1. Introduction

Optical metasurface with subwavelength feature size has become one of the most promising technologies in optics industry, due to its unprecedented versatility in light manipulation, compact form factor and process compatibility. Nowadays, it has been deployed in several commercial products, including 3D sensors, polarization cameras and optical filters [1,2]. Reconfigurable metasurface, with the capability to dynamically control the state of operation, is also advancing rapidly in commercialization [3,4]. A wide range of applications, such as beam steering for Lidar, displays and telecommunication modulators, have started to employ reconfigurable metasurface [5]. The mechanisms to realize the reconfigurability of metasurface have been explored extensively in the past decade. Several distinct methods have been proposed [6,7]. For example, liquid crystal (LC)-based dynamic metasurface has been reported for the applications of beam steering and light modulation [8–10]. Electrically controlled birefringence of LC provides tunable dielectric environment for metasurface, enabling tunable metasurface behaviour. Density modulation of free charge carriers in semiconducting material through electrical gating or photocarrier excitation was also realized for high-frequency active metasurface [11,12]. Active metasurface using phase change material (PCM), for instance vanadium dioxide and chalcogenide compounds, has demonstrated high potential for addressable and fast optical modulators [13,14]. Mechanically movable metasurface is another strong candidate for commercially viable reconfigurable metasurface, thanks to the mature manufacturing processes developed in micro-electro-mechanical systems (MEMS) devices [15]. Other methods like chemistry-based active metasurface [16] and thermally tunable metasurface [17] have proved unique advantages in certain applications.

Although plenty of options have been demonstrated through literature, it is challenging to meet the stringent requirements for industrial applications. One of the foremost requirements is the mass manufacturability and scalability. It is necessary for the fabrication processes of active metasurface to be compatible with existing Complementary Metal-Oxide-Semiconductor (CMOS) processes, which is the key to the low cost of modern electronic devices. It is also important that the assembly, packaging, testing, and calibration processes are friendly to existing semiconductor infrastructure. Device lifetime and working efficiency are also critical to commercial products. Moreover, inhomogeneous reconfigurability [7] or addressability of large pixel arrays [3] is required for most of the high-end light modulators. LC-based reconfigurable metasurface has a high readiness level in terms of scalability and reliability because of the mature ecosystem of LC devices such as liquid crystal display (LCD). The integration with silicon backplane is also well established in liquid crystal on silicon (LCoS) devices. Therefore, LC-based approach stands out as one of the most feasible options for reconfigurable metasurface with commercial prospect. In literature, both electrical and thermal stimuli have been reported [18,19]. In Ref. [18], the optical transmission was modulated thermally from around 45% to around 75% at 810 nm by elevating the measuring temperature from room temperature to 60°C. Electrical tunability was reported in Ref. [19], in which the measured optical transmission experienced a spectral tuning of around 110 nm in the near-infrared range when the applied AC voltage amplitude reached 2.7 V. In addition, devices for the visible wavelength have been developed [8,20]. Reference [8] reported a total phase modulation of 4π/3 in the wavelength range of 660 nm – 670 nm. In Ref. [20], electrical switching at 650 nm was recorded with the response time of 110 µs and 800 µs for the ‘on’ and ‘off’ processes, respectively.

In this work, we demonstrate an LC-based active metasurface device (LCAM) that provides amplitude modulation for near infrared wavelength. Sub-micrometre cell gap (875 nm) was realized, thanks to the subwavelength-thick metasurface layer. As a result, sub-millisecond response time was experimentally recorded at room temperature. Response time was further reduced by raising the operating temperature and > 3.4 KHz switching frequency was achieved at 70 °C. Another important feature of the LCAM device is that there is no alignment material or alignment process. LC is aligned directly by metasurface itself. This work demonstrates the key performance indicators of the LCAM device, showcasing the commercial prospect of LC-based reconfigurable metasurface.

2. Preparation of the LCAM cell

The silicon metasurface on ITO-coated glass substrate was fabricated through standard electron beam lithography and liftoff processes. The metasurface is composed by 1-dimensional gratings with grating dimension: width $w$ = 400 nm, height $h$ = 190 nm, and period $p$ = 800 nm, as is shown in Fig. 1(a). The LCAM cell was then created by assembling the metasurface substrate with a glass substrate and spacers. Nematic LC was filled through vacuum filling above its clearing temperature. Unlike conventional LC devices, no alignment material was applied in this case (to either metasurface substrate or glass substrate) and LC over the metasurface area was aligned naturally by the metasurface. The alignment quality was quantitatively evaluated by measuring the LCAM cell transmittance between a pair of crossed polarizers [10]. The transmitted light intensity ($I$) was measured at different in-plane angles ($\varphi $) of the LCAM cell. As is shown in Fig. 1(b), the microscope images of the LCAM cell positioned at $\varphi $ = 0° and $\varphi $ = 45° exhibit clear $\varphi $ dependence of I over the metasurface area ${I_{meta}}$. Generally, I satisfies $I \propto {\sin ^2}\varphi $ if birefringent material is placed between crossed polarizers. This was experimentally verified by measuring ${I_{meta}}$ with a photodiode. The normalized ${I_{meta}}$ is shown in Fig. 1(c) with blue circles. The measured data points were fitted with $I \propto {\sin ^2}\varphi $ and it shows good accuracy (coefficient of determination ${R^2}$ = 0.99946). The alignment quality can be revealed by the ratio ${I_{max}}/{I_{min}}$, where ${I_{max}}$ and ${I_{min}}$ are the maximum and minimum value of the fitted curve, respectively. The ${I_{max}}/{I_{min}}$ for metasurface is 33. The values of ${I_{max}}$, ${I_{min}}$, and ${I_{max}}/{I_{min}}$ provide important information on the true ‘dark state’ and the contrast [21–23]. As a comparison, ${I_{max}}/{I_{min}}$ can reach a few hundred for a normal LC device that is composed of a layer of LC and alignment material. For example, ${I_{max}}/{I_{min}}$ = 300 was recorded in [10] and ${I_{max}}/{I_{min}}$ = 196.3 was recorded in [9]. Both LC cells were photoaligned before filling the LC, although the value of ${I_{max}}/{I_{min}}$ are different because of different preparation conditions. For metasurface-integrated LC devices, ${I_{max}}/{I_{min}}$ = 71 was recorded for photoaligned metasurface in [10] and ${I_{max}}/{I_{min}}$ = 25.6 was recorded for self-aligning metasurface in [10].

 

Fig. 1. Characterization of LC alignment quality by metasurface. (a) Schematic of LCAM device structure. (b) Microscope images of the LCAM cell when the in-plane angle $\varphi $ = 0° and $\varphi $ = 45°, respectively. (b) Measured transmission intensity at different in-plane angles.

Download Full Size | PDF

It is also noticed that the LC area between bare glass substrates next to metasurface also shows $\varphi $ dependence. Similar measurements were performed for the LC area with the same field size. The normalized transmitted intensity through the LC area ${I_{LC}}$ is shown in Fig. 1(c) with red circles. The ${I_{max}}/{I_{min}}$ for LC is 1.4. The weak $\varphi $ dependence of LC is due to the unintentional alignment effect of the bare glass substrates induced during the preparation processes.

It is also evident from Fig. 1(b) that both the metasurface and LC area show spatial nonuniformity. The nonuniformity of metasurface is mainly due to the fabrication defects of the metasurface, which distort the desired LC alignment. The nonuniformity of surrounding LC is due to the lack of alignment and pretilt angle on both sides of LC. The competition between different forces, such as the weak anchoring force of the bare glass substrate and elastic force among LC molecules, lead to the rather random ${I_{LC}}$ distribution.

The cell gap ($t$) after LC filling was measured indirectly by extracting the phase retardation ($\Delta \varphi $) between orthogonal components of linearly polarized incident light. Phase retardation was then converted to cell gap, provided that the birefringence of LC was pre-determined. The measurement of $\Delta \varphi $ was carried out by placing the LCAM cell between crossed polarizers with the alignment direction of LC oriented at 45° to the polarization direction of both polarizers. A square wave AC voltage with peak value ${V_{peak}}$ = 0 V – 10 V and frequency of 2 KHz was applied across the LCAM cell. A spectrometer was used to capture the transmission ($T$) spectrum over the birefringent LC area without metasurface. The measured T spectrum for the wavelength range of $\lambda $ = 415 nm - 500 nm is shown in Fig. 2(a). The T spectrum has been normalized at each wavelength. The cross-section at $\lambda $ = 420 nm, 430 nm, and 440 nm (indicated by white dashed lines in Fig. 2(a)) are shown in Fig. 2(b). We assume that LC molecules are all vertically aligned along the cell gap ($\Delta \varphi $ = 0) when ${V_{peak}}$ = 10 V, since $T \approx $ 0 at ${V_{peak}}$ = 10 V. The maximum transmission (${T_{max}}$) at ${V_{max}}$ indicates $\Delta \varphi $ = $\pi $, where the LCAM cell functions as a halfwave plate. In the range of ${V_{peak}}$ = 0 V - ${V_{max}}$, $\Delta \varphi $ further increases on top of $\pi $. The increased phase shift ($\delta \varphi $) can be determined by ${T_{max}} - T({{V_{peak}} = 0V} )$ = ${\sin ^2}\delta \varphi $, where $T({{V_{peak}} = 0V} )$ represents the transmission at ${V_{peak}} = 0V$. The extracted total phase retardation $\Delta \varphi = \pi + \delta \varphi $ is shown in Fig. 2(c) at different measuring temperatures. It is evident that $\Delta \varphi $ decreases as the measuring temperature rises. This is due to the reduced birefringence of LC at higher temperatures. The birefringence was determined separately through similar measurements, as was described in [24]. A reference cell with the cell gap measured with the interferometric technique was used and then particular care were taken to make sure that the cell gap does not change after the cell was filled with the LC. After that the phase depth was measured using transmission spectra of the cell placed between crossed polarisers, similar to the procedures described above. The birefringence was then calculated for at least ten points at the wavelength range 400-700 nm and fit with the extended Cauchy model. The results are shown in Fig. 2(d) with red squares. Finally, the cell gap was calculated to be 875 nm ${\pm} $ 5.7 nm, as is shown in Fig. 2(d) with blue squares.

 

Fig. 2. Measurement of cell gap after LC filling. (a) Normalized transmission spectrum at different voltages. (b) Cross section view at $\lambda $ = 420 nm, 430 nm, and 440 nm. (c) Extracted phase retardation at different measuring temperatures. (d) Measured birefringence and calculated cell gap at different wavelengths.

Download Full Size | PDF

3. Electrical reconfigurability

The electrical tunability of the LCAM cell was experimentally and numerically investigated in this section. The reflection spectrum of the LCAM cell under different voltage levels (peak AC voltage ${V_{peak}}$ = 0 V - 10 V) was measured for the near infrared (NIR) wavelength. The normalized spectrum is shown in Fig. 3(a). When ${V_{peak}}$ = 0 V, strong reflection ($R$) occurs near 1500 nm, indicating the existence of resonance. As ${V_{peak}}$ increases, a blue shift is observed. To visualize the spectral movement of the resonance, the spectral position for the range of reflection that is > 95% ${\times} {R_{max}}$ is shown in Fig. 3(b) as shaded area. In addition, the spectral position of the peak reflection (${R_{max}}$) is also plotted in Fig. 3(b) as scattered black circles, with a tuneable range from 1500 nm to 1441 nm as ${V_{peak}}$ increases from 0 V to 10 V. It is evident that the spectral movement of the resonance is not linear with voltage. In order to find out the optimum wavelength for this LCAM device in terms of the largest modulation depth in transmission, we firstly plot the maximum reflection (${R_{max}}$) and the minimum reflection (${R_{max}}$) over the whole voltage range for each wavelength. The result is shown in Fig. 3(c) with black solid lines and black dashed lines, respectively. The contrast, defined here as ${R_{max}}$ / ${R_{min}}$, is also plotted in Fig. 3(c) with solid red line. The maximum contrast of ${R_{max}}$ / ${R_{min}}$ = 5.6 and hence the maximum modulation depth in amplitude of 91% occurs at 1375 nm. At 1375 nm, the voltage-controlled reflection after normalization is shown in Fig. 3(d). The reflection intensity increases from 0.178 to 1 as the peak voltage increases from 0 V to 10 V. This corresponds to the white dashed line in Fig. 3(a).

 

Fig. 3. Characterization of electrical reconfigurability. (a) Measured and normalized reflection spectrum at different voltage levels. (b) The spectral movement of the resonance against voltages. (c) The maximum, minimum and contrast in reflection at different wavelengths. (d) The reflection modulation with voltage at 1375 nm.

Download Full Size | PDF

Numerical simulation was also carried out using FDTD. Figure 4(a) shows the reflection spectrum when LC director is at different angles (${\theta _{LC}}$). As ${\theta _{LC}}$ increases, the resonance experiences a blue shift, which is consistent with experimental results in Fig. 3(a). The spectral position of the maximum reflection is plotted in Fig. 4(b). It is evident that the resonance tuning range in simulation (1491 nm - 1601 nm) is slightly different from that in experiments (1441 nm - 1500 nm). The difference in spectral position is highly likely caused by the deviation in grating dimension introduced in the fabrication process. The tuning range in experiments (59 nm) is also smaller than that in simulation (110 nm). This is because the simulations were carried out under different ${\theta _{LC}}$, assuming LC directors are homogeneous in the cell gap, while experiments were conducted under different ${V_{peak}}$, in which LC is inhomogeneous due to the nonuniform distribution of ${\theta _{LC}}$ in the cell gap. Therefore, simulation provides an overestimate of the tuning range. The field distribution ${|E |^2}$ when ${\theta _{LC}}$ = 0° and ${\theta _{LC}}$ = 90° is also plotted at the wavelength of 1491 nm in Fig. 4(c). Stronger field within the grating can be observed when LC is switched from ${\theta _{LC}}$ = 0° to ${\theta _{LC}}$ = 90°. In addition, the nearfield is also enhanced by the LC switching, indicating that the nonlocal effect is responsible for the resonant reflection. In order to further verify the origin of the resonance reflection, the reflection spectrum for the device without metasurface was numerically calculated. The result is shown in Fig. 4(d). The resonant peak disappears in this case, substantiating that metasurface is essential for the resonant reflection. Similar results were experimentally obtained in [10], in which normal LC cell without metasurface showed no resonant reflection and trivial tunability.

 

Fig. 4. Simulation results. (a) Simulated reflection spectrum at LC director angles. (b) The spectral position of the maximum reflection. (c) The distribution of ${|E |^2}$ when ${\theta _{LC}}$ = 0° and ${\theta _{LC}}$ = 90° at the wavelength of 1491 nm, as is indicated by the white dashed line in (a). (d) Reflection spectrum for the device with and without grating structures.

Download Full Size | PDF

 

Fig. 5. Measurement of response time. (a) Temporal response of the switching on and switching off processes at the temperature of 70°C. (b) Extracted 10% / 90% response time at different measuring temperatures.

Download Full Size | PDF

4. Response speed

Response speed has been one of the most significant bottlenecks for LC-based devices, especially those using nematic LC. Only limited options are available for conventional devices to improve the response speed, including reducing the cell gap, using new LC material, and employing customized electric driving scheme [24]. Metasurface typically has subwavelength thickness (190 nm in this case), which makes ultrathin cell gap possible. This potentially brings huge benefits to LC devices, such as increased response speed and reduced electric fringing field in pixellated devices, as in the case of LCoS devices. In this paper, the response time was measured for the wavelength of 1400 nm, close to the wavelength of the highest switching contrast, with a filter with 12 nm bandwidth. Square-wave voltage with frequency 4 KHz and amplitude 10 V was applied to drive the LCAM cell. The cell transmittance was measured with an InGaAs Photodiode (SM05PD5A, Thorlabs) and an amplifier with 150 KHz bandwidth (Edmund Optics). The normalized transmission measured at 70°C is shown in Fig. 5(a). The response time $\tau $ (10% / 90%) for the switching on and switching off process are ${\tau _{on}}$ = 135.3 µs and ${\tau _{off}}$ = 288.5 µs, respectively. This enables a high operating frequency of more than 3.4 KHz. The response time of the LCAM cell at various measuring temperatures was also determined, as is shown in Fig. 5(b). For lower measuring temperatures, both response time (${\tau _{on}}$ and ${\tau _{off}}$) increases. However, they are maintained sub-millisecond, as is shown on the left axis of Fig. 5(b). It is also noted that the switching on process is always faster than the switching off process (${\tau _{on}}$ < ${\tau _{off}}$) because the restoring force within LC (which dominates the switching off process) is weaker compared to electric force (which dominates the switching on process). The difference in response time ($\Delta \tau $) is also plotted on the right axis of Fig. 5(b).

In Ref. [10], 20% / 80% response time was used, and the maximum operation frequency was 1.1 KHz. If 20% / 80% is used in this case, the maximum operation frequency is 5.5 KHz, 5 times of that in Ref. [10]. This is mainly attributed to the thinner cell gap in this case.

5. Conclusion

In this work, we demonstrated a fast-switching LC-based active metasurface device with a maximum operation frequency of 3.4 KHz. The ultrathin metasurface enables sub-micrometre cell gap. Moreover, no alignment material was required, making the fabrication process substantially simplified. This can potentially reduce the overall cost of making reconfigurable metasurface devices. The alignment quality was reasonably good in terms of contrast and the modulation depth of 91% was realized at telecommunication wavelength. This work summarized the key performance indicators of the LCAM device, showcasing the commercial potential of LC-based reconfigurable metasurface. It can be envisaged that metasurface will be a powerful tool for next generation LC-based devices, such as metasurface-integrated LCoS (meta-LCoS) devices.