1. Introduction

The development of photonic crystal fibers (PCFs) opened up new possibilities for biochemical sensing [1–5]. These fibers possess hollow channels that run along their entire length, allowing for the incorporation of a liquid or gas analyte while the light is guided inside the fiber core [6–8]. Ideally, PCFs and more largely microstructured optical fibers could feature large air channels that enable effective liquid infiltration and circulation. Such fibers have been used in different types of sensing, including biosensing, chemosensing or electromagnetic field sensing [9–14] using refractive index, absorption/transmission, fluorescence or SERS readouts [10,15–17]. Standard SERS platforms feature a planar plasmonic substrate composed of a planar nanoroughened metal (silver or gold) surface or their colloidal nanoparticles. The enhancement of the Raman signal occurs in close vicinity of the plasmonic entities (less than 100 nm) and the interaction area between the light, the analyte in planar or colloidal platform and the plasmonic structures is limited to the focal point of the laser (few micrometers in diameter). For PCF-based SERS sensing, the light can interact with the analyte and the plasmonic structures for relatively long length, thereby increasing the interaction surface. In turns, this results in better sensitivity and reliability [18,19].

The PCF topologies used for SERS sensing can be separated into two main classes, hollow core PCFs (HC-PCFs) [20–23] and solid core PCFs (SC-PCFs) [21,24–26]. HC-PCFs have been initially used for SERS sensing since they offer a direct interaction between the light and the analyte [27–29]. However, they are mainly photonic bandgap fibers that suffer from a rather small transmission window, which is shifted when it is filled with liquid analyte [30–33], limiting their applications. On the other hand, in SC-PCFs, the light is guided in the core by total internal reflection. This results in a fixed broad transmission windows, even when the fibers are filled with liquid in cladding holes [3,34]. The light-analyte interaction occurs through a small portion of evanescent field, which extends into the holes of the fiber cladding. In addition, SC-PCFs exhibit stronger SERS enhancement compared to HC-PCFs [35]. The size of the fiber core directly influences the amount of evanescent field in the cladding. A smaller core size results in more light interacting with the analyte in the cladding holes [24]. However, for SERS sensing, decreasing the core diameter leads to fewer number of anchored nanoparticles inside the fiber, which results in the reduced interaction with the analyte. In addition, the core diameter also plays a key role in the coupling of the light between the fiber, the excitation laser and the Raman spectrometer. The larger the core, the more effectively one can maximize the amount of light coupled within the fiber. In a previous study, we had demonstrated that a fiber with an intermediate core diameter will maximize the SERS signal and the reliability of the signal readout [18]. This illustrates that a compromised core size has to be found to benefit from a large surface of interaction, a good coupling efficiency and a relatively large amount of evanescent field. In this prospect, we present a novel PCF design, called suspended ring core fiber (RCF) that benefits from the advantages of SC-PCFs while increasing the surface area of light-analyte interaction. We first illustrate the relevance of the new design through a numerical that can result in an improvement in sensitivity compared to our previous best PCF SERS sensor [18]. After successfully fabricating a RCF with parameters fitting the simulation, we experimentally confirm the improvement in sensitivity.

2. Numerical simulation

A suspended core PCF (SuC-PCF), shown in Fig. 1(a), is a standard SC-PCF used for SERS sensing. The idealistic case of this fiber design is a silica rod of few micrometers diameter (i.e. the core size) surrounded by the analyte or air (as illustrated in Fig. 1(b)). The struts are only needed for mechanically holding the structure to the fiber cladding. An amount of evanescent power larger than 1% is obtained for a rod diameter smaller than 1.5 µm (at λ = 785 nm). In this case, the surface of interaction, which is the external surface of the rod, is below 0.5 mm² for a fiber length of 10 cm (Fig. 1(c)). To maximize the surface of interaction while maintaining a good portion of evanescent field, we hypothesized that the best solution would be a thin silica ring core suspended in air, as shown in Fig. 1(b).

 

Fig. 1. (a) SEM picture of the SuC-PCF. Inset: mode field distribution of the light in the core. (b) Representations of the idealistic cases of the SuC-PCF and RCF. (c) Evolution of evanescent power for increasing rod diameters (red) and increasing thicknesses for 40 µm silica ring surrounded by air (black). The S values represent the surface of the core. Inset: Simulated modes field inside the silica rod and silica ring.

Download Full Size | PDF

The amount of evanescent power of mode propagated in ring cores of 40 µm in diameter with different thicknesses, and in a silica rod-core with several diameters, were simulated using a Finite Element method (Comsol Multiphysics). As plotted in Fig. 1(c), the amount of evanescent field increases with a thinner thickness of the ring core. Both cases exhibit close amount of evanescent field while the surface of the ring core is more than one order of magnitude larger. Here, the interaction surface refers to the surface of the inner and outer sides of the ring. The ring diameter of 40 µm results in an interaction surface of ∼25 mm² for a 10 cm long fiber, which is about 50 times larger compared to the rod (for 1% of evanescent field). This result validates that the ring core design can overcome the limitation inherent to standard SC-PCFs, i.e. the evanescent field can be controlled independently from the diameter of the core. It is therefore possible to have fibers with a large core and a good portion of evanescent field for light interaction.

Subsequently, a model developed by Chen et al [25] was used to estimate a factor proportional to the SERS intensity each fiber could produce, named f_RI in the following:

${R_{fiber}}$ and ${L_{fiber}}$ represent the radius of the core and the length of the fiber respectively. $\eta $ represents the portion of evanescent field in the effective layer and $\alpha $ the attenuation coefficient. The interactions between the guided light and the nanoparticles (NPs) (required for SERS effect) are considered through an effective layer with an effective complex permittivity that depends on the NPs diameter, the coverage density of NPs ($CD$) and the refractive index of the analyte. It is worth mentioning that this relation was simplified from the original model by considering that the fiber properties are equals at the excitation and resulting Raman signal wavelengths. This approximation is sufficient for comparing the relative SERS intensity of the ring core to a rod-core. In the following simulations, we used the same parameters as our best SERS-based PCF sensor [18,36]. The simulated rod, representing the SuC-PCF, had a 3.5 µm core and was 10 cm long. The diameter of the gold NPs (Au NPs) was 60 nm, leading to a thickness of 60 nm for the effective layer, and the coverage density was 30 particles/µm². The surrounding analyte was air (n ∼ 1) as the experiments were realized after drying the fibers. The excitation was considered at 785 nm. To estimate the f_RI of the ring core, the model was adapted by considering two effective layers (inside and outside the ring), thus, increasing the sensitivity by a factor two.

The calculated f_RI of ring cores with a thickness of 0.4 µm and different diameters are plotted in Fig. 2(a). The f_RI increases linearly with the ring diameter. This demonstrates that only the thickness of the ring plays a role in the propagation of the light. This is different from the case of the rod, where the diameter of the core has an impact over the light propagation. A ring core diameter of 100 µm leads to more than one order of magnitude improvement compared to a rod-core of 3.5 µm diameter (similar to our previous best SuC-PCF, with f_RI = 13.40 a.u. and f_RI = 0.68 a.u. respectively).

 

Fig. 2. (a) Evolution of f_RI for increasing values of ring diameter; here CD = 30 particles/µm², thickness= 0.4 nm and L_fiber = 10 cm. For comparison, we added our previous best sensor (rod core). (b) Variations of f_RI with the ring thickness and Au NPs coverage density for 40 µm core RCF. Inset: Variations of f_RI with the particles density for 0.4 µm thick ring.

Download Full Size | PDF

This enhancement could be further increased by optimizing the Au NPs coverage density, as there is a tradeoff between the SERS signal enhancement offered by NPs and their optical absorption. Considering a 10 cm long RCF, with a diameter of 40 µm, various thicknesses (0.4 µm to 5 µm) and NPs coverage densities (0.01 to 100 particles/µm²) were simulated. The results plotted in Fig. 2(b) confirm that high NPs coverage densities do not exhibit the strongest Raman signal. The best sensitivity is achieved for a ring core with a thickness of 0.4 µm and a coverage density of 0.1 particle/µm² (f_RI = 1375.82 a.u.) (inset Fig. 2(b)). The f_RImax for this fiber is more than three orders of magnitude compared to the simulation of the 3.5 µm rod core (f_RI = 0.14 a.u.). It means that by optimizing the ring core and the coverage density of anchored NPs, we could improve by three order of magnitude the sensitivity of our previous best PCF-SERS sensor. It is worth mentioning that the simulations are done considering a forward propagation setup while the measurements are done in backscattering configuration, since it gives a higher sensitivity and the optimization of the fiber length is easier.

3. Experimental results and discussion

3.1 Fiber fabrication and characterization

To fabricate the RCF, we used the stack-and-draw process and fiber fabrication facilities at XLIM research institute, France [19]. As mentioned, the simulation considers the ideal cases of silica rod and ring surrounded by air. Practically, this is not achievable since the core needs to be maintained in place by thin silica struts. These struts are formed by adding external capillaries around the core during the fabrication process. In the case of the RCF, this could lead to the creation of apexes at the contact point between the core and the capillaries. An illustration of this phenomenon is available in the inset of Fig. 3(a). In such fiber, the light is mainly guided inside the apexes and not in the entire ring (Fig. 3(a)). To avoid this, we fine-tuned the temperature and the applied pressure during the drawing of the fiber so that the surface tension allowed to maintain open apexes. Furthermore, if the thickness of the struts is the same as the core, the light could also be guided inside them. Thus, an extra care was taken to have struts thinner than the core. SEM image (SEM QUANTA 450 W from FEI) of the cross section of the fabricated RCF is shown in Fig. 3(b) and (c). The RCF exhibits a ring diameter of 40 µm with a thickness of 0.41 µm. The thickness of the struts is found to be 0.11 µm, i.e. smaller than that of the core. In addition, the apexes between the struts and the ring are well open, as visible in Fi. 3(c). A white light source is coupled to the ring core and Fig. 3(d) shows the near-field distribution of light at the output of the RCF with open apexes. During the fabrication of the fiber, extreme care was taken, which helps in the uniform complete excitation of entire ring compared to the fiber with closed apexes as in (Fig. 3(a)). This corresponds close to the simulated case of a circular ring suspended in air.

 

Fig. 3. (a) Measured output near-field distribution of light propagated through a RCF with closed apexes. Inset: Picture of RCF with close apexes. (b) SEM picture of the RCF and (c) a zoom-in on the RCF ring core. (d) Measured output near-field distribution of light propagated through a RCF.

Download Full Size | PDF

3.2 SERS sensing

To realize SERS measurements inside the fiber, we used reporter molecule (RM), 2-naphtalenethiol (2-NT, Sigma-Aldrich) solution first dissolved in ethanol and subsequently diluted in water. Initially, RCF was cut into 10 cm long fibers and made SERS-active by anchoring Au NPs over the entire inner surface of their holes as described previously [19]. Based on a previous study, the nanoparticle density was estimated to be ∼30 particles/µm² [36]. Then, we pumped 100 µM of 2-NT solution into the RCF using a syringe pump. 2-NT molecules are anchored onto the AuNPs through their thiol group. Finally, the fibers were washed, dried and then positioned under the 50X objective lens of the Raman spectrometer (Renishaw InVia) for measurements in back reflection configuration using 785 nm excitation. To properly excite the entire core of the RCF, we focused the excitation laser beam (∼35 mW) into a multimode fiber first (MMF). The RCF was then butt-coupled to the MMF so that the light exiting could be coupled inside the ring of the PCF. Finally, the SERS signal, generated inside the RCF, was coupled back to the MMF and into the Renishaw spectrometer. To minimize the coupling loss between the MMF and the RCF, we tested five MMFs with different core sizes (200 µm to 400 µm) and numerical apertures (NA, 0.22 to 0.5.). For each MMF, we optimized the butt-coupling conditions to achieve the highest SERS intensity coming from the RM. We acquired ten spectra and calculated the average intensity. The MMF that generated the strongest intensity was with 200 µm core diameter and a NA of 0.39. The coupling losses between both fibers were measured by estimating the power at the output of the MMF and RCF. A second MMF was butt-coupled after the RCF to measure the losses from the RCF to the MMF. The coupling losses were ∼15 dB from the MMF to the RCF and ∼3 dB for the RCF to MMF. It is worth mentioning that the coupling losses from the MMF to the SuC-PCF were measured to be ∼25 dB, which could be critical in creating an all-fibered Raman spectrometer. Such spectrometer would simplify the use for clinicians by removing the alignment of the fiber under a microscope objective.

Finally, to test the SERS response of the fiber, we compared the intensity of 2-NT (1380 cm^-1 peak) from the RCFs with a SuC-PCF having 3.5 µm core size, tested under the same conditions; i.e. same coupling to MMF, fixed laser power and integration time. In the measurement setup, initially SuC-PCF was tested and subsequently it was replaced by the RCF and the coupling with the MMF was optimized again. The seven spectra acquired using 10 s integration time are plotted in inset of Fig. 4(a) and the average spectrum across the seven measurements is plotted in Fig. 4(a) for each fiber. The red and blue regions are indicating the corresponding standard deviation across the measurement for the SuC-PCF and RCF respectively. It is clear that the variations are minimal and the average spectra is matching with the spectrum of 2-NT reported previously [37]. Figure 4(b) represents the normalized intensity of 2-NT obtained with RCFs and SuC PCFs. The experiment was repeated four times with newly prepared fibers (i.e. different fiber samples) to confirm the results. Each time, the reproducibility for both fibers was excellent as they exhibit a significantly low relative standard deviation of only ∼1%. This is due to the large surface area of interaction between analyte and light inside the PCF compared to planar/colloidal SERS platforms, as demonstrated previously [19]. Lastly, the RCF provides a signal twice higher than the SuC-PCF, which is highly significant as it experimentally validates the improvement of sensitivity with the RCF design over previous reported best PCF-SERS sensor. For this fiber length, and Au NPs coverage density within the fibers, the computed $\alpha $ coefficients were: ∼270 a.u. for the rod core and ∼6400 a.u. for the ring core. The resulting improvement should have been approximately eight-fold. These small discrepancies between the experimental and simulated results can be explained by the simplification made during the simulations, i.e. the presence of struts, the coupling efficiencies of the light into the fiber cores and the uniformity in the deposition of Au NPs.

 

Fig. 4. (a) Average SERS spectra obtained with a single SuC-PCF and RCF over seven measurements. Red and blue highlighting represents the standard deviation over seven measurements. Inset: Seven spectra measured for each fiber. (b) Normalized RI of 1380 cm^-1 peak of 2-NT obtained with RCFs and SuC-PCFs sensors. The different cases represent the iterations we made to confirm the results tendency.

Download Full Size | PDF

Finally, we demonstrated here some limitations of the SuC-PCF fiber topology, i.e. the trade-off between surface of interaction and amount of light interacting with the analyte and the coupling loss in butt-coupling (which can be a limiting factor to create an all-fibered Raman spectrometer). The design of the RCF improves simultaneously the two aspects and the entire sensor exhibits a better sensitivity, which is a step further toward a plug-and-play spectrometer, easily usable in a clinical environment.

4. Conclusion

In this paper, we conceived a PCF with a disruptive fiber design based on a ring core for SERS sensing. We successfully fabricated a RCF composed of ring core diameter of 40 µm with a thickness of 0.41 µm surrounded by six large air channels for allowing an efficient liquid infiltration. Light is effectively propagating within the ring core, close to the idealistic case of ring core in air. This key feature was achieved by avoiding light guiding inside the apexes (by maintaining them open) and by ensuring the struts are thinner than the ring core. This novel design enables to increase the surface area for SERS interactions (light – Au NPs – analyte) by more than one order of magnitude in comparison to reported other PCF-SERS probes. Simulation results showed that the new RCF could improve SERS signal by more than three orders of magnitude in comparison to previously reported best PCF-SERS sensor, by optimizing the RCF properties, the coverage density of Au NPs and the fiber length. The relevance of the RCF for SERS sensing was demonstrated by measuring the Raman spectrum 2-NT, showing great promise by improving the sensitivity by more than two fold. Controlling key parameters such as the ring diameter, its thickness and the coverage density of Au NPs will allow maximizing the enhancement inherent to the RCF design. In addition, improving the light coupling efficiency between the MMF and the RCF or between the RCF and the Raman spectrometer might allow to fully exploit the sensing performances of the RCF. Another important aspect of this novel fiber lies in the fact that the benefit of the ring core design is not limited to SERS sensing as it improves other types of biochemical sensing that are localized on the fiber core surface such as fluorescence or absorption/transmission. The surface of the fiber core can also be functionalized with antibodies or aptamers to make them ready-to-use and compatible with clinical translation requirements.