2.1.

Principle and Design

Figures 1(a) and 1(b) schematically illustrate the tunable functions of the proposed metalens. The achromatism is accomplished by delicately designing the integrated resonant phase (${\phi }_R$) and geometric phase (${\phi }_G$). Geometries of silicon pillar meta-units^36,37 vary from the center to the edge, enabling a frequency-dependent lens phase profile. The space-variant alignments of LCs³⁸ perform a frequency-independent ${\phi }_G$ modulation. Without the applied voltage, it works as a broadband achromatic lens, where light within the designed spectrum deflects to the same focal spot. When applying a saturated bias on the graphene electrodes³⁹ covered on both sides of the LC layer, LCs are reoriented perpendicularly to the substrates. Thus the ${\phi }_G$ modulation vanishes and only the ${\phi }_R$ of the metaunits works. Correspondingly, the device turns into a chromatic lens with an anomalous dispersion, i.e., the focal length decreases when the frequency increases.

Fig. 1

Schematic function tunability and designs of the metalens. (a) The broadband achromatic focusing state without bias. (b) The dispersive focusing state with a saturated bias. The superstrate depicts the dielectric metasurface while the blue ellipsoids between the superstrate and substrate denote LCs. (c) Dependencies of ${\phi }_R$ (red curve) and PCR of a silicon pillar metaunit on $f$. Inset shows the dimensions of the unit. Dependency of ${\phi }_R$ on $l$ and $w$ at (d) 0.9 and (e) 1.4 THz, respectively. (f) Dispersion ratio of the silicon pillar as a function of $l$ and $w$ under LCP incidence. Dependency of PCR on $l$ and $w$ at (g) 0.9 and (h) 1.4 THz. The region circled by the black dashed line indicates a PCR over 12% from 0.9 to 1.4 THz. (i) Designed $l$ and $w$ of the silicon pillars along $r$ from the center to the edge. (j) ${\phi }_A$ and ${\phi }_R$ at 0.9 (blue solid and dashed curves) and 1.4 THz (red solid and dashed curves) and residual ${\phi }_G$. (k) Target ${\phi }_G$ and corresponding LC director distributions are labeled by the orange lines.

The generalized phase profile of an achromatic lens is expressed as

From Eq. (2), at a certain $r$, the required ${\phi }_A$ shows a linear relationship with $f$. Such a linear dispersion can be satisfied by high-order waveguide resonances excited inside the dielectric pillars with a high aspect ratio and large dielectric constant⁴⁰ (Fig. S2 in the Supplementary Material). Apart from the former dispersion item achieved by the ${\phi }_R$, the constant $d$ is frequency-independent and can be accomplished through introducing the ${\phi }_G$, namely the Pancharatnam–Berry phase.⁴¹ It originates from the photonic spin–orbit interaction and can be created via controlling the geometric orientation angle $\theta$ of anisotropic media, e.g., LCs and metaresonators.^42,43 Theoretically, ${\phi }_G$ is twice $\theta$ and conjugated at left-handed and right-handed circular polarization (LCP/RCP) incidences.

As a proof-of-concept demonstration, an achromatic lens with $F=15.0\mathrm{mm}$ from 0.9 to 1.4 THz was designed. In Fig. 1(c), ${\phi }_R$ and the polarization conversion ratio (PCR, determining the efficiency of the metalens) of a silicon pillar metaunit with $l=50\mu \mathrm{m}$, $w=36\mu \mathrm{m}$, $h=200\mu \mathrm{m}$, and $p=60\mu \mathrm{m}$ are simulated and presented as an example. To simplify the simulation, we fix $h$ and $p$ and then vary $l$ and $w$ in the range of 10 to $50\mu \mathrm{m}$ to achieve a library of dispersion ratios (defined as $\mathrm{d}{\phi }_R/\mathrm{d}f$), as shown in Fig. 1(f). $\mathrm{d}{\phi }_R/\mathrm{d}f$ increases with $l$ and $w$ separately. The abnormal values along the diagonal are due to the near zero PCR when $l=w$. ${\phi }_R$ and PCR are also related to $l$ and $w$; the dependencies of ${\phi }_R$ on them at 0.9 and 1.4 THz are shown in Figs. 1(d) and 1(e), respectively. The $\pi$ phase difference of ${\phi }_R$ on mirrored sides of $l=w$ is due to the additional ${\phi }_G$ induced by a 90-deg rotation of the metaunit. To achieve the desired achromatic property, the geometry of metaunits at different $r$ should be deliberately selected. In addition to the dispersion factor $\mathrm{d}{\phi }_R/\mathrm{d}f$, ${\phi }_R$ should be further considered to maintain a lens profile. Figures 1(g) and 1(h) exhibit the dependency of PCR on $l$ and $w$. For optimized efficiency, only parameters in the black-dashed region are used. Due to the limited range of $\mathrm{d}{\phi }_R/\mathrm{d}f$ ($41.8\mathrm{rad}/\mathrm{THz}$ at $l=48\mu \mathrm{m}$, $w=10\mu \mathrm{m}$ to $49.5\mathrm{rad}/\mathrm{THz}$ at $l=50\mu \mathrm{m}$, $w=44\mu \mathrm{m}$), $r$ of the metalens is limited to $3240\mu \mathrm{m}$, i.e., 54 units. According to this design, the $l$ and $w$ of the silicon pillars along $r$ are presented in Fig. 1(i). It is observed that $l$ remains nearly unchanged around $50\mu \mathrm{m}$, whereas $w$ gradually decreases from the center to the edge. Figure 1(j) plots the ${\phi }_R$ and ${\phi }_A$ at 0.9 and 1.4 THz. The $f$-independent and $r$-dependent gap between ${\phi }_A$ and ${\phi }_R$ can be compensated for by ${\phi }_G$. The corresponding ${\phi }_G$ profile [Fig. 1(k)] can be precisely accomplished by locally varying the LC directors, as shown schematically in the overlay.

2.2.

Focusing Performance Characterizations

Figures 2(a)–2(d) present the SEM images of the fabricated metasurface. Figure 2(a) shows a part of the metasurface from the top view. The silicon pillars close to the center and the edge are labeled by orange and blue dashed squares, respectively. Corresponding zoomed-in images in Figs. 2(b) and 2(c) show well-defined microstructures, which faithfully follow the design [Fig. 1(i)]. The cross-section image of the metasurface is exhibited in Fig. 2(d). Some broken pillars resulting from the damage caused in the cutting process are observed. Due to the different etching speeds for the varying aspect ratio in the fabrication, pillars near the edge [the right side in Fig. 2(d)] are higher than those near the center (left side) (detailed in Fig. S3 in the Supplementary Material). The photopatterned LCs are observed under the microscope with crossed polarizers. The optical image of the whole LC region is displayed in Fig. 2(e), in which the gray scales are consistent with the ${\phi }_G$ shown in Fig. 1(k). As the ${\phi }_G$ is pixelated with the same periodicity of $60\mu \mathrm{m}$ as that of silicon metaunits, gray scale pixels are vividly revealed in the zoomed-in image in Fig. 2(f).

Fig. 2

Characterization of the metalens. (a) SEM image of the metasurface (top view). Scale bar: $300\mu \mathrm{m}$. Zoomed-in images of the silicon pillars labeled by the (b) orange and (c) blue squares in (a), respectively. (d) Cross-section image of the metasurface. Scale bars in (b)–(d): $100\mu \mathrm{m}$. (e) Optical image of the LC layer under crossed polarizers. (f) A zoomed-in image labeled by the orange square in (e). Scale bars in (e) and (f) are 1 mm and $500\mu \mathrm{m}$, respectively.

We use a scanning near-field THz microscope (SNTM) to characterize the performance of the LC integrated metalens. The normalized intensity distributions from 0.9 to 1.4 THz with an interval of 0.1 THz are simulated and measured at the bias OFF and saturated states, respectively (Fig. S4 in the Supplementary Material). Figure 3(a) selectively reveals the simulated THz fields at 0.9, 1.1, and 1.4 THz in the $xz$ plane at bias OFF. The measured $F$ remains unchanged around 12.0 mm within the designed spectrum [Fig. 3(b)], which is consistent with the simulation. Little distortions are observed at 0.9 and 1.1 THz, which is mainly attributed to the slightly tilt incidence of the THz beam. The THz fields in the $xy$ plane ($z=12.0\mathrm{mm}$) and the corresponding transverse intensity profiles are also exhibited, verifying the perfect achromatic focusing property. It is noticed that the $F$ is slightly smaller than the design as a result of the gradual etching depth deviation (see Sec. S2 in the Supplementary Material). This deviation is also considered in the simulation. After a square-wave alternating voltage (75 $V_{\mathrm{rms}}$, 1 kHz) is applied, both the simulated and measured results [Figs. 3(c) and 3(d)] show a significant $F$ dispersion among different $f$ from 0.9 to 1.4 THz. It is noticed that, as the ${\phi }_G$ is spin-dependent, the metalens only works at the LCP THz incidence when the bias is OFF, whereas it works at both LCP/RCP incidences when the bias is saturated due to the spin-independent ${\phi }_R$. LCs can be electrically tuned to match the half-wave conditions at different $f$ to generate a complete geometric phase modulation, contributing to the achromatic case. Mismatch from the half-wave condition causes the residual unmodulated wave, contributing to the dispersive case. Thus via tuning the applied voltage, the required circular polarization incidence of the metasurface is satisfied.

Fig. 3

Focusing performance of the metalens. (a) Simulated THz fields in the $xz$ plane and (b) measured THz fields in the $xz$ and $xy$ plane ($z=12.0\mathrm{mm}$) at 0.9, 1.1, and 1.4 THz, respectively, when no bias is applied. (c) Simulated and (d) measured THz fields in the $xz$ and $xy$ plane ($z=13.0\mathrm{mm}$) of the same sample with a saturated bias. Dashed lines in (a)–(d) label the corresponding focal planes. White curves in (b) and (d) depict the transverse intensity distributions at $z=12.0$ and 13.0 mm, respectively. (e) Simulated and measured $F$, (f) FWHM at $z=12.0$ and 13.0 mm in (b) and (d), separately, and (g) measured device efficiency, as a function of $f$. The black dashed line in (f) depicts the diffraction-limited FWHM.

The dependencies of $F$ on $f$ at both the bias OFF and saturated states are quantified in Fig. 3(e). The results satisfactorily verify the function tunability from achromatic to dispersive. Full-widths at half-maximum (FWHMs) of the intensity profile at different $f$ are shown in Fig. 3(f). The values at the achromatic state [$z=12.0\mathrm{mm}$ in Fig. 3(b)] are very close to the ideal diffraction limit ($\lambda /2$ NA), whereas distinct deviations are observed at the dispersive state [$z=13.0\mathrm{mm}$ in Fig. 3(d)] due to the defocusing at lower $f$. Figure 3(g) exhibits the measured efficiency, which is defined as the ratio of THz intensity at the focal plane to the incident LCP THz power. The average efficiency of the dispersive case ($\sim 33.9\%$) is higher than that of the achromatic one ($\sim 26.1\%$), which is attributed to the additional PCR loss caused by LCs. The total efficiency is determined by several factors: the PCR of the metaunits and LCs, the absorption and scattering of all transmitted layers, and the imperfect fabrication of the sample. These issues can be addressed by further optimizing the metaunit geometry and the LC layer thickness and introducing low-loss substrates (detailed analysis in Sec. S4 in the Supplementary Material).

2.3.

Dynamic THz Imaging

The imaging performance of the metalens is further characterized, utilizing the setup schematically illustrated in Fig. 4(a). After generation with a photoconductive antenna,⁴⁴ the linearly polarized THz wave is converted to LCP with a quarter-wave plate. The THz beam transmits the object and subsequently the metalens; then it is scanned by the probe. The object is a “smiling face” with a diffraction-limited resolution. The object distance $u$ and image distance $v$ are designed as both 24.0 mm to obtain an image with an equal size of the object when $F=12.0\mathrm{mm}$. The intensity and phase distributions from 0.9 to 1.4 THz with an interval of 0.1 THz are measured at the bias OFF and saturated states (Fig. S5 in the Supplementary Material). At the bias OFF state, a multispectral imaging is realized, where the smiling face can be distinctly observed at any $f$ within the spectrum. When the bias is saturated, the imaging quality becomes poorer at lower $f$ due to the deviation of $F$ from 12.0 mm. Figures 4(b)–4(e) show the corresponding intensity and phase distributions at 0.9 and 1.4 THz. The transverse intensities across two “eyes” and the “mouth” are clearly recognized in Figs. 4(b), 4(c), and 4(e), indicating a satisfactory imaging quality. The smiling face is distinguishable in corresponding phase distributions as well. However, the imaging effect is poor in Fig. 4(d) due to the remarkable defocusing. The imaging quality can be further enhanced via enlarging the NA of the metalens. A simple and effective method is increasing the $h$ of the silicon pillars while maintaining the other parameters (see Fig. S6 in the Supplementary Material).

Fig. 4

Broadband THz imaging. (a) Schematic of the SNTM setup. The inset exhibits the micrograph of the object “smiling face,” where the scale bar indicates 1 mm. Measured intensity and phase distributions at (b) 0.9 and (c) 1.4 THz without bias. The white and orange curves depict the intensity profiles along the white and orange dashed lines, respectively. The outline of the object “smiling face” is labeled by black dashed lines in the phase diagrams. Measured intensity and phase distributions at (d) 0.9 and (e) 1.4 THz with a saturated bias.

2.4.

Active Metadeflector

The proposed strategy can be rationally extended to the design of various active metadevices beyond the metalens. We numerically demonstrate a beam deflector with tunable dispersion in Fig. S8. It is composed of silicon metaunits with geometries varying along the $x$ axis and LCs with gradient orientation distributions. The combination of ${\phi }_R$ and ${\phi }_G$ contributes to a fixed deflection angle of 17 deg from 0.9 to 1.4 THz. When the LCs are reoriented along the $z$ axis, the deflection becomes dispersive (details in Sec. S5 in the Supplementary Material).

3.1.

Sample Fabrication

The fabrication process of the LC integrated metalens is schematically illustrated in Fig. S9 in the Supplementary Material. The metasurface is fabricated via the standard photolithography process followed by a reactive ion etching on a $500\text{-}\mu \mathrm{m}$-thick intrinsic silicon wafer. Then the residual photoresist is lifted off. Both the metasurface and silica substrate are ultrasonically cleaned and O-plasma treated for 10 min. Then they are transferred with few-layer graphene from copper foil (Six Carbon Technology, China). After that, they are spin-coated at 3000 rpm for 30 s with photoalignment agent SD1^45,46 (Dai-Nippon Ink and Chemicals, Japan) and baked at 100°C for 10 min to remove the solvent. Then the metasurface and silica substrate are assembled with a $250\text{-}\mu \mathrm{m}$-thick Mylar spacer to form a cell. A digital micromirror device-based dynamic microlithography^47,48 is employed to perform the photopatterning process. Then the LC NJU-LDn-4⁴⁹ with an average birefringence over 0.3 within the designed spectrum is infiltrated at 180°C and self-assembled to the desired orientations. The imaging object “smiling face” is fabricated via photolithography followed by an electron beam evaporation to deposit Au on the silicon substrate.

3.2.

Numerical Simulation

The parameter sweep and selection of silicon pillar metaunits and the focusing performance simulation of the metalens are carried out using a commercial software (Lumerical FDTD Solutions). Due to the geometric symmetry of the lens, the model along the radius is established to simplify the simulation. The LC layer is divided into small domains. Each domain is set as a $60\mu \mathrm{m}\times 60\mu \mathrm{m}\times 250\mu \mathrm{m}$ ($x\times y\times z$) cuboid. The LC is a diagonal dielectric material with $n_o=1.56$ (diagonal elements $xx$ and $yy$) and $n_e=1.88$ (diagonal element $zz$). The LC director distributions are set by an LC orientation module. An LCP plane wave with a spectrum range from 0.9 to 1.4 THz is incident along the $z$ axis.

3.3.

Characterizations

An SNTM setup (Terahertz Photonics Co., Ltd., China), which is based on the photoconductive THz generation and detection, is utilized to characterize the performance of the metalens. In the setup, a scanning tip fixed on a motorized stage [Fig. 4(a)] is utilized to record the $E_x$ field in the $xy$ plane with a step size of 0.25 mm. The sample moves along the $z$ axis with a step size of 1 mm to capture the $E_x$ field in the $xz$ plane. Then an interpolation algorithm is adopted to obtain the measured THz fields.