1. Introduction

Surface-enhanced Raman spectroscopy (SERS) [1] is a powerful tool for high-sensitivity detection applications in many fields such as environmental science [2], food safety [3], and biomedicine [4], because it can greatly enhance the “fingerprints” Raman spectral intensity of molecular by exploiting the localized field enhancement effect of noble metal nanoparticles [5]. The SERS fiber probes [6] formed by preparing noble metal nanoparticles on optical multimode fibers can not only simplify the optical design of SERS spectral instrument and improve the excitation and collection efficiencies, but are also expected to realize long-range SERS measurements [7]. So far, various SERS fiber probes have been developed from the original flat or inclined facet [8–10] and D-shaped fiber probes [11], to tapered, combined tapered and cavity-enhanced tapered fiber probes [12–14], as well as the probes based on hollow-core photonic crystal fibers (HC-PCFs) [15]. Even microfluidics techniques have also been used to construct the SERS fiber probes with specific properties [16].

For the tapered SERS fiber probe, by optimizing its cone angle, nanoparticles fabricated on the cone surface can be excited by the guided-wave modes of the excitation laser [17]. Due to the relatively low field intensity of the guided-wave mode distributed on the surface, it is possible to increase the excitation laser power without triggering photothermal damage to the nanoparticles. As a result, nanoparticles in a larger area on the cone surface can be effectively excited, thereby improving the SERS detection sensitivity. In addition, compared with the SERS probes based on HC-PCFs and microfluidic technique, tapered fiber probes also have the advantages of compatibility with standard fibers and simple fabrication process, thus attracting much attention in the past decade. To date, numerous methods such as evaporation sputtering [18], self-assembly [19], laser induced chemical deposition [20], and light induced evaporation deposition [21,22] have been developed to deposit noble metal nanoparticles on the cone fiber surface. Additionally, by optimizing the morphology and size of the nanoparticles, the tapered fiber probe has already exhibited a higher SERS detection sensitivity than that of ordinary substrate [23].

The high SERS detection sensitivity of the tapered fiber probe stems from the relatively low intensity of surface guided-wave of excitation laser, which allows the excitation laser power to be increased to a high level under the premise of avoiding photothermal damage to the nanoparticles. Obviously, the field distributions of the surface guided-wave for the SERS excitation on the tapered fiber probe depend on its cone angle. Previously, with the laser-induced method we experimentally confirmed that the tapered fiber probe has an optimal cone angle for the best SERS detection sensitivity [24]. Unfortunately, when the tapered SERS fiber probe is fabricated by the laser-induced method, the thickness of the nanoparticles on the cone surface depends on the field distribution of the induced laser on the cone surface. This makes it difficult to accurately design the optimal cone angle. To this end, we used the electrostatic adsorption method to prepare a single layer of cubic silver nanoparticles on the surface of the fiber taper, and found that the tapered fiber probe has the optimal cone angle for the best SERS detection sensitivity [25]. Recently, researchers have begun to design such optimal cone angles through theoretical simulations for the tapered SERS fiber probes [23,26].

All the previous results have shown that the SERS excitation efficiency and the detection sensitivity can be effectively improved by optimizing the cone angle of the fiber probe. However, an open question still lies in how to design the optimal cone angle. Although an ideal bare fiber taper only with evanescent wave distribution on the cone surface can be designed according to the adiabatic coupled mode theory [27], the actual tapered SERS fiber probe is wrapped with nanoparticles, which makes the optical field distribution on the surface extremely complicated. It not only includes the guided-wave modes confined in the cone fiber and distributed in the nanoparticle layer, but also involves many effects such as mode cutoff, mode coupling, and mode radiation when the excitation laser propagates along the tapered fiber, and even includes the transmission loss in the metal nanoparticles. All these effects ultimately complicate the physical mechanisms involved in the SERS excitation of tapered fiber probes, making it difficult to design the optimal cone angle, restricting the improvement of SERS detection sensitivity of tapered fiber probes.

In this paper, we use the electrostatic adsorption self-assembly method to prepare a single-layer uniform gold spherical nanoparticle (GSN) tapered fiber probe, so that the effect of the cone angle on the SERS detection sensitivity of the probe may be investigated conveniently. We reveal that the photothermal damage is the main limiting factor for the detection sensitivity of tapered fiber probes. Based on the equivalent composite dielectric (ECD) model, the relationship between the photothermal damage and the cone angle of the fiber probe is analyzed with the fundamental principles of geometric and guided wave optics. From this, we clarify that only when the cone angle is optimized, one can increase the excitation laser power to a higher level without triggering the photothermal damage, so that the interaction area between the excitation laser and the GSNs on the cone surface increases, and finally achieving a higher SERS detection sensitivity.

2. Fabrication of tapered SERS fiber probes

The multimode fiber used to prepare the tapered probe is a common silica fiber with a uniform core refractive index, a core diameter of 200 µm, and an NA of 0.22. The preparation of the fiber probe can be divided into two steps: preparation of bare fiber taper and growth of nanoparticles on its surface. The bare fiber taper is prepared by etching the fiber with hydrofluoric acid solution combined with pulling by a pulling mechanism [24]. The concentration of hydrofluoric acid solution is 40%, and the liquid level of hydrofluoric acid solution is covered with methyl silicone oil to prevent the volatilization of hydrofluoric acid. By changing the pulling speed, the bare fiber tapers with different cone angles can be prepared. Using this preparation process, we prepare a batch of bare fiber tapers with cone angles in the range of 5.2-19.2°. They are placed in pure water first and then ethanol for 30-min ultrasonic cleaning, and then stored for subsequent use. Next, using the electrostatic adsorption self-assembly method [19], the GSNs are grown on the surface of the bare fiber taper. The GSNs used were purchased from NanoSeedz, and the diameter was chosen to be 70 nm. Although there is a certain deviation between the LSPR peak of the 70-nm GSN and 785nm that we chose for the excitation light [28], the 70nm GSNs can also obtain good SERS enhancement (see the test results in the 4th section below), and it is easier to prepare a single-layer uniform GSN on the surface of the fiber taper by the electrostatic adsorption self-assembly method. Thus, we chose the diameter to be 70nm. In order to ensure that the GSNs can be uniformly adsorbed on the cone surface in a monolayer, the bare fiber taper is first soaked in piranha solution (70% H₂SO₄ and 30% H₂O₂) for 1 hour, so that the SiO₂ molecules on the surface of the bare fiber taper are uniformly hydroxylated. Then, the fiber taper is taken out and immersed in carboxyethylsilanetriol sodium salt (25wt.% solution in H₂O) solution (20 µL/mL) for another hour to achieve uniform and negatively charged carboxyl group distribution on the cone surface through the condensation reaction between hydroxyl groups [25]. Finally, the fiber tapers with uniform carboxyl groups on the surface are placed in the sol of positively charged GSNs coated by CTAC. With the help of the electrostatic adsorption effect [25], GSNs can be uniformly self-assembled to the surface of the fiber taper to form the monolayer nanoparticle tapered fiber probe. By changing the electrostatic adsorption time, tapered fiber probes with different nanoparticle densities can also be prepared [26].

Scanning electron microscope (SEM, Quanta 200FEG) observations of the prepared SERS fiber probes with different cone angles show that the GSNs are uniformly distributed in a single layer on the cone surface. Figure 1 shows the photographs of the GSN distributions of different regions on the cone surfaces with cone angles 5.2° and 16.9°, prepared by 2-hr electrostatic adsorption. It can be clearly seen that the GSNs show a good monolayer uniform distribution in all regions of the cone surface. Figure 2 shows the SEM images of the GSN distributions on the cone surface prepared under different electrostatic adsorption times. As seen, the longer the electrostatic adsorption time is, the higher the GSN density is.

 

Fig. 1. SEM images of the GSNs prepared on the cone surface of the tapered fiber probes. (a) and (e) Overall view for 5.2° and 16.9° fiber probes, respectively; (b)-(d) and (f)-(h) Zooms in the bottom, middle and tip of the fiber probes with 5.2° and 16.9° cone angles, respectively.

Download Full Size | PDF

 

Fig. 2. SEM images of the GSN distributions on the cone surface with different adsorption times for 13° tapered SERS fiber probe. (a), (b) and (c) Adsorption times are 1, 2, and 4 hours, respectively. The insets in each subfigure show their zoom up SEM images.

Download Full Size | PDF

3. Tests of tapered SERS fiber probes

We used a commercial portable Raman spectrometer (Mini-Ram, B&W TEK Opto-electronics, USA) to test the performance of the prepared tapered SERS fiber probe. The Raman spectrometer comes with a standard multimode fiber probe. The multimode fiber of the standard probe has the same parameters as those of the fiber used for the tapered fiber probe, and the wavelength of the excitation laser is 785 nm. The standard probe of the spectrometer is spatially collimated and focused on the output. It is first spatially coupled to a transitional fiber with a length of nearly 1 m. Then, the transitional fiber is fused with low loss to the tapered fiber probe to be measured. The fiber lengths of all tapered probes to be tested are about 15 cm. In this way, the excitation laser power can be calibrated during SERS fiber probe testing. The test sample is R6G solution with a concentration of 10⁻⁶M. The SERS spectrum of the fiber probe can be obtained by direct subtraction of the spectra measured with and without sample molecules. The spectrum measured without the sample molecule is the background spectrum contributed by the spectrometer probe pigtailed fiber, the transition fiber, and the tapered fiber probe to be tested, which can be measured by placing the tapered fiber probe in pure water. In actual operation, the fiber probe is immersed in pure water, and the background spectra under different excitation powers are measured first. Then the probe is taken out to dry and further immersed in the sample solution for measurements of the SERS spectra under the corresponding excitation powers. Thus, background correction can be performed on the measured SERS spectra for the fiber probes.

4. Results and discussions

We first measure the background-corrected SERS spectra of the fiber probes with 8 different cone angles in the range of 5.2°–19.2°, all prepared under 2-hr electrostatic adsorption. The results show that once the excitation power is greater than 6mW, the Raman characteristic peaks of R6G at 1312, 1363, and 1507cm^-1 [29] can be clearly measured for all fiber probes. Also, the amplitudes of these peaks increase with the excitation power. However, for the same excitation power, different spectral peak amplitudes are observed for different cone-angle fiber probes. Figure 3 shows the measured SERS spectra of fiber probes with cone angles of 5.2°, 13°, and 19.2° for different excitation powers, in which each spectrum is the average of the spectra measured for 5 fiber probes with the same cone angle. It can be seen that the spectral baselines are all near zero and relatively flattened, indicating that our background correction is effective. Moreover, the amplitudes of the characteristic peaks increase with the excitation laser power. In contrast, the characteristic peak amplitudes of the 13° fiber probe are obviously higher than those of 5.2° and 19.2°.

 

Fig. 3. Measured SERS spectra of fiber probes for different excitation powers, in which each spectrum is an average of the measured spectra for 5 fiber probes with the same cone angle. (a), (b) and (c) are for the fiber probes with cone angles of 5.2°, 13°, and 19.2°, respectively.

Download Full Size | PDF

Although the measured characteristic peak amplitudes of fiber probes with different cone angles all increase with the increase of excitation power, for any given cone angle, the SERS spectra are no longer stable if the excitation power exceeds a fixed value ${P_{{\rm{th}}}}$ . Moreover, if the excitation power continues to increase beyond ${P_{{\rm{th}}}}$, the measured characteristic peak amplitudes will decrease instead. Also, fiber probes with different cone angles give different ${P_{{\rm{th}}}}$. Figure 4 shows the average ${P_{{\rm{th}}}}$ of the fiber probes, measured by five fiber probes with the identical cone angle as a function of the cone angle. When the cone angle increases from 5.2° to 19.2°, ${P_{{\rm{th}}}}$ decreases approximately monotonically from 42 to 24mW. Figure 4 also shows the 1507cm^-1 peak amplitudes measured when the fiber probes are respectively excited with their own ${P_{{\rm{th}}}}$ as a function of the cone angle. As can be seen, the 1507cm^-1 peak amplitude measured for the 13° fiber probe is the highest while the corresponding ${P_{{\rm{th}}}}$ is not.

 

Fig. 4. Average ${P_{{\rm{th}}}}$ (black) and characteristic peak amplitudes at 1507cm^-1 (blue) excited with ${P_{{\rm{th}}}}$ of the fiber probes as a function of the cone angle, each average value is measured by five fiber probes with the same cone angle.

Download Full Size | PDF

When the excitation power exceeds ${P_{{\rm{th}}}}$ and stays there for more than about 1min, even if it is reduced back a value below ${P_{{\rm{th}}}}$, the measured spectral amplitude cannot be restored to the original ones. This suggests that the fiber probes may be irreversibly damaged once they have been excited with excitation powers over ${P_{{\rm{th}}}}$. To this end, we again characterize the surface topography of the fiber probes after the excitation power exceeds ${P_{{\rm{th}}}}$. The results show that, for a fiber probe with arbitrary taper angle, the surface morphology after the excitation power exceeds ${P_{{\rm{th}}}}$ will change significantly. The region on the cone surface where the GSN morphology changes depends on the cone angle of the fiber probe. As the cone angle becomes larger, the region where the topography changes move towards the bottom of the cone. Figure 5 shows the surface morphologies of the 5.2°, 13° and 19.2° fiber probes after the excitation power exceeds ${P_{{\rm{th}}}}$ for 2 mins. Clearly, the bottom surface of the 19.2° fiber probe no longer presents a uniform spherical nanostructure, and the morphology of GSNs has changed significantly, indicating that the GSNs on the bottom surface have truly been photothermally damaged [30]. For the 13° and 5.2° fiber probes, the damage regions are located near the middle and tip of the cone surface, respectively, with the damage weakened. Therefore, fiber probes have different photothermal damage threshold ${P_{{\rm{th}}}}$ with respect to their different cone angles. When excitation power is greater than ${P_{{\rm{th}}}}$, irreversible photothermal damage occurs on the cone surface, which makes the measured SERS spectra instable.

 

Fig. 5. SEM images of the cone surface after the fiber probe has been excited with an excitation power exceeding ${P_{{\rm{th}}}}$ for 2 mins. (a) the tip surface of the 5.2° fiber probes; (b) the middle surface of the 13° fiber probes; and (c) the bottom surface of the 19.2° fiber probe.

Download Full Size | PDF

In order to explain the reason why ${P_{{\rm{th}}}}$ and the photothermal damage region of the fiber probe changes with the cone angle, we regard the single layer of uniform metal GSNs on the cone surface as an ECD layer [31], and its thickness is rightly the diameter of a gold sphere, as shown in Fig. 6. Then, the complex refractive index of the ECD layer may be obtained as ${n_3} = {n^{\prime}_3} + i{n^{\prime\prime}_3}$ [32], where ${n^{\prime}_3}$ and ${n^{\prime\prime}_3}$ are the real and imaginary parts of the complex refractive index of the ECD layer, respectively. Thus, the critical angle of total reflection at the interface of the ECD layer on the cone surface, $\varphi$, can be determined [33]. Note that since total reflection actually occurs on both surfaces of the ECD layer [33], this results in the total reflection loss of the excitation laser being generally much smaller than the refraction loss from directly passing through the ECD layer. If assuming that the maximum incident angle of the meridional ray (for simplicity and without loss of generality, only meridional ray is considered here) determined by the probe fiber NA is $\theta _{\rm{1}}^{{\rm{in}}}$, and the incident angle of the meridional ray after the m^th total reflection in the fiber cone is $\theta _{\rm{m}}^{{\rm{in}}}$, then $\theta _{\rm{m}}^{{\rm{in}}} = \arcsin \left( {{{\sqrt {n_1^2 - N{A^2}} } \mathord{\left/ {\vphantom {{\sqrt {n_1^2 - N{A^2}} } {{n_1}}}} \right. } {{n_1}}}} \right) - \left( {2m - 1} \right)\alpha $ [34], where $\alpha$ is the half-cone angle. With the increase of the number of reflections m, $\theta _{\rm{m}}^{{\rm{in}}}$ gradually decreases. Therefore, only when all incident angles are greater than $\varphi$ can the input excitation laser reach the tip of the tapered fiber without refraction. Otherwise, as $\theta _{\rm{m}}^{{\rm{in}}}$ continues to decrease with m, the excitation laser may be refracted through the ECD layer somewhere on the cone surface. For large $\alpha$, which corresponds to a small $\theta _{\rm{1}}^{{\rm{in}}}$, when the excitation laser enters the tapered fiber and undergoes only a few total reflections, it may be refracted through the ECD layer. Also, while passing through the ECD layer, a large amount of heat may be generated due to relatively larger number of GSNs at the bottom of the fiber taper and the larger refraction loss of the ECD layer, resulting in photothermal damage on the bottom of the cone surface. This means that the allowable ${P_{{\rm{th}}}}$ is relatively low. With the decrease of $\alpha$, the number of total reflections in the tapered fiber for the excitation laser increases and due to the relatively low total reflection loss no photothermal damage occurs in this process, so that the longitudinal location where the excitation laser may refract through the ECD layer, e.g., the photothermal damage region, moves toward the tip of the tapered fiber. Due to the increase of the total reflection times, the excitation laser power at the light refraction has already been reduced by the total reflection loss that is less than the refraction loss, resulting in the increase of the allowable ${P_{{\rm{th}}}}$ . It is because of the above reasons that the damage threshold ${P_{{\rm{th}}}}$ decreases almost monotonically as the cone angle increases, as shown in Fig. 4.

 

Fig. 6. Schematic illustration of meridional light transmission in a tapered fiber wrapped with an equivalent composite dielectric layer. (b) Simulation of the electric field distribution and temperature distribution of the fiber probe with different taper angles under the same power of 785 nm excitation light, obtained with the FEM method. In the simulation, the excitation light power used is the same as 24 mW, which is a uniform plane wave transmitted along the x-direction and polarized in the y-direction. The diameter of the tapered optical fiber is 20 µm; The metal nanoparticles and the distance between the two particles are selected as 70 nm and 100 nm respectively, and the distance from the surface of the optical fiber is 1 nm, and the external medium is air. In (b), the overall electric field distribution of the 20°, 13°, and 5° fiber probes are shown in insets (i)-(iii), and the temperature distributions are shown in insets (iv)-(vi), respectively. In order to analyze the heat generation behavior of the excitation light in the fiber probe, three different positions of the fiber probe (10%, 50%, and 90% of the longitudinal length of the fiber probe) were obtained by using the simulated overall electric field distribution. The local electric field distribution is further used to simulate the local temperature distributions (vii)-(ix) at 10%, 50%, and 90% longitudinal positions of the fiber probe with different taper angles, which correspond to 20°, 13° and 5° fiber optic probes. Insets (x) and (xi) also give details of the local electric field distribution and local temperature distribution of the 20° probe, respectively

Download Full Size | PDF

Using guided-wave optics and its coupled-mode theory [27], the same conclusion as the above may be obtained. When the excitation laser is transmitted into a tapered multimode fiber wrapped with the ECD layer, a series of eigenmodes can be excited. When these eigenmodes propagate along the taper fiber, the optical field distributed in the ECD layer experiences transmission loss, and the higher-order guided modes are gradually cut off due to the decreasing radius of the fiber taper. If it is a slowly-varying tapered fiber with a small $\alpha$ (the adiabatic approximation is satisfied), the guided modes tend to meet the phase matching conditions required by mode coupling, and the energies carried by them may be coupled to the low-order wave-guided modes when they are cut off, otherwise, the higher-order modes emit the energies carried in the form of radiation wave when they are cut off because the phase matching condition is not satisfied if $\alpha$ is large [27]. In this way, when $\alpha$ is large, after the excitation laser enters the tapered fiber, the higher-order modes tend to be cut off after transmitting a short distance due to the rapid reduction of the fiber taper diameter along the longitudinal direction. Because the phase matching conditions are not met, these higher-order modes radiate the energies they carry in the form of radiation waves, resulting in photothermal damage at the bottom of the tapered fiber probe. Thus, the effective interaction area between the excitation laser and the GSNs on the cone surface becomes small, that is, the number of GSNs involved in the SERS effect is relatively small, giving a low SERS detection sensitivity. As $\alpha$ decreases, the high-order modes tend to couple part of their energies into the lower-order modes that continues to transmit when they are cut off, so that the location where the radiation modes generate, e.g., the photothermal damage region, move from the bottom towards the tip of the cone fiber. Due to the mode transmission loss in the ECD layer, the excitation power is allowed to be increased to a higher level without triggering the photothermal damage, and the SERS detection sensitivity is also improved accordingly. With the further reduction of $\alpha$, the higher-order modes tend to satisfy the mode phase matching conditions at the cut-off, and may couple most of their energies to the lower-order modes, so that the position where the radiation modes generate will be near the tip of the cone fiber. Moreover, the energies of the radiation waves have been greatly reduced due to the transmission loss of the ECD layer, thereby allowing the excitation power to be increased to a higher level. However, in addition to the transmission loss in the ECD layer itself, the external surface wave of the ECD layer accounts for a large energy ratio of the excitation laser when $\alpha$ is too small [27], which may lead to serious molecular scattering when the excitation laser is transmitted over a relatively long distance in the cone fiber. Since this part of the scattered light is difficult to couple and collect, the SERS detection sensitivity is reduced as well.

In addition, we have adopted the FEM method to simulate the electric field distribution and temperature field distribution of the tapered fiber probes of 20°, 13°, and 5° under the excitation of the 785nm excitation of the same laser power [35]. Since the field distribution simulation of the 200 µm core diameter fiber probe is too difficult to calculate, our simulation focuses on revealing the effect of different taper angles on the field distribution of tapered probe. Thus, the field distribution of the fiber probe with a core diameter of 20 µm is simulated. The simulation results are shown in Fig. 6(b). It can be seen that for a fiber probe with a given taper angle, the electric field strength at different positions is different; the larger the cone angle is, the higher the electric field strength at the bottom of the cone is. However, for fiber probes with different taper angles, the overall electric field strength of the fiber probe gradually decreases with the decrease of the taper angle, and the simulation results are consistent with our previous analysis results. Although the magnitude of the fiber probe electric field at any given taper angle is different at different positions, the overall temperature distribution difference for a given fiber probe is very small due to the strong heat exchange between the GSNs on the fiber probe surface (see (iv)-(vi) in Fig. 6(b)). However, for fiber probes with different taper angles, under the same excitation light power, the temperature in different positions is obviously different. The larger the taper angle is, the higher the temperature is ((iv)-(vi) in Fig. 6(b)). Considering the overall temperature distribution of the different taper-angle fiber probes shown in Fig. 6(b) (iv)-(vi) cannot accurately reflect the heat generation in different parts of the fiber probe. We also use the overall stable electric field distribution obtained by the simulation to obtain the local electric field distribution at different positions of the fiber probe, and then use the local electric field distribution to simulate the corresponding local temperature distribution, and the results are shown in Figs. 6(b) (vii)-(ix). It can be seen that the temperature of the bottom of the cone of the 20° fiber probe, the middle of the cone of the 13° fiber probe, and the cone tip of the 5° fiber probe are all relatively high, so it is prone to generate photothermal damage in these parts, which is consistent with the experimental results given in Fig. 5. Therefore, these simulation results further intuitively confirm the correctness of our conclusions obtained through geometric optics and guided wave optics analysis. Therefore, according to our above analyses, when the cone angle is too large, light refraction or radiation waves are easily generated to cause photothermal damage at the bottom of the tapered fiber probe, which limits the increase of the excitation power, thereby restricting the increase of the effective interaction area between the excitation laser and the GSNs on the cone surface, i.e., the number of GSNs involved in the SERS process, finally giving a relatively low SERS detection sensitivity. When the cone angle is too small, although the excitation laser power can be increased to a higher level for increasing the interaction area with the GSNs on the cone surface (the number of GSNs involved in the SERS process) without triggering the photothermal damage, the transmission loss and the external scattering loss of the ECD layer also restricts the SERS detection sensitivity. This makes it possible to obtain the best SERS detection sensitivity only when the cone angle is moderate. For the fiber probe shown in Fig. 4, the best cone angle measured is 13°.

Obviously, the optimal cone angle of the tapered SERS fiber probe depends on the complex refractive index of the ECD layer. When the morphology, size and density of the GSNs are different, the complex refractive can be changed [31]. Thus, the optical refraction or radiation wave behavior of the excitation laser and its loss during the transmission are also changed. These changes determine the photothermal damage and the SERS detection sensitivity of tapered fiber probes.

Figure 7(a) shows the measured ${P_{{\rm{th}}}}$ for the two kinds of SERS fiber probes, prepared respectively with 2.5 and 1.5-hr electrostatic adsorptions for different nanoparticle densities, as a function of the cone angle. As seen, ${P_{{\rm{th}}}}$ for these two fiber probes are monotonically decreased with the increase of the cone angle, the same as that of the probe prepared with 2-h electrostatic adsorption. However, as the nanoparticle density increases, both the real and imaginary parts of the refractive index of the ECD layer increase [32]. The larger the real part of the refractive index, the stronger the confining ability to light waves, and the more difficult it is to generate light refraction or radiation waves; the larger the imaginary part of the refractive index, the higher both the total reflection and the transmission loss for the ECD layer [27]. Therefore, for fiber probes with the same taper angle, higher nanoparticle density fiber probes have higher ${P_{{\rm{th}}}}$. Figure 7(b) shows the amplitudes of the measured characteristic peaks of 1507 cm^-1 for the two density SERS fiber probes excited by their ${P_{{\rm{th}}}}$ as functions of the cone angle. It can be seen that the optimal cone angles still exist for the two-density SERS fiber probes. For the higher density fiber probe, the optimal taper angle is also around 13°. However, because the higher density leads to an increase in the number of nanoparticles involved in the SERS interaction, which gives a larger transmission loss in the ECD layer, the detected 1507cm^-1 peak amplitude is slightly lower than that with 2-h electrostatic adsorption. For the lower density fiber probe, both the total reflection and the transmission losses of the ECD layer are lower, so that the ${P_{{\rm{th}}}}$ is also decreased. This leads to a more favorable excitation of SERS when the cone angle of the fiber probe becomes smaller, giving an optimal cone angle of 11°. Moreover, due to the decrease of ${P_{{\rm{th}}}}$, the detected 1507cm^-1 peak amplitude is also lower than that with the 2-h electrostatic adsorption.

 

Fig. 7. (a) Measured ${P_{{\rm{th}}}}$ for the tapered SERS fiber probe as a function of its cone angle; (b) measured 1507cm^-1 characteristic peak amplitude of fiber probes excited with ${P_{{\rm{th}}}}$ as a function of the cone angle. Blue and black curves in the figures correspond to the fiber probes prepared by 2.5 and 1.5-hr electrostatic adsorption, respectively.

Download Full Size | PDF

It should be pointed out that to reveal the effect of the cone angle on the SERS detection sensitivity, we specially adopted the electrostatic adsorption self-assembly method to prepare the single-layer uniform GSNs on the surface of the tapered fiber probe. However, the variations of ${P_{{\rm{th}}}}$ and SERS detection sensitivity of the tapered fiber probe with the cone angle and the origin of the light refraction or radiation wave in the tapered fiber probe obtained by the ECD layer analysis are also applicable for other fiber probes with different morphology and scale of nanoparticles, thus, they may be a key step towards optimally designing the cone angle of the tapered fiber SERS probe.

5. Conclusion

Using the electrostatic adsorption self-assembly method, we have prepared a single-layer uniform GSN tapered fiber probe. The effects of the taper angle on the SERS detection sensitivity have been studied through the ECD layer model. When the cone angle is large, light refraction or radiation waves are easily generated at the bottom surface of the fiber probe, and excessively refracted light or radiation waves can photothermally damage the GSNs on the cone surface, limiting the increase of the excitation laser power. This effect also limits the interaction area between the excitation laser and the metal nanoparticles on the cone surface, resulting in relatively low SERS detection sensitivity. When the cone angle is small, although the excitation laser power can be increased for increasing the interaction area with the nanoparticles on the cone surface, due to the transmission loss and the external scattering loss of the ECD layer, it is difficult to improve the SERS detection sensitivity. The best SERS detection sensitivity can be obtained only when the cone angle is moderate. The optimal taper angle depends on the complex refractive index of the ECD layer composed of GSNs on the taper surface. The test results of three SERS fiber probes show that the optimal cone angle varies between 11-13° due to the difference in the complex refractive index of the ECD layer. This method of studying the optimal cone angle with the ECD layer may be expected to serve as a reference for the design and fabrication of high-sensitivity SERS fiber probes.