Letter pubs.acs.org/NanoLett Metamaterial Perfect Absorber Based Hot Electron Photodetection Wei Li and Jason Valentine* Department of Mechanical Engineering, Vanderbilt University, Nashville, Tennessee 37212, United States S Supporting Information * ABSTRACT: While the nonradiative decay of surface plasmons was once thought to be only a parasitic process that limits the performance of plasmonic devices, it has recently been shown that it can be harnessed in the form of hot electrons for use in photocatalysis, photovoltaics, and photodetectors. Unfortunately, the quantum eﬃciency of hot electron devices remains low due to poor electron injection and in some cases low optical absorption. Here, we demonstrate how metamaterial perfect absorbers can be used to achieve near-unity optical absorption using ultrathin plasmonic nanostructures with thicknesses of 15 nm, smaller than the hot electron diﬀusion length. By integrating the metamaterial with a silicon substrate, we experimentally demonstrate a broadband and omnidirectional hot electron photodetector with a photoresponsivity that is among the highest yet reported. We also show how the spectral bandwidth and polarization-sensitivity can be manipulated through engineering the geometry of the metamaterial unit cell. These perfect absorber photodetectors could open a pathway for enhancing hot electron based photovoltaic, sensing, and photocatalysis systems. KEYWORDS: Plasmonics, metamaterial perfect absorber, hot electrons, infrared photodetection guides.13,14,17 However, the main drawback of previously demonstrated hot electron detectors is their low photoresponsivity, typically on the order of a few tens to 100 μA/ W.11,16 Recent studies19,30 have reported signiﬁcant advances in detector eﬃciency by exciting propagating surface plasmon polaritons (SPPs) using metal gratings19 or combining gratings with deep trench cavities.30 For the latter case, a large eﬃciency enhancement was achieved due to high optical absorption as a result of the cavity and the use of thin metal layers, allowing the electrons to diﬀuse to the metal-semiconductor interface before thermalization. However, due to the use of a grating to excite SPPs, the optical absorption and photocurrent will be sensitive to the incident angle in these designs. In this Letter, we present an approach to greatly enhance photoresponsivity of hot electron detectors through the use of metamaterial perfect absorbers (MPAs) that do not require an optically thick ground plane. This allows us to integrate the MPA with semiconductor layers that require high temperature processing or growth. In our case, the MPA architecture is integrated with n-type silicon (n-Si) to realize Schottky diode based hot-electron photodetectors with high photoresponsivity in the near-infrared region, well below the Si bandgap energy. We experimentally characterize numerous MPA-based photodetectors with varying degrees of polarization sensitivity and photoresponse bandwidth. By taking advantage of the strong optical absorption in MPAs, combined with ultrathin metal layers, we demonstrate photoresponsivities that are among the highest reported while also preserving an omnidirectional, polarization insensitive, and broadband response. Speciﬁcally, we demonstrate a detector with an absorption full width at half- Surface plasmons (SPs), coherent oscillations of electrons in metals that can be excited with electromagnetic waves, are a key component in routing and manipulating light at nanometer length scales.1−3 Because of their deep-subwavelength mode volumes and high-ﬁeld concentrations, SPs provide a means to realize numerous techniques, and optical devices such as nanoscale imaging,5 subwavelength light concentrators,4 ultracompact lens and waveplates,6 photodetectors and modulators,1 photothermal devices,7,8 metamaterials,9 and plasmonic enhanced photovoltaic devices.10 Following excitation, surface plasmons can either decay radiatively into re-emitted photons or nonradiatively into energetic electrons. In most cases, the nonradiative decay of plasmons is a parasitic process that limits the performance of plasmonic devices. While much eﬀort has been devoted to suppressing nonradiative SP decay, recent research has shown that the energetic, or hot, electrons can be harnessed for a number of applications including photodetection,11−20 photovoltaic devices,21,22 photocatalysis,23−26 and surface imaging.27,28 In terms of photodetection, hot electron devices are typically formed by placing the metal surface in contact with a semiconductor,11 forming a Schottky barrier. Hot electrons generated from the nonradiative decay of SPs can transport to the metal-semiconductor interface before thermalization and be injected into the conduction band of the semiconductor, resulting in photocurrent. In this case, the photodetector’s bandwidth is limited by the Schottky barrier height, instead of the bandgap of the semiconductor. This allows one to achieve broadband absorption at energies below the bandgap of the semiconductor,11,29 provided the plasmonic structure has a broad absorption spectrum. SP-based hot electron detectors have been demonstrated using a variety of structures including nanorods,11,29 nanowires,16,20 metal gratings,19,30 and wave© 2014 American Chemical Society Received: March 23, 2014 Published: May 16, 2014 3510 dx.doi.org/10.1021/nl501090w | Nano Lett. 2014, 14, 3510−3514 Nano Letters Letter Figure 1. (a) Schematic of the MPA unit cell with upper and lower resonators (gold) integrated with a semiconductor (blue). (b) Simulated electric ﬁeld distribution in the MPA under illumination with TM-polarized light. (c) Simulated optical reﬂection, transmission, and absorption for an MPA with L = 160 nm, P = 320 nm, H = 120 nm metallized with a 1 nm thick Ti adhesion and 15 nm thick Au layer. maximum (fwhm) up to 800 nm and a photoresponsivity larger than 1.8 mA/W from 1200 to 1500 nm. This design is also scalable to arbitrary wavelengths within the solar spectrum and could potentially open a pathway for enhancing the eﬃciency of hot electron based photovoltaic, sensing, and photocatalysis systems. To begin, we ﬁrst examine a polarization sensitive MPA photodetector composed of 1D metal stripes, a schematic of which is illustrated in Figure 1a. The MPA is composed of a prepatterned n-type Si substrate, forming Si stripes, with a 1 nm titanium adhesion layer and a 15 nm gold ﬁlm deposited on the surface. The metal ﬁlm is naturally separated into upper and lower plasmonic stripe resonators. When the MPA is illuminated with TM-polarized light (E-ﬁeld perpendicular to the stripe direction), localized surface plasmon resonances (LSPRs) are excited on the upper and lower resonators (Figure 1b). A Fabry−Perot resonance is also present in the cavity formed from the upper and lower stripe resonator layers. By carefully designing the resonator length L, cavity height H, and periodicity P, the resonator and cavity resonances can be overlapped, leading to near unity absorption as calculated using a full-wave simulation and shown in Figure 1c. Furthermore, the thickness of the plasmonic resonator is only 15 nm, less than the hot electron diﬀusion length,31 ensuring that the hot electrons have a high probability of diﬀusing to the metal− semiconductor interface. It should be pointed out that in contrast with most of the previously demonstrated MPA structures32−34 all the metal constituents in our proposed MPA design are ultrathin and an optically thick metallic ground plane is not required. This ensures that all of the optical absorption and hot electron generation occurs in the thin plasmonic resonator layers. This architecture also eliminates the need to grow a high quality semiconductor on top of a metal ﬁlm which would be challenging due to the high process temperatures required. To experimentally characterize the photodetector, fabrication began by ﬁrst deﬁning the stripe patterns on a double side Figure 2. (a) Schematic of the fabricated 1D MPA photodetector including a thin metal coating on the sidewall. Dimensions of the MPA detectors D1, D2, and D3 are L = 160, 170, and 170 nm and P = 320, 320, and 340 nm, respectively. The etching depth, H, for the three devices was 120 nm. (b) SEM image of a fabricated device. (c) Experimentally measured (solid lines) and simulated (dashed lines) absorption spectra of D1, D2, and D3 (red, green, and blue solid lines, respectively). (d) Experimentally measured (circles) and calculated (lines) photoresponsivity spectra of D1, D2, and D3 (red, green, and blue, respectively). polished n-type Si wafer using PMMA and electron beam lithography (EBL). The MPA detectors were patterned with an overall size of 85 μm × 85 μm. The PMMA was used as a mask for patterning the underlying n-type Si using reactive-ion etching (RIE). After the RIE process, the sample was immersed in a 10:1 buﬀered oxide etchant (BOE) for 1 min to completely remove the native oxide. The sample was then immediately transferred into an evaporation chamber followed by deposition of 1 nm Ti and 15 nm Au layers. More details regarding fabrication can be found in Section 1 and Figure S1 of the Supporting Information. Three devices, D1, D2, and D3, were fabricated (Figure 2a,b) with dimensions of L = 160, 170, and 170 nm and P = 320, 320, and 340 nm, respectively. The etching depth H for the three devices was 120 nm, which was measured using atomic force microscopy (AFM). The optical absorption spectra of the devices were acquired by measuring the reﬂection R and transmission T in an infrared microscope coupled to a grating spectrometer. The reﬂection spectra were normalized to a silver mirror and the transmission spectra were normalized to transmission through air with absorption calculated as A = 1 − T − R. The optical absorption of these devices is shown in Figure 2c (more details are provided in Section 2 and Figure S2 of the Supporting Information). The devices show broadband optical absorption in the near IR region with a maximum value of 95% and a fwhm of 400 nm. In the fabricated structure, there is a thin coating of 3511 dx.doi.org/10.1021/nl501090w | Nano Lett. 2014, 14, 3510−3514 Nano Letters Letter metal on the sidewall of the Si ridge due to the slight sidewall angle. As a result, the absorption peak positions of the fabricated structure are slightly blue shifted compared to the structure simulated in Figure 1. However, good agreement is observed between the simulated absorption spectrum with the sidewall coating and the measured absorption spectrum (Figure 2c). More details regarding the numerical simulations are provided in Section 3 and Figures S3−S5 of the Supporting Information. The connection between the upper and lower resonators also ensures that both of the metal layers are included in the electrical circuit and can contribute to the total photocurrent. To characterize the hot-electron transfer eﬃciency of the devices, the sample was wire bonded to a chip carrier and photocurrent measurements were performed on the MPA detectors across a wavelength range of 1200 to 1500 nm (Figure 2d). These MPA photodetectors have broad photoresponsivity spectra with peak values of 3.37 mA/W, 3.05 mW, and 2.75 mW for devices D1, D2, and D3, respectively. Furthermore, these measurements were performed without an external bias and the current varied linearly with the incident power (Supporting Information Figure S6), indicating that nonlinear processes are not playing a role in the response. These photoresponsivities are several orders of magnitude larger than previously demonstrated LSPR based nanoantenna photodetectors11 and over four times larger than the SPP-based metal grating photodetectors.19 The photoresponsivities of the MPA detectors are comparable with the recently reported deep trench cavity structure30 at short wavelengths. However, the MPA detectors have a broader bandwidth resulting in larger photoresponsivity at long wavelengths. The photocurrent in the MPA detectors can be understood by considering the hot electron transfer process over the Schottky barrier, which is dictated by a three-step model35,36 as shown in Figure 3a. (1) Plasmons nonradiatively decay into hot electrons (optical absorption), (2) hot electrons transport to the metal-semiconductor interface before thermalization, and (3) hot electrons inject into the conduction band of the semiconductor through internal photoemission. Here, we analyze the two diﬀerent components of the MPA detectors separately, the upper resonator and the lower resonator, which yield diﬀerent hot electron generation and quantum transmission probabilities. As a result, the photoresponsivity spectrum of MPA detectors can be calculated as R (ν ) = ∑ A1i(ν)η2(ν)η3i(ν) i=u,l Figure 3. (a) Schematic of the hot electron transfer process over the Schottky barrier formed by the metal−semiconductor interface. Steps 1 to 3 correspond to hot electron generation, diﬀusion to the Schottky interface, and transmission to the conduction band of semiconductor. (b) FDTD simulation of absorbed power density in a 1D MPA device at 1250, 1350, and 1500 nm (plotted on a logarithmic scale for better visualization). (c) Calculated photoresponsivity (solid lines) and absorption (dash lines) in a MPA device. The total (red), upper resonator (green), and lower resonator (blue) contributions are plotted separately. and mean free path of the hot electrons.31 ηi3(ν) is the internal photoemission of hot electrons across the Schottky interface for the upper and lower resonator and is calculated using the modiﬁed Fowler equation37,38 η3i(ν) = C Fi (3) hν where is the device-speciﬁc Fowler emission coeﬃcient, hν is the photon energy, and qϕB = 0.54 eV is the Schottky barrier height for the MPA devices which is extracted from the current−voltage characteristics of the device (Supporting Information Figure S7).38 Fitting the experimentally measured photoresponsivity spectrum with eq 1, where η2(ν) and CiF are treated as ﬁtting parameters, results in good agreement with the experimental data as shown in Figure 2d. The extracted ratio of ClF/CuF is around 3 for all three devices, indicating that hot electrons generated in the lower resonator have a higher quantum transmission probability compared with the upper resonator. This eﬀect is mainly due to the formation of a 3D Schottky interface between the semiconductor and the embedded bottom resonator, allowing the hot electrons to cross the interface over a larger emission cone.16 While total absorption drops at longer wavelengths, the bottom resonator layer’s contribution remains relatively constant, resulting in it dominating the photocurrent at long wavelengths (Figure 3c). Strong broadband absorption A1(ν) and high transport eﬃciency η2(ν) in the ultrathin metal ﬁlm ultimately ensure high photoresponsivity over a large bandwidth. CiF (1) where u and l represent the upper and lower resonators, respectively, and ν is frequency. Ai1(ν) is the optical absorptivity of each component, which can be determined by the measured total absorption spectrum of the MPA detectors (Figure 2c) along with the absorption distribution in each resonator layer. The absorption distribution in the MPA devices (Figure 3b) are calculated numerically using the local ohmic loss, given by Q (r , ω) = 1 ω Im(εm)|E ⃗(r , ω)|2 2 (hν − qϕB)2 (2) where E⃗ (r,ω) is the local electric ﬁeld and Im(εm) is the imaginary part of the metal permittivity. This allows us to determine the wavelength dependent absorptivity of the upper, Au1, and lower, Al1, resonator (dash lines in Figure 3c). η2(ν) is the probability that the hot electron will transport to the Schottky interface and is dependent on the spatial distribution 3512 dx.doi.org/10.1021/nl501090w | Nano Lett. 2014, 14, 3510−3514 Nano Letters Letter Figure 4. (a) Schematic of the polarization-independent MPA photodetectors. Dimensions of MPA photodetectors D4, D5, and D6 are L = 185, 195, and 195 nm and P = 340, 340, and 360 nm, respectively. The etching depth, H, for the three devices was 135 nm. (b) SEM image of a fabricated device. (c) Experimentally measured (solid lines) and simulated (dashed lines) absorption spectra of D4, D5, and D6 (red, green, and blue lines, respectively) (d) FDTD simulation of the angularly dependent optical absorption for p- (left) and s-polarized light (right). (e) Experimentally measured (circles) and calculated (lines) photoresponsivity spectra of D4, D5, and D6 (red, green, and blue, respectively). detector is larger than 1.8 mA/W from 1200 to 1500 nm (Figure 5d), making it a candidate for broadband, polarizationindependent, and omnidirectional silicon based near-IR photodetectors. The performance of the MPA detector could be further improved by engineering the Schottky barrier height, roughing the metal semiconductor interface,14 or embedding the top resonator layer within the semiconductor.16 Also, while we have focused on maximizing the bandwidth of the photoresponse, the tunability of the MPA architecture also oﬀers a possibility to achieve a narrowband photoresponse. For instance, a narrow Fano resonance can be realized in these structures by coupling a grating mode with the LSPR resonance, leading to an extremely narrow absorption band39 that could potentially lead to a plasmonic spectrometer with direct electrical readout.19 Furthermore, although the demonstrated MPA detectors work in the near IR region, they can be scaled to the visible regime by replacing Si with a higher bandgap semiconductor such as TiO2. In the visible regime, the use of ultrathin metal layers in this type of MPA design could be particularly beneﬁcial due to the drastically reduced hot electron diﬀusion length.31 To summarize, we have proposed and experimentally demonstrated the ﬁrst metamaterial perfect absorber hot electron photodetector. Realizing near unity absorption within ultrathin plasmonic nanostructures allows the eﬃciency of the hot electron transfer process to be signiﬁcantly enhanced, resulting in photoresponsivities that are among the highest reported. Furthermore, the MPA architecture allows us to preserve a broadband, omnidirectional, and polarization invariant response. This ultracompact, CMOS-compatible Sibased MPA detector can also be readily integrated into on-chip optoelectronics and can be scaled to the solar spectrum, leading to enhanced eﬃciencies in hot-electron based photodetection, photovoltaic, and photocatalysis systems. In order to achieve a polarization-independent photoresponse, the 1D stripe resonators can be replaced by square resonators, forming a 2D MPA structure as shown in Figure 4a,b. In this case, the upper resonator layer can still contribute to the photocurrent due to the thin metal sidewall coating. Experimental validation of this concept was accomplished using three devices D4, D5, and D6 with dimensions of L = 185, 195, and 195 nm and P = 340, 340, and 360 nm, respectively. The etching depth H for the three devices was 135 nm, conﬁrmed by AFM. These polarization-independent MPA structures still maintain near-unity absorption (Figure 4c) over a large bandwidth. Furthermore, a peak absorptivity above 90% is maintained for incident angles of up to 75 and 60° under p- and s-polarized light, respectively (Figure 4d). This LSPR driven response is in contrast to previously demonstrated metal grating19 and the deep trench cavity30 hot electron photodetectors which rely on the excitation of SPPs through grating coupling, resulting in an angularly sensitive response. As a result, these 2D MPA detectors show high photoresponsivities (Figure 4e) while also preserving a broadband, omnidirectional, and polarization-independent response. Taking advantage of the high photoresponsivity and tunability of the 2D MPA structures, we can further enhance the absorption bandwidth by implementing MPAs with multiple resonator sizes. To demonstrate this, we combine two diﬀerent 2D MPA unit cells, forming a bipartite checkerboard supercell, shown in Figure 5a,b. In this case, the separate resonances of the two MPA unit cells result in bandwidth enhancement, though there is a slight decrease in peak absorption. However, the broadband MPA detector with dimensions L1 = 185 nm, L2 = 225 nm, P = 680 nm, and H = 160 nm, still maintains absorptivity over 80% from 1250 to 1500 nm (Figure 5c). Although the experimental fwhm of the absorption was not measurable due to the limited detection range of the spectrometer, data from simulations show a fwhm of over 800 nm. The photoresponsivity of this broadband MPA 3513 dx.doi.org/10.1021/nl501090w | Nano Lett. 2014, 14, 3510−3514 Nano Letters Letter (5) McLeod, A.; Weber-Bargioni, A.; Zhang, Z.; Dhuey, S.; Harteneck, B.; Neaton, J. B.; Cabrini, S.; Schuck, P. J. Phys. Rev. Lett. 2011, 106, 037402. (6) Yu, N.; Capasso, F. Nat. Mater. 2014, 13, 139−150. (7) Neumann, O.; Feronti, C.; Neumann, A. D.; Dong, A.; Schell, K.; Lu, B.; Kim, E.; Quinn, M.; Thompson, S.; Grady, N.; Nordlander, P.; Oden, M.; Halas, N. J. Proc. Natl. Acad. Sci. U.S.A. 2013, 110, 11677− 11681. (8) Coppens, Z. J.; Li, W.; Walker, D. G.; Valentine, J. G. Nano Lett. 2013, 13, 1023−1028. (9) Valentine, J.; Zhang, S.; Zentgraf, T.; Ulin-Avila, E.; Genov, D. A.; Bartal, G.; Zhang, X. Nature 2008, 455, 376−379. (10) Atwater, H. A.; Polman, A. Nat. Mater. 2010, 9, 205−213. (11) Knight, M. W.; Sobhani, H.; Nordlander, P.; Halas, N. J. Science 2011, 332, 702−704. (12) Wang, F.; Melosh, N. A. Nano Lett. 2011, 11, 5426−5430. (13) Goykhman, I.; Desiatov, B.; Khurgin, J.; Shappir, J.; Levy, U. Nano Lett. 2011, 11, 2219−2224. (14) Goykhman, I.; Desiatov, B.; Khurgin, J.; Shappir, J.; Levy, U. Opt. Express 2012, 20, 28594−28602. (15) Fang, Z.; Liu, Z.; Wang, Y.; Ajayan, P. M.; Nordlander, P.; Halas, N. J. Nano Lett. 2012, 12, 3808−3813. (16) Knight, M. W.; Wang, Y.; Urban, A. S.; Sobhani, A.; Zheng, B. Y.; Nordlander, P.; Halas, N. J. Nano Lett. 2013, 13, 1687−1692. (17) Casalino, M.; Iodice, M.; Sirleto, L.; Rendina, I.; Coppola, G. Opt. Express 2013, 21, 28072. (18) Wang, F.; Melosh, N. A. Nat. Commun. 2013, 4, 1711. (19) Sobhani, A.; Knight, M. W.; Wang, Y.; Zheng, B.; King, N. S.; Brown, L. V.; Fang, Z.; Nordlander, P.; Halas, N. J. Nat. Commun. 2013, 4, 1643. (20) Chalabi, H.; Schoen, D.; Brongersma, M. L. Nano Lett. 2014, 14, 1374−1380. (21) McFarland, E. W.; Tang, J. Nature 2003, 421, 616−618. (22) Clavero, C. Nat. Photonics 2014, 8, 95−103. (23) Li, J.; Cushing, S. K.; Zheng, P.; Meng, F.; Chu, D.; Wu, N. Nat. Commun. 2013, 4, 2651. (24) Mubeen, S.; Lee, J.; Singh, N.; Krämer, S.; Stucky, G. D.; Moskovits, M. Nat. Nanotechnol. 2013, 8, 247−251. (25) Mukherjee, S.; Libisch, F.; Large, N.; Neumann, O.; Brown, L. V.; Cheng, J.; Lassiter, J. B.; Carter, E. A.; Nordlander, P.; Halas, N. J. Nano Lett. 2013, 13, 240−247. (26) Mukherjee, S.; Zhou, L.; Goodman, A. M.; Large, N.; AyalaOrozco, C.; Zhang, Y.; Nordlander, P.; Halas, N. J. J. Am. Chem. Soc. 2014, 136, 64−67. (27) Giugni, A.; Torre, B.; Toma, A.; Francardi, M.; Malerba, M.; Alabastri, A.; Proietti Zaccaria, R.; Stockman, M. I.; Di Fabrizio, E. Nat. Nanotechnol. 2013, 8, 845−852. (28) Schuck, P. J. Nat. Nanotechnol. 2013, 8, 799−800. (29) Ueno, K.; Misawa, H. NPG Asia Mater. 2013, 5, e61. (30) Lin, K.-T.; Chen, H.-L.; Lai, Y.-S.; Yu, C.-C. Nat. Commun. 2014, 5, 3288. (31) Sze, S. M.; Moll, J. L.; Sugano, T. Solid-State Electron. 1964, 7, 509−523. (32) Liu, X.; Starr, T.; Starr, A. F.; Padilla, W. J. Phys. Rev. Lett. 2010, 104, 207403. (33) Liu, N.; Mesch, M.; Weiss, T.; Hentschel, M.; Giessen, H. Nano Lett. 2010, 10, 2342−2348. (34) Liu, X.; Tyler, T.; Starr, T.; Starr, A. F.; Jokerst, N. M.; Padilla, W. J. Phys. Rev. Lett. 2011, 107, 045901. (35) Spicer, W. Phys. Rev. 1958, 112, 114−122. (36) Spicer, W. E. Appl. Phys. 1977, 12, 115−130. (37) Fowler, R. H. Phys. Rev. 1931, 38, 45−56. (38) Sze, S. M.; Ng, K. K. In Physics of Semiconductor Devices, 3rd ed.; John Wiley & Sons, Inc.; Hoboken, NJ, 2007; pp 164, 682. (39) Shen, Y.; Zhou, J.; Liu, T.; Tao, Y.; Jiang, R.; Liu, M.; Xiao, G.; Zhu, J.; Zhou, Z.-K.; Wang, X.; Jin, C.; Wang, J. Nat. Commun. 2013, 4, 2381. Figure 5. (a) Schematic of the broadband MPA photodetector. Dimensions of the MPA photodetector are L1 = 185 nm, L2 = 225 nm, P = 680 nm, H = 160 nm. (b) SEM image of the fabricated device. (c) Experimentally measured (solid) and simulated (dashed) absorption spectra of the broadband MPA detector. (d) Experimentally measured (circles) and calculated (line) photoresponsivity spectra of the broadband MPA photodetector. ■ ASSOCIATED CONTENT S Supporting Information * Details of device fabrication, modeling, characterization, and Figure S1−S7. This material is available free of charge via the Internet at http://pubs.acs.org. ■ AUTHOR INFORMATION Corresponding Author *E-mail: [email protected]. Notes The authors declare no competing ﬁnancial interest. ■ ACKNOWLEDGMENTS This work was supported by the National Science Foundation (NSF) under program CBET-1336455. Device fabrication was performed at the Vanderbilt Institute of Nanoscale Science and Engineering (VINSE); we thank the staﬀ for their support. ■ REFERENCES (1) Schuller, J. A.; Barnard, E. S.; Cai, W.; Jun, Y. C.; White, J. S.; Brongersma, M. L. Nat. Mater. 2010, 9, 193−204. (2) Gramotnev, D. K.; Bozhevolnyi, S. I. Nat. Photonics 2010, 4, 83− 91. (3) Stockman, M. I. Opt. Express 2011, 19, 22029−22106. (4) Stockman, M. I. Phys. Rev. Lett. 2004, 93, 137404. 3514 dx.doi.org/10.1021/nl501090w | Nano Lett. 2014, 14, 3510−3514