Ultra‑Efficient Silicon Metasurfaces for Polarization Beam Splitting and Vortex Beam Generation at Telecom Wavelengths

Background

Recent advances in electromagnetics have focused on the full control of light waves, with metamaterials emerging as a pivotal research area due to their engineered optical responses [1]. These artificial structures have enabled phenomena such as negative refraction, zero‑refraction, and slow light [2‑4]. However, conventional three‑dimensional metamaterials suffer from intrinsic losses and complex fabrication, limiting their practical deployment. Two‑dimensional metamaterials, or metasurfaces, mitigate these drawbacks by offering ultrathin, sub‑wavelength structures that are easier to fabricate and integrate [5‑7]. Metasurfaces manipulate the amplitude or phase of incident light through sub‑wavelength resonators, enabling functionalities such as tunable waveguides, wave‑plates, lenses, anomalous refraction, vortex generators, and holograms [8‑19]. While metasurfaces outperform their 3D counterparts in efficiency, losses remain a concern, especially when metallic components are involved. To overcome this, Huygens’ metasurfaces and all‑dielectric metasurfaces have been proposed. Huygens’ designs reduce losses but still face fabrication challenges due to their 3D nature [20]. All‑dielectric metasurfaces, by overlapping electric and magnetic resonances at the same frequency, achieve full 2π phase control with high transmission efficiency [21‑27]. Yet, most demonstrated devices operate in the first or second diffraction orders; high‑order diffraction modes remain underexplored, primarily due to low efficiencies caused by metal losses [28‑32]. In this work, we present an all‑dielectric metasurface capable of manipulating wavefronts in high‑order diffraction modes while maintaining transmission efficiencies above 88 %. Using this platform, we design two polarizing beam splitters that separate orthogonal polarizations into distinct directions with high efficiency, and two vortex‑beam generators that produce topological charges of 2 and 3, showcasing the versatility of high‑order diffraction control.

Methods

The metasurface design is illustrated in Fig. 1a. It consists of 900‑nm‑thick crystalline silicon nanobricks fabricated on a 200‑nm‑thick SiO₂ substrate (refractive indices 3.48 and 1.48, respectively). Silicon’s high refractive index confers strong resonances with minimal ohmic loss, and its fabrication can be realized via mature semiconductor techniques such as electron‑beam lithography (EBL) and focused ion beam (FIB) milling. The SiO₂ substrate was chosen to suppress reflection and absorption at the target telecom wavelength of 1500 nm. A lattice constant of S = 650 nm was adopted. Transmission phase control is governed by the nanobrick dimensions along the X and Y axes. Numerical simulations were performed using the finite‑difference time‑domain (FDTD) method, with perfectly matched layers (PML) at the top and bottom and periodic boundary conditions (PBC) laterally. The chosen operation wavelength is 1500 nm. Figure 1a–d show the co‑polarized transmission efficiency and phase variation for XLP and YLP light as functions of the nanobrick dimensions (a = width, b = length). The simulations reveal that, for the vast majority of geometries, co‑polarized transmission exceeds 88 %, while the phase can be tuned across the full 0–2π range. The symmetry of the structure ensures similar performance for both polarizations, enabling independent phase control for XLP and YLP light by adjusting a and b, respectively.

a Transmission efficiency and b corresponding phase of XLP light versus a and b. c Co‑polarized efficiency and d phase of YLP light versus a and b. Inset: unit cell schematic (silicon nanobricks on SiO₂). Thicknesses: 900 nm (silicon) and 200 nm (SiO₂).

Results and Discussion

Design of Polarizing Beam Splitters

On‑chip polarization control is essential for integrated photonics. The polarizing beam splitter (PBS) we propose exploits the distinct phase responses for XLP and YLP light, enabling separation into two spatially distinct beams with high efficiency.

Each PBS is constructed from 13 nanobricks arranged into a supercell. By discretizing the phase over 0–2π for XLP light and 2π–0 for YLP light with a step of ±2π/13, we achieve a high‑order phase gradient. The lateral dimensions of the 13 bricks are listed in Fig. 2a. Reordering these bricks yields higher‑order diffraction modes: the third‑order mode (M₃) spans 0–3×2π, while the fifth‑order mode (M₅) spans 0–5×2π. This simple rearrangement expands the phase coverage proportionally to the diffraction order.

Design of dielectric metasurfaces with three diffraction orders. a Lateral dimensions of the 13 nanobricks for M₁ (0–2π), M₃ (0–3×2π), and M₅ (0–5×2π). b Simulated transmission phases for XLP (black) and YLP (blue). c Brick dimensions and transmitted efficiencies for M₁: co‑polarized efficiencies remain above 88 % except two bricks near 80 %.

Simulations of the PBS under 45° linearly polarized illumination confirm the design. Extracted XLP and YLP fields (Fig. 3a) show distinct wavefronts with peak diffraction angles of –10.2° (XLP) and +10.2° (YLP). The corresponding co‑polarized efficiencies are 85.9 % (XLP) and 88.4 % (YLP). The diffraction angles match the generalized Snell’s law prediction (±10.22°), validating the high‑order beam steering capability and confirming that nearly all incident light is transmitted with minimal reflection.

a Electric field distributions of extracted XLP (left) and YLP (right) under 45° illumination. b Co‑polarized efficiencies versus transmitted angle for XLP and YLP.

For higher‑order PBSs, the theoretical diffraction angles of M₃ and M₅ are ±32.18° and ±62.56°, respectively. Figures 4a–d illustrate the well‑defined phase fronts and confirm these angles. Co‑polarized efficiencies drop modestly to 82–84 % (third order) and 73.5–78.4 % (fifth order), primarily due to coupling between bricks of differing geometry. Nevertheless, the metasurfaces maintain high transmission in all high‑order modes.

Electric field distributions for M₃ (a,b) and M₅ (c,d) under 45° illumination.

Co‑polarized efficiencies versus transmitted angle for M₃ (a) and M₅ (b) under XLP and YLP illumination.

Design of Optical Vortex Generators

Optical vortex beams, characterized by a helical phase front and orbital angular momentum lℏ, find applications in lithography, trapping, and communications [37‑44]. To generate vortex beams, we arrange the 13 bricks of M₁ into 13 azimuthal sectors, creating a continuous phase ramp of 2π/13 across the circumference. Figure 6a shows the intensity profile of the generated l = 1 vortex, featuring a central null and a 2π phase jump from –π to π.

a‑c Transmitted intensity distributions; d‑f phase wavefronts for l = 1, 2, 3 using M₁, M₂, M₃ under X‑polarized incidence.

To achieve higher topological charges, we modify the phase step between adjacent bricks to 4π/13 (l = 2) and 6π/13 (l = 3), yielding M₂ and M₃. The resulting intensity profiles (Fig. 6b,c) and phase maps (Fig. 6e,f) exhibit two and three clear 2π phase jumps, respectively. Switching the incident polarization from XLP to YLP reverses the helical sense while preserving the intensity pattern, owing to the symmetrical phase response. This demonstrates that our metasurfaces can generate arbitrary vortex orders by simple brick reconfiguration.

Conclusions

We have engineered dielectric metasurfaces composed of periodic silicon nanobricks that provide full 0–2π phase coverage with transmission efficiencies exceeding 88 % at the telecom wavelength of 1500 nm. Leveraging these properties, we realized high‑order polarizing beam splitters that separate orthogonal polarizations into distinct directions with high efficiency, and optical vortex generators that produce topological charges up to 3 in high‑order diffraction modes. The simplicity of the design—modifying brick dimensions and arrangement—facilitates rapid fabrication and paves the way for a new class of ultra‑efficient photonic devices.