Design of an Integrated Circularly Polarized HgCdTe Photodetector Based on Silicon Metasurfaces

Abstract

Compared with conventional detectors, a circularly polarized detector operating at 4.26 μm effectively suppresses background noise (e.g., solar scattering and atmospheric interference), enabling high-precision CO₂ monitoring across ecosystems like farmland, forests, and wetlands. This capability allows the precise quantification of carbon sink potential and ecosystem health. Our design employs a mid-wave HgCdTe detector—a well-established platform—combined with a CMOS-compatible Si/SiO₂ metasurface. Geometric displacements were applied to break C₂ symmetry, achieving a chiral design. Through multiparameter optimization, we realized a circularly polarized photodetector (CPPD) with a CPER of 18 dB, expected to demonstrate superior CO₂ monitoring performance. These advances may offer researchers and practitioners a robust tool for both fundamental studies and field deployments.

1. Introduction

Carbon dioxide (CO₂) [1], a primary greenhouse gas, contributes directly to global warming through its increasing atmospheric concentration [2]. This phenomenon drives climate change impacts including extreme weather events, sea level rise, and ecosystem disruption [3]. Infrared photodetectors operating at 4.26 μm—corresponding to CO₂’s strong absorption peak—provide unparalleled capabilities for high-precision gas detection and environmental monitoring. However, conventional intensity-based detectors suffer significant limitations. They are particularly vulnerable to background noise sources like sunlight scattering and atmospheric interference, which degrade signal-to-noise ratios and compromise measurement accuracy in complex environments [4]. Furthermore, atmospheric turbulence and multiple scattering effects impair signal stability, restricting their effectiveness for long-range monitoring applications [5]. In comparison, circularly polarized infrared detectors [6,7,8] demonstrate superior performance characteristics. They overcome the inherent limitations of conventional detectors by offering enhanced measurement accuracy, improved noise immunity, and greater signal stability—essential attributes for meeting the rigorous demands of modern environmental monitoring and scientific research.

Early implementations of circularly polarized infrared detectors relied on optical systems incorporating quarter-wave plates [9,10] and 45° linear polarizers [11,12]. However, these systems were plagued by bulky form factors and substantial mechanical noise. The emergence of metasurface technology has effectively overcome these limitations. Metasurfaces [13,14,15,16,17] provide numerous advantages including ultrathin and lightweight profiles, high design flexibility, broadband operational capability, planar integration compatibility, multifunctional performance, precise light field manipulation, environmental robustness, and cost-effective fabrication. Recently, based on circular dichroism [18,19,20,21,22] metasurfaces, some excellent circular polarization detectors have been gradually reported. In 2020, Wang et al. [23] developed a chip-scale polarimeter integrating four graphene–silicon photodetectors capable of extracting full Stokes parameters through inter-detector information processing. In 2023, Yang et al. [24] developed a mid-infrared circularly polarized photodetector using asymmetric silicon gratings, achieving 0.7 circular dichroism absorption at λ = 8.163 µm. In 2023, Yao et al. [25] implemented a metasurface-based full-Stokes imaging sensor by integrating a polarization filter array with CMOS technology, enabling single-shot polarization imaging with a 40° field of view and dual-color operation. In 2024, Chen et al. [26] created an on-chip full-Stokes polarimeter using four plasmonic metasurface subpixels, achieving < 1% root mean square error in full polarization reconstruction via machine learning algorithms. In 2025, Zhu et al. [27] demonstrated a detector combining helical nanostructures with perovskite photodiodes, achieving a remarkable photocurrent anisotropy factor of 1.96 at visible wavelengths for quantitative polarization state analysis.

In this work, we investigate the integration of chiral metasurfaces with commercial HgCdTe detectors [28,29] to optimize performance while maintaining fabrication simplicity. A circularly polarized photodetector (CPPD) was developed by vertically stacking a HgCdTe detection layer with a chiral silicon metasurface. The metasurface’s unit cells break chiral symmetry to enable polarization selectivity. Under right-circularly polarized (RCP) illumination, counter-rotating current loops are generated, inducing destructive interference that blocks RCP transmission. Using COMSOL Multiphysics simulations, we quantified the performance sensitivity to fabrication tolerances.

2. Materials and Methods

Figure 1a illustrates the three-dimensional structure of the CPPD, which consists of a HgCdTe detector at the bottom and a silicon metasurface on top. The HgCdTe detector, also referred to as the MCT detector, converts incident photons into electrons collected by electrodes but lacks intrinsic polarization selectivity. The unit cell of the silicon metasurface exhibits a typical chiral symmetric structure, enabling significantly higher transmission for left-circularly polarized (LCP) light compared with RCP. Figure 1b provides the xz cross-sectional view of the CPPD, detailing its structural specifics and material composition. Figure 1c shows the xy cross-sectional view of the silicon metasurface in the CPPD, while Figure 1d presents the mirror image of Figure 1c about the YZ plane, indicating the opposite circular polarization selection characteristics compared with Figure 1a, favoring RCP transmission. The incident light propagates along the negative Z-axis. The absorption properties of optical materials are typically quantified by the optical absorption length, $L_a$ = ${\lambda }_0/(4\times \pi \times \kappa )$, where $L_a$ denotes the distance over which the light intensity decays to 1/e, $\kappa$ represents the imaginary part of the refractive index, and ${\lambda }_0$ is the wavelength in free space. Clearly, smaller values of $\kappa$ correspond to larger $L_a$, implying weaker absorption characteristics. Figure 1e displays the optical refractive indices of SiO₂ and MCT, with real parts of approximately 1.398 and 3.447, respectively. As shown in Figure 1f, the blue circle represents the $L_a$ of SiO₂, which is the actual value reduced by a factor of one million. At the operating wavelength of 4.26 μm, the blue circle overlaps with the red line representing MCT’s absorption characteristics near an $L_a$ of 2.1 μm. Figure 1g outlines the COMSOL model used to simulate the optical properties of the CPPD. Periodic boundary conditions, commonly employed in photonic crystals and antenna arrays, confine the electromagnetic fields in the X and Y directions. Perfectly matched layers (PMLs), utilized to simulate infinite-space electromagnetic wave propagation, are placed at the top and bottom of the model to absorb outgoing waves. The maximum mesh sizes for MCT and silicon are set to 90 nm and 50 nm, respectively, and the PMLs comprise 30 layers to ensure rapid field decay within the layer. The device’s transmission and reflection rates are determined by integrating the Poynting vector over the inner boundaries of the PMLs, and the CPPD’s absorption rate is obtained by subtracting these rates from unity. In addition, the circularly polarized MCT can be prepared through the following steps: After the processing of the HgCdTe device is completed, the surface is cleaned first. Then, a layer of SiO₂ is grown on the ZnS passivation layer by PECVD. Subsequently, a silicon layer is deposited on the SiO₂ surface using low-temperature sputtering or PECVD. The passivation layer and low-temperature conditions can effectively prevent thermal damage to the MCT detector during the process. After that, the patterning of the metasurface is completed through electron beam lithography and inductively coupled plasma etching. Selective etching gas and etching endpoint detection technology can effectively avoid over-etching problems. Throughout the process, due to the poor thermal stability of HgCdTe, all processes are preferably completed at low temperatures (<200 °C), with priority given to ALD, low-temperature PECVD, and sputtering. Thermal treatment is also conducted using local heating (such as laser annealing) instead of global thermal treatment.

3. Results

Figure 2a illustrates the absorption spectrum and CPER spectrum of CPPD in the 4.0–4.5 μm wavelength range. Here, CPER is defined as $10\times \log \left(A_{LCP}/A_{RCP}\right)$, where $A_{LCP}$ and $A_{RCP}$ represent the absorption rates of CPPD for LCP and RCP, respectively. For LCP, the absorption rate initially increases with wavelength and then decreases, reaching a peak value of 0.82 at 4.27 μm. For RCP, the minimum absorption rate occurs at the operating wavelength of 4.26 μm, indicating that CPPD does not respond to RCP. The inset in Figure 2a depicts the 2D structural diagram of a conventional MCT detector. Since no filter is present on its top, it inherently lacks sensitivity to the polarization information of incident light. The light blue dots and dark blue solid line represent the absorption rates of the 2D MCT chip for RCP and LCP, respectively. These two curves exhibit complete overlap as expected. The red five-pointed star represents the CPER spectrum of CPPD, with a maximum value of 18 dB corresponding to the minimum absorption rate of RCP. Furthermore, it is observed that at the operating wavelength of 4.26 μm, the absorption rate of CPPD for LCP is lower than that of the 2D MCT chip. This suggests that while the silicon chiral metasurface enhances the ability to detect circularly polarized signals in spatial light, this advantage comes at the cost of reduced device efficiency. Therefore, this high-performance CPPD may be more suitable for applications requiring sensitivity to circularly polarized signals, such as quantum communication and quantum computing, atmospheric detection, and display technology and augmented reality. Given that the maximum CD value and the maximum CPER do not coincide at the same operating wavelength in Figure 2a, a more comprehensive performance metric for evaluating CPPD should be introduced. CPQF is a new metric resulting from the combined effects of CD and CPER, expressed as CPQF = CD × CPER, where CD = $A_{LCP}$ − $A_{RCP}$. Figure 2b shows the CPQF spectrum of CPPD, with its maximum value occurring at an operating wavelength of 4.26 μm, providing evidence of its potential application in CO₂ monitoring.

3.1. Investigation into the Origin of Chirality in Silicon Metasurfaces

In fact, the circular dichroism of CPPD can be attributed to the chirality exhibited by the silicon metasurface in the transmission space. For LCP, the silicon metasurface allows some photons to pass through the SiO₂ layer and enter the MCT device, where they are converted into photoelectrons that can subsequently be collected by the electrodes. In contrast, for RCP, the silicon metasurface behaves as a near-perfect reflector, with almost no photons transmitted into the region below. Typically, the chiral symmetry of the metasurface’s unit cell determines its CD characteristics. Here, chiral symmetry refers to a geometric structure that cannot be fully superimposed onto its mirror image via any combination of rotation or translation operations. Figure 3a illustrates the chiral evolution of the metasurface unit cell. The basic structure consists of two identical rectangles, with the upper rectangle highlighted by an orange border for clarity. As the upper rectangle shifts to the right, the unit cell transitions from a C₂ symmetric configuration to a chiral symmetric one. Figure 3b is primarily used to quantitatively evaluate the impact of the dislocation length Δx on the CD of the silicon metasurface. Specifically, $CD=T_{LCP}-T_{RCP}$, where $T_{LCP}$ and $T_{RCP}$ denote the transmission rates of the silicon metasurface for LCP and RCP, respectively. When Δx equals 0 nm, corresponding to a C₂ symmetric unit cell, the metasurface lacks chiral symmetry, resulting in zero CD across all wavelengths. As Δx increases, the CD peak exhibits a gradual redshift, with the peak intensity initially increasing and then decreasing. As evident in Figure 3b, there exists a strong correlation between the chiral symmetric structure and the observed CD. Furthermore, the alignment and bonding between the silicon metasurface and MCT typically rely on three processes: indium pillar alignment or glass via technology, dielectric support layers, and organic adhesive bonding. Under these processes, the medium between the silicon metasurface and MCT is generally air, dielectrics (such as silicon dioxide or silicon nitride), or epoxy resin. Figure 3c depicts the variation in CD of the silicon metasurface under different alignment processes, with the highest CD achieved using the dielectric support layer process, where SiO₂ serves as the dielectric material. Figure 3d presents the transmission spectrum and CPER spectrum of the silicon metasurface. Here, CPER is defined as 10 × log ($T_{LCP}$/$T_{RCP}$), where $T_{LCP}$ and $T_{RCP}$ represent the transmission rates of the silicon metasurface for LCP and RCP, respectively. The minimum transmission rate for RCP occurs at an operating wavelength of 4.26 μm, corresponding to a maximum CPER value of 23.5 dB.

For metasurfaces, near-field distribution maps are of critical importance [9,30]. These maps can not only reveal the localization characteristics and coupling effects of electromagnetic fields but also provide insights into the electromagnetic response mechanisms and energy dissipation processes. Figure 4 presents the electric field intensity distribution map and equivalent current vector direction map of the silicon metasurface in the xy cross-section. The labels 1/4, 1/2, and 3/4 above the figure correspond to the Z-axis positions at 1/4, 1/2, and 3/4 of the metasurface height, respectively. The six electric field intensity distribution maps share a common color scale, with the equivalent currents corresponding to LCP and RCP represented by black and red arrows, respectively. The thickness of the arrows indicating the equivalent current in the silicon metasurface is proportional to the current intensity. For LCP, the electric field is concentrated not only along the boundaries of the rectangle but also significantly localized at points 1 and 2. The equivalent currents at the three cross-sections of the silicon metasurface are relatively weak and can be neglected. In contrast, for RCP, the electric field localization is much weaker compared with the LCP case and is primarily confined to the interface between the rectangle and air. The intensity of the equivalent current indicated by the red arrows is substantial and cannot be ignored. The equivalent current rotates counterclockwise within the upper rectangle and clockwise within the lower rectangle. Based on the theory outlined in Reference [28], it is hypothesized that the superposition of the light wave generated by the point-source resonance (indicated by the red arrow) and the right-circularly polarized (RCP) wave at the transmission end results in a typical coherent destructive interference phenomenon in optics. This elucidates why almost no photon energy leaks into the underlying silica material.

3.2. Optimization of the Dielectric Support Layer Thickness (hs) in CPPD

In optics, when two metasurfaces are brought into close proximity or integrated, their interaction can lead to a range of optical phenomena, including Fabry–Perot resonance, mode coupling and hybridization, near-field coupling and energy transfer, as well as non-reciprocal transmission. However, in the case of the CPPD depicted in Figure 1a, the upper silicon metasurface and the lower MCT detector must function as independent optical components. Consequently, it is crucial to investigate how the thickness hs of the intermediate medium layer affects the performance of the CPPD. Figure 5a illustrates the impact of the SiO₂ spacer layer thickness hs on the absorption rate and CPER of the CPPD. As hs increases, the absorption rate of RCP initially decreases sharply before gradually tapering off, remaining relatively constant for hs > 2.7 μm. In contrast, the absorption rate of LCP exhibits irregular peaks and valleys, with maximum and minimum peak values of approximately 0.81 and 0.27, respectively. Furthermore, the CPER also displays multiple peaks, though its largest peak does not align with the design wavelength of 4.26 μm. To determine the optimal hs, CPQF appears to take precedence over absorption rate and CPER. As indicated by the red circle in Figure 5b, the CPPD achieves its highest CPQF peak at hs = 3 μm.

3.3. Analysis of Potential Process Errors

In the preceding chapters, we analyzed the origin of circular dichroism in silicon metasurfaces and the influence of the spacer thickness hs on the performance of CPPD. In this chapter, we primarily employ COMSOL 5.6 software to evaluate the extent of performance degradation in CPPD caused by potential fabrication errors introduced during the manufacturing process. As shown in Figure 6a, factors such as optical diffraction effects, the chemical properties of photoresist, or anisotropy during etching can cause the vertices of the rectangular structure to become rounded corners with a radius of r0. Figure 6b demonstrates that when r0 is less than 100 nm, the absorption rate of RCP remains nearly constant, whereas that of LCP exhibits a gradual increase. Figure 6c,e illustrate the processing errors Δh associated with under-etching and over-etching, which are primarily caused by insufficient or excessive etching times. For the under-etching case, the etching error Δh severely degrades CPPD performance, and when Δh reaches 30 nm, CPPD almost loses its chirality, as shown in Figure 6d. In contrast, for the over-etching case, the impact of Δh on CPPD performance is less detrimental compared with under-etching, as depicted in Figure 6f. As Δh increases, the absorption rate of LCP initially rises and then decreases, with the minimum CD remaining above 0.4. These results indicate that both under-etching and over-etching significantly impair device performance. To mitigate these effects, it is crucial to enhance etching selectivity, refine mask design, and implement real-time monitoring to improve the performance and yield of semiconductor devices. Additionally, due to the inherent challenges of avoiding focusing errors in lithography and issues such as non-uniform etching rates, insufficient selectivity, and lateral etching during actual fabrication, linewidth errors (as shown in Figure 6g) are inevitable. Figure 6h reveals that linewidth deviation Δl critically affects CPPD performance. To ensure Δl remains below 50 nm, stringent process control is required, otherwise, the optimal CPQF peak may exhibit wavelength drift near 4.26 μm. Fortunately, the design material for the metasurface is silicon, which benefits from highly mature fabrication processes (lithography, etching, thin-film deposition, and doping), process optimization and standardization, and global industrial collaboration. Therefore, we believe that the device we designed has a realistic chance of being manufactured with high fidelity.

Tilted incident light can break the apparent symmetry of the metasurface structure, thereby inducing an equivalent chiral optical response. Therefore, monitoring the influence of the incident angle on circular dichroism seems to be an issue that cannot be ignored. The tilted incidence situation can generally be simplified into two cases: the incident light in the XZ plane and in the YZ plane. When the incident angle is α, the typical characteristic wavelength of the PMLs is λ_0/cos(α), where λ_0 is the vacuum free wavelength. For the case where the incident light is in the XZ plane, the incident azimuth angle is 0°. As shown in Figure 7a, as the incident angle increases, the circular dichroism shows a tendency to disappear. However, when α is 20 degrees, the CPQF spectrum splits, and the spectral shape splits from a single sharp peak at α = 0° into multiple wide peaks. As shown in Figure 7b, when the incident plane is the YZ plane, it can be found that although the circular dichroism gradually attenuates when the incident angle is less than 30 degrees, it still plays a role. However, when the incident angle exceeds 30 degrees, the circular dichroism basically disappears in the wavelength band from 4.0 to 4.5 μm. It can be seen that the compact circular polarization detector we designed has a relatively high collimation requirement for the incident angle, which needs to be paid extra attention in practical application scenarios.

4. Discussion

Circularly polarized light demonstrates superior atmospheric stability compared with unpolarized or linearly polarized light due to its unique interactions with atmospheric components. We developed a 4.26 μm circularly polarized detector via COMSOL Multiphysics simulations, utilizing a chiral silicon metasurface for polarization discrimination. This technology enables precise real-time CO₂ monitoring, generating vital datasets for climate research and supporting evidence-based emission policy development.

5. Conclusions

In summary, we have designed a compact photodetector capable of directly recognizing circularly polarized light. By etching a chiral-symmetric amorphous silicon metasurface on top of the MCT detector, we achieved a CPER as high as 18 dB. This circular polarization detector has the advantages of a simple process and easy commercialization, and is likely to become a design prototype for large-scale focal plane polarization cameras. Future work should address the trade-off between optical crosstalk and array scale under small pixel pitch conditions.