Phase-only transmissive spatial light modulator based on tunable dielectric metasurface

See allHide authors and affiliations

Science  14 Jun 2019:
Vol. 364, Issue 6445, pp. 1087-1090
DOI: 10.1126/science.aaw6747

Dynamic metasurfaces

Nanostructured metasurfaces can function as many passive optical elements. Now, S.-Q. Li et al. demonstrate that metasurfaces can be combined with liquid crystals to provide active control over light beams. With a view toward developing near-eye augmented reality display technology, they combined a dielectric metasurface with a liquid crystal layer to produce a tiny spatial light modulator. The results present a path for the development of dynamic metasurfaces as a platform for miniaturized optical technology with advanced time-dependent functionality.

Science, this issue p. 1087


Rapidly developing augmented reality, solid-state light detection and ranging (LIDAR), and holographic display technologies require spatial light modulators (SLMs) with high resolution and viewing angle to satisfy increasing customer demands. Performance of currently available SLMs is limited by their large pixel sizes on the order of several micrometers. Here, we propose a concept of tunable dielectric metasurfaces modulated by liquid crystal, which can provide abrupt phase change, thus enabling pixel-size miniaturization. We present a metasurface-based transmissive SLM, configured to generate active beam steering with >35% efficiency and a large beam deflection angle of 11°. The high resolution and steering angle obtained provide opportunities to develop the next generation of LIDAR and display technologies.

Spatial light modulators (SLMs) have widespread applications ranging from three-dimensional video projection (1) and additive manufacturing (2) to quantum (3) and adaptive optics. (4) The most versatile phase-only SLMs are capable of reconfiguring the phase retardation of light transmitted through or reflected from each pixel without changing its intensity. Typically, SLMs achieve this using liquid crystals (LCs). The LC molecules, called directors, align with each other, endowing the LC medium with a large uniaxial anisotropy of refractive index, ∆n = neno, ranging from 0.2 to 0.4 in the visible spectral range (ne and no being the extraordinary and ordinary refractive indices, respectively). Moreover, these molecules rotate under the influence of an applied electric field, providing the means to dynamically control the refractive index along a given direction. One limitation of LC-based SLMs is their large pixel size, which is above 3 μm for the best reflective SLM devices and tens of micrometers for the transmissive ones. This limits the field of view (FOV) of the SLM, defined as the angular coverage of the first diffractive order. Further downsizing of the pixels while keeping the required thickness of the LC layer is limited by their mutual cross-talk (5). A typical commercial transmissive SLM (HOLOEYE LC 2012) has a pixel pitch of 36 μm, giving a maximum FOV of 0.7°, severely limiting its potential applications.

Recently, a new class of flat optical elements, called metasurfaces, have been developed. They abruptly modify the phase of light by designated amounts using subdiffractive optical elements called nanoantennas (6, 7). Although metasurfaces have been successfully applied to realize static optical components (811), for various applications it is important to modify the phase dynamically (1219). If each nanoantenna could be individually tuned by applying electrical voltages, the metasurface would act as an SLM with a subwavelength pixel size. One way to achieve tunability is by combining nanoantennas with LCs. Initial results on the integration of nanoantennas inside LC cells (14, 20, 21) showed the possibility of obtaining spectral shifts of resonances using thermally (20) and electrically (21) switched LCs, as well as on-off switchable devices (14). However, a universal active metasurface capable of arbitrary beam shaping through phase-only manipulation is yet to be realized.

Here, we demonstrate how to achieve this active metasurface by integrating the nanoantennas into an LC-SLM device. Modifying the LC orientation around the nanoantennas changes their local environment and resonances. In this case, the main phase accumulation happens inside the nanoantennas rather than in the LC layer, thus uncoupling it from the LC thickness. This helps to reduce the LC cell thickness, which helps to solve the cross-talk issue and to reduce the pixel size, without requiring major modifications to the existing SLM technology.

We use the Huygens’ metasurface concept, realized via dielectric nanoantennas supporting spectrally overlapped electric and magnetic dipole resonances (22, 23). They provide full-range phase shift from 0 to 2π around the resonances peaks (22) and, when the induced dipole moments have equal amplitudes and phases, suppression of the back-scattering (24), resulting in a close-to-unity transmission (22, 23, 25). To achieve high efficiencies in the visible spectral range, we design nanoantennas made of TiO2 (9, 26), with negligible absorption and sufficiently high refractive index (n ~ 2.5) to obtain resonances inside the LC environment (E7 from Merck; no ~ 1.5 and ne ~ 1.7) (see fig. S1). To find the optimum dimensions, we perform full-wave simulations (27), starting with a square lattice of cylindrical nanoantennas. The embedding LC (thickness hLC = 1500 nm) is sandwiched between two glass plates with its director oriented in-plane (θLC = 0°) (see Fig. 1A). The incident light impinges normally onto the device, with its electric field polarized parallel to the LC director, experiencing its extraordinary refractive index. We show maps of transmission and phase for different radii of the TiO2 nanoantennas, sweeping around the optimized dimensions and targeted wavelength of 660 nm (Fig. 1, A and B). The period in all cases is p = 360 nm, so that the unit cell is subdiffractive, and the nanoantenna height is ha = 205 nm. Huygens’ condition (high transmission and 2π-phase coverage) is achieved in a wide range of wavelengths from 650 to 675 nm, which is optimized for the chosen nanoantenna height, becoming narrower at different heights (figs. S2 and S3).

Fig. 1 Optimization of the nanoantenna geometry through simulations.

(A) Amplitude and (B) phase spectra of the zero-order transmission of nanoantenna arrays with different particle radii. The inset shows the schematic of the simulated unit cell. A square array is formed by translating the unit cell in both the x and y directions. (C) Calculated relative phase retardation experienced by a normally incident wave passing through the optimized unit cell (ha = 205 nm, radius R = 135 nm, p = 360 nm) as a function of the LC director rotation. The color of the marker indicates the rotation—from blue (in-plane) to orange to yellow-green (out-of-plane)—and the size indicates the associated transmittance, with respect to the bare LC layer, the ones in the legend representing 50% transmittance. (D) Transmission spectra of the three main diffraction orders (T−1, T0, and T+1) of a wave passing through the beam-deflecting configuration. A unit cell, containing three nanoantennas, is depicted in the inset. The individual bottom electrodes are separated by an additional gap g = 60 nm.

Next, we study the phase retardation of light transmitted through the optimized nanoantenna array when the LC director orientation is switched from in-plane to out-of-plane (Fig. 1C). This can be realized by applying an electrical bias between two electrodes sandwiching the LC (28). We see that above 670 nm the relative phase difference at different LC orientations does not exceed 0.8π. This is the spectral region in which the nanoantennas are not resonant and the phase retardation is attributed to mere propagation in the LC layer. Below 670 nm, strong phase variations are generated by spectral shifts of the nanoantenna resonances induced by the refractive index change of the surrounding LC. In Fig. 1C, we highlight three series of results corresponding to LC director orientations of 0°, 45°, and 90°. The phase in these three series is evenly spaced at around 2π/3 to each other, in the wavelength region between 660 and 670 nm. Furthermore, they have similar transmittance, ranging between 60 and 90%, thus satisfying the requirements for a three-phase-level transmissive SLM—evenly spaced, full-phase coverage and uncoupled, high transmission (see fig. S4 for the corresponding field distributions).

We then simulate a beam-steering SLM device having the unit cell shown in the inset of Fig. 1D. Instead of one nanoantenna per pixel, we use three. This gives the individually addressable (bottom) electrodes sufficient width to accommodate the fringing electric field and phase-broadening effects (29). Further increasing the number of particles per pixel does not strongly improve performance (see table S1). A supercell comprising three pixels (with different LC rotations and corresponding phase levels, in accordance with Fig. 1C) is periodically repeated to produce an infinite gradient metasurface. The numerically calculated diffraction efficiencies of the device (Fig. 1D) show a clear increase of the transmission into the −1 diffraction order at the wavelength of 665 nm, reaching a value of ~48%, together with a strong reduction of efficiencies into the 0 and +1 orders. This translates into the expected beam deflection, consistent with the phase analysis of homogeneous arrays (Fig. 1C and fig. S4).

To prove our concept, we fabricate a device comprising 28 individually addressable electrodes (27). An optical image of the finalized device and its computer-generated design, including the printed circuit board, are shown in Fig. 2, A and B. Scanning electron microscopy (SEM) images, showing the nanoantennas on top of the electrodes before LC cell assembly, are shown in Fig. 2C (see the device transmission spectra in fig. S5).

Fig. 2 Metasurface-based SLM: Demonstration of a dynamic beam deflection device.

(A) A photo of the device mounted for measurements. (B) Computer-generated design consisting of the printed circuit board and the nanofabricated part (gray, central area). “TL” stands for top-left (this notation is used to keep the correct sample orientation). (C) SEM images of the fabricated device, showing the individual electrodes with the nanoantennas on top. Scale bars: left panel, 5 μm; right panels, 1 μm. (D) Two-level addressing of the device, which acts as a grating with tunable periodicity. The left side shows the experimentally measured diffraction intensities as a function of the deflection angle for different electrode-addressing configurations, shown in the corresponding right side. Gray patches represent grounded electrodes, and blue patches represent biased ones (at 8 V). a.u., arbitrary units. (E) As in (D), but for the three-level addressing of the device, which acts as a beam deflector with tunable deflection angle. Gray patches represent grounded electrodes, blue patches represent biased ones inducing total rotation of the LC, and green patches represent intermediate-voltage-level ones inducing partial rotation of the LC. (F and G) Measured transmission efficiencies (transmission intensity normalized to the incident intensity) of the three main diffraction orders for the case in which the beam deflects at ~4° (F) and at the maximum achievable angle of ~11° (G).

To test the device, we first use a two-level scheme, in which we address alternating electrodes at ground voltage (keeping the in-plane LC alignment) and elevated voltage (6 to 8 V, inducing vertical LC alignment). In this situation, the device acts as a simple diffraction grating in which the periodicity, and thus the diffraction angles, can be tuned by choosing different electrode-addressing configurations. The configurations and their corresponding measured diffraction intensities are shown in Fig. 2D. The optical performance of the device in this and all subsequent cases is characterized using the spectrally resolved back focal plane imaging technique (8, 27).

Next, we introduce the intermediate-voltage-level electrode inducing the partial rotation of the LC. In this way, we obtain a three-level-addressing scheme in which we can realize the beam deflection configuration analyzed theoretically above. By choosing different addressing schemes, we can tune the deflection angle, as demonstrated in Fig. 2E. The diffraction efficiencies into the three main diffraction orders, −1, 0, and +1, are shown in Fig. 2, F and G, for two characteristic cases. They correspond, respectively, to a case in which the beam deflects at around 4° and the one giving the maximum deflection angle of 11°. Both plots show a clear region at around 650 nm in which the 0 order and the +1 order are suppressed and the −1 order is enhanced, reaching values >15%. In both cases, the optimum voltage for the intermediate-level electrode was found to be 3.5 V, whereas the high-level voltage was 7 and 8 V for 4° and 11° deflection, respectively.

The measured efficiencies are noticeably lower than the theoretical predictions (fig. S6). We attribute this to perturbation of the LC alignment at the edge of the active region and the limited number of electrodes fabricated. Both are related to the small sample size and can be mitigated upon further process improvement and integration of active-matrix electrodes, which can be addressed at a large scale. To corroborate this, we designed and fabricated a simpler but larger device in which two of three neighboring bottom electrodes are connected to the two terminals of an external voltage source. One of the connected electrodes and the top electrode are put to the ground state, and the other is biased to induce the out-of-plane switching of the LC directors (Fig. 3A). The third electrode is left floating (unconnected) to pick up an electric potential in between the biased and the grounded electrodes, so that the LC directors can attain the intermediate level of rotation leading to deflection of the incoming light beam. This design allows the reversal of the deflection direction by switching the applied voltages, as shown in Fig. 3A, but does not allow a change in deflection angle. On the other hand, it allows us to fabricate a device with a much larger number of electrodes and thus a much larger aperture size. SEM images of the fabricated device before the LC infiltration are shown in Fig. 3B.

Fig. 3 Large aperture, simplified metasurface-based SLM.

(A) Schematic drawings of the SLM and its operation concept. hLC denotes the thickness of the LC layer, w the width of the electrode, and g the gap between the electrodes. (B) SEM images showing the fabricated device. The unconnected (floating) electrodes appear brighter because of induced charging. Each device is 120 μm by 100 μm. Scale bars: left panel, 5 μm; central panel, 600 nm; right panel (tilted view), 250 nm. The area outlined in yellow in the left panel highlights a supercell of the device. (C) Measured transmission efficiencies (transmission intensity normalized to the incident intensity) of the three main diffraction orders for the case of 12-V left bias. (D) As in (C), but for the 13-V right bias.

In Fig. 3, C and D, we plot the measured diffraction efficiency spectra of the device at the optimum beam deflection conditions for each case (see figs. S7 and S8 for the voltage optimization). The efficiencies and the overall trends in this device show very good agreement with the simulated results plotted in Fig. 1D. In the best case, the experimental beam deflection efficiency at 660 nm reaches 36%, compared with 48% obtained from simulations. There are several effects not considered in the simulations, which may all contribute to this slight discrepancy, such as the fringing fields at the gaps between the electrodes, nonuniformities in nanoantenna sizes, and imperfections in the LC director alignment. Nevertheless, the results indicate that the moderate efficiencies obtained for the fully dynamic device (Fig. 2) can be improved by realizing a larger device size.

In this study, we demonstrate a one-dimensional phase-only nanoantenna-based transmissive SLM device that reaches an experimental efficiency of 36% with a pixel size of only ~1 μm and a FOV of 22°. The presence of the nanoantennas allows a reduction of the LC layer thickness required to achieve the necessary phase modulation by more than half compared with traditional SLMs. This, in turn, alleviates the phase broadening and fringing field effects, which are the main limiting factors preventing pixel-size reduction in traditional SLM devices. Our design can be further extended to two-dimensional, phase-only LC SLMs, providing opportunities to make ultra-high-resolution devices, which may perform faster and have a larger FOV, while keeping high efficiencies.

Supplementary Materials

Materials and Methods

Figs. S1 to S9

Table S1

Reference (30)

References and Notes

  1. Materials and methods are available as supplementary materials.
Acknowledgments: The authors acknowledge helpful discussions with Y. H. Fu, Y. F. Yu, Z. Pan, and S. T. Ha, and also the fabrication support provided by K. Goh, S. Yap, A. Huang, and Y. T. Toh. Funding: The authors acknowledge financial support from National Research Foundation of Singapore (grant NRF-NRFI2017-01), IET A F Harvey Engineering Research Prize 2016, A*STAR SERC Pharos program (grant 152 73 00025) (Singapore), and AME Programmatic Grant A18A7b0058 (Singapore). Author contributions: S.-Q.L. performed simulations, sample nanofabrication, device design, electro-optical characterization, and wrote the first draft; X.X. performed simulations and optical characterization; R.M.V. performed liquid crystal cell fabrication and characterization; V.V. performed SEM measurements; R.P.-D. conceived the idea and contributed to simulations and overall supervision of the project; A.I.K. conceived the idea and supervised the whole project. All authors discussed the results and worked on the manuscript. Competing interests: None declared. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper or the supplementary materials.

Stay Connected to Science

Navigate This Article