1. Introduction

Metallic nanostructures typically comprised of plasmonic metals such as gold or silver have provided a wealth of interesting physical phenomena, with both surface plasmon polaritons (SPPs) providing a mechanism for propagating optical photons within extreme sub-wavelength nanostructures and localized surface plasmons (LSPs) that exhibit extremely large localized electromagnetic (EM) fields capable of enhancing optical processes such as absorption or scattering cross-sections. One such process is the surface-enhanced Raman scattering (SERS) effect, initially reported by Fleishman [1] and later explained by Jeanmaire and van Duyne [2], whereby the large EM fields induced by the resonant optical stimulation of plasmonic nanoparticles leads to an approximate E⁴ enhancement of the Raman scattering due to the increase in absorption of the incident laser light and in the scattering cross-section of the adsorbed molecule. Work by Stockman et al. [3] illustrated that not only can large, localized EM fields be induced at such isolated nanoparticles, but in addition, when two or more particles are formed in an aggregate or placed within a few nm’s of one another, the EM fields induced between the particles can be amplified by several orders of magnitude [4,5]. These ‘hot-spots’ provided the means for single-molecule detection via SERS [6,7] and stimulated an entire field of study attempting to engineer ideal SERS substrates featuring high concentrations of these ‘hot-spots’ [8], with some of these attempts focused on aggregation of colloidal nanoparticles [9–12], arrays of pointed triangles through nanosphere lithography [13,14] and more recently work focused on random collections of gold colloids separated from an underlying gold film by dielectric spacers [15] and on periodic arrays of vertically-oriented holes that were lined with silver that featured sub-10 nm gaps [16]. Perhaps the tightest interparticle separations fabricated reproducibly to date were reported by Chen et al. [17] where the authors highlighted the capability to fabricate arrays of gold nanoparticles surrounded by alkanethiolate molecules of variable alkane chain lengths enabling interparticle gaps down to approximately 2 nm. However, these gaps were attained only with a surface coating of a self-assembled monolayer, therefore removing this approach from most SERS or fluorescence based sensor applications. In addition, in this and all other cases to date, these systems consisted of either randomly oriented or confined structures, thereby eliminating the potential for large-area periodic arrays of near-field coupled plasmonic structures where cooperative plasmonic oscillations could occur. Primarily, this difficulty has come about due to the limits of current lithography in writing periodic arrays of small metal features with inter-particle separations in the sub-10 nm regime. Furthermore, all but the latter approach result in pseudo-2D architectures, whereby high plasmonic fields are induced between the particles, but are confined to very tight spatial locations in the dimension perpendicular to the surface. By expanding into a 3D architecture, such as nanowires [18–20], nanorods [21,22] or the aforementioned vertically-oriented holes [16] a sensor would stand to gain the advantage that such large near-field coupled plasmonic fields would be distributed over the length of the structure involved, thereby increasing the likelihood that a molecule of interest would fall within the 'hot-spot' region. This effect is illustrated in Fig. 1 , where the spatial distribution of the calculated SERS enhancement (E⁴) is plotted for two 50 nm diameter silver (a) spheres and (b) cylinders that are spaced 5 nm apart. Note that in the case of the cylinders, while about a 20x reduction in peak enhancement is observed, an elongation of the plasmonic fields along the entire 300 nm long nanowire occurs. Thus, in the case of chemical sensing applications, combining such a large SERS-active area with the large enhancements attained through near-field plasmonic coupling, is highly desired. In this work, we demonstrate such an advancement through the fabrication and characterization of large (20-60 μm on a side), square arrays of Si nanopillars conformally coated with Ag with interparticle separations down to <2 nm. The arrays with the tightest interpillar separations, once treated with a self-assembled monolayer (SAM) of thiophenol, were observed to exhibit greater than an order of magnitude increase in the SERS intensity at 532, 633 and 785 nm incident with respect to those arrays with broader interparticle separations, thereby indicating that near-field plasmonic coupling is likely leading to these dramatic enhancements of the SERS response. These structures were enabled through the development of plasma enhanced atomic layer deposition (PEALD) of Ag [23,24], whereby conformal monolayers could be deposited on our periodic arrays of Si nanopillars with interparticle separations well within the range of standard e-beam lithography (>50 nm). This conformal coating was controlled at monolayer accuracy, therefore filling the gap between adjacent pillars two monolayers at a time until the desired interparticle separation is achieved. In principle, by initially measuring the dimensions of the periodic nanostructure used prior to Ag deposition, the monolayer accuracy of the PEALD process enables reproducible and controllable interparticle separations limited only by the statistical deviation in the fabricated nanoparticles size. Therefore, interparticle separations <10 nm can readily be achieved. Furthermore, because of the conformal nature of PEALD it is possible to create arrays of Ag-coated cylinder structures that will have long interaction regions extending along the entire nanopillar length similar to the model presented in Fig. 1(b).

 

Fig. 1 COMSOL simulations of two 50 nm diameter Ag (a) spheres and (b) 300 nm long nanowires, separated by 5 nm in air. The plots designate the predicted SERS enhancement (E⁴) as a function of position within the structures.

Download Full Size | PDF

2. Experimental

2.1 Sample preparation

For the work reported here, a combinatorial approach was implemented. A series of periodic arrays of 100x100 Si nanopillars each were fabricated using electron beam lithography and reactive ion etching. These arrays featured nanopillar diameters and interpillar separations ranging from ~70 to 288 nm and ~85 to 298 nm, respectively. The full description of the Si nanopillar fabrication can be found elsewhere [21]. Following formation of the ~300 nm tall nanopillars, the sample was over-coated with a conformal Ag PEALD film within a Beneq TFS 200 ALD Reactor using the HS500 solid-source cell [15]. The Ag(fod)(PEt₃) (fod = 2,2-dimethyl-6,6,7,7,8,8,8-heptafluorooctane-3,5-dionato) precursor was heated to 106°C and was alternated with hydrogen plasma cycles to strip the exposed ligands from the adsorbed Ag precursor. The PEALD cycles were repeated 500 times, resulting in an approximately 45 nm thick Ag film as analyzed from the increased diameter of the pillars. To ensure that no residual precursor residue persisted after completion of the ALD process, x-ray photoemission spectroscopy (XPS) was used and indicated that the as-deposited films contained <1% fluorine and phosphorus, the two primary contaminants from the precursors that would be expected. Furthermore, the SERS spectra did not indicate the presence of any molecule other than thiophenol. Once deposited, the sample was immersed in a 1x10⁻³ M solution of thiophenol in ethanol for approximately 18 hours to form the SAM that was used as the probe molecule in the SERS studies [21,25]. Presented in Fig. 2(a) is a 5x optical reflection image of the full combinatorial set of arrays after the Ag PEALD coating, with the final diameter and interpillar gaps following the Ag deposition provided in the label. As shown in the 50 kX SEM image of such a coated structure (imaged at 45 degrees) presented in Fig. 2(b), the as-deposited PEALD silver films are conformal, of uniform thickness and completely coat the Si pillar. However, a ‘mosaic-tile’ structure is observed where individual Ag islands meet. Plan-view 50kX SEM images of 196 nm diameter nanopillars with pre-Ag interpillar gaps of 85, 109, 222 and 298 nm are presented in Figs. 2(c)–(f), respectively, following Ag deposition. Note that the 45 nm thick Ag deposition was sufficient to reduce the gaps between nanopillars such that the two closest arrays (85 and 109 nm predeposition separation) were reduced to interpillar gaps of <2 and 16 nm, respectively, with both of these separations being within the range where near-field plasmonic coupling is anticipated [3,4,8]. A 100kX SEM image of the tightest-spaced array is provided in Fig. 2(g) to enable clear imaging of the film morphology and the interpillar separation.

 

Fig. 2 (a) 5x magnification optical reflection image of each of the arrays studied within this work. In this image, each square is a single 100x100 nanopillar array, with the corresponding nanopillar diameters and gaps provided on the axes. (b) A 65 kX magnification SEM image of Ag PEALD coated Si nanopillar arrays collected at 45° to illustrate the nanopillar structure and PEALD film morphology. (c)-(f) 50 kX magnification SEM images of ~200 nm diameter Ag PEALD coated, Si-nanopillars with interpillar gaps of 196, 124, 16 and <2 nm, respectively. A 100kX magnification image of the tightest spaced array is presented in (g).

Download Full Size | PDF

2.2 SERS measurements

In order to monitor the effective EM field intensities within these arrays, SERS measurements from the SAM of thiophenol on the Ag PEALD surface of these nanopillar arrays was carried out. SERS measurements were performed using the 532, 633 and 785 nm modules of a DeltaNu ExamineR μ-Raman system. Acquisition times used depended upon the array being measured and the incident wavelength, but they ranged from 0.1 to 2 s, with laser powers ranging from 2.2 to 8.2 mW. The neat Raman spectra were collected using the liquid sample cell on the microscope at the same power and several acquisition times to ensure reproducibility. All enhancement factors were calculated via the method outlined in Ref. [21] using the intensity of the 998 cm⁻¹ C-H wagging vibrational mode of the thiophenol molecule (marked by black arrow in Fig. 3 ). This mode was chosen because the C-H groups are spatially dislocated from the sulfur, which bonds to the Ag surface, thereby ensuring that any change in molecular polarizability induced due to the bonding, would have minimal impact upon the calculated enhancement [21,26,27].

 

Fig. 3 Neat Raman spectra of thiophenol (black trace), SERS spectra collected from the Ag PEALD film without nanopillars (green trace), and on arrays of ~200 nm diameter nanopillars with interpillar gaps of 198 (blue trace), 52 (light-blue trace) and <2 nm (red trace) gaps. Each spectra was normalized to account for both the incident laser power and corresponding acquisition time, while the corresponding number of molecules probed in each measurement is shown in the legend. The arrow in the figure denotes the position of the 998 cm⁻¹ mode (C-H wag) used in the enhancement factor calculations and in the SERS spatial plots presented in Fig. 4. Inset: Semi-logarithmic plot comparing the SERS spectra from the <2nm gap arrays and the neat spectra.

Download Full Size | PDF

3. Results and discussion

Presented in Fig. 3 are Raman spectra for a neat solution (black trace) of thiophenol in comparison to SERS spectra from the thiophenol SAM bound to the Ag PEALD films without nanopillars (green trace, ‘off-arrays’), and from three nanopillar arrays with diameters of ~200 nm and interpillar gaps of 198 (blue trace), 52 (light blue trace) and <2 nm (red trace)collected using 532 nm incident. In all cases, the Raman/SERS intensities were corrected for the incident laser power (3.6 mW) and acquisition time (5s for the Neat and off-array measurements, 0.5s for the 52 and 198 nm gap arrays and 0.1 s for the <2 nm gap array) to allow for them to be compared on the same scale. Note that in all cases the normalized SERS intensity was in excess of the Neat Raman spectral intensity, this despite the seven orders of magnitude difference in the number of molecules probed between the two measurements (1.0-5.2x10⁷ for SERS vs. 2.5x10¹⁴ for the neat Raman). A full description of this calculation can be found in a previous article from our group [21]. As shown in Fig. 3, the SERS spectra exhibited a broad fluorescence background that is not observed within the neat Raman spectra. As reported previously, this background is due to hybridized electronic states that are induced into the HOMO-LUMO gap of the thiophenol molecule upon binding of the sulfur to the Ag surface [18]. These hybridized states induce deep energy levels within the visible spectral range, thereby making the normally transparent thiophenol molecule emit visible light. To further illustrate the large SERS enhancements created via the plasmonic coupling in the tightest-spaced ‘hot-spot’ arrays, a semilog plot comparing the background corrected SERS spectra with the neat Raman spectra is presented in the inset of Fig. 3, clearly illustrating the 1-2 orders of magnitude increase in signal intensity from the SERS hot-spot arrays, despite the aforementioned >10⁷ difference in the number of probed molecules. In addition, over an order of magnitude increase in the SERS intensity is observed from the near-field coupled ‘hot-spot’ arrays in comparison to those with wider gaps. For these ‘hot-spot’ arrays, the total number of photons incident upon the sample within the acquisition time of the measurement was on the order of 9x10⁷, indicating that approximately two photons were required to get a Raman scattering event from each of the molecules (5.2x10⁷), thus implying that such systems are in principle sensitive enough for single molecule detection over large-area samples. Furthermore, as discussed in our previous work [21], the standard deviation in the SERS intensity across the array was less than 30% in all cases, with 15% variation being typical, thereby illustrating the large, reproducible distribution of plasmonic ‘hot-spots’ created across the array. Overall, we calculated SERS enhancement factors of 2x10⁸, 7.4x10⁷ and 1.7x10⁸ from the best performing near-field coupled arrays at 532, 633 and 785 nm, respectively, using the C-H wag mode at 998 cm⁻¹. In addition, a non-zero SERS intensity was detected from the PEALD Ag film in the absence of nanopillars (green trace, Fig. 3), presumably due to surface plasmons stimulated within the mosaic morphological features. However, the intensity of this response was 2-3 orders of magnitude reduced with respect to the arrayed nanopillar structures and thus would not be expected to influence the measured enhancements taken from the nanopillars structures, especially since the behavior of the EM fields as a function of interparticle spacing is a fundamental property of near-field plasmonic coupling and therefore is independent of the plasmonic material used.

The dependence of the SERS intensity as a function of interparticle gap at 532 nm incident for all nanopillar diameters explored is presented in Fig. 4(a) . In all cases a slow, decrease in the SERS intensity is observed as the interpillar separation is reduced from 200 nm, consistent with our previous results for Au-coated nanopillar arrays. Some variability in this decrease is observed, with these deviations potentially being due to variations in far-field diffractive coupling by the 2D nanopillars gratings and the variations between data sets due to the diameter-induced variations in the plasmonic response [21]. However, once the interpillar gap was reduced to ~16 nm, a large increase in the SERS intensity was observed, independent of diameter. In all cases, this increase was in excess of an order of magnitude, with a maximum twenty-four-fold increase observed from the largest diameter arrays. This increase in SERS intensity at such tight interpillar gaps is consistent with the creation of large near-field interparticle coupling between adjacent nanopillars and may imply long-range, cooperative near-field plasmonic coupling.

 

Fig. 4 (a) SERS intensity measured at 532 nm incident as a function of interparticle gap at each diameter as indicated in the figure. (b) Corresponding COMSOL simulations (red open squares; line provide as guide to the eye) of the coupling-induced enhancement of the SERS response from semi-infinite, periodic arrays of Ag-coated Si nanopillars as a function of interpillar gap. All data points are normalized to the SERS intensity of the array with the widest separation (210 nm gap) simulated. For comparison, the experimental normalized SERS enhancement results from the 279 nm diameter nanopillars are also provided (blue diamonds).

Download Full Size | PDF

In an effort to quantify the impact that plasmonic coupling would have within such a large-area periodic structure, COMSOL calculations of the predicted SERS enhancement at the center of a semi-infinite array of 150 nm diameter Si nanopillars over-coated with a conformal and continuous 45 nm thick Ag film were carried out and are presented as a function of interpillar gap in Fig. 4(b). The values presented are all normalized to the calculated SERS enhancement (E⁴) for the array featuring the widest interparticle gap (210 nm). This normalization is provided to clearly illustrate the role of the near-field coupling effects. Similar to the experimental results, the simulations indicate relatively little change in the SERS enhancement as the interpillar gap is reduced. Upon decreasing the gap below 50 nm, a dramatic increase in the SERS enhancement is observed. This increase continues as the gap is reduced further down to 2 nm, which was the tightest gap simulated. Plotted along with the simulations is the corresponding, normalized experimental data collected from the arrays with 279 nm diameter nanopillars (post-Ag deposition). As shown, the qualitative agreement between the generalized behavior of the experimental data with the simulations illustrates the fact that the SERS enhancement reported from the arrays with the tightest gaps is clearly due to the large collective EM fields that are induced due to near-field coupling from neighboring nanopillars. However, in contrast to the experiments, almost three orders of magnitude increase in the SERS enhancement is predicted from the simulations, although it is likely that this discrepancy is due to the inability of the simulations to correctly model the complex Ag PEALD morphology and/or due to capillary effects limiting thiophenol from completely coating the Ag nanopillar walls within the arrays with the tightest interpillar gaps.

Another key property of interparticle plasmonic coupling is the red-shift in the peak of the LSPR in addition to the increase in the amplitude discussed previously [3,4,8]. In order to maintain a constant resonant wavelength that is required to match the incident laser line for the optimal SERS response, any coupling-induced red-shift requires a compensating blue shift. Such a blue-shift can be attained through shrinking the nanopillar diameter. Thus, one would anticipate that at the onset of near-field plasmonic coupling that the optimal SERS response would shift to a smaller nanopillar diameter. This is observed in Figs. 5(a) and (b) where spatial plots of the SERS intensity as a function of both post-Ag diameter and interpillar gaps are presented for 532 and 785 nm incident light, respectively, (633 nm incident measurements were omitted for brevity). In these plots, each pixel corresponds to a given 100x100 nanopillar array, with the interpolation leading to the smoothed plots provided through Microcal Origin. Two very different dependences are observed at these two wavelengths. For the 785 nm incident measurements, the largest enhancements are found at a Ag-coated nanopillar diameter between 326 and 353 nm for arrays where the interpillar gap was larger than 35 nm, however, as the gap was reduced into the realm where near-field plasmonic coupling was observed, this optimal SERS enhancement shifted towards smaller nanopillar diameters, consistent with a plasmonic-coupling induced red-shift. On the contrary, at 532 nm incident, in all cases the most intense SERS response was observed at the tightest gaps, as illustrated in Fig. 4(a) and no distinct diameter dependence is identified. Thus, from the two spatial plots presented in Figs. 5(a) and (b) it can be ascertained that at 532 nm, the LSP resonance occurs at a diameter outside of the range explored here, and therefore any red-shift in the LSPR due to near-field plasmonic coupling would have minimal impact on the detected SERS response and only the increase in the overall amplitude would be observed. Further, as the enhancements from the off-resonant 532 nm measurements approached that of the on-resonant 785 nm values, it is anticipated that careful optimization for an on-resonant structure within the near-field coupling regime, would lead to significantly larger enhancements within the green region of the EM spectrum.

 

Fig. 5 Spatial plots of the SERS intensity measured as a function of nanopillar diameter and interpillar gap at (a) 532 and (b) 785 nm incident. The values plotted correspond to the average SERS intensity of the C-H wag mode of thiophenol (998 cm⁻¹) from a given array after being normalized to account for the laser power and acquisition time. All values are presented in units of countsW⁻¹s⁻¹.

Download Full Size | PDF

4. Conclusion

In summary, we report on the application of the plasma-enhanced atomic layer deposition (PEALD) of Ag on silicon nanopillar templates to achieve large-area, periodic ‘hot-spot’ arrays featuring tightly spaced (<2 nm) near-field coupled plasmonic nanopillars. This technique has enabled the fabrication of structures up to 60 μm on a side with interparticle gaps ranging from 196 down to < 2 nm. Using surface-enhanced Raman scattering (SERS), we have illustrated that the plasmonic coupling between these tightly-spaced plasmonic nanopillars leads to over an order of magnitude increase in the SERS intensity in comparison to similar nanostructures with interpillar gaps beyond the near-field coupling regime (>20nm). These measurements have shown average enhancement factors up to 2x10⁸, 7.4x10⁷ and 1.7x10⁸ from the best performing near-field coupled arrays at 532, 633 and 785 nm, respectively. To the best of our knowledge, this is the first demonstration of periodic arrays of plasmonic nanoparticles with consistent interparticle gaps down to 2 nm and based on the relaxed lithographic requirements for the fabrication outlined here, such structures should open the door to highly sensitive SERS and fluorescence sensors, as well as providing the potential for enhanced emitters and absorbers.