Open Access
Add to BibSonomy
Abstract
We experimentally demonstrate a polarization-insensitive optical filter (PIOF) using polarization rotator-splitters (PRSs) and microring resonators (MRRs) on the silicon-on-insulator (SOI) platform with complementary metal-oxide-semiconductor (CMOS) compatible fabrication process. The PRS consists of a tapered-rib waveguide and an asymmetrical directional coupler (ADC), which realize the polarization rotation and splitting, to ensure the connected MRRs-based optical filter operating at one desired polarization when light with different polarizations are launched into the device. The measured results show that the optical transmission spectra of the device are identical for TE and TM polarization input. The box-like filtering spectra are also achieved with a 3-dB bandwidth of ∼0.15 nm and a high extinction ratio (ER) over 30 dB.
© 2022 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement
1. Introduction
Optical interconnects possess the advantages of low loss, low dispersion, and absence of parasitic phenomena compared to electrical interconnects, making it very promising in increasing the link capacity of information interconnects [1–3]. With the opening of the post-Moore era, the silicon-on-insulator (SOI) platform has become one of the most applicable optical platforms to implement optical interconnects owing to its high integration density, low cost, and compatibility with mature complementary metal-oxide-semiconductor (CMOS) process [4]. Silicon photonic devices based on the SOI platform are ultra-compact due to the high refractive index contrast between Si and SiO2. However, the high refractive index contrast results in large birefringence of the waveguides [5,6], which makes silicon photonics devices polarization-sensitive. Silicon-based optical filter that provides ultra-fast linear photonic signal processing is one of the most critical devices for realizing various optical interconnect applications. Hence, it is quite important to achieve polarization-insensitive optical filters (PIOFs) on the SOI platform.
Several approaches are investigated to develop the PIOFs [7–15]. The direct way to solve polarization-dependent problem is using square waveguides. Deng et al. utilized the method with Mach-Zehnder interferometers (MZIs) to design PIOF, which can make TE and TM modes have same effective index. However, the nonstandard square waveguides are always obtained after the lithography and etching process, which still results in polarization-sensitive issues. Moreover, the scheme using MZIs-based filter has a large footprint [7].
Some reported works implemented the PIOF using gratings without additional polarization control components, which have compact structures but require a very high fabrication accuracy [8–10]. Yun et al. demonstrated a PIOF with phase-shifted polarization-rotating Bragg gratings (PRBGs). As a result of the fabrication deviation, the out-of-band rejection ratios of two sides around the resonance are inconsistent [8]. Okayama et al. also used the PRBG to realize the PIOF. Nonetheless, the fabrication processes of the non-vertical waveguides and asymmetrical grating system are difficult [9]. Liu et al. utilized dual-gratings to obtain the PIOF with a large 3 dB-bandwidth of ∼11 nm, however, the filtering center wavelengths of TE and TM polarizations have a difference of 0.7 nm, which may be caused by fabrication errors [10].
Another popular solution is to explore the polarization manipulating devices working in combination with optical filters [11–15]. Xiang et al. implemented a PIOF with two-dimensional (2D) grating coupler, which is equivalent to a polarization splitter, coupling the two orthogonal modes from a fiber into two paths of the device. The structure is ultracompact, but the PIOF’s extinction ratio (ER) is degraded and nonuniform due to the low coupling efficiency of 2D grating coupler as well as the fabrication deviation [11]. Fukuda et al. realized a PIOF using an off-axis double-core structure-based polarization rotator in combination with a microring resonator (MRR). The device suffers from the stringent requirement of high fabrication accuracy since the square waveguide is utilized to rotate the polarization [12]. The PIOF demonstrated by Li et al. in 2011 also faced the similar problem in fabrication accuracy because of the selection of L-shape waveguides as the polarization rotators [13]. In 2018, Wang et al. implemented a PIOF using multiple groups of 2nd-order MRRs together with polarization manipulating components. The device achieves a large free spectral range (FSR) of ∼90 nm at the sacrifice of the increased complexity in device configuration, which results in a relatively large footprint [14]. In addition, extra power consumption is required to compensate the fabrication induced filtering wavelength variations of multiple groups of optical filters.
In this work, we experimentally demonstrate a PIOF on the SOI platform using polarization rotator-splitters (PRSs) and 2nd-order MRRs. Only one MRRs-based optical filter is utilized in the PIOF since the PRSs are adopted to covert different polarizations to the desired state. It prohibits the fabrication induced uniformity problem caused by using multiple groups of optical filters [14]. The device is systemically designed by theoretical analysis and numerical simulation to achieve the box-like spectra response with a high ER, whilst possessing the compact size and relatively easy fabrication process. Thereafter it is fabricated on an 8-inch SOI wafer using CMOS-compatible technology. The measurement results show that the PIOF exhibits well polarization-insensitive feature and good filtering performance.
2. Working principle and device design
The schematic configuration of the proposed PIOF in three-dimensional (3D) view and top view are shown in Figs. 1(a) and 1(b), respectively. The PIOF comprises two PRSs and 2nd-order MRRs. The PRS is composed of a tapered-rib waveguide (L1 and L2 sections) and an asymmetrical directional coupler (ADC, L3 section). The blue and orange regions represent the fully and partially etched silicon waveguides, respectively. The working principle of the PIOF is described as follows. As shown in Fig. 1, when the fundamental TE (TE0) mode is launched into the input port P1, the light propagates along the tapered-rib waveguide and the upper waveguide of the ADC without mode conversion, then it is coupled to the MRRs-based optical filter and drops at port PT1. At last, the filtered TE0 mode light is received at output port P6. As for the fundamental TM (TM0) mode launched into port P1, it is converted to first-order TE (TE1) mode through the tapered-rib waveguide. The TE1 mode is then converted to TE0 mode as the light is progressively coupled to the lower waveguide of the ADC and emitted at port P4. Same as the first case mentioned above, the TE0 mode light is coupled into the 2nd-order MRRs but output at port PT2. The TE0 mode is converted to TE1 and thereafter to TM0 mode again by using the PRS in reverse. Finally, the filtered light propagates to the output port P6. In this way, the MRRs-based optical filter can operate at TE polarized state regardless of the polarization state of launched light.
Fig. 1. Schematic configuration of the proposed PIOF in (a) 3D view and (b) top view. The upper-cladding of the device is SiO2, which is not drawn to make the configuration clearer.
2.1 Tapered-rib waveguide mode convertor
Figure 2(a) shows the zoomed-in view of the L1 and L2 sections in Fig. 1(b). It is a tapered-rib waveguide that can provide a mode hybridization region to convert launched TM0 mode to TE1 mode [16]. To qualitatively verify the functionality, 3D finite-difference-time-domain (3D FDTD) simulation is performed, as shown in Figs. 2(b) and 2(c). It can be seen from Fig. 2(b) that light travels through the tapered rib waveguide without mode transformation when TE0 mode is launched. The middle-right inset is the electrical field intensity distribution of the cross section at the end position (the location of white dashed line shown in the top view) of L2 section, from which one can see that the mode remains as TE0 mode. On the contrary, when TM0 mode is launched, as seen in Fig. 2(c), light will be affected by the waveguide structure and gradually converted to TE1 mode, which has two maxima in the electrical field intensity distribution. The bottom-right inset proves that the TM0 mode has been converted to TE1 mode at the end of the tapered-rib waveguide. In the simulation, the dimension parameters are set as W1 = 500 nm, W2 = 1.2 µm, L1 = 19 µm.
Fig. 2. (a) Schematic configuration of the tapered-rib waveguide. The simulated electric field intensity distribution in the light propagation plane of tapered-rib waveguide when (b) TE0 mode or (c) TM0 mode is launched. The right insets are waveguide configuration and simulated electric field intensity distribution of the cross section at the end position of L2 section.
Since the length of the L2 section and the slab thickness hST determine the interaction time and coupling strength between the TM0 mode and TE1 mode, respectively, they play an important role in the mode conversion efficiency and loss [17]. The device performance as functions of L2 and hST is investigated by simulation in order to achieve high mode conversion efficiency, acceptable conversion loss and compact device size. The normalized transmission (T) of the tapered-rib waveguide with L2 increasing from 8 µm to 22 µm is shown in Fig. 3(a). One can see that when TE0 mode is launched, the normalized transmission of the transverse electrical field (TY–Pol) is ∼0 dB (black square line), while the normalized transmission of the vertical electrical field (TZ–Pol) is as low as ∼–28 dB (red dot line). It indicates that the TE0 mode transmits through the tapered-rib waveguide without being influenced. When the TM0 mode is launched into the tapered-rib waveguide, the TY–Pol is ∼0 dB (olive green triangle line) and TZ–Pol is less than –15 dB (blue triangle line). The result conveys that the TM polarization is converted to TE polarization, thus TY–Pol becomes the dominant component. It can also be seen in Fig. 3(a) that when L2 is larger than 16 µm, TZ-Pol is not further reduced obviously. It means the 16-µm length of L2 is enough for complete mode conversion. For the purpose of a minimized device size, L2 = 16 µm is selected for our mode convertor.
Fig. 3. The normalized transmission (T) of transverse/vertical electrical field varies with (a) length L2 and (b) slab thickness hST of the tapered-rib waveguide.
Figure 3(b) shows the normalized transmission of the tapered-rib waveguide with the hST increasing from 50 nm to 160 nm. As seen from Fig. 3(b), when TE0 mode is launched, the TY–Pol is ∼0 dB (black square line) and the TZ–Pol is as low as ∼–28 dB (red dotted line). It depicts that the change of hST has little influence on TE0 mode transmission. But when TM0 mode is launched, the TY–Pol keeps at ∼0 dB (olive green triangle line) only in the hST range of 60 nm to 130 nm. And the residual power of the undesired TZ–Pol has the lowest value of –20 dB at hST = 90 nm. Based on the above simulation results, the hST of 90 nm is adopted for our mode convertor.
2.2 Asymmetrical directional coupler (ADC)
The ADC is a fundamental component which can be designed for various functionalities, including the power splitting, polarization beam splitting, mode converting and mode-division multiplexing [18–20]. In our device, we design an ADC with tapered waveguides as the mode converter and splitter, as shown in Fig. 4(a), which is much less insensitive to the geometrical deviations compared to the conventional DC [21]. The starting section of the upper tapered waveguide and the ending section of the lower tapered waveguide in the ADC should support the stable propagation of TE1 mode and TE0 mode, respectively. Thus the widths of W3 and W6 are designed as 850 nm and 500 nm, respectively. In order to avoid the TE0 mode in the lower waveguide couples back to the upper waveguide, W4 should be larger than W6. Here, the value of 600 nm is selected for W4. The center-to-center distance (Wcenter) between the upper and lower waveguides is set to be 795 nm so that the gap between the two waveguides is not constant. The phase-matching condition between TE1 and TE0 modes can be optimized by varying the coupling length between the two waveguides, which will be discussed in later paragraphs. We simulate the ADC using variational finite-difference-time-domain (VarFDTD), which can keep the simulation accuracy of 3D FDTD and achieve faster simulation speed compared to 3D FDTD in waveguide and coupler designs. The size of the VarFDTD solver region is set to be 404 µm, 6 µm and 4 µm in the x, y and z axis, respectively. A mesh accuracy of 6 is used, which corresponds to 26 mesh points per wavelength. The default boundary condition of perfectly matched layer (PML) in x, y axis and metal in z axis is used in the simulation. The simulation results are shown in Figs. 4(b) and 4(c) with the TE0 mode and TE1 mode (coming from the tapered-rib waveguide) input to port P2, respectively. It can be seen from Fig. 4(b) that the TE0 mode travels through the upper waveguide and exits in the port P3 directly without mode transformation. On the contrary, as shown in Fig. 4(c), when TE1 mode is launched, light is efficiently coupled to the lower waveguide and exits at port P4. The bottom-right inset is the electrical intensity distribution of the cross section at the end position of L3 section, which proves that the TE1 mode has been converted to TE0 mode.
Fig. 4. (a) Schematic configuration of the ADC. The simulated electric field intensity distribution in the light propagation plane of ADC when (b) TE0 mode or (c) TE1 mode is launched. The right insets are waveguides configuration and simulated electric field intensity distribution of cross section at the end position of L3 section.
Coupling length is one of the most important parameters for ADC to control optical mode coupling [22]. In order to achieve desired mode conversion and high coupling efficiency, the ADC is simulated using the VarFDTD simulation with various coupling lengths L3 while keeping the input/output waveguide width, as well as the center distance between upper and lower waveguide. As shown in Fig. 5(a), when TE0 mode is launched, the normalized transmission at port P3 (black square line) and port P4 (red dotted line) remains ∼0 dB and is below –20 dB with L3 varying from 50 µm to 750 µm, respectively. It indicates that varying the coupling length of ADC has little influence on TE0 mode transmission. On the other hand, when TE1 mode is launched, though the normalized transmission at port P4 (olive green triangle line) is around 0 dB when L3 is changed from 100 µm to 750 µm, the undesired crosstalk at port P3 (blue triangle line) depends strongly on the length L3. It is seen that the lowest crosstalk is achieved at L3 = 400 µm.
Fig. 5. The normalized transmission (T) at port P3 and port P4 of ADC varies with (a) length L3 and (b) fabrication deviation ΔW.
The actual device dimension may deviate from the designed values due to fabrication process variation, which results in performance deterioration. The dominant dimension variation caused by fabrication error is the waveguide width in the etching processes. The ADC has the advantage of large fabrication tolerance of the waveguide width. Here the influence of waveguide width variation on the performance of our designed ADC is analyzed. We assume the widths of the ADC have the same deviation ΔW (ΔW = Wfabrication – Wdesign, Wfabrication is the actual fabrication width, Wdesign is the designed width.) induced by the fabrication error. Since the center-to-center distance between the two waveguides which composes the ADC is designed with a constant value, the gap between them varies accordingly with the changing of the waveguide widths. The simulation results are shown in Fig. 5(b), from which one can see that the crosstalk (blue triangle line and red dotted line), denoting how much light is transmitted to the undesired port. It is severely affected by the changing of ΔW. Over the range of ΔW from –80 nm to 20 nm, the crosstalk is below –18 dB. It depicts that the designed ADC can still work even the fabrication deviation of the waveguide width is at [–80, 20] nm range. It is noticed that the good tolerance of ADC improves the performance of the mode splitting, but other component such as the MRRs-based filter of the device is still sensitive to the fabrication error. The fabrication tolerance of the whole device will be investigated in future works.
2.3 2nd-order MRRs-based optical filter
The two MRRs with the same radii (R1 = R2 = 10 µm) are used to compose the filter for our PIOF shown in Fig. 1. The transmission spectra of the output port can be derived by the transform matrix method [23,24]:
The optical filters always desire a high ER and box-like spectral response [25]. The ki (i = 1,2) is the most important parameter for controlling the spectrum shape. Hence, by purposely design the coupling coefficients, box-like spectrum response with a high ER can be obtained [26]. Figure 6 shows the simulated relation of the amplitude coupling coefficients ki (i = 1,2) and the physical gap dimension in the 2nd-order MRRs-based filters. In the 3D FDTD simulation, the straight and ring waveguides have the same cross-sectional dimension of 450 nm×220 nm. According to the theoretical analysis using aforementioned Eq. (1) to Eq. (2), the optimized amplitude coupling coefficient k1 and k2 are 0.22 and 0.038, respectively, to achieve the box-like filtering spectrum with high ER. It is found in Fig. 6 that the corresponding gap between the straight and ring waveguides (Wg1), and the gap between the ring and ring waveguides (Wg2) are 180 nm and 380 nm, respectively. In the calculation, the Si waveguide loss is assumed to be 3 dB/cm, with equivalent α=0.9978.
Fig. 6. The relation of the amplitude coupling coefficients (k1 and k2) and the physical gap dimension in the 2nd-order MRRs-based filter (Wg1 and Wg2).
3. Device fabrication and measurement
The designed device is fabricated on an 8-inch SOI wafer with a 220 nm-thick top-silicon layer and a 3 µm-thick buried-dioxide layer. The 248-nm deep UV photolithography is used to pattern the device structure, followed by the inductively coupled plasma (ICP) etching process to transfer the hard mask and photoresist pattern to the silicon layer. Then a 3-µm-thick SiO2 upper-cladding layer is deposited using the plasma-enhanced chemical vapor deposition (PECVD) process. The optical microscopy image of the fabricated device is shown in Fig. 7(a). Limited by the field of view, the tapered-rib waveguide for TM0 to TE1 conversion is not captured. The scanning electron microscopy (SEM) images are taken instead for the tapered-rib waveguide. As seen in Figs. 7(b) and 7(c), the 220-nm-thick silicon waveguide and partially etched 90-nm-thick silicon slab are connected smoothly, which indicates the etching process is well developed. However, one can see the misalignment between the waveguide layer and slab layer in Fig. 7(c). This might increase the conversion loss and undesired crosstalk. Figure 7(d) shows the SEM image of the beginning part of the ADC, the narrow gap between the two waveguides is clearly opened by the ICP etching process. The polarization filtering functionality of the PIOF is characterized by using a broad-band amplified spontaneous emission (ASE) light source, a polarization controller and an optical spectrum analyzer (OSA). The lensed fibers with a mode diameter of 2.5 µm are used to couple light in and out of the chip.
Fig. 7. (a) Optical microscopy image of the fabricated PIOF. SEM images of (b)–(c) tapered-rib waveguide and (d) ADC. (Limited by the field of view of the microscope, the left side of ADC is the tapered-rib waveguide, which can not be captured in the image.)
The PRS composed of the tapered-rib waveguide and ADC is simulated as and fabricated in the same SOI wafer as an independent device. Figures 8(a) and 8(b) show the simulated and measured transmission spectra of the fabricated PRS. One can see from Fig. 8(a) that in the wavelength range from 1520 nm to 1560 nm, when light with TE polarization is launched into port P1, it transmits to port P3 (black line) directly with a low IL close to 0 dB, while the undesired crosstalk to port P4 is below -25 dB (blue line). On the other hand, the IL (olive green line) and crosstalk (red line) for injection light with TM mode are below 0.25 dB and –15 dB, respectively. And the crosstalk slightly decreases as the wavelength increases. The measurement results shown in Fig. 8(b) have good agreement with the simulation shown in Fig. 8(a) for the light with TM mode. Though the crosstalk of light with TE mode has a difference of ∼5 dB between the simulation and measurement result, the tendency of the normalized transmission vs wavelength is very similar. It should be noted that all the measured transmission results are normalized to the spectrum of the ASE light source.
In order to show the polarization insensitivity of the PIOF, an ordinary 2nd-order MRRs-based optical filter is also fabricated in the same batch for comparison. As seen in Fig. 9(a), the filtering spectra with TE input (black line and blue line) and TM input (olive green line and red line) are quite different without the polarization diversity. There is box-like filtering spectrum (blue line) at the output port around the wavelength of 1548.5 nm with TE input, while no effective filtering functionality is found in the wavelength regime of 1547.5 nm to 1549.6 nm with TM input. The measured transmission spectra of the fabricated PIOF are shown in Fig. 9(b). As seen from it, the box-like filtering spectra at 1548.6 nm with TE (blue line) and TM (red line) input light are almost identical above the normalized transmission of –35 dB, with a box-like spectral response which has 3-dB bandwidth of ∼0.15 nm, large ER over 30 dB and insertion loss (IL) of ∼3.6 dB, respectively. It reveals that the polarization-insensitive characteristic of the designed device is achieved. The measured IL of the PIOF is around 3.6 dB, which is higher than the simulated value of ∼1.3 dB. It is mainly attributed to the propagation loss in the ring waveguide, which is much larger than the assumed value in simulation and is estimated to be around 15 dB/cm. This can be mitigated by designing the racetrack cavity using larger radius bending and wider connection waveguides [27].
Fig. 9. The measured spectral responses of the (a) ordinary 2nd-order MRRs-based optical filter and the (b) PIOF.
4. Conclusion
In conclusion, we have designed and demonstrated a PIOF which is composed of PRSs and 2nd-order MRRs on an 8-inch SOI wafer using CMOS-compatible fabrication process. The experimental results show that the polarization-insensitive filtering function is achieved. The box-like filtering spectra at 1548.6 nm exhibit a 3-dB bandwidth of ∼0.15 nm and a large ER over 30 dB. This work provides a potential solution for building polarization-insensitive photonic circuits on the platform with high-index-contrast waveguides.
Funding
National Natural Science Foundation of China (62105376); Open Fund of Key Laboratory of Advanced Display and System Applications of Ministry of Education, Shanghai University.
Disclosures
The authors declare no conflicts of interest.
Data availability
Data presented in this paper are not publicly available at this time but may be obtained from the authors upon reasonable request.
References
1. R. W. Tkach, “Scaling optical communications for the next decade and beyond,” Bell Labs Tech. J. 14(4), 3–9 (2010). [CrossRef]
2. Z. Zhou, B. Yin, Q. Deng, X. Li, and J. Cui, “Lowering the energy consumption in silicon photonic devices and systems,” Photonics Res. 3(5), B28–46 (2015). [CrossRef]
3. M. Li, X. Chen, Y. Su, X. Wang, M. Chen, D. Dai, J. Liu, and N. Zhu, “Photonic integration circuits in China,” IEEE J. Quantum Electron. 52(1), 1–17 (2016). [CrossRef]
4. D. Dai, Y. Yin, L. Yu, H. Wu, D. Liang, Z. Wang, and L. Liu, “Silicon-plus photonics,” Front. Optoelectron. 9(3), 436–449 (2016). [CrossRef]
5. T. Barwicz, M. R. Watts, M. A. Popovic, P. T. Rakich, L. Socci, F. X. Kartner, E. P. Ippen, and H. I. Smith, “Polarization-transparent microphotonic devices in the strong confinement limit,” Nat. Photonics 1(1), 57–60 (2007). [CrossRef]
6. W. Bogaerts and L. Chrostowski, “Silicon photonics circuit design: methods, tools and challenges,” Laser Photonics Rev. 12(4), 1700237 (2018). [CrossRef]
7. X. Deng, L. Yan, H. Jiang, X. Feng, W. Pan, and B. Luo, “Polarization-insensitive and tunable silicon Mach–Zehnder wavelength filters with flat transmission passband,” IEEE Photonics J. 10(3), 1–7 (2018). [CrossRef]
8. H. Yun, L. Chrostowski, and N. A. F. Jaeger, “Narrow-band, polarization-independent, transmission filter in a silicon-on-insulator strip waveguide,” Opt. Lett. 44(4), 847–850 (2019). [CrossRef]
9. H. Okayama, Y. Onawa, D. Shimura, H. Yaegashi, and H. Sasaki, “Polarization rotation Bragg grating using Si wire waveguide with non-vertical sidewall,” Opt. Express 22(25), 31371–31378 (2014). [CrossRef]
10. D. Liu and D. Dai, “Silicon-based polarization-insensitive optical filter with dual-gratings,” Opt. Express 27(15), 20704–20710 (2019). [CrossRef]
11. L. Xiang, Y. Yu, Y. Qin, J. Zhou, B. Zou, and X. Zhang, “SOI based ultracompact polarization insensitive filter for PDM signal processing,” Opt. Lett. 38(14), 2379–2381 (2013). [CrossRef]
12. H. Fukuda, K. Yamada, T. Tsuchizawa, T. Watanabe, H. Shinojima, and S. Itabashi, “Silicon photonic circuit with polarization diversity,” Opt. Express 16(7), 4872–4880 (2008). [CrossRef]
13. C. Li, J. H. Song, J. Zhang, H. Zhang, S. Chen, M. Yu, and G. Q. Lo, “Silicon polarization independent microring resonator-based optical tunable filter circuit with fiber assembly,” Opt. Express 19(16), 15429–15437 (2011). [CrossRef]
14. H. Wang, J. Dai, H. Jia, S. Shao, X. Fu, L. Zhang, and L. Yang, “Polarization-independent tunable optical filter with variable bandwidth based on silicon-on-insulator waveguides,” Nanophotonics 7(8), 1469–1477 (2018). [CrossRef]
15. A. H. K. Park, H. Shoman, M. Ma, S. Shekhar, and L. Chrostowski, “Ring resonator based polarization diversity WDM receiver,” Opt. Express 27(5), 6147–6157 (2019). [CrossRef]
16. D. Chen, X. Xiao, L. Wang, W. Liu, Q. Yang, and S. Yu, “Highly efficient silicon optical polarization rotators based on mode order conversions,” Opt. Lett. 41(5), 1070–1073 (2016). [CrossRef]
17. D. Dai, Y. Tang, and J. E. Bowers, “Mode conversion in tapered submicron silicon ridge optical waveguides,” Opt. Express 20(12), 13425–13439 (2012). [CrossRef]
18. D. Dai and J. E. Bowers, “Novel concept for ultracompact polarization splitter rotator based on silicon nanowires,” Opt. Express 19(11), 10940–10949 (2011). [CrossRef]
19. E. S. Magden, N. Li, M. Raval, C. V. Poulton, A. Ruocco, N. Singh, D. Vermeulen, E. P. Ippen, L. A. Kolodziejski, and M. R. Watts, “Transmissive silicon photonic dichroic filters with spectrally selective waveguides,” Nat. Commun. 9(1), 3009 (2018). [CrossRef]
20. D. Dai and S. Wang, “Asymmetric directional couplers based on silicon nanophotonic waveguides and applications,” Front. Optoelectron. 9(3), 450–465 (2016). [CrossRef]
21. D. Dai, Z. Wang, and J. E. Bowers, “Ultrashort broadband polarization beam splitter based on an asymmetrical directional coupler,” Opt. Lett. 36(13), 2590–2592 (2011). [CrossRef]
22. W. D. Sacher, T. Barwicz, B. J. F. Taylor, and J. K. S. Poon, “Polarization rotator-splitters in standard active silicon photonics platforms,” Opt. Express 22(4), 3777–3786 (2014). [CrossRef]
23. A. Yariv, “Universal relations for coupling of optical power between microresonators and dielectric waveguides,” Electron. Lett. 36(4), 321–322 (2000). [CrossRef]
24. A. Yariv, “Critical coupling and its control in optical waveguide-ring resonator systems,” IEEE Photonics Technol. Lett. 14(4), 483–485 (2002). [CrossRef]
25. T. Hu, W. Wang, C. Qiu, P. Yu, H. Qiu, Y. Zhao, X. Jiang, and J. Yang, “Thermally tunable filters based on third-order microring resonators for WDM applications,” IEEE Photonics Technol. Lett. 24(6), 524–526 (2012). [CrossRef]
26. J. Yang, Q. Zhou, F. Zhao, X. Jiang, B. Howley, M. Wang, and R. T. Chen, “Characteristics of optical bandpass filters employing series-cascaded double-ring resonators,” Opt. Commun. 228(1-3), 91–98 (2003). [CrossRef]
27. H. Qiu, F. Zhou, J. Qie, Y. Yao, X. Hu, Y. Zhang, X. Xiao, Y. Yu, J. Dong, and X. Zhang, “A continuously tunable sub-gigahertz microwave photonic bandpass filter based on an ultra-high-Q silicon microring resonator,” J. Lightwave Technol. 36(19), 4312–4318 (2018). [CrossRef]
