Open Access
Add to BibSonomy
Abstract
We experimentally analyzed the internal reflection and loss of each component in a Si photonics frequency-modulated continuous-wave light detection and ranging (FMCW LiDAR) device using optical frequency domain reflectometry (OFDR) with a spatial resolution of better than 2.5 µm. Sweeping the incident laser wavelength by 120 nm, the reflections and losses of wire waveguides, widened waveguides, and optical switches on the chip were individually revealed. The slow-light grating (SLG) beam scanner, which has a limited working wavelength range, was evaluated with a spatial resolution of >10 µm by narrowing the wavelength sweep range. Consequently, a strong reflection was observed at the transition between the wire waveguide and the SLG, which can be a noise source in the FMCW LiDAR. Additionally, this study showed that the OFDR can be an important analysis tool for Si photonics integrated circuits. To our knowledge, this is the first demonstration, showing that the OFDR can be an important analysis tool for Si photonic integrated circuits.
© 2023 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement
1. Introduction
Silicon (Si) photonics technology integrates micro-optical components using a mature complementary metal-oxide-semiconductor process and has been used in data communications, sensing, and as a tool in research [1–4]. In recent years, solid-state nonmechanical light detection and ranging (LiDAR) using Si photonics has been studied actively and is expected to address conventional issues such as high cost, large size and weight, and instability. Here, Si photonics LiDARs are incompatible with the popular time-of-flight ranging method, which exploits high-intensity optical pulses, because of the two-photon absorption in Si at the wavelengths λ = 1.3–1.6 µm [5]. Therefore, the frequency-modulated continuous-wave (FMCW) method with a much lower light intensity is employed for Si photonics LiDARs [6–10]. The FMCW generates a temporally linear, frequency-swept optical signal and measures the distance of target objects from the beat frequency obtained via heterodyne detection of the returned light signal with a local reference light signal. By employing a triangularly frequency-swept signal, the velocity and vibration of objects can also be detected simultaneously. Optical phased arrays and focal plane arrays have been developed by many groups for use as nonmechanical beam scanners that can be integrated in the Si photonics FMCW LiDAR, and the LiDAR action has recently been demonstrated [8,9]. As an alternative, we have developed a device incorporating a slow-light grating (SLG) scanner, which enables high-resolution scanning using a much smaller number of components, thanks to the slow-light effect. We have reported large-scale, real-time, two-dimensional beam scanning and the acquisition of point cloud images of several thousand pixels [10–12].
In such Si photonics devices and circuits, the experimental characterization of each component on a fabricated chip is very important. In this paper, we performed the analysis and evaluation of our LiDAR chip using optical frequency domain reflectometry (OFDR). The principle of OFDR is similar to that for the FMCW LiDAR [13], where the frequency spectrum obtained via heterodyne detection of frequency-swept reflected light with local reference light provides the reflected light distribution. Because the range resolution of OFDR can be of micrometer order by extending the wavelength sweep range, it is quite effective for black-box analysis of large-scale photonic integrated circuits. It has been tested to Si photonics but only for grating couplers [14]. In this study, we evaluated the reflection points and insertion loss of the integrated components such as the singlemode wire waveguides, pseudo singlemode widened waveguides, Mach-Zehnder interferometer optical switches (MZI SWs), and the SLG. In particular, the reflection at the transition between the wire waveguide and the SLG were focused on as it can be a noise source in the FMCW ranging.
2. Device structure and setup
Figure 1 shows the top view of the fabricated LiDAR chip and its schematic. The red line shows the optical path considered in this measurement. Transverse-electric-polarized laser light is coupled from a polarization-maintaining (PM) core-shrunk fiber fixed by ultraviolet curing resin into a Si-inverse-tapered spot-size converter (SSC, A) with a tip width of 180 nm at the chip facet, and then the singlemode Si-wire waveguides of 400-nm width are wired. The thickness of Si was approximately 205 nm, which was common for all the following components. The light then passes through an ON/OFF MZI SW (B) and is split into a reference path and a signal path by a 2 × 2 coupler (C). The SW consists of wire waveguides and a 1 × 2 and a 2 × 2 coupler and is controlled via the thermo-optic (TO) effect using a TiN heater. The following SWs are identical. The couplers are multimode interference (MMI) type, which are designed to achieve a theoretical excess loss of less than 0.1 dB and a SW insertion loss of less than 0.2 dB. The reference light intensity is controlled by a similar MZI SW acting as an attenuator (ATT, E). The light signal is coupled into one SLG (H) selected by a left/right (L/R) SW (D), which determines the light incidence direction on the SLG, and a SW tree consisting of five-stage SWs (G). Between (D) and (G), zigzag waveguides are inserted to equalize the left and right optical path lengths. To suppress the loss in the straight sections as much as possible, wire waveguides and widened waveguides of 3-µm width are mixed through 30-µm adiabatic tapers. 32 SLGs, each consisting of a photonic crystal waveguide (PCW) with shallow-etched surface grating, are connected at the end of the SW trees and emit a fan beam into free space independently. This beam is scanned by directly applying a voltage to the p-i-p-doped Si heater in the SLG to produce the TO effect. When the temperature near the ends of the SLG becomes nonuniform, the light emission angle varies at different locations, resulting in beam divergence. To prevent this, the grating is omitted in the 100 µm range from two ends of the PCW. We call these regions nonradiative regions and the region with the grating, as radiative region. The SWs and the p-i-p heater were driven by a driver circuit controlled by a personal computer.
Fig. 1. (a) Fabricated LiDAR chip and (b) LiDAR chip schematic; the red line shows the optical path considered in the measurement.
Figures 2(a) and (b) show the measurement setup for the OFDR analysis. Light from a wavelength-swept source (Santec TSL-570) was input into the OFDR equipment (Santec OFDR-100) and coupled into the LiDAR chip via the PM core-shrunk fiber. The reflected light was returned to the OFDR equipment, and then the reflection spectrum was obtained, as shown in Fig. 2(c). The wavelength sweep range was set to λ = 1505–1625 nm (for a wavelength bandwidth 120 nm and a corresponding frequency bandwidth B = 14.7 THz). In general, the range resolution ΔR is given by c/2Bng for the speed of light c in vacuum and the group index ng in a medium. Therefore, ΔR = 10 µm/ng for this bandwidth. For example, ΔR = 2.3 and 0.5 µm for the typical ng = 4.3 for a Si-wire waveguide and ng = 20 for a PCW, respectively. However, the wavelength range mentioned above cannot be used fully due to the constraint of the group index bandwidth product of slow light. The working wavelength range of the PCW is approximately 20 nm for ng = 20 in the wavelength range mentioned above, and when the wavelength-swept range of the OFDR is limited to this range, the range resolution is 3 µm. Figure 2(c) shows an example of the reflection spectrum for ΔR = 10 µm/ng. The horizontal axis shows the optical path length converted to that in vacuum. The first and second reflection peaks correspond to the output port of the OFDR equipment and the fiber connector, respectively. The reflection intensity increased, and another peak appeared due to the connection to the PM core-shrunk fiber. Beyond this, the strong peaks were observed at the LiDAR chip, which are analyzed in detail in the following sections.
Fig. 2. (a) Measurement setup and (b) Equipment. (c) Measured reflection spectrum, where ΔR = 10 µm/ng.
3. Analysis of general-purpose components
Figure 3(a) shows the measured reflection spectrum of the LiDAR chip. The first reflection point (A) corresponds to the SSC, and the next two, (B) and (D), correspond to the ON/OFF SW and the L/R SW, respectively. At (C), between (B) and (D), the reflection intensity decreases by 6 dB due to the round-trip essential loss of the 2 × 2 coupler.
Fig. 3. (a) Measured reflection spectrum of the fabricated LiDAR chip, where ΔR = 10 µm/ng. (b) Magnified view of the reflection peak at point (F). Schematic of the waveguide bend is shown in the inset. (c) Magnified view of range (G). Light and dark lines show the raw data and moving average for 20 data points.
The reflection peak is observed at point (E) for the ATT in the reference path. The range (F) shows the zigzag waveguides. There are five turnarounds of wire waveguide bends, which caused five reflection peaks. Figure 3(b) shows the magnified view around the peak. As shown in the inset, the widened waveguide in the straight section is connected to the wire waveguide via the taper, and then folded back by the bend with a radius of 5 µm and connected to the widened waveguide via the taper again, and such a structure is repeated. The reflection intensity is weak for the widened waveguides and gradually increases toward a peak at the first taper and the bends. The peak at the second taper might be due to the slight excitation of higher order modes at the taper. However, the maximum intensity of the five peaks is almost unchanged. Moreover, the intensity is not so different from that around the SW consisting of wire waveguides in the front and rear sections. This suggests that the losses of the zigzag waveguide, including wire waveguides, bends, and tapers, are sufficiently small. The reflection intensity of the widened waveguides is approximately 25 dB lower than that of the wire waveguides. As the scattering loss is known to be inversely proportional to the cubic of the waveguide width [15], the scattering intensity is reasonable considering that the width of the two types of waveguides differs by a factor of 7.5. In the measurement of other test samples, including waveguides of different lengths, the propagation losses of the wire and widened waveguides were evaluated to be 3 and 0.3 dB/cm, respectively, and the loss of each taper, 0.02 dB. The losses shown in Fig. 3(a) are consistent with these values. Anyway, these results indicated that widened waveguides are effective for the overall reduction of on-chip losses. In fact, widened waveguides are employed in most of the long straight sections on this chip.
The range (G) in Fig. 3(a) corresponds to the SW tree, and Fig. 3(c) shows its magnified view. The round-trip loss of the SW tree, evaluated by taking a moving average of 20 data points to reduce the fading noise, was approximately 5 dB. This means that the single-pass insertion loss of one SW is 0.5 dB. In other words, 1 × 2 and 2 × 2 couplers might have approximately 0.2 dB loss, which are slightly larger than calculated minimum values of 0.06 dB for our design. As this difference causes a large accumulative loss in the multistage switches and the round-trip path, optimization for more robust and low-loss structures, even with fabrication errors, is desirable.
4. Analysis of SLG
The reflection point (H) in Fig. 3(a) corresponds to the SLG. However, as mentioned earlier, we estimated that the working wavelength range, Δλ, for low-dispersion slow light in the SLG is approximately 20 nm. Thus, the wavelength sweep range was reduced to 6 nm, which corresponds to a range resolution of 200 µm/ng. We measured the spectra again, shifting the wavelength range gradually, as shown in Fig. 4(a). The reflection intensity is overall low in the shortest wavelength range, which corresponds to the frequency range above the light line in the photonic band of the guided slow-light mode. Therefore, although light is slightly coupled to the SLG, it is emitted even in a nonradiative region and rapidly decreased. On the other hand, a strong reflection peak appears on the longest wavelength side because propagation is prohibited in the bandgap below the band-edge frequency of the PCW. The low and constant scattering intensity observed beyond the peak might be due to the coupling into a SiO2 cladding mode. The working wavelength range of the SLG is λ = 1523–1549 nm, and the short wavelength side (1523–1536 nm) produces low-dispersion slow light without being affected by the band-edge slow light. In this range, the strong reflection peak at the transition was suppressed and a smooth logarithmic decay was observed beyond this range. The decay slope increases from the nonradiative region indicated by gray color to the radiative region indicated by red color. But still a reflection peak remains at the transition.
Fig. 4. (a) Measured reflection spectra of the SLG, where ΔR = 200 µm/ng. Gray and red colored regions indicate nonradiative and radiative regions, respectively. (b) Normalized reflection spectra and fitting curves in the radiative region at λ = 1523–1529 nm and 1543–1549 nm.
Since the horizontal axis of the reflection spectra is converted to the optical path length in vacuum, the group index ng is derived from the ratio of the apparent length Lmeas and the actual length of the device L. Focusing on the radiative region, ng was obtained as Lmeas/L = 24 mm/1.5 mm ≈ 16. This value is reasonable because the designed ng was 15–20. Moreover, the round-trip losses in the radiative regions of 1.5 mm lengths at λ = 1523–1529 and 1543–1549 nm were 36 and 33 dB, respectively, which indicate the losses of the SLG to be 120 and 110 dB/cm, as shown in Fig. 4(b). The propagation loss of the PCW can also be obtained from the slope in the nonradiative region; however, its length (100 µm) is not sufficient to evaluate the propagation loss accurately. Therefore, we assumed a PCW loss of 30 dB/cm, which was evaluated separately by comparing the transmissions of PCWs of different lengths. Then, the radiation coefficient of the radiative region was evaluated to be 80–90 dB/cm, which is similar to the designed value.
Figure 5(a) shows the reflection spectra of the LiDAR chip for the limited wavelength range of λ = 1523–1529 nm. Although the reflection intensity at the transition was larger than that of the SSC in Fig. 3(a), it became lower than that of the SSC when a suitable wavelength range was set. However, the intensity is still slightly higher than that of the SW in the front, as shown at the bottom. Figure 5(b) shows the reflection intensity of the transition (red) and the SW (black) at each wavelength sweep range. Focusing on the working wavelength range of the SLG colored as gray, the reflection intensity of the transition is 2–8 dB higher than that of the SW on the short wavelength side and increases gradually toward the long wavelength side. In particular, the slope becomes large from λ = 1539–1545 nm, which suggests the effect of the band-edge slow light.
Fig. 5. (a) Measured reflection spectrum of the LiDAR chip and magnified view around the transition of SLG, where ΔR = 200 µm/ng. (b) Peak reflection intensity at the transition toward SLG (red) and at the SW (black) for different wavelength sweep ranges of 6 nm width. The gray region shows the working wavelength range of the SLG. The horizontal axis shows the wavelength sweep range in OFDR.
5. Discussion
Considering the principle of the FMCW LiDAR, it is important to suppress the noise from the laser source but difficult to completely suppress its influence even with balanced photodetection because light returned from the internal reflection points on the chip follows the same path as the signal light. This limits the signal-to-noise ratio and the measurable distance. Therefore, suppression of the reflection is a crucial issue. The reflection spectra show that strong reflection occurred at the transition between the input wire waveguide and the SLG even though the reflection intensity decreased with the increasing distance of the reflection points due to the round-trip loss on the chip. The transition structure employed in the measured chip was a slightly modified version of simple tapered type whose calculated and measured losses were 0.28 and 0.46 dB, respectively [16]. The back reflection of such transition was calculated to be 16% of the total loss amount, implying a 1.4% reflectivity. This transition structure has been optimized using automatic optimization algorithms previously, showing a theoretical loss of 0.12 dB [17,18]. If such a low loss is realized, the reflectivity will be suppressed to 0.4%.
Finally, we highlight the effectiveness of OFDR in characterizing photonic components and integrated circuits. Until now, test element groups (TEGs) for each component have been fabricated on separate chips, light has been coupled into devices with different lengths or numbers one by one, and the optical characteristics such as loss have been evaluated by comparing their transmission. This is a time-consuming task, and accuracy is limited because unstable coupling of light fluctuates the transmission intensity. Moreover, the transmission in large-scale integrated chips such as our LiDAR, which has many components cascaded in series, is subject to cumulative losses and optical properties, making individual evaluation impossible. Therefore, the transmission of such integrated circuits has been evaluated indirectly by multiplying the transmission of the TEG components. However, the detailed characteristics of each component may differ from those of TEG due to layout and proximity effect differences in the fabrication process. Small loss differences add up and can become unignorable. OFDR can evaluate each component independently from the spatial distribution of reflections obtained in a single measurement, greatly reducing the task and improving the accuracy for both TEG and integrated chips (note that there are no other methods for integrated chips). One drawback is the need for wide wavelength sweep. The sweep range must be narrower than the operating range, not just SLGs. The flat operating range of each component must be known in advance. Reducing the sweep range degrades spatial resolution. However, by setting the appropriate wavelength range, a good overview of the characteristics can be obtained even for components with the limited operating range like SLG. In any case, OFDR is a very useful tool for characterizing Si photonics devices.
Funding
New Energy and Industrial Technology Development Organization (JPNP14004); Japan Society for the Promotion of Science (22H00299).
Disclosures
The authors declare no conflicts of interest.
Data availability
Data underlying the results presented in this paper are not publicly available at this time, but may be obtained from the authors upon reasonable request.
References
1. B. Jalali and S. Fathpour, “Silicon Photonics,” J. Lightwave Technol. 24(12), 4600–4615 (2006). [CrossRef]
2. R. Soref, “The Past, Present, and Future of Silicon Photonics,” IEEE J. Select. Topics Quantum Electron. 12(6), 1678–1687 (2006). [CrossRef]
3. S. Y. Siew, B. Li, F. Gao, H. Y. Zheng, W. Zhang, P. Guo, S. W. Xie, A. Song, B. Dong, L. W. Luo, C. Li, X. Luo, and G.-Q. Lo, “Review of Silicon Photonics Technology and Platform Development,” J. Lightwave Technol. 39(13), 4374–4389 (2021). [CrossRef]
4. J. X. B. Sia, X. Li, J. Wang, W. Wang, Z. Qiao, X. Guo, C. W. Lee, A. Sasidharan, S. Gunasagar, C. G. Littlejohns, C. Liu, G. T. Reed, K. S. Ang, and H. Wang, “Wafer-Scale Demonstration of Low-Loss (∼0.43 dB/cm), High-Bandwidth (>38 GHz), Silicon Photonics Platform Operating at the C-Band,” IEEE Photonics J. 14(3), 1–9 (2022). [CrossRef]
5. A. D. Bristow, N. Rotenberg, and H. M. van Driel, “Two-photon absorption and Kerr coefficients of silicon for 850-2200 nm,” Appl. Phys. Lett. 90(19), 191104 (2007). [CrossRef]
6. P. A. M. Sandborn, “FMCW Lidar: Scaling to the Chip-Level and Improving Phase-Noise-Limited Performance,” Doctoral Dissertation, Univ. California Berkeley (2017).
7. A. Martin, D. Dodane, L. Leviandier, D. Dolfi, A. Naughton, P. O’Brien, T. Spuessens, R. Baets, G. Lepage, P. Verheyen, P. D. Heyn, P. Absil, P. Feneyrou, and J. Bourderionnet, “Photonic Integrated Circuit-Based FMCW Coherent LiDAR,” J. Lightwave Technol. 36(19), 4640–4645 (2018). [CrossRef]
8. C. V. Poulton, M. J. Byrd, P. Russo, B. Moss, O. Shatrovoy, M. Khandaker, and M. R. Watts, “Coherent LiDAR with an 8,192-Element Optical Phased Array and Driving Laser,” IEEE J. Select. Topics Quantum Electron. 28(5), 1–8 (2022). [CrossRef]
9. C. Rogers, A. Y. Piggott, D. J. Thomson, R. F. Wiser, I. E. Opris, S. A. Fortune, A. J. Compston, A. Gondarenko, F. Meng, X. Chen, G. T. Reed, and R. Nicolaescu, “A universal 3D imaging sensor on a silicon photonics platform,” Nature 590(7845), 256–261 (2021). [CrossRef]
10. T. Baba, T. Tamanuki, H. Ito, M. Kamata, R. Tetsuya, S. Suyama, H. Abe, and R. Kurahashi, “Silicon Photonics FMCW LiDAR Chip With a Slow-Light Grating Beam Scanner,” IEEE J. Select. Topics Quantum Electron. 28(5), 1–8 (2022). [CrossRef]
11. H. Ito, Y. Kusunoki, J. Maeda, D. Akiyama, N. Kodama, H. Abe, R. Tetsuya, and T. Baba, “Wide beam steering by slow-light waveguide gratings and a prism lens,” Optica 7(1), 47–52 (2020). [CrossRef]
12. T. Tamanuki, H. Ito, and T. Baba, “Thermo-Optic Beam Scanner Employing Silicon Photonic Crystal Slow-Light Waveguides,” J. Lightwave Technol. 39(4), 904–911 (2021). [CrossRef]
13. R. I. MacDonald, “Frequency domain optical reflectometer,” Appl. Opt. 20(10), 1840–1844 (1981). [CrossRef]
14. M. Tokushima and J. Ushida, “Demonstration of in-depth analysis of silicon photonics circuits using OFDR: waveguides with grating couplers,” Opt. Lett. 47(1), 162–165 (2022). [CrossRef]
15. D. Marcuse, Light Transmission Optics. (Van Nostrand Reinhold, New York, 1972).
16. Y. Terada, K. Miyasaka, K. Kondo, N. Ishikura, T. Tamura, and T. Baba, “Optimized optical coupling to silica-clad photonic crystal waveguide,” Opt. Lett. 42(22), 4695–4698 (2017). [CrossRef]
17. R. Shiratori, M. Nakata, K. Hayashi, and T. Baba, “Particle swarm optimization of silicon photonic crystal waveguide transition,” Opt. Lett. 46(8), 1904–1907 (2021). [CrossRef]
18. K. Hirotani, R. Shiratori, and T. Baba, “Si photonic crystal slow-light waveguide optimized by informatics technology,” Opt. Lett. 46(17), 4422–4425 (2021). [CrossRef]
