1. Introduction

Polarizers are one of the most basic optical devices in optics. They were conceived by Hertz during the very early stages of electromagnetism [1]; at the time, wire-grid polarizers were introduced as the first polarizers.

In general, there are two classes of polarizers in terms of performance, which is quantitatively indicated by extinction ratio for two linear cross-polarizations. The first class exhibits relatively low performance; e.g., wire-grid polarizers and polymer-based film polarizers belong to this class. To obtain larger extinction ratios, thicker polarizers are required in this class. Usually, these polarizers are inexpensive and are used in situations that do not need extinction ratios above 5000. The second class is usually made of prisms, exhibiting high performance exceeding 10000; e.g., Glan-Thompson and Glan-laser polarizers belong to this class. These polarizers are based on a polarization-dependent coupling between two prisms, exhibiting the high-end extinction ratios in a wide range of wavelength.

The high-end prism-based polarizers have the disadvantages of being expensive and needing long optical paths of up to cm dimensions, as shown in Fig. 1(a). Figure 1(a) also shows an ultrathin polarizer of a net thickness of 245 nm and a large area more than 1 cm², which was fabricated on a quartz substrate. This ultrathin polarizer is the central issue of this paper.

 

Fig. 1 (a) Image of high-end polarizers: conventional prism polarizer (left) and ultrathin polarizer (right) used in this study. Both the polarizers offer large extinction ratio more than 10000. The rule indicates that the thickness of the prism polarizer is more than 1 cm. The ultrathin polarizer fabricated on a quartz substrate is composed of a semi-transparent metallic film of approximately 45 nm. (b) Schematic of the ultrathin polarizer in (a), forming a SC structure in terms of metal (gray). Pale blue denotes transparent materials such as quartz. The x and y coordinate axes are set as indicated.

Download Full Size | PDF

Figure 1(b) illustrates the structure of the ultrathin polarizer, appearing in Fig. 1(a), and coordinate axes. It is composed of three layers located on a transparent substrate (pale blue). The top layer is a perforated metallic film (gray) and has II-shape air holes. The middle layer is made of a transparent material (pale blue) and has the II-shape air holes. The bottom layer consists of the metal and transparent material; the metallic part forms the array of the II shapes and the transparent part has the same shape as the middle layer. Focusing on the metallic structures, they form a stacked complementary (SC) structure, which was found, by numerical studies, to be a high-efficiency polarizer several years ago [2].

SC structures have been found to be a unique series in various types of plasmonic resonators [3]. The SC structures combine two heteroplasmons with almost identical resonant frequencies via the middle layer with appropriate thickness and form new coupled resonant states. The heteroplasmons originate from the structural complementarity as Babinet principle tells [4]. In anisotropic cases, the SC structures are able to enhance the optical anisotropic properties; the II-shape SC structure in Fig. 1 is one of the actual examples. In addition to the highly efficient designs for ultrathin polarizers [2,5,6], plasmon–photonic-guided-mode hybrid resonances were found in SC structures fabricated in waveguide silicon-on-insulator (SOI) substrates [7]. The SOI-based SC structures serve as high-emittance artificial surfaces, enabling highly enhanced fluorescence and Raman scattering [8–11].

One of the most extensive developments in photonics is being realized in integrated photonics that include ultrahigh speed signal processing of up to 50 Gbps [12]. It is currently based on Si photonics. However, more dense integration requires new elements other than Si because the minimum size in Si photonics is limited to the micrometers dimension (typically, tens of micrometers) due to the use of guided modes. One of the potential candidates to reduce the limit is metamaterials that include metallic nanostructures; deep subwavelength modes will enable the reduction of the size to submicrometers for optical waves. Indeed, designs of subwavelength basic optical devices such as polarizers and wave plates have been reported employing high-transmission and fairly low-loss metamaterials [13]. The thickness of the metamaterial-based optical devices is thinner than working wavelengths (in air); therefore, they can be called artificial surface structures on bulk substrates and are referred to as metasurfaces hereon. Metasurfaces are often single-layer structures on substrates and have complex unit cells [14,15], whereas here we address metasurfaces of three-layer structures, which have quite simple unit cells.

In this study, we report on the fabrication and experimental demonstration of ultrathin SC metasurface polarizers. Numerical examinations will also clarify the optimum performance exploring the structural parameters.

2. Methods

2.1. Nanoimprint lithography (NIL)

Figure 2(a) shows the design of a unit cell of SC metasurfaces, obtained in the previous numerical results [2]. The unit cell is a square of periodicity a. The scaling factors noted in the unit cell denote the dimensions of the II-shape structure; for example, 0.6 indicates that the length is 0.6a. The periodic arrays of the unit cells form the metasurfaces.

 

Fig. 2 (a) Design of the unit cell of the SC metasurfaces. The periodicity a is the same in the x and y directions. The scaling factors inside the unit cell indicate the dimensions of the II-shape structure as fractions of the periodicity a. (b) and (c) Top-view SEM images of Ag- and Au-deposited SC metasurfaces of a = 900 nm. White scale bars indicate 1 µm.

Download Full Size | PDF

Figures 2(b) and 2(c) show the top-view scanning electron microscopy (SEM) images of the SC metasurfaces, which comprised an array of II-shape. The deposited metals in Figs. 2(b) and 2(c) were Ag and Au, respectively. The thickness was set to 45 nm. The metasurfaces were produced using NIL with a large area of 1.2 × 1.2 cm², as shown in Fig. 1(a).

Figure 3 succinctly illustrates a section view of the NIL procedure conducted. UV light was illuminated on a quartz mold that was pressed onto a resist-coated quartz substrate, as illustrated in Fig. 3(a). The details of this UV NIL were reported in depth previously [8,9]. After removing the mold and treating the residual thin film at the bottom of the patterned resist, dry etching of the substrate was conducted to the designed depth, as shown in Fig. 3(b). In the present case, we set the depth to 200 nm. Furthermore, the remaining resist was removed, and metal was deposited at an angle that was nominally normal to the substrate, as shown in Fig. 3(c). The metals used in this study (Ag and Au) were deposited using an electron-beam evaporation instrument (RDEB-1206K, R-DEC Co. Ltd., Japan). Thicknesses of the metals were set to 45 nm. The SEM images in Figs. 2(b) and 2(c) were taken after the metal deposition.

 

Fig. 3 Nanoimprint lithography: a simplified procedure. (a) Nanoimprinting step. UV light was illuminated to make the resist solid. (b) Dry etching of quartz substrate with the patterned resist mask. After the etching, the left resist mask was removed. (c) Metal deposition, which completes the fabrication of SC structure.

Download Full Size | PDF

2.2. Optical measurement

Optical characterizations were conducted using a spectrometer (V-7200, JASCO, Japan) and a hand-made measurement setup for large extinction ratios. Ordinary transmittance (T) spectra were measured using the spectrometer; however, the measurement of large extinction ratio is out of reach for the spectrometer. We therefore prepared the setup for the large extinction ratio.

In the setup for the large extinction ratio, incidence comes from a wavelength-tunable pulsed laser (7 ns and 20 Hz) because large intensity contrast exceeding 10000 is required. The laser light was collimated and highly polarized using a Glan-laser prism, which kept the polarization degree to exceed 30000. The incidence was normal to the samples and shed on the top layer. Transmitted light was detected by a photodiode detector and simultaneously the incident light intensity was monitored by another photodiode detector. When measuring cross polarization, the samples were rotated by 90^◦. To maintain a linear response of the photodetectors, appropriate neutral density filters, which were calibrated by the spectrometer, were inserted in the light path. Thus, the extinction ratio was obtained as the ratio of two calibrated transmitted signals. We note that the ratio was evaluated in this setup and that each polarized T was not evaluated.

2.3. Numerical details

Numerical calculations for the optical spectra of the SC metasurfaces were implemented using the rigorous coupled-wave analysis method [16] that incorporated a scattering-matrix algorithm [17]. This method analyzes Fourier-transformed Maxwell equations, being a standard to evaluate the optical spectra of periodic photonic structures in the frequency domain. In the computations, relative permittivity is used as a material parameter of constituent materials. We simply refer to relative permittivity as permittivity from now on. The numerical implementations were conducted on supercomputers via multi-parallel implementation.

We aim at quantitative reproduction and evaluation of the ultrathin polarizers. Therefore, the measured permittivity of the metals was compared with permittivity taken from the literature [18].

Figures 4(a) and 4(b) show the complex permittivities of Ag and Au, respectively. Red curves show the permittivities from the literature, which were obtained by fitting the measured data with the Brendel-Bormann (BB) model, and showed better reproduction than the Lorentz-Drude model [18]. Black curves denote measured permittivities. The real parts of the permittivities are displayed with solid curves and the imaginary parts with dashed and dotted curves. Black curves represent measured permittivities using an ellipsometer (M-2000, J. A. Woollam Co., Inc., USA); the wavelength range was from 1688 to 248 nm. The permittivities were evaluated using thin Ag and Au films of 40 nm thickness, which were deposited on flat quartz substrates of 1 mm thickness. Surface roughness of the thin films is taken into account in the analysis procedure; the roughnesses are estimated to be less than 1.7 nm.

 

Fig. 4 Complex permittivity of (a) Ag and (b) Au. Measured permittivities are shown with black curves and permittivity from the literature [18] is represented with red curves. In both panels, the real parts are shown with solid curves.

Download Full Size | PDF

Comparing the measured permittivity with that from the literature, it is obvious that the measured values are smaller in the real part. The imaginary part of the measured permittivity is almost similar but slightly smaller than that from the literature. These results indicate that Ag and Au in these measurements have slightly lower optical losses than the noble metals analyzed in the literature. To realize better reproduction, we mainly use the measured permittivities in the following numerical calculations. Note that permittivity is generally dependent on preparation method of materials.

3. Results and Discussions

In this section, we first show the experimental results (Fig. 5) and then present the corresponding numerical results (Figs. 6 and 7). Furthermore, we examine and discuss the results and note the implications of this study.

 

Fig. 5 Measured optical spectra represented with wavelengths in nm. (a)–(c) Optical spectra of the SC metasurfaces made of Ag. (a) and (b) Polarized T spectra of periodicity a = 900 and 1000 nm, respectively. T_x’s exhibit large transmittance. The incidence was set to be normal. (c) Extinction ratio, defined by T_x/T_y. Blue and red lines with dots and open circles, respectively, show the measured ratios of a = 900 and 1000 nm. Green solid and black dashed lines are shown as guides to the eye, representing the interference-suppressed spectral shapes and are described in the text. (d)–(f) Optical spectra of the SC metasurfaces made of Au, which are displayed in a similar manner to (a)–(c), respectively.

Download Full Size | PDF

 

Fig. 6 Numerically calculated T and extinction-ratio spectra. Periodicity was set to 900 nm. Thickness of metals was fixed at 45 nm. (a) Measured permittivity of Ag (mAg) was used in the computation. (b) BB-model permittivity of Ag (BB Ag) was used. (c) Measured permittivity of Au (mAu) was used. (d) BB-model permittivity of Au (BB Au) was used. Polarized T_x’s are shown with blue solid lines and T_y’s with red dashed lines. The T spectra are plotted on the logarithmic scale. Extinction ratios are shown with black lines, plotted for the right axes.

Download Full Size | PDF

 

Fig. 7 Numerical examination of extinction ratios under various conditions. (a) and (b) Dependence on Ag and Au thickness, respectively. Dashed lines indicate 50 nm thickness, black solid lines 45 nm, and gray solid lines 40 nm. Periodicity was fixed at 900 nm. We note that the cases of 45 nm are in common with those in Figs. 6(a) and 6(b). (c) Dependence on middle-layer thickness M_t in Fig. 3(c), which was varied from 55 to 205 nm. The metal was assigned to be Ag. Periodicity and the Ag thickness were set to 900 and 45 nm, respectively. (d) Dependence on periodicity, which was varied from 500 to 1000 nm. These numerical calculations were implemented using measured permittivities of Ag and Au.

Download Full Size | PDF

3.1. Experimental results

Figure 5 shows the measured optical spectra of the SC metasurfaces fabricated by NIL. Figures 5(a) and 5(b) present the polarized T spectra of the SC metasurfaces of periodicity a = 900 and 1000 nm, respectively; x-polarized T (T_x) shown with blue and red curves is quite high and reaches 80% at the maxima whereas y-polarized T (T_y) shown with black lines is efficiently terminated. The T_y spectra are mostly less than 0.1% and under the detection limit of the spectrometer. Note that the coordinate was set in Fig. 1(b). The large T_x of the SC metasurfaces originates from a local plasmon associated with efficient local Poynting flux to the transmission direction; the local rotational flux is induced due to the locally enhanced ∇ × H [2]. Generally, when T_x at the top layer takes large values, the corresponding x-polarized refractance R_x at the top layer takes small values. Babinet’s principle tells that the R_x at the top layer is almost equal to T_y at the bottom layer [4]. In addition, T_y at the top layer takes small values because there is no resonance at the wavelength range in Fig. 5 to assist efficient transmission. In total, the SC metasurfaces exhibit small T_y, yielding large extinction ratios; a similar description was reported previously [5].

Figure 5(c) shows the extinction ratios measured using the wavelength-tunable laser, as noted in section 2.2. The blue line with closed circles shows the measured extinction ratio of the Ag SC metasurface of periodicity a = 900 nm and red line with open circles denotes that of a = 1000 nm. The extinction-ratio spectra have spiky structures that presumably originate from interference in the optical component(s); therefore, we plot smooth guides for the eye, shown by green solid and black dashed curves. The guides take Lorentzian shapes, fitting the measured data around the peaks. We refer to the spectral shape later (Fig. 6). The maxima of the extinction ratios reach 17000; moreover, the widths of the extinction-ratio peaks at a ratio of 10000 are more than 100 nm. It is thus evident that the SC metasurfaces are high-performance polarizers. We stress that the thickness is only 245 nm, being thinner than the 1/5 wavelength of the incident light.

Figures 5(d)–5(f) show measured T spectra and extinction ratios of the SC metasurfaces using Au, displayed similarly to Figs. 5(a)–5(c), respectively. Qualitatively, the spectral features are similar to those of the SC metasurfaces of Ag. The maxima of the extinction ratios reach 12000. Quantitatively, Au SC metasurfaces exhibit smaller T and extinction ratios whereas Au is chemically stable and probably has an advantage in extended usage.

3.2. Numerical examinations

Figure 6 shows numerically calculated T spectra and extinction ratios. These are shown to clarify how permittivity affects the numerical results. To keep the contents clear, we focus on the case of periodicity a = 900 nm and a metal thickness of 45 nm. The four panels present the numerical results in a similar way. Blue solid and red dashed curves represent T_x and T_y spectra, respectively, plotted on a logarithmic scale. Black solid curves denote extinction ratios, defined by T_x/T_y, which have the scales on the right axes.

Figures 6(a) and 6(b) are numerical results using measured and BB-model permittivity of Ag, respectively [see Fig. 4(a)]. Qualitatively, the results are similar to each other. However, there is a quantitative difference: the values of the extinction ratios. In Fig. 6(a), the maximum of the extinction ratio exceeds 20000, while, in Fig. 6(b), the maximum is at most 10000.

Figures 6(c) and 6(d) are based on measured and BB-model permittivity of Au, respectively [see Fig. 4(b)]. In the Au case, a quantitative discrepancy of extinction ratio appears that is similar to the Ag case; in Fig. 6(c), the peak reaches 20000, while, in Fig. 6(d), it is less than 4000. Thus, it is probable that the measured permittivity of the metals results in a better reproduction of the measured extinction ratios. Still, the measured extinction ratio and numerical one have some quantitative differences, which are examined in the next figure by considering other structural factors that are possibly different from the initial design.

We here refer to the spectral shape of the extinction ratios. It was assumed to be Lorentzian in Figs. 5(c) and 5(f). The reason for the derivation is described as follows. Figure 6 clearly shows that the spectral shape of extinction ratio is primarily determined by the shape of T_y which is resonantly terminated. As noted in section 3.1, the SC structures resonantly enhance and terminate transmission, depending on the polarizations. The high-T band is rather broad; as a result, the low-T resonance mainly contributes to the shape of the extinction ratio. Consequently, the Lorentzian spectral shapes are considered to appear.

Figure 7 shows numerical examinations of extinction ratios under various conditions. Figures 7(a) and 7((b) examine effects due to the metal thickness, which was varied from 50 to 40 nm in steps of 5 nm. The extinction ratio is quite sensitive to the metal thickness. As the thickness decreases, the extinction ratio rapidly decreases. Furthermore, the peak wavelength prominently shifts to longer wavelengths; indeed, as the metal thickness is reduced by 10 nm, the peak wavelength moves by approximately 100 nm. These results suggest that the fabricated SC metasurfaces have a thinner metal thickness than our design; in the case of Ag, the peak of the measured extinction ratio of 17000 is located between the peaks of the 45 and 40 nm metal thickness in Fig. 7(a) and, in the case of Au, the peak of the extinction ratio of 12000 can be reproduced by the metal thickness between 45 and 40 nm.

Figure 7(c) presents a set of extinction ratios of SC metasurfaces that have different thicknesses of the middle layer, M_t, which is indicated in Fig. 3(c). The thickness M_t was changed from 55 to 205 nm at a step fo 25 nm; each result is shown with a colored dashed or solid line. Other structural parameters were fixed such that the periodicity a = 900 nm and the metal thickness was 45 nm. It was confirmed that the maximum of the extinction ratio is realized by the middle-layer thickness between 130 and 155 nm and that the experimental result in Fig. 5 is almost at the optimum in terms of the middle-layer thickness.

Figure 7(d) shows extinction-ratio spectra dependent on periodicity, which was varied from 1000 to 500 nm; accordingly, the curves vary from black to green. The metal was assumed to be Ag and the thickness was set to 45 nm. The middle layer thickness M_t was fixed at 155 nm. These extinction-ratio spectra show that large extinction ratios of more than 10000 are realized for the telecommunication wavelength band from 1300 to 1600 nm, by simply changing the periodicity of the SC metasurfaces that have the unit cell in Fig. 2(a). Besides, Fig. 7(d) implies that the design of the II-shape unit cell is unlikely to be very good in the visible wavelengths, and one should explore other designs to obtain high-end ultrathin polarizers in the visible.

3.3. Comparison with other polarization-controlling devices

As we have shown, the SC metasurfaces include thin metal films of only 45 nm thickness. Still, the extinction ratio is larger than 10000. When one realizes such a large extinction ratio with wire-grid polarizers, the thickness becomes more than a micron for the visible and near-infrared, which is quite thick. In addition, such thicker wire-grid polarizers tend to exhibit lower T and are technically hard to be fabricated. Thus, when we think of polarizers of subwavelength thickness, wire-grid polarizers cannot be the solution.

SC wire-grid polarizers may be expected to show better performance than the usual single-layer wire-grid polarizers. This idea was tested so far [19]; as a result, the extinction ratio was at most 400 in the visible and near-infrared ranges. The performance was similar or less than the usual wire-grid polarizers.

Polarization-controlling metasurfaces are one of the issues in metasurface research. However, with single layer metasurfaces it is generally hard to manipulate polarization vectors in a highly efficient manner; the reason is in common with the wire-grid polarizers. In this sense, the ultrathin polarizers are quite unique.

In contrast, wave plates comprising single-layer metasurfaces are often reported [20–23]. Wave plates manipulate the phase of light waves, due to anisotropic effective refractive indices [20] and also by 2D modes inducing anisotropic polarization transmission, reflection, and scattering [21–23]. The selection of the wave plates will depend on the purpose and requirements (i.e., working wavelengths, feasibility in fabrication, and transmission and reflection efficiencies).

4. Conclusions

We have experimentally demonstrated ultrathin high-performance polarizers with extinction ratios exceeding 12000. They have SC metasurfaces, fabricated via a quite simple procedure including a single metal deposition. The design was numerically shown to be highly suitable for the telecommunication wavelengths of 1.3–1.6 µm. The present SC metasurface polarizers will be helpful to realize densely integrated, ultracompact photonic circuits working at definite telecommunication wavelengths.