1. Introduction

The plasmonic-enhanced metallic nanostructured substrates, which are capable of inducing localized surface plasmons (LSPs), have drawn a lot research attentions lately, because of their potential applications in optoelectronic devices, plasmonic crystals, nanolithography, subwavelength imaging and biomolecular detection [1–5]. Through various lithographic or synthetic approaches, the size, the shape, the compositions and the inter-particle spacing of the metallic nanostructures could be engineered to specifically enhance a local electromagnetic (EM) field of LSPs allowing ultra-sensitive detection [4–6]. Among the LSP based chemical- and bio-sensing techniques, the surface enhanced Raman scattering (SERS) has long been used to investigate the structural information of molecules adsorbed on the surfaces [7–9]. However, the cross section of Raman scattering is extremely small (typically about 10^-28 - 10^-30 cm²/molecule). Only those molecules adsorbed on the noble metals, such as gold and silver, are enhanced both electromagnetically and chemically to produce reasonable Raman signal for analytically purpose [5, 10]. SERS technique was not used in the ultra-sensitive detection until the discovery of the unusual large Raman cross sections for the molecules adsorbed on the aggregates of nanoparticles [11, 12]. It has been claimed that the cross section of Raman scattering could be enhanced up to 10¹⁴ on the so-called “hot-spots”. Later studies have suggested that the molecules in the junctions between nanoparticles, whose separation was about 1 nm, could exhibit unusual large Raman cross section allowing single molecular detection through the excitation of LSPs [13]. Since the discovery of the single molecule SERS (SM-SERS), a lot of research efforts have been focused on the development of ultra-sensitive chemical- and bio-sensors based on such “hot-spots” concept [5, 9, 14]. However, such “hot-spots” are rare in the SERS substrates, because the field enhancement is very sensitive to the relative position of the molecules in the “hot-spots” [15], which makes it very difficult to fabricate a reproducible SERS substrate for ultra-sensitive detection.

In the past few years, many lithographic approaches have been utilized to fabricate periodic particle array to obtain reproducible SERS active substrates with optimal SERS signal [5, 16, 17]. Among these techniques, nanosphere lithography developed by Van Duyne et al [18] has been very successful in preparing reliable SERS active substrates. It has been demonstrated that the silver film over nanosphere (AgFON) substrates were capable of providing reproducible Raman signal allowing rapid detection of glucose and anthrax [19, 20]. However, in a recent photochemical hole burning (PHB) study, it was reported that the site enhancement distribution for benzenethiol molecules on the AgFON substrates was highly inhomogeneous [21]. It was observed that only a small fraction (63 ppm) of molecules on the surfaces contributed to 24% of the overall SERS signals. It is counterintuitive to imagine that the distribution of the site enhancement on such type of regular nanostructures could be so inhomogeneous while providing reproducible SERS signal. Where are the locations of the “hot-spots” in the regular nanoparticle arrays? Can we increase the number of the “hot spots” in the regular nanoparticle arrays? The answers to these questions may lead us to design an optimal SERS substrate, which can produce strong and reliable SERS signal for ultrasensitive detection. To explore these questions, it requires the investigation of the spatial distribution of these “hot spots” and the topographic information around them. So far, the best spatially resolved Raman images were obtained by tip-enhanced Raman scatting, which has been demonstrated capable of measuring single molecular Raman scattering with 10 nanometer resolution [22]. However, the presence of the tip induces additional enhancement contribution, which may distort the enhancement site distribution on the nanostructured surfaces. An alternative approach to reveal the spatial distribution of Raman signal is the confocal Raman microscopy (CRM), which is capable of mapping the Raman signal with sub-wavelength spatial resolution. When the CRM is combined with high resolution microscopic tools such as atomic force microscopy (AFM) or scanning electron microscopy (SEM), both optical and topographic information of the nanostructured samples can be obtained on the same area. In this article, we report the investigation of the spatial distribution of the SERS site enhancement on the AgFON substrates by a combination of CRM, AFM and SEM, and the design of an optimal SERS substrate for rapid chemical and bio-sensing.

2. Experimental section

To measure the spatial distribution of the site enhancement on the AgFON substrates, the AgFON substrates were prepared following the procedure described previously [18]. In short, monolayers of 300 nm close packed polystyrene nanoparticles (Bang’s Lab) coated with 150 nm thin silver film were used in this experiment. A combined CRM and AFM (alpha 300, WITec Instruments Corp., Germany) was used to record the SERS images and the topographic images of the AgFON substrates. The wavelength of the excitation laser (Ar⁺ laser, Melles Griot, U.S.) was 488 nm and the laser power was around 19 μW. A 100X objective (Nikon) with a numerical aperture of 0.9 was used to focus the laser beam into a 0.4 μm spot and the Raman signal was collected through a 25 μm fiber. To eliminate the complication in measuring the enhancement factor, a monolayer of benzenethiol (Aldrich), which is known to exhibit minimum chemical and resonance enhancement at the laser wavelength [21], was prepared by dipping the AgFON substrates into 10^-4 M benzenethiol ethanol solution for 4 hours. To compare the Raman spectra, the intensity of Raman signals from different samples were normalized to 0.2 s, which was used in the Raman imaging experiment.

3. Results and discussion

Shown in Fig. 1(a) are the Raman spectra of benzenthiol molecules from the neat liquid and the AgFON surfaces prepared by 300 nm polystyrene nanoparticles. The Raman images from the AgFON surfaces were calculated by integrating SERS signal of the 1575 cm^-1 peak. A typical Raman image from AgFON substrate is shown in Fig. 1(b) where the scanning area is 10 × 10 μm² (150 × 150 pixel). It is obvious that the spatial distribution of SERS signal is very inhomogeneous. When we look closely into the spatial distribution of the Raman signal, it can be clearly seen that most of the strong SERS signals are located in the defects of the close packed nanospheres as measured in the topographic image (Fig. 1(c)). To calculate the enhancement factor, the Raman signals were normalized to the neat benzenethiols assuming that the packing density of the benzenethiol on the AgFON substrate was 6.8×10¹⁴ cm^-2 [23]. The ratio of the benzenethiol molecules within the confocal volume of the neat liquid to the benzenethiol molecules in the focused spot on the silver film was calculated to be 2000. The calculated distribution of the enhancement factors is depicted in Fig. 1(d). The averaged enhancement factor for the 300 nm AgFON substrates was measured to be 1.9 ± 1.0 × 10⁶, which agrees with a previous measurement within experimental error [21]. Despite of the fact that the spatial distributions of SERS signals from the AgFON substrates were very inhomogeneous, the averaged SERS intensity over the scanned area remained roughly the same for different samples. One simple explanation for such reproducible SERS intensity from an inhomogeneous enhancement distribution can be attributed to the relative constant density of defects in a monolayer of close packed nanospheres where the SERS signals are concentrated.

 

Fig. 1. (a) Raman spectra of benzenethiol molecules from neat liquid and AgFON surface. Black spectrum is the Raman spectrum from neat benzenethiol liquid. Exposure time 100 s. Red line is an averaged Raman spectrum on the AgFON substrate over 10 × 10 μm area (150×150 pixel). Exposure time: 0.2 s. A typical Raman spectrum on the hot site with an enchantment factor of 10⁸ is depicted in blue line. Exposure time 0.2 s. (b) SERS image of the 1575 cm^-1 peak for benzenthiol from the 300 nm AgFON substrate. Bar: 2 μm. (c) AFM image of the AgFON substrate. Bar: 2 μm. (d) Distribution of the measured SERS enhancement factor log(∣E∣⁴).

Download Full Size | PDF

The contribution to the overall SERS intensity from various sites is listed in Table 1. As we can see from the Table 1, the maximum measured enhancement was around 2 × 10⁹, which was one order of magnitude less than those measured in PHB measurement (4 × 10¹⁰) [21]. The spatial distribution for the site enhancement was also very inhomogeneous where the hot-sites (with a enhancement factor larger than 10⁸) only occupied less than 0.3% of the scanned area, however, they contributed to 27.5% of the overall SERS signal. One reasonable explanation for the lower maximum observed enhancement factor is that the population of such hot site is very rare. Therefore, the chance for finding such hot site in a 10 × 10 μm² area is very low. Another reason for the lower enhancement factor at the hottest site may be due to normalization. Since the spatial resolution for our measurement was about 0.4 μm, we could not distinguish how many molecules contributed to the Raman signal within the detection spot. Therefore, an average number of 3 × 10⁶ beneznethiol molecules in this area was used to calculate the enhancement factor. It is know that there are only a few molecules present in the hottest spot. If that is the case, the observed enhancement factors for the hottest sites could be several orders of magnitude larger than the number listed in Table 1.

Table 1. Distribution of the enhancement factors for benzenethiol molecules on the 300 nm AgFON substrate.

View Table

Knowing that the hot sites for the SERS signals are located along the defect lines between the close packed nanoparticles, one may ask what kind of topographic arrangement in these defects that leads to large SERS enhancements. Previous SERS experiments on the aggregates of nanoparticles have demonstrated that the “hot spots” were located in the junctions between nanoparticles [13]. Where are these junctions on the AgFON substrate? One may speculate that the junction could exist between close packed nanoparticles or the edge of the close packed nanoparticles (defects). To explore the location of junctions, we have investigated the SERS signal from benzenethiol molecules on the silver film over isolated nanoparticles or a line of close contacted nanoparticles, which was achieved by removing the polystyrene nanoparticles in the close packed structures with a PDMS stamp containing microchannels. When the PDMS stamp was in contact with the close packed polystyrene nanoparticles, nanoparticles would attach to the stamp surface leaving arrays of polystyrene nanoparticles with the shape of microchannel. By repeating this procedure once and slightly shifting the stamp, it was possible to obtain isolated nanoparticles or a line of close contacted nanoparticles as shown in Fig. 2. For visualization purpose, 460 nm polystyrene nanoparticles were used in this experiment. From the Raman images, it can be clearly seen that the enhancement of the Raman signal is concentrated on the edge of nanoparticles, not between nanoparticles. For isolated nanoparticles, the SERS signal is around the nanoparticle. Therefore, we conclude that the large enhanced sites are located between the nanoparticles and the flat area. The topographic information on these sites was examined by the cross-sectional SEM image as shown in Figs. 2(e), and 2(f). It is evident that the silver film was divided into two parts forming junctions at the touching edges of silver films over the nanoparticles and on the flat area. If the hot spots are located at the gap between the edge of the silver films over the nanospheres and on the flat area outside of the nanospheres, there may exist an optimal thickness for SERS enhancement. The SERS signals of benzenethiol molecules were measured on the 460 nm AgFON substrates with different silver film thickness as shown in Fig. 3. It was observed that the SERS signal peaked at 230 nm, which was equal to half of the diameter of the nanoparticles. This result is not surprising, since the bottom part of the silver film could reach the edge of the silver film over the upper part of nanospheres when the deposition thickness equals to the radius of the nanosphere. To simulate the field enhancement in such junctions, the electromagnetic (EM) field on single nanosphere of AgFON substrate was calculated via the Fullwave simulation software (RSoft Design Group, Inc.) based on the finite-difference time-domain (FDTD) method. The diameter of polystyrene sphere and thickness of silver film was 460 nm and 230 nm, respectively. The junction between silver film and nanosphere covered with silver film was set to 1 nm. The refractive indexes of silver, polystyrene sphere, and silicon substrate used in the simulation were 0.131 + 2.81j, 1.59, and 4.367 + 0.079j, respectively. The size of uniform spatial grid and temporal step (cT, c and T are optical velocity and simulated time step.) in FDTD simulation were set to 0.2 nm and 0.000141 μm. Figure 4 shows the EM field distribution of single nanosphere on the silicon substrate excited by different polarization of incident plane wave at a wavelength of 488 nm. If the polarization is parallel to the incident plane (TM mode), the Raman enhancement factor at the gap between silver films on the top of polystyrene and on the flat area was calculated to be large than 10⁸ through the excitation of the localized surface plasmon (LSP). However, if the polarization is perpendicular to incident plane (TE mode), the enhancement in the EM field is very low. Our simulation result indicated that up to 10⁸ enhancement factor for SERS signal can be achieved when the distance between the edges of silver films was 1 nm.

 

Fig. 2. (a), (c) Confocal Raman images of benzenethiol on the silver film over isolated nanospheres (460 nm) and a line of close-contacted nanospheres. (b), (d) SEM images of the nanosphere in the same area. Scale bar 500 nm in (a) and 1 μm in (b)(c)(d). (e), (f) Cross-sectional SEM image of silver film over nanospheres. Bar: 500 nm.

Download Full Size | PDF

 

Fig. 3. The measured SERS intensity for the 1575 cm^-1 peak on the 460 nm AgFON substrates as a function of film thickness.

Download Full Size | PDF

 

Fig. 4. The calculated SERS enhancement distribution log(∣E∣⁴) around the nanosphere for different polarizations. (a) The simulated model of the single nanosphere. (b) TE mode. (c) TM mode and (d) Enlarged view of field at the edge.

Download Full Size | PDF

Knowing that the SERS signals are concentrated on the edges of the nanoparticles where the nanoparticles are not in close contact, it should be possible to increase the SERS signal by producing non-contacted nanoparticle arrays. One simple approach to produce such type of nanoparticle arrays is to trim the contacting edge of polystyrene nanoparticles by oxygen plasma. It has been shown that such process can be used to produce size tunable nanopillar arrays [24]. Shown in Fig. 5 are the SERS and SEM images of the size trimmed nanoparticle arrays. The diameters of the polystyrene nanoparticles were reduced from 460 nm to 360 nm and 300 nm. These results clearly demonstrated that the area percentage of the SERS signal with enhancement factor larger than 10⁶ can be increased from around 10% to about 90% of the overall area. When compared with the conventional 460 nm AgFON substrates, the SERS intensity on size trimmed nanoparticle array with optimal silver film thickness was improved by ten times. Our result indicates that it is possible to fabricate reproducible substrates with optimal SERS signal by controlling the silver film thickness and the defect density.

 

Fig. 5. (a), (c) The SERS images of the 1575 cm^-1 peak for benzenethiol on the 300 nm and 360 nm size trimmed nanoparticle array substrates, respectively. (b), (d) the SEM images of the nanoparticle array substrates. Bar: 2 μm.

Download Full Size | PDF

4. Conclusions

In summary, we have investigated the spatial distribution of the Raman site enhancement on the AgFON substrates. It was observed that the site enhancements were highly inhomogeneous on the AgFON substrates. When the SERS images were combined with AFM and SEM images, it was found that the hot sites were located on the defects between close packed nanoparticles. Further study revealed that the hot sites were formed in the gaps between the edges of the silver films over the nanoparticles and on the surrounding flat area. As a result of geometric consideration, the maximum Raman signal can be obtained by controlling the film thickness and the defect density. Our results clear demonstrate that it is necessary to investigate the spatial distribution of the Raman scattering and the topographic information on the nanostructured surfaces for fabricating reproducible SERS substrates with optimal signal.