Gold sol was prepared by the Frens method.^[22] 100 ml of 0.01-wt% aqueous solution of HAuCl₄ was placed in a round-bottom flask and heated to boiling under stirring. Next, 1 ml of 0.25-mM trisodium citrate solution was added quickly into the boiling solution, and the mixture was kept boiling gently for 30 min with continuous stirring. Then, the solution was cooled to the room temperature and put in a dark place before further use. By changing the amount of trisodium citrate which was added, to obtain different sizes of gold NPs, the sizes of 15 nm, 25 nm, and 50 nm of gold NPs were prepared.

For coating gold NPs with amorphous silica shells, two methods to prepare Au@SiO₂ NPs were used. The first method according to classic Stober method:^[23] 8-ml gold sol was dissolved in 20-ml isopropanol, and stirred for 3 min–5 min. Next, 0.5-ml aqueous ammonia was added rapidly, and then 5 μl of tetraethylorthosilicate (TEOS) was added, then stirred for 3 h. Different shell thickness can be obtained by controlling the reaction time and adding a certain amount of TEOS. With this method, we obtained Au NPs (∼ 25 nm) coated with SiO₂ shell of thickness ca. 15 nm, 35 nm, and 50 nm, respectively.

The second method is presented by Li et al.,^[17] 0.4 ml of 1-mM aqueous solution of (3-aminopropyl)trimethoxysilane (APTMS) was added to 30 ml of gold sol and stirred at 25 °C for 15 min. Subsequently, 3.2 ml of freshly prepared diluted sodium silicate solution (0.54 wt% of SiO₂) adjusted with HCl to PH 9.6–PH 11 was added with stirring 3 min–5 min. Then, we placed the flask in a 90-°C water bath and stirred the contents for 10 min–60 min. After the reaction finished, the flask was placed in ice–water mixture to terminate the reaction quickly.

Au@SiO₂ NPs monolayer film was prepared according to Li’s method.^[24] 5 ml of cyclohexane was added in 15-ml Au@SiO₂ NPs colloid. Then 4 ml of ethanol was added with stirring 3 min–5 min. Then, cyclohexane was removed with a glue dropper head. The Au@SiO₂ NPs monolayer film was transferred onto gold film and silicon for further experiments. The scanning electron microscopy (SEM) images are shown in the Supporting Information (S1 and S2).

The gold film and silicon were purchased from German Tech Co., Ltd and Henghao Industrail, respectively. The average roughness of gold film was less than 3 nm. Prior to use, the gold film and silicon wafer were immersed in a mixture of concentrated sulfuric acid and hydrogen peroxide for 15 min, then rinsed by water and dried with nitrogen. The clean gold film and silicon were immersed in 10⁻⁶ mol/L R6G solution for 1 h–2 h and then rinsed with ethanol. After drying, the Au@SiO₂ NPs monolayer film was transferred onto it.

Experimentally, it is difficult to precisely control the distance between the Au NPs. In addition, the Raman signal will be distorted when the probe molecules directly contact bore Au NPs. Therefore, In order to analyze the effect of core size on the SERS signal, Au@SiO₂ NPs with 2 nm–3 nm in shell thickness, and ca. 15-nm, 25-nm, and 50-nm core sizes are prepared, respectively, the transmission electron microscopy (TEM) images are shown in Fig. 1(A). Then a monolayer film of Au@SiO₂ NPs was deposited on the silicon substrate surface after rinsing with ethanol (see the previous section). Figure 1(B) shows the SERS spectra (laser: 632.8 nm, 20 mW) of R6G with different core sizes of Au@SiO₂ NPs. Encouragingly, the SERS signal of R6G molecules adsorbed at the gap between Au@SiO₂ NPs (with ∼ 50 nm core/shell 2 nm–3 nm) exhibits a notable Raman enhancement. The enhancement signal is much higher than those of R6G molecules adsorbed at the gap between Au@SiO₂ NPs (core size 15 nm) and (core size 25 nm) referred to as spectra (curve a) and (curve b). It is apparent that the SERS signal intensity is attributed to the increase of the particle size.

Fig. 1.

	Figure Option ViewDownloadNew Window
	Fig. 1. TEM images of Au@SiO2 NPs with core sizes of (curve a) 15 nm, (curve b) 25 nm, and (curve c) 50 nm, and corresponding SERS spectra of Rh6G molecule adsorbed in the gap between Au@SiO2 NPs. Laser: 632.8 nm, 20 mW.

In our experiment, the Au NPs with 25-nm average diameter have much better monodispersity. In order to verify the effect of the interparticle spacing on the same particle size, the Au@SiO₂ NPs with core size 25 nm and different thicknesses are prepared as shown in Fig. 2(A) and Fig. S1. For a monolayer film of Au@SiO₂ NPs prepared by the above method, its Raman spectra (laser: 632.8 nm, 20 mW) are shown in Fig. 2(B). When comparing the samples, the SERS signal of R6G molecules adsorbed at the gap between Au@SiO₂ NPs (shell thickness 2 nm–3 nm), referring to spectrum (curve d), shows the largest enhancement. The enhancement signal is much greater than those of R6G molecules adsorbed at the gaps between Au@SiO₂ NPs with a shell thickness of 15 nm and those with shell thickness 35 nm, referring to spectra (curve c) and (curve b). However, for Au@SiO₂ NPs with a shell thickness of 50 nm, the enhancement signal still exists, which shows that a long tail effect occurs in the induced EM field,^[20] referring to spectrum (curve a) and the peaks in Fig. S5.

Fig. 2.

	Figure Option ViewDownloadNew Window
	Fig. 2. SEM images of Au@SiO2 NPs with shell thickness of (a) 50 nm, (b) 35 nm, (c) 15 nm, and (d) 3 nm, and corresponding SERS spectra of Rh6G molecule adsorbed in the gap between Au@ SiO2 NPs. Laser: 632.8 nm, 20 mW.

In the coupled NP system, an Au@SiO₂ nanosphere dimer is chosen to simulate the above SERS experiment and calculate the EM field distribution by the finite different time domain (FDTD) method. The calculation results show that the particles’ size and the distance between the NPs play vital roles in the Raman signal enhancement. As shown in Figs. 3(a)–3(c) and Fig. S3, where the 633-nm laser is used to simulate calculation, with the diameter of Au@SiO₂ NPs fixed to be 50 nm, the smaller the gap size of Au@SiO₂ NPs dimer is, the larger the maximum EM coupling intensity reaches. The reason is that the aggregation occurs because the NPs are very close to each other, and the absorption wavelength of the NPs is changed accordingly. When the absorption wavelength of the aggregated NPs is the same as the laser wavelength, the surface plasma resonance (SPR) strength will increase significantly, meanwhile, EM field coupling intensity becomes stronger. In addition, as shown in Figs. 3(b) and 3(c), and Fig. S4, when the gap is 4 nm, the EM field intensity is attributed to the increase of the particle size. Because the SPR is related to the charge density, the larger the size of Au NPs, the more free electrons on the surface of Au NPs and the stronger the SPR becomes, meanwhile, EM coupling becomes stronger. Moreover, the SPR peaks of the Au NPs are red-shifted obviously with NPs size increasing, and the size is larger, the red shift is more obvious. The appropriate laser wavelength will cause a stronger SPR between the large Au NPs dimers. Ultimately, we reach the conclusion that the larger the particle size and the narrower the gap, the stronger the EM field coupling intensity is. The results show that theoretical calculations are in good agreement with the experimental results.

Actually, very strong EM field is only observed in “hot spots” in junctions between two particles. As is well known, the maximum EM field is tightly confined near the NPs surface. Only the molecules which are absorbed on or very close to the gap can have a giant enhancement from SERS. The signal enhancement is determined by the NP size, the gap, and the position of the probe molecules in the “hot spots”. In addition, only the averaged surface enhancement factor (ASEF) can veritably and fully reflect the overall enhancement of the substrate. A simple calculation is offered to explain ASEF, only a few molecules (ca. 2%) in “hot spots” contribute to 98% of the total Raman signals, the average of all molecules Raman signal from the nanoparticle surface is still weak. How to obtain the strongest Raman signal and much more “hot spots” in the metallic nanostructures is still an enormous challenge in experiment.

Fig. 3.

Figure Option ViewDownloadNew Window

Fig. 3. EM field distributions of Au@SiO2 dimer, simulated by FDTD method. Directions of incidence and polarization of the 633-nm excitation laser are shown in the figure. The thickness of the silica shell is 2 nm, and the diameters of Au@SiO2 NPs are (a) 15 nm, (b) 25 nm, and (c) 50 nm. The diameter of the Au core is 25 nm, the thickness values of the silica shell are (d) 2 nm, (e) 4 nm, and (f) 8 nm.

In practice, glass sheets and silicon wafers are widely used as the supports of metal NPs. The FDTD method is used to simulate the EM field distribution of the Au@SiO₂ NPs on different substrates (parameters: diameter d_Au = 50 nm, shell thickness = 2 nm, λ = 633 nm). As shown in Fig. 4(a), for the Au@SiO₂ NPs dimer on silicon wafer, the presence of a dielectric material (silicon) around the NPs will lead to a shift of SPR peaks of metal NPs, but will not cause an obvious change of the EM field distribution.^[21] However, if a metal substrate is used, as shown in Fig. 4(b), for the Au@SiO₂ NPs dimer on a Au film, the large EM coupling is due to the “hot spots” transferred from the NP–NP gap to NP–surface gap, meanwhile, the more “hot spots” that occur, the stronger the EM coupling intensity that is obtained.

Fig. 4.

	Figure Option ViewDownloadNew Window
	Fig. 4. FDTD simulations showing the EM field distributions of Au@SiO2 dimers on gold film (a) and silicon substrate (b).

In order to verify the theoretical calculation and simulation accuracy, experimentally, the monolayer film of Au@SiO₂ NPs of ca. 50-nm diameter is transferred on the Au film and silicon substrate after rinsing with ethanol. The scanning electron microscopic images are shown in Figs. 5(a) and 5(b) and Fig. S2. Then the configuration is immersed in an ethanol solution containing R6G molecules. The Raman spectra (laser: 632.8 nm, 20 mW) are shown in Fig. 5(c), and we obtain SERS signal from the R6G molecules adsorbed in the gap between the NPs (b₁ and a₂ curves in Fig. 5(c)). It can be seen that the Raman enhancement properties of those two configurations are almost the same, showing that no matter what substrate is used, the EM coupling between the NP–NP gap still exists. Results show that experimental results are in good agreement with theoretical calculations. However, as can be seen in Fig. 4(b), for Au substrate, the EM coupling occurs not only in the NP–NP gap but also in the NP–surface gap. In order to further verify this simulation calculation, a Au@SiO₂ NPs monolayer film is transferred on the Au film and silicon substrate, separately. The Raman spectra are shown in Fig. 5(c), the SERS signals are only detected by molecules adsorbed on the Au film surface (b₂ curve in Fig. 5(c)). But the signal intensity is much higher than that of the EM coupling between the NP–NP gaps. This can be well explained by the fact that if Au@SiO₂ NPs disperse on the silicon substrate, the EM coupling only occurs in “hot spots” in junctions between Au NPs. Due to the fact that the presence of a metal material will change the charge arrangement of NPs, the interaction between laser and dimer, the dipole occurs, and because the free electrons on the NP surface oscillate along the dimer axis, a large number of free electrons are focused on the interparticle spacing, resulting in a large amount of positive charge at the bottom of the NPs. Therefore, a number of electrons are induced on the gold film surface by electrostatic effect. Then, the redistribution of dipole will occur. Ultimately, the large EM coupling leads to the “hot spots” transferred from the NP–NP gaps to NP–surface gaps. In addition, the light which is reflected from the metal surface will increase the probability of resonance. This new type of coupling not only increases the EM coupling intensity, but also obtains more “hot spots”. Meanwhile, this coupling mode has higher sensitivity and stability, higher ASEF than the coupling mode which occurs between Au NPs on silicon substrate (Figs. S6 and S7).

Fig. 5.

	Figure Option ViewDownloadNew Window
	Fig. 5. SEM images of Au@SiO2 NPs monolayer film on the silicon (a) and gold film (b). (c) Scheme for SERS detection with R6G molecules adsorbed in the gap between Au@SiO2 NPs and Si (curve a1), R6G molecules adsorbed in the gap between NPs (curve a2), R6G molecules adsorbed in the gap between NPs (curve b1), and R6G molecules adsorbed in the gap between NPs and Au film (curve b2).