Thin Layer Sensing with Multipolar Plasmonic Resonances - The

Aug 22, 2008 - First, the small modal volume makes them selectively sensitive to the first 10−30 nm distance from the gold. Second, each mode yields...
0 downloads 0 Views 199KB Size
J. Phys. Chem. C 2008, 112, 14355–14359

14355

Thin Layer Sensing with Multipolar Plasmonic Resonances ´ lvarez, and Maximilian Kreiter* Noelia L. Bocchio, Andreas Unger, Marta A Max-Planck-Institute for Polymer Research, Ackermannweg 10, D-55128 Mainz, Germany ReceiVed: May 9, 2008; ReVised Manuscript ReceiVed: June 12, 2008

Gold crescent-shaped nanoparticles, exhibiting multiple resonances in the visible and near-infrared spectral region, were characterized in terms of their optical response to thin dielectric coatings. A strong red shift of the resonances is observed upon the addition of material. The coating thickness at which saturation of the shift sets in is resonance dependent and is directly connected to the spatial extension of the near field associated to each of the plasmonic modes, which is estimated through simplified theoretical models. These structures are advantageous for sensing for two reasons. First, the small modal volume makes them selectively sensitive to the first 10-30 nm distance from the gold. Second, each mode yields independent information about the coating and therefore in principle allows the determination of more than one free parameter of the coating, for example, thickness and refractive index. Introduction Several sensing techniques make use of the detuning of optical resonances in systems containing noble metal films or particles, via the attachment of a layer of a material with a refractive index that differs from the one of the subphase. The most commonly used approach is based on propagating surface plasmons on plane metal surfaces, where a fraction of a monolayer is easily detectable.1–3 Similarly, the shift in the resonance wavelength of metallic particles in response to thin dielectric coatings can be used for sensing.4–6 The conceptual advantage of this approach lies in the highly localized sensing volume that allows for very small amounts of analyte to be detected.7 While the sensitivity of spherical particles to deposition of a thin dielectric layer is limited, several more complex structures such as hollow spheres,8 triangular pyramids,4,9–11 and rings12 have been shown to be more sensitive. Sensing based on particles until now has been limited mostly to geometries that support only one resonance, while structures that support several resonances are known.13–16 A system with multiple resonances allows for the simultaneous determination of more than one material-parameter in only one experiment, i.e., thickness and refractive index in conventional thin layer sensing approaches.3 These applications might be extended to study in a simple experiment, and in a highly localized manner, more complex systems like the optical properties of thin hydrogel films at different penetration depths,17 the collapse of biological brushes,18 or biorecognition events.12,19,20 In the following, we demonstrate the feasibility of thin layer sensing with crescentshaped metal structures, which support several modes. The sensitivity of the modes is determined and an estimate for the near field extension is given. Experimental Section Sample Preparation. Crescent-shaped gold nanoparticles were prepared according to a nanosphere lithographic method.15 Polystyrene nanospheres (Polystyrene Nanobead: NIST, Polysciences, Warrington, USA), 200 nm in diameter, were randomly dispersed on a piranha (H2SO4:H2O2, 2.5:1; Caution: piranha * To whom correspondence should be addressed. E-mail: kreiter@ mpip-mainz.mpg.de.

solutions are extremely dangerous and should be handled with extreme care) cleaned glass substrate (Objekttra¨ger Glas, Menzel Glaeser, Germany, approximately 20 mm × 25 mm, 1 mm). A 40 nm thick gold film was evaporated (Auto 306, Edwards, Sussex UK) on the colloid-covered substrates; the samples’ surface normal was tilted by 30° relative to the metal source. After the metal deposition, the gold-coated samples were exposed to an argon ion beam (RR-I SQ76, Roth & Rau, Wu¨stenbrand, Germany) perpendicularly incident to the substrate and etched. After the etching process, the colloidal mask was removed mechanically by means of an adhesive tape (Scotch Magic Tape 810: 19 mm × 33m, 3M France) and a second etching step was applied to ensure the removal of any extra sputtered material from the surface. A scanning electron micrograph (LEO 1530 Gemini) of a reference sample prepared on silicon is shown in Figure 1a. Thin films composed of polyelectrolyte multilayers21 were chosen as a model coating to study the response of the optical resonances to changes in the dielectric environment. They represent a very robust and reproducible coating strategy that has been used extensively in earlier studies. Before building the films on the gold crescents, their surface was functionalized by immersion in a 0.03 M aqueous solution of 3-mercaptopropionic acid (3-MPA, Sigma Aldrich) for 15 min. This thiol is expected to attach preferentially to gold and provide a site for electrostatic attachment of the polymeric layers. After rinsing with ultra-pure water (Milli-Q), alternating layers of poly(allylamine) hydrochloride (Mw: 15000, Aldrich) and poly(styrene sulfonate) sodium salt (Mw: 70000 Aldrich) were deposited on the samples by immersion for 20 min in the corresponding polyelectrolyte solutions, starting with polyallylamine.21–23 The concentration of the polyelectrolytes was 0.02 M (in monomer units). 0.5 M MnCl2 were added to the polystyrene sulfonate solution and 2 M for NaBr (in the polyallylamine solution). The pH of the polyelectrolyte solutions was set to 3 via the addition of hydrochloric acid. The samples were rinsed thoroughly with ultra pure water after each immersion step, dried under a stream of nitrogen, and stored. Characterization. Film Thickness. The thickness of the added layers was determined for the first 3 bilayers by using atomic force microscopy (AFM, Nanoscope IIIa, Veeco). The

10.1021/jp804099u CCC: $40.75  2008 American Chemical Society Published on Web 08/22/2008

14356 J. Phys. Chem. C, Vol. 112, No. 37, 2008

Figure 1. (a) Scanning electron micrograph of a crescent-shaped particle; the scale bar represents 200 nm. (b) Sketch of a cross section of a crescent covered with polyelectrolytes along the dashed line in part a; d is the thickness of the film, h and w are the height and width of the structure, respectively, and PE is the polyelectrolyte film. (c) Film thickness as a function of the number of bilayers. The vertical crosses (+) represent the AFM measurements, the tilted ones ( ×) are literature values.24 The circle represents an estimated intermediate value between these two regimes.

measurements were done by comparing the height of the structures before and after the attachment of the bilayers, taking the clean glass surface as zero height. This method works properly in the 1 to 3 bilayers regime; in that case, a contribution of approximately 4 nm per bilayer was determined, which is in good agreement with earlier investigations of the same system by optical methods and by small-angle X-ray reflectometry.23,24 A larger number of bilayers leads to full surface coverage of the glass substrate, which hinders the thickness determination by AFM. In this regime, we rely purely on the results from the optical investigations, which indicate a thickness of 8.4 nm per bilayer from the fifth bilayer on. We note that this transition from a regime with slower growth for the first few bilayers, followed by a regime with a larger, constant increase per bilayer is known for this type of system and has been discussed, e.g., by Buron et al.25 A sketch showing some typical dimensions on a cross section of a polymer-coated crescent is shown in Figure 1b. Here, w ) 65 nm and h ) 40 nm. The layer thickness as a function of deposited bilayers that was used in this work is presented in Figure 1c. UV-Vis/NIR spectroscopy. The optical properties were studied with UV-vis/NIR transmission spectroscopy, a common method to study particle resonances,4,9–11,15,20,26,27 where these are identified as maxima in the corresponding extinction spectra. Nonpolarized light spectra were recorded in an UV-vis/NIR spectrometer (Perkin-Elmer, Lambda 900) operating in transmission mode, before and after the addition of polyelectrolyte bilayers, sampling an area of the substrates of approximately 50 mm2. Peak position and width of the resonances were estimated by using a multipeak fitting routine (Igor Pro version 5.02), adjusting Lorentzian curves in a 500-2500 nm range14 and after polyelectrolyte deposition, yielding a displacement as their difference. The occurrence of an increasing absorption background upon the addition of polymers was taken into account in the fitting procedures by adding an extra Lorentzianlike decay in the 400-800 nm range.

Bocchio et al.

Figure 2. (a) Extinction spectra of four different samples before (black) and after (gray) the addition of 1, 4, 8, and 11 polyelectrolyte bilayers. The spectra have been displaced vertically for clarity; indication of the denomination of each resonance is included in the graph. (b) Simplified model of the charge distribution of the resonant modes.

Results and Discussion Figure 2a presents nonpolarized extinction spectra for 4 representative samples, before any chemical treatment (black) and after coating (gray). Four peaks are seen which are identified as the standing waves along the crescents’ contour of first (c1), second (u1) and third (c2) order. The fourth peak is assigned to a polarization of the crescents perpendicular to the contour (pp). Variations in peak strength are seen which are due to differences in crescent density on the surface. Only minor variations in peak position are observed, pointing toward a good sample-to-sample reproducibility. Slight variations of the resonances’ widths may indicate some inhomogeneous broadening, which does not affect the further discussion. Figure 2b presents a simplified charge distribution of these modes. They appear before adding the coating at wavelengths of roughly 1300, 850, 750, and 550 nm, respectively. A thorough discussion of the assignment of these peaks as well as the terminology to label them is given in an earlier publication.14 The spectra of all samples are available in the Supporting Information. Figure 2a shows that, upon the addition of the dielectric coating, all resonances progressively red shift. The observed displacements are different for each resonance. In addition to this, an increasing background in the 400-800 nm range sets in upon the addition of polymer, which cannot be assigned to a resonance of the crescents. Figure 3 displays the peak displacement as determined from the fitting for the c1, u1, and c2 resonances as a function of the coating thickness in the many-layers (a) and few-layers (b) regimes. Each set of data points associated to a defined number of bilayers corresponds to one independent sample, except for 1, 2, and 3 bilayers. For 1 and 3 bilayers, 2 independent samples were measured; for 2 bilayers the data are based on 3 samples. In these cases, the mean values are displayed (see the Supporting Information for the corresponding spectra). The data corre-

Thin Layer Sensing with Plasmonic Resonances

J. Phys. Chem. C, Vol. 112, No. 37, 2008 14357

Figure 4. Sketch of the simplified models: (a) crescent, (b) planar interface, and (c) rod.

Figure 5. Normalized peak shift, as calculated from the rod model, as a function of the distance to the rod’s surface (d).

Figure 3. Peak displacement of the main resonances in the full (a) and first layers (b) regimes. Resonance assignments are given in the graph. The dashed lines represent an exponential fit to the data. The solid markers in panel a correspond to the displacement obtained when immersing the samples in a n ) 1.524 oil.28 For a thickness of roughly 140 nm (18 bilayers), only the displacement corresponding to c1 can be determined unequivocally, and is included in the overall fitting. The gray marker gives an indication for the displacement of u1 for thick coatings, but this value is not included in further calculations.

TABLE 1: Fitted Parameters, According to Eq 1), for the Three Main Resonances c1 u1 c2

∆λm,i [nm]

∆λ0,i [nm]

dl,i [nm]

d(∆λi)/d(d)|d ) 0

236.9 111.4 51.6

-9.95 -17 -17.7

26.2 17.2 8.5

9.4 7.4 8.1

sponding to d ) 0 nm correspond to functionalization with 3-MPA only. The pp mode will be disregarded, since its displacement saturates rapidlysin the 5 to 10 nm range. Added to this, the increasing number of layers gives rise to a growing background, which in turns also makes the precise fitting of pp difficult. The shifts increase rapidly in the first layers (1 to 5) regime and saturate for further increasing thickness. The shift at saturation can be compared to the shifts observed when the crescents are surrounded by an infinitely thick medium with homogeneous refractive index, in analogy to an infinitely thick

coating. Since deposition of infinitely many multilayers is experimentally impossible and deposition of thick films leads to additional interference effects29 that complicate the analysis, this comparison is to be made with the crescents response when immersed in a liquid with a refractive index that equals that of the polyelectrolytes. The latter has been determined23,24 to be n ) 1.54 at λ ) 633 nm. Considering normal dispersion, a refractive index of n ) 1.5 can be used as an approximate value at the longer wavelengths where the crescents’ resonances occur. As a simple reference, a measurement with an n ) 1.524 index oil28 is taken (see Figure 3a). This oil gives rise to resonance displacements that are in reasonable agreement with the values obtained for an approximately 150 nm thick layer, indicating that, for this bilayer thickness, the saturation of the shifts is reached for all resonances. The peak displacement for each resonant mode (∆λi) can be quantified by fitting an exponential function of the form:

∆λi(d) ) ∆λm,i + (∆λ0,i - ∆λm,i)e-d ⁄ dl,i where ∆λm,i represents the resonance-specific (i) saturation limit, d the thickness of the layers, ∆λ0,i the peak displacement at d ) 0, and dl a typical decay length. The index “i” denotes the 3 main resonances (c1, u1, and c2). The results of this procedure are summarized in Table 1. An apparent negative displacement for the “zero layer” (∆λ0,i) resonances is observed. This effect could be related to a physical phenomenon-like structure annealing or thiol binding,30 or simply be the result of the scatter in the measurements, which is of the order of the observed displacements. It must be stressed here that the choice of an exponential function is not based on a physical model but represents an arbitrary and rather simple parametrization of a typical decay length. Since the exponentials describe the data quite accurately given the experimental scatter, the decay lengths as obtained from the fits can be used as a basis for the comparison with theoretical models. Both the displacements at saturation (∆λm,i) and the decay lengths (dl,i) are resonance-specific, as can be observed in Table 1. Higher order resonances (c2, u1) show much more confined

14358 J. Phys. Chem. C, Vol. 112, No. 37, 2008

Bocchio et al.

TABLE 2: Decay Lengths for the Different Modes (Resonances c1, u1, and c2) According to the Measurements (dl,exp) and the Different Models (dl,PI, dl,rod) wavelength [nm] kx [nm-1] kz [nm-1] dl,exp [nm] dI,PI [nm] dl,rod [nm]

c1

u1

c2

1364 0.00855 0.0072 26.2 70 44.5

897 0.0171 0.0156 17.2 32 25.5

742 0.02565 0.0242 8.5 20.6 17.6

near fields than the fundamental mode (c1), which leads to smaller decay lengths dl. The first 20 nm regime is most useful for sensing, since there the relative displacements of the resonances upon the addition of layers of polymer are most pronounced. Since the widths of the resonances are similar, we consider the derivative d(∆λi)/d(d) at d ) 0 from the fitting (see Table 1) as an estimate to the sensitivity of the modes upon the addition of the layers. We observe that c1 is the most sensitive, followed by c2 and u1. It should be noted, though, that all values are of the same order of magnitude, meaning that all resonances are comparable. A clear advantage of c1 is its large extinction cross-section and spectral separation from the other modes, which simplifies the simultaneous multipeak fitting. We note that frequently the usefulness for sensing is expressed in terms of a figure of merit (FOM), which is given as the peak displacement per change of the surrounding bulk refractive index, divided by the resonance width. Since we are focusing on the response to thin layers, a direct comparison to other particle shapes in terms of this quantity is not possible; earlier immersion experiments28 suggest a FOM that is similar to alternative structures. It is important to point out that, due to their different penetration depths, the modes give independent information about the surrounding environment. This is particularly important when considering that the simultaneous use of n resonances allows for the determination of n parameters that describe the film; for example, with two resonances one may determine thickness and refractive index simultaneously; a third resonance would allow in addition the determination of another parameter like a refractive index gradient in the film. This approach is much like the strategies that are used commonly in conventional sensing with dielectric waveguide modes31 but here, the modal volume is far below that determined by the order of magnitude of the wavelength of light, and the modes are spatially confined in three dimensions. Comparison to Simple Analytical Models. When the resonances on the crescents are interpreted as being composed of standing waves which are formed by the superposition of a forward and a backward propagating wave along the crescents’ contour, an intimate connection between the shift of the modes and their near field is directly evident. We consider a guided electromagnetic mode with known field distribution. A change of the refractive index (∆n) within the volume where the electric field (E) is nonvanishing will lead to a change ∆k in the propagation constant k of this mode. This change is to a first approximation obtained by mode coupling theory32

∆k ∝

∫ dV |E|2∆n2

where the integral is to be taken over the entire space. If, as is the case here, a material with a fixed refractive index is added so that it occupies some volume in the nearby area, and we furthermore realize that for a resonator with a given length (l)

a change in k will directly lead to a change in resonance wavelength, then

∆λ ∝ ∆n2

∫V

oc

dV |E|2

where the integral is to be taken over the volume (Voc) that is occupied with the high refractive index material. It should be stressed that this theory is only strictly valid for real dielectrics or perfect metals, which is not the case for gold in the visible range. Furthermore it is assumed that the perturbing change in the refractive index only leads to infinitesimal changes of the field, which is only true for small dielectric contrasts ∆n. Despite these limitations, we use this theory for an approximate description of the observed shifts. While a full numerical description of the system under study is tedious, good physical insight into the effect of different penetration depths can be gained by considering two simple model systems where analytical solutions for the field distributions of the eigenmodes are known; these models are the plane interface and the cylindrical rod in vacuum (see Figure 4 for details). One may compare the guided modes on the crescent with the ones on a plane interface (see Figure 4b for a simplified sketch), as a first rough approximation. In this case, the z-dependence of the electrical field is given as

E ∝ e-Im(kz)z where kz is purely imaginary and is obtained for guided modes from

kz2 + kx2 ) k02 with kx and k0 being the wave vector in the propagation direction and the modulus of the wave vector of plane light waves, respectively. It is possible to estimate kx as follows:

kx )

2π mπ ) λmode lcontour

m being the order of the resonance and lcontour being the contour length of the crescent, which is in turn identified with the size of the resonator. To estimate lcontour we chose, from geometrical considerations14

lcontour ) 1/2dcollR where R represents the angle expressed in radians, corresponding to an arc of longitude lcontour and dcoll the diameter of the colloids used to produce the crescents. Neglecting all imperfections and variations of the preparation process, for an ideal sample R is 1.17π. In this model, the peak position displacement as a function of the thickness (d), ∆λi(d), is obtained up to a prefactor, allowing us to calculate its value normalized to the displacement for an infinitely thick coating, ∆λm,

∆λ(d) ) 1 - e-2Im{kz}d ∆λm which directly yields the decay depth dl,PI,

dl,PI )

1 2Im{kz}

where the subscript PI denotes the plane interface model. Another model where field distributions are readily calculated analytically are cylindrical rods (see Figure 5b). For a propagating wave on an infinitely extending cylindrical rod, the radial electrical Er field outside the rod is given by

Thin Layer Sensing with Plasmonic Resonances

Er ∝ H1(1)(kzr) where H1(1) is a Hankel function of the first kind, and r is the distance measured from the center of the rod. The Er component of the electrical field is the normal component and dominant in the near field. If the crescents are modeled as finite rods, supporting a standing wave which is a superposition of two counter propagating modes of the free rod which are reflected at the ends, this leads to R+d

∆λ(d) ∝

∫ r|[H1(1)(kzr)]2| dr R

with R being the radius of the rod and d the thickness of the layer. We chose to model the crescent as a rod with R ) 20 nm corresponding to half the height of the crescents. The resulting normalized ∆λi(d) curves are displayed in Figure 5. A typical decay length (dl,rod) was obtained in this approximation by solving the equation ∆λ(dl,rod) ) (1 - e-1)∆λmax, ∆λmax being the shift for an infinitely thick coating. Table 2 summarizes the results for the plane interface model, the rod model and the experiments. These very simple geometrical models reproduce qualitatively the experimentally observed trend of a shrinking modal volume with increasing mode number. In particular, the direct connection to a simple analytical field distribution allows for an intuitive understanding of the observed effects. The expected field extensions dl for c2 and u1 are comparable for the two models, but higher than the experimental results; the rod model appears as particularly suitable to describe the response of the crescent structures. Several mechanisms may account for the quantitative difference between the models and the experiments. First, the geometrical simplifications made when describing the crescents as rods/plane layers are significant, in particular, at the corners of the crescents, where a high localization of the field occurs, which is not reflected in the models and can lead to a reduced dl. Second, the mode coupling theory itself is only an approximation. Finally, imperfections of the structures and the coating may also play a role, particularly for the first layers. Conclusion We presented a measurement of the optical response of crescent-shaped nanoparticles to the successive attachment of thin polymer layers, resulting in a clear mode-dependent red shift of the multiple crescents’ resonances upon layer growth. Typical decay lengths were extracted from exponential fits and used as estimates for the near field confinement. The sensitivity of all modes is similar. The system was described by two simple physical models, a plane interface and a rod, obtaining field confinements comparable to the experiments. Since it is possible to monitor several resonances in one single experiment, simultaneous determination of multiple material parameters should be possible in ultimately small modal volumes. While this requires a full quantitative description of the near field distribution, which is beyond the simple models discussed here, it will represent a major advantage for sensing applications.

J. Phys. Chem. C, Vol. 112, No. 37, 2008 14359 Acknowledgment. We thank the Bundesministerium fu¨r Bildung and Forschung (BMBF) (grant 03N8702) and Junta de Comunidades de Castilla la Mancha for financial support and J. Fischer for experimental support. Supporting Information Available: UV-vis/NIR transmission spectra before and after the attachment of polyelectrolytes, for all the samples analyzed and peak position and widths of crescent-shaped nanoparticles immersed in different index oils. This material is available free of charge via the Internet at http:// pubs.acs.org. References and Notes (1) Gordon, J. G.; Swalen, J. D. Opt. Commun. 1977, 22, 374. (2) Homola, J. Surface Plasmon Resonance Based Sensors; Springer: Berlin, Germany, 2006. (3) Knoll, W. Annu. ReV. Phys. Chem. 1998, 49, 569. (4) Haes, A. J.; Zou, S. L.; Schatz, G. C.; Van Duyne, R. P. J. Phys. Chem. B 2004, 108, 109. (5) Englebienne, P. Analyst 1998, 123, 1599. (6) Frederix, F.; Friedt, J. M.; Choi, K.-H.; Laureyn, W.; Campitelli, A.; Mondelaers, D.; Maes, G.; Borghs, G. Anal. Chem. 2003, 75, 6894. (7) McFarland, A. D.; Van Duyne, R. P. Nano Lett. 2003, 3, 1057. (8) Raschke, G.; Brogl, S.; Susha, A. S.; Rogach, A. L.; Klar, T. A.; Feldmann, J; Fieres, B.; Petkov, N.; Bein, T.; Nichtland, A.; Kurzinger, K. Nano Lett. 2004, 4, 4. (9) Haes, A. J.; Stuart, D. A.; Nie, S. M.; Van Duyne, R. P. J. Fluoresc. 2004, 14, 355. (10) Haes, A. J.; Van Duyne, R. P. J. Am. Chem. Soc. 2002, 124, 10596. (11) Haes, A. J.; Zou, S. L.; Schatz, G. C.; Van Duyne, R. P. J. Phys. Chem. B 2004, 108, 6961. (12) Larsson, E. M.; Alegret, J.; Kaell, M.; Sutherland, D. S. Nano Lett. 2007, 7, 1256. (13) Krenn, J. R.; Schider, G.; Rechberger, W.; Lamprecht, B.; Leitner, A.; Aussenegg, F. R.; Weeber, J. C. Appl. Phys. Lett. 2000, 77. (14) Rochholz, H.; Bocchio, N.; Kreiter, M. New J. Phys. 2007, 9. (15) Shumaker-Parry, J.; Rochholz, H.; Kreiter, M. AdV. Mater. 2005, 17, 2131. (16) Sherry, L.; Chang, S.-H.; Schatz, G.; Van Duyne, R. P. Nano Lett. 2005, 5, 2034. (17) Beines, P. W.; Klosterkamp, I.; Menges, B.; Jonas, U.; Knoll, W. Langmuir 2007, 23, 2231. (18) Lim, R. Y. H.; Fahrenkrog, B.; Ko¨ser, J.; Schwarz-Herion, K.; Deng, J.; Aebi, U. Science 2007, 318. (19) Kim, S.; Jung, J.-M.; Choi, D.-G.; Jung, H.-T.; Yang, S.-M. Langmuir 2006, 22, 7109. (20) Yonzon, C. R.; Jeoung, E.; Zou, S. L.; Schatz, G. C.; Mrksich, M.; Van Duyne, R. P. J. Am. Chem. Soc. 2004, 126, 12669. (21) Decher, G.; Schmitt, J. Colloid Polym. Sci. 1992, 89, 160. (22) Decher, G. Science 1997, 277, 1232. (23) Vasilev, K. Ph.D. Thesis, Martin-Luther-Universitaet, 2004. (24) Vasilev, K.; Knoll, W.; Kreiter, M. J. Chem. Phys. 2004, 120. (25) Buron, C. C.; Filiaˆtre, C.; Membrey, F.; Bainier, C.; Charraut, D.; Foissy, A. J. Colloid Interface Sci. 2007, 314, 358. (26) Larsson, E. M.; A., J.; Kaell, M.; Sutherland, D. S. Nano Lett. 2007, 7, 1256. (27) Whitney, A. V.; Elam, J. W.; Zou, S. L.; Zinovev, A. V.; Stair, P. C.; Schatz, G. C.; Van Duyne, R. P. J. Phys. Chem. B 2005, 109, 20522. (28) Rochholz, H. Ph.D. Thesis, 2005. (29) Murray, W. A.; Suckling, J. R.; Barnes, W. Nano Lett. 2006, 6, 1772. (30) Duval Malinsky, M.; Lance Kelly, K.; Schatz, G.; Van Duyne, R. J. Am. Chem. Soc. 2001, 123, 1471. (31) Van Os, M. T.; Menges, B.; Foerch, R.; Vancso, G. J.; Knoll, W. Chem. Mater. 1999, 11, 5. (32) Buckman, B. A. Guided waVe Photonics; Saunders College/Harcourt Brace Publishing: Fort Worth, TX, 1995.

JP804099U