Within the burgeoning field of nanotechnology, one of the

Transcription

1 pubs.acs.org/nanolett Direct Observation of Beamed Raman Scattering Wenqi Zhu, Dongxing Wang, and Kenneth B. Crozier* School of Engineering and Applied Sciences, Harvard University, Cambridge, Massachusetts 02138, United States *S Supporting Information ABSTRACT: Appropriately designed surface plasmon nanostructures enable the emission patterns of surface-enhanced Raman scattering to be modified to facilitate efficient collection, an effect sometimes termed beamed Raman scattering. Here, we demonstrate the direct and unambiguous observation of this phenomenon by separating the Raman emission pattern from the luminescent background using energy momentum spectroscopy. We observe beamed Raman scattering from two types of optical antennas: the first are Yagi Uda optical antennas, and the second are optical dimer antennas formed above a plasmonic substrate consisting of a gold film integrated with a one-dimensional array of gold stripes. For both antenna types, the emission patterns from different Raman lines are simultaneously measured. For the second antenna type, the emission patterns show signatures stemming from the bandstructure of the plasmonic substrate. KEYWORDS: Plasmonics, optical antenna, surface-enhanced Raman scattering, spectroscopy Within the burgeoning field of nanotechnology, one of the topics being pursued with great interest is the interaction between nanoscale photon emitters and optical antennas. These include fluorescent molecules, 1 quantum dots, 2 nitrogen vacancy centers, 3 and Raman molecules. 4 Optical antennas employ the localized surface plasmon resonances (LSPRs) of metal nanoparticles to enhance light matter interactions. 5 Engineered optical antennas boost the efficiency by which light can be focused into subwavelength regions. Similarly, in the reciprocal process, optical antennas boost the transmission of energy to the far-field. Recently, interest has emerged on shaping the emission patterns of nanoscale emitters with directional optical antenna designs. 2,6 9 A common goal of these efforts is to collimate emission into one direction, thereby enabling efficient detection even with low numerical aperture (NA) objective lenses. It is well established that emission patterns from nanoscale emitters can be directly imaged in the back focal planes of high-na objective lenses. 10 This technique has very high sensitivity, even at the single molecule level. Using the back focal plane technique, directional antennas, such as Yagi Uda (YU) designs and periodic plasmonic structures, have been shown to efficiently collimate the emission from quantum dots and fluorescent molecules. 2,8 Surface-enhanced Raman scattering (SERS) employs optical antennas to enhance Raman cross section of molecules by orders of magnitude. Since the first demonstrations of single molecule sensitivity, 11,12 much emphasis in SERS substrate design has been on finding nanostructures with larger electric field enhancement. This is because the electromagnetic component to the SERS enhancement factor (EF) is proportional to the fourth power of electric-field enhancement (assuming a Raman shift of zero). 13 Optical antenna designs with interesting possibilities for SERS include nanoparticle dimers with nanoscale gaps, self-similar nanoparticle chains, 17 Fano-type heptamers, 18,19 and plasmonic oligomers. 20 Significantly enhanced electric fields can be achieved, owing to both nanoscale gaps and coupling between the LSPRs of the particles. Recently, the importance of directionality in SERS has been recognized. That Raman scattering can be collimated by directional optical antennas has been termed beamed Raman scattering 21,22 and directional SERS (DERS). 23 These concepts have been employed in SERS substrates achieving very high EFs (average value ) 24 and for single molecule sensitivity. 25 One of the challenges facing the direct observation of beamed Raman scattering is that Raman cross sections are weak, typically orders of magnitude smaller than fluorescence cross sections. Shegai et al. addressed the problem of measuring the angular distribution of SERS from Rhodamine 6G molecules adsorbed on individual gold nanoparticle aggregates. 7 A microscope objective lens was used to focus laser light onto each aggregate and to collect the emission, with Rayleigh scattered laser light removed with a long pass filter (LPF). The angular distribution of emitted light could be directly monitored in the Fourier plane of the optical microscope. There was no provision, however, to distinguish between the emission patterns of the Raman lines and that of the broad background continuum. The latter comes from fluorescent emission of the Raman molecules or the substrate or even photoluminescence of the metallic structure itself. 26,27 Received: September 4, 2012 Revised: October 21, 2012 Published: October 26, American Chemical Society 6235

2 Figure 1. Raman emission pattern from dimer antennas on dielectric substrate. (a) SEM image of top view of one dimer antenna. (b) Cartesian xyzcoordinate that defines the geometry of the dimer structure and the air dielectric interface (xy-plane), and spherical ΘΨ-coordinate that defines the emission direction. For polar plots in (c) we specify the plane of emission and allow Θ to range from 0 to 360. For color plots from (d f) we use the standard spherical coordinate convention which allows Θ to range from 0 to 180 and Ψ to range from 0 to 360. (c) FEM simulation of the emission pattern in which an arbitrarily oriented Raman dipole is placed at the gap center of the dimer antenna. The region below the line from Θ = 90 to Θ = 270 represents the dielectric substrate; red: emission pattern in yz-plane (Ψ =90 ) and blue: emission pattern in xz-plane (Ψ =0 ). (d) Emission pattern of thiophenol 1074 cm 1 Raman line retrieved from EMS measurements. Color map represents the scattering intensity normalized by the maximum intensity. Inner and outer black dashed circles indicate the critical angle at the air glass interface and the NA of objective lens, respectively. Raman emission pattern for (e) 415 and (f) 1586 cm 1 Raman lines, retrieved from same measurement. An improved version of back focal plane technique, namely energy momentum spectroscopy (hereon referred as the EMS technique ), has recently been developed to measure both the spectral and angular characteristics of the emission patterns of nanoscale emitters. 28 A spectrometer measures the spectrum of each pixel in the back focal plane of the objective lens, enabling the emission pattern for each wavelength of interest to be determined (see Supporting Information for details). Here, we use the EMS technique to measure the radiation patterns of multiple Raman lines of three SERS substrates, directly and unambiguously observing the beamed Raman scattering effect. Raman Emission Pattern from Dimer Antennas. We demonstrate the effectiveness of the EMS technique by first studying the Raman emission pattern from optical dimer antennas. A scanning electron microscope (SEM) image of one such antenna structure is shown in Figure 1a. It consists of two gold nanorods that are both 117 nm long and 65 nm wide and are separated by a gap of 13 nm. The antennas are fabricated by electron-beam lithography on a glass coverslip coated with indium tin oxide (ITO). Large near-field enhancement within 6236

3 the small gap region ensures a large SERS enhancement factor. To increase the signal-to-background ratio, measurements are made from arrays of optical dimer antennas rather than single dimer antenna. The unit cell of the array (0.5 μm 1.5 μm) is chosen to be large in order to minimize near-field coupling between adjacent dimer antennas. To further reduce background fluorescence, a nonresonant Raman molecule (thiophenol) is used, formed on the antennas as a self-assembled monolayer (SAM). In the following simulations/experiments, the laser excitation wavelength is maintained at λ = 785 nm, meaning that the 1074 cm 1 Raman line occurs at λ = 857 nm. We also perform experiments and simulations for the Raman lines 415 cm 1 (λ = 811 nm) and 1586 cm 1 (λ = 896 nm), as a way of determining the bandwidths of our directional optical antennas. The antenna orientation and emission angles (Θ, Ψ) are schematically illustrated as Figure 1b. We begin by simulating the Raman radiation pattern at λ = 857 nm from an optical dimer antenna on glass using the finite-element method (FEM). Although in experiments, the thiophenol molecules coat the antenna surface uniformly, the largest Raman dipole is induced within the gap region, due to the greatly enhanced near-fields there. For the simulations, therefore, we place an electric Raman dipole in the center of the gap region with a random orientation. In the emission process, the Raman dipole excites the dipolar LSPR mode of the dimer antenna. 15 If the Raman dipole frequency coincides with LSPR frequency, the emission pattern of Raman scattering will be dominated by the dipolar mode of the dimer antenna. 6 Therefore, although the orientation of the Raman dipole is arbitrary, Figure 1c can be seen to follow the typical emission pattern for a horizontal dipole radiating above dielectric substrate. It can be seen that the majority of the radiated power goes into the substrate, is mainly within the yz-plane that is perpendicular to the dipole orientation, and occurs symmetrically into two lobes which are just beyond the critical angle for total internal reflection at the air dielectric interface. The measured SERS emission pattern at λ = 857 nm (1074 cm 1 Raman line) found using the EMS technique is shown as Figure 1d. An oil-immersion objective lens (Nikon 100, NA= 1.4) focuses polarized laser illumination through the glass coverslip substrate onto the antennas and collects the Raman emission. As described in the Supporting Information, the objective is incorporated into a home-built setup that permits the angular distribution of the Raman emission to be measured. As depicted in Figure 1b, the dimer antennas are oriented such that their longer axes are parallel to the laser polarization (Ψ = 0, x-axis). Figure 1d can be understood as a polar coordinate system, with the origin at the center of the circular pattern. 10 The radial coordinate r corresponds to the polar angle Θ of Raman emission according to r = n sin Θ, where n = 1.5 for the glass coverslip. The outline of the circle represents the NA of the objective lens (NA = 1.4). The angular coordinate corresponds to the azimuth angle Ψ of Raman emission. With the defined coordinates, we can estimate that the two main emission lobes of the Raman scattering from the dimer antennas peak in the following directions: (Θ = 139.5, Ψ = 100 ) and (Θ = 139.5, Ψ = 260 ). It can be seen that the azimuthal angles are close to the expected values of Ψ =90 and 270. Both measured polar angles Θ are close to the critical angle at the air glass interface (138.2 ) within the measurement resolution. The full width at half-maximum (fwhm) contours of the two beams have widths of (ΔΘ =22, ΔΨ = 48 ) and (ΔΘ =15, ΔΨ =48 ). We define the front-to-back (F/B) ratio as the ratio of integrated emission power within these two fwhm contours. The ratio between the power radiated into +y (Ψ =90,defined as forward direction) and y (Ψ = 270,defined as backward direction) is very close to unity (1.01), as expected from the symmetry of the dimer antenna. One powerful attribute of the EMS technique is that it allows reconstruction of the emission patterns of the different Raman lines from one set of measurements, with the only limitation being the spectrometer bandwidth. Figure 1e,f shows measured Raman emission patterns corresponding to the Stokes Raman lines of 415 and 1586 cm 1, respectively. For the Raman line at 415 cm 1, the radiation peaks at (Θ = 139.1, Ψ = 90.0 ) and (Θ = 139.1, Ψ = 273 ), with a measured F/B ratio of For the Raman line at 1586 cm 1, the radiation peaks at (Θ = 140.6, Ψ =87 ) and (Θ = 140.6, Ψ = 273 ), with a measured F/B ratio of Figure 1d f indicates that, for this structure, the different Raman lines have similar radiation patterns. This is because these Raman lines are enhanced by the same dipolar mode of the dimer antenna, whose radiation pattern varies little over the wavelengths of the different Raman lines. 6 Although dimer antennas are a very common design for SERS substrates, Figure 1 indicates that most of the enhanced Raman scattering occurs into the substrate, peaking around the critical angle. Therefore, in the absence of an oil-immersion objective with a high NA, this light is trapped by the substrate. One could roughen the backside of the substrate for better light extraction, but the objective lens would still need to have a large NA. In addition, the LSPR of dimer antennas is broad, with the different Raman lines therefore having quite similar emission patterns. If the different Raman lines had different emission patterns, then an extra degree of freedom in the design of sensor systems based on SERS would be provided. With these considerations in mind, we consider two directional antenna designs for SERS and measure their emission patterns with the EMS technique. The first design is a YU antenna, and the second consists of dimer antennas formed above a gold film, with integrated gold strips. Raman Emission Patterns from YU Antennas. At radio frequencies, YU antennas are frequently used for their directional properties. Nanoscale YU antennas 29,30 have been shown to produce directional emission from quantum dots. 2 Here we demonstrate beamed Raman scattering using YU antennas. Like its radio frequency counterpart, the nanoscale YU antenna has three components: the feed, the reflector, and the directors. The feed boosts the excitation of, and emission from, nanoscale emitters. The reflector and directors provide antenna directivity. As distinct from previous nanoscale YU antenna designs, we use a dimer antenna, rather than a single nanorod, as the feed antenna in order to use the gap mode 31 to further enhance Raman scattering. 15 This allows us to achieve a high signal-to-background ratio in the EMS measurements (see Supporting Information). The operating wavelength of YU antenna is designed to be 857 nm, which corresponds to the 1074 cm 1 Raman line. We note that this differs from the approach frequently taken in the design of SERS substrates, in which the operating wavelength is chosen to be at the midpoint between wavelengths of the laser excitation and Stokes scattering. 32 Because the main emphasis of our paper is on the direct observation of angular emission patterns, we choose the operating wavelength to exactly match the Stokes line. An array of YU antennas, with a unit cell of 0.5 μm 1.5 μm, is fabricated on an ITO-coated glass coverslip using electron- 6237

4 Figure 2. Raman emission pattern of YU antenna. (a) SEM image of top view of a YU antenna. (b) Measured transmittance of arrays of directors, feed antennas, and reflectors. Unit cell of each array is of μm. (c) FEM simulation of emission pattern resulting from an electric dipole (free space wavelength λ = 857 nm) being placed in center of gap in feed element of YU antenna; red: emission pattern in yz-plane and blue: emission pattern in xz-plane. (d) Emission pattern of thiophenol 1074 cm 1 Raman line retrieved from EMS measurements. Color map represents the normalized scattering intensity. Black dashed circles denote NA of objective lens. Measured emission patterns of (e) 415 and (f) 1586 cm 1 Raman lines. beam lithography, gold evaporation, and lift-off. Each YU antenna consists of five gold nanoparticles (or pairs), as shown in Figure 2a. The feed is a dimer antenna that has the same dimension as the one described in Figure 1, whose LSPR peaks at wavelength of λ = 857 nm as shown by the measurements of Figure 2b. The widths of both the reflector and directors are 65 nm, while the length is varied to achieve the YU functionality at an operating wavelength of 857 nm. The reflector has a length of 187 nm. The LSPR of this reflector is around λ = 902 nm, which is red-shifted compared to the operating wavelength of 857 nm. The three nanoparticles to the right of the feed antenna are directors. The length of each director is 139 nm. The LSPR of these directors is around λ = 765 nm, which is blue-shifted compared with the operating wavelength. The distances between the nanoparticles also play an important role in YU antenna design. 29 The reflector is 187 nm from the feed, while the distance between the feed and its adjacent director is 214 nm. The spacing between each director is also 214 nm. Figure 2c shows the simulated radiation pattern, found by situating an electric dipole with a free space wavelength of λ = 857 nm in the center of gap of the feed and monitoring the farfield intensities. It can be seen that most of the Raman scattering is predicted to occur in the forward direction, with the backward scattering being suppressed. The F/B ratio is predicted by the simulation to be 4.4. To observe directional emission from the YU antennas, we form a SAM of thiophenol on the feed sections only. This is achieved by coating the sample with a layer of e-beam resist [poly(methyl methacrylate), PMMA], then performing electron-beam lithography and development to open windows in the resist on the feed sections. Figure 2d shows the measured emission pattern of the 1074 cm 1 Raman line. It is for this wavelength (λ = 857 nm) that the YU antenna is designed. An underlying ring shape pattern can be seen, similar to that occurring for a dipolar mode radiating above a dielectric substrate (Figure 1d). One key difference of Figure 2d from 6238

5 Figure 3. Functions of components of YU antenna. (a) SEM and (b) measured emission pattern at 1074 cm 1 Raman line from YU antenna without directors. (c) SEM and (d) measured emission pattern at 1074 cm 1 Raman line from YU antenna without reflectors. (e) FEM simulation of emission pattern from YU antenna without directors; red: emission pattern in yz-plane and blue: emission pattern in xz-plane. (f) FEM simulation of emission pattern from YU antenna without reflector. Figure 1d, however, is that the Raman scattering is now predominantly emitted into the direction at (Θ = 137.5, Ψ = 90.0 ), with the fwhm contour having widths of ΔΘ =42 and ΔΨ =70. The angle of Ψ is consistent with the forward direction of the YU antenna (pointing from reflector to director). The measured F/B ratio is 2.21, indicating the YU antenna effectively directs Raman scattering into the forward direction. We also study the directivity of the emission patterns of other Raman lines. Figure 2e,f shows the measured emission patterns for the Raman lines at 415 and 1586 cm 1, respectively. The radiation in the forward direction is still larger than that in backward direction in both cases. However, the measured F/B ratio for Figure 2e,f is 1.34 and 1.15, respectively. From these measurements, we can estimate that our design of YU antennas has a F/B ratio above unity over a bandwidth of 85 nm. This is the same order of magnitude as the fwhm of the LSPR of each nanoparticle, or pair of nanoparticles, of the YU antenna. YU antennas rely on the near-field coupling between the LSPRs on the feed, directors, and reflector to achieve directional radiation. The high EF provided by the feed dimer antenna enables us to study the functionalities of each component of the YU antenna. Figure 3a shows a YU antenna with the same design as before but without the directors. Again, a SAM of thiophenol is formed solely on the feed dimer part. Figure 3b shows the emission pattern of the 1074 cm 1 Raman line found using the EMS technique. It can be seen that a majority of the scattering radiates into forward direction with a F/B ratio of The azimuthal fwhm of ΔΨ = 85 is, 6239

6 however, larger than that of the full YU design. The results confirm that the reflector reflects radiation into the forward direction, thereby increasing the F/B ratio. We next perform experiments on a YU antenna with the same design as before but without the reflector (Figure 3c). The emission pattern of the 1074 cm 1 Raman line is found by the EMS technique (Figure 3d). Due to the absence of reflector, a significant amount of Raman scattering is visible in the backward direction, with the F/B ratio reduced to However, the radiation is more confined in the azimuthal direction, with the azimuthal fwhm being ΔΨ =71, a reduction of 16% over the directorfree YU antenna. These trends are also confirmed by simulations. The YU antenna without reflectors (Figure 3e) is predicted to achieve a F/B ratio of 3.75, while the YU antenna without directors (Figure 3f) is predicted to achieve a F/B ratio of This demonstrates that the reflectors increase the F/B ratio, while the directors shape the radiation pattern of the main lobe to reduce its angular spread. Radiation Emission Pattern of Device Combining Dimer Antennas with Plasmonic Substrate. The combination of dimer antennas with propagating surface plasmon polaritons on planar metal surfaces 22 or metal insulator metal waveguides 23,24 has been studied as a means for achieving beamed Raman scattering or DERS. Here we use the EMS technique to directly image beamed Raman scattering. Figure 4a shows the device under investigation, consisting of dimer antennas on a plasmonic substrate that consists of a gold film (120 nm thick), a SiO 2 spacer (50 nm thick), and gold strips (110 nm wide and 30 nm thick). The underlying substrate is silicon. The dimer antennas are in a square array with a periodicity of 730 nm. The gold strips are periodic in one dimension (periodicity of 730 nm). The SERS performance of this device were previously studied, 24 but the directional nature of the Raman scattering was not measured. In this experiment, an air objective lens (Nikon 100, NA = 0.8) above the sample is used to excite the Raman scattering and to measure the emission pattern. Figure 4b shows a top view SEM of the device. We form a SAM of thiophenol on the device, solely on the dimer antenna surfaces. This is achieved by depositing 2 nm SiO 2 layer on the gold strips before fabricating the dimer antennas. The dimer antennas are located at the midpoints between the strips. It should be noted that the dimer antennas in this device have different dimensions (length of 85 nm for each rod and gap width of 4 nm) from those of the earlier parts of this paper (Figures 1 3). As before, the design optimizes the collimation at λ = 857 nm, appropriate for the 1074 cm 1 Raman line with the laser wavelength of λ = 785 nm. Figure 4c shows the direct imaging of the emission pattern of the 1074 cm 1 Raman line. It is observed that the largest Raman scattering is at Θ =0 and decreases rapidly at larger angles. The fwhm polar angle of the main lobe is ΔΘ =6 in the x-direction (Ψ =0 ) and ΔΘ =57 in the y-direction (Ψ = 90 ). This substantial collimation of the Raman scattering in the x-direction is due to the fact that both the incident laser polarization and the gold strip periodicity are in this direction. An important consequence is that objective lenses with low NAs can be used with this device without much loss in light collection. In addition to the emission peak around Θ =0, Figure 4c shows bright and dimmed circular contours. These contours are confirmed in the simulation of Figure 4d. This simulation predicts the emission pattern at λ = 857 nm that would result, if a point dipole was placed in the gap of a dimer antenna. Rather than simulating this configuration directly, we simulate the fields in the gap that result when plane waves are incident on the device from various angles. We then use these results with the optical reciprocity theorem to predict the emission pattern 22 (see also Supporting Information for details). We now discuss the physical interpretation of the emission pattern. To begin, we note that the enhanced Raman scattering from the dimer antenna can be coupled to the surface modes of the plasmonic substrate that can then be diffracted into free space by the 2D dimer antenna array. Therefore, considering this diffraction effect, conservation of momentum in the xyplane requires k ( ω) =± mg ± ng + k ( ω) in plane x y s (1) where ω is the frequency of Raman scattering, k in plane is the wavevector of Raman scattering projected in xy-plane, m and n are integers, G x ( G y ) is the reciprocal lattice wavevector of the 2D dimer array along the x (y) axis, and k s is the wavevector of the surface modes supported by the plasmonic substrate. To facilitate physical interpretation, we simplify the problem as much as possible. Rather than using 3D modeling to find k s( ω) along all directions within the xy-plane, we instead solely consider waves propagating along the x- and y-axes. We calculate the bandstructures of the plasmonic substrate that shows the dispersion relations k sx (ω) and k sy (ω) along each of these axes using 2D modeling (see Supporting Information). The modeled structure contains the following layers: air/sio 2 / gold/silicon. The gold dimer antennas are not included explicitly in the modeled structure. Their periodicity, however, is implicitly included through the use of G x and G y. For waves propagating along the x-axis, the gold strips serve as strong periodic scatterers. We therefore include these strips in the modeled structure when finding k sx (ω). On the other hand, for waves propagating along the y-axis, the gold strips do not serve as periodic scatterers, and we do not include them in the modeled structure. In both cases, the modes are excited by an electric dipole placed close to the plasmonic substrate (see Supporting Information). The excitation of surface modes is monitored via sample points randomly positioned within the simulation domain, whose sum is plotted as the color scale in Figure 4e,f (see Supporting Information). The modeled bandstructure along the y-axis is shown as Figure 4e. In a manner analogous to the empty lattice approximation, 33 two dispersion curves can be seen to cross at k y = 0. These originate from the surface waves shifted by reciprocal lattice vectors ( ± G y,0). The modeled bandstructure along the x-axis is shown as Figure 4f. It can be seen that there are two dispersion curves that again originate from the surface waves shifted by reciprocal lattice vectors ( ± G x,0). This time, there is an anticrossing at k x = 0. This is akin to what is seen in dielectric photonic crystals 34 and arises from the strong scattering provided by the gold strips. For any given frequency ω, the right-hand side of eq 1 represents dispersion contours k s( ω) of the surface modes offset by integral multiples of the reciprocal lattice vectors of the structure. Figure 4g depicts the case for the wavelength of λ = 857 nm. Here, we approximate the dispersion contours as ellipses with semiaxes of k sx (ω) and k sy (ω). The detection process is limited by the objective lens NA, marked as the 6240

7 Figure 4. Raman emission pattern of device combining dimer antennas with plasmonic substrate. (a) Schematic diagram of device and objective lens. (b) Top view SEM image of the device, showing dimer antennas and gold strips. (c) Emission pattern of thiophenol 1074 cm 1 Raman line retrieved from EMS measurements. Color map represents the normalized scattering intensity. Black dashed circle denotes NA of objective lens. (d) Simulated emission pattern (normalized). (e) Band structure of the plasmonic substrate along y-axis. Color map represents the normalized intensity of the coupled surface modes in log scale. (f) Band structure of the plasmonic substrate along x-axis. (g) Diagram of conservation of momentum in xy-plane according to eq 1. Dashed black circle represents the NA of objective lens. dotted black circle. Note that only the relevant dispersion contours are shown, i.e., those with features within the objective lens NA. At the wavelength of λ = 857 nm, Figure 4f indicates that the two dispersion curves centered at ( ± G x,0) meet at k x = 0, due to the fact that the surface wave propagation constant in the x-direction k sx of the plasmonic substrate matches G x. Similarly, at the same wavelength, Figure 4e indicates that k sy is smaller than G y. Therefore the 2D momentum diagram in Figure 4g shows that two ellipses whose centers lie on the x-axis tangentially contact at the origin. This is clearly observed in Figure 4c,d. Furthermore, the two ellipses whose centers lie on the y-axis do not touch at the origin. This feature is more difficult to see but is present in Figure 4c,d. Figure 4g also indicates that parts of the elliptical momentum contours centered at ( ± G, ± G x y) are also within the objective lens NA. These features are seen as the four partial ellipses around Ψ = 45, 135, 225, and 315 in Figure 4c,d. 6241

8 Figure 5. Bandwidth study of the device with plasmonic substrate. (a) Measured and (b) simulated emission pattern of thiophenol 415 cm 1 Raman line retrieved from EMS measurements. (c) Measured and (d) simulated emission pattern of thiophenol 1586 cm 1 Raman line retrieved from EMS measurements. Grating structures usually diffract different frequency components into different directions. As the wavelength moves away from that for which the design is optimized (λ = 857 nm), the Raman scattering from the device starts to deviate from being largely surface normal. This effect is studied in Figure 5. Figure 5a shows the EMS measurement results for the Raman line of 415 cm 1. The peak radiation is at Θ =0, but the fwhm is increased to ΔΘ =12. This agrees well with the simulation result shown in Figure 5b, which shows peak radiation at Θ =0 and a fwhm of ΔΘ =15. Interestingly, both the measured and simulated emission patterns at this wavelength appear to show two intersecting ellipses. This can be understood by considering the surface modes of the plasmonic substrate according to eq 1. As shown in Figure 4f, at wavelength of λ = 811 nm (415 cm 1 Raman line), the surface wave propagation constant in the x-direction k sx of the plasmonic substrate becomes larger than G x, with k x = k sx G x = 0.065G x. Therefore, in the 2D momentum space along xyplane, the two ellipses along x-axis shown in Figure 4g now have semimajor axes longer than G x. Therefore the 2D emission patterns in Figure 5a,b show two intersecting ellipses. According to eq 1, this momentum mismatch would result in two Raman beams being emitted at angles of Θ = ± 4.1. However, due to the surface modes having finite linewidths, the two ellipses overlap, resulting in the maximum at Θ =0. Figure 5c shows the EMS measurement result for the Raman line at 1586 cm 1. Two radiation branches can be clearly observed with peaks at Θ = ±4, in reasonable agreement with the simulations (Figure 5d) that predict radiation peaks at Θ = ±5. We again interpret this phenomenon by referring to 6242 Figure 4f. At a wavelength of λ = 896 nm (1586 cm 1 Raman line), the surface wave propagation constant in the x-direction k sx of the plasmonic substrate becomes smaller than G x, with k x = k sx G x = 0.080G x. Thus the emission patterns of Figure 5c,d show two ellipses that do not touch. The value of the inplane momentum predicted by Figure 4f suggests that Raman scattering would be mainly diffracted into directions of Θ = ±5.6, which is consistent with the experiments (Figure 5c) and the predictions of simulations (Figure 5d). In conclusion, we have used the energy momentum spectroscopy technique to directly observe the beamed Raman scattering effect. This technique allows Raman scattering to be distinguished from the broad luminescent background continuum that typically occurs in SERS, thereby enabling the Raman emission patterns from different plasmonic structures to be unambiguously measured. It should be noted that an investigation of whether the angular emission patterns of the Raman scattering and the luminescent background are similar or different 35 is beyond the scope of this paper. Using this technique, we demonstrate that directional Raman scattering can be achieved by placing additional plasmonic structures around the feed dimer antennas. In the first case studied, the YU antenna, these additional structures are directors and reflector. In the second case studied, these additional structures are a gold film, a spacer, and gold strips. In both cases, the Raman scattering is shaped into a narrow beam, permitting its efficient collection by a low NA objective lens. This could be advantageous for practical applications in which SERS substrates are used for sensing. The present work is the first direct observation of beamed Raman scattering to the best

9 of our knowledge, and we believe this technique will prove helpful for developing SERS substrates with high collection efficiency. ASSOCIATED CONTENT *S Supporting Information Experimental setup; SERS enhancement factor of YU antenna; energy momentum spectroscopy; methods of simulations for Raman emission patterns; bandstructures of plasmonic substrate. This material is available free of charge via the Internet at AUTHOR INFORMATION Corresponding Author * Notes The authors declare no competing financial interest. ACKNOWLEDGMENTS This work was supported in part by the Harvard Quantum Optics Center. This work was also supported in part by the Center for Excitonics, an Energy Frontier Research Center funded by the U.S. Department of Energy, Office of Science and Office of Basic Energy Sciences under award no. DE- SC Fabrication was performed at the Harvard Center for Nanoscale Systems, which is supported by the National Science Foundation. We thank three anonymous reviewers for insightful criticism of an earlier version of this paper. REFERENCES (1) Kinkhabwala, A.; Yu, Z.; Fan, S.; Avlasevich, Y.; Mu llen, K.; Moerner, W. E. Nat. Photonics 2009, 3, (2) Curto, A. G.; Volpe, G.; Taminiau, T. H.; Kreuzer, M. P.; Quidant, R.; van Hulst, N. F. Science 2010, 329, (3) Wang, C.; Kurtsiefer, C.; Weinfurter, H.; Burchard, B. J. Phys. B: At., Mol. Opt. Phys. 2006, 39, (4) Le Ru, E. C.; Etchegoin, P. G. Annu. Rev. Phys. Chem. 2012, 63, (5) Crozier, K. B.; Sundaramurthy, A.; Kino, G. S.; Quate, C. F. J. Appl. Phys. 2003, 94, (6) Taminiau, T. H.; Stefani, F. D.; Segerink, F. B.; van Hulst, N. F. Nat. Photonics 2008, 2, (7) Shegai, T.; Brian, B.; Miljkovic, V. D.; Ka ll, M. ACS Nano 2011, 5, (8) Aouani, H.; Mahboub, O.; Devaux, E.; Rigneault, H.; Ebbesen, T. W.; Wenger, J. Nano Lett. 2011, 11, (9) Lee, K. G.; Chen, X. W.; Eghlidi, H.; Kukura, P.; Lettow, R.; Renn, A.; Sandoghdar, V.; Go, S. Nat. Photonics 2011, 5, (10) Lieb, M. A.; Zavislan, J. M.; Novotny, L. J. Opt. Soc. Am. B 2004, 21, (11) Kneipp, K.; Wang, Y.; Kneipp, H.; Perelman, L. T.; Itzkan, I.; Dasari, R. R.; Feld, M. S. Phys. Rev. Lett. 1997, 78, (12) Nie, S.; Emory, S. R. Science 1997, 275, (13) Le Ru, E. C.; Etchegoin, P. G. Chem. Phys. Lett. 2006, 423, (14) Talley, C. E.; Jackson, J. B.; Oubre, C.; Grady, N. K.; Hollars, C. W.; Lane, S. M.; Huser, T. R.; Nordlander, P.; Halas, N. J. Nano Lett. 2005, 5, (15) Zhu, W.; Banaee, M. G.; Wang, D.; Chu, Y.; Crozier, K. B. Small 2011, 7, (16) Osberg, K. D.; Rycenga, M.; Harris, N.; Schmucker, A. L.; Langille, M. R.; Schatz, G. C.; Mirkin, C. A. Nano Lett. 2012, 12, (17) Stockman, M.; Faleev, S.; Bergman, D. Phys. Rev. Lett. 2001, 87, (18) Fan, J. A.; Wu, C.; Bao, K.; Bao, J.; Bardhan, R.; Halas, N. J.; Manoharan, V. N.; Nordlander, P.; Shvets, G.; Capasso, F. Science 2010, 328, (19) Ye, J.; Wen, F.; Sobhani, H.; Lassiter, J. B.; Van Dorpe, P.; Nordlander, P.; Halas, N. J. Nano Lett. 2012, 12, (20) Hentschel, M.; Saliba, M.; Vogelgesang, R.; Giessen, H.; Alivisatos, A. P.; Liu, N. Nano Lett. 2010, 10, (21) Wang, D.; Yang, T.; Crozier, K. B. Opt. Express 2011, 19, (22) Chu, Y.; Zhu, W.; Wang, D.; Crozier, K. B. Opt. Express 2011, 19, (23) Ahmed, A.; Gordon, R. Nano Lett. 2011, 11, (24) Wang, D.; Zhu, W.; Chu, Y.; Crozier, K. B. Adv. Mater. 2012, 24, (25) Ahmed, A.; Gordon, R. Nano Lett. 2012, 12, (26) Fromm, D. P.; Sundaramurthy, A.; Kinkhabwala, A.; Schuck, P. J.; Kino, G. S.; Moerner, W. E. J. Chem. Phys. 2006, 124, (27) Beversluis, M.; Bouhelier, A.; Novotny, L. Phys. Rev. B 2003, 68, (28) Taminiau, T. H.; Karaveli, S.; van Hulst, N. F.; Zia, R. Nat. Commun. 2012, 3, 979. (29) Kosako, T.; Kadoya, Y.; Hofmann, H. F. Nat. Photonics 2010, 4, (30) Dregely, D.; Taubert, R.; Dorfmu ller, J.; Vogelgesang, R.; Kern, K.; Giessen, H. Nat. Commun. 2011, 2, 267. (31) Schnell, M.; Garcia-Etxarri, A.; Huber, A. J.; Crozier, K.; Aizpurua, J.; Hillenbrand, R. Nat. Photonics 2009, 3, (32) McFarland, A. D.; Young, M. a; Dieringer, J. a; Van Duyne, R. P. J. Phys. Chem. B 2005, 109, (33) Kittel, C. Introduction To Solid State Physics; 8th ed.; John Wiley & Sons, Inc.: Hoboken, NJ, (34) Crozier, K.; Lousse, V.; Kilic, O.; Kim, S.; Fan, S.; Solgaard, O. Phys. Rev. B 2006, 73, (35) Weber, M. L.; Litz, J. P.; Masiello, D. J.; Willets, K. A. ACS Nano 2012, 6,