1. Introduction

Raman scattering is an extremely useful spectroscopic tool because of the high level of molecular specific information obtainable. For this reason, Raman spectra are often referred to as “molecular barcodes”. Raman scattering is an inelastic scattering process occurring when a photon interacts with a molecule and loses or gains an amount of energy equal to that of a characteristic molecular vibration of the molecule. The Raman cross section of a molecule, σ_R, which describes the ratio of Raman scattered photons to incident photons, is typically a relatively modest value of 10^-29 cm² per molecule. However, this probability of Raman scattering can be dramatically increased by more than 13 orders of magnitude to around 10^-16 cm² per molecule by attaching the molecule to a metallic nanoparticle or surface with nano-sized features [1]. This scattering process is known as surface enhanced Raman scattering (SERS). This enhancement is primarily due to a local increase in the intensity of the exciting radiation as a result of plasmon excitations in the metal [2]. A further enhancement of 3–4 orders of magnitude can be gained when the molecule is excited by light that lies within, or close to, an electronic absorption band of a chromophore in that molecule. This is known as surface enhanced resonance Raman scattering (SERRS).

Biochemical sensing is one area that has benefited from the detailed information obtainable through SERS spectroscopy. The shift in wavenumber of the scattered radiation is related to the exact conformation in which the molecule resides. Hence information on the distribution of conformations of the analyte, as well as concentrations, can be revealed. The peaks of a Raman spectrum are typically well defined, and this, combined with the fact that there are as many peaks as there are Raman active vibrations, makes SERS an excellent platform for multivariate sensing. A common solvent for biological systems is water, which has an exceedingly small Raman signal, and therefore will contribute minimally to any noise background. Furthermore, as a result of being in close proximity to a metal surface, the analyte suffers almost no photodamage [3].

So just how good has SERS proven to be as a bioanalytical sensor? One of the most striking achievements of the SERS spectroscopy techniques is the ability to detect single molecules. SERS spectra of single DNA base [4], hemoglobin [5], myoglobin [6], and a protein [7] molecules have been obtained. Another achievement is the use of SERS to probe the interior components of live cells through excitation of gold nanoparticles inside the cells [8]. Remarkably, the innermost confines of the cell, including the nucleus, may be probed by nanoparticles grown inside living cells [9].

A drawback of SERS for sensing is that almost all effective SERS systems, and certainly all systems where single molecules have been detected, consist of interacting metallic particles most commonly in the form of nanoparticle aggregates. Raman enhancements of approximately 6 orders of magnitude greater than ordinary SERS signals can be obtained for molecules that lie within an interparticle interstice [2], known as a “hot-spot”. However, such environments cannot be created with regularity, and additional constraints such as the requirement that the electric field vector is polarized along the interparticle axis, are incurred. Hence, the huge variation in signal strength from particles in different locations in a typical sample limits the potential of quantitative SERS sensing.

Raman scattered photons can be emitted in all directions [10], making collection of most or all of the Raman scattered photons difficult. Raman scattered light emitted from particles inside an optical fiber, where the excitation light is also guided by the fiber, has been shown to increase the detection limit of some systems by up to 1000 times [11]. The fiber effectively captures and directs the scattered light to the detection system. A liquid core optical fiber can thus be used to increase the proportion of the signal collected. Historically, liquid core optical fibers have been fabricated from a glass capillary and filled with a higher refractive index liquid to allow total internal reflection (TIR) guidance of both the pump and scattered radiation. It has been possible to have aqueous analytes forming the core of index guiding optical fibers since the development of Teflon-AF by DuPont, as this material has a lower refractive index than that of water, as well as most buffer solutions and solvents used in biochemical applications. Detection of Raman scattering from the liquid core of Teflon fiber has recently been reported [12]. In another recent report, a SERRS signal was collected from the liquid core of an index guiding microstructured optical fiber (MOF), in which the hybrid air-silica cladding had a lower average refractive index than that of the aqueous core [13].

Hollow core microstructured optical fibers (HC-MOF) commonly referred to as photonic bandgap fibers [14], may also be used for aqueous analyte sensing [15]. These fibers possess a microstructured cladding that can confine light to a hollow core by a process of coherent reflections from interfaces in the cladding. Strong reflection by the cladding is observed for some wavelengths of light leading to photonic bandgaps. One can think of the light being confined in a HC-MOF because radiation in the hollow core is not easily able to couple to features in the cladding, hence no loss can occur. In this way, the core of HC-MOF does not need to be of higher refractive index than the cladding.

In the case of isotropic emission by particle in the core of an optical fiber, the fraction of emission collected depends on the N.A. of the optical fiber. Radiation emitted at an angle less than the critical angle will not be guided by total internal reflection and hence will not contribute to the detected signal, placing limitations on sensing schemes such as those described in [13]. An increase in the solid angle of collection beyond what is currently attainable in index guiding fibers is possible through the use of HC-MOF. If a high refractive index contrast exists between the materials of the cladding, then strong reflection from the cladding will occur over a wide range of emission angles from a Raman scatterer in the core of the fiber. This effect is due to increased band splitting with increased refractive index contrast [16]. An example of a HC-MOF with a high index contrast is the multilayer photonic bandgap fiber consisting of alternating layers of arsenic triselinide and poly(ether imide) [17].

SERRS in the core of a HC-MOF has not been demonstrated to date, however collection of a SERRS signal from the core of a liquid filled planar waveguide [18] has been obtained, although comparison with a bulk measurement which is indicative of the signal enhancement due to the waveguide was not made. HC-MOFs have, however, been shown to significantly enhance stimulated Raman scattering [19]. We note that Yan, et al., [20] have coated a HC-MOF with rhodamine/silver nanoparticles for SERRS. However, in this case the fiber did not guide at the excitation or scattered wavelengths and so the fiber served merely as a platform on which to deposit the sample. In this work, we demonstrate for the first time, to the best of our knowledge, that the SERRS signal obtained from a solution inside a HC-MOF exceeds that of the equivalent free space system.

Apart from capture efficiency, the other significant advantage that HC-MOFs provide is that the long interaction length allows the signals from particles along the entire length of the fiber to be collected in a single scan, automatically giving an averaged signal from both hot-spots and other particles and thereby improving scan-to-scan reproducibility. This provides a stark contrast to the often used system of scanning nanoparticles deposited on a solid substrate in a “hit-or-miss” exercise. Other advantages of using optical fibers as spectroscopy cells include the fact that the pump light is guided and tightly confined to the sample. This allows for higher intensities over longer lengths, compared with the free space alternative, and hence a reduction in the amount of sample required. It also reduces the need for high power lasers. Finally optical fibers are light, flexible and portable, and polymer optical fibers are suitable for in vivo applications.

2. Experimental

Rhodamine 6G (R6G) adsorbed onto silver nanoparticles is one of the most extensively studied SERRS systems and was chosen for this proof-of-concept experiment. The nanoparticles were fabricated by the citrate reduction method, and the suspension of nanometer sized particles, or sol, exhibited a wide absorption band indicating a large range of particle sizes. This was confirmed by a TEM image of the particles, which show both a large size and shape distribution, as seen in Fig. 1. The particles ranged in diameter from about 30–80 nm, and were in the form of spheres and rods. The sol had strongest absorption between 380–470 nm.

 

Fig. 1. Silver nanoparticles used in this experiment. TEM taken by Tich Lam Nguyen

Download Full Size | PDF

 

Fig. 2. SEM of the HC-MOF with a kagome lattice cladding used in this experiment.

Download Full Size | PDF

Rhodamine 6G (R6G), 95 %, and sodium chloride 99.8 % (NaCl) were purchased from Sigma-Aldrich and used without further purification. NaCl is used to activate the R6G/nanopaticle mixture [21]. An aqueous solution of R6G, silver nanoparticles and NaCl was made using milli-Q water. The concentrations of the components in the final solution were as follows: R6G 210 nM, silver nanoparticles 5.1 × 10^-11 M, NaCl 2.19 nM, where the nanoparticle concentration was estimated using the volume of a silver ion and the average diameter, assuming all particles were spheres. By volume, silver occupies less than 0.001 % of the solution, so we assume that the refractive index of the solution is not changed from that of water.

The fiber used in this experiment was a 3-ring HC-MOF with a kagome lattice cladding, which was fabricated from PMMA [22]. This fiber is shown in Fig. 2. For further discussions of the kagome fiber see [22, 23]. These fibers have broad transmission windows and are hence well suited to in-fiber spectroscopy applications. The fiber used in this experiment was first filled with mili-Q water using a pressure cell with an optical window, and a transmission spectrum using a supercontimuum source was recorded. Wavelengths between 480–840 nm were guided in the aqueous core. The fiber used was approximately 20 cm long, had a core size of 30 μm.

The water was removed from the fiber sample, and air was pumped through to dry it. Using the pressure cell, the sample solution was then pumped through the fiber. Filling time was of the order of a few minutes. Filling with too high a pressure sometimes resulted in damage to the delicate kagome structure, where the polymer struts in the cladding are less than 1 μm thick. With one end of the fiber secured in the pressure cell, which was slowly but continuously pumping solution, the other end was positioned along the length of a 5 cm mount. A glass coverslip was attached to the end of this mount, and the fiber endface was brought almost in contact with the coverslip. A droplet of the solution bridged the small gap between the fiber endface and the coverslip at all times during the experiment, ensuring consistent launch conditions. This mount was then placed under a microscope attached to an In Via Renishaw Raman spectrometer. After this, the fiber was bent sharply, and therefore we estimate the effective guiding length of the fiber to be 5 cm.

The 514.5 nm line of the Argon-ion laser was chosen as the excitation wavelength in this experiment. This wavelength is absorbed by both the nanoparticles and R6G itself, giving rise to resonance Raman signals. The laser light was launched into the core of the fiber through a 20x microscope objective with an N.A. of 0.4. The power of the excitation light at the foremost glass coverslip was only 0.13 mW. The back-scattered Raman signal was collected through the same microscope objective, and detected by a Rencam CCD camera after passing through a notch filter to remove the Rayleigh scattered light. The microscope was used in normal, rather than confocal, mode.

The Renishaw InVia Raman spectrometer allows one to view the sample using white light launched through the same objective as the laser light. Hence, it was possible to view the fiber endface, and accurately determine where the laser light was being launched. Three different measurements were made. The first was the intensity of the SERS signal obtained from the fiber when the pump light was focused into the core of the fiber, and the second was when the pump light was focused randomly in the cladding. The final measurement was made in a free-space geometry with the same solution in a cuvette. The plastic cuvette had a glass coverslip on top, and the solution in the cuvette was in contact with this coverslip, ensuring that the launch conditions in all cases were equivalent. Each spectrum acquired was the average of 50 scans of 10 s duration each, i.e. 500 s total integration time. Shorter integration times are possible with higher pump intensities; however the CCD camera used in this experiment saturated at even low intensities. No intensity reduction of the SERS signal was noted between the scans, indicating that photodegradation was not occurring in the sample. A detailed spectrum from a more concentrated solution with the same component ratios as the solution used in the experiment was also obtained for reference using a free-space geometry.

3. Results and discussion

The sample contained a slight excess of R6G compared to the nanoparticle concentration, resulting in a large fluorescent background in all signals obtained. Figure 3. shows the SERRS spectra obtained from the fiber core, fiber cladding and free-space measurements after fluorescence baseline removal. Apart from this correction, no further processing of the data was carried out. The curves have been separated vertically for clarity, but have not been scaled relative to one another. All three signals display a relatively high noise level due to the low concentrations and laser powers used. The signal obtained from the core of the fiber displays far better resolved Raman peaks than that from either the cladding or the free space cuvette measurement. In the case of the cladding measurement, the signal is barely resolvable above the noise of the system, whereas no signal can be distinguished in the free space system.

 

Fig. 3. SERRS spectra obtained from the fiber core (top), fiber cladding (middle) and free space (bottom) geometries.

Download Full Size | PDF

The wavenumber position of the Raman peaks in the fiber core spectrum agree well with literature values [24, 25], as well as the peaks obtained in the reference measurement. The largest discrepancy between literature and measured values was only 3 cm^-1.

The enhanced signal from the core of the fiber, when compared to the free space measurement, is attributed to the guidance of scattered photons back to the detector via the core of the fiber since both pump and Raman shifted wavelengths are guided by this solution filled fiber. The intensity and definition of the Raman peaks is clearly superior to those obtained from either the cladding or cuvette, leading us to conclude that the fiber is not merely acting as substrate or container for the SERRS particles.

The spectrum obtained from the fiber cladding displays weak Raman peaks, indicating that the cladding also gives rise to some signal enhancement when compared to a free space measurement. This is because there is always a small amount of resonant reflection leading to weak guidance in a hollow capillary [26], as well as antiguiding.

There are several characteristics of the HC-MOF that can be manipulated to yield a larger SERRS signal. The first is that a fiber fabricated from a material with higher refractive index then that of PMMA can be used to collect a greater portion of light emitted in the transverse direction, and hence guide more light to the detector. A fiber with a more uniform structure and a greater number of rings of holes would reduce the loss of guided modes thereby increasing the measured signal.

Finally, the signal obtained in our experiments was collected through only 5 cm of fiber, since only 5 cm of the total 20 cm length was held straight on the mount under the microscope. Based on the promising results obtained from this short piece of fiber we believe that the use of longer pieces would lead to extremely good signal enhancement.

The effect of the loss of the sol is also an important consideration in the assessment of liquid filled HC-MOF for SERS. If the liquid in the core of the fiber is highly absorbing, then the backscattered intensity through the fiber will be small and little or no advantage is gained by using the fiber geometry. However, to detect trace quantities in dilute aqueous solution the liquid core optical fiber is certainly useful because the transparency of the solution allows for confinement over longer path lengths than in free space.

A final interesting point to note is that when obtaining the cuvette measurements, the scan-to-scan reproducibility was low. A high SERRS signal would often coincide with an aggregate ‘floating’ across the beam path and then several scans with a minimal Raman signal would ensue. In the fiber case, however, there was no significant difference between one scan and the next. Further work will aim to quantify the scan-to-scan reproducibility in each case, as higher reproducibility allows SERRS to be used as a quantitative analytical tool even at low concentrations.

4. Conclusions

SERRS of R6G adsorbed on colloidal silver generated in the core of a HC-MOF has resulted in a significant increase of the detection limit of this analyte when compared with equivalent free-space geometry. It was shown that the HC-MOF guides both pump and Raman scattered wavelengths when filled with water. Hence, the increased signal from the HC-MOF is attributed to both better confinement of the pump beam to the sample, as well as guidance of scattered photons to the detector. This is the first time, to the best of our knowledge, that SERRS signals have been enhanced by guidance in a HC-MOF. By demonstrating the feasibility of SERRS in the core of a HC-MOF, we have taken the first step in realizing the unparalleled collection efficiency of such scattering events occurring in the core of the HC-MOF. To this end, further work by us and other groups should focus on the use of higher contrast microstructures.