Dual-Band Unidirectional Emission in a Multilayered Metal-Dielectric Nanoantenna.

Controlling the emission efficiency, direction, and polarization of optical sources with nanoantennas is of crucial importance in many nanophotonic applications. In this article, we design a subwavelength multilayer metal-dielectric nanoantenna consisting of three identical gold strips that are separated by two dielectric spacers. It is shown that a local dipole source can efficiently excite several hybridized plasmonic modes in the nanoantenna, including one electric dipole (ED) and two magnetic dipole (MD) resonances. The coherent interplay between the ED and MDs leads to unidirectional emissions in opposite directions at different wavelengths. The relative phase difference between these resonant modes determines the exact emission direction. Additionally, with a proper spacer thickness and filling medium, it is possible to control the spectral positions of the forward and backward unidirectional emissions and to exchange the wavelengths for two unidirectional emissions. An analytical dipole model is established, which yields comparable results to those from the full-wave simulation. Furthermore, we show that the wavelength of the peak forward-to-backward unidirectionality is essentially determined by the MD and is approximately predictable by the plasmonic wave dispersion in the corresponding two-dimensional multilayer structure. Our results may be useful to design dual-band unidirectional optical nanoantennas.

Introduction

Optical nanoantennas have been studied widely over the past decades, mostly for their ability to confine the optical fields into volumes of size far below the wavelength of the incoming light.[1] Such nanoantennas are often characterized by different metrics, including scattering efficiency, directivity, emission pattern, and operating frequency bandwidth.[2−4] Among these metrics, directional nanoantennas are particularly useful for modulating and steering the energy radiated by a single quantum emitter (e.g., a quantum dot or nitrogen-vacancy center) and enable the creation of an efficient single-photon source.[5−8] Moreover, a highly unidirectional receiving nanoantenna is a suitable platform for creating a local field with an extreme spatial gradient, which is useful to control quantum source dynamics.[9,10] Sufficiently high directionality has been demonstrated theoretically and experimentally for numerous nanophotonic applications,[11] using Yagi–Uda antennas,[12] core–shell nanoparticle arrays,[13] dielectric nanoparticle arrays,[14] ultracompact antennas,[15−17] and even single metallic or dielectric nanoparticles.[18−22]

Of special interest in this context are nanoantennas exhibiting both electric and magnetic resonances. In general, the interferences between the radiated optical fields from the induced electric dipole (ED) and magnetic dipole (MD) present an additional degree of freedom to tailor light emission characteristics, in terms of frequency, efficiency, phase, direction, and polarization. More specifically, these interactions allow for predictable highly unidirectional light emission if comparable strength of magnetic and electric responses is achieved. These interactions raise interesting possibilities for tailoring optical scattering and emission, such as controlling the coupling between a single-photon emitter and receiver,[18,23] as well as unique processes arising from nonlinear optical interactions.[24]

However, few natural materials exhibit an intrinsic MD transition,[25] and even in those that do, the response is significantly weaker than that of their ED counterparts. Alternatively, optical magnetic responses can be achieved using nonmagnetic materials by properly structuring noble metals or high-index dielectrics, which behave as optical meta-atoms. The properties of such meta-atoms can be tuned via composition, geometry, and size. For active dipolar sources close to these structures with proper position and polarization, the magnetic resonance may become remarkably excitable. In such a scenario, the structure generates ED- and MD-type responses at a comparable level.[26] As such, unidirectional light emission resulting from ED−MD interference becomes possible.[5] In this regard, plasmonic particles are attractive for such a purpose because plasmon resonance occurs with strongly confined and substantially controllable local fields. On this basis, one then expects large enhancement of the optical local density of states (LDOS) that can accelerate the spontaneous emission decay and substantial shaping of the emission patterns.

Recently, Pakizeh and Käll proposed a kind of nanoantenna made of a gold nanodisk dimer.[17] In their system, an ED source can efficiently excite magnetic resonance in the dielectric gap and the induced ED and MD meet the Kerker condition.[27] The local source–antenna coupled system radiates predominantly in one particular direction. Essentially, the directional emission is achieved by tuning the resonance wavelengths of the antenna elements, thereby adjusting the relative phase differences in their radiated fields. For an appropriate configuration, this may result in constructive interference of radiation in a particular direction and destructive interference in the opposite direction. Furthermore, in many applications, it is of great interest to provide similar functional operations at different and selective wavelengths simultaneously.[28,29]

In this work, we show that in contrast to the dimer system a multilayer metal–dielectric–metal (MDM) nanoantenna can support more than one hybridized electric and magnetic dipolar modes. When excited by a properly positioned dipole source, the phase relationship between the different ED and MD resonant modes can be engineered and the overall light emission is controlled predominately toward opposite directions at different wavelengths. Furthermore, it is possible to switch the emission directions for the same wavelengths simultaneously. We demonstrate such a possibility by an optical nanoantenna consisting of three gold strips that are separated by two different dielectric layers, as schematically shown in Figure a. When such a nanoantenna is excited by a nearby dipole source (e.g., a quantum dot), it allows the localized surface plasmon resonances (LSPRs) sustained by the system to strongly interact, leading to excitation of the three hybridized plasmonic resonant modes. The detailed excitation strength and phase relations crucially depend on the retardation effect.[30] More specifically, the parallel gold strips favor coupled electric resonance that is formed by in-phase currents, whereas the dielectric spacers can sustain magnetic resonances formed by out-of-phase (antigoing) currents on the neighboring metallic strips.[31] This article shows that the interference of the scattered fields of these induced electric and magnetic resonances leads to controllable emission with high directionality. Moreover, we demonstrate that the emission direction can switch from one direction to the opposite direction at different wavelengths, along the normal multilayer structure. Furthermore, by judiciously varying the thickness and refractive index of the dielectric spacer, it is possible to switch optical emission from the backward (forward) direction to the forward (backward) direction at the fixed dual-band wavelengths (see the schematic in Figure b,c). The designed multilayered MDM antenna is quite flexible for steering the dipole emission. We expect that this bidirectional antenna can be integrated with dielectric waveguides and can provide an on-chip solution for two-color routing, which may lead to compact devices for wavelength-selective waveguide demultiplexing.[32]

Figure 1

Illustration of unidirectional emissions by the ultracompact optical nanoantenna. (a) Schematic figure of the nanoantenna consisting of three identical gold nanostrips with length l = 100 nm, width w = 100 nm, and thickness t = 20 nm, and they are separated by two different dielectric spacers with thicknesses d1 and d2 and refractive indices n1 and n2. An ED source is positioned at the edge of the middle strip and is polarized along the y axis. The distance between the dipole source and the upper surface of the antenna is 10 nm. (b) Schematic representation of the color-switched directionality effect after setting d1 = 15 nm, d2 = 25 nm, n1 = 1, and n2 = 1. The high directionality is switched from the backward direction (+x) toward the forward direction (−x) as the operating wavelength is changed from λ = 664 to 758 nm. (c) Schematic representation of the color-switched directionality when the parameters are as follows: d1 = 47 nm, d2 = 25 nm, n1 = 1.26, and n2 = 1.32. As the relative phase of ED and MD goes from π to 0 at the resonance of λ = 664 nm and from 0 to π at the resonance of λ = 758 nm, the switching from forward (backward) to backward (forward) scattering occurs at these two wavelengths.

Results and Discussion

Design and Characterization of the Nanoantenna

Figure a shows the antenna geometry, which consists of gold strips with length l = 100 nm and thickness t = 20 nm, separated by two dielectric gap layers of thicknesses d1 and d2 and refractive indices n1 and n2. To examine the emission properties of a nearby dipole source coupled to this antenna (e.g., marked by the red arrow in Figure a), full-wave electromagnetic calculations based on the finite element method (FEM, COMSOL Multiphysics[33]) are performed. The permittivity of gold is taken from Johnson and Christy.[34] To simplify the analysis, the whole structure is assumed to be freestanding in air (ε0 = 1). The source is assumed to be an ideal classical dipole (no initial damping) oscillating at the emission frequency with a dipole moment amplitude, p0 = 1/2πf C m, where f is the working frequency. As schematically shown in Figure b,c, we would like to achieve a flexible control on the highly directional emission by this MDM antenna that is specifically coupled to the nearby ED source.

First, we consider the case in which d1 = 15 nm, d2 = 25 nm, and n1 = n2 = 1. Such a nanoantenna can greatly enhance the LDOS, which remarkably affects the spontaneous decay rate of the ED source.[35] A figure of merit to describe the enhancement is the radiative decay rate enhancement, Γrad = Prad/P0, where Prad is the radiated power flowing through a closed surface surrounding both the dipole source and the antenna and P0 is the radiated power of the dipole source in free space.[30]Figure a shows Prad spectra for the nanoantenna. To justify the calculated results by the FEM, we have superimposed the results of Prad obtained by the discrete dipole approximation (DDA).[36] Perfect agreement between FEM (black solid curve) and DDA (green squares) is seen in Figure a. It is further seen that Prad is enhanced more than 10 times at the two distinct peaks, λ = 672 and 746 nm, as well as at a relatively indistinguishable peak, λ ≈ 621 nm (see the black solid line and the squares). This enhancement is stronger than that in the three-layer core–shell particles.[37] The three peaks correspond to three optical resonances excited in the structure. When the nanoantenna is illuminated by a normally incident plane wave from either the +x axis or −x axis that is linearly polarized along the y direction, the peak at λ = 621 nm can be observed clearly in the corresponding scattering spectrum. The resonance modes at λ = 672 and 746 nm are weakly excitable (see Figure S1 in Supporting Information). This indicates that the resonance at λ = 621 nm corresponds to a superradiant (relatively bright) mode, whereas the other two resonances at λ = 672 and 746 nm result from the subradiant (relatively dark) modes. Furthermore, all of these modes partially overlap in the spectrum and may have different radiation interferences that can shape the emission pattern of a nearby active source. The differential directionality (measured in decibels) of an antenna reads as followswhere θ and φ are the spherical angles and S(θ, φ) is the radiated power in the given directions θ and φ. Here, we are primarily interested in the forward–backward (FB) ratio, GFB = DF – DB = 10 log 10(SF/SB), where SF and SB are the far-field radiated powers in the forward (θ = 90°, φ = 90°) and backward directions (θ = 90°, φ = 270°), respectively. The FB ratio, GFB, allows a rough assessment of the directionality of the antennas. In our case, it reaches about −21 dB at λ1 = 664 nm and nearly 20 dB at λ2 = 758 nm (see the green solid line in Figure b). Notice that positive (negative) GFB means that forward (backward) emission dominates. Figure a,b further shows that the spectral positions of extreme GFB (at λ1 = 664 nm and λ2 = 758 nm) are very close to those of the two lower-energy subradiant modes (i.e., λ = 672 and 746 nm).

Figure 2

Simulation results of the antenna with d1 = 15 nm, d2 = 25 nm, n1 = 1, and n2 = 1. (a) Radiative decay rate enhancement Γrad. (b) Far-field forward-to-backward directionality GFB (green curve) by FEM numerical and KFB (red curve) defined by [eq ], respectively. (c) The radiated powers of the ED and MD moments and their relative phase difference Δφ (green dashed curve). (d)–(f) The real part of E and the normalized H field distributions for the peak wavelengths of the ED and MD moment radiated powers in (c). “M” and “D” represent the metal and dielectric layers, respectively. (g) Schematic diagram illustrating the hybridization between the monomer on the left and the dimer plasmonic strips on the right. The solid arrows in the particles represent the EDs, and the green dashed arrows inside the gaps represent the MDs. Three-dimensional (3D) radiation patterns for peak GFB are depicted in the bottom, at wavelengths λ = 664 and 758 nm.

Because the multilayer MDM nanoantenna may sustain multipole resonances,[30] it is essential to identify the multipolar moment contributions to the far-field at the resonant wavelength (i.e., λ1 = 664 nm and λ2 = 758 nm). The radiated power of the electric and magnetic multipoles can be calculated by the induced volume current density j(r) inside the nanostructure.[38] As expected, the contributions from the induced ED and MD are significantly larger than those of the magnetic quadrupole (MQ) and electric quadrupole (EQ; see Figure S2). Therefore, it is reasonable to neglect the higher-order multipole moments that have negligible effects on the radiative decay rate and the radiation pattern.

Figure c shows the radiated powers, PED (violet solid line) and PMD (pink solid line), of the induced dipole moments p and m. It is seen that PMD maximizes at λ = 672 and 746 nm, whereas a relatively large PED covers nearly the entire visible spectrum. The broad PED spectrum indicates that the ED resonance has a low quality factor. The ED resonance position is at about λ = 621 nm, identified from the scattering spectrum by plane-wave excitation (see Figure S1). To gain further insight into the resonant dipole modes, we plot the near fields (E or H) for the three peak wavelengths in Figure d–f. Careful examination of the E distribution shows that the first peak at λ = 621 nm corresponds to an electric resonance in which the LSPRs on the three plasmonic strips oscillate in-phase (see Figure d). Figure e,f shows the H distribution in the nanoantenna for the two magnetic resonances at λ = 672 and 746 nm, respectively. It is evident that the magnetic fields are mainly localized and enhanced inside the left and right dielectric gaps, respectively. The electric fields over the left two (right two) strips oscillate out of phase at λ = 672 nm (λ = 746 nm), which forms strong circulating displacement currents in the left (right) dielectric gap, highlighting strong cavity magnetic resonance modes (for details, see Figure S3). As a result, it is reasonable to assume the induced ED to be oriented along the emitter polarization direction (i.e., y axis) and the MD to be oriented along the z direction. A corresponding three-layer metal–dielectric strip nanoantenna with similar parameters can support one ED and one MD, giving rise to highly directional emission at only one specific wavelength (see Figure S4).

Dipole Interaction Model

To qualitatively understand the roles of the two kinds of dipoles in achieving the dual-band unidirectional emission control, a schematic figure of the dipolar interference is sketched in Figure g. The hybridization occurs between a strip monomer and a strip dimer. When the monomer and the dimer are coupled strongly to each other, three hybridization cases can be characterized by one ED (p) and two MDs (m(1) and m(2)). The solid arrows in the particles represent the EDs, and the green dashed arrows inside the gaps represent the MDs. For excitation by a dipole source at the proper position and polarization, the excitation spectra of p and m overlap partially, which enables the dual-wavelength unidirectional emission, as reflected by the GFB spectrum. The 3D radiation power pattern for the dip of GFB at λ1 = 664 nm also is depicted in Figure g, showing an almost complete cancellation of the radiation toward the right-half space (+x direction) and a predominated emission to the left-half space (−x direction). In sharp contrast, for the low-energy peak of GFB at λ2 = 758 nm, the nanoantenna–dipole source coupling system emits almost completely toward the +x direction. For a three-layer MDM strip nanoantenna with similar geometry and materials (i.e., the parameters are the same as in Figure ), the spectral position of GFB matches precisely with the spectral position of the MD resonance (see Figure S4b,c). For the five-layer MDM nanoantenna, the spectral positions of extreme GFB values slightly deviate from the resonance wavelengths of the corresponding two MDs. This deviation is partially due to the coupling effect (although weak) between the two magnetic cavity modes. Moreover, in this configuration, the nanoantenna ED and MD are significantly larger than the active dipole source because of the strong LSPR. As such, the total system’s far field can be expanded into a multipolar series that takes into account only the first two leading terms (i.e., ED and MD). With dipole moments p and m, the antenna’s far field reads[19]where k is the wave vector, n = k/|k| is the unit vector in the emission direction, and r is the coordinate vector. According to eq , zero forward or backward scattering requires the following: (1) the two dipoles (ED and MD) to be orthogonal to each other; (2) the radiated powers to be nearly identical, and (3) the phase difference Δφ = φ – φ = 0 (forward scattering) or Δφ = ±π (backward scattering), where φ and φ are the relative phases of the induced p and m in the nanoantenna. In our case, when Δφ = φ – φ ≈ 0 is fulfilled (i.e., the first Kerker condition), most emission is directed to the right-half space (+x direction). Likewise, when Δφ = φ – φ ≈ ±π (i.e., the second Kerker condition), the emission dominates in the left-half space (−x direction).

In our system, with comparable amplitudes (see the solid curves in Figure c) and relatively small phase difference, Δφ = 0.13π, of the induced p and m(2) at λ = 758 nm (see the green dashed curve in Figure c), the first Kerker condition is approximately fulfilled. Similarly, the second Kerker condition is nearly met at λ = 664 nm with Δφ = 0.94π. We can have Δφ ≈ ±π and Δφ ≈ 0 in more than one position. However, highly unidirectional emission does not happen, as seen in the spectra of GFB, because the radiated power of ED is much larger than that of MD at those spectral positions. Furthermore, Δφ is sensitive to the active dipole position. As a result, the sign of the extreme value of GFB strongly depends on the local source position (see Figure S5).

Next, to demonstrate the mechanism of directional emission more clearly, we reconstruct the two-dimensional (2D) radiation patterns using an ED and MD interference model (D-M model). The induced ED p and MD m are calculated numerically as p = 6.74 e0.45π × 10–17 C m (2.48 e0.55π × 10–17 C m) and m = 18.39 e–0.49π × 10–9 A m2 (7.84 e0.42π × 10–9 A m2) for λ1 = 664 nm (λ2 = 758 nm). Then, eq is used to calculate the far-field and the radiation patterns. The results from the full FEM calculation and the D-M model are compared in Figure a–d. The results are in good agreement, and the slight differences are attributed to negligible contributions from the higher-order multipoles, which are not accounted for in the D-M model.

Figure 3

(a) Normalized 2D radiation patterns in the xz plane calculated by the FEM (solid curves) and the D–M model (dashed curves) for λ = 664 nm. (b) Radiation patterns in the xy plane calculated by the FEM (solid curves) and the D–M model (dashed curves) for λ = 664 nm. (c, d) Radiation patterns in the xz and xy planes calculated by the FEM (solid curves) and the D–M model (dashed curves) for λ = 758 nm, respectively.

On the basis of eq , the complementary FB directionality, KFB, can be defined as follows[19]As shown in Figure b, the spectrum of KFB (red solid curve) matches overall, but it is slightly red-shifted, with respect to that of GFB (black solid curve). This deviation is ascribed partially to the dissipation in the nanoantenna structure. Figure b also shows that the extreme KFB wavelength matches the spectral positions where Δφ = 0.03π and 0.96π. Similar results arise in plasmonic systems with ED and EQ interference.[19]

Theoretical Analysis of Cavity Resonances

On the basis of the previous analysis, it becomes possible to control the peak positions of GFB by tuning the magnetic resonances. The magnetic cavity modes in the MDM structure are considered to result from the gap plasmonic waves.[31] On the basis of the Fabry–Pérot (FP) model, the resonant condition of the cavity modes in our structure is approximately represented as follows[39,40]where kgsp denotes the gap surface plasmon polariton (SPP) propagation constant, p and q are the quantum numbers in the y and z directions (p = 1 and q = 0 in our case). Here, Leff is the effective side length of this resonator, which accounts for the fringe field effect that may be considered as a phase shift in the gap plasmonic wave upon reflection at the cavity edges. Similarly, in the circular plasmonic patch nanoantennas,[30,31] we can replace Leff with l + d in eq as the first-order approximation, where d is the gap thickness. The peak position of GFB for three-layer MDM strip nanoantennas can be approximated by eq directly (see Figure S6). The five-layer MDM structure can be viewed effectively as two intertwined three-layer MDM nanoantennas. In our case, the fringe field effect is expected to be different in the two dielectric layers (see Figure S7) and the phase shift increases as the gap thickness increases.[41] Thus, Leff = l + d1 (or Leff = l + d2) is considered to be the effective side length of the FP resonator at dielectric layer d1 (or d2).[42]Figure a shows the numerically calculated resonance frequencies of the MD cavity modes versus Leff = l + d1 (or Leff = l + d2). Figure b shows the GFB spectrum with the extreme positions labeled by squares and circles. The data were collected for different geometry lengths, l. As the side length increases from l = 100 to 250 nm, the MD resonances are red-shifted and the peak positions of GFB experience the same varying trend. For a larger size (e.g., l = 200–250 nm), higher-order multipole components contribute significantly to the scattered field, leading to enhanced directionality at high frequency (see the green and orange curves in Figure b).

Figure 4

(a) Magnetic resonant frequencies (symbols) vs π/(l + d1) or π/(l + d2) for different side lengths l of the nanoantenna, superimposed with the dispersion relation (solid curve) of the corresponding 2D multilayer structure. (b) GFB spectra for different side lengths l of the nanoantenna. The rest of the parameters are the same as in Figure .

According to eq , the resonant frequency and the size of the metal–dielectric resonator are correlated with kgsp (ω), which can be examined by the transfer matrix method for the corresponding 2D multilayer system.[30,31,43,44]Figure a shows that the peak positions of extreme GFB approximately fall on the dispersion relation, kgap (ω) (i.e., the black solid curves). The dispersion has several branches, and the lowest two are in accordance with the symmetric plasmonic gap modes in the two gap layers of thicknesses d1 and d2, respectively. More interestingly, it is confirmed that all of the magnetic modes excited in the same dielectric layer fall on the same dispersion branch, justifying the validity of the 2D analysis in eq . Figure clearly shows that the peak positions of GFB can be approximately predicted by eq , by properly setting the effective side length Leff.

Switching of Unidirectional Emission

Because the dispersion relation, kgsp (ω), depends on a bunch of parameters, including the material dielectric function (n1 and n2) and the geometry (e.g., l, t, d1, and d2), the hybridized MDs certainly not only depend on l but also depend on the gap thickness and refractive index of the dielectric spacer.[45] As a result, the directionality of GFB can be remarkably controlled by d1, d2, n1, and n2.

First, we examine the case of gradual variation of the refractive index, n2, from 1 to 1.5, while keeping the rest of the parameters the same as in Figure . The directionality of GFB, the radiated power of ED and MD, and their relative phase difference Δφ are shown in Figure a–r. As seen in the first column of Figure , the spectral position of extreme GFB experiences a slight redshift and the peak amplitude is gradually reduced when n2 changes from 1 to 1.2. Simultaneously, the amplitude of MD is gradually reduced because of the weak interactions of the three metallic particles, whereas the amplitude of ED remains almost unchanged (see the second column of Figure ). As a result, the mismatch between the ED and MD amplitudes increases, leading to decreased GFB. Because the reduction is rather gradual, it allows for certain flexibility in the choice of parameters. More interestingly, when the refractive index approaches n2 = 1.3, the signs of the two extreme GFB values become negative, indicating that Δφ values at both these two wavelengths are close to π (see Figure i). Furthermore, when n2 increases to 1.4 (see Figure m–o), the high-energy peak of GFB disappears because the MD becomes off-resonance and its amplitude is almost extinguished. Further increasing n2 to 1.5, it is seen that the signs of peak GFB at the short wavelength (λ = 701 nm) and the long wavelength (λ = 804 nm) flip, changing from negative (positive) to positive (negative), compared to those in the initial case of n2 = 1 (the first row of Figure ).

Figure 5

Far-field forward-to-backward directionality of GFB [see panels (a), (d), (g), (j), (m), and (p)], amplitude of the induced ED and MD moments [see panels (b), (e), (h), (k), (n) and (q)], and relative phase difference of p and m [see panels (c), (f), (i), (l), (o), and (r)] for various refractive indices of the spacer from n2 = 1 to 1.5. (s, t) Normalized radiation patterns in the xy plane for the peak wavelengths when n2 varies; the rest of the parameters remain the same as in Figure c.

To demonstrate the unidirectional emission switching more clearly, the normalized 2D radiation patterns at the xy plane for different n2 values at the two peaks of GFB (i.e., the shorter and longer wavelengths labeled λ1 and λ2) are plotted in Figure s,t, respectively. It is seen that the refractive index of the dielectric spacer plays a crucial role in determining the directionality of the nanoantenna. One can switch the emission between the forward (backward) and backward (forward) directions with a single element for two independent wavelengths. A similar behavior can be observed by properly changing the thickness of the dielectric spacer (see Figure S8). The nature of this switching can be attributed to differences in the phase retardation effect on the left two (or right two) metal strips when excited by the dipole source, for systems characterized by different parameters (e.g., refractive index and thickness of the dielectric spacer). Such a phase retardation variation can be seen more clearly through careful examination of the near field. For example, |H| is localized mainly inside the left and right dielectric gaps at the shorter and longer wavelengths, respectively, for the two MD resonances in the case of n2 = 1.5, which is opposite to the situation of n2 = 1, as shown in Figure e,f.

The switching between forward (backward) and backward (forward) directionalities when n2 = 1 is changed to n2 = 1.5 is predicted by eq . Figure shows that by properly setting the effective length, Leff, which depends on the excitation of m(1) (Leff = l + d1) or m(2) (Leff = l + d2), the resonant frequencies of MDs for n2 = 1 (red symbols) and n2 = 1.5 (green symbols) approximately fall on the corresponding dispersion curves of the gap SPPs (see the red curve for n2 = 1 and the green curve for n2 = 1.5). Furthermore, the positive peak of GFB (circles) falls on the low-energy branch for n2 = 1 but on the high-energy branch for n2 = 1.5. Correspondingly, the negative peak (dip) of GFB (triangles) falls on the high-energy branch for n2 = 1 and on the low-energy branch for n2 = 1.5. Moreover, we also examined the gradual variation of the refractive index, n1, from 1 to 1.5 (see Figure S9a), while keeping the rest of the parameters the same as those in Figure . The results show that the positive GFB peak position is gradually red-shifted and the negative positions remain almost unchanged because the high-energy dispersion branch is relatively more insensitive to n1 (see Figure S9b). We also study the forward and backward scattering directionalities of GFB for different emitter polarizations (see Figure S10). The peaks of GFB have a spectral shift, whereas peak signs do not change at the short or long wavelength for different emitter orientations. This phenomenon can be explained by the Kerker condition, and color-switching can be achieved by different emitter polarizations. To this end, we have demonstrated the switching between forward (backward) and backward (forward) directionalities by properly tuning one parameter (thickness or refractive index of the dielectric layer), as shown in Figure (also see Figure S8). However, the operating wavelengths cannot be fixed in these schemes. In view of the dispersion band dependence on both the geometry and material, we attempt to utilize gap thicknesses d1 and d2 simultaneously (or n1 and n2) to obtain the emission direction switching for two fixed wavelengths. Figure a shows the effects of gap thicknesses d1 and d2 on GFB. Four cases of d1 = 10 nm (d2 = 20 nm), d1 = 15 nm (d2 = 25 nm), d1 = 25 nm (d2 = 35 nm), and d1 = 50 nm (d2 = 60 nm) are shown. It is evident that by varying the gap thickness, d1 or d2, the emission directionality can be altered remarkably. Because an increase in d1 and d2 leads to weaker near-field couplings (∼1/d3) of the LSPRs,[46] a blueshift for both the forward and backward directionalities is observed. This blue-shifting saturates for further increased gap thickness because the near-field interactions may diminish at a sufficiently large separation. In a similar strategy, a noticeable redshift can be seen in Figure b when we increase n1 or n2 from 1 to 1.5 (i.e., silicon dioxide). For all such cases, several combinations of d1 (d2) and n1 (n2) exist that make it possible to tune the directional emission between the forward (backward) and backward (forward) directionalities at fixed wavelengths. For example, we can first change n2 to 1.5 while retaining n1 = 1 to switch the sign of GFB at the peak positions and then we can properly increase d1 and d2 to compensate the wavelength shift.

Figure 6

Cavity resonant frequencies vs π/(l + d1) or π/(l + d2) for the nanoantenna with n2 = 1 and n2 = 1.5. The rest of the parameters remain the same as in Figure c.

Figure 7

Directionality of GFB tuned by the gap layer thickness and the refractive index. (a) Effects of thicknesses d1 and d2. (b) Effects of refractive indices n1 and n2.

To better illustrate this point, we carefully adjust and eventually set d1 = 47 nm, d2 = 25 nm, n1 = 1.26, and n2 = 1.32 and compare this nanoantenna to the one studied in Figure a. Figure shows that the lowest two dispersion branches of these two cases coincide. In a similar strategy as shown in Figure , it is easy to demonstrate the exchange between forward (backward) and backward (forward) directionalities at a fixed wavelength. Figure a shows the spectrum of GFB for this case (see the green dotted line) compared to that for the case of d1 = 15 nm, d2 = 25 nm, n1 = n2 = 1 (see the black dotted line). The forward-to-backward emission direction exchange at the same wavelength, λ = 758 nm (λ = 664 nm), is seen clearly. The normalized radiation pattern in the xy plane for both antennas is shown in Figure b. According to these patterns, for properly selected parameters d1 = 15 nm, d2 = 25 nm, n1 = 1, and n2 = 1, at λ = 664 nm, the dipole source emission is highly directed (GFB = −21 dB) toward the −x direction (the solid line in Figure b). However, in a dramatic contrast, at the wavelength of λ = 758 nm, the emission is highly directed (GFB = +20 dB) toward the +x direction (the solid line in Figure c). Conversely, for properly set parameters d1 = 47 nm, d2 = 25 nm, n1 = 1.26, and n2 = 1.32, an orange-color emission is directed toward the +x direction (the dashed line in Figure b) at λ = 664 nm and a red-color emission is directed toward the −x direction at λ = 758 nm (the dashed line in Figure c). This unambiguously demonstrates the color-switching behavior as schematically shown in Figure b,c. In this way, the proposed nanoantenna acts like a two-channel optical demultiplexer. Therefore, it shows great potential for applications in integrated optical circuits, nano-optics, and nanophotonics.

Figure 8

Cavity resonant frequencies vs π/(l + d1) or π/(l + d2) for two specific nanoantennas with d1 = 15 nm, d2 = 25 nm, n1 = 1, and n2 = 1 and d1 = 47 nm, d2 = 25 nm, n1 = 1.26, and n2 = 1.32.

Figure 9

Bi-unidirectional switching with gap thickness and refractive index of the dielectric spacers. (a) Far-field forward-to-backward directionalities of GFB sign reversal at fixed wavelengths for two nanoantennas with different geometries and dielectric parameters. (b, c) Normalized 2D radiation patterns in the xy plane at the peak wavelengths, λ = 664 and 758 nm, for the two cases shown in (a).

It is important to study the directivity when the antenna is placed on a dielectric substrate because this is the typical experimental realization for optical antennas.[29,47] The emitter–antenna system depicted in Figure is placed on a glass substrate (ε = 2.25; see Figure a). The positive and negative directivity peaks also show up at 758 and 664 nm because the electromagnetic fields are localized mostly inside the antenna. Figure b,c shows the resulting angular far-field directionality for the two wavelengths. The emission is directed mainly toward the substrate[47] and is scattered into the +x (−x) direction at λ = 758 nm (λ = 664 nm). Thus, the present antenna can be fabricated to experimentally realize the bidirectional color routing.

Figure 10

Emitter–antenna system in Figure placed on a glass substrate (ε = 2.25). (a) Schematic figure of the nanoantenna on the glass substrate. (b, c) Normalized 2D radiation patterns in the xy plane of the peak wavelengths, λ = 664 and 758 nm. The substrate fills the half-space from θ = 180 to 360°.

Conclusions

We design and numerically demonstrate a dual-band ultracompact plasmonic nanoantenna with controllable high unidirectionality for a nearby dipole emitter. The nanoantenna supports two spectrally tunable MD modes with a strength that is comparable to that of the overall ED resonance. Superior forward and backward radiation patterns are achieved, with forward-to-backward directionality reaching +20 and −21 dB, respectively. Analyses based on a simple dipole model are performed, and the reconstructed radiation patterns agree well with the full-wave simulation results. Furthermore, we show that the spectrum of the extreme directionality can be approximated by the SPP dispersion in the corresponding 2D metal–dielectric multilayered structure and can demonstrate the possibility of color-switching. Our findings significantly broaden the versatility of plasmonic devices, and we envision the possible practical implementations in meta-materials, photon couplers, and chemo- or biosensors.

Methods

Transfer Matrix Method

First, we extend the radius to infinity (i.e., l → ∞) and treat the system as a seven-layer film system with air on the top and at the bottom, consisting of alternating three homogeneous metallic layers and two homogeneous dielectric layers. For TM polarization, the transfer matrix that relates electric fields across the interface from layer i to j readswhere η = kε/kε; here, ε is the permittivity in material i and k is the x-component of the wave vector in material i and can be written as , where k0 and k∥ are the free space wave vector and the in-plane propagation constant, respectively. The field at x + Δx can relate to this at position x in material j by matrix ϕThen, the total transfer matrix to describe the nine-layer structure can be written as M = D12ϕ2D23ϕ3D34ϕ4D45ϕ5D56ϕ6D67 and the reflection coefficient as rc = M2,1/M1,1. When the reflection coefficient, rc, is undefined (i.e., M1,1 = 0), it corresponds to the eigenmodes of the structure, and then, the gap plasmon dispersion relation (kgsp) of this nanostructure can be obtained.

FEM Modeling

The radiative decay rate spectrum and the multipole analysis for the multilayered antenna were conducted using the radio frequency module of commercial FEM software COMSOL Multiphysics. The ED emitter was modeled as an electric current dipole moment with amplitude 1/(2πf), where f is the frequency. To absorb light scattered in all directions, the entire simulation space was then surrounded by a perfectly matched layer (PML), whose thickness equals λmax/2, where λmax is the longest wavelength in the frequency domain study. Scattering boundary conditions were imposed on PML boundaries to preclude reflections. Far-field calculations were performed on a sufficiently large spherical surface (radius, R = 1 m) to ensure the accuracy of the near-field to far-field transformation. Test runs with different mesh sizes and domain sizes were conducted until convergence was reached.

DDA Method

DDSCAT 7.1 was chosen for calculating the radiative decay rate using the DDA method. The total system contains around 1.4 × 107 dipoles, and test runs with different dipoles were also conducted until convergence was reached.