1. Introduction

Surface-enhanced Raman scattering (SERS) is a powerful label-free molecular fingerprinting technique that has been broadly used in various research fields [1–4]. SERS sensitivity is directly related to the materials and micro-/nano-structures of the substrates. Thus far, noble metals [5], transition metals [6], and some semiconductor materials [7] in the form of nanoparticles [8–10], nanowires [11,12], or even more complexed nanostars [13–15] and nanoflowers [16–18] display a notable SERS performance. For trace-level analytes in liquids, high SERS sensitivity relies on the analytes to diffuse closely to SERS-active sites. Preconcentration based on physical drying or chemical adsorption is always necessary, both of which are time-consuming. Therefore, a major challenge in planar SERS substrate is to realize real-time and in-situ detection. To overcome the limitation, three-dimensional SERS platforms, which provide more SERS-active sites within the detection volume, have been extensively investigated in recent years. Capillaries [19], microstructure fibers [20,21], metal-organic frameworks [22,23], porous aluminum membranes [24], and PDMS microchannels [25,26] have been reported. At the same time, some novel microcavity coupling structures have also been developed [27,28], with a detection limit of rhodamine 6 G reaching 10⁻¹¹mol/L. Microstructure fibers, which possess both fluidic channels and low loss waveguides [29], significantly increase light-analyte interaction length and hence lead to unparalleled signal enhancement. Light guided in the fiber core leaks into the liquid channels in the form of evanescent field and interacts with the SERS-active sites and analytes on the inner surface. Among various structures of fibers, suspended-core fibers are of special interest [21,30]. On one hand, they have large liquid channels, which are conducive to liquid injection and self-assembly of SERS-active sites. On the other hand, they have a small core, which effectively enhances the evanescent field. After the decoration of nanoparticles on the inner surface, fiber mode fields alter with the density of the nanoparticles, and the confinement loss changes accordingly. Especially when avoided-crossing between the surface mode and core mode occurs, the intensity near the inner surface enhances remarkably [31,32]. In brief, the mode distribution in microstructure fibers greatly affects the interaction strength and distance, and hence the SERS sensitivity. Nevertheless, to the best of our knowledge, the relationship between the mode distribution and the SERS performance has rarely been studied.

In this paper, we numerically and experimentally studied the modes in microstructure fibers decorated with gold nanoparticles (AuNPs). With an optimized density of AuNPs, avoided crossing between the core mode and the surface mode occurs at the wavelength of excitation laser light. The interaction field changes from a weak evanescent wave to a strong surface wave, resulting in an improved SERS sensitivity. A detection limit of 10⁻⁹ M and an enhancement factor as high as 10⁸ for crystal violet (CV) have been achieved.

2. Materials and methods

2.1 Materials

Chloroauric acid (HAuCl₄) was purchased from Sinopharm Chemical Reagent Co., Ltd. Sodium citrate (SC) was purchased from Aladdin. CV, malachite green (MG) and sodium hydroxide solution (NaOH) were purchased from Macklin. 3-Aminopropyltriethoxysilane (APTES) was purchased from Beijing Huaxia Ocean Technology Co., Ltd. Milli-Q deionized water was used in all the experiments.

2.2 Synthesis and characterization of AuNPs

The synthesis of AuNPs was achieved by reducing HAuCl₄ with SC [33]. Firstly, the seed solution was prepared by adding 150 mL of SC aqueous solution (2.2 mM) to a round bottom flask and heated to boiling, and then 1 mL of HAuCl₄ aqueous solution (25 mM) was added. During the process, the solution color changed from yellow to soft pink, and finally to wine red, indicating the synthesis of Au seeds. The solution was cooled and kept at 90 °C. 1 mL of HAuCl₄ (25 mM) was added and the reaction lasted 30 min. This process was repeated twice to complete the synthesis of the first generation of AuNPs. The solution was diluted by extracting 55 mL of the sample and adding 53 mL of Milli-Q deionized water and 2 mL of SC (60 mM). This solution was used as a seed solution for the synthesis of next-generation of AuNPs. The above procedure was repeated to obtain AuNPs with particle sizes ranging from 20 nm to 110 nm. With the increment of particle size, the color of the solution gradually darkened. The TEM image of AuNPs was obtained by high-resolution transmission electron microscopy (HRTEM). The absorption spectra of CV and MG aqueous solution were obtained using UV-vis spectroscopy.

2.3 SERS measurement

One tip of the AuNPs decorated fiber was immersed in the liquid to be tested for 30 s to fulfill the microchannels by capillary force. The SERS spectra were measured using a confocal Raman spectroscopy system (WITec). During the measurement, the fiber was fixed on a substrate using tape and vertically placed under the objective lens. A 50× objective lens (NA = 0.80, Olympus) coupled the laser light into the core of the fiber and collected the backscattered Raman signal. The laser coupling was optimized according to the SERS intensity. The integration time for each spectrum was set to 5 seconds, and the integration was performed once. The baseline of spectral data was removed using the software of Origin.

2.4 Simulation of fiber eigenmodes

The eigenmodes of the suspended-core fiber were analyzed by finite element method using the electromagnetics module in COMSOL Multiphysics software. The scanning electron microscope (SEM) image of suspended-core fiber was converted to a binary image using Matlab and then imported to COMSOL software as a fiber structure for 2D mode analysis. Scattering boundary condition was used on the exterior boundaries. AuNPs were simplified as a circle with a diameter of 50 nm and arranged along the inner surface. The air holes were filled with water to simulate directly liquid detection. The refractive indices of gold, water, and silica were taken from the material library of the software, which were the experimental data of Johnson and Christy [34], Daimon and Masumara [35], and Malitson [36], respectively.

2.5 Simulation of the extinction spectrum of AuNPs

A finite-difference time-domain (FDTD) method (Lumerical, Inc.) was used to simulate the extinction spectrum of AuNPs. AuNP was simplified as a circle with a diameter of 50 nm. The background material was set to water with a refractive index of 1.33. A plane wave was used to illuminate the nanostructure and perfect matched layer was added as the absorbing boundary condition.

3. Design and decoration of AuNPs in suspended-core fiber

Figure 1(a) shows schematically the self-assembly of AuNPs on the inner surfaces of the suspended-core fiber. The liquid was injected into the air channels in the fiber under the pressure of N₂ gas. The fiber with a length of 10 cm was first cleaned with NaOH solution for 30 min to remove organic materials and hydroxylate the inner surface. Then, 5% APTES methanol solution was pumped into the fiber for about 6 hours to functionalize the inner surface with amine groups. After that, the fiber was fulfilled with AuNPs aqueous solution and AuNPs were bound to amine groups and immobilized on the inner surface. Finally, the fiber was rinsed with deionized water and dried with N₂ gas. The effective fiber length for SERS sensing was measured based on the cut-back technique. Raman scattering intensity remained unchanged with a fiber length longer than 1 cm. Therefore, the fiber was cut into segments of 1 cm for further SERS sensing.

 

Fig. 1. (a) Schematic illustration of the decoration procedures of AuNPs on the inner surface of suspended-core fiber. The fiber was cleaned with NaOH solution, functionalized with APTES solution, and AuNPs were immobilized on the fiber inner surface by bounding to amine groups. (b) Schematic illustration of the SERS probe based on AuNPs decorated suspended-core fiber. The SERS signal was collected in a backscattering configuration. (c) Schematic illustration of the power distribution of the core mode. (d) Schematic illustration of the power distribution of surface mode. (e) TEM image of synthesized AuNPs. (f) SEM image of the end face of suspended-core fiber decorated with AuNPs. Inset: SEM image of the end face of pristine suspended-core fiber. (g) and (h) SEM images of the inner surface of the fiber decorated with AuNPs. To expose the inner surface, the fiber was cut obliquely with a blade. Insets: Statistical plots of particle spacing.

Download Full Size | PDF

Figure 1(b) shows schematically the design and basic principle of the fiber SERS probe. Laser light is coupled into the fiber core and guided along the fiber length. Part of the light leaks out into the liquid microchannels and interacts with AuNPs and analytes. The SERS signal was collected in a backscattering configuration. The fiber core mode mainly localizes in the silica region with a weak evanescent field in the liquid (Fig. 1(c)). In contrast, the surface mode has a dominant field distribution near the fiber inner surface where AuNPs are decorated (Fig. 1(d)). The electromagnetic field will be further enhanced by the localized surface plasmon resonance (LSPR) in AuNPs.

AuNPs were synthesized by reducing HAuCl₄ with SC. The shape and size of AuNPs are uniform with an average diameter of 50 nm, as shown in Fig. 1(e). The end-face SEM image of the suspended-core fiber used in the experiment is shown in the inset of Fig. 1(f). The core diameter is 3 µm and the cladding air hole diameter is ∼30 µm. The flow rate of AuNPs aqueous solution determines the uniformity of AuNPs distribution. With the pressure of N₂ gas in the range of 0.1-0.2 MPa, AuNPs are evenly decorated on the fiber inner surface, as shown in Fig. 1(f). The density of AuNPs was controlled by the retention time of the AuNPs solution in the microchannels. When the retention time is 5 min (Fig. 1(g), Fiber I), the density of AuNPs on the inner surface is low with an average spacing of 60 nm. With the retention time extends to 10 min (Fig. 1(h), Fiber II), the density of AuNPs increases obviously with an average spacing of 20 nm.

4. Numerical simulation of avoided crossing between the core mode and surface mode

Figure 2(a) shows the evolution of two transverse mode profiles with wavelength. The spacing between AuNPs was set to 20 nm to simulate a fiber structure of Fiber II. At a wavelength of 540 nm, mode 1 mainly localizes in the core with less than 3.0% energy distributed in the liquid channels that can interact with analytes. Therefore, the generated SERS signal is weak, while the Raman background of silica is strong. In contrast, mode 2 is a surface mode with more than 50% energy localized near the inner surface of the liquid channels. Nevertheless, surface mode cannot be effectively excited by the incident laser due to the mode mismatch. With an increase of wavelength, avoided crossing occurs between the effective refractive index curves of the two modes (Fig. 2 (b)). Strong mode coupling makes mode 1 evolve to a mode profile similar to mode 2 around the wavelength of 630 nm and the power ratio in the liquid channels for mode 1 increases to a maximum. In other words, mode 1 has a high energy ratio and power density localized near the surface of the liquid channels and can be directly excited by the optical coupling. Therefore, when the fiber is excited with a laser wavelength of 633 nm, SERS sensitivity will be improved.

 

Fig. 2. (a) The simulated profiles of two eigenmodes in Fiber II at different wavelengths. (b) The evolution of the effective index (black lines) and power ratio (red lines) in the liquid channels of two eigenmodes in Fiber II. (c) The linear relationship between the avoided crossing wavelength and the AuNPs spacing. (d) The simulated transmission loss of mode 1 in Fiber I and Fiber II.

Download Full Size | PDF

The density of AuNPs will greatly affect the equivalent refractive index of fiber core and cladding. Therefore, the avoided crossing wavelength depends on not only the fiber structure but also the density of AuNPs. AuNPs spacing of 20 nm, 30 nm, 40 nm, 50 nm, and 60 nm were simulated and the results are shown in Fig. 2(c). The avoided crossing wavelength shifts linearly to blue with an increase of AuNPs spacing, which centers at 599 nm for Fiber I and shifts to 630 nm for Fiber II. When avoided crossing occurs, the confined core mode strongly couples to the leaky surface mode. Therefore, the density of AuNPs also affects the transmission loss of the fiber, thus affecting the interaction length between light and analytes. As shown in Fig. 2(d), the transmission loss of mode 1 increases to a maximum at the avoided crossing wavelength. Although Fiber II has a higher density of AuNPs, its transmission loss is lower than Fiber I around 599 nm.

5. SERS performance and its dependent factors

Figure 3 shows the SERS spectra of CV solution excited by 532 nm and 633 nm lasers in Fiber I and Fiber II, respectively. The insets show the corresponding mode profiles and their power ratio in the liquid channels. A higher power ratio in the liquid channels corresponds to a higher SERS intensity. Fiber II excited with 633 nm laser, which has the highest power ratio in the liquid channels, gets the highest SERS intensity. The experimental results agree well with the numerical results and fully demonstrate the importance of mode distribution in fiber SERS performance.

 

Fig. 3. SERS spectra of CV solution (10⁻⁶ M) excited by 532 nm and 633 nm lasers in Fiber I and Fiber II, respectively. Insets: the corresponding mode profiles along with the power ratio in the liquid channels.

Download Full Size | PDF

To investigate the SERS detection capability of Fiber II, CV solutions with concentration range from 10⁻⁹ M to 10⁻⁵ M were measured, as shown in Fig. 4(a). The detection limit is as low as 10⁻⁹ M. Figure 4(b) shows the dependence of the peak intensity at 1616 cm⁻¹ on the concentration of CV in the range of 10⁻⁵−10⁻⁹ M. It is nonlinearity but fits well with the Freundlich isotherm. The corresponding determination coefficient (R²) is 0.988. This indicates that the adsorption of CV molecules onto AuNPs in Fiber II obeys the Freundlich isotherm. In order to evaluate the Raman enhancement ability of Fiber II, the normal Raman scattering spectrum of CV solution in liquid pool was measured and the detection limit is only 10⁻² M, as shown in Fig. 4(a). Therefore, the enhancement factor can be calculated to be 1.02 × 10⁸.

 

Fig. 4. (a) SERS spectra of CV with concentration range from 10⁻⁹ M to 10⁻⁵ M detected using Fiber II. (b) The linear relationship between SERS intensity at 1616 cm⁻¹ and concentration of CV. (c) Simulated extinction spectra of a single AuNP and two AuNPs separated by a spacing of 20 nm. The purple line is the experimentally measured extinction spectrum of AuNPs aqueous solution. Inset: the electric field distribution at a wavelength of 532 nm and 633 nm, respectively. (d) The experimentally measured absorption spectrum of CV and MG aqueous solution. (e) SERS spectra of MG with concentration range from 10⁻¹⁰ M to 10⁻⁵ M detected using Fiber II. (f) SERS intensity at 1612 cm⁻¹ as a function of MG concentration.

Download Full Size | PDF

Except for the power ratio in liquid channels, there are many factors contributing to the SERS performance. Firstly, the morphology and density of AuNPs greatly affect the strength of LSPR. In Fiber I, the average spacing between AuNPs is large, which can be simulated by a single nanoparticle. In fiber II, the decreased average spacing enhances the interaction between nanoparticles, which can be simulated by a dimer. Figure 4(c) shows the simulated extinction spectra of a single AuNP with a diameter of 50 nm and a dimer composed of two AuNPs with a spacing of 20 nm. The extinction peak locates at 515 nm for a single AuNP and red shifts to 537 nm for the dimer. The extinction spectrum of AuNPs aqueous solution was also experimentally measured. The extinction peak is slightly red shifted compared to the simulated result, because the diameters of actual AuNPs are not strictly equal to 50 nm and the distance between AuNPs in water is random. Therefore, compared to a wavelength of 633 nm, the laser with a wavelength of 532 nm can more effectively excite the LSPR in AuNPs both in Fiber I and Fiber II. Secondly, the excitation efficiency depends on the matching degree between the photon energy and the molecular energy band. The laser with a wavelength of 532 nm falls within the absorption spectrum of CV (Fig. 4(d)), which matches well with the molecular energy band to excite the resonance Raman scattering. Thirdly, the interaction length between the light and the analytes depends on the transmission loss of the fiber. As shown in Fig. 2(d), the transmission loss at 532 nm is much lower than that at 633 nm both for Fiber I and Fiber II. Nevertheless, the SERS signal excited by 633 nm laser in Fiber II is much stronger than that excited by 532 nm laser. These sufficiently indicate the importance of mode field distribution in fiber SERS probe. The mismatching between these dependent factors results in a relatively high detection limit of CV. By comparison, MG has an absorption band centered near 633 nm (Fig. 4(d)), and its detection limit can be as low as 10⁻¹⁰ M (Fig. 4(e)) and the SERS peak intensity at 1612 cm⁻¹ also fits well with Freundlich isotherm (Fig. 4(f)). In our future work, we will optimize the fiber structure or the size of AuNPs to match the wavelength of avoided crossing with the LSPR of AuNPs. The synergistic action of high power ratio on fiber inner surface and resonance exciting of AuNPs and molecules will greatly improve the SERS performance.

6. Conclusions

In conclusion, we numerically and experimentally studied the field enhancement in fiber SERS probe based on avoided crossing between the fiber core mode and the surface mode. The avoided crossing wavelength depends linearly on the nanoparticle spacing and shifts to the excitation laser wavelength of 633 nm with a nanoparticle spacing of 20 nm. The light-analyte interaction field changes from a weak evanescent wave to a strong surface wave and hence greatly improves the SERS performance. The detection limit of CV is 10⁻⁹ M, and the enhancement factor reaches 10⁸. It is worth to mention that the detection was carried out directly in the static liquid phase without pre-concentration or processing of the solution, ensuring the rapid trace detection in liquid possible. In addition to the power ratio in the liquid channels, the LSPR in AuNPs, transmission loss of fiber, and the excitation of molecules are all laser wavelength dependent. Collaborative optimization of them will further improve the SERS performance.