1. Introduction

Surface-enhanced Raman spectroscopy (SERS) is a powerful spectroscopic technique that can provide molecular or lattice vibrational fingerprint information [1,2]. It can be used for detection and identification of trace amounts of molecules with high specificity and sensitivity, even down to monolayer or single-molecule level [3,4]. The name itself emphasizes the key concept of amplification of Raman signals of trace molecules at surfaces and interfaces. There exist two generally accepted mechanisms for SERS enhancement: electromagnetic enhancement (electromagnetic field enhancement) and chemical enhancement (mainly charge transfer enhancement). The available results show that in the majority of cases, the former is much stronger than the latter [5]. In electromagnetic case, when the incident light wavelength matches with the resonance frequency of the plasma oscillations in the metal nanoparticle, free electrons in particle driven by laser would accumulate on the surface and lead to the charge density increasing significantly. As known as localized surface plasmon resonance (LSPR), this phenomenon would excite a spatially localized and strongly enhanced electromagnetic field called “hotspot”. The hotspots can in turn amplify the inelastic light-emission process, i.e. Raman scattering. The Raman signal of molecules at or in close vicinity of hotspots of nanoparticles can be dramatically amplified by an often-stated ${(E/{E_0})^4}$ enhancement approximation, where $E/{E_0}$ represents the local field enhancement factor at hotspots [6]. Typically, the hotspots occur in the closed interstices between nanoparticles that are much stronger than the surface of individual particle. Thus, one of the most significant strategies to improve the sensitivity of SERS is to create a sufficient number of dense hotspots on a large-scale substrate.

To this purpose, there are massive SERS substrates fabricated with regular morphology of nanostructure, such as nanoholes [7], nanodiscs [8], nanorods [9], and nanocaps [10], etc., and modified with metal nanoparticles capable of supporting LSPR to further provide powerful SERS effects. However, among the emerging hotspot-dependent SERS substrates, there are still problems of limited and unstable number of hotspots. To this, more complicated towards 3D SERS substrates have attracted particular attention, whose building blocks are usually made of cylinders, nanowires or nanopillars decorated with metallic nanoparticles [11–17]. This kind of SERS architecture with extension in the vertical dimension provides larger surface area for the noble metallic nanoparticles as well as target molecules to attach to. It can potentially increase the distribution of narrow spatial gaps in all three dimensions and form 3D denser hotspots to achieve much high Raman signal sensitivity. The building block materials of these 3D SERS substrates commonly include element semiconductors (Si) [13,18–20] and metallic oxides ($\textrm{ZnO}$ [21–23], $\textrm{Ti}{\textrm{O}_2}$ [24–26]). For example, it was reported that both Au-coated vertically aligned $\textrm{ZnO}$ nanorods and Ag-modified $\textrm{Ti}{\textrm{O}_2}$ nanorods scaffold showed a detection limit down to ${10^{ - 12}}\textrm{M}$ for the target molecules methylene blue [23] and malachite green [26], respectively. Especially, silicon-based SERS substrates have been much more investigated due to its unique properties. Silicon material shows both good biocompatibility and CMOS-compatibility, and has promising applications in designing advanced clinical testing devices. More importantly, silicon nanomaterial with a typical high optical refractive index can support multiple low-loss optical resonances including electric/magnetic Mie resonances [27], anapole states [28] and bound states in the continuum [29]. It enables hybrid coupling between silicon-based optical modes and metallic plasmonic resonances, and can dramatically engineer the light field on the nanoscale both inside and outside the nanostructure. This provides the capability to generate higher hotspot densities per unit volume. These features render high-refractive index silicon nanostructure into a key ingredient for Raman enhancement. The recent nanofabrication methods, such as electron-beam/photo lithography and nanoparticle self-assembly, have also led to the emergence of various excellent silicon-based SERS substrates. These include hexagonal-packed silicon nanowire arrays decorated with Ag or Au nanoparticles (AuNPs) [18,20,30], spatial-tunable Au-coated Silicon nanorods (SiNRs) arrays [31,32] and heterostructured $\textrm{ZnO}/\textrm{Si}$ nanostructures decorated with metallic nanoparticles [33,34]. Many of those SERS substrates achieved high enhancement of Raman signals for regular small analyte molecules (such as rhodamine 6 G (R6G), crystal violet and 4-aminothiophenol), and even long-chain biomacromolecules (such as protein fibrils and DNA molecules). Despite these demonstrations, it is still desired and necessary to further develop new morphology of silicon-based SERS substrate with high sensitivity, reproducibility and stability. Moreover, most of the previous approaches require complicated and expensive preparation steps. To reduce cost, many methods have been explored to recycle SERS substrates. These include using the photocatalytic properties of molecules [23,35] or removing metallic particles adsorbed by molecules [36] to achieve self-cleaning of substrates. To find other recyclable alternative strategy remains to be explored.

In this paper, a novel and effective SERS-active substrate that can be used repeatedly while maintaining high Raman sensitivity was proposed. This SERS substrate is based on large-scale 3D Si hierarchical nanoarrays decorated with homogeneous AuNPs, fabricated through a simple and low-cost top-down nanofabrication method. The building block of 3D Si hierarchical nanoarrays composes of a thick SiNR in the middle, surrounded by a few thin SiNRs leaning on, like an umbrella-frame structure. This umbrella-frame structure can not only provide a larger adhering area for target molecules and AuNPs, but also generate sufficient 3D plasmonic gaps due to hybrid coupling between silicon optical trapped mode and AuNPs LSPR as well as that between AuNPs LSPR. Finite-difference time-domain (FDTD) simulations illustrate strong field enhancement around Au-coated thick SiNR and thin leaning SiNR, and the dense 3D hotspots existing on the whole SERS substrate surface. Due to these, the Au-coated Si-based umbrella-frame substrate presents excellent SERS performance with high sensitivity and spatial uniformity. The detection limit for target R6G can be as low as ${10^{ - 14}}\textrm{M}$. The relative standard deviation (RSD) reaches about 11% indicating good SERS reproducibility. More importantly, a simple erase-and-reuse solution by plasma cleaning was further explored to recycle SERS substrate. The experimental results demonstrated the proposed substrate has high stability and long shelf life, making it a promising SERS platform in real application scenarios.

2. Materials and methods

2.1 Materials

The suspensions of polystyrene spheres (PS, $650\;\textrm{nm}$ and $220\;\textrm{nm}$ in diameters) were purchased from Duke Scientific (USA). The Si wafers (n-type, $0.01\mathrm{\;\ \Omega } \cdot \textrm{cm}$) were purchased from MTI (China). Acetone, methanol, ${\textrm{H}_2}\textrm{S}{\textrm{O}_4}$, ${\textrm{H}_2}{\textrm{O}_2}$ and HF were purchased from Sinopharm Chemical Reagent (China). Thiol functionalized methoxyl polyethylene glycol, PEG thiol (mPEG-SH, Mw 10000) was purchased from Sigma-Aldrich. The AuNPs dispersed in water solution ($20\;\textrm{nm}$, $0.05\;\textrm{mg}/\textrm{mL}$) were purchased from Nanjing XFNANO Materials Tech Co., Ltd (China). The ultrapure water (resistivity $18.2{\;\ \mathrm{M}\mathrm{\Omega} } \cdot \textrm{cm}$) was obtained from an ultrafiltration system (Milli-Q, Millipore, Marlborough, MA).

2.2 Fabrication of umbrella-frame SiNRs@AuNPs substrate

Firstly, the crystalline Si wafer was ultrasonically cleaned in acetone, methanol and ultrapure water for $5\;\textrm{min}$, respectively. Then, the Si wafer was immersed in ${\textrm{H}_2}\textrm{S}{\textrm{O}_4}/{\textrm{H}_2}{\textrm{O}_2}$ (volume ratio: $4:1$) for $10\;\textrm{min}$, following by a $5\;\textrm{min}$ ultrapure water washing. Finally, the Si wafer surface was hydrophobized with a 30 s hydrofluoric acid dip and washed with ultrapure water for $5\textrm{min}$. The cleaned Si wafer was prepared to the next fabrication process for the AuNPs-decorated umbrella-frame SiNRs (SiNRs@AuNPs) substrate, as schematically shown in Fig. 1. (1) A monolayer of two kinds of PS spheres (diameters: $650\;\textrm{nm}$ and $220\;\textrm{nm},$ volume ratio: $1:3$) was self-assembled onto the cleaned Si wafer; (2) The PS spheres were then etched by reactive ion etching (RIE, power 50 W and chamber pressure 70 mTorr, Trion Technology) with oxygen gas (flow rate 20 sccm); (3) The gold film with a thickness of about $20\,\textrm{nm}$ was deposited onto the PS-covered substrate by electron beam evaporation (EBE, Kurt J. Lesker, LAB18). This gold film mainly acts as a catalyst to etch Si wafer by the following metal-assisted chemical etching (MACE) method [37]; (4) The Au/PS-covered substrate was then immersed in $\textrm{HF}({40\textrm{wt\%}} )/{\textrm{H}_2}{\textrm{O}_2}\textrm{}({30\textrm{wt\%}} )$ mixed solution with a volume ratio of $4:1$. During the MACE process, the Au/PS-covered Si wafer could not be etched due to the mask role of PS spheres, while only Au-covered substrate could be etched vertically, forming SiNRs. Especially, due to the surface tension of the solvent evaporates, the thin SiNRs corresponding to the small PS spheres could not support themselves and lean on the thick SiNRs corresponding to the large PS spheres, forming Au/PS-coated umbrella-frame SiNRs. The length of umbrella-frame SiNRs could be controlled by the MACE times and the diameters correspond to the PS sphere diameters controlled by RIE time in step 2. (5) Afterwards, the remaining Au film and PS spheres were removed by $\textrm{KI}/{\textrm{I}_2}$ solution ($\textrm{KI}10\textrm{g},\textrm{}{\textrm{I}_2}\textrm{}2.5\textrm{g},\textrm{}{\textrm{H}_2}\textrm{O}100\;\textrm{mL}$) and tetrahydrofuran solution, respectively. A large-area umbrella-frame SiNRs arrays substrate was obtained. (6) To decorate AuNPs onto the surface of umbrella-frame SiNRs, a “grafting onto” strategy was utilized. The fabricated SiNRs substrate was functionalized with a Si hydroxyl group by soaking in an aqueous solution of mPEG-SH ($0.04\textrm{wt\%}$) for $24\;\textrm{h}$ at room temperature. Then, about $50{\ \mathrm{\mu} \mathrm{L}}$ of AuNP solution was dropped onto the surface of PEG-treated substrate. Through the bridging effect of mPEG-SH molecule, where -OCH₃ on one side is chemically linked with the hydroxyl group on the SiNR surface while the sulfhydryl on the other side is tightly connected to AuNPs, the SiNRs are well conjugated with sufficient AuNPs [38,39]. Finally, the sample was further washed with ultrapure water and cleaned with oxygen plasma to remove residual PEG and excess reactants of AuNP solution. The AuNPs-decorated umbrella-frame SiNRs substrate was fabricated successfully.

 

Fig. 1. Schematic diagram of the fabrication of the umbrella-frame SiNRs@AuNPs substrate.

Download Full Size | PDF

2.3 Structural characterization and spectra measurement

The size and monodispersity of the AuNPs and the morphology of the fabricated SERS substrates were characterized by scanning electron microscope (SEM, Sigma 300) and high-resolution TEM (HRTEM, JEM-F200). For the HRTEM measurement, the AuNPs sample was prepared by dropping a drop of AuNPs solution onto a holey carbon-coated molybdenum grid (200 mesh) and evaporating the solvent. Then the shape and size of AuNPs sample were measured by using HRTEM operating at 200 kV. The power X-ray diffraction (XRD) analysis was performed by synchrotron radiation XRD (SR-XRD) at beamline BL17B1 of the Shanghai Synchrotron Radiation Facility (SSRF) with the incident photon energy of 18 keV. For the SR-XRD measurement, 3 mL of AuNPs solution was centrifuged at 8000 rpm for 5 min, washed several times with distilled water, and then dried overnight to yield a small amount of AuNPs power material. The AuNPs power sample was then sealed in Kapton film for SR-XRD measurements. The UV−vis absorption spectra were measured by home-built angle-resolved optical spectroscopy with a Xenon lamp as light source. For AuNPs, the absorption spectrum was measured by transferring the AuNPs solution into a regular micro-quartz cuvette with a 10 mm optical path length. The transmission spectrum through air was used as a reference. For the absorption of the fabricated SERS substrates, a standard white board (250–1500 nm, average reflectance > 98%) was used as a reference.

2.4 SERS measurements

The Raman spectra were recorded by a confocal microscope/Raman spectrometer system (Renishaw, inVia-Reflex Raman) equipped with a $633\textrm\;{nm}$ laser (output power $1.7\textrm\;{mW}$) and a 532 nm laser (output power $2.5\; \textrm{mW}$). In the measurements, the spot diameter of the laser focused on the substrate was about 1.0 μm through 100× objective (numerical aperture 0.85). The spectral resolution was approximately $1\;\textrm{c}{\textrm{m}^{ - 1}}$ using a $1800\textrm{lines}/\textrm{mm}$ grating. Each Raman measurement was accumulated for 10 times with 10 seconds laser exposure time for per accumulation. To investigate the SERS performance, R6G was used as the probe molecule. The diluted solutions of R6G with concentrations ranging from ${10^{ - 5}}\textrm{M}$ to ${10^{ - 14}}\textrm{M}$ were prepared in ethanol. $10{\;\ \mathrm{\mu} \mathrm{L}}$ of R6G solution was drop-cast onto the SERS substrates. Due to the weak coffee effect of ethanol solvent and good hydrophilicity of the substrate surface, R6G molecules can uniformly disperse all over the SERS substrates.

2.5 Numerical simulations

Numerical simulations were performed with commercial 3D finite-difference time-domain calculations (FDTD, Lumerical Solutions Inc.). The simulation model matching the geometrical parameters of the as-fabricated SERS substrate was simplified in Supplement 1, Fig. S6. The plane wave source was incident on the SERS substrate from the head (top to bottom, z-axis) with linear polarization (x-axis) along the direction of the umbrella ribs leaning umbrella shaft. The boundary conditions on all three dimensions of the simulation region were set to perfectly matched layers. To achieve accurate results, non-uniform meshing with mesh accuracy 2 was used, and the finest mesh step of $2\;\textrm{nm}$ was set over the umbrella-frame structure. The permittivity of AuNPs was taken from the experimental data of Johnson and Christy [40]. To simulate plasmon resonance of AuNPs in water, the refractive index of surrounding medium was set to 1.33 to mimic water. A total field-scattered field source with $500 - 700\;\textrm{nm}$ wavelengths was used. A 3D power monitor enclosing the AuNP was used to simulate plasmon spectrum. The simulated plasmon resonance wavelength agrees well with the experimental result. For both SERS substrates and a single AuNP in water, 2D frequency domain field profile monitors were used to analyze the near-field intensity distribution. The simulation times were $10000\;\textrm{fs}$ and $500\;\textrm{fs}$, respectively, ensuring convergence.

3. Results and discussion

One key point in the prepared umbrella-frame SiNRs@AuNPs substrates is the conjugation of AuNPs with the SiNRs. This directly affects the formation and distribution of hotspots. The morphology of AuNPs was measured by using HRTEM. As shown in Fig. 2(a), the AuNPs with spherical shapes are observed and show a good monodispersity with diameter $20\;\textrm{nm}$. The HRTEM image of individual AuNP shown in Supplement 1, Fig. S1 shows the distinct lattice fringes with an interlayer spacing of about 0.24 nm, which is assigned to the (111) plane of Au. Moreover, the XRD pattern of AuNPs exhibits strong diffractivity, and confirms the face centered cubic (fcc) crystalline geometry of AuNPs (shown in Supplement 1, Fig. S2). The UV-vis absorption spectrum of AuNPs in water solution (Fig. 2(b)) shows a strong absorption peak at $530\;\textrm{nm}$, the inset shows the electric dipole field at LSPR for an individual AuNP in water [41,42]. Typical SEM images during the SiNRs@AuNPs substrates fabrication are shown in Fig. 2(c)–2(e). As shown in Fig. 2(c), two kinds of PS spheres mixed on Si wafer self-assembled into a good monolayer structure. The larger PS spheres (diameter $650\;\textrm{nm}$) are closely surrounded by the smaller PS spheres (diameter $220\;\textrm{nm}$) with a good homogeneous distribution of both particles. Through the 2D Fourier transform of Fig. 2(c) (The inset), a ring-like feature of the Fourier components is observed. It reveals the self-assembly structure of mixed PS spheres is amorphous photonic structure with the well-defined short-range order. It is different from the periodic case of single-size PS spheres with both long- and short-range orders (The SEM images of the PS spheres self-assembly with $650\textrm{nm}$ and $220\textrm{nm}$ are exhibited in Supplement 1, Fig. S3, respectively). This is because the introduction of PS spheres with different sizes can break down the long-range ordered arrangement of single-size PS spheres. Figure 2(d) shows the SEM image of a final 3D umbrella-frame SiNRs@AuNPs substrates. The diameters of the thick and thin SiNR are about $400\;\textrm{nm}$ and $50\;\textrm{nm}$, respectively. The SiNRs arrays have a length of about $1.2{\;\ {\mu} \mathrm{m}}$, which makes the thin SiNRs unable to support themselves and lean on the adjacent thick SiNR. The thick SiNR in the middle shapes like an “umbrella shaft”, while the surrounding thin SiNRs shape like “umbrella ribs”. The whole structure is like an “umbrella frame”. SEM images of SiNRs arrays corresponding to single-size $650\;\textrm{nm}$ (SiNRs-650) and $220\;\textrm{nm}$ (SiNRs-220) are also shown in Supplement 1, Fig. S4. For a fair comparison, the fabrication conditions of the SiNRs-650 and SiNRs-220 arrays are the same as the umbrella-frame case. The inset in Fig. 2(d) shows its 2D Fourier transform. It indicates that this umbrella-frame SiNRs@AuNPs structure is more like a random structure whose unit umbrella frame acts as a single scatter. Figure 2(e) shows the zoomed-in view of Fig. 2(d). It can be seen that a large density of AuNPs are modified on the surface of SiNRs with a good spatial uniformity, which is further confirmed by the EDS elemental maps (Supplement 1, Fig. S5). The absorption spectra of these Si-based SERS substrates are compared in Fig. 2(f). It clearly shows that the SiNRs-220 arrays substrate presents a nearly 95% absorbance at a broadband wavelength from $500\;\textrm{nm}$ to $800\;\textrm{nm}$, higher than the SiNRs-650 case. This is mainly attributed to the structure periodicity parameters. The SiNRs-220 arrays periodicity is lower than the incident light wavelength and the light is more likely to enter the structure and be absorbed, while the SiNRs-650 arrays has a periodicity comparable to the light wavelength, and is prone to Bragg scattering. So, a typical high reflection near $600\;\textrm{nm}$ for the SiNRs-650 arrays is observed. Due to the presence of the two effects mentioned above, the absorption of the umbrella-frame SiNRs formed by the mixture of these two sizes of PS spheres is in between, but still up to 90%. After coating with AuNPs, the absorption of the umbrella-frame SiNRs can be dramatically enhanced to nearly 98%, indicating a strong enhanced plasmonic effect by the umbrella-frame SiNRs@AuNPs substrates. This superb ability to trap light over a wide spectral range can facilitate Raman scattering performance.

 

Fig. 2. Morphology and absorption spectra of the SERS substrates at different fabrication process stages. (a) HRTEM image of synthesized AuNPs. Scale bar: $20\;\textrm{nm}$. (b) UV-vis absorption spectrum of AuNPs in water solution. The inset exhibits the simulated electric-field distribution at LSPR for an individual AuNP in water. SEM images of (c) the self-assembly structure of the mixed PS spheres and (d) umbrella-frame SiNRs@AuNPs array. The inset in (c) and (d) exhibit the corresponding 2D Fourier transform. (e) Zoomed-in SEM image of umbrella-frame SiNRs@AuNPs substrate. Scale bars in (c) and (d): 1 μm, scale bar in (e): $500\;\textrm{nm}$. (f) UV-vis absorption spectrum of SiNRs-220@AuNPs, SiNRs-650@AuNPs, umbrella-frame SiNRs, and umbrella-frame SiNRs@AuNPs substrates.

Download Full Size | PDF

3.1 Sensitivity

To investigate the sensitivity of the proposed SERS substrate, firstly, there is a comparison of Si signal between planar Si wafer substrate, umbrella-frame SiNRs and umbrella-frame SiNRs@AuNPs substrate. As shown in Fig. 3(a), the different enhancements of the Si signal ($520.7\;\textrm{c}{\textrm{m}^{ - 1}}$) for these SERS substrates are observed. It can be clearly seen that the signal of bare umbrella-frame SiNRs is stronger than the planar Si wafer. This originates from the strong absorption enhancement of incident light induced by the multiple reflections between the SiNRs. After coating with AuNPs, the Si signal intensity is enhanced dramatically compared to bare umbrella-frame SiNRs, indicating further strong field enhancement of the SiNRs structure enabled by the LSPR effect of AuNPs.

 

Fig. 3. Raman measurements on different SERS substrates. (a) Si Raman signal ($520.7\;\textrm{c}{\textrm{m}^{ - 1}}$) measured on the planar Si wafer, umbrella-frame SiNRs, and umbrella-frame SiNRs@AuNPs substrates. (b) Raman signal of ${10^{ - 7}}\textrm{M}$ R6G molecules measured on the planar Si wafer, SiNRs-220@AuNPs, SiNRs-650@AuNPs, umbrella-frame SiNRs, and umbrella-frame SiNRs@AuNPs substrates.

Download Full Size | PDF

Next, R6G, as a typical probe molecule in standard SERS measurements, was used to evaluate the Raman enhancement performance of the umbrella-frame SiNRs@AuNPs substrate in detail. The Raman measurement results of ${10^{ - 7}}\textrm{M}$ R6G are shown in Fig. 3(b). It can be seen that because of the ultralow concentration of R6G molecules, the R6G signal is too weak to be distinguished on the planar Si wafer. Interestingly, the R6G signal can be enhanced on the bare umbrella-frame SiNRs substrate, but the enhancement factor is lower than that of Si in Fig. 3(a). This phenomenon is understandable. In contrast to the plasmonic field enhancement on the metal surface, the high-refractive-index SiNR mainly produces large internal field enhancements. Therefore, the signal of R6G molecules adsorbed on the bare SiNRs surface is limitedly enhanced. Obviously, it is significantly enhanced on three types of Au-decorated SiNRs substrates. For two single-size SiNRs arrays substrates, the SiNRs-220@AuNPs substrate achieves a higher signal than SiNRs-650@AuNPs, which is consistent with the absorbance intensity results in Fig. 2(f). Especially, for the umbrella-frame SiNRs@AuNPs substrate, its Raman enhancement is much higher than that of the other two (SiNRs-650@AuNPs and SiNRs-220@AuNPs). To quantify the enhancement capability of the umbrella-frame SERS-active substrate, the enhancement factor (EF) is calculated as follows [43]

To further evaluate the detection sensitivity of the umbrella-frame SiNRs@AuNPs substrate, a series of Raman spectra were measured with concentrations of R6G molecules gradient from ${10^{ - 5}}\textrm{M}$ to ${10^{ - 14}}\textrm{M}$. As a decrease in the R6G concentration, the number of R6G molecules adsorbed on the SERS structure decreases, naturally depressing SERS performance. Figure 4(a) shows that the Raman intensity decreases with R6G concentration, as expected. Specifically, it can be clearly seen that the distinct Raman peaks of R6G located around $613\;\textrm{c}{\textrm{m}^{ - 1}}$, $774\;\textrm{c}{\textrm{m}^{ - 1}}$, $1184\;\textrm{c}{\textrm{m}^{ - 1}}$, $1310\;\textrm{c}{\textrm{m}^{ - 1}}$, $1362\;\textrm{c}{\textrm{m}^{ - 1}}$, $1507\;\textrm{c}{\textrm{m}^{ - 1}}$, and $1649\;\textrm{c}{\textrm{m}^{ - 1}}$ exhibit a strong Raman signal at ${10^{ - 5}}\textrm{M}$. With decreasing the concentration of R6G molecules even down to ${10^{ - 14}}\textrm{M}$, the R6G Raman signals decrease but could still be clearly identified, indicating the excellent Raman sensitivity of the umbrella-frame SiNRs@AuNPs structure. It can also be found that the variation of the Raman signal of different R6G characteristic peaks is different with the concentration from ${10^{ - 5}}\textrm{M}$ to ${10^{ - 14}}\textrm{M}$. The Raman peak at $774\;\textrm{c}{\textrm{m}^{ - 1}}$ is first decreased to an almost indistinguishable signal at the concentration of ${10^{ - 14}}\textrm{M}$. The Raman signal at $1184\;\textrm{c}{\textrm{m}^{ - 1}}$ is decreased by about 22 times. The smallest change is in the peak at $1310\;\textrm{c}{\textrm{m}^{ - 1}}$, with only about 7.9 times. To further investigate the relationship between the R6G Raman signal and its concentration, four characteristic peaks of R6G ($1184\;\textrm{c}{\textrm{m}^{ - 1}}$, $1362\;\textrm{c}{\textrm{m}^{ - 1}}$, $1507\;\textrm{c}{\textrm{m}^{ - 1}}$, and $1649\;\textrm{c}{\textrm{m}^{ - 1}}$) were specifically counted, as shown in Fig. 4(b). A good linear relationship between the R6G signal intensity and the logarithm of its concentration in the range of ${10^{ - 5}}\textrm{M}$ to ${10^{ - 8}}\textrm{M}$ is obtained, with correlation coefficients ${\textrm{R}^2}$ of 0.946, 0.996, 0.994, and 0.991, respectively. This indicates the potential applications of the proposed SERS-active substrate in quantitative determination. At concentrations below ${10^{ - 8}}\textrm{M}$, the decreasing rate of Raman intensity with concentration does not keep the previous linear correlation and becomes sharply slow. There is a tendency for the Raman intensity to be a limited value as the R6G concentration down to ${10^{ - 14}}\textrm{M}$. This may be attributed to the limited number of molecules (several or even single molecules) available at very low concentrations. The overall trend of Raman intensity with concentration is similar to previous work [44–46], except that the different concentration conditions for the linear relationship, which may be related to the different adsorption capacities and enhancement factors of the molecules on the SERS-active substrates.

 

Fig. 4. Sensitivity measurements on the umbrella-frame SiNRs@AuNPs substrate. (a) Raman spectra of R6G with concentrations gradient from ${10^{ - 5}}\textrm{M}$ to ${10^{ - 14}}\textrm{M}$. (b) Raman intensities at four characteristic peaks of R6G molecules ($1184\;\textrm{c}{\textrm{m}^{ - 1}}$, $1362\;\textrm{c}{\textrm{m}^{ - 1}}$, $1507\;\textrm{c}{\textrm{m}^{ - 1}}$, $1649\;\textrm{c}{\textrm{m}^{ - 1}}$) as a function of R6G concentration. The inset exhibits the corresponding linear relationship and correlation coefficients ${\textrm{R}^2}$.

Download Full Size | PDF

3.2 Enhancement mechanism

To elucidate the enhancement mechanism of the proposed SERS-active substrate, a 3D FDTD simulation was performed to calculate the electric field distribution. Based on the 2D Fourier transform in the inset of Fig. 2(d), the unit umbrella frame behaves more like a single scatter. Thus, the perfectly matched layers were set on the three dimensions of the simulation models (Supplement 1, Fig. S6). The geometric parameters of the structure are unified with the as-fabricated SERS substrate. The simulated absorption spectrum of the umbrella-frame SiNRs@AuNPs structure is in good agreement with the experimental result, also maintaining a broadband ∼93% absorption (Supplement 1, Fig. S7). The electric field distributions of the bare umbrella-shaft SiNR, umbrella-frame SiNRs and SiNRs@AuNPs substrate for incidence light wavelength of $633\;\textrm{nm}$ were compared in Fig. 5. It can be found that the antinodes of the enhancement fields are regularly distributed over the surface of the umbrella-shaft SiNR due to the excited leaky guided mode of SiNR [47]. By comparing the fields of the bare umbrella-shaft SiNR (Fig. 5(a)) and the umbrella-frame SiNRs (Fig. 5(b)), the umbrella-ribs SiNRs can further confine the light field to the SiNRs structure and reduce its spatial decay length. This is very beneficial for improving the Raman signal of the molecules adsorbed on the structure surface. After coupling with the AuNPs-induced LSPR, the electric fields at the antinodes become strong and local, forming a large number of hotspots (Fig. 5(c)). Figure 5(d) shows the decay of the field intensity in each antinode. It clearly shows that the leaky guided mode of SiNR, which originally decays gradually with distance towards air, is dramatically enhanced due to the introduction of the umbrella-ribs SiNRs and, in particular, AuNPs. Especially in the gap between the umbrella-shaft SiNR and the umbrella-ribs SiNRs, these hotspots are greatly enhanced. Even the zero-gap fields are generated in the close inter-rod space (marked with arrows in Fig. 5(d)). The field distributions in the x-y plane at each antinode are shown in Fig. 5(e-h). It shows that since the umbrella-ribs SiNRs are distributed around the umbrella-shaft SiNR, the field coupling also exists in another dimension, forming intensive 3D local hotspots. The x-y plane field distribution of the umbrella-shaft SiNR and umbrella-ribs SiNRs were also investigated (Supplement 1, Fig. S8). Additionally, the x-z plane field distribution at the incident wavelength of $532\;\textrm{nm}$ (Supplement 1, Fig. S9) implies that high SERS performance can also be obtained with $532\;\textrm{nm}$ laser incidence. A series of Raman spectra of R6G molecules with concentrations gradient from ${10^{ - 6}}\textrm{M}$ to ${10^{ - 12}}\textrm{M}$ under 532 nm excitation light were measured. The SERS spectra are shown in Supplement 1, Fig. S10. We can find that the Raman intensity of R6G decreases with its concentration, which is similar with that of 632 nm excitation light. When the concentration of R6G molecules decreases even down to ${10^{ - 12}}\textrm{M}$, the R6G Raman signals decrease but could still be clearly identified, indicating the excellent Raman sensitivity of the SERS substrate under 532 nm excitation light. As a result, a highly sensitive SERS-active substrate with multi-wavelength response was achieved based on the AuNPs-coated silicon-based umbrella-frame structure. For the proposed umbrella-frame SiNRs@AuNPs substrate, the high sensitivity of the achieved SERS performance is mainly attributed to electromagnetic enhancement (EM). In fact, we can further optimize our SERS performance by introducing chemical enhancement (CM) into our SERS substrate. CM can simply be seen as corresponding to the charge transfer between the target molecules and the SERS substrate [48]. Guided by this, 2D materials are a kind of promising SERS materials due to their high carrier mobility, excellent photoelectricity activity, and large specific surface area [49,50]. We can integrate the umbrella-frame SiNRs@AuNPs SERS substrate with 2D materials. Through developing this unique nanostructure boosted by synergistic resonance enhancement of CM and EM, it is worthwhile to expect to obtain a remarkable SERS sensitivity for single-molecule detection, and has a potential application in the practical bio-application of SERS technology, such as pesticide residue analysis and hazardous substance detection [51].

 

Fig. 5. FDTD simulation of the electric field distributions of different SERS substrates under $633\;\textrm{nm}$ incident light. (a-c) show the x-z plane electric field distributions of the bare umbrella-shaft SiNR, umbrella-frame SiNRs, and umbrella-frame SiNRs@AuNPs substrates, respectively. (d) Line profiles of electric field intensity along the x-axis in four antinodes for three substrates. Four positions are depicted by z₁, z₂, z₃ and z₄ in (c), respectively. The range of the line profile of field intensity along the x-axis is denoted by the black dashed line in (c). The gray shaded area in (d) represents the field in the umbrella-shaft SiNR. The yellow circle represents AuNP. (e-f) show the x-y plane electric field distributions of the umbrella-frame SiNRs@AuNPs substrates at z₁, z₂, z₃ and z₄, respectively. Scale bars in (a-c, e-h): $500\;\textrm{nm}$.

Download Full Size | PDF

3.3 Reproducibility and uniformity

The reproducibility and uniformity are also major concerns in practical applications for any SERS-active substrate. This can be investigated by spot-to-spot Raman scanning. Figure 6(a) shows the stacking SERS spectra of ${10^{ - 6}}\textrm{M}$ R6G on 10 different active-sites over a $100 \times 100{\;\ \mathrm{\mu} }{\textrm{m}^2}$ area. It can be seen that 10 SERS spectra exhibit relatively high uniformity and stability. To quantify the spot-to-spot reproducibility, one typically introduces the relative standard deviation (RSD), i.e., the ratio of the standard deviation to the mean, of the signal-to-base peak intensity from the multiplex SERS spectra. Figure 6(b) shows that the RSDs of several characteristic peaks ($613\;\textrm{c}{\textrm{m}^{ - 1}}$, $774\;\textrm{c}{\textrm{m}^{ - 1}}$, $1184\;\textrm{c}{\textrm{m}^{ - 1}}$, $1310\;\textrm{c}{\textrm{m}^{ - 1}}$, $1362\;\textrm{c}{\textrm{m}^{ - 1}}$, $1507\;\textrm{c}{\textrm{m}^{ - 1}}$, and $1649\;\textrm{c}{\textrm{m}^{ - 1}}$) from 30 spectra with a range of 10.81% ∼ 12.15%. These values are slightly higher than the RSDs of the periodic structure, but extremely lower than that of the random structure, indicating a high reproducibility of the close-packed umbrella-frame SERS-active structure. Moreover, the temporal stability of the SERS-active substrate was investigated. The Raman spectra of R6G molecules evolving with time on the same position of the substrate were measured and the time evolution of the signal peak intensities are shown in Supplement 1, Fig. S11. The R6G signal intensities by repeated measurements during the continuous laser irradiation (45 min, at 5 min intervals) remain fairly stable, indicating that the photothermal effect is negligible for this SERS-active substrate. And the Raman signal still retained 98% of the original value even after a long time (2 weeks, shown in Supplement 1, Fig. S12). This means that the SERS structure is hardly affected by environmental influences such as oxidation. Thus, the proposed SERS-active substrate exhibits good photothermal stability and long-term stability.

 

Fig. 6. Reproducibility measurements on the umbrella-frame SiNRs@AuNPs substrate. (a) Ten Raman spectra of ${10^{ - 6}}\textrm{M}$ R6G located on different active-sites over a $100 \times 100{\;\ \mathrm{\mu} }{\textrm{m}^2}$ area. (b) Relative standard deviation from 30 Raman spectra measured on the same SERS-active substrate.

Download Full Size | PDF

3.4 Recyclability

The potential recyclability of the SERS-active substrate is worth investigating due to the obvious economic benefit. Several methods have been proposed in previous works. The photocatalytic activity of the SERS-active substrate has been often utilized to remove the analyzed molecules. This approach relies on the photocatalytic efficiency of the SERS structure and requires a relatively long irradiation time, typically in the order of hours, to achieve a complete molecule degradation. Here, a simple, rapid and low-cost alternative by oxygen plasma cleaning is used to erase the prepared umbrella-frame SiNRs@AuNPs substrate. Its principle is based on the rapid thermal degradation of the molecules. Figure 7(a) shows the multi-process SERS activities of the erase-and-reuse SERS-active substrate for R6G molecules via oxygen plasma cleaning. The Raman spectra of each process were mediated over 15 different spatial positions covering the entire substrate. It can be seen that with increasing plasma cleaning time, the R6G signals become gradually weaker (upper in Fig. 7(a)), and completely vanish after 9 min (middle in Fig. 7(a)), much shorter than the photodegradation time. It indicates that the by-products generated during the oxygen plasma bombardment of the molecules are easily desorbed from the SERS substrate. Subsequently, the cleaned substrate was reused to detect R6G molecules at the same concentration, obtaining R6G signal intensity comparable to that before plasma erasing (lower in Fig. 7(a)). The SEM image of the umbrella-frame SiNRs@AuNPs substrate after erasing also shows no noticeable damage to the SERS structure by oxygen plasma (Supplement 1, Fig. S13). Furthermore, this erase-and-reuse cycle was repeated over 10 times and still achieved highly reproducible SERS activity, as show in Fig. 7(b). It should be noted that because the oxygen plasma cleaning strategy is based on the rapid thermal degradation of the molecules, it should have the same effect on other molecules, such as crystal violet (CV). As shown in Supplement 1, Fig. S14, the erase-and-reuse SERS activity for CV molecules are similar to that of R6G. This demonstrates that the proposed umbrella-frame SiNRs@AuNPs substrate exhibits excellent recyclability for robust SERS functionality.

 

Fig. 7. (a) Recyclability measurements of the umbrella-frame SiNRs@AuNPs substrate for R6G molecules via oxygen plasma cleaning. (b) Relative Raman intensities of R6G at $1649\;\textrm{c}{\textrm{m}^{ - 1}}$ obtaining on the umbrella-frame SiNRs@AuNPs substrate after 10 times erase-and-reuse cycles.

Download Full Size | PDF

4. Conclusion

In summary, an umbrella-frame SiNRs array decorated with homogeneous AuNPs was proposed served as an efficient, stable and recyclable SERS-active substrate. The substrate was fabricated by using self-assembly and lithography techniques, a simple and cost-effective top-down nanofabrication method. This SERS-active substrate exhibits excellent Raman performance for analyte molecules (R6G) with high sensitivity and reproducibility. The detection limit of R6G molecules can be as low as ${10^{ - 14}}\textrm{M}$, with an EF of up to ${10^7}$. This is mainly attributed to the following two reasons. One is much more active-sites provided by the umbrella-frame structure for adsorption of target molecules and AuNPs. The other, more important, is sufficient hotspots in three dimensions of the umbrella-frame structure. Both the umbrella-ribs SiNRs and AuNPs can dramatically confine the light field to the SiNRs surface and generate strongly enhanced and even zero-gap fields in 3D space. Moreover, the proposed SERS-active substrate exhibits good photothermal stability, long-term stability and typically excellent recyclability for robust SERS activity. A simple erase-and-reuse solution by oxygen plasma cleaning was further exploited to recycle SERS substrate. These results make the umbrella-frame SiNRs@AuNPs substrate exciting potential for SERS detection of molecules and biomolecules and provide a reliable strategy for the development of new pattern-based SERS platform.