1. Introduction

Metal nanoparticles (MNPs) can have surface plasmon resonances that enhance the local electric field leading to surface enhanced Raman spectroscopy (SERS) [1–8]. MNPs offer added functionality in many applications, for example, with Raman bio-labels. These labels can be functionalized and bound to a target, such as cell surface markers, to provide a high-degree of multiplexed detection [7,8]. MNPs can be defined lithographically [9–11] or grown in sollution [12]. The wet chemical approach allows for substantial shape control, and even single-crystal structures with lower scattering losses. Compared with randomly generated SERS substrates, the SERS response from MNPs is reliable and reproducible [3].

Despite the many advantages of MNPs, the SERS enhancements reported so far have been typically orders of magnitude less than randomly roughened substrates or aggregates [13,14]. Even reports of 10⁹ enhancement factors in core-shell structures are still three orders of magnitude smaller than generally accepted for single-molecule Raman demonstrations on random structures [3]. Therefore, a general means to increase MNP SERS by orders of magnitude would be a transformative step for the field.

Previous works show that a dielectric spaced metal layer could enhance the SERS signal for evaporated silver islands about 10-fold by tuning their plasmon resonance frequency [15–19] and changing the local density of states [20–24]. However, further improvements are required to boost the SERS to the level of the randomly roughened surface case and ultimately to detect the single molecule Raman signal [1,2,25].

In this paper, we use the colloidally synthesized silver nano-prisms on top of a gold ground plane spaced by a TiO₂ dielectric layer to coherently enhance the SERS signal of rhodamine 6G (R6G). Over 50 × SERS enhancement is achieved. Theoretical calculations and finite difference time domain (FDTD) simulations verify the experimental results and indicate more room for further SERS amplification with this configuration.

2. SERS measurement with multilayer substrates

2.1 SERS experimental setup

Figure 1(a) shows the schematic of the multilayer SERS substrate. An optically thick 100 nm Au layer was used as a ground plane (EMF Corp.). This was coated with a TiO₂ spacing layer evaporated by 7.5 kV electron beam source in an Angstrom Engineering physical vapor deposition system. The TiO₂ layer refractive index,$n_d$, was measured to be 2.19 via white light reflection measurements. The purpose of the spacer layer was to tune the phase of the reflected light from the gold mirror as a function of thickness, t. For each thickness, we fabricated three different samples to ensure reproducibility.

 

Fig. 1 Silver nano-prisms over the multilayer SERS substrate. (a) Schematic of silver nano-prisms on TiO₂ spacer layer over optically thick Au layer, where t is the thickness of TiO₂ and d is the side length of a nano-prism. The illumination pattern is not to scale and the actual experiment has ~30 MNPs within the focus. (b) The SEM of the multilayer SERS substrate surface. The inset shows a TEM image of a single silver nano-prism.

Download Full Size | PDF

Silver nano-prisms were synthesized in water by white-light assisted conversion of spherical nanoparticles [12]. This yielded an ensemble of prisms with average length on a side, d, of 80 nm, shown as the inset in Fig. 1(b). A solution containing silver nano-prisms and dye were drop-cast (0.02 mL) on the substrates, where the concentration of the R6G dye was 1 μM. The sample was then allowed to dry for 5 hours. Ultra-pure water with a resistivity of 18.2 MΩ cm (from Barnstead NANOpure Diamond water purification system) was used throughout the experiments.

The Raman spectra of the dye were taken using a Renishaw inVia Raman microscope with a 785 nm diode laser of 0.5 mW power illumination and an estimated density of 30 nano-prisms within the laser focus, as determined by scanning electron micrograph (SEM) studies of the surface – shown in Fig. 1(b). The backward Raman scattered light was collected by a 20× objective (NA = 0.4) with a total integration time of 30 s. All the measurements were repeated at least 4 times in each experiment and all the experiments were repeated on several different days in order to ensure the consistency and the stability of the results.

2.2 Extinction spectrum of silver nano-prisms

The spectral dependence of the plasmon resonances were examined in solution, because of their relevance to SERS [26,27]. Figure 2 shows the extinction spectrum (Cary 5 UV-VIS-NIR Spectrophotometer) of the nano-prisms used in the experiment. Three extinction peaks were observed at 337 nm, 413 nm, and 673 nm, of which the 673 nm peak has the strongest extinction.

 

Fig. 2 Extinction spectrum of the silver nano-prisms used in the experiment in an aqueous environment, where the 673 nm extinction peak is clearly visible.

Download Full Size | PDF

2.3 Theoretical calculation on phase reflection

The primary objective of this experiment was to find the optimized dielectric spacer layer thickness for SERS enhancement. To coherently enhance the SERS, the reflected light from the ground plane should constructively interfere with the incident light beam. Upon reflection at the interface of a perfect electric conductor (PEC) and a dielectric, there is a ${180}^{\circ }$ phase shift of the electric component. The optimal thickness is corresponding to the in-phase reflection configuration which is expected to be [28]:

2.4 SERS measurement results

Figure 3(a) shows a sample SERS spectrum from the multilayer SERS substrate. Three Raman shift peaks were observed, which are 1312 cm⁻¹, 1364 cm⁻¹, and 1509 cm⁻¹, respectively. The full analysis was performed on the 1509 cm⁻¹ peak because no deconvolution was necessary (although the other peaks showed the same general enhancement behavior).

 

Fig. 3 Experimental SERS spectra. (a) An example Raman spectra for the R6G dye using the silver nano-prisms. (b) Enhancement of SERS using silver nano-prisms for the 1509 cm⁻¹ Stokes shift peak as a function of dielectric layer thickness, normalized by the SERS signal from a bare glass substrate. The blue bands indicate the first order and the second order SERS enhancement peaks.

Download Full Size | PDF

Figure 3(b) shows the enhancement of SERS using silver nano-prisms as a function of dielectric layer thickness. The enhancement is with respect to a bare glass substrate with the same drop-casting of silver nano-prisms and dye. The uncertainty for each thickness was calculated from the standard deviation from at least four SERS measurements at different locations on the sample. Furthermore, the measurements were repeated on two additional samples, each with the same thickness, showing the same results. It can be seen that the enhancement factor changes with thickness variation, and the peak enhancements (45.4 ± 1.6 and 51.6 ± 4.7) were achieved when the TiO₂ equaled 40 nm and 200 nm, respectively. The uncertainty in these values comes from standard deviation over multiple measurements over randomly distributed nano-prisms. The difference in the thickness between the two peaks is 160 nm, which is close to the theoretical prediction of 179 nm. Also, the values are offset from the prediction of Eq. (1), which will be discussed further below.

To ensure the generality of the enhancement for different MNPs, we repeated the experiments using Au nano-rods [29] instead of Ag nano-prisms, and similar enhancement factors and dielectric thickness dependencies were observed (not shown).

3. FDTD simulation results

In the experiment, the situation is complicated from the simple picture presented above by other factors such as the absorption of the metal ground-plane, coupling into the modes of the finite dielectric layer underneath the MNPs, multiple reflections by the dielectric layer and ground-plane and a finite collection aperture. In addition, the finite wavelength difference between the excitation and Stokes wavelengths should be considered. For a more comprehensive understanding, we used FDTD numerical analysis for comparison with experiments.

The Au permittivity values were taken from a previous work [30], and the Ag permittivity values were taken from a different work [31], and these are two commonly used references for those materials, respectively. A 2 nm mesh was also used around the silver nano-prism to ensure that plasmonic effects were accurately captured, as verified by convergence studies. The simulation region was enclosed with perfectly-matched layer boundaries in the direction perpendicular to the gold ground plane, and with 200 nm periodic boundaries in the directions parallel to the ground plane. The theoretical results were invariant to changes in periodicity. A plane wave was injected from the top of the structure.

Figure 4 . shows the resulting simulation of the SERS signal and its TiO₂ thickness dependence. To obtain the theoretical enhancement factor, we compare the near-field intensity of the nano-prism above the Au ground using different TiO₂ layer thicknesses to the control where the nano-prism was placed directly on a glass substrate without the Au ground. The SERS intensity $I_{sers}$ is proportional to the localized electric field intensity both at the excitation wavelength $E_{\text{ex}}^2$ and the Raman wavelength $E_{Raman}^2$ [32,33]:

 

Fig. 4 Finite difference time domain simulations of enhancement factor, for 80 nm side nano-prism in the same configuration as in Fig. 3(b).

Download Full Size | PDF

Figure 5 . shows the electric field intensity around a 80nm prism (the average size), the dielectric layer and the reflector (outlined with dashed white lines) in the xz-plane. The local field intensity in the resonant cases of 80 nm and 260 nm thicknesses are one order of magnitude larger than the off-resonant case (160 nm). This results from the constructive and destructive interference of the image excitation created by the Au ground plane reflector.

 

Fig. 5 Simulated local electric field intensity distributions close to a nano-prism for varying dielectric thicknesses (t = 80 nm, 160 nm, 260 nm) shown on a logarithmic scale. The dashed lines show the interfaces of the silver nano-prism, the dielectric layer and the gold ground plane.

Download Full Size | PDF

4. Discussion

The experiment and calculations give comparable enhancement factors and spacer layer thickness dependencies. Aside from the finite penetration of light into the metal, the additional offset with respect to the theoretically expected optimal spacer layer thickness values remains an uncertainty in the experiment, but can be attributed to (at least in part) dye accumulation beneath the MNPs, off-axis excitation by the focusing objective and uncertainty in the dielectric thickness. The SERS signal peak values of experiment results are close to those given by the FDTD simulation. In the measurements, we measured locations of the sample where there was no obvious aggregation, as observed under the optical microscope. Even so, some aggregation may be present, and we cannot accurately capture that effect within our simple simulation

Moreover, it is possible to envisage more advanced multilayer SERS schemes [34–36], such as a right corner reflector, which in the optimal geometry leads to an enhancement factor of 4 in terms of the localized field around the nano-prisms. As compared to the present system, the flat ground plane only has an enhancement factor of 2 [37]. Considering the SERS signal intensity is approximately proportional to the fourth power of the localized field [32,33], then 16 times greater SERS signal enhancement is expected for the corner reflector substrate. Even more advanced schemes, such as the Yagi-Uda shaped MNPs, may be implemented to increase the enhancement still further. Based on these considerations, it is expected that at least 3 orders of magnitude enhancements in the MNP Raman should be possible through configuration optimization. Such boosts in the electric field could make single-molecule Raman demonstrations viable with MNPs [1,2,25].

5. Conclusion

We have demonstrated that the combination of simple substrate engineering and silver nano-prisms can enhance SERS by a factor of 50. Both the experiment and the theoretical calculations give comparable enhancement factor dependence on the dielectric spacer layer thickness. Similar results were also observed for Au nano-rods which indicate this multilayer substrate is a generic approach to boost the SERS signal for different MNPs grown in solution. FDTD simulations also verified this SERS enhancement quantitatively. With more advanced schemes utilizing corner reflectors or Yagi-Uda antennas, it is expected that MNP SERS can be enhanced by 3 orders of magnitude to the regime of signal molecule detection. Obvious benefits will arise from this sensitivity boost for many applications including the use of MNPs as Raman biolabel markers, the development of more reliable Raman-enhanced templates, and the improvement of Raman-based pathogen sensors.