1. Introduction

Electromagnetic metamaterials consisting of subwavelength structures have exotic optical responses such as electromagnetic cloaking [1], negative and even zero refractive indices [2,3], and perfect lenses [4], which are not available in nature [5]. These exotic optical responses are realized by the independent control of the permittivity and permeability of metamaterials. The control of the materials’ parameters such as permeability was the main goal during the early study of metamaterials. In particular, extensive efforts were devoted to decreasing or compensating the losses of parameters associated with the resonance of subwavelenth structures [6,7]. On the basis of these studies, new optics called flat optics were proposed [8]. In flat optics, electromagnetic waves can be tailored in a subwavelength region called a metasurface. Metasurfaces typically consist of an array of subwavelength antennae which control the electromagnetic wavefront. To describe the electrodynamics in metasurfaces, the law of refraction is generalized to include the phase gradient of the surface, and light refraction can be described and realized by a more general scheme [9,10],within which the basic control of light such as focusing by lenses has been demonstrated in the subwavelength region [11,12].

Control of the polarization state of light by metasurfaces is one of the key issues in flat optics [13–18]. Polarizers can be also replaced by metasurfaces. Among various kinds of metasuface polarizers, stacked complementary structures made of a metallic material have a unique working principle [19]. The top metallic structure is designed to have a plasmonic resonance with high transmittance for horizontal polarization. Following Babinet’s principle [20,21], the bottom complementary structure is off-resonance, with low reflectance (high transmittance) for the same polarization. The top structure is designed to be off-resonance with low transmittance for vertical polarization, and the bottom structure has resonance with high reflectance (low transmittance) for the same polarization. When the top and bottom structures are stacked, the overall optical response is given by the product of each optical response, resulting in very high transmittance for the horizontal polarization and very low transmittance for the vertical polarization. Thus, the metasurface serves as a high-performance polarizer. In this scenario, extinction ratios exceeding $10^4$ have been demonstrated numerically and experimentally [22–26]. Moreover, a numerical study revealed that the extinction ratios would be much enhanced by combining Babinet’s principle and Fano resonance between propagating and local surface plasmon polaritons (SPPs). Because of an extremely low transmittance induced by Fano resonance, ultrahigh extinction ratios are possible in the C-band (extinction ratio $>3\times 10^9$) and the O-band (extinction ratio $>200\times 10^9$) of telecommunication wavelengths [27].

However, the performance of the metasurface polarizer incorporating Babinet’s principle diminishes drastically with the increase in metallic losses. Hence, an ultra-high extinction ratio has been demonstrated numerically only in the telecommunication wavelengths where the metallic losses of noble metals are not very high. The demonstration of a high extinction ratio in the optical region is a challenge.

Here, we describe a high-performance metasurface polarizer in the deep ultraviolet (DUV) region (from 200 to 320 nm), where wire grid polarizes (WGPs) were extensively studied in the previous studies [28,29]. The working principle of the WGPs in the DUV region is basically material absorptance of semiconductors. The design of the WGPs strongly depends on the type of materials and the performance of the WGPs made of a metallic material such as Al remains low in the DUV region. In this work, we show that it is possible to realize high-performance polarizers comprised of a conventional plasmonic material in the DUV region such as Al. The working principle of the polarizers depends solely on an interference controlled by the structural parameters and conventional plasmonic materials can be utilized as a constituent material for the polarizers, leading to a high degree of freedom in design. We show the design of a reflection-type metasurface polarizer with extinction ratios of $\ge 6.2\times 10^6$. The physics behind the high performance is discussed. On the basis of calculations, a metasurface polarizer was prepared and the signature of the increase in extinction ratio was demonstrated experimentally.

2. Metasurface polarizer in the deep ultraviolet region

Figure 1(a) depicts a schematic of a metasurface polarizer. The metasurface consists of three layers (Fig. 1(b)). The top layer is a rectangular metallic grating of Al (aluminum) with thickness $t_g$. The middle dielectric layer consists of a thin $\mathrm {Al_2O_3}$ (alumina) film with thickness $t_d$. The bottom layer consists of a thin Al film with thickness $t_m$. The period and width of the groove of the grating are $P_x$ and $w=P_x/2$, respectively. The metasurface polarizer sits on a quartz substrate. This kind of metal grating/insulator/metal (MIM) structure is very well known but we show that such an MIM structure serves as a high-performance polarizer in the DUV region. To calculate the optical responses of the polarizers, we used rigorous coupled wave analysis [30] with Li’s inverse algorithm [31,32]. The calculation used 161 reciprocal lattice vectors. The refractive indices of $\mathrm {Al_2O_3}$ [33] and quartz [34] were obtained from the literature. The permittivity of Al is described by the Drude-Lorentz model and was obtained from the literature [35].

 

Fig. 1. (a) Bird’s eye view and (b) cross section view of the metasurface polarizer.

Download Full Size | PDF

First, we focused on the optical response of the polarizer at normal incidence. In this configuration, a half mirror is required to separate the polarization states of light (Fig. 2(a)). With structural parameters set to $P_x = 150$ nm, $t_g = 20$ nm, $t_d = 15$ nm, and $t_m = 25$ nm, we calculated the optical spectra at normal incidence. There is a very sharp dip in the reflectance spectrum ($R_x$) around 259 nm (red curve): the minimum value is extremely low at about $1.14\times 10^{-7}$ (Fig. 2(b)). This value creates a sharp dip within a broader dip. The polarizer has high absorptance at near 1 around the position of the sharp dip, but the peak width is much broader than that of the reflectance, indicating that the sharp dip originates not just from absorption but also from another factor. Figure 2(c) shows the optical spectra for $y$ polarization. There are no distinct features. The reflectance ($R_y$) is high in the DUV region, indicating that the metasurface serves as a mirror with high reflectance over the whole spectral range when the polarization of the incident light is parallel to the grooves. Figure 2(d) shows the extinction ratio spectrum calculated as $R_y/R_x$. There is a sharp peak around 259 nm, which corresponds to the sharp dip in the reflectance. The maximum value of the extinction ratio is about $6.2\times 10^6$, comparable with those of prism-based polarizers. These calculation results show that the metasurface polarizer has an extremely low reflectance for $x$-polarized light and a high reflectance for $y$-polarized light, resulting in a much enhanced extinction ratio of $\ge 6.2 \times 10^6$ in the DUV region. Around the wavelengths where the extinction ratio is much enhanced, the spectral shape has an asymmetry (inset in Fig. 2(d)). On the other hand, the enhanced extinction ratio around 620 nm has a typical Lorentzian shape. This comparison indicates that the spectral shape of the much enhanced extinction ratio is asymmetric. In the following section, we consider the dispersion relation of the eigenmodes in the metasurface and reveal the physics behind the high extinction ratio.

 

Fig. 2. (a) Schematic of a metasurface polarizer combined with a half mirror working at normal incidence. (b, c) Optical spectra for (b) $x$ and (c) $y$ polarizations. (d) Extinction ratio spectrum. The inset shows the enlarged spectrum around 260 nm.

Download Full Size | PDF

3. Discussion

To consider the dispersion relation in the metasurface, we calculated the angle-resolved reflectance spectra. Figure 3(a) shows a pseudo color plot of the angle-resolved reflectance spectra for the p-polarization excitation. Note that we used the notations of $x/y$-polarization and p/s-polarization to refer the polarization state of incident light at normal incidence and at oblique incidence, respectively. The vertical axis plots the incident photon energy and the horizontal axis plots the incident wavevector parallel to the periodic direction ($k_\parallel$) normalized to the reciprocal lattice vector $G=2\pi /P_x$. The spectra show three main features. The first is the mode with a cross point around 4.8 eV ($\sim$259 nm) at the $\Gamma$ point (normal incidence). This mode is assigned to the propagating SPPs at the metal-substrate interface, and is associated with the first order reciprocal lattice vector. The second is the mode with a broad line width around the cross point of the propagating SPPs (4.8 eV). This mode has a relatively flat dispersion curve compared with that of the propagating SPPs. The third is the mode with a broad line width around 2 eV ($\sim$620 nm). It has a very flat dispersion curve, similar to the second. The remaining linear dispersion curve in the higher energy region represents the opening of the first order diffraction in the air. The dispersion curves and the pseudo color show that the reflectance becomes very low near where two modes cross, indicating that the extremely low reflectance is induced by the destructive interference between the two modes.

 

Fig. 3. Angle-resolved reflectance spectra for (a) p- and (b) s-polarizations. The horizontal axis shows the wavevetor parallel to the periodic direction, normalized to the reciprocal lattice vector $G=2\pi /P_x$. The color bar indicates the reflectance intensity. The white dashed lines are drawn for eye guide.

Download Full Size | PDF

Figure 3(b) shows the angle-resolved reflectance spectra for the s-polarization excitation. The metasurface has high reflectance for s-polarization and a featureless optical response compared with the p-polarization excitation. Only the linear dispersion curve representing the opening of the first order diffraction in the air can be found in the spectra. As we have shown, the metasurface polarizer has an extremely low reflectance for $x$-polarized light and a high reflectance for $y$-polarized light. The extremely low reflectance is caused by the destructive interference between two modes, known as Fano resonance. The metasurface has a high reflectance under s-polarized excitation because it has no mode in the non-diffraction regime and serves as a simple mirror. As a result, the metasurface has a high performance with an extinction ratio exceeding $6.2\times 10^6$.

To reveal the eigenmodes causing the ultra-high extinction ratio, we calculated the near field distribution patterns in the metasurface polarizer. Figure 4(a) shows the reflectance spectrum at $45^\circ$ incidence. Figure 4(b) shows a snap shot of the magnetic field distribution at 589.75 nm, where the reflectance spectrum has a very broad dip indicated by the black arrow (1) in Fig. 4(a). The magnetic fields are concentrated only in the gap region, indicating that the gap SPPs are excited. Figure 4(c) shows a snap shot of the magnetic field distribution at 281.05 nm, where the reflectance spectrum has a very broad dip indicated by the black arrow (2) in Fig. 4(a). The field distribution pattern is similar to that of Fig. 4(b), but the number of antinodes in the gap region is different. The modes in Fig. 4(b) and (c) are classified into the gap SPPs of the first and second order, respectively. Figures 4(d) and (e) show the magnetic field distribution at 352 nm and 329.65 nm, where the reflectance spectrum has the dips indicated by the black arrows (3) and (4) in Fig. 4(a). In Fig. 4(d, e), the magnetic fields are concentrated in the $\mathrm {Al_2O_3}$ region. This field distribution pattern is a feature of the gap SPPs. The magnetic field is also concentrated at the interface between the Al film and the quartz substrate. This is a feature of the propagating SPPs. These two features define that these modes as hybridized modes of the gap and propagating SPPs, which split into two in a manner that depends on the energy level of each mode. Analysis of the electric field distributions indicates that two different types of SPPs are involved in the high performance. These two types can interfere with each other, resulting in Fano resonance [36,37]. The asymmetry in the extinction spectra in the inset of Fig. 2(d) can be attributed to Fano interference. The analysis also indicates that the extinction ratio remains low when only the gap SPPs are excited. On the basis of the dispersion curve and field distribution analyses, we attribute the extremely high extinction ratio to Fano resonance between the propagating and gap SPPs.

 

Fig. 4. (a) Calculated optical spectra at $45^\circ$. (b-e) Snap shots of the magnetic field distributions at (b) 598.75 nm, (c) 281.05 nm, (d) 352 nm, and (e) 329.65 nm. The color bar indicates the value of the magnetic field.

Download Full Size | PDF

Following the analysis of the eigenmodes and dispersion relations, we consider the influence of structural parameters on the optical responses and discuss the design guidelines of the metasurface polarizer. We varied three structural parameters of the thickness of the $\mathrm {Al}_2\mathrm {O}_3$ layer ($t_d$), period ($P_x$), and height of the grating ($t_g$). When sweeping one parameter, we fixed the other parameters to the original values. As for the optical responses, we focused on $R_x, R_y$, and extinction ratio defined by $R_y/R_x$. In addition to the discussion about the design guidelines, we show the reason why we chose the parameters in the first place of this paper through the sweeps.

Figure 5(a, b, c) shows the pseudo color plot of the optical spectra as a function of $t_d$. In $R_y$ (Fig. 5(b)), there are three radial straight lines, which correspond to the excitation of Fabry-Pérot resonance in the $\mathrm {Al}_2\mathrm {O}_3$ layer. There are two curves with sharp linewidths in the DUV region. These modes have a feature of cutoff, indicating that the sharp curves originate from the excitation of waveguided modes in the $\mathrm {Al}_2\mathrm {O}_3$ layer. As well as $R_y$, there are radial lines in $R_x$, which correspond to the excitation of Fabry-Pérot resonance but the lines are strongly deformed by the interaction with other modes. There is a line indicated by a red arrow, which does not shift its position with the variation of the thickness of the $\mathrm {Al}_2\mathrm {O}_3$ layer. This mode corresponds to the propagating SPPs. The other modes are assigned to be the local SPPs. Near the cross point of the two different types of SPPs (for example, around 260 nm and $t_d \sim 20$ nm), $R_x$ becomes extremely low. This low reflectance is caused by Fano resonance between the two SPPs. Following the significant decrease in $R_x$, the extinction ratio becomes extremely high. The metasurface polarizer should work as a mirror with high reflectance for $y$-polarization in a broad spectral range and have an extremely low reflectane for $x$-polarization. Therefore, $t_d$ should be thinner than 20 nm.

 

Fig. 5. Pseudo color plots of $R_x, R_y$, and extinction ratio spectra as a function of the thickness of $\mathrm {Al}_2\mathrm {O}_3$ layer (a, b, c), period (d, e, f), height of the grating (g, h, i). A white dashed line in each figure indicates the position of the original value of the swept parameter.

Download Full Size | PDF

Figure 5(d, e, f) shows the pseudo color plot of the optical spectra as a function of $P_x$. In $R_x$ (Fig. 5(d)), there are two broad lines with low reflectance. These two modes originates from the local SPPs. Moreover, there is a sharp curve in the ultraviolet region (wavelength $\leq 400$ nm). With the increase in the period, the resonance wavelength of the sharp curve shifts towards the longer wavelength region. This mode originates from the propagating SPPs. Near the cross point of the two SPPs, $R_x$ becomes extremely low due to the destructive interference. On the other hand, $R_y$ has no characteristics. The metasurface works as a mirror with high reflectance for $y$-polarization, which does not depend on the period. Due to the featureless optical response for $y$-polarization, the variation in the extinction ratio solely depends on $R_x$. As a result, the extinction ratios in the DUV region become extremely high due to the decrease in $R_x$ caused by the destructive interference. These calculation results indicate that the period of the metasurface should be adjusted so that the propagating SPPs are excited near the desired wavelength region.

Figure 5(g, h, i) shows the pseudo color plot of the optical spectra as a function of $t_g$. Around 260 nm in $R_x$, there is a line independent of the height of the grating, which is indicated by a red arrow. This mode is assigned to be the propagating SPPs. There are several lines with broad widths which depend on the height of the grating. These modes are assigned to be local SPPs. Near the cross points of the two different types of SPPs, $R_x$ becomes extremely low due to the destructive interference between the two SPPs. Such an interference is possible even when the height of the grating is only 20 nm. On the other hand, there is no characteristics in $R_y$ since the grating does not serve as a coupling layer to $y$-polarized light. Due to the featureless optical response for $y$-polarization, the variation of the extinction ratio solely depends on $R_x$. The significant decrease in $R_x$ results in the gigantic enhancement of the extinction ratio. These calculation results indicate that the height of the grating should be adjusted to lower $R_x$ by the destructive interference.

From the study on the variations of the structural parameters, the design guidelines of high performance metasurface polarizers are summarized as follows. The thickness of the $\mathrm {Al}_2\mathrm {O}_3$ layer should be thin so that no mode is excited for $y$-polarization but should be thick so that local SPPs are excited in the gap region between the grating and bottom Al layer for $x$-polarization. The period of the metasurface polarizer should be adjusted so that the propagating SPPs are excited in a desired wavelength region for $x$-polarization. The height of the grating should be adjusted to extremely lower $R_x$ by the destructive interference between the propagating and local SPPs. In addition to the aforementioned parameters, we refer to the thickness of the bottom Al layer. Too thick Al layer prevents an efficient excitation of propagating SPPs at the Al/Quartz interface. Therefore, the thickness of the bottom Al layer should be comparable with the SPP penetration depth of Al (about 15 nm in the DUV region). On the basis of the design guidelines, we optimized the structural parameters to be $t_d =$ 15 nm, $P_x=$ 150 nm, $t_g=$ 20 nm, and $t_m=$ 25 nm. The total thickness of the polarizer (60 nm) is sufficiently thin compared to the incident wavelength ($\sim 260$ nm) and an extremely high extinction ratio is realized in a subwavelength region, which allows us to refer the metal/insulator/metal structure as a metasurface.

We have designed a metasurface polarizer with an extinction ratio exceeding $6.2\times 10^6$ in the DUV region. The calculated ratio is comparable to those of existing prism-based polarizers. We also showed the design guidelines of the polarizer. However, it is designed to operate at normal incidence. In this case, a half mirror is required to separate the $x-$ and $y-$polarized light (Fig. 2(a)). After passing through the half mirror twice, the light intensity decreases to one-fourth of original. Moreover, the bandwidth of the extinction ratio is extremely narrow: that of an extinction ratio of $\ge 100$ is about 13 nm. These characteristics are not preferable for practical applications. To overcome these drawbacks, we designed a DUV metasurface polarizer with a much wider bandwidth working at an incident angle of 45$^\circ$. In this configuration, there is no need to use a half mirror to separate the polarization state (Fig. 6(a)). Setting $P_x = 125$ nm, $t_g = 20$ nm, $t_d = 18$ nm, and $t_m = 25$ nm, we calculated the optical spectra at an incident angle of 45$^\circ$.

 

Fig. 6. (a) Schematic of a metasurface polarizer working at $45^\circ$ incidence. (b, c) Optical spectra for (b) $x$ and (c) $y$ polarizations. (d) Extinction ratio spectrum.

Download Full Size | PDF

The colored curves in Fig. 6(b) show the reflectance ($R_p$), transmittance, and absorptance spectra for the p-polarization. As well as the reflectance spectrum in Fig. 4(a), the reflectance spectrum in Fig. 6(b) has four dips. The lowest and highest energy dips originate from the gap SPPs of the first and second orders, respectively. The split modes around 270 and 305 nm are the hybridized modes of the gap and propagating SPPs. Owing to the interference between the two modes, the reflectance becomes very low around 300 nm. Figure 6(c) shows the optical spectra for s-polarization. As at normal incidence, the metasurface polarizer has a featureless spectrum with high reflectance for s-polarizerd light ($R_s$), indicating that the polarizer works as a mirror with high reflectance for s-polarization. Figure 6(d) shows the extinction spectrum. The maximum values of the higher and lower energy peaks around 300 nm are $2.78\times 10^4$ and $1.29\times 10^4$, respectively. The basic optical response is the same as that at normal incidence and the high extinction ratio originates from Fano resonance between the propagating and gap SPPs. Even though the maximum value of the extinction ratio is lower than that at normal incidence, an extinction ratio around $10^4$ is enough for practical applications. Moreover, the bandwidth of an extinction ratio of $\ge 100$ is improved to about 31 nm, 2.4 times that at normal incidence.

4. Experiment

We prepared a metasurface polarizer working at 45$^\circ$ incidence in the DUV region. We prepared the bottom 25-nm-thick Al layer by radiofrequency magnetron sputtering (NMS-2000, ULVAC, Japan) and the middle 18-nm-thick Al$_2$O$_3$ layer by atomic layer deposition (R-200 Advanced, Picosun, Finland). Electron-beam resist (ZEP520A, ZEON corporation, Japan) spin-coated on the Al$_2$O$_3$ layer was patterned by electron beam lithography (ELS7500-EX, ELIONIX, Japan) and subsequent development processes. Finally, the top Al grating layer was prepared by electron beam evaporation (EB-350T, EIKO engineering, Japan) and a subsequent lift-off process. Figure 7 shows the optical microscopy (a) and scanning microscopy (b, c) (S-5000, HITACHI, Japan) images of the metasurface. The edge of the grating was rounded and the cross-section profile of the grating differed from that of lamellar grating. The quartz substrate measured 10 mm $\times$ 10 mm $\times$ 1 mm. The patterned area was 4.5 mm $\times$ 4.5 mm.

 

Fig. 7. (a) Optical microscope and (b, c) scanning electron microscope images of the prepared metasurface polarizer. Figures (b) and (c) are the top and cross-section views, respectively.

Download Full Size | PDF

The reflectance spectra were measured by an ultra-violet (UV) / visible (VIS) / near-infrared (NIR) spectrophotometer (V-770ST with ARMN-920, JASCO, Japan). The spot size and angular divergence of the incident beam were 3 mm$\phi$ and below 0.1$^\circ$, respectively. An aperture was inserted in front of the sample, which reduced the beam spot size. We measured the reflectance spectra from $5^\circ$ to $60^\circ$ in $1^\circ$ steps. Figure 8 shows the angle-resolved reflectance spectra displayed with offset for p-(a) and s-polarizations (b). In Fig. 8(a), the spectra show three features. The first is the broad dip around 450 nm at 5$^\circ$ which does not shift strongly with the incident angle. This feature is also found in the numerical calculation, indicating that the dip originates from the gap SPPs of the first order. The other two features are found below 300 nm at 5$^\circ$. The lower energy mode is assigned to gap SPPs of the second order and the higher energy mode is assigned to propagating SPPs at the Al/substrate interface. The experiment reproduced the overall features of the calculations (Fig. 8(c)), although with small deviations in the resonance wavelengths and the spectral shapes owing to imperfections in the sample. On the other hand, the reflectance spectra for s-polarization (Fig. 8(b)) in the visible region show an almost featureless optical response. The reflectance significantly decreases in the UV region ($<$ 400 nm). In particular, the decrease in the DUV region ($<$ 320 nm) is very strong because of the strong light scattering, resulting in the decrease in the specular reflectance. This decrease in the spectral intensity has little dependence on incident angle. The overall features in the reflectance for s-polarization are consistent with the calculation results (Fig. 8(d)).

 

Fig. 8. (a, b) Measured and (c, d) calculated angle-resolved reflectance spectra displayed with offset for (a,c) p- and (b,d) s-polarizations. The reflectance spectra are highlighted in blue in $10^\circ$ steps.

Download Full Size | PDF

Figure 9(a) shows the angle-resolved extinction ratio spectra displayed with offset. There are two main peaks around 300 nm, which shift with the incident angle. The higher energy peak originates from the propagating SPPs at the Al-substrate interface. The lower energy peak originates from the gap SPPs. These two modes interfere with each other, causing Fano interference [36–38]. This decreases the extinction ratio from $30^\circ$ to $40^\circ$ and, strongly affects its magnitude. To clearly show the features of the spectral shape, we show the extinction ratio spectra at incident angles of $5^\circ$, $15^\circ$, $45^\circ$, $50^\circ$, and $60^\circ$ in Fig. 9(c). When the incident angle is below $45^\circ$, the width of the main peak does not vary strongly. When it is $50^\circ$, the width of the main peak becomes very narrow near the resonance with a broad width around 260 nm. In the presence of two resonances with broad and sharp line widths, a very sharp resonance with asymmetric spectral shape appears in a specific condition. This is a feature of Fano resonance. When the incident angle is $60^\circ$, the extinction ratio around 300 nm diminishes significantly because of the absence of Fano resonance. The extinction ratios around 400 nm become higher following the increase in the incident angle. This feature also agrees with that of the calculation result (Fig. 9(b)). Even though the details of the measured spectra are different from those of the calculations owing to imperfections in the sample (Fig. 7(c)), the main features agree. Thus, the observed extinction ratio spectra show that the device served as a polarizer, and the performance is strongly affected by Fano resonance. The maximum value of the extinction ratio is about 150 at $50^\circ$ (Fig. 9(c)), much smaller than calculated. The difference can be attributed to three reasons. One is the strong light scattering in the DUV region. As we discussed in the explanation of the measured reflectance in Fig. 8(b), the measured $R_s$ in the DUV region decreases strongly because of the strong light scattering. Following the decrease in $R_s$, the extinction ratio $R_s/R_p$ decreases. The second is the dynamic range of our experimental method. Since extremely low reflectance such as $10^{-5}$ cannot be measured by the spectrophotometer, the much enhanced extinction ratio shown in Fig. 6(d) cannot be demonstrated in our experimental setup. The last is the angular divergence of the incident beam, which degrades the purity of the interference and the performance of the polarizer. To measure the ultra-high extinction ratio shown in the numerical calculations, special experimental techniques using a laser light source [26] are required. This issue will be discussed elsewhere.

 

Fig. 9. (a) Measured and (b) calculated angle-resolved extinction ratio spectra. The spectra are displayed with offset and highlighted in blue in $10^\circ$ steps. The logarithm of the extinction ratio is plotted in the calculation results to clearly show the spectral features. (c) Measured extinction spectra at incident angles of $5^\circ$, $10^\circ$, $45^\circ$, $50^\circ$, and $60^\circ$.

Download Full Size | PDF

5. Conclusion

We numerically and experimentally studied a metasurface polarizer in the DUV region. The polarizer consists of a trilayer of Al grating/Al$_2$O$_3$/Al on a quartz substrate. The calculated extinction ratio exceeds $6.2\times 10^6$ around 259 nm at normal incidence. The metasurface serves as a mirror with high reflectance for $y$-polarization and extremely low reflectance induced by interference between the propagating and gap SPPs for $x$-polarization, serving as a high-performance polarizer. The numerical calculations suggest that the bandwidth of the polarizer working at $45^\circ$ incidence would be improved by adjusting the incident angle, period, and thickness of the $\mathrm {Al_2O_3}$ thin layer of the metasurface. The maximum value of the extinction ratio exceeds $2.78\times 10^4$, which is enough for practical applications. The bandwidth at an extinction ratio of $\ge$ 100 is about 31 nm. Following the numerical studies, we prepared a DUV metasurface polarizer and successfully observed the signature of the increase in the extinction ratio by Fano resonance. We numerically showed that high-performance polarizers are possible in the DUV and experimentally demonstrated the increase in the extinction ratios by Fano resonance between the propagating and local SPPs for the first time. Our study contributes to the progress of high-performance metasurface polarizers in the DUV region.