Three-dimensional (3D) Ag–Au/SiO₂ periodic nanogrids with various thickness ratios of Ag and Au were designed and fabricated. Hydrogen silsesquioxane (HSQ XR-1541–006, Dow Corning, USA) was firstly spin-coated on silicon (100) substrates with the thickness of 198 nm. Patterning was realized by using electron beam lithography (EBL, Vistec EBPG 5000 plus ES, Raith Company, Germany) with an accelerating voltage of 100 kV and a beam current of 2 nA, followed by the development. The typical width of SiO₂ sidewalls for all nanogrids was controlled to be around 11 nm (Fig. A1).^[26,27] Alternate deposition of Ag and Au with m cycles, along with a 1-nm-thick Cr adhesion layer was performed on an electron-beam evaporator (OHMIKER-50B, Cello-Tech Company, Taiwan, China) with the constant total thickness of 36 nm. The rate of deposition was 0.4 Å/s for Cr and 0.2 Å/s for both Ag and Au. The fabricated structures were observed using a scanning electron microscope (NOVA NanoSEM 430, FEI Company, USA). Upon annealing, the samples were heated to 200 °C at a rate of 10 °C/min and kept for 1 h in nitrogen in a fast alloy furnace (RTP3, Shentong-TOEC Company, China), followed by cooling down to room temperature.

The optical constants of Ag–Au composite layers were determined using a spectroscopic ellipsometer (SE 850 DUV, Sentech Company, Germany). Finite- difference time-domain (FDTD) method was used to calculate the absorption spectra and spatial distributions of the electromagnetic fields. For simplicity, we used the rough Ag–Au/SiO₂ models with periodically arranged semiellipsoid-like Ag–Au composite nanoparticles with a smooth Ag–Au composite film of 2 nm thick on the sidewalls of SiO₂ nanogrids to model the real Ag–Au/SiO₂ nanogrids in which the sizes and center distance of particles were semi-principal axes, a = b = 20.0 nm, c = 10.0 nm, and d = 40.0 nm for the calculations of absorption spectra.^[25] The width and height of SiO₂ grids were 11 and 198 nm, respectively. The structures were illuminated with a plane wave with x-direction linear polarization incident along the –z axis. Periodic boundary conditions for the xz and yz planes were applied to simulate an infinite array of periodic nanogrids or nanowalls. Perfectly matched layer (PML) boundary conditions were used in the z-direction. The mesh size used in the simulation region was 2 nm for the calculations of absorption spectra and 0.5 nm for the calculations of the electromagnetic fields.

The samples were first immersed into 4,4’-Bipyridine (Bpy) absolute ethanol solution with a concentration of 10⁻⁵ M for 4 h, and then dried naturally in air as SERS probes for Hg ions detection. 35 μL of Hg ion solutions with different concentrations of 5.0 × 10⁻¹² (1 ppt), 5.0 × 10⁻¹¹ (10 ppt), 5.0 × 10⁻⁹, 5.0 × 10⁻⁷, and 5.0 × 10⁻⁵ M were dropped onto the SERS probes, respectively, then kept for 10 min, and finally dried in air. Likewise, 35 μL deionized water was prepared with the same procedure as SERS probes for the blank control group. The SERS measurements were performed using a 514 nm laser with a power of 0.5 mW and the x-polarization on a Raman microscope (Renishaw in-Via, Renishaw company, UK) equipped with a 20× objective (NA = 0.4) and an integration time of 10 s. For each sample, measurements on at least five different positions were taken.

We designed and fabricated alternate Ag and Au layers with various thicknesses deposited on hexagonal 3D periodic SiO₂ nanogrids with a total thickness of 36 nm to maximize the SPP effects based on the proposed design method for exploration of SERS activity and stability.^[25] Here, 2-nm-thick Ag and 2-nm-thick Au deposited alternately on hexagonal SiO₂ nanogrids with a grid length of 188 nm and a height of 198 nm for 9 cycles is labeled by (2 nm Ag–2 nm Au)-9c/198 nm SiO₂_h188. The scanning electron microscopy (SEM) images of these samples are shown in Fig. 1, which reveals Ag–Au nanoparticles to form the rough sidewalls of nanogrids and the roughness of sidewalls appears to be very similar for Ag–Au layers with different thickness ratios. The statistical analysis of SEM images shown in Fig. 1 gives the almost same average widths of hybrid nanogrids with different atomic percents of Ag of about 28 nm with about 8.5-nm-thick Ag–Au layers on each side of about 11-nm-thick SiO₂ nanogrids, as shown in Fig. A1.

Fig. 1.

	Figure Option ViewDownloadNew Window
	Fig. 1. Tilt SEM images of SiO2 and hybrid hexagonal nanogrids_h188 with a height of 198 nm. (a) SiO2 nanogrids. (b) 36 nm Au/SiO2 nanogrids. (c) (2 nm Ag–2 nm Au)-9c/SiO2 nanogrids. (d) (5 nm Ag–1 nm Au)-6c/SiO2 nanogrids. (e) (11 nm Ag–1 nm Au)-3c/SiO2 nanogrids. (f) 36 nm Ag/SiO2 nanogrids. Scale bars: 200 nm.

In order to obtain optical properties of the Ag–Au composite layers above, their equivalent refractive indexes n and extinction coefficients k within the wavelength range from 360 nm to 930 nm were derived by fitting the spectroscopic ellipsometric curves using effective medium approximation method (EMA),^[28] as shown in Figs. 2(a) and 2(b), respectively. It is evident that the equivalent optical constants change with Ag–Au composite layers even for the thickness ratio of Ag to Au, such as (5 nm Au-1 nm Ag)-6c and (10 nm Au-2 nm Ag)-3c.

Fig. 2.

Figure Option ViewDownloadNew Window

Fig. 2. (color online) (a), (b) Refractive indexes and extinction coefficients of different Ag–Au composite layers derived from the measurements by spectroscopic ellipsometry, respectively. (c) Changes of experimental Raman intensities of the peak positioned at 1610 cm−1 of the Bpy-decorated 36 nm Ag–Au/198 nm SiO2 nanogrids_h188 with different thickness ratios of Ag and Au (atomic percentage of Ag) for 514 nm wavelength laser. (d) The calculated absorption spectra of 36 nm Ag–Au/198 nm SiO2 nanogrids_h188 with different thickness ratios of Ag and Au.

To evaluate the performance, the Raman intensity of the peak at 1610 cm⁻¹ of Bpy molecules is recorded to investigate SERS activity here, and its dependences on atomic ratio of Ag for 36 nm Ag–Au/198 nm SiO₂ nanogrids_h188 are presented in Fig. 2(c). The intensity is found to firstly increase more rapidly with the increase of Ag proportion and reach the maximum at 84.3% (i.e., (5 nm Ag-1 nm Au)-6c), and then decrease gradually but increase suddenly for pure Ag, indicative of the material properties dependences. Interestingly, the intensity of the (10 nm Ag–2 nm Au)-3c sample is found to be far lower than that of (5 nm Ag–1 nm Au)-6c sample though with the same percentage of Ag atoms. We believe that the SERS signal intensities were mainly related to their LSPR peaks which depend on the optical properties of materials (Figs. 2(a) and 2(b)) and structures. LSPRs are excited at the gaps between bottom and sidewalls, and at those on rough sidewalls under the light illumination, but those on the bottom film are negligible due to the weak electric field enhancements.^[25]

To get plasmon resonance modes for the hybrid structures with different Ag–Au components, we performed finite-difference time-domain (FDTD) simulations to derive the corresponding absorption spectra by assuming the same sidewall roughness and using the measured equivalent optical constants. The results are shown in Fig. 2(d) from which we can see three groups of peaks in the spectra. For the 36 nm Au/SiO₂ structure, one of the absorption peaks is near 675 nm and blue shifts to 523 nm with the increase of Ag atomic ratio. With the Ag atomic ratio above 84% (i.e., (5 nm Ag–1 nm Au)-6c), the position of the peak changes little but its intensity decreases. For the 36 nm Ag/SiO₂ structure, the resonance peak red shifts slightly but its intensity increases significantly. In addition, the resonance peak of the (10 nm Ag–2 nm Au)-3c structure red shifts a little compared with the (5 nm Ag–1 nm Au)-6c structure. Therefore, the changes of the absorption peak intensity under excitation of 514 nm laser can explain the Ag atomic ratio dependence of the experimental Raman intensity well.

For periodic and rough Ag–Au/SiO₂ hybrid nanogrids, SPP waves and their interference effects, LSPR effects, standing wave effect of incident light and coupling effects between LSPR and SPP are involved.^[25] For 514 nm light, a height with 36% of the incident light wavelength, 185 nm, is adequate for the strongest intrinsic EMF enhancement due to the standing wave effect (the height here is 198 nm which can meet the condition).^[25] The interference effects of SPP waves due to Fabry Perot (FP)-like resonance caused by multiple reflections in periodic structures,^[29,30] can excite LSPR as a secondary source (incident light as primary sources), like Au/SiO₂ nanogrid structures.^[25] According to the calculation formula of SPP wavelength and the equivalent optical constants of the Ag–Au composite layers (Figs. 2(a) and 2(b)),^[31] the wavelengths of SPP waves excited at these metal/air interfaces for the 514 nm excitation light are shown in Table A1. The SPP wavelengths range from 474 nm to 488 nm for Ag–Au composite layers with different thickness ratios, with 485 nm for the pure Ag and 375 nm for the pure Au. Here, we employ 488 nm for structure design.

The general design principles for 3D structures also hold for Ag–Au/SiO₂ nanogrids structures here.^[25] It is known that the SPP interference effects depend mainly on the geometries of structure especially the sidewall spacing (materials independent) and the reflectance of SPP wave reflected by the sidewall (materials dependent).^[25] Therefore, we need to know the sidewall spacing dependence of LSPR and the reflectance for Ag–Au/SiO₂ nanogrids. Taking (5 nm Ag–1 nm Au)-6c/198 nm SiO₂ as an example, sidewall spacing L_w dependences of the maximum and averaged |E/E₀|⁴ from LSPR on the rough nanowalls at TE mode calculated by FDTD solutions are given in Fig. A2. The maximum and average electric field enhancements (|E/E₀|⁴) are both maximized at L_w ∼ 300 nm. Its reflectance of SPP wave, R = 0.46 calculated by FDTD on the model with 8.5-nm-thick Ag–Au composite layers on each side of SiO₂ sidewall for the 488 nm SPP wave, is larger than that of 36 nm Au/SiO₂ sidewall, R = 0.29, implying the stronger interference intensity of SPP wave in Ag–Au/SiO₂ nanogrids compared to Au/SiO₂ nanogrids. Therefore, the geometric dimension dependence of SPP wave interference effect (FP-like resonance) still plays a dominant role in the relationship between the total electric field enhancement and the size of the Ag–Au/SiO₂ nanogrids. The hexagonal Ag–Au/SiO₂ nanogrids with the sidewall spacing of half SPP wavelength is the optimum structure with the strongest intrinsic electric field enhancement, which will be confirmed by the following SERS experiments.

SEM image of the optimized hexagonal 36 nm Ag–Au/198 nm SiO₂ nanogrids_h157 samples with 28 nm sidewall width and 244 nm sidewall spacing (half of 488 nm SPP wavelength) is compared with those h87, h101, h105, h188, h194, h227, h274, and h282, as shown in Fig. A3. Their SERS intensities and corresponding normalized intensities are shown in Figs. 3(a) and 3(b), respectively. Both the intensity and normalized intensity of the (5 nm Ag–1 nm Au)-6c/198 nm SiO₂ nanogrids SERS probe are the largest compared to those of (3 nm Ag–1 nm Au)-9c, (5 nm Ag–1 nm Au)-6c and (8 nm Ag–1 nm Au)-4c with the same grid length, which is attributed to the positions and intensities of the absorption peaks as discussed above (Fig. 2(d)). With the grid length, the normalized intensity shows the quasi-periodic characteristics and reaches the maximum at h157 for three kinds of Ag–Au structures, which indicates the maximum intrinsic electric field enhancement of h157. This reveals that our design method for 3D periodic SERS probes is generalized which can be applied to different materials.^[25]

Fig. 3.

	Figure Option ViewDownloadNew Window
	Fig. 3. (color online) (a) Changes of experimental SERS intensities of hexagonal Ag–Au/198 nm SiO2 nanogrids with different thickness ratios of Ag and Au against grid length for 514 nm laser. (b) Changes of their corresponding normalized SERS intensities.

In order to study the detection performances of the Ag–Au and Ag/SiO₂ nanogrid SERS probes for Hg ions, SERS measurements were conducted on the SERS probes onto which Hg ion solutions with different concentrations were dropped. Meanwhile, deionized water instead of Hg ion solutions was prepared as the blank control experiments with the same procedure. The intensity changes of Ag–Au and Ag/SiO₂ nanogrid probes with the concentration of Hg ions and their SEM images after SERS measurements are shown in Fig. A4. The probes with the higher percentages of Au, such as the pure Au^[25] and (2 nm Ag–2 nm Au)-9c samples are not sensitive enough to the extremely low Hg ions concentrations, and thus their detection limits are only 5.0 × 10⁻¹¹ M. Their morphologies changed little after SERS measurements. In sharp contrast, the morphology of the pure Ag probe underwent significant changes after Hg ion detection, and its detection limit is also only 5.0 × 10⁻¹¹ M with a poorly linear relationship between intensity variations and concentrations which should be attributed to the instability of Ag in the atmosphere.^[32] For the (5 nm Ag–1 nm Au)-6c and (11 nm Ag–1 nm Au)-3c/SiO₂ nanogrid probes, their detection limits for Hg ions are 5.0 × 10⁻¹² M (1 ppt) with the linear relationships between intensity variations and Hg concentrations. This indicates that the Ag–Au composite layers with proper configurations would show the strengths of high sensitivity Ag and the high stability of Au simultaneously, which is therefore of great significance for practical applications.

We found that the change of signal intensity for (10 nm Ag–2 nm Au)-3c/SiO₂ nanogrids probe during the storage of 10 days is smaller than that of (5 nm Ag–1 nm Au)-6c/SiO₂ nanogrids probe, which indicates the better stability with 2 nm Au surface layer. Based on this, we fabricated (7 nm Ag–2 nm Au)-4c and (16 nm Ag–2 nm Au)-2c/198 nm SiO₂ nanogrids_h157 probes with 2 nm thick Au outermost layer to study the stability of the unannealed and annealed probes after 10 days of storage. We annealed the Au, Ag, and Ag–Au/198 nm SiO₂ nanogrids structures, and their SEM images are shown in Fig. 4. We found that the changes of morphologies for structures with thickness ratios of Ag and Au less than 8:1 after annealed are not obvious, but those of the structures with a higher percentage of Ag are significant. These SERS probes annealed were also used to detect the Hg ions, without obvious changes of morphologies, indicating that the annealing treatment makes the SERS probes more stable.

Fig. 4.

	Figure Option ViewDownloadNew Window
	Fig. 4. Tilt SEM images of annealed Au, Ag–Au and Ag/198 nm SiO2 nanogrids. (a)–(f) 36 nm Au, (3 nm Ag–1 nm Au)-9c, (5 nm Ag–1 nm Au)-6c, (11 nm Ag–1 nm Au)-3c, (17 nm Ag–1 nm Au)-2c, and 36 nm Ag/198 nm SiO2_h188, respectively. (g)–(i) (2 nm Ag–2 nm Au)-9c, (7 nm Ag–2 nm Au)-4c, and (16 nm Ag–2 nm Au)-2c/198 nm SiO2_h157, respectively. Scale bars: 200 nm.

Experimental SERS intensities of unannealed and annealed hexagonal (2 nm Ag–2 nm Au)-9c, (7 nm Ag–2 nm Au)-4c, and (16 nm Ag–2 nm Au)-2c/198 nm SiO₂ nanogrids_h157 probes are shown in Fig. 5(a). Their SERS intensities changed by -34.7%, 27.7%, and 24.4% after annealing. Relative intensity changes of unannealed, annealed, and 10-day stored hexagonal 36 nm Ag–Au/198 nm SiO₂ nanogrid probes with different thickness ratios of Ag and Au are shown in Fig. 5(b). The SERS intensity for the pure Au probe stored for 10 days decreased by 32.2%, possibly related to the surface adsorption of C, O, S, and other elements.^[33] The SERS intensity for the Ag–Au probes with higher percentages of Ag increased significantly by 200%–370% after 10 days of storage, which is mainly due to the formation of silver sulfide on silver during the exposure to air.^[32] Exposure to air for a short time would lead to a Ag₂S layer as thin as about 1 nm which could enhance the average electric field at hot spots dramatically. However, the longer time exposure would increase the thickness of Ag₂S layer,^[32] decreasing the average electric field enhancement greatly again, as demonstrated by FDTD calculations in Fig. A5. In addition, few changes of SERS intensities for the Ag–Au composite layers with proper proportions of Ag changed little after 10 days of storage are considered to probably result from a synergistic effect of the enhancement effect caused by a thin Ag₂S layer and the weakening effect caused by the adsorption of C, O, S and other elements. Again, for the same atomic ratios of Ag the intensities for the probes with a 2-nm-thick Au outermost layer changed less than those a 1-nm-thick Au outermost layer, which is mainly due to the better stability of Au.^[24] Therefore, increasing the thickness of outermost Au layer may facilitate the stability of probes. It can also be observed that annealing leads to the less change of intensity for SERS probes after 10- day storage, which stabilizes the probes by releasing the stress and stabilizing the microstructure.^[34] By comparison, annealed (7 nm Ag–2 nm Au)-4c/198 nm SiO₂ nanogrids_h157 probe shows both the highest intensity and the best stability with only 4.0% change in SERS intensity after 10-day storage, as illustrated in Fig. 5(a).

Fig. 5.

	Figure Option ViewDownloadNew Window
	Fig. 5. (color online) (a) Experimental SERS intensities of annealed and unannealed hexagonal Ag–Au/198 nm SiO2 nanogrids with different Ag and Au thickness ratios. (b) Relative changes of SERS intensities of the above probes stored for 10 days.

Here, the annealed hexagonal (7 nm Ag–2 nm Au)-4c/198 nm SiO₂ nanogrids_h157 SERS probes with 10-day and without storage (0 day) were used to detect Hg ions, as shown in Figs. 6(a) and 6(b), respectively. The relationships between intensity variation and the concentration (5.0 × 10⁻¹² M to 5.0 × 10⁻⁵ M) for both probes are found to be linear (Figs. 6(a) and 6(b)). Both probes detected 5.0 × 10⁻¹² M (1 ppt) trace Hg ions in water successfully, which is outstanding among those reported, about three orders of magnitude lower than the US threshold value (10 nM or 2000 ppt) for drinkable water.^[35,36] The SERS mapping of the annealed probe gives a low RSD of 5.78% (Fig. 6(c)), normally much smaller than those of the reported SERS substrates,^[13,37,38] which is quite significant for SERS probes. The excellent selectivity of the optimized SERS probes to Hg ions shown in Fig. 6(d) was demonstrated by comparing with other metal ions with the same concentration of 10⁻⁶ M, such as Mg²⁺, Co²⁺, Ni²⁺, Li⁺, Cu²⁺, Zn²⁺, Fe³⁺, Al³⁺, In³⁺, Mn²⁺, Pb²⁺, and mixed (without and with Hg ions) metal ions. The superior selectivity of the probes renders it possible to be applied for practical detections of trace Hg ions.

Fig. 6.

Figure Option ViewDownloadNew Window

Fig. 6. (color online) (a), (b) Detection of Hg ions by the annealed (7 nm Ag–2 nm Au)-4c/198 nm SiO2 nanogrids_h157 probes stored for 0 and 10 days, respectively. Raman spectra of Bpy molecules and the relationships between Raman signal intensities and concentrations of Hg ions. (c) Raman intensity mapping of the annealed hexagonal (7 nm Ag–2 nm Au)-4c/198 nm SiO2 nanogrids_h157 decorated by Bpy as SERS probes. (d) Selectivity of the optimized SERS probes for the detection of Hg ions over other metal ions.