentioned. As discussed in a previous work [33], the SERS characteristics of the nanocomposite, factor (i), are attributed to a combination of the EM and CM. The EM occurs by the adsorption of Rhodamine molecules in the nanocomposite’s surface, leading to high signal enhancement in the plasmonic hot spots and an enhancement factor varying between 10⁶ and 10⁸ [33]. The interaction of the Rhodamine molecules with the GO surface may also help detection through fluorescence background quenching. Indeed, Li et al. [23] showed that fluorescent dyes may interact with GO through electrostatic attraction, hydrogen bonding, hydrophobic interaction and π → π interactions, resulting in fluoresce quenching. Thus, as the quenching mechanism may include static quenching combined with dynamic quenching (Fӧrster resonance energy transfer) [23], we speculate that the CM may also play an effective role to the overall sensitivity increase of the system. Indeed, when light is launched directly at the center of capillary hole, as discussed below for RH640 [Fig. (4)], only luminescence is detected, indicating that the nanocomposite strongly contributes to the clear detection of the SERS signals.

 

Fig. 4 (a) Optical microscope image of the capillary fiber cross section. The marked points refer to the positions where the spectra shown in (b) were taken. (b) Spectra obtained at the points marked with the same color in (a). Incident powers: 105 µW (red); 975 µW (black and blue). (c) Output intensity profile when a 633-nm laser was launched at the silica wall/capillary hole interface (power: 23.7 µW; capillary length: 1.8 cm).

Download Full Size | PDF

Capillary confinement effects, factor (ii), also contribute to the observed high sensitivity. Light confinement, obtained via partial reflection in the water-silica interface and by total internal reflection in the silica-air outer interface, makes the detected signal be the sum of the signal from several AuNR hotspots, which is believed to contribute to the excellent spectral reproducibility observed from the several tested fiber samples. Also, for a fiber filled with the RH640 solution and positioned perpendicularly to the laser beam, no SERS signal is detected at the low excitation powers as used for the detection along the fiber axis [Fig. (3)], further supporting that the capillary geometry and guiding properties highly contribute to the results. In addition, as wave guiding accumulates the signal from many plasmonic hot spots along the fiber, the SERS signal is readily acquired with the capillary vertically held, while for the capillary positioned perpendicularly to the laser beam it is necessary to look for a hot spot for the signal to be observed. Analyte confinement, in turn, drastically increases the surface-to-volume ratio, allowing for a higher number of molecules to be close to the SERS substrate. We observed a strong SERS signal with RH640 concentrations down to 10⁻⁹ M, which, considering a capillary internal volume of 150 nanoliters, amounts ~75 femtograms of analyte. As a comparison, on a flat SERS substrate (i.e., the nanocomposite on a Si/SiO₂ wafer) approximately 20 µL were required for the measurement and the lowest detectable RH640 concentration was 10⁻¹⁰ M [33], corresponding to 980 femtograms of analyte. Therefore, a sensitivity enhancement factor of ~10 times can be estimated owing to analyte confinement effects.

As mentioned, all measurements shown so far were obtained with light launched at the inner wall of the capillary. In fact, the measured spectrum varied significantly when the light incidence position was varied. Figure 4(a) shows the optical image of the capillary cross section. The black, red and blue marks indicate the approximate locations at which the RH640 spectra, shown in Fig. 4(b) with the same respective colors, were obtained (with a 50 × objective lens). It is possible to observe that the SERS signature is only visible in the blue curve, taken with light launched at the nanocomposite-coated inner wall. For light incident at the center of the capillary and at the silica cladding (away from the inner wall), only RH640 luminesce and the characteristic silica Raman spectrum (with the Si-O bending vibration [60] at 440 cm⁻¹) are observed, respectively.

To analyze the excitation intensity distribution obtained at the different launching positions, a 633-nm laser beam was launched by a 40 × objective lens into a 1.8-cm-long coated capillary fiber filled with water, while the output intensity distribution was imaged with a 10 × objective lens and a CCD camera. A beam splitter before the first objective lens and a microscope setup allowed for the determination of the exact incidence position. The result for light launched at the nanocomposite-coated inner wall is show in Fig. 4(c). It is possible to observe that both the silica cladding and the capillary are illuminated, with the region near the capillary’s wall significantly brighter than when light was launched at the cladding or at the hole center, which is believed to promote the observation of the SERS signal.

We highlight that the capillary coating procedure described here is extremely fast and simple, and that long fibers (of at least tens of centimeters) can be produced and stored for months. These features, together with the real-time sensing of molecules at low concentrations and small sample volumes, point to applications that require quick and reliable detection of analytes in small sample quantities, such as in clinical analysis. Integration with optofluidic fiber systems [61] is feasible and would further facilitate the platform’s use outside the laboratory environment.

4. Conclusion

In conclusion, we presented what we believe to be the first optofluidic SERS substrate based on a GO/AuNR nanocomposite-coated capillary fiber. We showed that the GO/AuNRs coating can be continuously created along the capillary’s inner wall without surface functionalization by a simple two-step method consisting of the creation of a GO coating followed by the AuNRs electrostatic anchoring. The AuNRs presence on the GO surface is easily confirmed through the observation of the Au-Br stretching mode in the SERS spectrum, offering a non-destructive technique for the identification of the gold nanorods. The nanocomposite-coated capillary was successfully tested for the detection of R6G and RH640 down to concentrations of 10⁻⁷ M and 10⁻⁹ M, respectively, with the latter corresponding to ~75 femtograms of analyte. It also allowed for the detection of molecules at low optical excitation powers and spectral integration times as low as 0.05 s in extremely low (~150 nanoliters) volumes of sample. Therefore, the demonstrated capillary fibers arise as an emerging stable and compact platform for optofluidic SERS, in which optics and microfluidics can be combined to achieve high levels of sensitivity, integration and handling, as well as real-time detection.