1. Introduction

Surface plasmons are highly-localized electromagnetic excitations that arise from coupling of light with collective oscillations of electrons at the metal/dielectric interfaces [1]. They can be either propagating surface waves at the extended, possibly nanostructured, metallic boundaries (surface plasmon polaritons, SPPs) or localized modes supported by metallic nanoparticles (localized surface plasmons, LSPs). The ability of surface plasmons to confine, control and enhance light at a subwavelength scale together with their high sensitivity to environmental changes provide unique prospects for light manipulation at the nanoscale [2–5], development of highly-integrated photonic integrated circuits [6,7], optoelectronic devices [8,9] and artificial materials with an engineered or greatly enhanced optical response [10–15], together with advances in nonlinear optics [16,17] and chemo-/bio-sensing [18–20].

As plasmonic resonances are highly dependent on the shape of the metallic nanostructures, symmetry breaking of the nanostructure geometry results in pronounced sensitivity of the plasmonic response to the polarization of the incident light. Compared to conventional methods, polarization manipulation with nanostructures provides both polarization and phase control with high spatial resolution. In this respect, extensive efforts have been made to develop various nanostructure designs with polarization-dependent optical behavior, particularly using Fano resonances [21–25]. For example, properly positioned pairs of rectangular nanoholes can be used for polarization-sensitive tunable directional coupling of light to SPPs [26,27], generation of a holographic contour [28,29] and control of Airy plasmons [30]. Along with this, a plasmonic filter based on polarization-controlled reproduction of colors was demonstrated with arrays of asymmetric cross-shaped nanoapertures in a thin aluminum film [31]. The double-nanoring pattern can realize polarization-insensitive as well as polarization-sensitive dual-band absorption. A surface plasmon polarization-controlled beam splitter based on slot waveguides and integrated in a plasmonic circuitry was realized for applications in optical signal processing and optical computation [32]. However, most of these devices need to be fabricated using rather expensive lithographic techniques or ion beam milling, which hinders their large-scale manufacturing.

In this paper, we employ a robust and inexpensive self-assembly method based on nanosphere lithography to realize large-scale arrays of anisotropic coaxial plasmonic nanocavities with an engineered plasmonic response and achieve polarization management of the reflected light in both visible and NIR spectral regions. Designing nanocavity surface plasmon resonances supported by the nanocavities provides the means to control absorption efficiency for different polarization states of the incident light over a wide range of incident angles. The contrast in the reflection for orthogonal polarizations is determined by the geometrical parameters of nanoellipsoids producing the core of the coaxial nanocavities and can be easily defined at the fabrication stage. Moreover, by tuning the nanoellipsoid height, tilt angle and semi-axes values it is possible to control not only the far field polarization state, but also the local vectorial distribution of the electromagnetic field. The proposed structure, featuring inexpensive and scalable self-assembly fabrication, can find applications in polarization-sensitive optical switches, spatial light modulators and optical sensing devices.

2. Self-assembly fabrication of anisotropic nanocavity arrays

The arrays of anisotropic plasmonic nanocavities (APNs) were fabricated using a nanosphere lithography method combined with tilted metal deposition. The fabrication process includes four steps (Fig. 1(a)). Firstly, hexagonal polystyrene sphere (PS) arrays (p = 690 nm) embedded in a silica gel layer were formed by self-assembly of the spheres upon drying from a water solution with an added silica gel (20 h, 1%vol PS, 0.01%vol silica precursor solution). Secondly, the PS spheres were partially etched by O₂ reactive ion etching (RIE) to form shallow ring gaps between the remaining polystyrene core and the silica gel matrix (not affected by the etching). Thirdly, a silver film was deposited on the etched structures by magnetron sputtering. Adjusting the time of the magnetron sputtering, one can control the vertical base height h of the nanocavities (Fig. 1(b)). Finally, a tilted thermal evaporation was employed for the directional metal deposition to obtain the APNs. During this process asymmetric core structures in a form of nanoellipsoids defining the geometry of the plasmonic nanocavity are formed. The evaporation angle (70°, Fig. 1(a)) defines the nanoellipsoids’ tilt angle θ, while the evaporated film thickness and PS etching degree determine their aspect ratio s = a/b, where a and b are the nanoellipsoid semi-axes.

 

Fig. 1. (a) Fabrication process of the APN arrays: (1) hexagonal PS spheres arrays (period p = 690 nm) embedded in a silica gel are formed by self-assembly of the spheres, (2) the PS spheres are etched using RIE, (3) a silver film is deposited by magnetron sputtering, (4) tilted nanoellipsoids are formed at the center of the nanocavity using directional (θ = 70°) evaporation. (b) Cross-sections of the obtained APN designs with separated (APN-S) and attached (APN-A) nanoellipsoids. (c) SEM images of the APN arrays with various geometrical parameters. All scale bars are 500 nm.

Download Full Size | PDF

Importantly, as a result of the tilted deposition, the nanoellipsoids are oriented in the plane of the sample, producing a hexagonal array of plasmonic coaxial nanocavities with an anisotropic geometry and therefore an anisotropic optical response (Fig. 1(a), (b). Depending on the fabrication conditions, the nanoellipsoids can be either separated from the outer walls of the nanocavity or attached to them by one end. The obtained APN arrays with various geometrical parameters are shown in Fig. 1(c). Particularly, four combinations of magnetron sputtered silver film thicknesses (h = 50 nm and h = 100 nm) and evaporated film thicknesses (80 nm (s = 1.37) and 160 nm (s = 2)) are presented. From the top SEM images one can estimate the lengths of the semi-axes and the distances between nanoellipsoids and cavity walls along a semi-axis, e.g. 2a·sinθ = 304 nm, 2b = 235 nm, L₁ = 130 nm and L₂ = 148 nm for the top-left structure in Fig. 1(c). As the thermal evaporation time increases, a and L₁ increase, while b and L₂ are kept almost constant, which results in the increase of s. Thus, the geometry parameters h, s, a, b and θ can be continuously tuned during the fabrication process, resulting in the robust control of the polarization-dependent optical response of the nanostructure. The high uniformity of the fabricated samples can be seen in the large-area SEM images presented in Fig. SI-1 of the Supplement 1.

3. Nanostructure operation

3.1 Polarization-sensitive optical response

To understand the polarization-dependent optical response of the nanocavity arrays, numerical simulations of their reflectivity at normal incidence were performed using a finite element method (COMSOL Multiphysics software package). As the optical response largely depends on whether the nanoellipsoids are separated or attached, both cases with characteristic parameters taken from fabrication were numerically simulated, particularly APN-S (separated) and APN-A (attached). In either case the center of the ellipsoid nanoparticle was lifted above the top surface of the magnetron-deposited silver layer (having the height h, Fig. 1(b)) at a distance of 80 nm. The obtained reflection spectra, presented in Fig. 2, show pronounced minima, corresponding to the resonant excitation of plasmonic modes. As a result of the symmetry breaking in the individual nanostructures and the consequent anisotropy of the array, different reflection spectra are observed for the two principal orthogonal polarizations of the incident light (φ = 0° and φ = 90°, defined in Fig. 1(a)).

 

Fig. 2. Numerically simulated normal-incidence reflectance spectra of the coaxial arrays with (a) separated (APN-S, a = 200 nm, b = 120 nm, h = 50 nm, θ = 70°, s = 1.66) and (b) attached (APN-A, a = 240 nm, b = 120 nm, h = 100 nm, θ = 70°, s = 2.33) core nanoellipsoids. (c), (d) Cross-sections of (a) APN-S and (b) APN-A, respectively, presenting reflectance spectra at φ = 0° and 90°.

Download Full Size | PDF

3.2 Origin of the plasmonic modes

To identify the origin of the excited plasmonic modes related to the reflectivity minima, both the electromagnetic field and current density maps at the resonance wavelengths were analyzed (Fig. 3). Particularly, the distribution of the electric field in the gaps between the nanoellipsoids and the nanocavity walls indicates the resonant excitation of nanocavity plasmonic modes unavoidably hybridized with the plasmonic modes of the nanoellipsoid, as can be further observed in the corresponding surface charge distributions. These modes can have predominantly localized (S1 and S5/S6) or coupled/hybrid (to some extent collective, the rest of the modes) nature, which was revealed by simulation of arrays with artificially enlarged periods (see Supplement 1, Fig. SI-2 – 4). In the case of separated nanoellipsoids, high-order nanocavity modes S1, S2 and S4 keep their spectral positions with the change of the polarization angle φ from 0° to 90°, while their excitation magnitudes can be modified (Fig. 2). On the other hand, the spectral position of the nanocavity mode hybridized with the nanoellipsoid dipolar resonances along the short (S5) and the long (S6) nanoellipsoid semi-axes (Fig. 3(a)) significantly depends on the incident polarization. In addition, for the polarization along the nanoellipsoid long semi-axis a localized gap mode appears (S3), confined in the narrowest point of the nanocavity.

 

Fig. 3. Numerically simulated distributions of the electric field component collinear with the polarization direction of the incident plane wave, presenting the excitation of resonant plasmonic modes (a) in the case of isolated core nanoellipsoids (APN-S) with the corresponding charge distributions on the metallic surface presented in (b), and (c) in the case of the nanoellipsoids attached to the outer nanocavity wall (APN-A), modelled for the normal incidence and geometrical parameters from Fig. 1. The correspondence of the excited plasmonic resonances to the polarization state of the incident light φ = 0° (black curves in Fig. 2) and φ = 90° (red curves in Fig. 2) is reflected by the color of the mode labels.

Download Full Size | PDF

In the case of nanoellipsoids attached to the cavity walls (case APN-A), as φ changes from 0° to 90°, reflectance dip A1 changes the spectral position from 559 nm to 596 nm, and new modes A3, A4 and A6 appear at 723 nm, 950 nm and 1504 nm, respectively, while mode A5 disappears. Assessing the field map presented in Fig. 3, A5 is the dipolar-type nanoellipsoid mode (an analogue of mode S5), which in this case can be excited only for one polarization across the elliptical nanoparticle. Its counterpart in the orthogonal direction disappeared, as the nanoparticle is merged with the outer nanocavity wall. Examining the field maps of the modes in the APN-A case, one can see that although these modes are strictly speaking not the coaxial nanocavity modes, as the nanocavity now has an interruption produced by the attached nanoellipsoids, they are related and the former can be evolutionarily traced to the latter. Finally, it can be noted that any intermediate polarization of the incident light can be decomposed into two orthogonal components at 0° and 90°, which will result in the simultaneous excitation of both sets of modes with the corresponding magnitudes and phase delays.

3.3 Influence of geometrical parameters

In order to demonstrate the possibility of engineering of the optical response of the anisotropic nanocavity arrays via the control of their geometry at the fabrication stage, numerical simulations of their reflectivity as a function of various geometrical parameters were performed (Fig. 4). Taking the attached-nanoparticle APN-A case with h = 50 nm, θ = 70°, a = 280 nm and b = 120 nm as an initial design, each of the above parameters was parametrically swept while the other kept constant, monitoring the reflectivity for both incident polarizations (φ = 0° and φ = 90°). For φ = 90°, as the values of h and b increase, wavelengths of all plasmonic modes experience a linear red shift. With the increase of θ, the spectral position of mode A3 exhibits a blue shift, while A4 and A6 move to longer wavelengths. It is important to note that there is a characteristic anti-crossing in the modal dispersion at θ = ∼57° where it happens a transition between the separated and the attached elliptical nanoparticle cases, which may be a result of hybridization of the nanocavity modes with the localized plasmonic modes appearing in the narrow gap between the nanoellipsoid tip and the wall. With the increase of a, all plasmonic modes experience a red shift. Here, the transition from separated case to attached case accompanied with the mode anti-crossing happens at a = ∼250 nm. For the illumination with φ = 0° polarization, the changes of h, θ and a show little influence on the spectral positions of all plasmonic modes. However, as b increases, modes A1, A2, and A5 exhibit significant red shifts, as this is the parameter significantly affecting the width of the nanocavity gap and therefore the optical properties of these modes localized in this region, as can be seen from their field distributions presented in Fig. 3(b).

 

Fig. 4. Numerically simulated reflectance of the APN arrays as a function of the nanocavity parameters: nanoellipsoid base height h (a, e), tilt angle θ (b, f), semi-axes a (c, g) and  b (d, h) for two orthogonal polarizations φ = 90° (a–d) and φ = 0° (e–h). Unless varied, the structural parameters are h = 50 nm, θ = 70°, a = 280 nm and b = 120 nm.

Download Full Size | PDF

Overall, one can see that the polarization-sensitive reflectivity happens across the entire visible to near-infrared spectral region and therefore the difference between the reflectivity for φ = 0° and φ = 90° can be efficiently engineered. By comparing of simulation results for the cases APN-S and APN-A, it is clear that the plasmonic resonances for the case of the attached nanoellipsoid (APN-A) are more sensitive to the polarization variation than that for the case of separated one (APN-S), which can be attributed to a higher level of anisotropy of the structure.

4. Experimental investigation

4.1 Polarization-dependent optical properties

We proceed with experimental investigation of the nanocavity arrays. The experimental setup for measurement of the reflection spectra at normal incidence is presented in Fig. 5(a). A 50 W halogen lamp was used as a broadband optical source. The emitted light was collimated using a pair of lenses before passing through a diaphragm with an aperture, which reduces the size of the light spot to 2 mm. Subsequently, a tunable polarizer (wavelength range 200 – 4000 nm, resolution 2 degrees) was introduced into the optical path to linearly polarize the light. The resulting light beam was incident normally on the sample. The angle in respect to the major anisotropy axes of the sample φ was changed by in-pane rotation of the sample. In order to capture the reflected signal, a half-transparent mirror was positioned in front of the sample. The reflected light is then collected using a 10× objective lens and focused into a multimode optical fiber. Finally, the light signal was analyzed using optical spectrum analyzers (Ocean QE65000 and Ocean NQ 512NIR covering 200 – 1000 nm and 900 – 1700 nm ranges, respectively). Reflection from an Ag film was used as a reference to calculate the sample reflectivity. For measurement of reflection spectra at various angles of incidence α, the half mirror was removed and the lamp together with the lenses, diaphragm and polarizer was placed on a rail positioned at the top of a rotation plate (Fig. 5(b)). The sample was placed on another rotation plate. The rail and sample rotation plate were properly rotated around the same axis to vary the incidence angle with the resolution of 2° and the 10× objective was used to collect the reflected light.

 

Fig. 5. Experiment setup for measurement of the reflectance spectra at (a) normal and (b) oblique incidence.

Download Full Size | PDF

Figures 6(a) and 6(b) present the polarization-dependent reflectance spectra of the fabricated separated APN arrays with s = 1.37, h = 50 nm, s = 1.37 and h = 100 nm. At h = 50 nm, as φ changes from 0° to 90° the excitation of mode S5 gradually decreases, with mode S6 appearing instead, which leads to an effective red shift of the corresponding reflectivity minimum from 1010  nm to 1100 nm with a polarization angle sensitivity (PAS) of ∼1.2 nm/deg. When h = 100 nm, the PAS increases to ∼2.1 nm/deg, which is accompanied, though, with its larger effective FWHM. The measured results are consistent with the simulation counterparts with a slight discrepancy that can be attributed to the difference between the idealized geometry of the models and the exact geometry of the measured samples. Similar trends are observed for a larger thickness of the sputtered film h = 100 nm.

 

Fig. 6. Experimentally measured reflectance of the separated (a, b) and almost attached (c, d) APN arrays as a function of polarization of the incident light for APNs with a large aspect ratio (s = 2) for h = 50 nm (a, c) and h = 100 nm (b, d). The measurements were performed using visible (400 – 1000 nm) and NIR (1000 – 1600 nm) spectrometers.

Download Full Size | PDF

When the evaporated film thickness is increased to 160 nm, resulting in s = 2, the upper side of the nanoellipsoids is almost attached to the cavity wall, which corresponds to the APN-A scenario. In this case, much more pronounced reflectivity changes with the variation of the incident polarization can be observed (Fig. 6(c), (d). With the change of the polarization angle φ, there is an obvious change of the set of the excited modes from (A1, A2, A5) at 0° to (A3, A4, A6) at 90°. The change of the reflection at a given wavelength due to the polarization-sensitive mode excitation reaches many tens of percent, which can be used to realize an optical switch controlled by the polarization of the incident light. Furthermore, as can be seen in Fig. 6(c) and 5(d), the operational wavelength of the switch can be adjusted throughout the visible and near-IR regions by changing only one fabrication parameter h. The measured results (h = 50 nm in Fig. 6(c) and (h) = 100 nm in Fig. 6(d)) agree quite well with the simulation counterparts, as shown in the line graphs showing the comparison for the two major polarization cases of φ = 0° and φ = 90°, with all the plasmonic resonances could be clearly identified. At the same time, the experimental resonances show substantially larger line widths, which can be explained by structure-to-structure inhomogeneity of the sample and higher losses of the deposited Ag in comparison to the tabulated data used in the simulations. The latter factor can also contribute to the lower experimental reflectance at shorter wavelengths.

4.2 Illumination-dependent optical properties

The dependence of reflectance spectra on the angle of incidence for both φ = 0° and φ = 90° polarization directions were experimentally measured for the two cases of the base height h (Fig. 7). Due to the in-plane asymmetry of the structure for both polarizations, the measurements were taken in the two principal planes, as illustrated in the top row of Fig. 7(a). While the plasmonic resonances excited with the electric field in the incidence plane containing the long semi-axis (φ = 90°, Fig. 7(b), (c), (f), (g)) are quite insensitive to the incident angle α, the situation for the orthogonal polarization is quite different. Mode S5 corresponding to the dipolar excitation along the short semi-axis of the nanoellipsoid (φ = 90°) experiences a substantial spectral shift with the change of the incidence angle in both incidence planes (Fig. 7(d), (f), (h), (i)) with angular sensitivities of 2.9 nm/deg (d), 4.5 nm/deg (e) for h = 50 nm and 1.5 nm/deg (h), 2.2 nm/deg (i) for h = 100 nm.

 

Fig. 7. Experimentally measured reflectance of the almost attached APN arrays (with the same parameters as in Fig. 6(d)) as a function of the incidence angle for four principal combinations of the plane of incidence (containing either long or short semi-axes of the nano-ellipsoid) and the incident light polarization.

Download Full Size | PDF

4.3 Refractive index sensing

The potential of the APN arrays for sensing of the refractive index (RI) of the environment was estimated using numerical simulations (Fig. 8). As the RI increases, modes A3, A4 and A6 experience a linear red shift with the sensitivities of 670 nm/RIU, 898 nm/RIU and 1335  nm/RIU, respectively. In addition, the modes have spatially different positions of electric field enhancement in the cavity, which makes it possible to independently detect RI changes in different areas at the same time. Particularly, the cavity space can be divided into three regions: 1, 2 and 3 (Fig. 8(b)). The evolution of the spectral positions and the width of the resonances corresponding to modes A3, A4 and A6 with the RI change in these regions is shown in Fig. 8(c)–(e). As the RI in region 1 increases, A3, A4 and A6 experience a linear red shift, with RISs of 383 nm/RIU, 500 nm/RIU and 916 nm/RIU, respectively. As the RI in region 2 increases, modes A4 and A6 experience a linear red shift, with RISs of 216 nm/RIU and 550 nm/RIU, while the spectral position of A3 remains constant. For the RI increase in region 3, mode A4 shows a RIS of 250 nm/RIU, while the spectral position of mode A6 remains constant and mode A3 disappears. The observed levels of sensitivity are typical for the sensors using individual localized resonances [33].

 

Fig. 8. Numerically simulated reflectance of the ANP-A arrays (with the same parameters as in Fig. 5(d)) at normal incidence and φ = 90° polarization as a function of (a) uniform change of the RI of the environment and (c – e) RI change in three nanocavity regions shown in (b).

Download Full Size | PDF

5. Conclusion

We present a low-cost self-assembly method based on nanosphere lithography for fabrication of large-area arrays of plasmonic nanocavities featuring an anisotropic optical response. The arrays support plethora of hybridized plasmonic modes related to the nanoellipsoid nanocavity core and the coaxial nanocavity itself, which due to the in-plane anisotropy of the structure lead to a pronounced polarization-sensitive behavior of the arrays. The modes can be individually tuned by changing the geometrical parameters of the nanoellipsoids at the fabrication stage. The optical response of the arrays is sensitive to the incident angle when the polarization direction is along the short semi-axis of the core nanoellipsoids. In addition, simulation results show that the arrays offer sensitivity to the refractive index of the surroundings as high as 1335 nm/RIU. Furthermore, it is possible to simultaneously and independently detect RI changes in various regions of the cavity by monitoring spectral shifts of the plasmonic modes. Such optical response makes asymmetric nanocavity arrays to be prospective building elements for integrated photonic devices targeting a variety of applications, including polarization-controlled optical switches, spatial light modulators and RI sensors.