1. Introduction

Polarization imaging can simultaneously acquire the spatial distribution and polarization state of a target scene and obtain information regarding the shapes and textures of the reflective surfaces, target contours, or optical activity of various materials [1–3]. In addition, the polarization information is well maintained in harsh environments and is not sensitive to environmental factors. Therefore, polarization imaging is widely used in target detection [4], the life sciences [5,6], underwater detection [7,8], remote sensing and communication [9,10], and other fields. Polarization splitting and filtering devices are the core components of polarization imaging systems. However, conventional devices are bulky, poorly integrated, susceptible to interference, and unable to acquire both linear and circular polarization states [11]. This leads to polarization imaging systems with complex structures, large image alignment errors, and the inability to obtain the full-Stokes vector parameters, restricting the development and application of polarization imaging. Metasurfaces are two-dimensional arrays of subwavelength scatterers or nanocells that can arbitrarily change the phase, polarization state, and intensity of electromagnetic waves on the subwavelength scale [12,13]. Dielectric metasurfaces avoid absorption losses and are compatible with standard micromanufacturing technologies; thus, they have attracted widespread attention and research [14]. Polarization devices based on dielectric metasurfaces are small, lightweight, highly efficient, and easy to integrate [15], providing new possibilities for the realization of highly integrated full-Stokes polarization imaging systems.

Metasurfaces have been designed for full-Stokes polarization imaging. Some literature has proposed polarization beam splitters capable of simultaneously measuring the full Stokes component of incident waves [11,16–18]. However, those devices require the combination of lenses to achieve polarization imaging, making the system more complex. Besides, most of them are reflective devices, which are not suitable for imaging applications. In the literature [1,19,20,21], polarization-sensitive metalens arrays that allow real-time polarization imaging were investigated, which can split and focus 0° linearly polarized light, 90° linearly polarized light, 45° linearly polarized light, 135° linearly polarized light, left-handed circularly polarized light, and right-handed circularly polarized light to different focal spots. However, the aforementioned metalens arrays can generally only be used for monochromatic detection. For measuring multiple objects with similar polarization states under complex conditions, although monochromatic detection can achieve the measurement of the polarization contrast of the target scene, the gray value has only one degree of freedom, which significantly limits the amount of measured information of the object; thus, the complete information of the polarization state of the target cannot be acquired. Broad-spectrum polarization color imaging can provide more image degrees of freedom and more accurately characterize the polarization information of multitarget object scenes in complex environments. Therefore, it is important to design metalenses that can realize broadband achromatic polarization imaging. There is literature on the design of achromatic polarization-sensitive metalenses, but the research is relatively limited. In addition, the design bands are generally in the infrared [21,22] or terahertz bands [23], or the research goal is not to achieve full-Stokes polarization imaging [22–24]. Furthermore, in most studies, the period and height of the nano-units were directly selected, mainly owing to the limitations of the current metasurface fabrication techniques [25,26]. However, the height and aspect ratio of the fabricated nano-units have been improved in recent years [27,28]. Therefore, studies on the effects of the height and period of the nano-units composing the metalenses on their focusing efficiency are starting to become necessary.

Considering the above background, the objective of this study is to design a metalens array that can achieve full-Stokes vector measurements in the broadband visible range (460–680 nm). Titanium dioxide (TiO₂) with a transparent window in the visible spectral range is selected as the design material. To design metalenses that can achieve polarization splitting and efficient focusing, the height and period of the nano-units, which affect the focusing efficiency, are simulated and optimized in this study. The designed metalens array can be used for polarization imaging systems such as multispectral polarization imaging and multispectral polarization light field imaging systems.

2. Theory

The metalenses explored in this study is realized using elliptical TiO₂ nano-units exhibiting birefringence (as shown in Fig. 1(a)), with period p, height h, and radii d_x in the x-axis direction and d_y in the y-axis direction. The phase change of each nano-unit to TM polarized incident light is φ_TM, and that to TE polarized incident light is φ_TE. According to Huygens’ principle, the scattering antenna at the interface is equivalent to several secondary wave sources, and the electromagnetic radiation of the scattering antenna excited by the incident light comprises the secondary waves generated by these secondary wave sources. Because nano-units with different size parameters can produce different phase transitions, the envelope surface formed by secondary waves with different time delays is no longer a plane. Therefore, the desired nonplanar wavefront can be realized at the interface by controlling the characteristic parameters of the nano-units (as shown in Fig. 1(b)). The spherical wavefront can realize aberration-free focusing (as shown in Fig. 1(c)); thus, the radius of the secondary wave at position r = (x,y) can be expressed as $R(x,y) = \sqrt {{{(x - {x_0})}^2} + {{(y - {y_0})}^2} + {f^2}} - f$, where f represents the focal length and r₀ = (x₀,y₀) represents the designated focal spot. The phase delay provided by the scattering antenna at r = (x,y) should satisfy $\varphi (x,y) ={-} kR(x,y) ={-} \frac{{2\pi {n_b}}}{\lambda }(\sqrt {{{(x - {x_\textrm{0}})}^2} + {{(y - {y_\textrm{0}})}^2} + {f^2}} - f)$, where λ represents the wavelength of the incident light and n_b represents the refractive index of the background medium (the medium is a vacuum in this study, i.e., n_b = 1). The above equation also indicates that the condition that the phase delay should meet varies with respect to the wavelength. However, the target device is achromatic. Therefore, we need to find a nano-unit with appropriate structural parameters at each r = (x,y) position, whose phase response can satisfy the phase delay required in all design bands. Khorasaninejad et al. [29] proposed the addition of a constant phase factor c(λ_i) to the ideal phase. This phase can be considered as a type of adjusting wrench, which can help determine the structural parameters of the nano-unit at each r = (x,y) position more effectively. Therefore, to achieve achromatic polarization splitting and focusing, the phase provided by the nano-units at each position of the metalenses to the differently polarized incident light should meet the following requirements:

 

Fig. 1. Basic structure of the nano-unit, wavefront control by the phase gradient interface and the phase profile for metalenses based on Huygens’ principle. (a) basic structure of the nano-unit; (b) wavefront control by the phase gradient interface; (c) phase profile for metalenses based on Huygens’ principle.

Download Full Size | PDF

In the case of linearly polarized light, the phase provided by the nano-unit to the incident light is provided by the transmission phase only; that is, the phase that the nano-unit provides to the TM-/TE-polarized incident light must satisfy Eq. (1). In the case of circularly polarized light, the phase provided by the nano-unit to the incident light is provided by both the transmission phase and the Pancharatnam–Berry phase (geometric phase). Therefore, the phase that the nano-unit must provide to the TM-/TE-polarized incident light must satisfy Eq. (3), and the rotation angle of the nano-unit must satisfy Eq. (4) (Supplement 1).

However, as the rotation angle of the nano-unit is fixed at each position, the geometric phase can only be used at a certain wavelength. To achieve an achromatic aberration, the transmission phase provided by the nano-unit must compensate for the phase difference caused by the geometric phase ($\Delta {\varphi _{\textrm{shift}}}^{need}(x,y,{\lambda _i})$); i.e., the transmission phase that the nano-unit must provide to the TM-/TE-polarized incident light must satisfy Eq. (5), and the rotation angles of the nanopillars must satisfy Eq. (6).

However, from Eq. (5), if the transmission phase provided by the nano-unit to the TM-/TE-polarized incident light can compensate for the phase difference caused by the geometric phase, the transmission phase provided by the nano-unit to the TM-/TE-polarized incident light cannot meet the basic requirement for establishing the geometric phase (${\varphi _{\textrm{TM}}}^{need}(x,y,{\lambda _i}) - {\varphi _{\textrm{TE}}}^{need}(x,y,{\lambda _i}) = \pi$). Therefore, the transmission phase provided by the nano-units for TM-/TE-polarized incident light cannot compensate for the phase difference caused by the geometric phase while satisfying the basic requirement for establishing the geometric phase. Here, we propose that the condition for establishing the geometric phase is a necessary condition. Therefore, we set the transmission phase provided by the nano-unit to the TM-polarized incident light to compensate for the phase difference caused by the geometric phase; then, the transmission phase provided by the nano-unit to the TE-polarized incident light is the above phase plus π. Equation (5) is replaced with Eq. (7). This design method leads to a deviation between the actual focus position and the designated position under right-handed circularly polarized light incidence. To mitigate this phenomenon, in contrast to previous studies, the intermediate wavelength (rather than the maximum wavelength) of the designed band is used to determine the geometric phase, that is, λ_base = λ_intermediate. This processing method can reduce the standard deviation of the offset between the actual focus position and the designated position when broadband right-handed circularly polarized light is incident. Reducing the standard deviation of the offset is beneficial for calibrating and processing images obtained using the full-Stokes polarization imaging system.

When the nano-unit period and height are selected, the process of determining the additional constant phase required for different wavelengths c(λ_i)/$c^{\prime}({\lambda _i})$ and the size parameters of the nano-unit at each position is as follows:

3. Results and discussion

To design metalenses that can achieve polarization splitting and focusing with high focusing efficiency, we simulate and optimize the height and period of the nano-units. Because the phase equations that the nano-unit must provide for TM-/TE-polarized light under orthogonal linearly polarized light incidence differ from those under orthogonal circularly polarized light incidence, we determine the height and period of the nano-units that form a metalens array capable of detecting the full-Stokes vector by optimizing the height and period of the nano-units under orthogonal linearly polarized light incidence. Three discrete wavelengths (λ1 = 460 nm, λ2 = 570 nm, and λ3 = 680 nm) are selected for the optimized design. The bandwidth is discretized into three wavelengths to simplify the operation. This is because it was found through simulations that the focusing of the metalenses is almost constant when the design wavelength interval is small. In addition, ΔG is used as an index to evaluate the total focusing efficiency η (η = η_TM,λ1+η_TM,λ2+η_TM,λ3+η_TE,λ1+η_TE,λ2+η_TE,λ3) of the designed achromatic polarization sensitive metalenses, in addition to the simulation results for the total focusing efficiency of the designed metalenses based on the finite-difference time-domain (FDTD) method. Thus, we can investigate the effects of the height and period of the nano-unit on the focusing efficiency of achromatic polarization multiplexing metalenses in a shorter time.

The metalenses to be designed contains 41 × 41 nano-units, with a numerical aperture (NA) of 0.2. Here, $\textrm{NA} = {n_b} \cdot \sin \left[ {\frac{1}{2}\left( {\arctan (\frac{{R + {r_\textrm{0}}}}{f}) + \arctan (\frac{{R - {r_\textrm{0}}}}{f})} \right)} \right]$, where R represents the radius of the metalenses, r₀ represents the distance from the focal spot to the central axis of the metalenses, and f represents the focal length. The focal spot position is set as (${\pm}$ 2 µm, 0 µm) under the incidence of a 460–680 nm TM/TE wave. The specific optimization process is as follows:

First, we use FDTD Solutions to simulate the responses of nano-units of different sizes to different incident light. The nano-unit structure is elliptical TiO₂ (the refractive index is set to a fixed value of 2.35), and the substrate is silicon dioxide (SiO₂) (the refractive index n_sub follow the Palik model included in the FDTD software). Three selected wavelengths are 0.46, 0.57, and 0.68 µm. Periodic boundary conditions are set in the x- and y-directions, and perfectly matched layer boundary conditions are set in the z-direction. The incident light is a TM-/TE-polarized plane wave perpendicular to the z-axis. By scanning and calculating the phase control of nano-units with heights of h = 600, 800, 1000, 1200, and 1400 nm; periods of p = 200, 250, 300, 350, and 400 nm; and ellipse radii of $0.05 \times p \le {d_x},{d_y} \le 0.46 \times p$ to different wavelengths of TM-/TE-polarized incident light, as well as the transmission efficiency of the different wavelengths of TM-/TE-polarized incident light passing through the nano-units with different sizes, we can obtain 25 groups of small datasets. According to the process described in Section 2, we can obtain the optimized c(λ_i) and ΔG for each dataset (Table 1, Fig. 2).

 

Fig. 2. ΔG optimized from each dataset (note: when the period is different, the focal length needs to be adjusted for the fixed NA).

Download Full Size | PDF

Table 1. c(λ_i) and ΔG Optimized from 25 Groups of Small Datasets

View Table

As shown in Fig. 2, when the period is 200 and 250 nm, a larger height of the nano-unit corresponded to a smaller ΔG and a higher total focusing efficiency of the metalenses η. When the period is 300 nm, the total focusing efficiency of the metalenses η generally increases with the height. When the period is 200, 250, and 300 nm and the height is ≤1000 nm, the total focusing efficiency of the metalenses η changes rapidly with the change in height, whereas the change rate is reduced when the height is >1000 nm. This is mainly because the optical response range provided by the nano-unit with a height of 1000 nm for different polarized incident light of different wavelengths is similar to that provided by the nano-unit with a height of >1000 nm (Supplement 1). When the period is 350 nm or 400 nm, the influence of the height of the nano-unit on the total efficiency of the metalenses is insignificant. This may be because when the period of the nano-unit is close to the incident wavelength, the multi-order diffraction caused by the periodicity will increase the reflection efficiency and reduce the transmission efficiency (Supplement 1). However, the wavelength range where multi-order diffraction is generated by nano-units of different heights is different, which will lead to irregular changes in the transmission efficiency when incident light of the three selected wavelengths (460, 570, and 680 nm) passes through nano-units of different heights, finally causing the total focusing efficiency to change irregularly with respect to the height of the nano-units. Therefore, according to the design requirements and settings of this study, we select the nano-unit height to be 1000 nm. Additionally, the period of 350 nm and 400 nm is not considered in the design of this study. Figure 2 shows that when the height of the nano-unit is 1000 nm, the total focusing efficiency η is the highest when the period is 250 nm. However, when the period of the nano-unit is 200, 250, and 300 nm, ΔG is similar and is small. Considering the limitations of the manufacturing process, we choose the period of the nano-unit to be 300 nm. In the design of multifunctional metalenses, shorter periods and larger heights do not result in high-performance devices. Instead, the parameters should be optimized during the design process to obtain the size parameters of the nano-units that yield the expected responses.

The electric-field intensity distribution at z = 30 µm of the metalense that can split and focus orthogonal linearly polarized light (designed according to Section 2), which consists of nano-units with a height of 1000 nm and a period of 300 nm under the incidence of 460-, 570-, and 680-nm TM-/TE-polarized light, is presented in Fig. 3. As shown, the 460-, 570-, and 680-nm TM-/TE-polarized incident light converge to the designated position with a slight error. This is mainly because the nano-units in the datasets do not fully correspond to the optical requirements. Additionally, electromagnetic coupling occurs between different nano-units, which cannot be considered in the forward design process.

 

Fig. 3. Electric-field intensity distributions at z = 30 µm of NA ≈ 0.2 metalenses consisting of nano-units with a height of 1000 nm and a period of 300 nm under the incidence of 460-, 570-, and 680-nm (a–c) TE-polarized light and (d–f) TM-polarized light.

Download Full Size | PDF

The electric-field intensity distribution at z = 30 µm of the metalense that can split and focus orthogonal circularly polarized light (designed according to Section 2), which consists of nano-units with a height of 1000 nm and a period of 300 nm under the incidence of left/right-handed circularly polarized light, is presented in Fig. 4 (intermediate wavelengths is used to determine the geometric phase). As shown, the left-handed circularly polarized light in all three bands converges to the designated focus position (2 µm, 0 µm), while the right-handed circularly polarized incident light in all three bands converges to a position slightly shifted from the designated one (–2 µm, 0 µm). For comparison, we design another metalense that can split and focus orthogonal circularly polarized light whose geometric phase is determined according to the maximum wavelength. The electric-field intensity distribution of this metalense at z = 30 µm under the incidence of 460-, 570-, and 680-nm left/right-handed circularly polarized light is presented in the Supplement 1. The Supplement 1 also presents the deviation distance diagram between the actual focus position and the designated position of the designed metalenses, whose geometric phase is determined according to the maximum wavelength or intermediate wavelength. The standard deviations of the offset distances are 0.343 and 0.209, respectively. Thus, for designing metalenses that can split and focus orthogonal circularly polarized light, the standard deviation of the offset distance can be reduced by using the intermediate wavelength of the designed band (rather than the maximum wavelength) to determine the geometric phase.

 

Fig. 4. Electric-field intensity distributions at z = 30 µm of NA ≈ 0.2 metalenses consisting of nano-units with a height of 1000 nm and a period of 300 nm under the incidence of 460-, 570-, and 680-nm (a–c) right-handed circularly polarized light and (d–f) left-handed circularly polarized light.

Download Full Size | PDF

In addition, we calculate the focusing efficiency of the designed metalenses at the focal length of 30 µm under the incidence of 460-, 570-, and 680-nm orthogonally polarized light. Here, we define the focusing efficiency as the ratio of the power within a circle having a diameter three times the full width at half maximum (FWHM) to the total power on the focal plane [30,31]. In this study, the power on the focal plane is obtained by calculating $\frac{1}{2}{\textbf E} \times {\textbf H}$ (where E and H represent the electric and magnetic fields, respectively) instead of calculating $\frac{1}{2}\sqrt {\frac{\varepsilon }{\mu }} {|E |^2}$ (where ɛ and µ represent the dielectric constant and permeability in vacuum, respectively, and E represents the electric-field intensity). This is because the electric and magnetic fields are not perpendicular to each other everywhere. The focusing efficiencies are shown in Fig. 5(a).

 

Fig. 5. (a) Focusing efficiency and (b) FWHM at the focal length under 460–680 nm orthogonally polarized incident light.

Download Full Size | PDF

The average focusing efficiency reaches 59.2% under 460–680-nm TM linearly polarized light, 53.1% under TE linearly polarized light, 58.8% under left-handed circularly polarized light, and 52.7% under right-handed circularly polarized light. The FWHM at the focal point under different polarized light incidence is shown in Fig. 5(b). This value is smaller than the focal spot size given by the diffraction theory of physical optics (0.5×λ/NA). Therefore, the designed metalenses can achieve or even surpass the diffraction-limited focusing.

The metalens array used for broadband full-Stokes polarization imaging consists of the designed metalense that can split and focus orthogonal linearly polarized light, the above metalense with a 45° rotation angle, and the designed metalense that can split and focus orthogonal circularly polarized light, as shown in Fig. 6. The electric-field intensity distributions at z = 30 µm under the incidence of 460-, 570-, and 680-nm different polarized light are presented in Fig. S7 (Supplement 1). As shown, the intensity distribution at the focal plane of the proposed metalense array is consistent with the theory, and the array can realize real-time full-Stokes polarization imaging in the visible broadband range.

 

Fig. 6. Metalens array for broadband full-Stokes polarization imaging.

Download Full Size | PDF

To verify that the designed array model can achieve full Stokes vector measurement, the focusing intensity $I^{\prime} = [{{I_{0^\circ }}\mathrm{^{\prime}\ }{I_{90^\circ }}\mathrm{^{\prime}\ }{I_{45^\circ }}\mathrm{^{\prime}\ }{I_{135^\circ }}\mathrm{^{\prime}\ }{I_L}\mathrm{^{\prime}\ }{I_R}\mathrm{^{\prime}}} ]$ at each design focal point position of the focal plane under the incidence of 0°, 90°, 45°, 135° linearly polarized and left/right-handed circularly polarized light are calculated. The focusing intensity is defined as the power in a circle whose diameter is three times the full width at half maximum (FWHM) on the focal plane, and the power is also calculated by the equation $\frac{1}{2}{\textbf E} \times {\textbf H}$. According to the focusing intensity and Eq. (10), the Stokes vector of the incident light can be reconstructed. The reconstruction results and the relative errors are shown in Table S2-S4 (Supplement 1).

The relative error is defined as [22]

To correct the reconstructed Stokes parameter results, we use the transfer matrix M to establish a calculation relationship between the Stokes parameter obtained from the focusing intensity and Eq. (10) and the theoretical Stokes parameter, which can be written in the following form:

By introducing the Stokes vectors of 0°, 90°, 45° linearly polarized and right-handed circularly polarized light obtained from the focusing intensity and Eq. (10) into the right side of Eq. (12) and corresponding theoretical Stokes vectors into the left side of Eq. (12), the specific values of each parameter in the transfer matrix can be obtained. To verify that the transfer matrix can effectively correct the reconstructed Stokes vectors, we calculate the focusing intensity at each design focal point position of the focal plane under the incidence of four other types of polarized light with different Stokes vectors (the Stokes vectors are ${\left[ {1\textrm{ 0 }\frac{{\sqrt 3 }}{2}\textrm{ }\frac{1}{2}} \right]^\textrm{T}}$, ${\left[ {1\textrm{ 0 }\frac{{\sqrt 2 }}{2}\textrm{ }\frac{{\sqrt 2 }}{2}} \right]^\textrm{T}}$, ${\left[ {1\textrm{ - }\frac{3}{5}\textrm{ }\frac{{2\sqrt 2 }}{5}\textrm{ }\frac{{2\sqrt 2 }}{5}} \right]^\textrm{T}}$, ${\left[ {1\textrm{ - }\frac{4}{5}\textrm{ }\frac{{3\sqrt 3 }}{{10}}\textrm{ }\frac{3}{{10}}} \right]^\textrm{T}}$ respectively). Based on the focusing intensity and Eq. (10), the Stokes vectors of the four polarized incident light can be reconstructed. Then the corrections of the reconstructed Stokes vectors can be made according to Eq. (12). The position of the corrected reconstruction results in the Poincaré sphere is shown in Fig. 7, and the relative error is shown in Table S5-S7 (Supplement 1). The mean relative errors of the above four polarization states are 6.53%, 3.76%, and 5.38% at the wavelengths of 460, 570, and 680 nm, respectively. The results show that the transfer matrix can effectively correct the reconstructed Stokes vectors.

 

Fig. 7. The corrected reconstruction Stokes parameters in the Poincaré sphere under the incidence of (a) 460-, (b) 570-, and (c) 680-nm polarized light (green, yellow, orange, and purple represents ${\left[ {1\textrm{ 0 }\frac{{\sqrt 3 }}{2}\textrm{ }\frac{1}{2}} \right]^\textrm{T}}$, ${\left[ {1\textrm{ - }\frac{3}{5}\textrm{ }\frac{{2\sqrt 2 }}{5}\textrm{ }\frac{{2\sqrt 2 }}{5}} \right]^\textrm{T}}$, ${\left[ {1\textrm{ - }\frac{4}{5}\textrm{ }\frac{{3\sqrt 3 }}{{10}}\textrm{ }\frac{3}{{10}}} \right]^\textrm{T}}$, ${\left[ {1\textrm{ 0 }\frac{{\sqrt 2 }}{2}\textrm{ }\frac{{\sqrt 2 }}{2}} \right]^\textrm{T}}$ respectively. ‘o’ means the theoretical results,while ‘×’ means the reconstruction results).

Download Full Size | PDF

In this study, to determine the effect of the period and height of the nano-unit on the efficiency of the designed metalenses and to exclude the impact of dispersion on the efficiency of the designed metalenses, we assumed that the refractive index of TiO2 does not change with wavelength and is always 2.35. The refractive index of 2.35 is fixed according to the experimental measurement data in Ref. [28] at the wavelength of 460 nm. The choice of fixing the material refractive index based on the minimum wavelength is because the phase modulation of incident light by nanopillars of the same height is greatly affected by the material refractive index when incident by electromagnetic waves with small wavelengths. However, in reality, the specific design needs to optimize the selection of nano-units based on the dielectric constant including dispersion of the materials used in the experiment. Nevertheless, the design process is consistent with that described in Section 2. In addition, we did not consider the manufacturing process. However, this study provides inspiration for manufacturing techniques, and it is not the case that smaller periods and larger heights will result in high-performance devices when designing multifunctional metalenses. We must select the parameters of the nano-units according to the specific requirements.

4. Conclusions

To design metalenses that can achieve broadband polarization splitting and focusing with high focusing efficiency, we simulate and optimize the height and period of the nano-units, which affect the focusing efficiency.

When the difference between the nano-unit period and the minimum wavelength of the incident light is large, the following results are obtained. 1) In general, a larger nano-unit height corresponded to a higher total focusing efficiency of the designed metalenses; however, the increasing rate of the total focusing efficiency of the metalenses decreases with an increase in the height. Therefore, there is a turning point in the height of the nano-unit, which may represent a compromise between the focusing efficiency and the manufacturing process. 2) The period has little effect on the total focusing efficiency of the metalenses.

When the nano-unit period is close to the minimum wavelength of the incident light, the influence of the nano-unit height on the total efficiency of the metalenses has no obvious trend. This may be because the multi-order diffraction caused by the periodicity increases the reflection efficiency and reduces the transmission efficiency. However, the wavelength range where multi-order diffraction is generated by nano-units of different heights is different, which will lead to irregular changes in the transmission efficiency when light of the three selected wavelengths (460, 570, and 680 nm) passes through nano-units of different heights, finally causing the total focusing efficiency to change irregularly with respect to the height of the nano-units.

To achieve broad-spectrum polarization imaging, we design a metalens array that can achieve full-Stokes polarization imaging in the broadband visible range, where the design process of metalenses for splitting and focusing broadband orthogonal circularly polarized light is emphasized. The designed metalens array can split and diffraction-limited focus the orthogonally polarized light to the designated position. The average focusing efficiencies reach 59.2% under 460–680 nm TM linearly polarized light, 53.1% under TE linearly polarized light, 58.8% under left-handed circularly polarized light, and 52.7% under right-handed circularly polarized light. The designed metalenses can be applied to imaging systems such as polarization imaging and polarization light-field imaging systems.