1. Introduction

Si photonics achieves high-performance optical devices and allows their large-scale photonic integration via a complementary metal oxide semiconductor (CMOS) process. Recently, researchers attempted to realize a compact three-dimensional sensor, i.e., light detection and ranging (LiDAR), based on Si photonics for use in advanced driver assistance systems (ADAS), robots, drones, and various internet of things devices [1]. A crucial component for such LiDAR is a non-mechanical beam-steering device. Conventional mechanical devices have disadvantages such as bulkiness, high assembly cost of components, low speed, and low tolerance to mechanical vibrations. Microelectromechanical system mirrors [2,3] can be cheaply mass-produced; however, if their diameter is designed to be several millimeters to obtain a sufficiently sharp optical beam, their scanning speed is significantly reduced to below 1 kHz. Therefore, non-mechanical solid-state devices such as optical phased arrays (OPAs) [4–8] and waveguide diffraction gratings [9–11] were studied.

OPAs consisting of arrayed optical antennas were expected to form an arbitrary beam pattern via interference between radiated light from antennas with appropriate phase differences. However, a slight phase error in each antenna significantly degrades the beam quality. Searching for and correcting different phase errors in all the antennas from a so-formed beam profile is a complex and time-consuming process, particularly when the scale of the array is above several hundreds. Because of this, it remains challenging to obtain high resolution, which is achieved using a sharp beam with wide range steering. Another problem associated with OPAs with respect to high-efficiency beaming is the suppression of higher order diffraction. For this, a narrow antenna pitch close to the optical wavelength is required, but this increases unwanted optical coupling and thermal crosstalk between the antennas, whose phase is controlled by thermo-optic (TO) heaters. Intentionally disordering the array arrangement has suppressed the higher order diffraction peaks after complex optimization [8]. However, this does not concentrate unwanted power components into the main beam but only equalizes them. Therefore, this does not improve the efficiency but merely raises the background noise level; the demonstrated peak to background ratio did not surpass 10 dB. These will be critical problems for LiDAR, which requires high efficiency and a wide dynamic range.

Waveguide diffraction gratings have much simpler structures and allow a sharp beam to be formed and steered without complex control. However, the beam steering range is too narrow. Due to their small wavelength and refractive index sensitivity of the steering angle, a broadband wavelength-swept laser source with Δλ > 70 nm and/or a large index change of Δn > 0.5 is necessary for obtaining a steering angle of Δθ > 10°.

Another common and crucial issue for any type of LiDAR is constraint of the sensing throughput by the round-trip time of light. When sensing an object at a distance of 150 m or more, the round-trip time is of the order of μs. This means that, if the number of pixels in the acquired image is, e.g., 100 k, the frame rate of LiDAR cannot exceed 10 fps even if the sensing duration is ignored.

To solve these problems, we propose and demonstrate here the allocation of different wavelengths to each beam-steering range using wavelength-division multiplexing (WDM). Consequently, even if the steering angle of each wavelength is small for the index change, a wide overall angle can be covered. Furthermore, the sensing throughput is improved even when detecting at a long distance by parallel operation at different wavelengths. In this study, we adopted a doubly periodic two-dimensional (2D) bulk photonic crystal waveguide (BPCW), which acts as a beam-steering device operated by TO heaters without complex tuning. In this study, we observed one-dimensional (1D) steering of the spot beam in the WDM scheme by integrating coupled microring wavelength multiplexers (MUXs) and loading a collimator lens. We finally achieved 2D beam steering by switching one BPCW from the array. The number of resolution points in this experiment was found to be 448.

2. Design and fabrication

Figure 1 shows a schematic of the proposed WDM optical beam steering device. Light of different wavelengths is coupled to an input Si wire waveguide with transverse electric (TE) polarization, multiplexed to a bus waveguide through each coupled microring MUX, coupled to the BPCW via a taper, and radiated into the air by doubly periodic structural modulation. The figure also shows devices fabricated via a 200 mm silicon-on-insulator (Si layer thickness 210 nm, SiO₂ BOX layer thickness 2 μm) and a Si photonics CMOS process with a minimum feature size of < 130 nm obtained by combining KrF excimer laser exposure with a phase shift mask.

 

Fig. 1 WDM beam steering device. (a) Schematic. (b)–(e) Fabricated device. (b), (c) WDM MUX circuit. (d), (e) Thermally controlled doubly periodic BPCW beam steering device.

Download Full Size | PDF

The details of the coupled triangular microring MUX are described in [13]. This is composed of a 400 nm wide Si wire waveguide embedded with SiO₂ cladding. Each side of the triangle is designed with a 5 μm straight waveguide and a 120° bent waveguide with a 2 μm radius, and a 20 nm offset is introduced at the connection between the straight section and the bend to reduce bending loss. The round-trip length of the microring is 27.6 μm, and the corresponding free spectral range (FSR) is 19 nm at λ ≈1.55 μm. A TiN heater for phase tuning, utilizing the TO effect with a resistance of 70 Ω and a rated power of 18 mW, is located in the SiO₂ cladding on one side of the triangle of each microring. The temperature shift at the center of the waveguide cross section of the microring due to heating, calculated by the finite element method (FEM), is dT/dP ≈10 K/mW (this value varies slightly depending on the position). The TO coefficient of the Si index is dn_Si/dT = 1.86 × 10⁻⁴ RIU/K. If n_Si = 3.45 and n_SiO2 = 1.45 at λ ≈1.55 μm, the modal equivalent index change for the change in n_Si is calculated as dn_eq/dn_Si ≈1.05. Using these relations and the experimentally evaluated relation between the change in n_eq and the wavelength shift, dλ/dn_eq = 570 nm/RIU, we finally obtain dλ/dP = 1.1 nm/mW. Since the wavelength shift when heated at the rated power is ~20 nm, the entire FSR can be covered. We designed a Butterworth transfer function with a box-like spectral response at the passband, and the full width at half maximum Δλ = 2.4 nm, where the power coupling coefficient between the bus waveguide and microring, κ_b, and that between microrings, κ_r, were set at 0.4 and 0.08, respectively. The optimum inter-waveguide gaps estimated by three-dimensional finite-difference time-domain calculations were 204 nm for the former and 347 nm for the latter.

As shown in Fig. 2(a), the BPCW is formed by arranging perforated circular holes in a square lattice in a wide Si waveguide of width W = 9.8 μm and length L = 980 μm embedded in SiO₂ cladding. Compared with a shallow-etched diffraction grating, the BPCW does not require precise etch depth control, so high fabrication tolerance is expected. Furthermore, due to light radiation into the air, the hole diameter was modulated as 2r₁ = 2r₀ + Δr and 2r₂ = 2r₀ − Δr with a double period of twice the lattice constant a of the photonic crystal along the light propagation direction, similar to that in [14]. The light radiation coefficient can be controlled by the hole radius difference, Δr = r₁ − r₂, and an appropriate radiating aperture length was obtained. Figure 2(b)–2(e) shows calculated photonic bands showing the relationship between the normalized frequency a/λ and propagation constant β, the group index n_g spectrum, the beam angle θ in the propagation direction (longitudinal direction), and the radiation coefficients for Δr = 5, 7.5, and 10 nm, respectively. In the photonic band calculation, we adopted the unit cell model shown by the orange rectangle in Fig. 2(a). Targeting the vicinity of the band edge at λ ≈1.55 μm, we set a = 380 nm and 2r₀ = 190 nm. The angle θ is given by θ = sin⁻¹(β/k₀ − λ/2a), using β and the wave number in the air, k₀ ≈2π/λ [12]. Even when Δr is changed, the operating band hardly changes, and a linear band with n_g ≈4 is generated. In this structure, since the remarkable slow-light effect described in [12] does not appear, the wavelength sensitivity of θ is dθ/dλ = 0.15°/nm. Since n_g abruptly increases at longer wavelengths approaching the band edge, the sensitivity also increases to 0.61°/nm.

 

Fig. 2 Theoretical characteristics of BPCW beam steering device. (a) Schematic. (b) Photonic band. The gray area represents the light cone of the SiO₂ cladding. (c) Group index n_g spectrum. (d) Beam angle θ in longitudinal direction. (e) Radiation coefficient α_rad.

Download Full Size | PDF

The angle θ depends on the refractive index n_Si of the Si layer as well as the wavelength [15]. The refractive index sensitivity is dθ/dn_Si = (λ/n_Si)(dθ/dλ) ≈90°/RIU from the wavelength sensitivity. From the refractive index TO coefficient of Si, dθ/dT ≈90 × 1.86 × 10⁻⁴ = 0.017°/K. As shown in Fig. 1(d), a TiN heater (resistance 1.2 kΩ, rated power 9.6 W) was located on both sides of the BPCW in the fabricated device. Therefore, the BPCW can be heated by injecting current into the heater in the direction along the BPCW. Using the FEM calculation, ΔT = 400 K was estimated for heating at 2.9 W and considering the above index sensitivity, Δθ = 6.8° was expected. This heating is not so efficient and fast because the BPCW is wide and heated indirectly from the heaters buried in the SiO₂ cladding having a low thermal conductivity. The efficiency and speed will be improved by forming the heaters on the Si layer having a much higher thermal conductivity and employing a rib-type structure to define the lateral optical mode of the BPCW, so that the heat is conducted through the Si layer to the BPCW.

The radiation coefficient α_rad in Fig. 2(e) was optimized for a BPCW length of 1.4 mm by using the method described in [14]. The coefficient varies significantly with a small change in Δr and can be adjusted within the range 20–200 dB/cm. From the calculation result given in [14], if we set α_rad = 100 dB/cm, the loss due to light scattering inside the fabricated BPCW with some disordering (α_scat = 35 dB/cm is assumed), and light passing through without radiation from the BPCW, can be reduced to 1 dB or less; furthermore, it is possible to suppress the longitudinal divergence angle δθ (full width at half maximum) to <0.09°. If the waveguide length L is set to extend to 3 mm, it is possible to suppress δθ to <0.04°. In the experiment described below, due to the radiation coefficient varying slowly with wavelength, we set Δr = 7.5 nm.

3. Measurement

The continuous-wave light from the tunable wavelength laser source was controlled to the TE polarization and coupled to a spot size converter (SSC) at the end of the chip. The coupling loss from the fiber to the Si wire waveguide via the SSC was about 2 dB. We evaluated the demultiplexing characteristics of the coupled microring MUX, as shown in Fig. 3. Here, the blue and red lines represent the through and drop spectra, respectively, and the light-colored lines represent those of the as-fabricated device. Since the resonance wavelengths of the two microrings were slightly detuned, the drop spectrum was split, and the transmission intensity decreased. When the resonance wavelength was corrected with the heater for each microring, the box-like spectrum shown by the dark-colored line in the magnified view of Fig. 3(b) was obtained, and the maximum intensity improved to −1.4 dB at this point. As described above, the operating wavelength can be adjusted while maintaining this box-like spectrum.

 

Fig. 3 Through spectrum (blue line) and drop spectrum (red line) of coupled microring MUX. (a) Wide wavelength range. (b) Magnified view of λ ≈1.55 μm. The light and dark colors represent before and after thermal tuning, respectively.

Download Full Size | PDF

Figure 4 shows the beam-steering characteristics of the fabricated BPCW; Fig. 4(a) shows the near-field pattern of light radiation observed with an InGaAs camera (Hamamatsu Photonics, C14041-10U, QVGA, pixel pitch 20 μm) from the top of the device through a microscope at wavelength λ = 1.55 μm, which is almost the center of the propagation band of the linear photonic band. Gradual attenuation of propagated light was observed, and the decay coefficient obtained from curve fitting was α = α_rad + α_scat = 186 dB/cm, which was slightly larger than the calculated value, even considering that α_loss = 35 dB/cm was evaluated independently for a device without double periodicity. This might be due to the hole radius difference Δr of the fabricated device being slightly above the designed value, which increased α_rad. Figure 4(b) shows the corresponding far-field pattern (FFP), which was observed using a far-field microscope with an InGaAs camera (Raptor, OWL1280 BIS-SWR, SXGA, pixel pitch 10 μm) having a high resolution of 0.0288°. It can be seen that a sharp beam, δθ < 0.2°, is formed in the longitudinal direction. The profile perpendicular to the propagation direction (lateral direction) was evaluated by integrating within the range δθ = ± 10° from the peak value. The angle spread in the lateral direction, δϕ, was within a range of approximately ± 10°, reflecting a waveguide width of W = 9.8 μm. Figure 4(c) shows 1D beam steering in the θ-direction observed by wavelength sweep. The wavelength sensitivity was dθ/dλ ≈0.2°/nm, which roughly corresponded to the theoretical value shown in Fig. 2(d). Figure 4(d) shows the beam steering caused by heating when the wavelength is fixed at λ = 1.55 μm. Since the photonic band shifts to lower frequency due to heating, it follows the same operating principle as the short wavelength shift at the wavelength sweep. A beam steering of Δθ = 4.4° was observed with a heating power of 7.4 W, and the sensitivity to heating was dθ/dP ≈0.6°/W. These values can be increased by employing either a structure that can obtain an even higher n_g or a heater close to the waveguide core. Figure 4(e) shows a comparison of the beam divergence angles δθ for wavelength sweeping and heating. In the vicinity of the band edge on the long wavelength side, the radiation coefficient reached >500 dB/cm, and δθ increased as the radiation aperture length decreased. However, δθ was 0.14° on the short wavelength side corresponding to the linear band. This was consistent with the diffraction limit δθ obtained from the radiation coefficient shown in Fig. 4(a). The fan-shaped beam spread in the ϕ-direction can be collimated to a spot beam by aligning a cylindrical lens above the chip in the direction along the waveguide [14]. Figure 4(f) shows the spot beam collimated with a rod lens with a diameter of 5 mm and a focal length f = 3.8 mm for various wavelengths. Sharp spots were formed, in which the lateral beam divergence was δϕ < 0.1° on average.

 

Fig. 4 Characteristics of BPCW. (a) Top view of device and NFP for λ = 1.55 μm. The lower graph is the intensity position profile, attenuated exponentially. (b) FFP for λ = 1.55 μm. θ-direction angular profile by (c) wavelength sweep and (d) heating BPCW. (e) Comparison of beam divergence angles. (f) Steering of spot beam formed with rod lens.

Download Full Size | PDF

To observe the WDM optical beam-steering operation, the 4-ch WDM MUX circuit was adjusted with each heater. CW light at different wavelengths was coupled into the chip through each SSC and multiplexed to the bus waveguide through the coupled microring MUX. Since the same design was applied to microring MUXs of all channels, the resonance wavelengths were observed as shown in Fig. 5(a). The drop spectrum was deformed by slight detuning between two microrings. Therefore, the TiN heater in each microring was turned on, and box-like spectra were obtained as shown in Fig. 5(b). Four channels were aligned in a FSR of 20 nm by controlling the heating power.

 

Fig. 5 Thermal tuning of 4-ch WDM circuit. (a) As fabricated. (b) After tuning. Each color corresponds to one channel.

Download Full Size | PDF

Figure 6(a) shows a FFP of the radiation beams observed when each channel was set to a resonance wavelength indicated by each color band shown in Fig. 5(b). The beams of different channels were radiated at different angles, and the angles were further controlled by heating the BPCW. Figure 6(b) shows the change in the angular beam profile in the θ-direction observed in each channel when heated in the range 0–7 W. Each color represents a given wavelength, and the heating power (the number on the right) is increased in order from the top. Although there were slight differences in the wavelength sensitivities of the steering angles due to the different n_g in each channel, we obtained beam-steering reaching 4°–5° for each channel, and Δθ = 16° in total with simple control of the heating power. The beam divergence of each channel at this time is shown in Fig. 6(d). The average δθ < 0.15° for all channels, and the number of resolution points N = Δθ/δθ > 107. If the number of WDM channels is increased, the Δθ range can be expanded accordingly, and simultaneous parallel operation is also possible. Figure 6(c) shows the change of θ with the heating power of each channel P_i. Since the operating wavelength of each channel was initially set close to the photonic band edge, the θ−P_i characteristics were slightly different between channels. Although they exhibited a small nonlinearity, it can be compensated by calibrating the heating power. Figure 6(e) shows the superposition of FFPs of the respective channels when the rod lens is aligned on top of the waveguide. A spot beam was formed, and beam steering was clearly confirmed. The end and start points of different wavelengths were shifted because the inclinations of the BPCW and the rod lens were not completely parallel. It will be reduced by fixing all the component on a sub-mount carrier.

 

Fig. 6 4λ-multi-beam steering. (a) Observed FFPs of radiated beam. (b) Longitudinal angular profile at each channel for heating power P_i. (c) θ–P_i characteristics. (d) Longitudinal divergence angle. (e) Steering of spot beam formed by rod lens. Threshold processing was performed to suppress the background.

Download Full Size | PDF

We also integrated multiple BPCWs and selected one of them using a 1 × 4 heater-controlled Mach-Zehnder optical switch [14]. Due to the change in the relative position between each PCW and the collimator lens, the beam was steered in the ϕ-direction as shown in Fig. 7. In this experiment, the spots were clearly separated in the ϕ direction, demonstrating the lateral beam steering. This separation can be easily controlled by changing the pitch between the BPCWs and the focal length of the collimator lens. Since the photonic bands of the four BPCWs were not perfectly the same due to fabrication errors, their θ were different by 0.4°–0.9° even at the same wavelength (see the start points slightly shifted between four blue plots, for example). This difference can easily be compensated by calibrating their heater powers. In this measurement, N was confirmed to be above 112 × 4 = 448.

 

Fig. 7 2D beam steering observed by a combination of the wavelength sweep and switching of the BPCW.

Download Full Size | PDF

4. Discussion

We have proposed and demonstrated a WDM beam-steering device integrating thermally controlled doubly periodic BPCWs and coupled microring MUXs. By dividing the steering range into channels operating at different wavelengths, a wide total steering range can be covered, even if each steering angle is small. Furthermore, if we apply this to sensors such as LiDAR, it is possible to improve the sensing throughput, which is usually constrained by the round-trip time of light, by simultaneous parallel sensing.

In this study, we fabricated a 4-ch WDM device and observed beam-steering reaching 4°–5° for each wavelength and 16° in total for four wavelengths, which achieved N > 107 1D resolution points. We also demonstrated 2D steering by selecting one BPCW from a four-BPCW array, achieving N > 448. We will be able to increase N further by the following three improvements. (1) By elongating the BPCW to 3 mm, the beam divergence will be reduced to <0.04°. (2) The beam angle takes a negative value when light is incident on the opposite end of the BPCW. Therefore, the beam-steering range is easily doubled by switching the direction of light incidence. (3) The BPCWs can easily be increased; a 64 BPCW array can be integrated in a 3 × 3 mm2 footprint. These improvements will increase N by 2.9, 2, and 16 times, respectively, resulting in N > 41,000. Even for such a large N, precise control such as that required for OPAs is unnecessary, and single beam formation with a wide dynamic range can be realized by optimizing the fabrication process. These are significant advantages for LiDAR applications requiring a large dynamic range.

Finally, we discuss whether such a WDM device can be used for LiDAR, mounted on an ADAS system, a robot, or a drone, for example, in a severe temperature environment. Assuming that a GaInAsP semiconductor distributed feedback laser is used as a light source, the temperature dependency of the operating wavelength is approximately 0.1 nm/K. Regarding the Si photonics device, however, the optical field from Si penetrates the SiO₂ cladding, having a reduced temperature dependency on the refractive index, so the temperature dependency of the operating wavelength is slightly reduced to 0.08 nm/K. Therefore, considering that the temperature fluctuates in this environment, a deviation of 0.02 nm/K occurs between the laser and Si photonics components. The transmitted bandwidth of the coupled microring MUX design in this study was 2 nm, and if we assume that a deviation of ± 1 nm from the center wavelength is allowed, the allowable temperature change is ± 50 K. Although it is necessary to adjust the temperature to some extent to stabilize the laser, it is still considered that the proposed WDM device works well with considerably coarse adjustment of the order of ± 10 K.

5. Conclusion

We proposed and demonstrated a WDM optical beam-steering device consisting of a thermally controlled doubly periodic Si BPCW and coupled microring MUXs. Beam forming and steering while maintaining a sharp profile is much easier in this device than with optical phased arrays. By dividing the range of beam-steering angles into different wavelength channels, it is possible to cover a wide range of angles, even when each angle is small. In this study, we fabricated a device with four wavelength channels, each of which showed beam steering of 4°–5° as a result of heating, resulting in a total of 16°. Two-dimensional steering is also achieved by loading a collimator lens and selecting one waveguide from those arrayed. We evaluated 112 resolution points with four wavelengths and 448 points in total by switching four waveguides. If this WDM concept is introduced into light detection and ranging and the number of wavelengths is increased, it will be possible to increase the sensing throughput, which is usually constrained by the round-trip time of light, by simultaneous parallel operation.