Methodology
Figure 1a presents the structure of our CPL photodetector. The basic unit of the device comprises a pair of electrodes positioned on Te nanosheets, with a series of etched V-shaped grooves symmetrically situated between them. Array devices are dual-port devices composed of multiple basic units, which will be discussed later. Unless otherwise stated, all measurements in this study were conducted under zero bias.
The device operates through chirality-resolved near-field modes in achiral structures, leveraging a phenomenon known as hidden chirality. Contrary to the conventional belief that chiral geometries are required for CPL discrimination, recent advancements in plasmonic metasurfaces have demonstrated that achiral structures can exhibit different near-field modes under LCP and RCP light, despite their identical far-field responses24,36. This concept has been extended to dielectric nanostructures, as illustrated in Fig. 1b and 2b. Specifically, in carefully designed V-grooves, exposure to LCP or RCP light results in light localization at the right or left arms, respectively, leading to photoresponses of equal magnitude but opposite directions. This chiral-dependent mode induces differential temperature increases on either side of the V-grooves, depicted in Fig. 1c. By integrating electrodes on both sides of the V-groove and employing the PTE effect, this temperature difference generates photovoltage in opposing directions. The broadband nature of these near-field modes endows the device with broadband response capabilities. Such mode arises from the interference of x-polarized (LP x) and y-polarized (LP y) light excitation modes, depicted in Fig. 2b. When LP x and LP y light excite the structure, the resulting near-field modes distribute symmetrically, resulting in zero net photovoltage. Under CPL light, these two modes are mixed with different phases, enabling chirality-dependent near-field patterns.
The CPL detection scheme discussed is not limited to specific materials, optoelectronic mechanisms, or geometry structures, thus offering great versatility. In this study, Te nanosheets were employed as the photothermoelectric material due to their straightforward synthesis process, high refractive index, and superior thermoelectric properties37,38. Otherwise, any material with a sufficient refractive index to support Mie resonance modes and capable of effectively converting light into a photoelectric response can facilitate CPL detection through achiral structures. Apart from PTE, any mechanism capable of inducing a photoelectric response at the electrode-material interface is applicable, such as the photovoltaic effect widely existing in the semiconductor-metal and semiconductor-semiconductor interface. For instance, achiral silicon structures under CPL excitation can exhibit chirality-sensitive modes that asymmetrically illuminate the electrodes, causing an uneven distribution of photogenerated carriers on two sides. These carriers are then separated by the built-in field of the Schottky junction, resulting in a photoresponse directed by the light chirality. Furthermore, the utilization of V-shaped geometric structures is not the sole alternative. Any geometry that maintains a singular symmetry plane, such as T-shaped grooves (see Supplementary Figure S1), could fulfill a similar function. The flexibility of this achiral-geometry-based CPL detection method opens up a broad spectrum of potential applications.
Design of the achiral nanostructure
The geometric parameters of the nanostructure were optimized through numerical simulations. Near field chirality, defined as the difference between the mean light intensities on the structure's left and right sides under CPL excitation, serves as a metric to assess the discrimination capacity of geometric structures. The optimized geometric parameters of the V-groove for 520 nm wavelength are arm length l = 300 nm, arm width w = 100 nm, and the included angle θ = 70°, as shown in Fig. 2d and Supplementary Figure S2. Additionally, we investigated the influence of Te nanosheet thickness and groove depth on near-field chirality. According to the calculations, increasing groove depth enhances chiral resolution, necessitating thicker nanosheets.
We further conducted a mode analysis of the V-shaped groove. Figure 2a illustrates the resonance modes of the optimized V-shaped groove when subjected to various polarized lights. As expected, the achiral nanostructure shows an identical scattering cross-section under the LCP and RCP light. Typically, the wavelength shift observed in resonance peaks under LP x and LP y light indicates near-field chirality, which is mutually confirmed with the aforementioned parameter optimization results. The near-field modes of the V-shaped groove under various wavelengths and polarizations are shown in Supplementary Figure S3, 4. Under the CPL excitation, the light field focuses on one arm of the V-shaped groove, showing highly asymmetric, whereas linearly polarized light yields symmetrical near-field modes. Under LP x light, the light field distributes evenly on both arms of the V-shape. With LP y light excitation, the field localizes at the V-shaped groove's apex. Such mode feature spans the entire visible and near-infrared band, endowing the device with a theoretical broadband response capability (Fig. 2c and Supplementary Figure S3). In structures with weak chiral resolution (l = 500 nm, w = 100 nm), negligible wavelength discrepancy exists between the resonance peaks for LP x and LP y light, suggesting an absence of chirality-dependent mode interference (Supplementary Figure S5).
To verify the numerical calculations, PEEM was employed for near-field measurement. The operational principle of PEEM is depicted in Fig. 2e. The light interacts with the sample, generating free electrons via the two-photon photoelectric effect. These electrons are then imaged onto a CCD after passing through a series of electromagnetic lenses, enabling high spatial resolution detection of near-field modes. The near-field modes corresponding to various polarizations are illustrated in Fig. 2g-j, aligned closely with simulations. Figure 2f presents a scanning electron microscope (SEM) image of the corresponding area, showcasing V-shaped groove arrays with depths incrementing from right to left. As expected, deeper grooves correlate with heightened asymmetry in the near-field mode under CPL excitation. Notably, a 410 nm laser beam was employed in the PEEM measurements to fulfill the necessary condition for electrons to surpass the vacuum energy level.
CPL-sensitive photovoltage in unit devices
An SEM image of our device is shown in Fig. 3a. Te nanosheets were synthesized by hydrothermal method39 and dropped to SiO2-Si substrate. The electrodes were fabricated through electron beam lithography, followed by focused ion beam (FIB) etching to create V-shaped grooves onto the Te nanosheets. The electrodes were oriented perpendicular to the Te nanosheets’ long axis, corresponding to the direction of Te atomic chains, as confirmed by angle-resolved Raman spectroscopy (Supplementary Figure S6). Laser beams with wavelengths of 405 nm, 520 nm, and 638 nm were directed through a polarizer and an achromatic quarter-wave plate (QWP), then concentrated onto the devices with focused spot diameters around 1 µm. Devices were mounted on a piezo displacement stage. Scanning this stage enables photocurrent mapping, as shown in Fig. 3b. The photoresponse is believed to originate from the PTE effect, which is indicated by the linear power-dependence curve. Electrical tests confirm the absence of Schottky junctions between the electrodes and Te nanosheets (Supplementary Figure S7), ruling out the possibility of a built-in electric field induced photocurrent, consistent with previous literature37. When the light spot position moves perpendicular to the electrodes, it induces photoresponses in opposite directions on each side. With the light spot positioned between the electrodes—where the y-polarization response turns to zero (see Fig. 3c)—the device obtains an equilibrium state and exhibits photovoltage directed by the light field's chirality. During this phase, both linearly polarized and unpolarized light induce counteracting photoresponses on the electrodes, allowing the CPL-sensitive photovoltage to dominate. The light polarization state can be continuously modulated by rotating the QWP, facilitating polarization-dependent photovoltage measurements. The photovoltage curves varying with the QWP rotation angle φ in the equilibrium state are shown in Fig. 3d-f. The top panels show the outcomes before FIB etching, while the bottom panels show the results after etching V-grooves.
For the three testing wavelengths, the devices demonstrate photovoltage in contrasting directions under LCP and RCP light, differing from unetched devices' responses. A high-discrimination, broadband, calibration-free CPL photodetector has been realized. Notably, despite the intrinsic chirality of one-dimensional Te atomic chains, which are expected to cause varying absorption for different CPL40. The interaction between chiral light and these atomic structures is extremely weak in the visible band. Our experiments showed that unpatterned devices failed to demonstrate any appreciable circular polarization response. The device has a CPL responsivity of 0.37 V/W at 520 nm and a DR of 4.1, significantly surpassing the typical DR of reported visible CPL detectors.
Although the CPL-sensitive component, VC, dominates the photoresponse at the equilibrium position, the overall photovoltage includes a polarization-insensitive term, V0, and a linear polarization-sensitive term, VL. The whole photoresponse is expressed as Vph= V0 + VL cos(2φ + ψ1) + VC cos(4φ + ψ2), which accurately describes the experimental observations. ψ1 and ψ2 are phase constants. V0 and VL cause a discrepancy in the photovoltage amplitudes induced by LCP and RCP light, yielding a limited DR. Additionally, the polarization-dependent curves reveal that photovoltage transition points do not fully align with chirality transition points, indicating that the sign of the photovoltage cannot completely reflect the chirality of the elliptically polarized light, as shown in Fig. 3e. To further suppress the V0 and VL terms, their origins were carefully analyzed.
Notably, the polarization-insensitive component of photovoltage emerges from the incomplete elimination of the asymmetrical response generated by the incident light at the electrodes. Theoretically, a centrally positioned light spot should result in equal light intensity at both electrodes for a device without V-grooves. This leads to photovoltage of identical magnitude but opposite directions, achieving an equilibrium state with zero net photovoltage. However, in practice, disparities in electrode-nanosheet interfaces, fabrication imperfections, and alignment inaccuracies between the spot and the device prevent the total elimination of the net photovoltage (Supplementary Figure S8). Additionally, alignment inaccuracies during V-groove fabrication may offset the structure from the device's central axis, causing LCP and RCP photoexcitation to impact the electrodes unevenly. These factors jointly contribute to the formation of the polarization-insensitive component of photoresponse.
The linear polarization sensitive term VL has more complex origins, mainly derives from the modulation of baseline voltage V0 induced by the differential absorption of variously polarized lights. Two competing mechanisms play a role in this absorption modulation: scattering at the gold electrode edges and the anisotropic absorption of Te material. Te nanosheets, composed of one-dimensional atomic chains bonded by van der Waals forces, exhibit strong anisotropy. Typically, light polarized perpendicular to the atomic chains, that is, LP y light, has a higher absorption coefficient41. On the contrary, the scattering effect at the electrode edges favors LP x light, facilitating more efficient coupling of light polarized perpendicular to the electrodes into the device. Generally, the scattering effect of the electrodes is the dominant factor, offsetting the absorption differences in Te, resulting in smaller photovoltages under LP y light excitation, as shown in Fig. 3d-f and Supplementary Figure S9. Moreover, V-grooves demonstrate uneven near-field distributions when exposed to obliquely polarized light, which influences the VL term. However, it is not the primary contributor (Supplementary Figure S10).
Performance optimization with array devices
Based on the above analysis, array devices were designed and fabricated to further reduce the V0 and VL terms, achieving an impressive discrimination ratio across the visible band. This innovative design reduces the necessity for precise alignment of the light spot, enhancing the device's tolerance to spot position variations. The V0 component of photovoltage arises from electrode asymmetries due to variations in contact properties, fabrication imperfections, and other factors. While VL mainly originates from V0 modulation via alternate absorption. Even with uniform illumination, the electrodes on either side of a unit device may respond differently, leading to a net photovoltage, as detailed in Supplementary Figure S8. This random asymmetry can be neutralized by employing multiple electrode pairs, reducing the polarization-insensitive photovoltages and the corresponding linear polarization-dependent terms. The refined device structure, along with its SEM images, is presented in Fig. 4a, b. The device incorporates interdigitated electrodes flanked by V-shaped grooves, with adjacent rows of V-grooves oriented in opposite directions. This configuration guarantees that one side of the interdigitated electrodes consistently experiences a higher near-field mode heating under CPL. The experiments employed a broad, uniform light spot to cover the entire device, coupled with the rotation of QWP to capture the polarization-dependent photovoltage curves. Results at the wavelengths of 405 nm, 520 nm, and 638 nm are shown in Fig. 4c-e. Compared to unit devices under a focused light spot in an equilibrium state, the optimized devices exhibited suppressed polarization-insensitive and linear terms, achieving a maximum DR of 107 at 405 nm wavelength. Furthermore, satisfactory DR values were also obtained at 520 nm (20) and 638 nm (32). The CPL-sensitive photovoltage dominates the responses, and high-precision CPL distinguish is achieved. Figure 4f, g displays the normalized spectral distribution of the photovoltage under a uniform light field for unit and array devices, respectively. The components at angular frequencies 0, 2/π, and 4/π represent V0, VC and VL terms. In the unit device, the random asymmetry-induced polarization-insensitive term completely masks the CPL-sensitive photovoltage, leading to mere magnitude variations without directional distinction under opposite CPL illumination (Supplementary Figure S11). To attain a notable discrimination ratio, it is necessary to employ a focused light spot and align it to the equilibrium state position, thereby reducing the V0 term and emphasizing the Vc term. Conversely, in the array device, the CPL-sensitive term dominates under uniform excitation, substantially reducing spatial alignment demands. It is believed that further increases in the number of interdigitated fingers can lead to additional suppression of the V0 and VL terms in the array devices, resulting in enhanced DR. Therefore, we have successfully demonstrated a high DR, broadband CPL photodetector using all-dielectric nanostructures. Array devices, with further improved DR and reduced alignment requirements, have emerged as promising candidates for on-chip CPL detection.
The array device also exhibits exceptional noise levels and response speed. It showcases a rise time of 290 µs and a fall time of 600 µs, yielding an estimated − 3dB bandwidth of 688 Hz (Supplementary Figure S12). Previous reports suggest that the photothermoelectric effect in Te can achieve much faster response speeds38. The parasitic capacitance within the electrodes and the electrical measurement systems are to blame for the reduced response speed, which can be significantly enhanced through design optimization. Operating without bias voltage, the device demonstrates a remarkably low noise level, reaching down to 28 nV Hz− 1/2 around its cut-off frequency, with a noise-equivalent power (NEP) of merely 0.60 µW Hz− 1/2 (Supplementary Figure S13). In addition, our CPL photodetectors demonstrate remarkable atmospheric stability, with their performance remaining virtually unchanged after a 3-month exposure.