Surface enhanced Raman spectroscopy (SERS) has been widely exploited in the fields of chemistry ^[1], biology ^[2], materials ^[3], superresolution imaging ^[4], etc., due to possessing the ability of significantly improving the detection limit of Raman spectroscopy, which is called the fingerprint of molecules. Based on the localized surface plasmon resonance (LSPR) effect ^[5–7], the great enhancement of the electromagnetic (EM) field in the vicinity of metallic nanostructures dominates the SERS mechanism ^[8–10]. By optimizing the surface morphology of metallic nanostructures and illumination methods, SERS sensitivity can reach even the single molecule level ^[11–13].

Up to now, many metallic nanostructures, i.e., nanospheres ^[14], nanorods ^[15], and nanocubes ^[16], have been employed as SERS substrates ^[17–19]. Nanoparticles (NPs) are some of the most commonly used nanostructures because of their excellent activity and easy preparation ^[17]. The NPs-based SERS substrates, such as dimers ^[20], trimers ^[21], or NPs-on-film structures ^[22], have attracted wide attention for presenting strong electric field enhancement induced by the gap mode ^[23] in nanogaps between adjacent NPs or between NPs and a film. Importantly, when NPs-based structures are deposited on the surface of a tapered fiber ^[24–27], an SERS fiber probe is formed and can be used to increase the sensitivity of Raman examination with the advantages of high efficiency, remote examination capability, and cost effectiveness. Different geometries and structures ^[28,29] of fiber probes have been proposed to optimize SERS performance. However, there lacks research on different modes of in-fiber excitation light, which would be an important breakthrough to enhance the SERS signal due to dependence of excitation of localized surface plasmon modes on the polarization of light ^[30,31].

In free space, the spatially structured light field has been adopted to improve metallic nanostructure-based Raman scattering intensity due to the great influences of illumination methods on the excitation of localized surface plasmon modes ^[32,33]. An azimuthal vector beam (AVB) is a typical structured light field in which the polarization direction is tangential on the cross section of the beam ^[34]. The AVB has been used in tip-enhanced Raman spectroscopy (TERS) and shows considerable improvement when it excites the NP at the apex of the atomic force microscope (AFM) tip ^[35,36]. However, in fiber, the AVB has not been used in enhancing Raman spectrum detection due to difficulty in generating the AVB in fiber caused by the boundary condition constraint. Considering the special polarization distribution of the AVB, it is promising to design an active NPs-based plasmonic tip for AVB excitation to enhance fiber SERS sensitivity and systematic research to optimize and understand that the AVB-enhanced fiber SERS is necessary for high-sensitivity detection.

In addition, the fiber probes are usually made of multimode fiber ^[37] with thousands of vector modes in the core ^[38]. The superposition among so many vector modes leads to a very complex and uncertain polarization distribution and low mode purity, which makes it impossible to accurately explain the electric field enhancement mechanism. Thus, it is necessary to guarantee the high mode purity of excitation light in fiber.

In this paper, an SERS fiber probe is fabricated by using a few-mode fiber (FMF) internally excited via an AVB, which is directly generated in the FMF by using an acoustically induced fiber grating (AIFG). The theoretical analysis shows that the gap mode can be effectively and uniformly generated on the surface of the Ag-NPs-coated fiber tip excited via an AVB. In the experiment, we have achieved approximately eight times stronger Raman spectrum via an AVB than that of a linearly polarized beam (LPB) with excellent time stability and repeatability, where malachite green (MG) is used as the target analyte. The detection sensitivity under AVB excitation can reach 10−11M(1 M=1 mol/L), which cannot be achieved by LPB excitation.

The commercial software MODE Solutions was used to calculate the dispersion curves of the LPB (HE11x/y) and AVB (TE01) as the radius of the bare tapered fiber decreases, as shown in Fig. 1(c). The energy distributions of the LPB and AVB along the Ag-film-coated tapered fiber without NPs deposited on the film surface, as shown in Figs. 1(d) and 1(e), respectively, were simulated using the three-dimensional finite-difference time domain (3D FDTD, Lumerical) method. Furthermore, the electric intensity distributions of the local surface plasmon modes on the surface of the Ag-NPs-coated fiber probe, as shown in Figs. 2(d)–2(f), were also simulated with the 3D FDTD method.

The SERS fiber tip was fabricated by a step-index FMF, with a core radius of 4.5μm and a cladding radius of 62.5μm, in the visible band. The heating–pulling method was used to prepare the tapered fiber ^[39]. The photochemical deposition method ^[40] was adopted to deposit the Ag NPs onto the surface of the tapered fiber. The bare tapered fiber was vertically immersed into the reaction mixture while being internally illuminated by a laser beam. The reaction solution was a mixture including silver ammonia solution (0.02 g/mL AgNO3 solution and a moderate amount of ammonia water) and potassium sodium tartrate (0.1 g/mL) with a volume ratio of 1:1. The laser power was kept at 500 μW during the deposition process of 30 min. The shape of the tapered fiber was characterized by an optical microscope. The surface topography of the Ag-NPs-coated fiber tip was characterized by a scanning electron microscope (SEM) [Fig. 3(c)]. Energy dispersive spectroscopy (EDS) was used to analyze the chemical component of the fabricated plasmonic tip.

The AIFG ^[41–43] was adopted to generate the AVB in the FMF. The laser beam (He–Ne laser, 632.8 nm, CW, 20 mW) was coupled into the FMF using a 25× micro-objective. The polarization orientation of the laser beam was adjusted through a half-wave plate. A mode stripper was used to purify the light to the LPB in the core by filtering out the high-order modes through critical bending loss. The diameter of the unjacketed FMF for forming the AIFG was etched to 30 μm by hydrofluoric acid to adjust the resonance wavelength and increase the overlap between the acoustic and optical waves, thus increasing the coupling efficiency of the AIFG within the 50 mm long etched segment. An acoustic transducer was used to generate an acoustic flexural wave to induce the AIFG. Based on the polarized dependence vector mode coupling characteristics, the AIFG could convert the LPB to an AVB, and the mode intensity distribution of the AVB was recorded by a CCD.

The Ag-NPs-coated fiber tip was immersed in the MG (dispersed in alcohol) solutions for 5 min and then taken out for natural air drying. The in-fiber AVB was used to excite Raman scattering of MG adsorbed at the apex of the tip. A laser line filter was put after the He–Ne laser beam to eliminate the side modes. The sketch map of the Ag-NPs-coated fiber probe internally excited via the AVB for SERS activity examination is shown in Fig. 4(a), in which the SERS signal was collected via a micro objective (MO; 40×) and then coupled into a spectrometer (Andor SR-500i-B1-R-1F1).

Figure 3(a) is the optical microscope image of the tapered fiber fabricated by the heating–pulling method ^[39]. Ag NPs, as an excellent plasmonic material in the visible band, are transferred onto the surface of the tapered fiber by using a laser-induced chemical deposition method ^[40]. The optical microscope image of the Ag-NPs-coated tapered fiber is shown in Fig. 3(b). Figure 3(c) is the SEM image of the tip area of the fabricated SERS fiber probe. The partial enlargement of the surface of the Ag-NPs-coated fiber tip is shown in Fig. 3(d); note that an Ag film is coated on the surface of the tapered fiber, and the Ag NPs, with radius of ∼15nm, are deposited on the surface of the silver film. In addition, according to the EDS shown in Fig. 3(e), the chemical component of NPs coated at the surface of the tip is silver.

Figure 1(a) is the sketch of the Ag-NPs-coated fiber probe established based on the morphology characterization depicted in Fig. 3. In order to simplify the theoretical analysis, on a cross section of the tapered fiber region, it can be considered that there is an Ag film on the surface of the tapered fiber, and Ag NPs with radius R≈15nm are distributed on the surface of the Ag film, as shown in Fig. 1(b). Figure 1(c) depicts the dispersion curve of the two lowest-order vector modes (HE11x/y, TE01) with the decrease in the fiber radius at a wavelength of 632.8 nm. Note that the effective refractive index of HE11x/y is larger than that of TE01, and two vector modes are, respectively, cut off at R=75nm (HE11x/y) and 225 nm (TE01), revealing that the energy distribution of HE11x/y is more concentrated than that of TE01, and the cutoff radius of HE11x/y is smaller than that of TE01. Furthermore, the energy distributions of the HE11x/y and TE01 modes along the Ag-film-coated tapered fiber are also simulated at a wavelength of 632.8 nm, as, respectively, shown in Figs. 1(d) and 1(e), which is coincident with the theoretical analysis depicted in Fig. 1(c).

To compare the electric field enhancement characteristic of the local surface plasmon modes excited by HE11x/y and TE01 modes, the cross section of the Ag-NPs-coated fiber probe with a tapered fiber radius of R=250nm is selected because the HE11x/y and TE01 modes are not cut off at this position, as shown in Fig. 1(c). Figures 2(a)–2(c) show the electric field enhancement characteristics of the local surface plasmonic modes on the surface of the Au-NPs-coated fiber probe calculated in three cases, because HE11x/y is a strictly double-degenerated vector mode including HE11x and HE11y with orthogonally linearly polarized directions, and TE01 has azimuthal polarization distribution. In the case of HE11x excitation, the gap mode can be excited between two adjacent Ag NPs with a nanoscale of g=2nm, as shown in Fig. 2(d), when the linear polarization direction of HE11x is approximately parallel to the center line of two adjacent Au NPs. As exhibited in Fig. 2(e), the surface and gap modes can be excited simultaneously with the linear polarization direction of HE11y approximately perpendicular to the center line of two adjacent Au NPs, but the electric field enhancement factor is smaller than that obtained by parallel excitation, as shown in Fig. 2(d). Based on the simulation results shown in Figs. 2(d) and 2(e), the gap mode can possess a large electric field enhancement factor only when the linearly polarized excitation light has a polarization direction parallel to the center line of two adjacent Au NPs. In other words, the gap mode with an optimized electric field enhancement factor cannot be effectively excited by the linearly polarized modes (HE11) at any position of the cross section of the Ag-NPs-coated fiber probe.

Furthermore, the electric field intensity distribution between two adjacent Ag NPs excited by TE01 mode is shown in Fig. 2(f). The gap mode between any two adjacent Ag NPs on a cross section of the Ag-NPs-coated fiber probe can be effectively excited by the TE01 mode due to possessing the azimuthally polarized distribution, leading to significant enhancement compared with HE11, even though the gap mode is not increased compared with that shown in Fig. 2(d). In other words, TE01 mode can excite the gap mode with the uniform electric field enhancement factor on the cross section of the Ag-NPs-coated fiber probe, but only the HE11 mode with linear polarization direction parallel to the center line of two adjacent Ag NPs can excite the gap mode with an electric field intensity equal to TE01 excitation. Therefore, the Raman signal intensity obtained by TE01 mode excitation is larger than that by HE11 mode excitation, because exciting the Ag-NPs-coated fiber probe via the TE01 mode can obtain a larger SERS active surface than by HE11 mode excitation.

The sketch of the experimental setup for SERS detection using the Ag-NPs-coated fiber probe internally excited via an AVB is shown in Fig. 4(a), in which the SERS signal was collected via an MO (40×) and then coupled into a spectrometer (Andor SR-500i-B1-R-1F1). In this experiment, the AVB is directly generated in an FMF via an AIFG reported in our previous work ^[41–43]. Based on the polarized dependence vector mode coupling characteristic of the AIFG, the linearly polarized fundamental vector mode HE11y, as shown in Fig. 4(b), is coupled as the azimuthal vector mode (TE01), whose transverse mode intensity distribution is depicted in Fig. 4(c), and the polarization distribution examination results are shown in Figs. 4(d)–4(g), respectively.

After the generation of the TE01 mode in the FMF via the AIFG, an Ag-NPs-coated fiber probe, as shown in Fig. 3(b), is fabricated at the output terminal of the FMF. As a widely accepted target analyte, Raman spectra of MG with concentrations 10−5M adsorbed on the surface of the Au-NPs-coated fiber probe are internally excited using TE01 and HE11 modes to reveal the electric field enhancement characteristic. As shown in Fig. 4(h), the red and black curves are the Raman spectra excited using TE01 and HE11 modes. The excitation power is 3.5 mW measured from the output terminal before the fiber-tapering process, and the integration time is 10 s. Note that Raman spectrum intensity under TE01 excitation is approximately eight times stronger than that of HE11 excitation, revealing that exciting the Ag-NPs-coated fiber probe via TE01 can obtain a larger SERS active surface than that of HE11 mode excitation, and this experiment result verifies the rationality of the theoretical analysis, as shown in Fig. 2. The magnification (∼8×) is better than that of other AVB excited Raman detection systems (∼3×) ^[35].

Furthermore, to reveal the SERS activity of the fabricated Au-NPs-coated fiber probe, the Raman spectra of MG with lower concentrations (10−9M, 10−10M, and 10−11M) adsorbed on three Ag-NPs-coated fiber probes are, respectively, measured, and the corresponding Raman spectra of MG with different concentrations are shown in Fig. 4(i). Note that all the characteristic peaks of MG can be distinguished clearly even with a concentration down to 10−11M under TE01 mode excitation, revealing the high SERS activity of the fabricated Ag-NPs-coated fiber probe excited by the TE01 mode. For comparison, the Raman spectrum of 10−11M MG solution adsorbed on the surface of the Ag-NPs-coated fiber probe is also measured under HE11 excitation, as shown in the pink curve in Fig. 4(i). Note that no SERS signal is detected under HE11 mode excitation.

The time stability of the Ag-NPs-coated fiber probe is also examined. As shown in Fig. 5(a), the red and black curves are the Raman spectra excited using TE01 and HE11 modes. After that, an Ag-NPs-coated fiber probe is stored in a lab environment for 3 h; note that the Raman spectrum intensity under TE01 mode excitation is still approximately eight times stronger than that of HE11 mode excitation. Figure 5(b) shows the Raman spectra as a function of storage time, recorded at 10 min intervals for 60 min. Note that all characteristic peaks can be distinguished clearly after preservation for 3 h, even though there exists slight intensity fluctuation. The time variation of the intensity of the 1613.5cm−1 band is shown in Fig. 5(c). Note that the SERS signal is quite stable for 60 min. The standard deviation (SD) and the corresponding relative SD (RSD) are calculated to be 182.06 and 5.8%, respectively, revealing excellent time stability.

Moreover, three Ag-NPs-coated fiber probes were fabricated to examine the repeatability of the Ag-NPs-coated fiber probe. Raman spectra of MG (10−5M) absorbed on the surface of three Ag-NPs-coated fiber probes excited with HE11 and TE01 modes are, respectively, shown in Figs. 6(a)–6(c). The variation of the magnification calculated with the intensity of the 1613.5cm−1 band of the three SERS fiber probes, respectively, excited using TE01 and HE11 modes is shown in Fig. 6(c). The SD and the corresponding RSD from the three SERS fiber probes are calculated to be 0.32 and 4.4% ^[22], revealing excellent repeatability.

In conclusion, a method was presented for highly sensitive SERS detection based on a plasmonic tip internally excited via an AVB. Theoretical analysis shows that the Ag-NPs-coated fiber probe internally excited via the TE01 mode can provide a larger SERS active surface than that of HE11 mode excitation. The experimental result shows that the intensity of the Raman spectrum of MG by internally exciting the Ag-NPs-coated fiber tip via an AVB is about eight times as strong as that of LPB excitation, thereby verifying the rationality of the theoretical analysis. The detection sensitivity under AVB excitation can reach 10−11M, which cannot be achieved by LPB excitation. The time stability and reliability are determined to be excellent. This work gives an effective and reliable way to further improve SERS fiber probe technology.