Design of a High-Efficiency Near-Infrared Circular Polarization Filter Responding to Dual Wavelengths Based on Twisted Bilayer Plasmonic Metasurfaces

Abstract

Circular dichroism is at the core of chiral spectroscopy and polarization light manipulation. However, achieving metal-based devices with high efficiency, compactness, and easy integration in the near-infrared band remains a significant challenge. Traditional metal chiral microstructures, such as broken open rings, helical lines, or waveplates–polarizers I confirm., are limited to circular dichroism values below 50% due to their inherent ohmic losses, severely restricting practical applications. To overcome this bottleneck, this paper proposes a twisted double-layer plasmonic metasurface composed of two anisotropic metal metasurfaces. This design breaks the mirror symmetry of the structure by precisely controlling the in-plane twist angle between the layers, inducing strong coupling and interference effects in the chiral optical response. Simulation results show that this device achieves excellent multi-wavelength circular dichroism control. At a wavelength of 1660 nm, the circular dichroism value reaches 0.48, and it further increases to 0.84 at 2200 nm, significantly surpassing the performance limits of traditional metal structures. This work not only provides a simple and scalable design paradigm for high-performance chiral optical devices but also opens up new avenues for advanced applications such as chiral molecular sensing, polarization coding, and quantum optics in the near-infrared band.

1. Introduction

Circularly polarized light, with its unique chiral characteristics, has shown broad application prospects in fields such as quantum optics [1], spin photonics [2], chiral molecule sensing [3], and three-dimensional display [4]. Unlike linearly polarized light, the electric field vector of circularly polarized light rotates helically during propagation, a property that enables it to serve as a probe to reveal the chiral characteristics of substances or as an information carrier for polarization multiplexing encoding [5]. In these applications, circularly polarized filters play a core functional role, selectively transmitting specific chiral circularly polarized light. The development of traditional circularly polarized filters can be traced back to the field of photography, with a typical structure consisting of a linear polarizer and a quarter-wave plate in series [6]. When non-polarized light passes successively through a linear polarizer and a quarter-wave plate with its optical axis at a 45° angle to the linear polarization direction, circularly polarized light of a specific chirality can be generated. Conversely, this combination can also be used to identify and filter incident circularly polarized light. However, this cascaded structure has inherent drawbacks [7]. Firstly, the device is relatively large, making it difficult to meet the compactness requirements of photonic integrated circuits. Secondly, the dispersion characteristics of the quarter-wave plate limit its phase retardation condition to a narrow band, restricting its operational bandwidth.

To fundamentally avoid the shortcomings of traditional circular polarization filtering devices in terms of performance and integration, chiral metasurfaces [8,9,10,11,12,13], as a cutting-edge artificial electromagnetic interface, have come into the research spotlight. This structure arranges subwavelength-scale “artificial atoms” on a two-dimensional plane to construct electromagnetic responses that natural materials do not possess, thereby precisely customizing the amplitude, phase, and polarization state of light waves at the deep subwavelength scale [14,15,16,17,18,19,20]. Its core advantage lies in the fact that through the chiral geometric design of artificial atoms, strong chiral optical effects can be excited within an extremely thin material layer. Examining the development trajectory of plasmonic chiral metasurfaces, a clear evolution from structurally complex to functionally integrated can be observed. Initially, researchers emulated chiral molecules in nature by constructing macroscopic three-dimensional helical arrays [21,22,23,24,25]. Although this structure could generate circular dichroism (CD) through intuitive geometric chirality, its fabrication process relied on complex micro-nano processing techniques, which were not only costly but also difficult to achieve uniform and repeatable preparation on a wafer scale. To break this deadlock, research focus gradually shifted to two-dimensional and quasi-three-dimensional structures, such as vertically stacked bilayer nanorods [26,27,28,29] or helical structure [30,31,32,33] with in-plane asymmetry. These structures utilized interlayer near-field coupling or in-plane symmetry breaking to effectively achieve chiral responses, greatly simplifying the fabrication process and enhancing the degree of freedom in controlling the polarization state. However, such metal-based metasurfaces have always been unable to bypass the physical limitations imposed by the materials themselves. Due to the inherent Joule losses of metals in the near-infrared band, the circular dichroism value of the devices in the transmission mode is often limited to below 50% [34]. This efficiency ceiling determined by material properties directly weakens the competitiveness of the devices in application scenarios that require high photon utilization, such as low-threshold chiral laser emission or high-performance polarization state converters, becoming a key obstacle hindering their practical application.

This paper proposes a chiral structure design based on twisted bilayer plasmonic (TBP) metasurfaces. This scheme differs from the traditional “waveplate–polarizer” cascading approach and is also distinct from three-dimensional helical configurations that rely on the intrinsic chirality of the structure. The designed device consists of two stacked anisotropic plasmonic metasurfaces, each of which does not possess intrinsic chirality. However, by precisely controlling the in-plane twist angle between the two layers, the mirror symmetry of the overall structure can be effectively broken, thereby inducing strong CD in the transmission spectrum. Simulation results show that the device achieves multi-wavelength CD enhancement in the near-infrared band. At a wavelength of 1660 nm, the CD value reaches 0.48, and it further increases to 0.84 at 2200 nm, significantly surpassing the 50% performance limit of traditional metal structures. This work provides a new design paradigm for efficient and compact chiral optoelectronic devices and is expected to promote the practical application of chiral photonics on integrated optical platforms.

2. Materials and Methods

Figure 1a schematically shows the unit cell structure of the proposed TBP metasurface. This metasurface is composed of a bottom SiO₂ substrate, a bottom plasmonic metasurface (metasurface b), an intermediate SiO₂ spacer layer, and a top plasmonic metasurface (metasurface a), which are stacked and infinitely extended in the x and y directions with sub-wavelength periods. The structure adopts a gold–SiO2 material system and has the potential to be compatible with commercial CMOS processes. Figure 1b presents a front view of the unit cell of the TBP metasurface, with the thickness information of each functional layer labeled. Figure 1c shows the corresponding top view, where the key geometric degree of freedom—the twist angle θ, which is the in-plane angle between the long axes of metasurface a and metasurface b—is defined. Figure 1d illustrates the combined configurations when the twist angle θ is 0° and 90°, respectively. The structures in these two states have mirror symmetry in the YZ plane. When θ takes any angle between 0° and 90° (excluding the endpoints), the overall structure no longer coincides with its mirror image, thus presenting chiral geometric features. The function of this TBP metasurface is similar to that of a circular polarization filter: it has a high transmission rate for left circularly polarized (LCP) light and a low transmission rate for right circularly polarized (RCP) light. The difference in transmission rates between the two is defined as the CD of the device. The transmission rate data were obtained through COMSOL software (Version 5.6) based on the finite element method, and the value is equal to the area integral of the Poynting vector at the transmission port. In the simulation model, due to the periodic arrangement of the unit cell of metasurfaces, the boundaries in the x and y directions were set as periodic boundary conditions. Perfectly matched layers (PML) were added to the top and bottom of the model to absorb boundary reflections, with a PML thickness of 4000 nm. The dispersion parameters of gold were taken from reference [35], and the maximum size in the meshing was 30 nm. The refractive index of SiO₂ in the target band was set to 1.46. The polarization state and intensity of the incident light were controlled through the incident port. The potential testing and fabrication methods of metasurfaces can be found in the Supplementary Materials.

3. Results

Figure 2a shows the transmission spectra of the proposed TBP metasurface for LCP and RCP in the wavelength range from 1500 nm to 3000 nm. For LCP, the transmission spectrum exhibits a double-peak feature within the studied wavelength range, with a spectral profile resembling the letter “M”. Two typical transmission peaks (labeled as peak a and peak c) are located near 1640 nm and 2240 nm, respectively, with corresponding peak transmission rates of approximately 0.89 and 0.84. For the case of RCP incidence, the transmission spectrum shows a more complex spectral profile, but overall, it is approximately mirror-symmetric to the LCP spectrum. The two transmission minima corresponding to RCP are labeled as valley b and valley d. Figure 2b calculates the CD spectrum of the device in the corresponding wavelength band, where CD is defined as the absolute value of the difference between the transmission rates of LCP and RCP, that is, CD = |T_LCP–T_RCP|. The results show that the metasurface has two significant CD peaks in the studied wavelength range, with peak values located at 1660 nm and 2200 nm, respectively, and corresponding CD values reaching 0.48 and 0.84. As shown in

Table 1. A comparison of plasmonic metasurfaces that emerged in the same period of work.

Structure	Number of Layers	CD	Number of CD Peaks
Four holes [31]	one	~3.8%@700 nm ~1.4%@630 nm	two
Interlaced rectangles [33]	one	30%1550 nm	one
Twisted metamaterials [22]	two	~40%770 nm	one
Moiré gratings [26]	two	40%@800 nm	one
Cascaded gratings [27]	two	10%@3800	one
Broken rings [20]	three	~70%@2200 nm ~20%@3800 nm	two
This work	two	48%@1660 nm 84%@2200 nm	two

, most of the work has focused on chiral control at a single wavelength, and the intensity of CD is not high. Although the broken-ring structure has achieved high-efficiency chiral control, it requires the stacking of three layers of metasurfaces, which is highly challenging in terms of process. The TBP metasurface we designed has significant advantages in terms of the intensity of the CD peak and the number of controllable CD peaks.

3.1. The Influence of Interlayer Twist Angle θ

To explore the physical origin of the chiral optical response of the TBP metasurface, the evolution of the CD spectrum with the interlayer twist angle θ is systematically analyzed. As shown in Figure 3a, the metasurface is composed of two vertically stacked anisotropic elliptical cylindrical gold metasurfaces. When the two layers are in-plane aligned (θ = 0° or 90°), the overall structure is mirror-symmetric, and its CD spectrum is strictly zero throughout the entire studied wavelength range. As the twist angle θ increases from 0° to 90° in 10° steps, the structure gradually evolves into a chiral configuration, and the CD spectrum shows a non-zero response, presenting two characteristic peaks throughout the entire wavelength range. Notably, the intensities of the two CD peaks first increase and then decrease as the twist angle θ increases, with the maximum value of the peaks corresponding to θ = 50° and θ = 30°, respectively. This evolution clearly indicates that the mirror symmetry breaking introduced by the interlayer twist is the fundamental cause of the chiral optical resonance in this system, rather than the intrinsic properties of the individual metasurface layers. To further reveal the physical mechanism of the two CD peaks, we analyzed the light absorption distribution of the device at the two characteristic wavelengths. Figure 3b shows the light absorption intensities of the top metasurface (metasurface a) and the bottom metasurface (metasurface b) at wavelengths λ₁ and λ₂, with all data have been uniformly normalized based on the value of the sixth group of data from left to right (the data on the far right) in the bar chart. The results indicate that the total light absorption intensity at λ₂ is approximately three times that at λ₁. From the wavelength spatial distribution perspective, at λ₁, the absorption loss is mainly concentrated in the top metasurface a. At λ₂, both the top and bottom metasurfaces contribute significantly to the absorption. This difference reveals that the two CD peaks correspond to resonant modes with different electromagnetic field localization characteristics. λ₁ may correspond to a localized mode dominated by the top layer, while λ₂ involves a strong coupling mode between the top and bottom layers.

3.2. The Equivalent Current Analysis at Characteristic Wavelengths

To reveal the physical mechanisms corresponding to the two CD peaks, Figure 4 shows the electric field intensity distribution and equivalent current distribution within the metasurface at the wavelengths of peak a and peak c in the LCP spectrum. The red arrows below each small figure in the figure are simplified diagrams of the equivalent current. At the incident wavelength corresponding to peak a, the electric field energy is mainly localized at the boundary between the bottom metasurface and the surrounding silica medium, and its intensity density is significantly higher than that at the top metasurface. From the equivalent current distribution, the top metasurface can be regarded as a pair of line currents oscillating in opposite directions along the Z-axis, which is a typical out-of-plane magnetic dipole resonance feature, as shown in Figure 4a. The bottom metasurface, on the other hand, exhibits two parallelly arranged electric dipoles, with their dipole moments existing in the XY plane and forming an angle of approximately −40° with the X-axis (Figure 4b). This current configuration indicates that peak a results from the coherent interference of the scattering field generated by the top magnetic dipole and the bottom tilted electric dipole with the incident light. At the incident wavelength corresponding to peak c, the distribution of the electric field intensity is similar to that corresponding to peak a. However, the response of the current distribution presents a completely different pattern. The equivalent current within the top metasurface is mainly distributed in the XZ plane and shows an oscillation feature of approximately 45° crossing (Figure 4c). In addition, Figure 4d shows that the bottom metasurface, on the other hand, exhibits a collection of two electric dipoles oscillating in the same direction and reinforcing each other. Therefore, peak c can be attributed to the cooperative coupling effect between the 45° crossing line current in the top and the parallel reinforcing electric dipoles in the bottom.

To further clarify the physical origin of the two transmission minima (valley b and valley d) in the RCP transmission spectrum, Figure 5 shows the electric field intensity distribution and equivalent current distribution of the metasurfaces at the corresponding wavelengths. Generally speaking, the electric energy density of the bottom metasurface is significantly higher than that of the top metasurface, which is mainly attributed to the difference in their mode volumes. In addition, unlike Figure 4, the electric field distribution in Figure 5 exhibits typical behavior of Localized Surface Plasmon Resonances (LSPRs). It can be observed that, regardless of whether it is valley b or valley d, the elliptical cylinders at the top metasurface have very sharp shapes at the two ends in the x-direction, which is an excellent location for amplifying the electric field intensity. A very significant coupling phenomenon of LSPRs occurs between adjacent unit cells, corresponding to the red vertical stripes on both sides in Figure 5a,c. For the bottom metasurface, it also has the LSPRs effect, but the locations of its electric field hotspots are different from the uniformly distributed situation in Figure 4b,d. It shows a significant anisotropic feature, mainly concentrated at the positions of “Region 1” and “Region 2”. The mutual coupling of LSPRs between the bottom and top metasurfaces may have a crucial impact on the extinction effect of RCP. Moreover, the current distribution map also contains a lot of physical information. At the wavelength of valley b, the equivalent current of the top metasurface is mainly distributed in the XZ plane. It presents as a pair of oppositely oscillating electric dipoles along the X-axis and two coherently oscillating electric dipoles along the Z-axis (Figure 5a). This current configuration indicates that the top metasurface supports a complex multipole hybrid mode at this wavelength. Meanwhile, the equivalent current of the bottom metasurface shows two semi-ring distributions with opposite rotations (Figure 5b), which is a typical feature of the vertical electric dipole. At the wavelength of valley d, the electromagnetic response mode changes significantly. The equivalent current of the top metasurface simplifies to a single-oriented electric dipole (Figure 5c), while the bottom metasurface presents as an electric dipole at an approximately 40° angle to the X-axis (Figure 5d). Therefore, the formation of valley b and valley d can be attributed to the destructive interference between specific multipole modes in the upper and lower metasurface, thereby significantly suppressing the transmission of LCP. Figure S1 in the Supplementary Materials also analyzes the optical resonance phenomenon from the perspective of the two-dimensional CD spectrum of the period and the wavelength.

3.3. Error Analysis in Chiral Mirrors

To ensure the reliable convergence of optical simulation, we systematically investigated the influence of material parameters and model configuration on the CD spectrum. Figure 6a assesses the impact of the optical refractive index parameters of gold, with four sets of dispersion data sourced from different literature [35,36,37,38]. The results indicate that minor perturbations in material parameters only cause weak changes in the CD peak and do not alter the fundamental characteristics of the chiral optical response, confirming the robustness of the designed structure to material parameter tolerances. Figure 6b examines the effect of the thickness variation Δh2 of the dielectric layer between the upper and lower metasurfaces, which can result from thermal expansion due to temperature fluctuations or polishing errors during fabrication. The analysis reveals that a shift in the dielectric layer thickness does not lead to the disappearance of the circular dichroism phenomenon but causes a slight wavelength drift of the characteristic peak positions. Figure 6c investigates the influence of the grid size of the gold material. In this model, the default grid size is set to 30 nm. When the grid size varies within the range of 20 nm to 45 nm, the circular dichroism spectra remain stable, indicating that the calculation results have converged to a grid-independent solution. Figure 6d verifies the impact of the thickness of the perfectly matched layer on the simulation accuracy. The results show that when the thickness of the perfectly matched layer exceeds 2500 nm, the circular dichroism values no longer change with it, demonstrating that the current model’s setting of 4000 nm for the perfectly matched layer is sufficient to eliminate numerical errors caused by boundary reflections. In addition, the analyses of the lateral dislocation and etching tilt angle can all be found in the Figure S2 of the Supplementary Materials.

4. Discussion

The twisted bilayer plasmonic metasurface proposed in this paper introduces the interlayer twist angle as a new degree of freedom, providing a design concept different from traditional techniques for achieving efficient circular dichroism. Unlike the intrinsic chiral schemes that rely on three-dimensional helical structures, this work adopts a stacked structure of two non-chiral anisotropic metasurfaces, inducing chiral optical responses by breaking the mirror symmetry of the overall structure. This design not only avoids the complex three-dimensional micro-nano processing technology but also realizes the tunability of chiral responses through the precise control of the twist angle.

It is worth noting that, compared with the traditional “waveplate–polarizer” cascaded scheme, this design adds the new degree of freedom of twist angle control, achieving high circular dichroism values at two different wavelengths. At a wavelength of 1660 nm, the CD value is 0.48, which is close to the theoretical limit of traditional metal chiral structures. At a wavelength of 2200 nm, the CD value is as high as 0.84, significantly breaking through the 50% efficiency bottleneck. This phenomenon can be attributed to the wavelength dependence of chiral resonance modes in the multilayer structure. The strength of interlayer coupling changes in different bands, thus forming enhanced chiral optical responses at specific wavelengths. This multi-wavelength characteristic makes the device potentially valuable in wavelength division multiplexing systems. Compared with chiral structures based on dielectric metasurfaces, although metal metasurfaces have ohmic losses, by optimizing geometric parameters and twist angles, significant enhancement of chiral responses can be achieved in specific bands, thereby offsetting the negative impact of losses. The simulation results of this work verify the feasibility of this design concept, laying a foundation for subsequent experimental verification.

5. Conclusions

In summary, this study numerically demonstrates an efficient CD device based on a twisted double-layer metal metasurface. By precisely engineering the in-plane twist angle between the two anisotropic metasurface layers, the mirror symmetry of the structure is broken, giving rise to strong coupling and interference effects that significantly enhance the CD response at multiple wavelengths. Simulation results indicate that the device achieves CD values of 0.48 at 1660 nm and 0.84 at 2200 nm, surpassing the 50% efficiency bottleneck inherent to conventional metallic chiral structures. This approach eliminates the need for complex three-dimensional helical geometries as well as traditional “waveplate–polarizer” cascades, offering a simple, fabrication-compatible, and scalable design strategy for high-performance chiral photonic devices. These findings may open up new avenues for applications such as near-infrared high-resolution chiral molecular sensing, dual-wavelength dynamic polarization coding, and low-loss integrated quantum photonics.