1. Introduction

Surface plasmon resonance (SPR) biosensors have become a central technology for the investigation of biomolecular interactions and have been increasingly applied for the detection of chemical and biological species [1]. Although most SPR sensors are based on surface plasmons propagating along continuous metal films, in recent years metallic nanoparticles and nanostructures have been increasingly exploited [2]. Metallic nanoparticles have been used in conventional SPR sensors as amplification agents [3] or colloidal aggregation based biosensors [4]. Moreover, localized surface plasmons (LSP) on metallic nanoparticles and nanostructures have been used in LSP-based biosensors [5,6]. In spectroscopic LSP-based biosensors, refractive index changes induced by biomolecular interactions are measured by tracking the resonant wavelength in the spectrum of scattered or transmitted light [7]. LSP-based biosensors provide a tool for probing localized effects on the scale comparable with the dimensions of individual molecules [8]. This high localization is associated with the profile of the electromagnetic field of LSPs [9]. For instance, small spherical nanoparticles (<100 nm) support LSPs with electromagnetic fields localized within a few tens of nanometers from the surface of the nanoparticle. For spheroidal or nanorod particles the electric field is more enhanced at the tips of the nanoparticle [10] and an increase in the particle aspect ratio results in an increase in the LSP sensitivity to refractive index changes [7]. Therefore nanorod particles offer higher sensitivity and better performance comparing to spherical nanoparticles [11,12]. Extremely high enhancements of electromagnetic fields were achieved in very small volumes at sharp edges and tips of isolated nanoparticles such as nano-pyramids [8]. Confinement of the LSP field in gaps between metallic particles [11,13,14], or chains of coupled nanoparticles [15] represents a convenient approach to controlling the spatial distribution of the electromagnetic field of surface plasmons. An enhancement of the sensitivity of LSP-based biosensors due to selective binding of target molecules to areas with concentrated electromagnetic field has been recently demonstrated [16] and the dependence of the sensitivity on the distance from the nanoparticle surface has been studied experimentally [17].

In this paper we experimentally map the sensitivity to refractive index changes localized on the surface of nanoparticles supporting LSPs (local sensitivity). Specifically, we investigate nanostructures consisting of an array of metallic nanorod chains on a planar dielectric substrate. We measure the LSP resonance shift in response to localized refractive index changes to characterize the sensitivity to a local variation of the refractive index.

2. Nanoparticle array

Arrays of chains of metallic nanorods were fabricated by electron beam lithography using the positive resist poly(methyl methacrylate) (PMMA) on glass substrates covered with 10 nm of indium-tin-oxide (ITO). Following exposure, the PMMA layer was developed and the substrates were coated with a gold layer by means of thermal evaporation. A 0.5 nm thick chromium layer was used to promote adhesion of gold to the substrate. The preparation of the nanoparticle arrays was finished by lift-off in acetone.

The nanoparticle arrays were characterized using scanning electron microscopy (SEM). The inset in Fig. 1 shows a part of a resulting nanoparticle array. The thickness of the gold layer of 30 nm was measured using a stylus profilometer (Alphastep, Tencor Instruments). Nanoparticle dimensions were measured from SEM image to 35 nm x 75 nm. The gap between neighboring nanoparticles in one chain was 25 nm. The SEM image allowed measuring all dimensions with the precision of around 2 nm. The size of each array was 50 μm × 50 μm and up to 81 arrays were prepared on each substrate.

 

Fig. 1 Extinction spectra of the nanostructure measured in air and in water. The inset shows an SEM image of the nanoparticle array.

Download Full Size | PDF

Microspectrometry employing a conventional optical microscope connected to a spectrometer was used to measure extinction spectra from an area with a diameter of 10 μm. White light from a halogen lamp was polarized along the axis of the nanoparticle chain and the spectrum of the light transmitted through the nanoparticle array was measured. Figure 1 shows extinction spectra of the nanostructure in contact with air and water, respectively. The sharp extinction peak corresponds to the excitation of LSPs. Changing the medium surrounding the nanoparticles from air to water (Δn = 0.32 RIU – refractive index units) results in a shift in the extinction peak wavelength of 45 nm. This yields a sensitivity to bulk refractive index (RI) changes in the area surrounding the nanoparticles (excluding the substrate) of 140 nm/RIU.

A theoretical analysis of the equivalent periodic array of gold nanoparticles on an ITO coated glass substrate was performed using the finite-difference time-domain (FDTD) method (FDTD Solutions, Lumerical Solutions, Inc., Canada). The bulk refractive index sensitivity as calculated from the theoretical model was 201 nm/RIU. The difference in the theoretical sensitivity and experimental result is believed to be in a large part due to the uncertainties of the geometry and material parameters of the structure assumed in simulations.

3. Sensitivity to partial coating of nanoparticles

Next, the LSP change associated with a change in RI taking place at different areas of the nanoparticle surface was characterized and its contribution to the bulk RI sensitivity was investigated. A periodic stripe mask made of PMMA with the same periodicity as the metallic nanoparticle array was used to cover specific areas of the nanoparticle surface. First, the nanoparticle array was coated with a 60 nm thick PMMA layer. Then the mask was fabricated in a second step of electron beam lithography and the alignment with the first array was achieved using a set of alignment marks. Reference areas comprising (i) nanoparticle arrays completely coated with the PMMA mask and (ii) uncoated nanoparticle array with lithographically completely removed PMMA were prepared on each tested substrate. The width of the stripes was controlled by varying the exposure dose during lithography and their lateral positions were centered either on nanoparticles or in the gap between nanoparticles.

Figure 2 shows SEM images of (a) an array of nanoparticles with the mask covering centers of nanoparticles (“center mask”) and (b) an array of nanoparticles with the mask covering the tips of the nanoparticles and the gaps in between (“gap mask”). Center masks were prepared on 27 arrays, gap masks were prepared on 27 arrays and 27 arrays were used for the preparation of reference areas. Although the contrast of the dielectric mask is lower in the SEM image, it allows for measuring the position relative to the nanoparticle array with a precision of around 2 nm. The widths of the PMMA stripes was either 45 nm or 60 nm (Fig. 2 shows stripes with a width of 60 nm).

 

Fig. 2 SEM images of nanoparticle arrays covered with PMMA masks: (a) center mask, and b) gap mask. The upper right parts of the images show the original (maskless) nanoparticle arrays for comparison. The dimensions of the nanoparticles are 35 nm × 75 nm, the width of one mask stripe is 60 nm.

Download Full Size | PDF

Figures 3(a) and 3(b) depict the extinction spectra corresponding to different positions of the mask on the array in comparison with the reference measurements on bare and fully covered samples. The shift in the position of the resonance peak between the nanoparticle array without the mask and the nanoparticle array completely coated with PMMA was 68 nm. Considering the refractive index of PMMA as 1.49 (according to manufacturer specification and literature [18]), the deduced refractive index sensitivity of 138 nm/RIU is consistent with the bulk refractive index sensitivity for water as the surrounding medium. This result confirms that the thickness of the PMMA layer exceeds the range of LSP sensitivity which agrees well with previous experimental and theoretical studies [17].

 

Fig. 3 Extinction spectra of the nanostructure coated with PMMA masks at different positions (solid black curve). Experimental results are plotted in (a) for a 45 nm wide center mask and (b) for a 60 nm wide gap mask; SEM images of the respective configurations are attached. Results of FDTD simulations are plotted in (c) for a 50 nm wide center mask (d) for a 50 nm wide gap mask. Extinction spectra corresponding to reference arrays are shown for comparison (i) nanoparticles array completely coated with the mask (blue dash-and-dot curve) and (ii) nanoparticle array without the mask (red dashed curve).

Download Full Size | PDF

For a center mask of 45 nm width the extinction peak shifts by Δλ = 15 nm (Fig. 3(a)) which is only 23% of the shift due to the completely coated array. A gap mask of the same width yields the peak shift of 43 nm or 65% (data not shown). Figure 3(b) shows the effect of the 60 nm gap mask resulting in a shift of Δλ = 50 nm which is 75% of the total shift due to the complete coating. These results indicate that the sensitivity to RI changes is localized mostly around the nanoparticle tips and the interparticle gaps. The 60 nm center mask yields the peak shift of 33 nm or 50% (data not shown). For each mask position, 20 arrays were measured of which at least 15 arrays were found to exhibit the peak shift within 1 nm from the average value. The decrease in the maximum extinction for center mask (Fig. 3(a)) is believed to be caused mostly by the variability of the mask/nanostructure alignment across the measurement area. This is supported by the observation that the highest dispersion in the peak position occurred for the 45 nm center mask.

The FDTD approach was used to simulate changes in the extinction spectra due to stripe masks covering complementary areas of nanoparticles. Both center and gap masks were assumed to have a refractive index of 1.49, a width of 50 nm, and a layer thickness of 60 nm. Structures without a mask or completely coated with a PMMA film were modeled for reference. The results of the simulations are plotted in Figs. 3(c) and 3(d). The maximum shift of the LSP peak of 100 nm corresponds to the mask covering the whole nanostructure surface. Nanostructures coated with center mask and gap mask correspond to LSP wavelength shifts of 27 nm and 74 nm, respectively. The width of masks of 50 nm for both positions of the mask used in the theoretical model, allowed comparing peak shifts corresponding to complementary areas of nanoparticles. When these two shifts are summed up, the resulting value is very close to the shift corresponding to the completely coated nanoparticle (difference less than 0.5%).

The experimental results obtained with 45 nm and 60 nm wide masks confirmed quite well the additive effect of local refractive index changes. If the 5 nm overlap of 60 nm and 45 nm masks is considered negligible, the combined peak shift due to the complementary positions of the gap and center masks results in 98% and 115% of the experimental peak shift due to the completely coated nanoparticle.

A change in the mask position out of the nanoparticle symmetry resulted in an increasing difference between the sum of partial mask peak shifts and the peak shift for the completely coated nanoparticle array. Consistently with the experiment, simulations involving a mask aligned asymmetrically with the nanoparticle array (covering the area between the center of the nanoparticle and the center of the gap) resulted in a extinction peak shift of 38 nm, rather than 50 nm (half of the sensitivity of the completely coated array). This difference is probably associated with the more substantial field redistribution of the LSP due to asymmetrically aligned coating.

In summary, the local sensitivity of LSP on the nanoparticle array can be reconstructed from the contributions of local refractive index changes to the LSP peak shift; however, changes in RI should be distributed on the nanostructure in such a way, that the symmetry of the nanostructure is maintained.

4. Deconvolution of local sensitivity

The shift in the extinction peak was normalized to the width and RI of the mask. Horizontal bars depicted in Fig. 4(a) (grey) indicate different positions of the mask and corresponding peak shift (after the normalization). The peak shift due to mask positions at different overlapping intervals allowed for the deconvolution of the local sensitivity. This deconvolution from the experimental results relied on iterative numerical fitting to a set of integral conditions in Matlab (MathWorks, Inc.). The conditions used for the fit included: (i) the integral across each interval equals the corresponding normalized peak shift and (ii) the average of the local sensitivity across the whole period (0 – 100 nm) is equal to the bulk RI sensitivity (140 nm/RIU) – e.g. the vertical position of each horizontal bar in Fig. 4(a) corresponds to the integral of the sensitivity curve in the corresponding horizontal interval. In each cycle, the sensitivity curve was adjusted uniformly inside and outside of the interval corresponding to one mask position. In order to achieve smoothness and continuity of the local sensitivity, the fitted curve was smoothed with a 3 nm boxcar aperture in each iteration cycle. The standard deviation of the fit was 3.5% calculated from differences between measured peak shifts and peak shift calculated from the fitted local sensitivity.

 

Fig. 4 (a) Local sensitivity as deduced experimentally using the mask method. Measured peak shifts are plotted as grey horizontal bars – their horizontal position indicates the mask overlap with the nanoparticle and their vertical position indicates the peak shift normalized to mask width and RI; the average (bulk) sensitivity to bulk RI is indicated. The deconvoluted distribution of the sensitivity is plotted as the solid line for one period of the nanoparticle array. (b) Local sensitivity modeled using the FDTD method (solid line). The profile of the electric field intensity is shown for comparison (dashed line). The minor feature at the positions 30 nm and 70 nm is a numerical artifact associated with the discretization of the rounded nanoparticle tips. The nanoparticle is visualized in the graphs for reference.

Download Full Size | PDF

A similar deconvolution was carried out using the results of theoretical simulations. In order to achieve a higher spatial resolution of the simulated local sensitivity, narrower stripe masks were utilized in the simulations. A set of two masks with a width of 3 nm was modeled to symmetrically cover the nanoparticles and the distance from the nanoparticle center was sampled. The extinction peak shift due to the mask was normalized to the width of the mask and to the RI change in the same fashion as experimental results.

The resulting local sensitivity deduced from the measurements is plotted in Fig. 4(a) (solid line) showing a strong enhancement of local sensitivity around the tips of nanoparticles. The used experimental method allowed mapping of the local sensitivity with resolution of around 5 nm in the area of the most rapid sensitivity variation while in the center of the nanoparticle, the resolution is limited to several tens of nanometers. This local sensitivity enhancement was confirmed by the simulation carried out using a spatial resolution of 3 nm. The modeled local sensitivity is plotted in Fig. 4(b) (solid line). Integrals of simulated and measured local sensitivities differ by a factor of 1.4 which corresponds to the difference in the theoretical and experimental sensitivity (201 vs. 140 nm/RIU, see above). However, the position of maximum sensitivity and the confinement of high sensitivity agree well with the theoretical model, within the limits determined by the spatial resolution of the experimental procedure.

The variation of the local refractive index sensitivity along the nanoparticle was compared to the profile of electric field intensity of the LSP. The dashed line in Fig. 4(b) depicts the electric field intensity averaged in the plane perpendicular to the nanoparticle long axis, excluding all points within the particle and the substrate. Both experimental and theoretical results imply that the sensitivity is strongly correlated with the profile of the LSP electric field. This confirms the general assumption that the confinement of the electrical field in the proximity of geometrically sharp features of nanoparticle or narrow gaps between nanoparticles results in a similar confinement of the local sensitivity. Figure 4(b) suggests that the local sensitivity follows very closely the profile of the intensity of electric field.

The response to a refractive index change (e.g. due to the binding of biomolecules) is strongly dependent on the area of the nanoparticle, where the refractive index change takes place. Therefore, the sensor response to the binding of biomolecules will depend not only on the amplitude of the binding-induced refractive index change, but also on its spatial overlap with the strongly varying electric field of LSP. This may make quantification of the binding difficult, in particular, for very low numbers of binding molecules. This issue could be, for instance, reduced by employing functionalization approaches confining the interacting molecules to the areas near the tip of the nanoparticle.

5. Conclusion

The local distribution of the sensitivity to refractive index changes in the proximity of a plasmonic nanostructure was investigated experimentally. This is the first time an experimental method was employed to directly observe and map the localized sensitivity of plasmonic nanostructures. Results obtained experimentally were compared with simulations performed using the FDTD method. We demonstrate that the spatial distribution of sensitivity matches closely the profile of electric field intensity. For the model nanostructure comprising an array of nanorod chains the enhancement of sensitivity to RI changes localized at nanorod tips was confirmed both in the theoretical model and in experiment. This study suggests that simple analysis of field distribution in plasmonic nanostructures may provide a very powerful tool for optimization of the refractive index sensitivity of the plasmonic nanostructure-based sensors.