1. Introduction

Since recently, biosensors based on surface-enhanced Raman scattering [1] have gained increasingly relevance because this method has become one of the key tools for the detection of dangerous diseases pathogens in biomedical analyzes [2–7].

Traditionally, Raman spectroscopy has been widely used for visible wavelengths of laser excitation (450-750 nm) not only because of the availability of corresponding optical sources and detectors but also due to the deeply physical dependence of the Raman intensity on the fourth power of excitation frequency. A drawback of using shorter excitation wavelengths is however a typical problem associated with a strong background fluorescence as well as a quick degradation of organic molecules due to the pronounced light absorption. A natural way of avoiding these complications is to use longer excitation wavelengths in the infrared region [8]. However, for the particularly useful approach of enhancing the Raman scattering by the presence of the metal interface (SERS, surface enhanced Raman scattering) there appears another fundamental limitation arising from the increased detuning from the surface plasma resonance. Despite the sharp decrease of the surface plasma waves attenuation [9,10], the inherent property of a tight electromagnetic field localization close to the flat metal surface becomes steadily lost at lower operating frequencies. A widely recognized way of overcoming this limitation is an engineering of more complex metal-dielectric objects which support modified plasma excitations at the designed frequencies [11]. Recently we demonstrated how a metal-insulator-metal structure can extend the operating range of the traditional SERS-active layer of a thin silver film into the red part of the visible spectrum [12], however this approach was not successful for significantly longer wavelengths. Nanoscale engineering of the resonant metal antennas is another well-known approach [13], but it still lacks in terms of an affordable mass-scale production. It should be additionally stressed that manufacturing of SERS-substrates designed for longer wavelengths could potentially be less demanding in terms of the needed spatial resolution and this can significantly reduce their production cost.

Therefore, one of the most urgent needs of biotechnology is a development of stable and affordable SERS-active structures designed for the near infrared operation region and enabling a maximum Raman signal enhancement [13–15].

It was suggested earlier that periodically arranged optical resonators can realize an additional enhancement of local electromagnetic fields in SERS-active structures [16,17]. Most experimental studies in this direction dealt with nano-structured objects with period values much smaller compared to the light wavelength [18–20]. Previously we reported the design and properties of combined metal-dielectric resonant nanostructures for SERS at visible wavelengths [21,22]. The periodic arrays of dielectric pillars with a height of 10 nm-200 nm were created on Si/SiO₂ substrates by electron beam lithography and plasma etching to have a planar size a and a period 2a in the range a = 50-1500 nm. The structure was then complemented by a subsequent deposition of a relatively thick metal layer. It was demonstrated that for a fixed pillar height the Raman enhancement factor reveals prominent oscillations in dependence on the array period. The commensurability between the wavelength of the pumping laser and the pillar array period determined the optimal enhancement conditions. The spectral measurements were then carried out for 488–580 nm laser excitation wavelengths and provided a ground for the advancement into the region of longer excitation wavelengths.

2. Experimental

A thermally oxidized silicon substrate (oxide thickness 1200 nm) was patterned by a combination of an electron beam lithography and plasma etching. It represented 140 active fields with size of 50 × 50 microns separated by flat inactive fields of the same size. Each active field had rectangular SiO₂ pillars of height h, planar size a and period (a + d), where d is the gap between the pillars. The planar size a of pillars was varied from 100 nm to 620 nm. The gap between the pillars d was changed from 40 nm to 2070 nm. As a result, the ratio γ = a/(a+ d) varied from 0.23 to 0.71 for different active fields. The pillar height h value was changed for different structures from 100 nm to 600 nm. Both active and inactive fields were then uniformly covered by a silver layer with a thickness of t from 20 nm to 120 nm with the help of resistive evaporation.

Investigation of the spatial distribution of the Raman scattered intensity was carried out using a scanning Raman microscope allowing a spatial resolution up to 1 micron. However, the optimum spot diameter of the focused laser beam was set to 10 microns along with a scanning step set to 10 microns also. It enabled acquiring a large enough amount of data for reliable averaging of the Raman signal from both the active and inactive fields during a reasonable time period.

A typical studied SERS-structure is displayed in Fig. 1 under various magnifications. The upper part of the figure shows one of the SERS-active fields containing 51 × 51 pillars. In the lower part of the figure one can see individual pillars (a = 760 nm, d = 510 nm, h = 250 nm, t = 40 nm). For the studied metal thickness values exceeding 10 nm it was confirmed by SEM that metal surface has no separated metal islands.

 

Fig. 1. SEM pictures of a typical SERS structure active fields under various magnifications. In the lower part a = 760 nm, d = 510 nm, h = 250 nm, t = 40 nm.

Download Full Size | PDF

3. Results and discussion

Several organic substances (4-aminobenzeniol, thiophenol, R6G, adenine and acetaminophen) were used as analytes for studying the SERS effect at a laser excitation wavelength of 1064 nm. It was found that the bulk and surface enhanced Raman scattered spectra basically coincided for all substances used. This fact allowed for the most direct comparison of the scattered signal intensities and obtaining reliable results for the magnitude of the Raman signal enhancement. In Fig. 2 Raman spectra of thiophenol (C₆H₆S) at a 1064 nm laser excitation are compared for the case of bulk substance and when a drop of a very dilute thiophenol solution (concentration of 10⁻⁷) has dried on the studied SERS substrate.

 

Fig. 2. Raman spectra of thiophenol (C₆H₆S), measured for a 1064 nm laser excitation for the case of (a) bulk liquid thiophenol (100% concentration), and (b) when a drop of a thiophenol solution with a concentration of 10⁻⁷ has dried on a designed SERS substrate. The low energy tail in SERS-spectrum originates from the silicon substrate supporting the active layers.

Download Full Size | PDF

It is clearly seen that the two Raman spectra perfectly match and demonstrate the same Raman modes and comparable signal levels despite the remarkably lower analyte concentration for SERS measurements. In this particular case, the spectrum was obtained from an active field with the parameter values a = d = 500 nm, h = 200 nm, t = 40 nm that afterwards proved to be non-optimal. To find the maximum of the SERS enhancement, the design parameter space was scanned in the directions of changing values for a, d, h, t. The color-plot dependence of the Raman signal intensity on parameters a and γ is displayed in Fig. 3.

 

Fig. 3. The color-plot dependence of the Raman signal intensity measured for different active fields and averaged inside a selected field, on parameters a and γ. Brighter color corresponds to the increased intensity.

Download Full Size | PDF

Experiments clearly reveal two maxima of the Raman enhancement each corresponding to the specific value of the structure period (a + d). For the first maximum (mode 1) this value is close to 1000 nm and thus it matches the laser wavelength in vacuum. One more enhancement maximum occurs at (a + d) = 600 nm (mode 2) and this value is close to the laser wavelength in the SiO₂ material. This additional resonance condition resembles a secondary peak in the transmission spectra of quartz substrates supporting a silver metal layer perforated by a square lattice of holes [23].

Figure 4 shows the dependence of the SERS signal intensity on period (a + d), measured for various parameters γ. It follows from this plot that two SERS intensity maxima are again observed and their positions are weakly dependent on γ. The ratio of enhancements for modes 1 and 2 is however somewhat changing with parameter γ.

 

Fig. 4. The dependence of the SERS signal intensity on the period (a + d) for various parameter γ values.

Download Full Size | PDF

Furthermore, the resonant condition for the ratio between the SERS structure period and the laser wavelength is not a single prerequisite for the optimal Raman enhancement. It follows from Fig. 3 that when moving along the direction (a + d) = 1000 nm or (a + d)  =600 nm the absolute maximum of SERS signal is realized for a certain the value of the parameter γ, and deviation from this optimal value leads to a significant decrease of the Raman scattering.

The dependence of the SERS signal on the parameter γ value for both modes 1 and 2 is displayed in Fig. 5. The striking observation is that, contrary to the hot spot conception suggesting small gaps and thus γ parameter values approaching 1 for maximum SERS enhancement, the optimal enhancement condition corresponds to the γ value about 0.5 for both modes. More rigorously, it corresponds to γ = 0.42 for the mode 1 independently of the height of the dielectric pillars, and to γ = 0.5 for the mode 2.

 

Fig. 5. The dependence of the SERS signal intensity on the parameter γ for both modes 1 and 2.

Download Full Size | PDF

A discrete set of silver metal thickness values t = 10, 20, 40, 60, 80, 100 and 120 nm was investigated. It was found that the most prominent SERS enhancement takes place at t = 80 nm. To determine its absolute magnitude, the standard procedure was used [15]. According to it, the absolute value of the SERS enhancement is defined as [I_SERS/N_SERS]/[I_bulk/N_bulk], where I_SERS and I_bulk are the intensity of the Raman signal from the SERS-substrate and from the bulk material accordingly, and N_SERS and N_bulk - the number of molecules in the laser beam for each case.

First, the Raman signal intensity was measured from the bulk liquid samples with the concentration of thiophenol changing by several orders of magnitude. The linear relationship between the intensity of the Raman scattered signal and thiophenol concentration was confirmed. Then it was found that the bulk Raman signal falls below the measurement noise at thiophenol concentrations of about one percent. Since the volume illuminated by the laser beam is known one can calculate the total number of probed molecules. In Fig. 6 dark symbols show a measured linear dependence of a bulk Raman signal on the thiophenol concentration along with the calculated number of thiophenol molecules in a probed volume (the length of the probed area was set to the depth of focus). The dependence of Raman signal intensity from the homogeneous SERS substrates of a large size (5 mm x 5 mm) with non-optimized and optimized designs on the thiophenol concentration was then obtained. A specified volume of the corresponding solution was evaporated from the substrate and formed a barely visible circle with a diameter of about 2 mm.

 

Fig. 6. The SERS signal intensity dependence on thiophenol concentration measured on large size substrates (5 × 5 mm) for optimized (yellow circles) and non-optimized (squares) designs. Black circles show a measured linear dependence of the Raman signal on the bulk thiophenol concentration. The upper axis demonstrates the calculated number of the optically probed thiophenol molecules.

Download Full Size | PDF

The surface density of thiophenol molecules can be calculated in this case. Figure 6 demonstrates that with the proposed SERS substrate design it is possible to measure the Raman signal from thiophenol solutions with concentrations down to 10⁻⁹ and below. This finding clearly suggests the exceptional enhancement of the Raman signal at the excitation wavelength of 1064 nm. Accurate measurements show that at optimal parameter values a = 420 nm, d = 590 nm, h = 260 nm, t = 80 nm the absolute SERS enhancement is as large as 2·10⁸. It should be noted that further optimization of the metal coating through the metal layer combinations of different thicknesses could allow further advancing in the scattering enhancement and this research direction is in progress now.

The full theoretical modeling of the proposed SERS-active design is very challenging due to the comparable pillar/metal dimensions and the light wavelength and therefore goes beyond the scope of this report. The obtained SERS enhancement exceeding 2·10⁸ at 1064 nm excitation looks unexpectedly high, however this finding can be in part explained by a frequency dependence of the metal permittivity. As briefly mentioned in the introduction, the surface plasma waves attenuation actually decreases for both silver and gold when the frequency is lowered. This parameter is determined by the generic plasmon quality factor Q = ɛ₁²/ɛ₂ [14] where ɛ₁ and ɛ₂ are real and complex parts of the dielectric permittivity of the metal which depend substantially on the frequency [10]. For example, for gold at a laser wavelength 532 nm ɛ₁ = −4.7, ɛ₂ = 2.4, therefore the corresponding quality factor is around 9. At the same time, for a laser wavelength of 1064 nm ɛ₁ = −48, ɛ₂ = 3.6, and this ensures the quality factor improvement of about 70. Since the SERS enhancement is proportional to the square of the quality factor when the frequency is close to the fundamental surface plasma resonance [14], one could expect an increase in the local electric field enhancement for a gold film by up to three orders of magnitude when the laser wavelength is changed from 532 nm to 1064 nm - but only when the structure is designed to support the ‘spoof’ plasma resonance at this operation frequency. This improvement in the plasma wave properties can easily compensate a 16-fold decrease in the Raman scattering probability as mentioned in the introduction. Similar considerations for a silver material show that in this case a local electric field increase by more than 2 orders of magnitude could be expected. An especially notable experimental finding is a requirement for the structure period to be comparable to the excitation wavelength for the maximum SERS enhancement. It suggests that the incident electromagnetic wave diffraction is intimately related with the initial conversion into surface plasma modes and the subsequent back-conversion of the Raman scattered components [16]. Finally, the pillar height appears to be most responsible for the spoof plasmon resonance position as its optimal value of 260 nm closely resembles the quarter of the laser wavelength [24].

4. Conclusions

To summarize, the periodic dielectric structures coated with a thick layer metal were shown to support the surface enhanced Raman scattering (SERS) with more than eight orders of magnitude signal enhancement at a laser excitation wavelength 1064 nm. The SERS-substrate operation in an IR range has become feasible thanks to the excitation of the ‘spoof’ plasmon-polariton excitations at the spatially modulated metal surface. The experimentally found optimal parameters for an absolute maximum enhancement are as follows: (a) the period of the structure is approximately equal to the wavelength of the laser excitation; (b) the ratio the planar size of the oxide pillar to the structure period is about 0.42; (c) the oxide pillar height is 260 nm; (d) the silver metal thickness is 80 nm. These parameters can be easily implemented with the least-demanding sub-micron photolithography techniques and this ensures the direct applicability of the suggested design for the mass scale production.