Highly Efficient and Tunable Linear-to-Circular Polarization Conversion Enabled by Topological Unidirectional Guided Resonance

Abstract

The absence of natural sources of circularly polarized light has created a critical demand for high-performance polarization converters in emerging technologies such as quantum information processing, biomedical diagnostics, and advanced magneto-optical recording systems. Despite significant efforts, linear-to-circular polarization converters that combine high efficiency with tunability remain scarce. In this paper, we present a multifunctional and tunable device with 99.9% efficiency for converting linear to circular polarization in the telecom wavelength range. It can also function as a phase-only compensator with a controllable phase range of 1.93$\pi$ while maintaining a reflectance above 99%. The angular sensitivity analysis indicates that the device can tolerate incidence angles ranging from 8 to 10.4 degrees. This high-efficiency polarization conversion device offers significant advantages in integrated optics due to its multifunctionality and tunability and has promising applications in optical phased-array radars and quantum communication.

1. Introduction

Circularly polarized light (CPL) is fundamental to a wide range of applications, including optical communications [1], quantum computing [2], all-optical magnetic recording [3], and materials characterization [4]. The prominence of CPL arises from its superior information-carrying capacity and enhanced resistance to interference in certain scenarios, compared to linearly polarized light (LPL). Unfortunately, most conventional light sources do not directly emit CPL, necessitating the use of quarter-wave plates. These plates are traditionally fabricated from birefringent crystals or polymers to convert LPL into CPL. However, these techniques are often hampered by limitations such as device size, operational bandwidth, and lack of tunability. Although achromatic polarization conversion can be achieved using prisms that leverage total internal reflection like Fresnel rhombs [5], their large size hinders device compactness and integration.

In contrast, metasurfaces can exhibit strong optical anisotropy through carefully designed meta-atoms that possess distinct, polarization-dependent responses. This capability facilitates the development of miniaturized planar waveplates [6,7,8,9,10,11]. Various plasmonic metasurfaces based on metallic meta-atoms have been functionalized as ultrathin waveplates [12,13,14,15,16,17]. However, waveplates based on plasmonic metasurfaces typically exhibit non-ideal polarization conversion efficiency at optical wavelengths due to the inherent ohmic losses in metals. In contrast, dielectric resonant metasurfaces, which minimize absorption losses, can achieve significantly higher overall conversion efficiency. For instance, Yang et al. demonstrated a reflection-mode half-wave plate with 98% efficiency in the near-infrared (NIR) regime, leveraging Mie resonances in 2D structured silicon rectangular resonators on a silver ground plane [18]. Separately, Kruk et al. reported an all-dielectric transmissive quarter-wave plate achieving 90% transmittance and 99% conversion efficiency [19].

These high-efficiency polarization converters are typically limited to a single functionality and a fixed operating wavelength range. Consequently, a strong demand exists for a compact, tunable polarization conversion device capable of multi-polarization manipulation and wavelength control. Wu et al. proposed a near-infrared active metasurface, where indium tin oxide (ITO) was used as an active element for dynamic polarization conversion [20]. However, its linear-to-circular conversion efficiency was only 8%. Recently, an all-dielectric metasurface based on quasi-bound states in the continuum (Q-BIC) resonance has been proposed to achieve tunable transmitted polarization conversion [21]. Nevertheless, the use of doped silicon introduces substantial insertion losses, resulting in a maximum transmittance of only 0.77 and consequently reduced conversion efficiency.

In this work, we present the design of a multifunctional, tunable, and highly efficient polarization conversion device. This device can convert linear to circular polarization, rotate linear polarization, and modulate the reflective phase across the telecom wavelength range. The device comprises double-layer gratings that leverage the unique properties of topological unidirectional guided resonance (UGR) and guided-mode resonance (GMR) to achieve broadband nearly perfect reflection within the NIR regime while introducing a controllable phase retardation between orthogonal polarization states. The active element is antimony selenide (Sb₂Se₃), a phase change material (PCM) known for its ultralow optical loss, reversible phase transition, and non-volatility in the NIR regime [22,23]. By exploiting the properties of UGRs and GMRs, we realize a linear-to-circular polarization conversion efficiency of 99.9% with a tunable range of 25 nm. Additionally, the device can function as a phase-only compensator, providing a phase shift range (bandwidth) of 1.93$\pi$ with near-zero insertion loss over the wavelength band from 1550 nm to 1570 nm. We also investigated the angular tolerance of the structure. The angular sensitivity of the gratings constrains the optimal performance to an incident angle range of 8° to 10.4°.

2. Results and Discussion

The proposed metamaterial design features a bilayer grating architecture integrated with a phase change material, as illustrated in Figure 1a. The upper layer incorporates an inverted trapezoidal silicon grating integrated with a slab of Sb₂Se₃. The lower structure consists of silicon gratings embedded within a silicon oxide substrate. The upper cladding of the structure is air. Although inverted-trapezoidal designs pose additional fabrication challenges, particularly in the etching process, structures with asymmetric tilted sidewalls supporting UGRs have been successfully fabricated, demonstrating good consistency between experimental and simulation results [24].

Full-wave electromagnetic simulations were performed using the finite element method (COMSOL Multiphysics®). In our simulations, the incident wave vector was set parallel to the X-direction. The refractive indices used were 3.45 for silicon, 1.46 for silica, and 3.285 for a-Sb₂Se₃, with the latter assumed to have negligible loss [22]. This assumption of negligible loss is justified by the experimentally measured ultralow extinction coefficient ($k<{10}^{-5}$) of Sb₂Se₃ in both its amorphous and crystalline phases across the telecom wavelength range [22,23]. The structural design is based on the theory of UGRs [25,26,27], which are typically realized by breaking the symmetry of the photonic-crystal slab to create polarization singularities. The grating period was first estimated by applying the grating equation along with the structural configuration to establish a preliminary parameter range. We then performed joint parameter sweeps in COMSOL to maximize the asymmetric radiation contrast exceeding 60 dB. The optimized period of the structure is $P=714$ nm, and other geometric parameters are annotated in Figure 1b.

The device achieves a peak polarization conversion efficiency exceeding 99.9% by simultaneously satisfying two critical conditions: near-perfect reflection for both polarization states and a tunable phase delay between them. To validate this, we first simulated the structure’s intrinsic modal properties. Figure 2a shows the band structure and asymmetric radiation ratio for transverse-electric (TE) polarization ($E_{\Vert }$ in Figure 1a) along the $\Gamma -X$ direction, revealing an asymmetric radiation contrast of up to 74.26 dB at $k_{x}P/\left(2\pi \right)=0.07293$. We also calculated the far-field polarization states of the downward radiation in momentum space, as shown in Figure 2b. Notably, the position of the bound state in the continuum (BIC), which carries a topological charge of −1, coincides with the point of maximum asymmetric radiation (74.26 dB). This confirms that the high asymmetric radiation ratio originates from the UGR mode [24,26,28,29,30]. Furthermore, Figure 2c displays the electric field amplitude profile along the Z-direction under TE polarization and the magnetic field amplitude profile under transverse-magnetic (TM) polarization ($E_{\perp }$ in Figure 1a), respectively. The former mode is primarily confined to the upper silicon grating layer, whereas the latter is predominantly confined to the lower silicon gratings layer. Figure 2d and Figure 2e present the reflectance and reflection phase spectra for TE and TM polarizations, respectively. The reflectance exceeds 99% for both polarization states across the wavelength range from 1529 nm to 1557 nm, indicating broadband, nearly perfect reflection in the NIR with ultra-low insertion loss. Regarding the reflection phase, the TE mode exhibits an abrupt phase jump near the resonance center, while the TM mode shows a gradual, normal dispersion. This results in a controllable phase delay between the two modes, ranging from approximately $-0.25\pi$ to $0.89\pi$ over the wavelength interval of 1531 nm to 1551 nm.

The distinct polarization-dependent responses originate from fundamentally different resonance mechanisms. Under TE polarization, the resonance is predominantly a UGR of the upper grating, characterized by broadband nearly total reflection and an abrupt phase jump [25,27]. In contrast, under TM polarization, the resonance follows a leaky-mode mechanism in the lower grating, which is also characterized by broadband nearly total reflection [31,32].

After obtaining the complex reflection coefficients (amplitude and phase) for TE and TM polarizations, the circular polarization conversion efficiency of the structure can be retrieved as follows [8]:

$$\begin{array}{c}\hfill r_{\mathrm{RCP}}={\displaystyle \frac{1}{2}}\left|r_{\Vert }{e}^{i{\varphi }_{\Vert }}+i{r}_{\perp }{e}^{i{\varphi }_{\perp }}\right|\\ \hfill r_{\mathrm{LCP}}={\displaystyle \frac{1}{2}}\left|r_{\Vert }{e}^{i{\varphi }_{\Vert }}-i{r}_{\perp }{e}^{i{\varphi }_{\perp }}\right|\end{array}$$

(1)

where $r_{\mathrm{RCP}}$ and $r_{\mathrm{LCP}}$ denote the reflection coefficients for right-handed and left-handed circular polarization (RCP and LCP), respectively. Here, $r_{\Vert }$ and $r_{\perp }$ represent the amplitude reflectivities for the cases where the incident electric field is parallel and perpendicular to the gratings, corresponding to TE and TM polarization, respectively. ${\varphi }_{\Vert /\perp }$ are the reflection phases. The corresponding power conversion efficiency $\eta$ for linear-to-circular polarization is then given by

$$\begin{array}{c}\hfill {\eta }_{\mathrm{RCP}}=r_{\mathrm{RCP}}^2={\displaystyle \frac{1}{4}}\left[R_{\Vert }+R_{\perp }+2\sqrt{R_{\perp }{R}_{\Vert }}sin(\Delta \varphi )\right]\\ \hfill {\eta }_{\mathrm{LCP}}=r_{\mathrm{LCP}}^2={\displaystyle \frac{1}{4}}\left[R_{\Vert }+R_{\perp }-2\sqrt{R_{\perp }{R}_{\Vert }}sin(\Delta \varphi )\right]\end{array}$$

(2)

where R is the reflectance and $\Delta \varphi ={\varphi }_{\Vert }-{\varphi }_{\perp }$ is the phase retardation. The solid lines in Figure 2f depict the conversion efficiencies retrieved from the complex reflection coefficients of the TE and TM polarizations. At 1541.75 nm, the conversion efficiency reaches 99.9%, demonstrating the structure’s capability for highly efficient, lossless linear-to-circular polarization conversion.

To validate the correctness of Equation (2), direct simulations of linear-to-circular polarization conversion were performed. In these simulations, the incident electric field was oriented at a $45°$ angle relative to the incident plane (X-direction). The power ratios of the reflected light converted into RCP and LCP were then monitored. The results, depicted as dashed lines in Figure 2f, align with the retrieved efficiency.

The tunability of the designed structure is mainly through the phase transition of the Sb₂Se₃ slab between the amorphous state and the crystalline state. During the phase-transition process, the refractive index change in the layer is not a gradual and homogeneous increase from 3.285 (n(a-Sb₂Se₃)) to 4.2 (n(c-Sb₂Se₃)). The process is characterized by localized, complete crystallization within the laser spot, where the refractive index shifts abruptly from 3.285 to 4.2. This results in the coexistence of amorphous and crystalline phases in the Sb₂Se₃ layer, with their proportion being adjusted dynamically according to the desired tuning. The effective refractive index of the entire Sb₂Se₃ layer in this hybrid state is typically estimated using effective medium theories, such as the Lorentz–Lorenz relation [33]:

$${\displaystyle \frac{{\epsilon }_{\mathrm{h}}\left(\lambda \right)-1}{{\epsilon }_{\mathrm{h}}\left(\lambda \right)+2}}=m{\displaystyle \frac{{\epsilon }_{\mathrm{c}-{\mathrm{Sb}}_2{\mathrm{Se}}_3}\left(\lambda \right)-1}{{\epsilon }_{\mathrm{c}-{\mathrm{Sb}}_2{\mathrm{Se}}_3}\left(\lambda \right)+2}}+(1-m){\displaystyle \frac{{\epsilon }_{\mathrm{a}-{\mathrm{Sb}}_2{\mathrm{Se}}_3}\left(\lambda \right)-1}{{\epsilon }_{\mathrm{a}-{\mathrm{Sb}}_2{\mathrm{Se}}_3}\left(\lambda \right)+2}}$$

(3)

where ${\epsilon }_{\mathrm{h}}\left(\lambda \right)$ is the effective dielectric of the hybrid state and m is the crystallization fraction of Sb₂Se₃. ${\epsilon }_{\mathrm{c}-\begin{array}{c}\hfill {\mathrm{Sb}}_2{\mathrm{Se}}_3\end{array}}\left(\lambda \right)$ and ${\epsilon }_{\mathrm{a}-\begin{array}{c}\hfill {\mathrm{Sb}}_2{\mathrm{Se}}_3\end{array}}\left(\lambda \right)$ are the wavelength-dependent dielectric constant of Sb₂Se₃ in crystalline and amorphous state, respectively. Considering that the experimental results of Sb₂Se₃’s refractive index curve is almost flat in the C-band [22,23], for computational simplicity, here we adopt Sb₂Se₃’s refractive index at 1550 nm as a constant value throughout the 1520–1570 nm spectral range in the simulations.

The phase transition in Sb₂Se₃ is thermodynamically driven, requiring heating above its crystallization temperature (200 °C) with controlled cooling rates. This can be achieved via thermal annealing, pulsed laser irradiation, or electrical pulses through transparent electrodes [22,23,34]. For pulsed laser method, a 638 nm pulsed diode laser operating higher than 90 mW can meet the requirement. The experiment by M. Delaney et al. shows that 100 ms pulsed width with power larger than 60 mW can transform localized amorphous regions to crystalline states while 400 ns pulsed width with power larger than 60 mW can cause amorphization of the crystallized region [22]. In addition, the degree of crystallization can be monitored in situ using spectroscopic ellipsometry.

Figure 3a and Figure 3b illustrate the evolution of the structure’s reflectance in response to the gradual crystallization of a-Sb₂Se₃ for TE and TM polarizations, respectively. For TE polarization, the resonant wavelength redshifts as the refractive index increases. This redshift occurs because the resonant mode under TE polarization is strongly confined near the Sb₂Se₃ layer. Consequently, an increase in the refractive index induces a redshift of the UGR. In contrast, for TM polarization, the resonant energy is located far from the Sb₂Se₃ layer, resulting in the resonant wavelength being largely independent of n(Sb₂Se₃), as evidenced in Figure 3b. The black solid line in Figure 3c marks the phase retardation of 0.5$\pi$, demonstrating that a refractive index increase of 0.1 in Sb₂Se₃ layer induces a 11 nm redshift in the resonance wavelength at the 0.5$\pi$ phase position.

Figure 3c illustrates the phase retardation between two orthogonal electric fields with increments in n(h-Sb₂Se₃). The solid black line denotes a phase delay of 0.5$\pi$ of TM relative to TE, the dashed black line indicates a delay of −0.5$\pi$, and the solid yellow line represents a delay of −$\pi$. All three lines exhibit a linear response to changes in the refractive index, indicating that the desired phase retardation can be customized by controlling n(h-Sb₂Se₃) while maintaining nearly perfect reflectivity. Figure 3d and Figure 3f depict the conversion efficiency from LP to RCP and LP to LCP, respectively. As shown in Figure 3d, by adjusting the refractive index from 3.285 to 3.51, the operating wavelength for LP-to-RCP conversion shifts from 1541.75 nm to 1566.75 nm. This provides a tunable range of 25 nm while maintaining a conversion efficiency above 99%, demonstrating the excellent tunability of the device. A similar tuning effect is observed for LCP conversion, as shown in Figure 3e. This result implies that across the wavelength range from 1541.75 nm to 1566.75 nm, the structure can be dynamically configured to generate either RCP or LCP with minimal insertion loss. Such dual-output capability is uncommon in conventional linear-to-circular polarization converters [8,9]. In addition, owing to its flexible phase retardation control, the proposed structure can also serve as a half-wave plate, effectively achieving nearly lossless rotation of linearly polarized light. For instance, it allows x-polarized light to be converted to y-polarized light, a function crucial for polarization control applications. Moreover, when excited by a TE-polarized (y-polarized) incident field, the structure functions as a phase-only compensator with reflectivity close to 100%. We also computed the variation in the structure’s reflection phase with the refractive index of Sb₂Se₃ at fixed wavelengths within the near-infrared range. As shown in Figure 3f, the reflection phase can be tuned over a range approaching 2$\pi$ across the wavelength band from 1550 to 1570 nm.

For reflective conversion devices, angular tolerance is an important metric. To this end, we calculated the reflectance and conversion efficiency of the structure under incident angles ranging from $7°$ to $11°$, as depicted in Figure 4. To maintain broadband near-total reflection (R > 0.995) for both TE and TM polarizations, the incident angle must be confined to a range of $8°$ to $10.4°$ (Figure 4a,b). Due to the sensitivity of the grating diffraction order to the angle of incidence, the calculated LP-to-RCP conversion efficiency exceeds 99% only within this $8°$–$10.4°$ range, corresponding to an angular working tolerance of $2.4°$ (Figure 4d). At a fixed operating wavelength (e.g., 1542 nm), the device exhibits an operational angular stability of approximately ±0.5° with less than 10% efficiency degradation. This limited angular tolerance stems primarily from the working principle of gratings, specifically the high sensitivity of the diffraction conditions to the incident angle. Future design iterations could explore advanced grating profiles or dispersion-engineering techniques to improve angular tolerance, albeit potentially at the cost of other performance parameters.

In micro–nano fabrication, overlay errors between different process layers are common, typically ranging from several nanometers to tens of nanometers. To numerically model overlay errors, the effect of misalignment between the upper and lower grating layers on conversion efficiency is calculated. Figure 4e illustrates the lateral offset (d) between the central axes of the gratings within one period. We simulated the reflectance and phase retardation for lateral offsets ranging from 0 to 150 nm and evaluated the corresponding LP-to-RCP conversion efficiency as an example. As shown in Figure 4f, conversion curves of different offsets are almost identical, showing the misalignment-insensitivity of the structure, which significantly reduces fabrication difficulty. This robustness can be understood intuitively from the underlying physics. The proposed structure achieves polarization conversion primarily through two distinct resonant modes separated by a sufficiently large space interval (2 μm), ensuring little near-field influence between each other. The reflectance of each mode depends solely on its respective geometric parameters. The phase difference, meanwhile, stems from both the intrinsic phase response of the resonant modes and the geometric phase introduced by the propagation path difference between them. Crucially, both factors are inherently independent of lateral misalignment between the grating layers. While lateral misalignment robustness is demonstrated, the impact of other fabrication imperfections, such as etch depth variation, sidewall angle, and line-width roughness, constitute an essential area for future investigation as the design progresses towards experimental realization.

3. Conclusions

In summary, we have developed a multifunctional, highly efficient and tunable polarization conversion device. Our design incorporates double-layer gratings that leverage both unidirectional guided resonance (UGR) and guided-mode resonance (GMR). This configuration, combined with a phase-change material, enables flexible control of the phase difference between orthogonal electric field components. This device achieves an impressive conversion efficiency of 99.9% for the transformation from linear to circular polarization with a tunable range of 25 nm (1541.75–1566.75 nm). Additionally, it can also function as a half-wave plate, allowing for the rotation of linear polarization at arbitrary angles. Moreover, the structure can serve as a phase-only compensator with a phase shift range of 1.93$\pi$. The device operates effectively within an incident angle range of 8 to 10.4 degrees. The device can also be adapted for visible or mid-infrared wavelengths by adjusting the geometric parameters accordingly. Owing to its multifunctionality and tunability, this high-efficiency polarization conversion device holds significant promise for integrated optics, with potential applications in quantum communication, optical phased-array radars, and photonic-crystal-based surface-emitting lasers.