Boosting Metasurface Polarizer Performance: Numerical Study of Degradation and Optimization

Background

Controlling light at the nanoscale underpins modern photonics, from high‑Q photonic‑crystal cavities [1] to ultra‑small plasmonic resonators [6,7]. Metasurfaces, engineered 2‑D arrays of subwavelength scatterers, have emerged as versatile platforms for manipulating refraction [9], reflection [10], photoluminescence [11], wave plates [15], and beam splitting [16]. In particular, polarization control is a critical functionality, and a stacked complementary metasurface polarizer has been shown to achieve extinction ratios of ~10,000 in the near‑IR [23–26]. These structures rely on Babinet’s principle: a resonant complementary pair delivers high transmission for one polarization while suppressing the orthogonal component. However, the use of silver, which oxidizes and degrades in air, raises questions about long‑term reliability. Alternatives such as gold reduce loss but compromise performance. This work addresses the stability issue by identifying the physical origin of degradation and proposing design adjustments that enhance both performance and durability.

Methods/Experimental

The extinction ratio was measured with a pulsed OPO source (7 ns, 20 Hz) filtered by a Glan‑laser prism. Transmitted and reference signals were captured by an InGaAs detector, with a beam sampler and neutral‑density filter used to stabilize the intensity per pulse. The extinction ratio η is obtained directly from the ratio of the high‑ and low‑transmission signals (Equation (2) in the original manuscript).

A schematic of the extinction‑ratio measurement setup. M mirror, PH pinhole, L lens, BS beam sampler, BD beam damper, NDF neutral density filter, GLP Glan‑laser prism, D detector

Numerical validation employed RCWA with scattering‑matrix integration [29,30] and an inverse Fourier method [31]. Metal and dielectric permittivities were taken from established optical data [32,33]. Rough‑surface effects were modeled in COMSOL Multiphysics: (i) direct structural changes (bulk permittivity) and (ii) increased loss (modified imaginary permittivity). Calculations converged to <1 % tolerance.

Results and Discussion

The fabricated device is a three‑layer metasurface (Figure 2). The top and bottom silver layers (45 nm) host complementary rectangular‑hole pairs (150 × 540 nm, period 900 nm), while a 200 nm silica spacer separates them. SEM images (Figure 3) confirm the intended geometry.

Device architecture: (a) three‑layer stack; (b) complementary hole pair; (c) unit‑cell geometry; (d) layer thicknesses.

SEM images of the metasurface polarizer.

Spectrophotometer measurements (Figure 4) reveal high transmission for the x‑polarization and negligible transmission for y, confirming the ultra‑high extinction ratio. The custom setup (Figure 1) extends the dynamic range, yielding an extinction peak >20,000 at ~1640 nm (Figure 5b). Numerical RCWA reproduces these spectra with a peak of 15,000, validating the experimental data.

Measured transmittance for x (blue) and y (green) polarizations.

(a) Transmittance spectra for x and y. (b) Extinction ratio spectrum.

Time‑dependent measurements (Figure 7) show rapid degradation: after 6 days the extinction remains >20,000, but after 9 days it falls to ~500 with a broadened, blue‑shifted peak. This suggests a loss‑driven mechanism rather than a geometric change.

Extinction ratio evolution over 6–9 days.

Two rough‑surface models were explored: (1) a sinusoidal surface with Gaussian noise and (2) randomly distributed hemispherical nanoparticles. Both simulations (Figures 9 and 11) show that surface roughness alone does not significantly affect the extinction ratio; Babinet’s principle preserves the complementary response. However, incorporating an increased imaginary permittivity (parameter C in Eq. (3)) reproduces the dramatic drop and blue shift observed experimentally (Figure 12). Thus, increased metallic loss—stemming from surface scattering and grain‑boundary effects—dominates degradation.

Effect of increased loss (C = 1–5) on extinction spectra.

To counteract the loss sensitivity of the low‑transmission dip, we varied the thickness of the bottom silver layer while keeping the top layer fixed at 45 nm. Numerical results (Figure 14) show that increasing the bottom thickness to 60–65 nm aligns the high‑transmission peak with the low‑transmission dip, yielding a ~20% boost in extinction ratio. Fabrication can realize this by successive metal deposition and selective removal of the top layer.

(a) Electric field at high‑transmission peak. (b) Magnetic field at low‑transmission dip.

Transmittance and extinction spectra for varying bottom‑layer thicknesses (35–65 nm).

Conclusions

We demonstrated that a silver‑based metasurface polarizer can achieve extinction ratios exceeding 20,000 but that its performance decays rapidly due to increased metallic loss. Surface roughness alone is benign, but the associated rise in imaginary permittivity drives the degradation. By engineering the complementary‑layer thicknesses, the extinction ratio can be both maximized and stabilized. These insights guide the design of high‑performance, long‑lasting metasurface devices for photonic integration.

Abbreviations

BS

Beam sampler

GLP

Glan‑laser prism

NDF

Neutral density filter

OPO

Optical parametric oscillator

PhC

Photonic crystal

SEM

Scanning electron microscope

YAG

Yttrium iron garnet